fs/btrfs/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 #include <linux/sched.h>
  19 #include <linux/pagemap.h>
  20 #include <linux/writeback.h>
  21 #include <linux/blkdev.h>
  22 #include <linux/sort.h>
  23 #include <linux/rcupdate.h>
  24 #include <linux/kthread.h>
  25 #include "compat.h"
  26 #include "hash.h"
  27 #include "ctree.h"
  28 #include "disk-io.h"
  29 #include "print-tree.h"
  30 #include "transaction.h"
  31 #include "volumes.h"
  32 #include "locking.h"
  33 #include "free-space-cache.h"
  34
  35 static int update_reserved_extents(struct btrfs_root *root,
  36                                    u64 bytenr, u64 num, int reserve);
  37 static int update_block_group(struct btrfs_trans_handle *trans,
  38                               struct btrfs_root *root,
  39                               u64 bytenr, u64 num_bytes, int alloc,
  40                               int mark_free);
  41 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
  42                                 struct btrfs_root *root,
  43                                 u64 bytenr, u64 num_bytes, u64 parent,
  44                                 u64 root_objectid, u64 owner_objectid,
  45                                 u64 owner_offset, int refs_to_drop,
  46                                 struct btrfs_delayed_extent_op *extra_op);
  47 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
  48                                     struct extent_buffer *leaf,
  49                                     struct btrfs_extent_item *ei);
  50 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
  51                                       struct btrfs_root *root,
  52                                       u64 parent, u64 root_objectid,
  53                                       u64 flags, u64 owner, u64 offset,
  54                                       struct btrfs_key *ins, int ref_mod);
  55 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
  56                                      struct btrfs_root *root,
  57                                      u64 parent, u64 root_objectid,
  58                                      u64 flags, struct btrfs_disk_key *key,
  59                                      int level, struct btrfs_key *ins);
  60
  61 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
  62                           struct btrfs_root *extent_root, u64 alloc_bytes,
  63                           u64 flags, int force);
  64
  65 static noinline int
  66 block_group_cache_done(struct btrfs_block_group_cache *cache)
  67 {
  68         smp_mb();
  69         return cache->cached == BTRFS_CACHE_FINISHED;
  70 }
  71
  72 static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
  73 {
  74         return (cache->flags & bits) == bits;
  75 }
  76
  77 /*
  78  * this adds the block group to the fs_info rb tree for the block group
  79  * cache
  80  */
  81 static int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
  82                                 struct btrfs_block_group_cache *block_group)
  83 {
  84         struct rb_node **p;
  85         struct rb_node *parent = NULL;
  86         struct btrfs_block_group_cache *cache;
  87
  88         spin_lock(&info->block_group_cache_lock);
  89         p = &info->block_group_cache_tree.rb_node;
  90
  91         while (*p) {
  92                 parent = *p;
  93                 cache = rb_entry(parent, struct btrfs_block_group_cache,
  94                                  cache_node);
  95                 if (block_group->key.objectid < cache->key.objectid) {
  96                         p = &(*p)->rb_left;
  97                 } else if (block_group->key.objectid > cache->key.objectid) {
  98                         p = &(*p)->rb_right;
  99                 } else {
 100                         spin_unlock(&info->block_group_cache_lock);
 101                         return -EEXIST;
 102                 }
 103         }
 104
 105         rb_link_node(&block_group->cache_node, parent, p);
 106         rb_insert_color(&block_group->cache_node,
 107                         &info->block_group_cache_tree);
 108         spin_unlock(&info->block_group_cache_lock);
 109
 110         return 0;
 111 }
 112
 113 /*
 114  * This will return the block group at or after bytenr if contains is 0, else
 115  * it will return the block group that contains the bytenr
 116  */
 117 static struct btrfs_block_group_cache *
 118 block_group_cache_tree_search(struct btrfs_fs_info *info, u64 bytenr,
 119                               int contains)
 120 {
 121         struct btrfs_block_group_cache *cache, *ret = NULL;
 122         struct rb_node *n;
 123         u64 end, start;
 124
 125         spin_lock(&info->block_group_cache_lock);
 126         n = info->block_group_cache_tree.rb_node;
 127
 128         while (n) {
 129                 cache = rb_entry(n, struct btrfs_block_group_cache,
 130                                  cache_node);
 131                 end = cache->key.objectid + cache->key.offset - 1;
 132                 start = cache->key.objectid;
 133
 134                 if (bytenr < start) {
 135                         if (!contains && (!ret || start < ret->key.objectid))
 136                                 ret = cache;
 137                         n = n->rb_left;
 138                 } else if (bytenr > start) {
 139                         if (contains && bytenr <= end) {
 140                                 ret = cache;
 141                                 break;
 142                         }
 143                         n = n->rb_right;
 144                 } else {
 145                         ret = cache;
 146                         break;
 147                 }
 148         }
 149         if (ret)
 150                 atomic_inc(&ret->count);
 151         spin_unlock(&info->block_group_cache_lock);
 152
 153         return ret;
 154 }
 155
 156 /*
 157  * We always set EXTENT_LOCKED for the super mirror extents so we don't
 158  * overwrite them, so those bits need to be unset.  Also, if we are unmounting
 159  * with pinned extents still sitting there because we had a block group caching,
 160  * we need to clear those now, since we are done.
 161  */
 162 void btrfs_free_pinned_extents(struct btrfs_fs_info *info)
 163 {
 164         u64 start, end, last = 0;
 165         int ret;
 166
 167         while (1) {
 168                 ret = find_first_extent_bit(&info->pinned_extents, last,
 169                                             &start, &end,
 170                                             EXTENT_LOCKED|EXTENT_DIRTY);
 171                 if (ret)
 172                         break;
 173
 174                 clear_extent_bits(&info->pinned_extents, start, end,
 175                                   EXTENT_LOCKED|EXTENT_DIRTY, GFP_NOFS);
 176                 last = end+1;
 177         }
 178 }
 179
 180 static int remove_sb_from_cache(struct btrfs_root *root,
 181                                 struct btrfs_block_group_cache *cache)
 182 {
 183         struct btrfs_fs_info *fs_info = root->fs_info;
 184         u64 bytenr;
 185         u64 *logical;
 186         int stripe_len;
 187         int i, nr, ret;
 188
 189         for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
 190                 bytenr = btrfs_sb_offset(i);
 191                 ret = btrfs_rmap_block(&root->fs_info->mapping_tree,
 192                                        cache->key.objectid, bytenr,
 193                                        0, &logical, &nr, &stripe_len);
 194                 BUG_ON(ret);
 195                 while (nr--) {
 196                         try_lock_extent(&fs_info->pinned_extents,
 197                                         logical[nr],
 198                                         logical[nr] + stripe_len - 1, GFP_NOFS);
 199                 }
 200                 kfree(logical);
 201         }
 202
 203         return 0;
 204 }
 205
 206 /*
 207  * this is only called by cache_block_group, since we could have freed extents
 208  * we need to check the pinned_extents for any extents that can't be used yet
 209  * since their free space will be released as soon as the transaction commits.
 210  */
 211 static u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
 212                               struct btrfs_fs_info *info, u64 start, u64 end)
 213 {
 214         u64 extent_start, extent_end, size, total_added = 0;
 215         int ret;
 216
 217         while (start < end) {
 218                 ret = find_first_extent_bit(&info->pinned_extents, start,
 219                                             &extent_start, &extent_end,
 220                                             EXTENT_DIRTY|EXTENT_LOCKED);
 221                 if (ret)
 222                         break;
 223
 224                 if (extent_start == start) {
 225                         start = extent_end + 1;
 226                 } else if (extent_start > start && extent_start < end) {
 227                         size = extent_start - start;
 228                         total_added += size;
 229                         ret = btrfs_add_free_space(block_group, start,
 230                                                    size);
 231                         BUG_ON(ret);
 232                         start = extent_end + 1;
 233                 } else {
 234                         break;
 235                 }
 236         }
 237
 238         if (start < end) {
 239                 size = end - start;
 240                 total_added += size;
 241                 ret = btrfs_add_free_space(block_group, start, size);
 242                 BUG_ON(ret);
 243         }
 244
 245         return total_added;
 246 }
 247
 248 static int caching_kthread(void *data)
 249 {
 250         struct btrfs_block_group_cache *block_group = data;
 251         struct btrfs_fs_info *fs_info = block_group->fs_info;
 252         u64 last = 0;
 253         struct btrfs_path *path;
 254         int ret = 0;
 255         struct btrfs_key key;
 256         struct extent_buffer *leaf;
 257         int slot;
 258         u64 total_found = 0;
 259
 260         BUG_ON(!fs_info);
 261
 262         path = btrfs_alloc_path();
 263         if (!path)
 264                 return -ENOMEM;
 265
 266         atomic_inc(&block_group->space_info->caching_threads);
 267         last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
 268         /*
 269          * We don't want to deadlock with somebody trying to allocate a new
 270          * extent for the extent root while also trying to search the extent
 271          * root to add free space.  So we skip locking and search the commit
 272          * root, since its read-only
 273          */
 274         path->skip_locking = 1;
 275         path->search_commit_root = 1;
 276         path->reada = 2;
 277
 278         key.objectid = last;
 279         key.offset = 0;
 280         btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 281 again:
 282         /* need to make sure the commit_root doesn't disappear */
 283         down_read(&fs_info->extent_commit_sem);
 284
 285         ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
 286         if (ret < 0)
 287                 goto err;
 288
 289         while (1) {
 290                 smp_mb();
 291                 if (block_group->fs_info->closing > 1) {
 292                         last = (u64)-1;
 293                         break;
 294                 }
 295
 296                 leaf = path->nodes[0];
 297                 slot = path->slots[0];
 298                 if (slot >= btrfs_header_nritems(leaf)) {
 299                         ret = btrfs_next_leaf(fs_info->extent_root, path);
 300                         if (ret < 0)
 301                                 goto err;
 302                         else if (ret)
 303                                 break;
 304
 305                         if (need_resched() ||
 306                             btrfs_transaction_in_commit(fs_info)) {
 307                                 leaf = path->nodes[0];
 308
 309                                 /* this shouldn't happen, but if the
 310                                  * leaf is empty just move on.
 311                                  */
 312                                 if (btrfs_header_nritems(leaf) == 0)
 313                                         break;
 314                                 /*
 315                                  * we need to copy the key out so that
 316                                  * we are sure the next search advances
 317                                  * us forward in the btree.
 318                                  */
 319                                 btrfs_item_key_to_cpu(leaf, &key, 0);
 320                                 btrfs_release_path(fs_info->extent_root, path);
 321                                 up_read(&fs_info->extent_commit_sem);
 322                                 schedule_timeout(1);
 323                                 goto again;
 324                         }
 325
 326                         continue;
 327                 }
 328                 btrfs_item_key_to_cpu(leaf, &key, slot);
 329                 if (key.objectid < block_group->key.objectid)
 330                         goto next;
 331
 332                 if (key.objectid >= block_group->key.objectid +
 333                     block_group->key.offset)
 334                         break;
 335
 336                 if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
 337                         total_found += add_new_free_space(block_group,
 338                                                           fs_info, last,
 339                                                           key.objectid);
 340                         last = key.objectid + key.offset;
 341                 }
 342
 343                 if (total_found > (1024 * 1024 * 2)) {
 344                         total_found = 0;
 345                         wake_up(&block_group->caching_q);
 346                 }
 347 next:
 348                 path->slots[0]++;
 349         }
 350         ret = 0;
 351
 352         total_found += add_new_free_space(block_group, fs_info, last,
 353                                           block_group->key.objectid +
 354                                           block_group->key.offset);
 355
 356         spin_lock(&block_group->lock);
 357         block_group->cached = BTRFS_CACHE_FINISHED;
 358         spin_unlock(&block_group->lock);
 359
 360 err:
 361         btrfs_free_path(path);
 362         up_read(&fs_info->extent_commit_sem);
 363         atomic_dec(&block_group->space_info->caching_threads);
 364         wake_up(&block_group->caching_q);
 365
 366         return 0;
 367 }
 368
 369 static int cache_block_group(struct btrfs_block_group_cache *cache)
 370 {
 371         struct task_struct *tsk;
 372         int ret = 0;
 373
 374         spin_lock(&cache->lock);
 375         if (cache->cached != BTRFS_CACHE_NO) {
 376                 spin_unlock(&cache->lock);
 377                 return ret;
 378         }
 379         cache->cached = BTRFS_CACHE_STARTED;
 380         spin_unlock(&cache->lock);
 381
 382         tsk = kthread_run(caching_kthread, cache, "btrfs-cache-%llu\n",
 383                           cache->key.objectid);
 384         if (IS_ERR(tsk)) {
 385                 ret = PTR_ERR(tsk);
 386                 printk(KERN_ERR "error running thread %d\n", ret);
 387                 BUG();
 388         }
 389
 390         return ret;
 391 }
 392
 393 /*
 394  * return the block group that starts at or after bytenr
 395  */
 396 static struct btrfs_block_group_cache *
 397 btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr)
 398 {
 399         struct btrfs_block_group_cache *cache;
 400
 401         cache = block_group_cache_tree_search(info, bytenr, 0);
 402
 403         return cache;
 404 }
 405
 406 /*
 407  * return the block group that contains the given bytenr
 408  */
 409 struct btrfs_block_group_cache *btrfs_lookup_block_group(
 410                                                  struct btrfs_fs_info *info,
 411                                                  u64 bytenr)
 412 {
 413         struct btrfs_block_group_cache *cache;
 414
 415         cache = block_group_cache_tree_search(info, bytenr, 1);
 416
 417         return cache;
 418 }
 419
 420 void btrfs_put_block_group(struct btrfs_block_group_cache *cache)
 421 {
 422         if (atomic_dec_and_test(&cache->count))
 423                 kfree(cache);
 424 }
 425
 426 static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
 427                                                   u64 flags)
 428 {
 429         struct list_head *head = &info->space_info;
 430         struct btrfs_space_info *found;
 431
 432         rcu_read_lock();
 433         list_for_each_entry_rcu(found, head, list) {
 434                 if (found->flags == flags) {
 435                         rcu_read_unlock();
 436                         return found;
 437                 }
 438         }
 439         rcu_read_unlock();
 440         return NULL;
 441 }
 442
 443 /*
 444  * after adding space to the filesystem, we need to clear the full flags
 445  * on all the space infos.
 446  */
 447 void btrfs_clear_space_info_full(struct btrfs_fs_info *info)
 448 {
 449         struct list_head *head = &info->space_info;
 450         struct btrfs_space_info *found;
 451
 452         rcu_read_lock();
 453         list_for_each_entry_rcu(found, head, list)
 454                 found->full = 0;
 455         rcu_read_unlock();
 456 }
 457
 458 static u64 div_factor(u64 num, int factor)
 459 {
 460         if (factor == 10)
 461                 return num;
 462         num *= factor;
 463         do_div(num, 10);
 464         return num;
 465 }
 466
 467 u64 btrfs_find_block_group(struct btrfs_root *root,
 468                            u64 search_start, u64 search_hint, int owner)
 469 {
 470         struct btrfs_block_group_cache *cache;
 471         u64 used;
 472         u64 last = max(search_hint, search_start);
 473         u64 group_start = 0;
 474         int full_search = 0;
 475         int factor = 9;
 476         int wrapped = 0;
 477 again:
 478         while (1) {
 479                 cache = btrfs_lookup_first_block_group(root->fs_info, last);
 480                 if (!cache)
 481                         break;
 482
 483                 spin_lock(&cache->lock);
 484                 last = cache->key.objectid + cache->key.offset;
 485                 used = btrfs_block_group_used(&cache->item);
 486
 487                 if ((full_search || !cache->ro) &&
 488                     block_group_bits(cache, BTRFS_BLOCK_GROUP_METADATA)) {
 489                         if (used + cache->pinned + cache->reserved <
 490                             div_factor(cache->key.offset, factor)) {
 491                                 group_start = cache->key.objectid;
 492                                 spin_unlock(&cache->lock);
 493                                 btrfs_put_block_group(cache);
 494                                 goto found;
 495                         }
 496                 }
 497                 spin_unlock(&cache->lock);
 498                 btrfs_put_block_group(cache);
 499                 cond_resched();
 500         }
 501         if (!wrapped) {
 502                 last = search_start;
 503                 wrapped = 1;
 504                 goto again;
 505         }
 506         if (!full_search && factor < 10) {
 507                 last = search_start;
 508                 full_search = 1;
 509                 factor = 10;
 510                 goto again;
 511         }
 512 found:
 513         return group_start;
 514 }
 515
 516 /* simple helper to search for an existing extent at a given offset */
 517 int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len)
 518 {
 519         int ret;
 520         struct btrfs_key key;
 521         struct btrfs_path *path;
 522
 523         path = btrfs_alloc_path();
 524         BUG_ON(!path);
 525         key.objectid = start;
 526         key.offset = len;
 527         btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 528         ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path,
 529                                 0, 0);
 530         btrfs_free_path(path);
 531         return ret;
 532 }
 533
 534 /*
 535  * Back reference rules.  Back refs have three main goals:
 536  *
 537  * 1) differentiate between all holders of references to an extent so that
 538  *    when a reference is dropped we can make sure it was a valid reference
 539  *    before freeing the extent.
 540  *
 541  * 2) Provide enough information to quickly find the holders of an extent
 542  *    if we notice a given block is corrupted or bad.
 543  *
 544  * 3) Make it easy to migrate blocks for FS shrinking or storage pool
 545  *    maintenance.  This is actually the same as #2, but with a slightly
 546  *    different use case.
 547  *
 548  * There are two kinds of back refs. The implicit back refs is optimized
 549  * for pointers in non-shared tree blocks. For a given pointer in a block,
 550  * back refs of this kind provide information about the block's owner tree
 551  * and the pointer's key. These information allow us to find the block by
 552  * b-tree searching. The full back refs is for pointers in tree blocks not
 553  * referenced by their owner trees. The location of tree block is recorded
 554  * in the back refs. Actually the full back refs is generic, and can be
 555  * used in all cases the implicit back refs is used. The major shortcoming
 556  * of the full back refs is its overhead. Every time a tree block gets
 557  * COWed, we have to update back refs entry for all pointers in it.
 558  *
 559  * For a newly allocated tree block, we use implicit back refs for
 560  * pointers in it. This means most tree related operations only involve
 561  * implicit back refs. For a tree block created in old transaction, the
 562  * only way to drop a reference to it is COW it. So we can detect the
 563  * event that tree block loses its owner tree's reference and do the
 564  * back refs conversion.
 565  *
 566  * When a tree block is COW'd through a tree, there are four cases:
 567  *
 568  * The reference count of the block is one and the tree is the block's
 569  * owner tree. Nothing to do in this case.
 570  *
 571  * The reference count of the block is one and the tree is not the
 572  * block's owner tree. In this case, full back refs is used for pointers
 573  * in the block. Remove these full back refs, add implicit back refs for
 574  * every pointers in the new block.
 575  *
 576  * The reference count of the block is greater than one and the tree is
 577  * the block's owner tree. In this case, implicit back refs is used for
 578  * pointers in the block. Add full back refs for every pointers in the
 579  * block, increase lower level extents' reference counts. The original
 580  * implicit back refs are entailed to the new block.
 581  *
 582  * The reference count of the block is greater than one and the tree is
 583  * not the block's owner tree. Add implicit back refs for every pointer in
 584  * the new block, increase lower level extents' reference count.
 585  *
 586  * Back Reference Key composing:
 587  *
 588  * The key objectid corresponds to the first byte in the extent,
 589  * The key type is used to differentiate between types of back refs.
 590  * There are different meanings of the key offset for different types
 591  * of back refs.
 592  *
 593  * File extents can be referenced by:
 594  *
 595  * - multiple snapshots, subvolumes, or different generations in one subvol
 596  * - different files inside a single subvolume
 597  * - different offsets inside a file (bookend extents in file.c)
 598  *
 599  * The extent ref structure for the implicit back refs has fields for:
 600  *
 601  * - Objectid of the subvolume root
 602  * - objectid of the file holding the reference
 603  * - original offset in the file
 604  * - how many bookend extents
 605  *
 606  * The key offset for the implicit back refs is hash of the first
 607  * three fields.
 608  *
 609  * The extent ref structure for the full back refs has field for:
 610  *
 611  * - number of pointers in the tree leaf
 612  *
 613  * The key offset for the implicit back refs is the first byte of
 614  * the tree leaf
 615  *
 616  * When a file extent is allocated, The implicit back refs is used.
 617  * the fields are filled in:
 618  *
 619  *     (root_key.objectid, inode objectid, offset in file, 1)
 620  *
 621  * When a file extent is removed file truncation, we find the
 622  * corresponding implicit back refs and check the following fields:
 623  *
 624  *     (btrfs_header_owner(leaf), inode objectid, offset in file)
 625  *
 626  * Btree extents can be referenced by:
 627  *
 628  * - Different subvolumes
 629  *
 630  * Both the implicit back refs and the full back refs for tree blocks
 631  * only consist of key. The key offset for the implicit back refs is
 632  * objectid of block's owner tree. The key offset for the full back refs
 633  * is the first byte of parent block.
 634  *
 635  * When implicit back refs is used, information about the lowest key and
 636  * level of the tree block are required. These information are stored in
 637  * tree block info structure.
 638  */
 639
 640 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
 641 static int convert_extent_item_v0(struct btrfs_trans_handle *trans,
 642                                   struct btrfs_root *root,
 643                                   struct btrfs_path *path,
 644                                   u64 owner, u32 extra_size)
 645 {
 646         struct btrfs_extent_item *item;
 647         struct btrfs_extent_item_v0 *ei0;
 648         struct btrfs_extent_ref_v0 *ref0;
 649         struct btrfs_tree_block_info *bi;
 650         struct extent_buffer *leaf;
 651         struct btrfs_key key;
 652         struct btrfs_key found_key;
 653         u32 new_size = sizeof(*item);
 654         u64 refs;
 655         int ret;
 656
 657         leaf = path->nodes[0];
 658         BUG_ON(btrfs_item_size_nr(leaf, path->slots[0]) != sizeof(*ei0));
 659
 660         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 661         ei0 = btrfs_item_ptr(leaf, path->slots[0],
 662                              struct btrfs_extent_item_v0);
 663         refs = btrfs_extent_refs_v0(leaf, ei0);
 664
 665         if (owner == (u64)-1) {
 666                 while (1) {
 667                         if (path->slots[0] >= btrfs_header_nritems(leaf)) {
 668                                 ret = btrfs_next_leaf(root, path);
 669                                 if (ret < 0)
 670                                         return ret;
 671                                 BUG_ON(ret > 0);
 672                                 leaf = path->nodes[0];
 673                         }
 674                         btrfs_item_key_to_cpu(leaf, &found_key,
 675                                               path->slots[0]);
 676                         BUG_ON(key.objectid != found_key.objectid);
 677                         if (found_key.type != BTRFS_EXTENT_REF_V0_KEY) {
 678                                 path->slots[0]++;
 679                                 continue;
 680                         }
 681                         ref0 = btrfs_item_ptr(leaf, path->slots[0],
 682                                               struct btrfs_extent_ref_v0);
 683                         owner = btrfs_ref_objectid_v0(leaf, ref0);
 684                         break;
 685                 }
 686         }
 687         btrfs_release_path(root, path);
 688
 689         if (owner < BTRFS_FIRST_FREE_OBJECTID)
 690                 new_size += sizeof(*bi);
 691
 692         new_size -= sizeof(*ei0);
 693         ret = btrfs_search_slot(trans, root, &key, path,
 694                                 new_size + extra_size, 1);
 695         if (ret < 0)
 696                 return ret;
 697         BUG_ON(ret);
 698
 699         ret = btrfs_extend_item(trans, root, path, new_size);
 700         BUG_ON(ret);
 701
 702         leaf = path->nodes[0];
 703         item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
 704         btrfs_set_extent_refs(leaf, item, refs);
 705         /* FIXME: get real generation */
 706         btrfs_set_extent_generation(leaf, item, 0);
 707         if (owner < BTRFS_FIRST_FREE_OBJECTID) {
 708                 btrfs_set_extent_flags(leaf, item,
 709                                        BTRFS_EXTENT_FLAG_TREE_BLOCK |
 710                                        BTRFS_BLOCK_FLAG_FULL_BACKREF);
 711                 bi = (struct btrfs_tree_block_info *)(item + 1);
 712                 /* FIXME: get first key of the block */
 713                 memset_extent_buffer(leaf, 0, (unsigned long)bi, sizeof(*bi));
 714                 btrfs_set_tree_block_level(leaf, bi, (int)owner);
 715         } else {
 716                 btrfs_set_extent_flags(leaf, item, BTRFS_EXTENT_FLAG_DATA);
 717         }
 718         btrfs_mark_buffer_dirty(leaf);
 719         return 0;
 720 }
 721 #endif
 722
 723 static u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 offset)
 724 {
 725         u32 high_crc = ~(u32)0;
 726         u32 low_crc = ~(u32)0;
 727         __le64 lenum;
 728
 729         lenum = cpu_to_le64(root_objectid);
 730         high_crc = crc32c(high_crc, &lenum, sizeof(lenum));
 731         lenum = cpu_to_le64(owner);
 732         low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
 733         lenum = cpu_to_le64(offset);
 734         low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
 735
 736         return ((u64)high_crc << 31) ^ (u64)low_crc;
 737 }
 738
 739 static u64 hash_extent_data_ref_item(struct extent_buffer *leaf,
 740                                      struct btrfs_extent_data_ref *ref)
 741 {
 742         return hash_extent_data_ref(btrfs_extent_data_ref_root(leaf, ref),
 743                                     btrfs_extent_data_ref_objectid(leaf, ref),
 744                                     btrfs_extent_data_ref_offset(leaf, ref));
 745 }
 746
 747 static int match_extent_data_ref(struct extent_buffer *leaf,
 748                                  struct btrfs_extent_data_ref *ref,
 749                                  u64 root_objectid, u64 owner, u64 offset)
 750 {
 751         if (btrfs_extent_data_ref_root(leaf, ref) != root_objectid ||
 752             btrfs_extent_data_ref_objectid(leaf, ref) != owner ||
 753             btrfs_extent_data_ref_offset(leaf, ref) != offset)
 754                 return 0;
 755         return 1;
 756 }
 757
 758 static noinline int lookup_extent_data_ref(struct btrfs_trans_handle *trans,
 759                                            struct btrfs_root *root,
 760                                            struct btrfs_path *path,
 761                                            u64 bytenr, u64 parent,
 762                                            u64 root_objectid,
 763                                            u64 owner, u64 offset)
 764 {
 765         struct btrfs_key key;
 766         struct btrfs_extent_data_ref *ref;
 767         struct extent_buffer *leaf;
 768         u32 nritems;
 769         int ret;
 770         int recow;
 771         int err = -ENOENT;
 772
 773         key.objectid = bytenr;
 774         if (parent) {
 775                 key.type = BTRFS_SHARED_DATA_REF_KEY;
 776                 key.offset = parent;
 777         } else {
 778                 key.type = BTRFS_EXTENT_DATA_REF_KEY;
 779                 key.offset = hash_extent_data_ref(root_objectid,
 780                                                   owner, offset);
 781         }
 782 again:
 783         recow = 0;
 784         ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 785         if (ret < 0) {
 786                 err = ret;
 787                 goto fail;
 788         }
 789
 790         if (parent) {
 791                 if (!ret)
 792                         return 0;
 793 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
 794                 key.type = BTRFS_EXTENT_REF_V0_KEY;
 795                 btrfs_release_path(root, path);
 796                 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 797                 if (ret < 0) {
 798                         err = ret;
 799                         goto fail;
 800                 }
 801                 if (!ret)
 802                         return 0;
 803 #endif
 804                 goto fail;
 805         }
 806
 807         leaf = path->nodes[0];
 808         nritems = btrfs_header_nritems(leaf);
 809         while (1) {
 810                 if (path->slots[0] >= nritems) {
 811                         ret = btrfs_next_leaf(root, path);
 812                         if (ret < 0)
 813                                 err = ret;
 814                         if (ret)
 815                                 goto fail;
 816
 817                         leaf = path->nodes[0];
 818                         nritems = btrfs_header_nritems(leaf);
 819                         recow = 1;
 820                 }
 821
 822                 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 823                 if (key.objectid != bytenr ||
 824                     key.type != BTRFS_EXTENT_DATA_REF_KEY)
 825                         goto fail;
 826
 827                 ref = btrfs_item_ptr(leaf, path->slots[0],
 828                                      struct btrfs_extent_data_ref);
 829
 830                 if (match_extent_data_ref(leaf, ref, root_objectid,
 831                                           owner, offset)) {
 832                         if (recow) {
 833                                 btrfs_release_path(root, path);
 834                                 goto again;
 835                         }
 836                         err = 0;
 837                         break;
 838                 }
 839                 path->slots[0]++;
 840         }
 841 fail:
 842         return err;
 843 }
 844
 845 static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
 846                                            struct btrfs_root *root,
 847                                            struct btrfs_path *path,
 848                                            u64 bytenr, u64 parent,
 849                                            u64 root_objectid, u64 owner,
 850                                            u64 offset, int refs_to_add)
 851 {
 852         struct btrfs_key key;
 853         struct extent_buffer *leaf;
 854         u32 size;
 855         u32 num_refs;
 856         int ret;
 857
 858         key.objectid = bytenr;
 859         if (parent) {
 860                 key.type = BTRFS_SHARED_DATA_REF_KEY;
 861                 key.offset = parent;
 862                 size = sizeof(struct btrfs_shared_data_ref);
 863         } else {
 864                 key.type = BTRFS_EXTENT_DATA_REF_KEY;
 865                 key.offset = hash_extent_data_ref(root_objectid,
 866                                                   owner, offset);
 867                 size = sizeof(struct btrfs_extent_data_ref);
 868         }
 869
 870         ret = btrfs_insert_empty_item(trans, root, path, &key, size);
 871         if (ret && ret != -EEXIST)
 872                 goto fail;
 873
 874         leaf = path->nodes[0];
 875         if (parent) {
 876                 struct btrfs_shared_data_ref *ref;
 877                 ref = btrfs_item_ptr(leaf, path->slots[0],
 878                                      struct btrfs_shared_data_ref);
 879                 if (ret == 0) {
 880                         btrfs_set_shared_data_ref_count(leaf, ref, refs_to_add);
 881                 } else {
 882                         num_refs = btrfs_shared_data_ref_count(leaf, ref);
 883                         num_refs += refs_to_add;
 884                         btrfs_set_shared_data_ref_count(leaf, ref, num_refs);
 885                 }
 886         } else {
 887                 struct btrfs_extent_data_ref *ref;
 888                 while (ret == -EEXIST) {
 889                         ref = btrfs_item_ptr(leaf, path->slots[0],
 890                                              struct btrfs_extent_data_ref);
 891                         if (match_extent_data_ref(leaf, ref, root_objectid,
 892                                                   owner, offset))
 893                                 break;
 894                         btrfs_release_path(root, path);
 895                         key.offset++;
 896                         ret = btrfs_insert_empty_item(trans, root, path, &key,
 897                                                       size);
 898                         if (ret && ret != -EEXIST)
 899                                 goto fail;
 900
 901                         leaf = path->nodes[0];
 902                 }
 903                 ref = btrfs_item_ptr(leaf, path->slots[0],
 904                                      struct btrfs_extent_data_ref);
 905                 if (ret == 0) {
 906                         btrfs_set_extent_data_ref_root(leaf, ref,
 907                                                        root_objectid);
 908                         btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
 909                         btrfs_set_extent_data_ref_offset(leaf, ref, offset);
 910                         btrfs_set_extent_data_ref_count(leaf, ref, refs_to_add);
 911                 } else {
 912                         num_refs = btrfs_extent_data_ref_count(leaf, ref);
 913                         num_refs += refs_to_add;
 914                         btrfs_set_extent_data_ref_count(leaf, ref, num_refs);
 915                 }
 916         }
 917         btrfs_mark_buffer_dirty(leaf);
 918         ret = 0;
 919 fail:
 920         btrfs_release_path(root, path);
 921         return ret;
 922 }
 923
 924 static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
 925                                            struct btrfs_root *root,
 926                                            struct btrfs_path *path,
 927                                            int refs_to_drop)
 928 {
 929         struct btrfs_key key;
 930         struct btrfs_extent_data_ref *ref1 = NULL;
 931         struct btrfs_shared_data_ref *ref2 = NULL;
 932         struct extent_buffer *leaf;
 933         u32 num_refs = 0;
 934         int ret = 0;
 935
 936         leaf = path->nodes[0];
 937         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 938
 939         if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
 940                 ref1 = btrfs_item_ptr(leaf, path->slots[0],
 941                                       struct btrfs_extent_data_ref);
 942                 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
 943         } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
 944                 ref2 = btrfs_item_ptr(leaf, path->slots[0],
 945                                       struct btrfs_shared_data_ref);
 946                 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
 947 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
 948         } else if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
 949                 struct btrfs_extent_ref_v0 *ref0;
 950                 ref0 = btrfs_item_ptr(leaf, path->slots[0],
 951                                       struct btrfs_extent_ref_v0);
 952                 num_refs = btrfs_ref_count_v0(leaf, ref0);
 953 #endif
 954         } else {
 955                 BUG();
 956         }
 957
 958         BUG_ON(num_refs < refs_to_drop);
 959         num_refs -= refs_to_drop;
 960
 961         if (num_refs == 0) {
 962                 ret = btrfs_del_item(trans, root, path);
 963         } else {
 964                 if (key.type == BTRFS_EXTENT_DATA_REF_KEY)
 965                         btrfs_set_extent_data_ref_count(leaf, ref1, num_refs);
 966                 else if (key.type == BTRFS_SHARED_DATA_REF_KEY)
 967                         btrfs_set_shared_data_ref_count(leaf, ref2, num_refs);
 968 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
 969                 else {
 970                         struct btrfs_extent_ref_v0 *ref0;
 971                         ref0 = btrfs_item_ptr(leaf, path->slots[0],
 972                                         struct btrfs_extent_ref_v0);
 973                         btrfs_set_ref_count_v0(leaf, ref0, num_refs);
 974                 }
 975 #endif
 976                 btrfs_mark_buffer_dirty(leaf);
 977         }
 978         return ret;
 979 }
 980
 981 static noinline u32 extent_data_ref_count(struct btrfs_root *root,
 982                                           struct btrfs_path *path,
 983                                           struct btrfs_extent_inline_ref *iref)
 984 {
 985         struct btrfs_key key;
 986         struct extent_buffer *leaf;
 987         struct btrfs_extent_data_ref *ref1;
 988         struct btrfs_shared_data_ref *ref2;
 989         u32 num_refs = 0;
 990
 991         leaf = path->nodes[0];
 992         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 993         if (iref) {
 994                 if (btrfs_extent_inline_ref_type(leaf, iref) ==
 995                     BTRFS_EXTENT_DATA_REF_KEY) {
 996                         ref1 = (struct btrfs_extent_data_ref *)(&iref->offset);
 997                         num_refs = btrfs_extent_data_ref_count(leaf, ref1);
 998                 } else {
 999                         ref2 = (struct btrfs_shared_data_ref *)(iref + 1);
1000                         num_refs = btrfs_shared_data_ref_count(leaf, ref2);
1001                 }
1002         } else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
1003                 ref1 = btrfs_item_ptr(leaf, path->slots[0],
1004                                       struct btrfs_extent_data_ref);
1005                 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
1006         } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
1007                 ref2 = btrfs_item_ptr(leaf, path->slots[0],
1008                                       struct btrfs_shared_data_ref);
1009                 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
1010 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1011         } else if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
1012                 struct btrfs_extent_ref_v0 *ref0;
1013                 ref0 = btrfs_item_ptr(leaf, path->slots[0],
1014                                       struct btrfs_extent_ref_v0);
1015                 num_refs = btrfs_ref_count_v0(leaf, ref0);
1016 #endif
1017         } else {
1018                 WARN_ON(1);
1019         }
1020         return num_refs;
1021 }
1022
1023 static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans,
1024                                           struct btrfs_root *root,
1025                                           struct btrfs_path *path,
1026                                           u64 bytenr, u64 parent,
1027                                           u64 root_objectid)
1028 {
1029         struct btrfs_key key;
1030         int ret;
1031
1032         key.objectid = bytenr;
1033         if (parent) {
1034                 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
1035                 key.offset = parent;
1036         } else {
1037                 key.type = BTRFS_TREE_BLOCK_REF_KEY;
1038                 key.offset = root_objectid;
1039         }
1040
1041         ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1042         if (ret > 0)
1043                 ret = -ENOENT;
1044 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1045         if (ret == -ENOENT && parent) {
1046                 btrfs_release_path(root, path);
1047                 key.type = BTRFS_EXTENT_REF_V0_KEY;
1048                 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1049                 if (ret > 0)
1050                         ret = -ENOENT;
1051         }
1052 #endif
1053         return ret;
1054 }
1055
1056 static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans,
1057                                           struct btrfs_root *root,
1058                                           struct btrfs_path *path,
1059                                           u64 bytenr, u64 parent,
1060                                           u64 root_objectid)
1061 {
1062         struct btrfs_key key;
1063         int ret;
1064
1065         key.objectid = bytenr;
1066         if (parent) {
1067                 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
1068                 key.offset = parent;
1069         } else {
1070                 key.type = BTRFS_TREE_BLOCK_REF_KEY;
1071                 key.offset = root_objectid;
1072         }
1073
1074         ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
1075         btrfs_release_path(root, path);
1076         return ret;
1077 }
1078
1079 static inline int extent_ref_type(u64 parent, u64 owner)
1080 {
1081         int type;
1082         if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1083                 if (parent > 0)
1084                         type = BTRFS_SHARED_BLOCK_REF_KEY;
1085                 else
1086                         type = BTRFS_TREE_BLOCK_REF_KEY;
1087         } else {
1088                 if (parent > 0)
1089                         type = BTRFS_SHARED_DATA_REF_KEY;
1090                 else
1091                         type = BTRFS_EXTENT_DATA_REF_KEY;
1092         }
1093         return type;
1094 }
1095
1096 static int find_next_key(struct btrfs_path *path, int level,
1097                          struct btrfs_key *key)
1098
1099 {
1100         for (; level < BTRFS_MAX_LEVEL; level++) {
1101                 if (!path->nodes[level])
1102                         break;
1103                 if (path->slots[level] + 1 >=
1104                     btrfs_header_nritems(path->nodes[level]))
1105                         continue;
1106                 if (level == 0)
1107                         btrfs_item_key_to_cpu(path->nodes[level], key,
1108                                               path->slots[level] + 1);
1109                 else
1110                         btrfs_node_key_to_cpu(path->nodes[level], key,
1111                                               path->slots[level] + 1);
1112                 return 0;
1113         }
1114         return 1;
1115 }
1116
1117 /*
1118  * look for inline back ref. if back ref is found, *ref_ret is set
1119  * to the address of inline back ref, and 0 is returned.
1120  *
1121  * if back ref isn't found, *ref_ret is set to the address where it
1122  * should be inserted, and -ENOENT is returned.
1123  *
1124  * if insert is true and there are too many inline back refs, the path
1125  * points to the extent item, and -EAGAIN is returned.
1126  *
1127  * NOTE: inline back refs are ordered in the same way that back ref
1128  *       items in the tree are ordered.
1129  */
1130 static noinline_for_stack
1131 int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
1132                                  struct btrfs_root *root,
1133                                  struct btrfs_path *path,
1134                                  struct btrfs_extent_inline_ref **ref_ret,
1135                                  u64 bytenr, u64 num_bytes,
1136                                  u64 parent, u64 root_objectid,
1137                                  u64 owner, u64 offset, int insert)
1138 {
1139         struct btrfs_key key;
1140         struct extent_buffer *leaf;
1141         struct btrfs_extent_item *ei;
1142         struct btrfs_extent_inline_ref *iref;
1143         u64 flags;
1144         u64 item_size;
1145         unsigned long ptr;
1146         unsigned long end;
1147         int extra_size;
1148         int type;
1149         int want;
1150         int ret;
1151         int err = 0;
1152
1153         key.objectid = bytenr;
1154         key.type = BTRFS_EXTENT_ITEM_KEY;
1155         key.offset = num_bytes;
1156
1157         want = extent_ref_type(parent, owner);
1158         if (insert) {
1159                 extra_size = btrfs_extent_inline_ref_size(want);
1160                 path->keep_locks = 1;
1161         } else
1162                 extra_size = -1;
1163         ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
1164         if (ret < 0) {
1165                 err = ret;
1166                 goto out;
1167         }
1168         BUG_ON(ret);
1169
1170         leaf = path->nodes[0];
1171         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1172 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1173         if (item_size < sizeof(*ei)) {
1174                 if (!insert) {
1175                         err = -ENOENT;
1176                         goto out;
1177                 }
1178                 ret = convert_extent_item_v0(trans, root, path, owner,
1179                                              extra_size);
1180                 if (ret < 0) {
1181                         err = ret;
1182                         goto out;
1183                 }
1184                 leaf = path->nodes[0];
1185                 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1186         }
1187 #endif
1188         BUG_ON(item_size < sizeof(*ei));
1189
1190         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1191         flags = btrfs_extent_flags(leaf, ei);
1192
1193         ptr = (unsigned long)(ei + 1);
1194         end = (unsigned long)ei + item_size;
1195
1196         if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1197                 ptr += sizeof(struct btrfs_tree_block_info);
1198                 BUG_ON(ptr > end);
1199         } else {
1200                 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
1201         }
1202
1203         err = -ENOENT;
1204         while (1) {
1205                 if (ptr >= end) {
1206                         WARN_ON(ptr > end);
1207                         break;
1208                 }
1209                 iref = (struct btrfs_extent_inline_ref *)ptr;
1210                 type = btrfs_extent_inline_ref_type(leaf, iref);
1211                 if (want < type)
1212                         break;
1213                 if (want > type) {
1214                         ptr += btrfs_extent_inline_ref_size(type);
1215                         continue;
1216                 }
1217
1218                 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1219                         struct btrfs_extent_data_ref *dref;
1220                         dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1221                         if (match_extent_data_ref(leaf, dref, root_objectid,
1222                                                   owner, offset)) {
1223                                 err = 0;
1224                                 break;
1225                         }
1226                         if (hash_extent_data_ref_item(leaf, dref) <
1227                             hash_extent_data_ref(root_objectid, owner, offset))
1228                                 break;
1229                 } else {
1230                         u64 ref_offset;
1231                         ref_offset = btrfs_extent_inline_ref_offset(leaf, iref);
1232                         if (parent > 0) {
1233                                 if (parent == ref_offset) {
1234                                         err = 0;
1235                                         break;
1236                                 }
1237                                 if (ref_offset < parent)
1238                                         break;
1239                         } else {
1240                                 if (root_objectid == ref_offset) {
1241                                         err = 0;
1242                                         break;
1243                                 }
1244                                 if (ref_offset < root_objectid)
1245                                         break;
1246                         }
1247                 }
1248                 ptr += btrfs_extent_inline_ref_size(type);
1249         }
1250         if (err == -ENOENT && insert) {
1251                 if (item_size + extra_size >=
1252                     BTRFS_MAX_EXTENT_ITEM_SIZE(root)) {
1253                         err = -EAGAIN;
1254                         goto out;
1255                 }
1256                 /*
1257                  * To add new inline back ref, we have to make sure
1258                  * there is no corresponding back ref item.
1259                  * For simplicity, we just do not add new inline back
1260                  * ref if there is any kind of item for this block
1261                  */
1262                 if (find_next_key(path, 0, &key) == 0 &&
1263                     key.objectid == bytenr &&
1264                     key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
1265                         err = -EAGAIN;
1266                         goto out;
1267                 }
1268         }
1269         *ref_ret = (struct btrfs_extent_inline_ref *)ptr;
1270 out:
1271         if (insert) {
1272                 path->keep_locks = 0;
1273                 btrfs_unlock_up_safe(path, 1);
1274         }
1275         return err;
1276 }
1277
1278 /*
1279  * helper to add new inline back ref
1280  */
1281 static noinline_for_stack
1282 int setup_inline_extent_backref(struct btrfs_trans_handle *trans,
1283                                 struct btrfs_root *root,
1284                                 struct btrfs_path *path,
1285                                 struct btrfs_extent_inline_ref *iref,
1286                                 u64 parent, u64 root_objectid,
1287                                 u64 owner, u64 offset, int refs_to_add,
1288                                 struct btrfs_delayed_extent_op *extent_op)
1289 {
1290         struct extent_buffer *leaf;
1291         struct btrfs_extent_item *ei;
1292         unsigned long ptr;
1293         unsigned long end;
1294         unsigned long item_offset;
1295         u64 refs;
1296         int size;
1297         int type;
1298         int ret;
1299
1300         leaf = path->nodes[0];
1301         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1302         item_offset = (unsigned long)iref - (unsigned long)ei;
1303
1304         type = extent_ref_type(parent, owner);
1305         size = btrfs_extent_inline_ref_size(type);
1306
1307         ret = btrfs_extend_item(trans, root, path, size);
1308         BUG_ON(ret);
1309
1310         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1311         refs = btrfs_extent_refs(leaf, ei);
1312         refs += refs_to_add;
1313         btrfs_set_extent_refs(leaf, ei, refs);
1314         if (extent_op)
1315                 __run_delayed_extent_op(extent_op, leaf, ei);
1316
1317         ptr = (unsigned long)ei + item_offset;
1318         end = (unsigned long)ei + btrfs_item_size_nr(leaf, path->slots[0]);
1319         if (ptr < end - size)
1320                 memmove_extent_buffer(leaf, ptr + size, ptr,
1321                                       end - size - ptr);
1322
1323         iref = (struct btrfs_extent_inline_ref *)ptr;
1324         btrfs_set_extent_inline_ref_type(leaf, iref, type);
1325         if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1326                 struct btrfs_extent_data_ref *dref;
1327                 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1328                 btrfs_set_extent_data_ref_root(leaf, dref, root_objectid);
1329                 btrfs_set_extent_data_ref_objectid(leaf, dref, owner);
1330                 btrfs_set_extent_data_ref_offset(leaf, dref, offset);
1331                 btrfs_set_extent_data_ref_count(leaf, dref, refs_to_add);
1332         } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1333                 struct btrfs_shared_data_ref *sref;
1334                 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1335                 btrfs_set_shared_data_ref_count(leaf, sref, refs_to_add);
1336                 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1337         } else if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
1338                 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1339         } else {
1340                 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
1341         }
1342         btrfs_mark_buffer_dirty(leaf);
1343         return 0;
1344 }
1345
1346 static int lookup_extent_backref(struct btrfs_trans_handle *trans,
1347                                  struct btrfs_root *root,
1348                                  struct btrfs_path *path,
1349                                  struct btrfs_extent_inline_ref **ref_ret,
1350                                  u64 bytenr, u64 num_bytes, u64 parent,
1351                                  u64 root_objectid, u64 owner, u64 offset)
1352 {
1353         int ret;
1354
1355         ret = lookup_inline_extent_backref(trans, root, path, ref_ret,
1356                                            bytenr, num_bytes, parent,
1357                                            root_objectid, owner, offset, 0);
1358         if (ret != -ENOENT)
1359                 return ret;
1360
1361         btrfs_release_path(root, path);
1362         *ref_ret = NULL;
1363
1364         if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1365                 ret = lookup_tree_block_ref(trans, root, path, bytenr, parent,
1366                                             root_objectid);
1367         } else {
1368                 ret = lookup_extent_data_ref(trans, root, path, bytenr, parent,
1369                                              root_objectid, owner, offset);
1370         }
1371         return ret;
1372 }
1373
1374 /*
1375  * helper to update/remove inline back ref
1376  */
1377 static noinline_for_stack
1378 int update_inline_extent_backref(struct btrfs_trans_handle *trans,
1379                                  struct btrfs_root *root,
1380                                  struct btrfs_path *path,
1381                                  struct btrfs_extent_inline_ref *iref,
1382                                  int refs_to_mod,
1383                                  struct btrfs_delayed_extent_op *extent_op)
1384 {
1385         struct extent_buffer *leaf;
1386         struct btrfs_extent_item *ei;
1387         struct btrfs_extent_data_ref *dref = NULL;
1388         struct btrfs_shared_data_ref *sref = NULL;
1389         unsigned long ptr;
1390         unsigned long end;
1391         u32 item_size;
1392         int size;
1393         int type;
1394         int ret;
1395         u64 refs;
1396
1397         leaf = path->nodes[0];
1398         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1399         refs = btrfs_extent_refs(leaf, ei);
1400         WARN_ON(refs_to_mod < 0 && refs + refs_to_mod <= 0);
1401         refs += refs_to_mod;
1402         btrfs_set_extent_refs(leaf, ei, refs);
1403         if (extent_op)
1404                 __run_delayed_extent_op(extent_op, leaf, ei);
1405
1406         type = btrfs_extent_inline_ref_type(leaf, iref);
1407
1408         if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1409                 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1410                 refs = btrfs_extent_data_ref_count(leaf, dref);
1411         } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1412                 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1413                 refs = btrfs_shared_data_ref_count(leaf, sref);
1414         } else {
1415                 refs = 1;
1416                 BUG_ON(refs_to_mod != -1);
1417         }
1418
1419         BUG_ON(refs_to_mod < 0 && refs < -refs_to_mod);
1420         refs += refs_to_mod;
1421
1422         if (refs > 0) {
1423                 if (type == BTRFS_EXTENT_DATA_REF_KEY)
1424                         btrfs_set_extent_data_ref_count(leaf, dref, refs);
1425                 else
1426                         btrfs_set_shared_data_ref_count(leaf, sref, refs);
1427         } else {
1428                 size =  btrfs_extent_inline_ref_size(type);
1429                 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1430                 ptr = (unsigned long)iref;
1431                 end = (unsigned long)ei + item_size;
1432                 if (ptr + size < end)
1433                         memmove_extent_buffer(leaf, ptr, ptr + size,
1434                                               end - ptr - size);
1435                 item_size -= size;
1436                 ret = btrfs_truncate_item(trans, root, path, item_size, 1);
1437                 BUG_ON(ret);
1438         }
1439         btrfs_mark_buffer_dirty(leaf);
1440         return 0;
1441 }
1442
1443 static noinline_for_stack
1444 int insert_inline_extent_backref(struct btrfs_trans_handle *trans,
1445                                  struct btrfs_root *root,
1446                                  struct btrfs_path *path,
1447                                  u64 bytenr, u64 num_bytes, u64 parent,
1448                                  u64 root_objectid, u64 owner,
1449                                  u64 offset, int refs_to_add,
1450                                  struct btrfs_delayed_extent_op *extent_op)
1451 {
1452         struct btrfs_extent_inline_ref *iref;
1453         int ret;
1454
1455         ret = lookup_inline_extent_backref(trans, root, path, &iref,
1456                                            bytenr, num_bytes, parent,
1457                                            root_objectid, owner, offset, 1);
1458         if (ret == 0) {
1459                 BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID);
1460                 ret = update_inline_extent_backref(trans, root, path, iref,
1461                                                    refs_to_add, extent_op);
1462         } else if (ret == -ENOENT) {
1463                 ret = setup_inline_extent_backref(trans, root, path, iref,
1464                                                   parent, root_objectid,
1465                                                   owner, offset, refs_to_add,
1466                                                   extent_op);
1467         }
1468         return ret;
1469 }
1470
1471 static int insert_extent_backref(struct btrfs_trans_handle *trans,
1472                                  struct btrfs_root *root,
1473                                  struct btrfs_path *path,
1474                                  u64 bytenr, u64 parent, u64 root_objectid,
1475                                  u64 owner, u64 offset, int refs_to_add)
1476 {
1477         int ret;
1478         if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1479                 BUG_ON(refs_to_add != 1);
1480                 ret = insert_tree_block_ref(trans, root, path, bytenr,
1481                                             parent, root_objectid);
1482         } else {
1483                 ret = insert_extent_data_ref(trans, root, path, bytenr,
1484                                              parent, root_objectid,
1485                                              owner, offset, refs_to_add);
1486         }
1487         return ret;
1488 }
1489
1490 static int remove_extent_backref(struct btrfs_trans_handle *trans,
1491                                  struct btrfs_root *root,
1492                                  struct btrfs_path *path,
1493                                  struct btrfs_extent_inline_ref *iref,
1494                                  int refs_to_drop, int is_data)
1495 {
1496         int ret;
1497
1498         BUG_ON(!is_data && refs_to_drop != 1);
1499         if (iref) {
1500                 ret = update_inline_extent_backref(trans, root, path, iref,
1501                                                    -refs_to_drop, NULL);
1502         } else if (is_data) {
1503                 ret = remove_extent_data_ref(trans, root, path, refs_to_drop);
1504         } else {
1505                 ret = btrfs_del_item(trans, root, path);
1506         }
1507         return ret;
1508 }
1509
1510 #ifdef BIO_RW_DISCARD
1511 static void btrfs_issue_discard(struct block_device *bdev,
1512                                 u64 start, u64 len)
1513 {
1514         blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_KERNEL,
1515                              DISCARD_FL_BARRIER);
1516 }
1517 #endif
1518
1519 static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
1520                                 u64 num_bytes)
1521 {
1522 #ifdef BIO_RW_DISCARD
1523         int ret;
1524         u64 map_length = num_bytes;
1525         struct btrfs_multi_bio *multi = NULL;
1526
1527         /* Tell the block device(s) that the sectors can be discarded */
1528         ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
1529                               bytenr, &map_length, &multi, 0);
1530         if (!ret) {
1531                 struct btrfs_bio_stripe *stripe = multi->stripes;
1532                 int i;
1533
1534                 if (map_length > num_bytes)
1535                         map_length = num_bytes;
1536
1537                 for (i = 0; i < multi->num_stripes; i++, stripe++) {
1538                         btrfs_issue_discard(stripe->dev->bdev,
1539                                             stripe->physical,
1540                                             map_length);
1541                 }
1542                 kfree(multi);
1543         }
1544
1545         return ret;
1546 #else
1547         return 0;
1548 #endif
1549 }
1550
1551 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1552                          struct btrfs_root *root,
1553                          u64 bytenr, u64 num_bytes, u64 parent,
1554                          u64 root_objectid, u64 owner, u64 offset)
1555 {
1556         int ret;
1557         BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID &&
1558                root_objectid == BTRFS_TREE_LOG_OBJECTID);
1559
1560         if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1561                 ret = btrfs_add_delayed_tree_ref(trans, bytenr, num_bytes,
1562                                         parent, root_objectid, (int)owner,
1563                                         BTRFS_ADD_DELAYED_REF, NULL);
1564         } else {
1565                 ret = btrfs_add_delayed_data_ref(trans, bytenr, num_bytes,
1566                                         parent, root_objectid, owner, offset,
1567                                         BTRFS_ADD_DELAYED_REF, NULL);
1568         }
1569         return ret;
1570 }
1571
1572 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1573                                   struct btrfs_root *root,
1574                                   u64 bytenr, u64 num_bytes,
1575                                   u64 parent, u64 root_objectid,
1576                                   u64 owner, u64 offset, int refs_to_add,
1577                                   struct btrfs_delayed_extent_op *extent_op)
1578 {
1579         struct btrfs_path *path;
1580         struct extent_buffer *leaf;
1581         struct btrfs_extent_item *item;
1582         u64 refs;
1583         int ret;
1584         int err = 0;
1585
1586         path = btrfs_alloc_path();
1587         if (!path)
1588                 return -ENOMEM;
1589
1590         path->reada = 1;
1591         path->leave_spinning = 1;
1592         /* this will setup the path even if it fails to insert the back ref */
1593         ret = insert_inline_extent_backref(trans, root->fs_info->extent_root,
1594                                            path, bytenr, num_bytes, parent,
1595                                            root_objectid, owner, offset,
1596                                            refs_to_add, extent_op);
1597         if (ret == 0)
1598                 goto out;
1599
1600         if (ret != -EAGAIN) {
1601                 err = ret;
1602                 goto out;
1603         }
1604
1605         leaf = path->nodes[0];
1606         item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1607         refs = btrfs_extent_refs(leaf, item);
1608         btrfs_set_extent_refs(leaf, item, refs + refs_to_add);
1609         if (extent_op)
1610                 __run_delayed_extent_op(extent_op, leaf, item);
1611
1612         btrfs_mark_buffer_dirty(leaf);
1613         btrfs_release_path(root->fs_info->extent_root, path);
1614
1615         path->reada = 1;
1616         path->leave_spinning = 1;
1617
1618         /* now insert the actual backref */
1619         ret = insert_extent_backref(trans, root->fs_info->extent_root,
1620                                     path, bytenr, parent, root_objectid,
1621                                     owner, offset, refs_to_add);
1622         BUG_ON(ret);
1623 out:
1624         btrfs_free_path(path);
1625         return err;
1626 }
1627
1628 static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
1629                                 struct btrfs_root *root,
1630                                 struct btrfs_delayed_ref_node *node,
1631                                 struct btrfs_delayed_extent_op *extent_op,
1632                                 int insert_reserved)
1633 {
1634         int ret = 0;
1635         struct btrfs_delayed_data_ref *ref;
1636         struct btrfs_key ins;
1637         u64 parent = 0;
1638         u64 ref_root = 0;
1639         u64 flags = 0;
1640
1641         ins.objectid = node->bytenr;
1642         ins.offset = node->num_bytes;
1643         ins.type = BTRFS_EXTENT_ITEM_KEY;
1644
1645         ref = btrfs_delayed_node_to_data_ref(node);
1646         if (node->type == BTRFS_SHARED_DATA_REF_KEY)
1647                 parent = ref->parent;
1648         else
1649                 ref_root = ref->root;
1650
1651         if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1652                 if (extent_op) {
1653                         BUG_ON(extent_op->update_key);
1654                         flags |= extent_op->flags_to_set;
1655                 }
1656                 ret = alloc_reserved_file_extent(trans, root,
1657                                                  parent, ref_root, flags,
1658                                                  ref->objectid, ref->offset,
1659                                                  &ins, node->ref_mod);
1660                 update_reserved_extents(root, ins.objectid, ins.offset, 0);
1661         } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1662                 ret = __btrfs_inc_extent_ref(trans, root, node->bytenr,
1663                                              node->num_bytes, parent,
1664                                              ref_root, ref->objectid,
1665                                              ref->offset, node->ref_mod,
1666                                              extent_op);
1667         } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1668                 ret = __btrfs_free_extent(trans, root, node->bytenr,
1669                                           node->num_bytes, parent,
1670                                           ref_root, ref->objectid,
1671                                           ref->offset, node->ref_mod,
1672                                           extent_op);
1673         } else {
1674                 BUG();
1675         }
1676         return ret;
1677 }
1678
1679 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
1680                                     struct extent_buffer *leaf,
1681                                     struct btrfs_extent_item *ei)
1682 {
1683         u64 flags = btrfs_extent_flags(leaf, ei);
1684         if (extent_op->update_flags) {
1685                 flags |= extent_op->flags_to_set;
1686                 btrfs_set_extent_flags(leaf, ei, flags);
1687         }
1688
1689         if (extent_op->update_key) {
1690                 struct btrfs_tree_block_info *bi;
1691                 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
1692                 bi = (struct btrfs_tree_block_info *)(ei + 1);
1693                 btrfs_set_tree_block_key(leaf, bi, &extent_op->key);
1694         }
1695 }
1696
1697 static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
1698                                  struct btrfs_root *root,
1699                                  struct btrfs_delayed_ref_node *node,
1700                                  struct btrfs_delayed_extent_op *extent_op)
1701 {
1702         struct btrfs_key key;
1703         struct btrfs_path *path;
1704         struct btrfs_extent_item *ei;
1705         struct extent_buffer *leaf;
1706         u32 item_size;
1707         int ret;
1708         int err = 0;
1709
1710         path = btrfs_alloc_path();
1711         if (!path)
1712                 return -ENOMEM;
1713
1714         key.objectid = node->bytenr;
1715         key.type = BTRFS_EXTENT_ITEM_KEY;
1716         key.offset = node->num_bytes;
1717
1718         path->reada = 1;
1719         path->leave_spinning = 1;
1720         ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key,
1721                                 path, 0, 1);
1722         if (ret < 0) {
1723                 err = ret;
1724                 goto out;
1725         }
1726         if (ret > 0) {
1727                 err = -EIO;
1728                 goto out;
1729         }
1730
1731         leaf = path->nodes[0];
1732         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1733 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1734         if (item_size < sizeof(*ei)) {
1735                 ret = convert_extent_item_v0(trans, root->fs_info->extent_root,
1736                                              path, (u64)-1, 0);
1737                 if (ret < 0) {
1738                         err = ret;
1739                         goto out;
1740                 }
1741                 leaf = path->nodes[0];
1742                 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1743         }
1744 #endif
1745         BUG_ON(item_size < sizeof(*ei));
1746         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1747         __run_delayed_extent_op(extent_op, leaf, ei);
1748
1749         btrfs_mark_buffer_dirty(leaf);
1750 out:
1751         btrfs_free_path(path);
1752         return err;
1753 }
1754
1755 static int run_delayed_tree_ref(struct btrfs_trans_handle *trans,
1756                                 struct btrfs_root *root,
1757                                 struct btrfs_delayed_ref_node *node,
1758                                 struct btrfs_delayed_extent_op *extent_op,
1759                                 int insert_reserved)
1760 {
1761         int ret = 0;
1762         struct btrfs_delayed_tree_ref *ref;
1763         struct btrfs_key ins;
1764         u64 parent = 0;
1765         u64 ref_root = 0;
1766
1767         ins.objectid = node->bytenr;
1768         ins.offset = node->num_bytes;
1769         ins.type = BTRFS_EXTENT_ITEM_KEY;
1770
1771         ref = btrfs_delayed_node_to_tree_ref(node);
1772         if (node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1773                 parent = ref->parent;
1774         else
1775                 ref_root = ref->root;
1776
1777         BUG_ON(node->ref_mod != 1);
1778         if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1779                 BUG_ON(!extent_op || !extent_op->update_flags ||
1780                        !extent_op->update_key);
1781                 ret = alloc_reserved_tree_block(trans, root,
1782                                                 parent, ref_root,
1783                                                 extent_op->flags_to_set,
1784                                                 &extent_op->key,
1785                                                 ref->level, &ins);
1786                 update_reserved_extents(root, ins.objectid, ins.offset, 0);
1787         } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1788                 ret = __btrfs_inc_extent_ref(trans, root, node->bytenr,
1789                                              node->num_bytes, parent, ref_root,
1790                                              ref->level, 0, 1, extent_op);
1791         } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1792                 ret = __btrfs_free_extent(trans, root, node->bytenr,
1793                                           node->num_bytes, parent, ref_root,
1794                                           ref->level, 0, 1, extent_op);
1795         } else {
1796                 BUG();
1797         }
1798         return ret;
1799 }
1800
1801
1802 /* helper function to actually process a single delayed ref entry */
1803 static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
1804                                struct btrfs_root *root,
1805                                struct btrfs_delayed_ref_node *node,
1806                                struct btrfs_delayed_extent_op *extent_op,
1807                                int insert_reserved)
1808 {
1809         int ret;
1810         if (btrfs_delayed_ref_is_head(node)) {
1811                 struct btrfs_delayed_ref_head *head;
1812                 /*
1813                  * we've hit the end of the chain and we were supposed
1814                  * to insert this extent into the tree.  But, it got
1815                  * deleted before we ever needed to insert it, so all
1816                  * we have to do is clean up the accounting
1817                  */
1818                 BUG_ON(extent_op);
1819                 head = btrfs_delayed_node_to_head(node);
1820                 if (insert_reserved) {
1821                         if (head->is_data) {
1822                                 ret = btrfs_del_csums(trans, root,
1823                                                       node->bytenr,
1824                                                       node->num_bytes);
1825                                 BUG_ON(ret);
1826                         }
1827                         btrfs_update_pinned_extents(root, node->bytenr,
1828                                                     node->num_bytes, 1);
1829                         update_reserved_extents(root, node->bytenr,
1830                                                 node->num_bytes, 0);
1831                 }
1832                 mutex_unlock(&head->mutex);
1833                 return 0;
1834         }
1835
1836         if (node->type == BTRFS_TREE_BLOCK_REF_KEY ||
1837             node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1838                 ret = run_delayed_tree_ref(trans, root, node, extent_op,
1839                                            insert_reserved);
1840         else if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
1841                  node->type == BTRFS_SHARED_DATA_REF_KEY)
1842                 ret = run_delayed_data_ref(trans, root, node, extent_op,
1843                                            insert_reserved);
1844         else
1845                 BUG();
1846         return ret;
1847 }
1848
1849 static noinline struct btrfs_delayed_ref_node *
1850 select_delayed_ref(struct btrfs_delayed_ref_head *head)
1851 {
1852         struct rb_node *node;
1853         struct btrfs_delayed_ref_node *ref;
1854         int action = BTRFS_ADD_DELAYED_REF;
1855 again:
1856         /*
1857          * select delayed ref of type BTRFS_ADD_DELAYED_REF first.
1858          * this prevents ref count from going down to zero when
1859          * there still are pending delayed ref.
1860          */
1861         node = rb_prev(&head->node.rb_node);
1862         while (1) {
1863                 if (!node)
1864                         break;
1865                 ref = rb_entry(node, struct btrfs_delayed_ref_node,
1866                                 rb_node);
1867                 if (ref->bytenr != head->node.bytenr)
1868                         break;
1869                 if (ref->action == action)
1870                         return ref;
1871                 node = rb_prev(node);
1872         }
1873         if (action == BTRFS_ADD_DELAYED_REF) {
1874                 action = BTRFS_DROP_DELAYED_REF;
1875                 goto again;
1876         }
1877         return NULL;
1878 }
1879
1880 static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
1881                                        struct btrfs_root *root,
1882                                        struct list_head *cluster)
1883 {
1884         struct btrfs_delayed_ref_root *delayed_refs;
1885         struct btrfs_delayed_ref_node *ref;
1886         struct btrfs_delayed_ref_head *locked_ref = NULL;
1887         struct btrfs_delayed_extent_op *extent_op;
1888         int ret;
1889         int count = 0;
1890         int must_insert_reserved = 0;
1891
1892         delayed_refs = &trans->transaction->delayed_refs;
1893         while (1) {
1894                 if (!locked_ref) {
1895                         /* pick a new head ref from the cluster list */
1896                         if (list_empty(cluster))
1897                                 break;
1898
1899                         locked_ref = list_entry(cluster->next,
1900                                      struct btrfs_delayed_ref_head, cluster);
1901
1902                         /* grab the lock that says we are going to process
1903                          * all the refs for this head */
1904                         ret = btrfs_delayed_ref_lock(trans, locked_ref);
1905
1906                         /*
1907                          * we may have dropped the spin lock to get the head
1908                          * mutex lock, and that might have given someone else
1909                          * time to free the head.  If that's true, it has been
1910                          * removed from our list and we can move on.
1911                          */
1912                         if (ret == -EAGAIN) {
1913                                 locked_ref = NULL;
1914                                 count++;
1915                                 continue;
1916                         }
1917                 }
1918
1919                 /*
1920                  * record the must insert reserved flag before we
1921                  * drop the spin lock.
1922                  */
1923                 must_insert_reserved = locked_ref->must_insert_reserved;
1924                 locked_ref->must_insert_reserved = 0;
1925
1926                 extent_op = locked_ref->extent_op;
1927                 locked_ref->extent_op = NULL;
1928
1929                 /*
1930                  * locked_ref is the head node, so we have to go one
1931                  * node back for any delayed ref updates
1932                  */
1933                 ref = select_delayed_ref(locked_ref);
1934                 if (!ref) {
1935                         /* All delayed refs have been processed, Go ahead
1936                          * and send the head node to run_one_delayed_ref,
1937                          * so that any accounting fixes can happen
1938                          */
1939                         ref = &locked_ref->node;
1940
1941                         if (extent_op && must_insert_reserved) {
1942                                 kfree(extent_op);
1943                                 extent_op = NULL;
1944                         }
1945
1946                         if (extent_op) {
1947                                 spin_unlock(&delayed_refs->lock);
1948
1949                                 ret = run_delayed_extent_op(trans, root,
1950                                                             ref, extent_op);
1951                                 BUG_ON(ret);
1952                                 kfree(extent_op);
1953
1954                                 cond_resched();
1955                                 spin_lock(&delayed_refs->lock);
1956                                 continue;
1957                         }
1958
1959                         list_del_init(&locked_ref->cluster);
1960                         locked_ref = NULL;
1961                 }
1962
1963                 ref->in_tree = 0;
1964                 rb_erase(&ref->rb_node, &delayed_refs->root);
1965                 delayed_refs->num_entries--;
1966
1967                 spin_unlock(&delayed_refs->lock);
1968
1969                 ret = run_one_delayed_ref(trans, root, ref, extent_op,
1970                                           must_insert_reserved);
1971                 BUG_ON(ret);
1972
1973                 btrfs_put_delayed_ref(ref);
1974                 kfree(extent_op);
1975                 count++;
1976
1977                 cond_resched();
1978                 spin_lock(&delayed_refs->lock);
1979         }
1980         return count;
1981 }
1982
1983 /*
1984  * this starts processing the delayed reference count updates and
1985  * extent insertions we have queued up so far.  count can be
1986  * 0, which means to process everything in the tree at the start
1987  * of the run (but not newly added entries), or it can be some target
1988  * number you'd like to process.
1989  */
1990 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
1991                            struct btrfs_root *root, unsigned long count)
1992 {
1993         struct rb_node *node;
1994         struct btrfs_delayed_ref_root *delayed_refs;
1995         struct btrfs_delayed_ref_node *ref;
1996         struct list_head cluster;
1997         int ret;
1998         int run_all = count == (unsigned long)-1;
1999         int run_most = 0;
2000
2001         if (root == root->fs_info->extent_root)
2002                 root = root->fs_info->tree_root;
2003
2004         delayed_refs = &trans->transaction->delayed_refs;
2005         INIT_LIST_HEAD(&cluster);
2006 again:
2007         spin_lock(&delayed_refs->lock);
2008         if (count == 0) {
2009                 count = delayed_refs->num_entries * 2;
2010                 run_most = 1;
2011         }
2012         while (1) {
2013                 if (!(run_all || run_most) &&
2014                     delayed_refs->num_heads_ready < 64)
2015                         break;
2016
2017                 /*
2018                  * go find something we can process in the rbtree.  We start at
2019                  * the beginning of the tree, and then build a cluster
2020                  * of refs to process starting at the first one we are able to
2021                  * lock
2022                  */
2023                 ret = btrfs_find_ref_cluster(trans, &cluster,
2024                                              delayed_refs->run_delayed_start);
2025                 if (ret)
2026                         break;
2027
2028                 ret = run_clustered_refs(trans, root, &cluster);
2029                 BUG_ON(ret < 0);
2030
2031                 count -= min_t(unsigned long, ret, count);
2032
2033                 if (count == 0)
2034                         break;
2035         }
2036
2037         if (run_all) {
2038                 node = rb_first(&delayed_refs->root);
2039                 if (!node)
2040                         goto out;
2041                 count = (unsigned long)-1;
2042
2043                 while (node) {
2044                         ref = rb_entry(node, struct btrfs_delayed_ref_node,
2045                                        rb_node);
2046                         if (btrfs_delayed_ref_is_head(ref)) {
2047                                 struct btrfs_delayed_ref_head *head;
2048
2049                                 head = btrfs_delayed_node_to_head(ref);
2050                                 atomic_inc(&ref->refs);
2051
2052                                 spin_unlock(&delayed_refs->lock);
2053                                 mutex_lock(&head->mutex);
2054                                 mutex_unlock(&head->mutex);
2055
2056                                 btrfs_put_delayed_ref(ref);
2057                                 cond_resched();
2058                                 goto again;
2059                         }
2060                         node = rb_next(node);
2061                 }
2062                 spin_unlock(&delayed_refs->lock);
2063                 schedule_timeout(1);
2064                 goto again;
2065         }
2066 out:
2067         spin_unlock(&delayed_refs->lock);
2068         return 0;
2069 }
2070
2071 int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
2072                                 struct btrfs_root *root,
2073                                 u64 bytenr, u64 num_bytes, u64 flags,
2074                                 int is_data)
2075 {
2076         struct btrfs_delayed_extent_op *extent_op;
2077         int ret;
2078
2079         extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
2080         if (!extent_op)
2081                 return -ENOMEM;
2082
2083         extent_op->flags_to_set = flags;
2084         extent_op->update_flags = 1;
2085         extent_op->update_key = 0;
2086         extent_op->is_data = is_data ? 1 : 0;
2087
2088         ret = btrfs_add_delayed_extent_op(trans, bytenr, num_bytes, extent_op);
2089         if (ret)
2090                 kfree(extent_op);
2091         return ret;
2092 }
2093
2094 static noinline int check_delayed_ref(struct btrfs_trans_handle *trans,
2095                                       struct btrfs_root *root,
2096                                       struct btrfs_path *path,
2097                                       u64 objectid, u64 offset, u64 bytenr)
2098 {
2099         struct btrfs_delayed_ref_head *head;
2100         struct btrfs_delayed_ref_node *ref;
2101         struct btrfs_delayed_data_ref *data_ref;
2102         struct btrfs_delayed_ref_root *delayed_refs;
2103         struct rb_node *node;
2104         int ret = 0;
2105
2106         ret = -ENOENT;
2107         delayed_refs = &trans->transaction->delayed_refs;
2108         spin_lock(&delayed_refs->lock);
2109         head = btrfs_find_delayed_ref_head(trans, bytenr);
2110         if (!head)
2111                 goto out;
2112
2113         if (!mutex_trylock(&head->mutex)) {
2114                 atomic_inc(&head->node.refs);
2115                 spin_unlock(&delayed_refs->lock);
2116
2117                 btrfs_release_path(root->fs_info->extent_root, path);
2118
2119                 mutex_lock(&head->mutex);
2120                 mutex_unlock(&head->mutex);
2121                 btrfs_put_delayed_ref(&head->node);
2122                 return -EAGAIN;
2123         }
2124
2125         node = rb_prev(&head->node.rb_node);
2126         if (!node)
2127                 goto out_unlock;
2128
2129         ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2130
2131         if (ref->bytenr != bytenr)
2132                 goto out_unlock;
2133
2134         ret = 1;
2135         if (ref->type != BTRFS_EXTENT_DATA_REF_KEY)
2136                 goto out_unlock;
2137
2138         data_ref = btrfs_delayed_node_to_data_ref(ref);
2139
2140         node = rb_prev(node);
2141         if (node) {
2142                 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2143                 if (ref->bytenr == bytenr)
2144                         goto out_unlock;
2145         }
2146
2147         if (data_ref->root != root->root_key.objectid ||
2148             data_ref->objectid != objectid || data_ref->offset != offset)
2149                 goto out_unlock;
2150
2151         ret = 0;
2152 out_unlock:
2153         mutex_unlock(&head->mutex);
2154 out:
2155         spin_unlock(&delayed_refs->lock);
2156         return ret;
2157 }
2158
2159 static noinline int check_committed_ref(struct btrfs_trans_handle *trans,
2160                                         struct btrfs_root *root,
2161                                         struct btrfs_path *path,
2162                                         u64 objectid, u64 offset, u64 bytenr)
2163 {
2164         struct btrfs_root *extent_root = root->fs_info->extent_root;
2165         struct extent_buffer *leaf;
2166         struct btrfs_extent_data_ref *ref;
2167         struct btrfs_extent_inline_ref *iref;
2168         struct btrfs_extent_item *ei;
2169         struct btrfs_key key;
2170         u32 item_size;
2171         int ret;
2172
2173         key.objectid = bytenr;
2174         key.offset = (u64)-1;
2175         key.type = BTRFS_EXTENT_ITEM_KEY;
2176
2177         ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
2178         if (ret < 0)
2179                 goto out;
2180         BUG_ON(ret == 0);
2181
2182         ret = -ENOENT;
2183         if (path->slots[0] == 0)
2184                 goto out;
2185
2186         path->slots[0]--;
2187         leaf = path->nodes[0];
2188         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
2189
2190         if (key.objectid != bytenr || key.type != BTRFS_EXTENT_ITEM_KEY)
2191                 goto out;
2192
2193         ret = 1;
2194         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
2195 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
2196         if (item_size < sizeof(*ei)) {
2197                 WARN_ON(item_size != sizeof(struct btrfs_extent_item_v0));
2198                 goto out;
2199         }
2200 #endif
2201         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
2202
2203         if (item_size != sizeof(*ei) +
2204             btrfs_extent_inline_ref_size(BTRFS_EXTENT_DATA_REF_KEY))
2205                 goto out;
2206
2207         if (btrfs_extent_generation(leaf, ei) <=
2208             btrfs_root_last_snapshot(&root->root_item))
2209                 goto out;
2210
2211         iref = (struct btrfs_extent_inline_ref *)(ei + 1);
2212         if (btrfs_extent_inline_ref_type(leaf, iref) !=
2213             BTRFS_EXTENT_DATA_REF_KEY)
2214                 goto out;
2215
2216         ref = (struct btrfs_extent_data_ref *)(&iref->offset);
2217         if (btrfs_extent_refs(leaf, ei) !=
2218             btrfs_extent_data_ref_count(leaf, ref) ||
2219             btrfs_extent_data_ref_root(leaf, ref) !=
2220             root->root_key.objectid ||
2221             btrfs_extent_data_ref_objectid(leaf, ref) != objectid ||
2222             btrfs_extent_data_ref_offset(leaf, ref) != offset)
2223                 goto out;
2224
2225         ret = 0;
2226 out:
2227         return ret;
2228 }
2229
2230 int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
2231                           struct btrfs_root *root,
2232                           u64 objectid, u64 offset, u64 bytenr)
2233 {
2234         struct btrfs_path *path;
2235         int ret;
2236         int ret2;
2237
2238         path = btrfs_alloc_path();
2239         if (!path)
2240                 return -ENOENT;
2241
2242         do {
2243                 ret = check_committed_ref(trans, root, path, objectid,
2244                                           offset, bytenr);
2245                 if (ret && ret != -ENOENT)
2246                         goto out;
2247
2248                 ret2 = check_delayed_ref(trans, root, path, objectid,
2249                                          offset, bytenr);
2250         } while (ret2 == -EAGAIN);
2251
2252         if (ret2 && ret2 != -ENOENT) {
2253                 ret = ret2;
2254                 goto out;
2255         }
2256
2257         if (ret != -ENOENT || ret2 != -ENOENT)
2258                 ret = 0;
2259 out:
2260         btrfs_free_path(path);
2261         return ret;
2262 }
2263
2264 #if 0
2265 int btrfs_cache_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2266                     struct extent_buffer *buf, u32 nr_extents)
2267 {
2268         struct btrfs_key key;
2269         struct btrfs_file_extent_item *fi;
2270         u64 root_gen;
2271         u32 nritems;
2272         int i;
2273         int level;
2274         int ret = 0;
2275         int shared = 0;
2276
2277         if (!root->ref_cows)
2278                 return 0;
2279
2280         if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
2281                 shared = 0;
2282                 root_gen = root->root_key.offset;
2283         } else {
2284                 shared = 1;
2285                 root_gen = trans->transid - 1;
2286         }
2287
2288         level = btrfs_header_level(buf);
2289         nritems = btrfs_header_nritems(buf);
2290
2291         if (level == 0) {
2292                 struct btrfs_leaf_ref *ref;
2293                 struct btrfs_extent_info *info;
2294
2295                 ref = btrfs_alloc_leaf_ref(root, nr_extents);
2296                 if (!ref) {
2297                         ret = -ENOMEM;
2298                         goto out;
2299                 }
2300
2301                 ref->root_gen = root_gen;
2302                 ref->bytenr = buf->start;
2303                 ref->owner = btrfs_header_owner(buf);
2304                 ref->generation = btrfs_header_generation(buf);
2305                 ref->nritems = nr_extents;
2306                 info = ref->extents;
2307
2308                 for (i = 0; nr_extents > 0 && i < nritems; i++) {
2309                         u64 disk_bytenr;
2310                         btrfs_item_key_to_cpu(buf, &key, i);
2311                         if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
2312                                 continue;
2313                         fi = btrfs_item_ptr(buf, i,
2314                                             struct btrfs_file_extent_item);
2315                         if (btrfs_file_extent_type(buf, fi) ==
2316                             BTRFS_FILE_EXTENT_INLINE)
2317                                 continue;
2318                         disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
2319                         if (disk_bytenr == 0)
2320                                 continue;
2321
2322                         info->bytenr = disk_bytenr;
2323                         info->num_bytes =
2324                                 btrfs_file_extent_disk_num_bytes(buf, fi);
2325                         info->objectid = key.objectid;
2326                         info->offset = key.offset;
2327                         info++;
2328                 }
2329
2330                 ret = btrfs_add_leaf_ref(root, ref, shared);
2331                 if (ret == -EEXIST && shared) {
2332                         struct btrfs_leaf_ref *old;
2333                         old = btrfs_lookup_leaf_ref(root, ref->bytenr);
2334                         BUG_ON(!old);
2335                         btrfs_remove_leaf_ref(root, old);
2336                         btrfs_free_leaf_ref(root, old);
2337                         ret = btrfs_add_leaf_ref(root, ref, shared);
2338                 }
2339                 WARN_ON(ret);
2340                 btrfs_free_leaf_ref(root, ref);
2341         }
2342 out:
2343         return ret;
2344 }
2345
2346 /* when a block goes through cow, we update the reference counts of
2347  * everything that block points to.  The internal pointers of the block
2348  * can be in just about any order, and it is likely to have clusters of
2349  * things that are close together and clusters of things that are not.
2350  *
2351  * To help reduce the seeks that come with updating all of these reference
2352  * counts, sort them by byte number before actual updates are done.
2353  *
2354  * struct refsort is used to match byte number to slot in the btree block.
2355  * we sort based on the byte number and then use the slot to actually
2356  * find the item.
2357  *
2358  * struct refsort is smaller than strcut btrfs_item and smaller than
2359  * struct btrfs_key_ptr.  Since we're currently limited to the page size
2360  * for a btree block, there's no way for a kmalloc of refsorts for a
2361  * single node to be bigger than a page.
2362  */
2363 struct refsort {
2364         u64 bytenr;
2365         u32 slot;
2366 };
2367
2368 /*
2369  * for passing into sort()
2370  */
2371 static int refsort_cmp(const void *a_void, const void *b_void)
2372 {
2373         const struct refsort *a = a_void;
2374         const struct refsort *b = b_void;
2375
2376         if (a->bytenr < b->bytenr)
2377                 return -1;
2378         if (a->bytenr > b->bytenr)
2379                 return 1;
2380         return 0;
2381 }
2382 #endif
2383
2384 static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
2385                            struct btrfs_root *root,
2386                            struct extent_buffer *buf,
2387                            int full_backref, int inc)
2388 {
2389         u64 bytenr;
2390         u64 num_bytes;
2391         u64 parent;
2392         u64 ref_root;
2393         u32 nritems;
2394         struct btrfs_key key;
2395         struct btrfs_file_extent_item *fi;
2396         int i;
2397         int level;
2398         int ret = 0;
2399         int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *,
2400                             u64, u64, u64, u64, u64, u64);
2401
2402         ref_root = btrfs_header_owner(buf);
2403         nritems = btrfs_header_nritems(buf);
2404         level = btrfs_header_level(buf);
2405
2406         if (!root->ref_cows && level == 0)
2407                 return 0;
2408
2409         if (inc)
2410                 process_func = btrfs_inc_extent_ref;
2411         else
2412                 process_func = btrfs_free_extent;
2413
2414         if (full_backref)
2415                 parent = buf->start;
2416         else
2417                 parent = 0;
2418
2419         for (i = 0; i < nritems; i++) {
2420                 if (level == 0) {
2421                         btrfs_item_key_to_cpu(buf, &key, i);
2422                         if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
2423                                 continue;
2424                         fi = btrfs_item_ptr(buf, i,
2425                                             struct btrfs_file_extent_item);
2426                         if (btrfs_file_extent_type(buf, fi) ==
2427                             BTRFS_FILE_EXTENT_INLINE)
2428                                 continue;
2429                         bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
2430                         if (bytenr == 0)
2431                                 continue;
2432
2433                         num_bytes = btrfs_file_extent_disk_num_bytes(buf, fi);
2434                         key.offset -= btrfs_file_extent_offset(buf, fi);
2435                         ret = process_func(trans, root, bytenr, num_bytes,
2436                                            parent, ref_root, key.objectid,
2437                                            key.offset);
2438                         if (ret)
2439                                 goto fail;
2440                 } else {
2441                         bytenr = btrfs_node_blockptr(buf, i);
2442                         num_bytes = btrfs_level_size(root, level - 1);
2443                         ret = process_func(trans, root, bytenr, num_bytes,
2444                                            parent, ref_root, level - 1, 0);
2445                         if (ret)
2446                                 goto fail;
2447                 }
2448         }
2449         return 0;
2450 fail:
2451         BUG();
2452         return ret;
2453 }
2454
2455 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2456                   struct extent_buffer *buf, int full_backref)
2457 {
2458         return __btrfs_mod_ref(trans, root, buf, full_backref, 1);
2459 }
2460
2461 int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2462                   struct extent_buffer *buf, int full_backref)
2463 {
2464         return __btrfs_mod_ref(trans, root, buf, full_backref, 0);
2465 }
2466
2467 static int write_one_cache_group(struct btrfs_trans_handle *trans,
2468                                  struct btrfs_root *root,
2469                                  struct btrfs_path *path,
2470                                  struct btrfs_block_group_cache *cache)
2471 {
2472         int ret;
2473         struct btrfs_root *extent_root = root->fs_info->extent_root;
2474         unsigned long bi;
2475         struct extent_buffer *leaf;
2476
2477         ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
2478         if (ret < 0)
2479                 goto fail;
2480         BUG_ON(ret);
2481
2482         leaf = path->nodes[0];
2483         bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
2484         write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
2485         btrfs_mark_buffer_dirty(leaf);
2486         btrfs_release_path(extent_root, path);
2487 fail:
2488         if (ret)
2489                 return ret;
2490         return 0;
2491
2492 }
2493
2494 static struct btrfs_block_group_cache *
2495 next_block_group(struct btrfs_root *root,
2496                  struct btrfs_block_group_cache *cache)
2497 {
2498         struct rb_node *node;
2499         spin_lock(&root->fs_info->block_group_cache_lock);
2500         node = rb_next(&cache->cache_node);
2501         btrfs_put_block_group(cache);
2502         if (node) {
2503                 cache = rb_entry(node, struct btrfs_block_group_cache,
2504                                  cache_node);
2505                 atomic_inc(&cache->count);
2506         } else
2507                 cache = NULL;
2508         spin_unlock(&root->fs_info->block_group_cache_lock);
2509         return cache;
2510 }
2511
2512 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
2513                                    struct btrfs_root *root)
2514 {
2515         struct btrfs_block_group_cache *cache;
2516         int err = 0;
2517         struct btrfs_path *path;
2518         u64 last = 0;
2519
2520         path = btrfs_alloc_path();
2521         if (!path)
2522                 return -ENOMEM;
2523
2524         while (1) {
2525                 if (last == 0) {
2526                         err = btrfs_run_delayed_refs(trans, root,
2527                                                      (unsigned long)-1);
2528                         BUG_ON(err);
2529                 }
2530
2531                 cache = btrfs_lookup_first_block_group(root->fs_info, last);
2532                 while (cache) {
2533                         if (cache->dirty)
2534                                 break;
2535                         cache = next_block_group(root, cache);
2536                 }
2537                 if (!cache) {
2538                         if (last == 0)
2539                                 break;
2540                         last = 0;
2541                         continue;
2542                 }
2543
2544                 cache->dirty = 0;
2545                 last = cache->key.objectid + cache->key.offset;
2546
2547                 err = write_one_cache_group(trans, root, path, cache);
2548                 BUG_ON(err);
2549                 btrfs_put_block_group(cache);
2550         }
2551
2552         btrfs_free_path(path);
2553         return 0;
2554 }
2555
2556 int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr)
2557 {
2558         struct btrfs_block_group_cache *block_group;
2559         int readonly = 0;
2560
2561         block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
2562         if (!block_group || block_group->ro)
2563                 readonly = 1;
2564         if (block_group)
2565                 btrfs_put_block_group(block_group);
2566         return readonly;
2567 }
2568
2569 static int update_space_info(struct btrfs_fs_info *info, u64 flags,
2570                              u64 total_bytes, u64 bytes_used,
2571                              struct btrfs_space_info **space_info)
2572 {
2573         struct btrfs_space_info *found;
2574
2575         found = __find_space_info(info, flags);
2576         if (found) {
2577                 spin_lock(&found->lock);
2578                 found->total_bytes += total_bytes;
2579                 found->bytes_used += bytes_used;
2580                 found->full = 0;
2581                 spin_unlock(&found->lock);
2582                 *space_info = found;
2583                 return 0;
2584         }
2585         found = kzalloc(sizeof(*found), GFP_NOFS);
2586         if (!found)
2587                 return -ENOMEM;
2588
2589         INIT_LIST_HEAD(&found->block_groups);
2590         init_rwsem(&found->groups_sem);
2591         spin_lock_init(&found->lock);
2592         found->flags = flags;
2593         found->total_bytes = total_bytes;
2594         found->bytes_used = bytes_used;
2595         found->bytes_pinned = 0;
2596         found->bytes_reserved = 0;
2597         found->bytes_readonly = 0;
2598         found->bytes_delalloc = 0;
2599         found->full = 0;
2600         found->force_alloc = 0;
2601         *space_info = found;
2602         list_add_rcu(&found->list, &info->space_info);
2603         atomic_set(&found->caching_threads, 0);
2604         return 0;
2605 }
2606
2607 static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
2608 {
2609         u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
2610                                    BTRFS_BLOCK_GROUP_RAID1 |
2611                                    BTRFS_BLOCK_GROUP_RAID10 |
2612                                    BTRFS_BLOCK_GROUP_DUP);
2613         if (extra_flags) {
2614                 if (flags & BTRFS_BLOCK_GROUP_DATA)
2615                         fs_info->avail_data_alloc_bits |= extra_flags;
2616                 if (flags & BTRFS_BLOCK_GROUP_METADATA)
2617                         fs_info->avail_metadata_alloc_bits |= extra_flags;
2618                 if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
2619                         fs_info->avail_system_alloc_bits |= extra_flags;
2620         }
2621 }
2622
2623 static void set_block_group_readonly(struct btrfs_block_group_cache *cache)
2624 {
2625         spin_lock(&cache->space_info->lock);
2626         spin_lock(&cache->lock);
2627         if (!cache->ro) {
2628                 cache->space_info->bytes_readonly += cache->key.offset -
2629                                         btrfs_block_group_used(&cache->item);
2630                 cache->ro = 1;
2631         }
2632         spin_unlock(&cache->lock);
2633         spin_unlock(&cache->space_info->lock);
2634 }
2635
2636 u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
2637 {
2638         u64 num_devices = root->fs_info->fs_devices->rw_devices;
2639
2640         if (num_devices == 1)
2641                 flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
2642         if (num_devices < 4)
2643                 flags &= ~BTRFS_BLOCK_GROUP_RAID10;
2644
2645         if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
2646             (flags & (BTRFS_BLOCK_GROUP_RAID1 |
2647                       BTRFS_BLOCK_GROUP_RAID10))) {
2648                 flags &= ~BTRFS_BLOCK_GROUP_DUP;
2649         }
2650
2651         if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
2652             (flags & BTRFS_BLOCK_GROUP_RAID10)) {
2653                 flags &= ~BTRFS_BLOCK_GROUP_RAID1;
2654         }
2655
2656         if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
2657             ((flags & BTRFS_BLOCK_GROUP_RAID1) |
2658              (flags & BTRFS_BLOCK_GROUP_RAID10) |
2659              (flags & BTRFS_BLOCK_GROUP_DUP)))
2660                 flags &= ~BTRFS_BLOCK_GROUP_RAID0;
2661         return flags;
2662 }
2663
2664 static u64 btrfs_get_alloc_profile(struct btrfs_root *root, u64 data)
2665 {
2666         struct btrfs_fs_info *info = root->fs_info;
2667         u64 alloc_profile;
2668
2669         if (data) {
2670                 alloc_profile = info->avail_data_alloc_bits &
2671                         info->data_alloc_profile;
2672                 data = BTRFS_BLOCK_GROUP_DATA | alloc_profile;
2673         } else if (root == root->fs_info->chunk_root) {
2674                 alloc_profile = info->avail_system_alloc_bits &
2675                         info->system_alloc_profile;
2676                 data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile;
2677         } else {
2678                 alloc_profile = info->avail_metadata_alloc_bits &
2679                         info->metadata_alloc_profile;
2680                 data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
2681         }
2682
2683         return btrfs_reduce_alloc_profile(root, data);
2684 }
2685
2686 void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *inode)
2687 {
2688         u64 alloc_target;
2689
2690         alloc_target = btrfs_get_alloc_profile(root, 1);
2691         BTRFS_I(inode)->space_info = __find_space_info(root->fs_info,
2692                                                        alloc_target);
2693 }
2694
2695 /*
2696  * for now this just makes sure we have at least 5% of our metadata space free
2697  * for use.
2698  */
2699 int btrfs_check_metadata_free_space(struct btrfs_root *root)
2700 {
2701         struct btrfs_fs_info *info = root->fs_info;
2702         struct btrfs_space_info *meta_sinfo;
2703         u64 alloc_target, thresh;
2704         int committed = 0, ret;
2705
2706         /* get the space info for where the metadata will live */
2707         alloc_target = btrfs_get_alloc_profile(root, 0);
2708         meta_sinfo = __find_space_info(info, alloc_target);
2709
2710 again:
2711         spin_lock(&meta_sinfo->lock);
2712         if (!meta_sinfo->full)
2713                 thresh = meta_sinfo->total_bytes * 80;
2714         else
2715                 thresh = meta_sinfo->total_bytes * 95;
2716
2717         do_div(thresh, 100);
2718
2719         if (meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
2720             meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly > thresh) {
2721                 struct btrfs_trans_handle *trans;
2722                 if (!meta_sinfo->full) {
2723                         meta_sinfo->force_alloc = 1;
2724                         spin_unlock(&meta_sinfo->lock);
2725
2726                         trans = btrfs_start_transaction(root, 1);
2727                         if (!trans)
2728                                 return -ENOMEM;
2729
2730                         ret = do_chunk_alloc(trans, root->fs_info->extent_root,
2731                                              2 * 1024 * 1024, alloc_target, 0);
2732                         btrfs_end_transaction(trans, root);
2733                         goto again;
2734                 }
2735                 spin_unlock(&meta_sinfo->lock);
2736
2737                 if (!committed) {
2738                         committed = 1;
2739                         trans = btrfs_join_transaction(root, 1);
2740                         if (!trans)
2741                                 return -ENOMEM;
2742                         ret = btrfs_commit_transaction(trans, root);
2743                         if (ret)
2744                                 return ret;
2745                         goto again;
2746                 }
2747                 return -ENOSPC;
2748         }
2749         spin_unlock(&meta_sinfo->lock);
2750
2751         return 0;
2752 }
2753
2754 /*
2755  * This will check the space that the inode allocates from to make sure we have
2756  * enough space for bytes.
2757  */
2758 int btrfs_check_data_free_space(struct btrfs_root *root, struct inode *inode,
2759                                 u64 bytes)
2760 {
2761         struct btrfs_space_info *data_sinfo;
2762         int ret = 0, committed = 0;
2763
2764         /* make sure bytes are sectorsize aligned */
2765         bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
2766
2767         data_sinfo = BTRFS_I(inode)->space_info;
2768 again:
2769         /* make sure we have enough space to handle the data first */
2770         spin_lock(&data_sinfo->lock);
2771         if (data_sinfo->total_bytes - data_sinfo->bytes_used -
2772             data_sinfo->bytes_delalloc - data_sinfo->bytes_reserved -
2773             data_sinfo->bytes_pinned - data_sinfo->bytes_readonly -
2774             data_sinfo->bytes_may_use < bytes) {
2775                 struct btrfs_trans_handle *trans;
2776
2777                 /*
2778                  * if we don't have enough free bytes in this space then we need
2779                  * to alloc a new chunk.
2780                  */
2781                 if (!data_sinfo->full) {
2782                         u64 alloc_target;
2783
2784                         data_sinfo->force_alloc = 1;
2785                         spin_unlock(&data_sinfo->lock);
2786
2787                         alloc_target = btrfs_get_alloc_profile(root, 1);
2788                         trans = btrfs_start_transaction(root, 1);
2789                         if (!trans)
2790                                 return -ENOMEM;
2791
2792                         ret = do_chunk_alloc(trans, root->fs_info->extent_root,
2793                                              bytes + 2 * 1024 * 1024,
2794                                              alloc_target, 0);
2795                         btrfs_end_transaction(trans, root);
2796                         if (ret)
2797                                 return ret;
2798                         goto again;
2799                 }
2800                 spin_unlock(&data_sinfo->lock);
2801
2802                 /* commit the current transaction and try again */
2803                 if (!committed) {
2804                         committed = 1;
2805                         trans = btrfs_join_transaction(root, 1);
2806                         if (!trans)
2807                                 return -ENOMEM;
2808                         ret = btrfs_commit_transaction(trans, root);
2809                         if (ret)
2810                                 return ret;
2811                         goto again;
2812                 }
2813
2814                 printk(KERN_ERR "no space left, need %llu, %llu delalloc bytes"
2815                        ", %llu bytes_used, %llu bytes_reserved, "
2816                        "%llu bytes_pinned, %llu bytes_readonly, %llu may use "
2817                        "%llu total\n", (unsigned long long)bytes,
2818                        (unsigned long long)data_sinfo->bytes_delalloc,
2819                        (unsigned long long)data_sinfo->bytes_used,
2820                        (unsigned long long)data_sinfo->bytes_reserved,
2821                        (unsigned long long)data_sinfo->bytes_pinned,
2822                        (unsigned long long)data_sinfo->bytes_readonly,
2823                        (unsigned long long)data_sinfo->bytes_may_use,
2824                        (unsigned long long)data_sinfo->total_bytes);
2825                 return -ENOSPC;
2826         }
2827         data_sinfo->bytes_may_use += bytes;
2828         BTRFS_I(inode)->reserved_bytes += bytes;
2829         spin_unlock(&data_sinfo->lock);
2830
2831         return btrfs_check_metadata_free_space(root);
2832 }
2833
2834 /*
2835  * if there was an error for whatever reason after calling
2836  * btrfs_check_data_free_space, call this so we can cleanup the counters.
2837  */
2838 void btrfs_free_reserved_data_space(struct btrfs_root *root,
2839                                     struct inode *inode, u64 bytes)
2840 {
2841         struct btrfs_space_info *data_sinfo;
2842
2843         /* make sure bytes are sectorsize aligned */
2844         bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
2845
2846         data_sinfo = BTRFS_I(inode)->space_info;
2847         spin_lock(&data_sinfo->lock);
2848         data_sinfo->bytes_may_use -= bytes;
2849         BTRFS_I(inode)->reserved_bytes -= bytes;
2850         spin_unlock(&data_sinfo->lock);
2851 }
2852
2853 /* called when we are adding a delalloc extent to the inode's io_tree */
2854 void btrfs_delalloc_reserve_space(struct btrfs_root *root, struct inode *inode,
2855                                   u64 bytes)
2856 {
2857         struct btrfs_space_info *data_sinfo;
2858
2859         /* get the space info for where this inode will be storing its data */
2860         data_sinfo = BTRFS_I(inode)->space_info;
2861
2862         /* make sure we have enough space to handle the data first */
2863         spin_lock(&data_sinfo->lock);
2864         data_sinfo->bytes_delalloc += bytes;
2865
2866         /*
2867          * we are adding a delalloc extent without calling
2868          * btrfs_check_data_free_space first.  This happens on a weird
2869          * writepage condition, but shouldn't hurt our accounting
2870          */
2871         if (unlikely(bytes > BTRFS_I(inode)->reserved_bytes)) {
2872                 data_sinfo->bytes_may_use -= BTRFS_I(inode)->reserved_bytes;
2873                 BTRFS_I(inode)->reserved_bytes = 0;
2874         } else {
2875                 data_sinfo->bytes_may_use -= bytes;
2876                 BTRFS_I(inode)->reserved_bytes -= bytes;
2877         }
2878
2879         spin_unlock(&data_sinfo->lock);
2880 }
2881
2882 /* called when we are clearing an delalloc extent from the inode's io_tree */
2883 void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
2884                               u64 bytes)
2885 {
2886         struct btrfs_space_info *info;
2887
2888         info = BTRFS_I(inode)->space_info;
2889
2890         spin_lock(&info->lock);
2891         info->bytes_delalloc -= bytes;
2892         spin_unlock(&info->lock);
2893 }
2894
2895 static void force_metadata_allocation(struct btrfs_fs_info *info)
2896 {
2897         struct list_head *head = &info->space_info;
2898         struct btrfs_space_info *found;
2899
2900         rcu_read_lock();
2901         list_for_each_entry_rcu(found, head, list) {
2902                 if (found->flags & BTRFS_BLOCK_GROUP_METADATA)
2903                         found->force_alloc = 1;
2904         }
2905         rcu_read_unlock();
2906 }
2907
2908 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
2909                           struct btrfs_root *extent_root, u64 alloc_bytes,
2910                           u64 flags, int force)
2911 {
2912         struct btrfs_space_info *space_info;
2913         struct btrfs_fs_info *fs_info = extent_root->fs_info;
2914         u64 thresh;
2915         int ret = 0;
2916
2917         mutex_lock(&fs_info->chunk_mutex);
2918
2919         flags = btrfs_reduce_alloc_profile(extent_root, flags);
2920
2921         space_info = __find_space_info(extent_root->fs_info, flags);
2922         if (!space_info) {
2923                 ret = update_space_info(extent_root->fs_info, flags,
2924                                         0, 0, &space_info);
2925                 BUG_ON(ret);
2926         }
2927         BUG_ON(!space_info);
2928
2929         spin_lock(&space_info->lock);
2930         if (space_info->force_alloc) {
2931                 force = 1;
2932                 space_info->force_alloc = 0;
2933         }
2934         if (space_info->full) {
2935                 spin_unlock(&space_info->lock);
2936                 goto out;
2937         }
2938
2939         thresh = space_info->total_bytes - space_info->bytes_readonly;
2940         thresh = div_factor(thresh, 6);
2941         if (!force &&
2942            (space_info->bytes_used + space_info->bytes_pinned +
2943             space_info->bytes_reserved + alloc_bytes) < thresh) {
2944                 spin_unlock(&space_info->lock);
2945                 goto out;
2946         }
2947         spin_unlock(&space_info->lock);
2948
2949         /*
2950          * if we're doing a data chunk, go ahead and make sure that
2951          * we keep a reasonable number of metadata chunks allocated in the
2952          * FS as well.
2953          */
2954         if (flags & BTRFS_BLOCK_GROUP_DATA) {
2955                 fs_info->data_chunk_allocations++;
2956                 if (!(fs_info->data_chunk_allocations %
2957                       fs_info->metadata_ratio))
2958                         force_metadata_allocation(fs_info);
2959         }
2960
2961         ret = btrfs_alloc_chunk(trans, extent_root, flags);
2962         if (ret)
2963                 space_info->full = 1;
2964 out:
2965         mutex_unlock(&extent_root->fs_info->chunk_mutex);
2966         return ret;
2967 }
2968
2969 static int update_block_group(struct btrfs_trans_handle *trans,
2970                               struct btrfs_root *root,
2971                               u64 bytenr, u64 num_bytes, int alloc,
2972                               int mark_free)
2973 {
2974         struct btrfs_block_group_cache *cache;
2975         struct btrfs_fs_info *info = root->fs_info;
2976         u64 total = num_bytes;
2977         u64 old_val;
2978         u64 byte_in_group;
2979
2980         /* block accounting for super block */
2981         spin_lock(&info->delalloc_lock);
2982         old_val = btrfs_super_bytes_used(&info->super_copy);
2983         if (alloc)
2984                 old_val += num_bytes;
2985         else
2986                 old_val -= num_bytes;
2987         btrfs_set_super_bytes_used(&info->super_copy, old_val);
2988
2989         /* block accounting for root item */
2990         old_val = btrfs_root_used(&root->root_item);
2991         if (alloc)
2992                 old_val += num_bytes;
2993         else
2994                 old_val -= num_bytes;
2995         btrfs_set_root_used(&root->root_item, old_val);
2996         spin_unlock(&info->delalloc_lock);
2997
2998         while (total) {
2999                 cache = btrfs_lookup_block_group(info, bytenr);
3000                 if (!cache)
3001                         return -1;
3002                 byte_in_group = bytenr - cache->key.objectid;
3003                 WARN_ON(byte_in_group > cache->key.offset);
3004
3005                 spin_lock(&cache->space_info->lock);
3006                 spin_lock(&cache->lock);
3007                 cache->dirty = 1;
3008                 old_val = btrfs_block_group_used(&cache->item);
3009                 num_bytes = min(total, cache->key.offset - byte_in_group);
3010                 if (alloc) {
3011                         old_val += num_bytes;
3012                         cache->space_info->bytes_used += num_bytes;
3013                         if (cache->ro)
3014                                 cache->space_info->bytes_readonly -= num_bytes;
3015                         btrfs_set_block_group_used(&cache->item, old_val);
3016                         spin_unlock(&cache->lock);
3017                         spin_unlock(&cache->space_info->lock);
3018                 } else {
3019                         old_val -= num_bytes;
3020                         cache->space_info->bytes_used -= num_bytes;
3021                         if (cache->ro)
3022                                 cache->space_info->bytes_readonly += num_bytes;
3023                         btrfs_set_block_group_used(&cache->item, old_val);
3024                         spin_unlock(&cache->lock);
3025                         spin_unlock(&cache->space_info->lock);
3026                         if (mark_free) {
3027                                 int ret;
3028
3029                                 ret = btrfs_discard_extent(root, bytenr,
3030                                                            num_bytes);
3031                                 WARN_ON(ret);
3032
3033                                 ret = btrfs_add_free_space(cache, bytenr,
3034                                                            num_bytes);
3035                                 WARN_ON(ret);
3036                         }
3037                 }
3038                 btrfs_put_block_group(cache);
3039                 total -= num_bytes;
3040                 bytenr += num_bytes;
3041         }
3042         return 0;
3043 }
3044
3045 static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
3046 {
3047         struct btrfs_block_group_cache *cache;
3048         u64 bytenr;
3049
3050         cache = btrfs_lookup_first_block_group(root->fs_info, search_start);
3051         if (!cache)
3052                 return 0;
3053
3054         bytenr = cache->key.objectid;
3055         btrfs_put_block_group(cache);
3056
3057         return bytenr;
3058 }
3059
3060 int btrfs_update_pinned_extents(struct btrfs_root *root,
3061                                 u64 bytenr, u64 num, int pin)
3062 {
3063         u64 len;
3064         struct btrfs_block_group_cache *cache;
3065         struct btrfs_fs_info *fs_info = root->fs_info;
3066
3067         if (pin)
3068                 set_extent_dirty(&fs_info->pinned_extents,
3069                                 bytenr, bytenr + num - 1, GFP_NOFS);
3070
3071         while (num > 0) {
3072                 cache = btrfs_lookup_block_group(fs_info, bytenr);
3073                 BUG_ON(!cache);
3074                 len = min(num, cache->key.offset -
3075                           (bytenr - cache->key.objectid));
3076                 if (pin) {
3077                         spin_lock(&cache->space_info->lock);
3078                         spin_lock(&cache->lock);
3079                         cache->pinned += len;
3080                         cache->space_info->bytes_pinned += len;
3081                         spin_unlock(&cache->lock);
3082                         spin_unlock(&cache->space_info->lock);
3083                         fs_info->total_pinned += len;
3084                 } else {
3085                         int unpin = 0;
3086
3087                         /*
3088                          * in order to not race with the block group caching, we
3089                          * only want to unpin the extent if we are cached.  If
3090                          * we aren't cached, we want to start async caching this
3091                          * block group so we can free the extent the next time
3092                          * around.
3093                          */
3094                         spin_lock(&cache->space_info->lock);
3095                         spin_lock(&cache->lock);
3096                         unpin = (cache->cached == BTRFS_CACHE_FINISHED);
3097                         if (likely(unpin)) {
3098                                 cache->pinned -= len;
3099                                 cache->space_info->bytes_pinned -= len;
3100                                 fs_info->total_pinned -= len;
3101                         }
3102                         spin_unlock(&cache->lock);
3103                         spin_unlock(&cache->space_info->lock);
3104
3105                         if (likely(unpin))
3106                                 clear_extent_dirty(&fs_info->pinned_extents,
3107                                                    bytenr, bytenr + len -1,
3108                                                    GFP_NOFS);
3109                         else
3110                                 cache_block_group(cache);
3111
3112                         if (unpin)
3113                                 btrfs_add_free_space(cache, bytenr, len);
3114                 }
3115                 btrfs_put_block_group(cache);
3116                 bytenr += len;
3117                 num -= len;
3118         }
3119         return 0;
3120 }
3121
3122 static int update_reserved_extents(struct btrfs_root *root,
3123                                    u64 bytenr, u64 num, int reserve)
3124 {
3125         u64 len;
3126         struct btrfs_block_group_cache *cache;
3127         struct btrfs_fs_info *fs_info = root->fs_info;
3128
3129         while (num > 0) {
3130                 cache = btrfs_lookup_block_group(fs_info, bytenr);
3131                 BUG_ON(!cache);
3132                 len = min(num, cache->key.offset -
3133                           (bytenr - cache->key.objectid));
3134
3135                 spin_lock(&cache->space_info->lock);
3136                 spin_lock(&cache->lock);
3137                 if (reserve) {
3138                         cache->reserved += len;
3139                         cache->space_info->bytes_reserved += len;
3140                 } else {
3141                         cache->reserved -= len;
3142                         cache->space_info->bytes_reserved -= len;
3143                 }
3144                 spin_unlock(&cache->lock);
3145                 spin_unlock(&cache->space_info->lock);
3146                 btrfs_put_block_group(cache);
3147                 bytenr += len;
3148                 num -= len;
3149         }
3150         return 0;
3151 }
3152
3153 int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy)
3154 {
3155         u64 last = 0;
3156         u64 start;
3157         u64 end;
3158         struct extent_io_tree *pinned_extents = &root->fs_info->pinned_extents;
3159         int ret;
3160
3161         while (1) {
3162                 ret = find_first_extent_bit(pinned_extents, last,
3163                                             &start, &end, EXTENT_DIRTY);
3164                 if (ret)
3165                         break;
3166
3167                 set_extent_dirty(copy, start, end, GFP_NOFS);
3168                 last = end + 1;
3169         }
3170         return 0;
3171 }
3172
3173 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
3174                                struct btrfs_root *root,
3175                                struct extent_io_tree *unpin)
3176 {
3177         u64 start;
3178         u64 end;
3179         int ret;
3180
3181         while (1) {
3182                 ret = find_first_extent_bit(unpin, 0, &start, &end,
3183                                             EXTENT_DIRTY);
3184                 if (ret)
3185                         break;
3186
3187                 ret = btrfs_discard_extent(root, start, end + 1 - start);
3188
3189                 /* unlocks the pinned mutex */
3190                 btrfs_update_pinned_extents(root, start, end + 1 - start, 0);
3191                 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3192
3193                 cond_resched();
3194         }
3195
3196         return ret;
3197 }
3198
3199 static int pin_down_bytes(struct btrfs_trans_handle *trans,
3200                           struct btrfs_root *root,
3201                           struct btrfs_path *path,
3202                           u64 bytenr, u64 num_bytes, int is_data,
3203                           struct extent_buffer **must_clean)
3204 {
3205         int err = 0;
3206         struct extent_buffer *buf;
3207
3208         if (is_data)
3209                 goto pinit;
3210
3211         buf = btrfs_find_tree_block(root, bytenr, num_bytes);
3212         if (!buf)
3213                 goto pinit;
3214
3215         /* we can reuse a block if it hasn't been written
3216          * and it is from this transaction.  We can't
3217          * reuse anything from the tree log root because
3218          * it has tiny sub-transactions.
3219          */
3220         if (btrfs_buffer_uptodate(buf, 0) &&
3221             btrfs_try_tree_lock(buf)) {
3222                 u64 header_owner = btrfs_header_owner(buf);
3223                 u64 header_transid = btrfs_header_generation(buf);
3224                 if (header_owner != BTRFS_TREE_LOG_OBJECTID &&
3225                     header_transid == trans->transid &&
3226                     !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
3227                         *must_clean = buf;
3228                         return 1;
3229                 }
3230                 btrfs_tree_unlock(buf);
3231         }
3232         free_extent_buffer(buf);
3233 pinit:
3234         btrfs_set_path_blocking(path);
3235         /* unlocks the pinned mutex */
3236         btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
3237
3238         BUG_ON(err < 0);
3239         return 0;
3240 }
3241
3242
3243 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
3244                                 struct btrfs_root *root,
3245                                 u64 bytenr, u64 num_bytes, u64 parent,
3246                                 u64 root_objectid, u64 owner_objectid,
3247                                 u64 owner_offset, int refs_to_drop,
3248                                 struct btrfs_delayed_extent_op *extent_op)
3249 {
3250         struct btrfs_key key;
3251         struct btrfs_path *path;
3252         struct btrfs_fs_info *info = root->fs_info;
3253         struct btrfs_root *extent_root = info->extent_root;
3254         struct extent_buffer *leaf;
3255         struct btrfs_extent_item *ei;
3256         struct btrfs_extent_inline_ref *iref;
3257         int ret;
3258         int is_data;
3259         int extent_slot = 0;
3260         int found_extent = 0;
3261         int num_to_del = 1;
3262         u32 item_size;
3263         u64 refs;
3264
3265         path = btrfs_alloc_path();
3266         if (!path)
3267                 return -ENOMEM;
3268
3269         path->reada = 1;
3270         path->leave_spinning = 1;
3271
3272         is_data = owner_objectid >= BTRFS_FIRST_FREE_OBJECTID;
3273         BUG_ON(!is_data && refs_to_drop != 1);
3274
3275         ret = lookup_extent_backref(trans, extent_root, path, &iref,
3276                                     bytenr, num_bytes, parent,
3277                                     root_objectid, owner_objectid,
3278                                     owner_offset);
3279         if (ret == 0) {
3280                 extent_slot = path->slots[0];
3281                 while (extent_slot >= 0) {
3282                         btrfs_item_key_to_cpu(path->nodes[0], &key,
3283                                               extent_slot);
3284                         if (key.objectid != bytenr)
3285                                 break;
3286                         if (key.type == BTRFS_EXTENT_ITEM_KEY &&
3287                             key.offset == num_bytes) {
3288                                 found_extent = 1;
3289                                 break;
3290                         }
3291                         if (path->slots[0] - extent_slot > 5)
3292                                 break;
3293                         extent_slot--;
3294                 }
3295 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
3296                 item_size = btrfs_item_size_nr(path->nodes[0], extent_slot);
3297                 if (found_extent && item_size < sizeof(*ei))
3298                         found_extent = 0;
3299 #endif
3300                 if (!found_extent) {
3301                         BUG_ON(iref);
3302                         ret = remove_extent_backref(trans, extent_root, path,
3303                                                     NULL, refs_to_drop,
3304                                                     is_data);
3305                         BUG_ON(ret);
3306                         btrfs_release_path(extent_root, path);
3307                         path->leave_spinning = 1;
3308
3309                         key.objectid = bytenr;
3310                         key.type = BTRFS_EXTENT_ITEM_KEY;
3311                         key.offset = num_bytes;
3312
3313                         ret = btrfs_search_slot(trans, extent_root,
3314                                                 &key, path, -1, 1);
3315                         if (ret) {
3316                                 printk(KERN_ERR "umm, got %d back from search"
3317                                        ", was looking for %llu\n", ret,
3318                                        (unsigned long long)bytenr);
3319                                 btrfs_print_leaf(extent_root, path->nodes[0]);
3320                         }
3321                         BUG_ON(ret);
3322                         extent_slot = path->slots[0];
3323                 }
3324         } else {
3325                 btrfs_print_leaf(extent_root, path->nodes[0]);
3326                 WARN_ON(1);
3327                 printk(KERN_ERR "btrfs unable to find ref byte nr %llu "
3328                        "parent %llu root %llu  owner %llu offset %llu\n",
3329                        (unsigned long long)bytenr,
3330                        (unsigned long long)parent,
3331                        (unsigned long long)root_objectid,
3332                        (unsigned long long)owner_objectid,
3333                        (unsigned long long)owner_offset);
3334         }
3335
3336         leaf = path->nodes[0];
3337         item_size = btrfs_item_size_nr(leaf, extent_slot);
3338 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
3339         if (item_size < sizeof(*ei)) {
3340                 BUG_ON(found_extent || extent_slot != path->slots[0]);
3341                 ret = convert_extent_item_v0(trans, extent_root, path,
3342                                              owner_objectid, 0);
3343                 BUG_ON(ret < 0);
3344
3345                 btrfs_release_path(extent_root, path);
3346                 path->leave_spinning = 1;
3347
3348                 key.objectid = bytenr;
3349                 key.type = BTRFS_EXTENT_ITEM_KEY;
3350                 key.offset = num_bytes;
3351
3352                 ret = btrfs_search_slot(trans, extent_root, &key, path,
3353                                         -1, 1);
3354                 if (ret) {
3355                         printk(KERN_ERR "umm, got %d back from search"
3356                                ", was looking for %llu\n", ret,
3357                                (unsigned long long)bytenr);
3358                         btrfs_print_leaf(extent_root, path->nodes[0]);
3359                 }
3360                 BUG_ON(ret);
3361                 extent_slot = path->slots[0];
3362                 leaf = path->nodes[0];
3363                 item_size = btrfs_item_size_nr(leaf, extent_slot);
3364         }
3365 #endif
3366         BUG_ON(item_size < sizeof(*ei));
3367         ei = btrfs_item_ptr(leaf, extent_slot,
3368                             struct btrfs_extent_item);
3369         if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
3370                 struct btrfs_tree_block_info *bi;
3371                 BUG_ON(item_size < sizeof(*ei) + sizeof(*bi));
3372                 bi = (struct btrfs_tree_block_info *)(ei + 1);
3373                 WARN_ON(owner_objectid != btrfs_tree_block_level(leaf, bi));
3374         }
3375
3376         refs = btrfs_extent_refs(leaf, ei);
3377         BUG_ON(refs < refs_to_drop);
3378         refs -= refs_to_drop;
3379
3380         if (refs > 0) {
3381                 if (extent_op)
3382                         __run_delayed_extent_op(extent_op, leaf, ei);
3383                 /*
3384                  * In the case of inline back ref, reference count will
3385                  * be updated by remove_extent_backref
3386                  */
3387                 if (iref) {
3388                         BUG_ON(!found_extent);
3389                 } else {
3390                         btrfs_set_extent_refs(leaf, ei, refs);
3391                         btrfs_mark_buffer_dirty(leaf);
3392                 }
3393                 if (found_extent) {
3394                         ret = remove_extent_backref(trans, extent_root, path,
3395                                                     iref, refs_to_drop,
3396                                                     is_data);
3397                         BUG_ON(ret);
3398                 }
3399         } else {
3400                 int mark_free = 0;
3401                 struct extent_buffer *must_clean = NULL;
3402
3403                 if (found_extent) {
3404                         BUG_ON(is_data && refs_to_drop !=
3405                                extent_data_ref_count(root, path, iref));
3406                         if (iref) {
3407                                 BUG_ON(path->slots[0] != extent_slot);
3408                         } else {
3409                                 BUG_ON(path->slots[0] != extent_slot + 1);
3410                                 path->slots[0] = extent_slot;
3411                                 num_to_del = 2;
3412                         }
3413                 }
3414
3415                 ret = pin_down_bytes(trans, root, path, bytenr,
3416                                      num_bytes, is_data, &must_clean);
3417                 if (ret > 0)
3418                         mark_free = 1;
3419                 BUG_ON(ret < 0);
3420                 /*
3421                  * it is going to be very rare for someone to be waiting
3422                  * on the block we're freeing.  del_items might need to
3423                  * schedule, so rather than get fancy, just force it
3424                  * to blocking here
3425                  */
3426                 if (must_clean)
3427                         btrfs_set_lock_blocking(must_clean);
3428
3429                 ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
3430                                       num_to_del);
3431                 BUG_ON(ret);
3432                 btrfs_release_path(extent_root, path);
3433
3434                 if (must_clean) {
3435                         clean_tree_block(NULL, root, must_clean);
3436                         btrfs_tree_unlock(must_clean);
3437                         free_extent_buffer(must_clean);
3438                 }
3439
3440                 if (is_data) {
3441                         ret = btrfs_del_csums(trans, root, bytenr, num_bytes);
3442                         BUG_ON(ret);
3443                 } else {
3444                         invalidate_mapping_pages(info->btree_inode->i_mapping,
3445                              bytenr >> PAGE_CACHE_SHIFT,
3446                              (bytenr + num_bytes - 1) >> PAGE_CACHE_SHIFT);
3447                 }
3448
3449                 ret = update_block_group(trans, root, bytenr, num_bytes, 0,
3450                                          mark_free);
3451                 BUG_ON(ret);
3452         }
3453         btrfs_free_path(path);
3454         return ret;
3455 }
3456
3457 /*
3458  * when we free an extent, it is possible (and likely) that we free the last
3459  * delayed ref for that extent as well.  This searches the delayed ref tree for
3460  * a given extent, and if there are no other delayed refs to be processed, it
3461  * removes it from the tree.
3462  */
3463 static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
3464                                       struct btrfs_root *root, u64 bytenr)
3465 {
3466         struct btrfs_delayed_ref_head *head;
3467         struct btrfs_delayed_ref_root *delayed_refs;
3468         struct btrfs_delayed_ref_node *ref;
3469         struct rb_node *node;
3470         int ret;
3471
3472         delayed_refs = &trans->transaction->delayed_refs;
3473         spin_lock(&delayed_refs->lock);
3474         head = btrfs_find_delayed_ref_head(trans, bytenr);
3475         if (!head)
3476                 goto out;
3477
3478         node = rb_prev(&head->node.rb_node);
3479         if (!node)
3480                 goto out;
3481
3482         ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
3483
3484         /* there are still entries for this ref, we can't drop it */
3485         if (ref->bytenr == bytenr)
3486                 goto out;
3487
3488         if (head->extent_op) {
3489                 if (!head->must_insert_reserved)
3490                         goto out;
3491                 kfree(head->extent_op);
3492                 head->extent_op = NULL;
3493         }
3494
3495         /*
3496          * waiting for the lock here would deadlock.  If someone else has it
3497          * locked they are already in the process of dropping it anyway
3498          */
3499         if (!mutex_trylock(&head->mutex))
3500                 goto out;
3501
3502         /*
3503          * at this point we have a head with no other entries.  Go
3504          * ahead and process it.
3505          */
3506         head->node.in_tree = 0;
3507         rb_erase(&head->node.rb_node, &delayed_refs->root);
3508
3509         delayed_refs->num_entries--;
3510
3511         /*
3512          * we don't take a ref on the node because we're removing it from the
3513          * tree, so we just steal the ref the tree was holding.
3514          */
3515         delayed_refs->num_heads--;
3516         if (list_empty(&head->cluster))
3517                 delayed_refs->num_heads_ready--;
3518
3519         list_del_init(&head->cluster);
3520         spin_unlock(&delayed_refs->lock);
3521
3522         ret = run_one_delayed_ref(trans, root->fs_info->tree_root,
3523                                   &head->node, head->extent_op,
3524                                   head->must_insert_reserved);
3525         BUG_ON(ret);
3526         btrfs_put_delayed_ref(&head->node);
3527         return 0;
3528 out:
3529         spin_unlock(&delayed_refs->lock);
3530         return 0;
3531 }
3532
3533 int btrfs_free_extent(struct btrfs_trans_handle *trans,
3534                       struct btrfs_root *root,
3535                       u64 bytenr, u64 num_bytes, u64 parent,
3536                       u64 root_objectid, u64 owner, u64 offset)
3537 {
3538         int ret;
3539
3540         /*
3541          * tree log blocks never actually go into the extent allocation
3542          * tree, just update pinning info and exit early.
3543          */
3544         if (root_objectid == BTRFS_TREE_LOG_OBJECTID) {
3545                 WARN_ON(owner >= BTRFS_FIRST_FREE_OBJECTID);
3546                 /* unlocks the pinned mutex */
3547                 btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
3548                 update_reserved_extents(root, bytenr, num_bytes, 0);
3549                 ret = 0;
3550         } else if (owner < BTRFS_FIRST_FREE_OBJECTID) {
3551                 ret = btrfs_add_delayed_tree_ref(trans, bytenr, num_bytes,
3552                                         parent, root_objectid, (int)owner,
3553                                         BTRFS_DROP_DELAYED_REF, NULL);
3554                 BUG_ON(ret);
3555                 ret = check_ref_cleanup(trans, root, bytenr);
3556                 BUG_ON(ret);
3557         } else {
3558                 ret = btrfs_add_delayed_data_ref(trans, bytenr, num_bytes,
3559                                         parent, root_objectid, owner,
3560                                         offset, BTRFS_DROP_DELAYED_REF, NULL);
3561                 BUG_ON(ret);
3562         }
3563         return ret;
3564 }
3565
3566 static u64 stripe_align(struct btrfs_root *root, u64 val)
3567 {
3568         u64 mask = ((u64)root->stripesize - 1);
3569         u64 ret = (val + mask) & ~mask;
3570         return ret;
3571 }
3572
3573 /*
3574  * when we wait for progress in the block group caching, its because
3575  * our allocation attempt failed at least once.  So, we must sleep
3576  * and let some progress happen before we try again.
3577  *
3578  * This function will sleep at least once waiting for new free space to
3579  * show up, and then it will check the block group free space numbers
3580  * for our min num_bytes.  Another option is to have it go ahead
3581  * and look in the rbtree for a free extent of a given size, but this
3582  * is a good start.
3583  */
3584 static noinline int
3585 wait_block_group_cache_progress(struct btrfs_block_group_cache *cache,
3586                                 u64 num_bytes)
3587 {
3588         DEFINE_WAIT(wait);
3589
3590         prepare_to_wait(&cache->caching_q, &wait, TASK_UNINTERRUPTIBLE);
3591
3592         if (block_group_cache_done(cache)) {
3593                 finish_wait(&cache->caching_q, &wait);
3594                 return 0;
3595         }
3596         schedule();
3597         finish_wait(&cache->caching_q, &wait);
3598
3599         wait_event(cache->caching_q, block_group_cache_done(cache) ||
3600                    (cache->free_space >= num_bytes));
3601         return 0;
3602 }
3603
3604 enum btrfs_loop_type {
3605         LOOP_CACHED_ONLY = 0,
3606         LOOP_CACHING_NOWAIT = 1,
3607         LOOP_CACHING_WAIT = 2,
3608         LOOP_ALLOC_CHUNK = 3,
3609         LOOP_NO_EMPTY_SIZE = 4,
3610 };
3611
3612 /*
3613  * walks the btree of allocated extents and find a hole of a given size.
3614  * The key ins is changed to record the hole:
3615  * ins->objectid == block start
3616  * ins->flags = BTRFS_EXTENT_ITEM_KEY
3617  * ins->offset == number of blocks
3618  * Any available blocks before search_start are skipped.
3619  */
3620 static noinline int find_free_extent(struct btrfs_trans_handle *trans,
3621                                      struct btrfs_root *orig_root,
3622                                      u64 num_bytes, u64 empty_size,
3623                                      u64 search_start, u64 search_end,
3624                                      u64 hint_byte, struct btrfs_key *ins,
3625                                      u64 exclude_start, u64 exclude_nr,
3626                                      int data)
3627 {
3628         int ret = 0;
3629         struct btrfs_root *root = orig_root->fs_info->extent_root;
3630         struct btrfs_free_cluster *last_ptr = NULL;
3631         struct btrfs_block_group_cache *block_group = NULL;
3632         int empty_cluster = 2 * 1024 * 1024;
3633         int allowed_chunk_alloc = 0;
3634         struct btrfs_space_info *space_info;
3635         int last_ptr_loop = 0;
3636         int loop = 0;
3637         bool found_uncached_bg = false;
3638
3639         WARN_ON(num_bytes < root->sectorsize);
3640         btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
3641         ins->objectid = 0;
3642         ins->offset = 0;
3643
3644         space_info = __find_space_info(root->fs_info, data);
3645
3646         if (orig_root->ref_cows || empty_size)
3647                 allowed_chunk_alloc = 1;
3648
3649         if (data & BTRFS_BLOCK_GROUP_METADATA) {
3650                 last_ptr = &root->fs_info->meta_alloc_cluster;
3651                 if (!btrfs_test_opt(root, SSD))
3652                         empty_cluster = 64 * 1024;
3653         }
3654
3655         if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) {
3656                 last_ptr = &root->fs_info->data_alloc_cluster;
3657         }
3658
3659         if (last_ptr) {
3660                 spin_lock(&last_ptr->lock);
3661                 if (last_ptr->block_group)
3662                         hint_byte = last_ptr->window_start;
3663                 spin_unlock(&last_ptr->lock);
3664         }
3665
3666         search_start = max(search_start, first_logical_byte(root, 0));
3667         search_start = max(search_start, hint_byte);
3668
3669         if (!last_ptr)
3670                 empty_cluster = 0;
3671
3672         if (search_start == hint_byte) {
3673                 block_group = btrfs_lookup_block_group(root->fs_info,
3674                                                        search_start);
3675                 /*
3676                  * we don't want to use the block group if it doesn't match our
3677                  * allocation bits, or if its not cached.
3678                  */
3679                 if (block_group && block_group_bits(block_group, data) &&
3680                     block_group_cache_done(block_group)) {
3681                         down_read(&space_info->groups_sem);
3682                         if (list_empty(&block_group->list) ||
3683                             block_group->ro) {
3684                                 /*
3685                                  * someone is removing this block group,
3686                                  * we can't jump into the have_block_group
3687                                  * target because our list pointers are not
3688                                  * valid
3689                                  */
3690                                 btrfs_put_block_group(block_group);
3691                                 up_read(&space_info->groups_sem);
3692                         } else
3693                                 goto have_block_group;
3694                 } else if (block_group) {
3695                         btrfs_put_block_group(block_group);
3696                 }
3697         }
3698
3699 search:
3700         down_read(&space_info->groups_sem);
3701         list_for_each_entry(block_group, &space_info->block_groups, list) {
3702                 u64 offset;
3703                 int cached;
3704
3705                 atomic_inc(&block_group->count);
3706                 search_start = block_group->key.objectid;
3707
3708 have_block_group:
3709                 if (unlikely(block_group->cached == BTRFS_CACHE_NO)) {
3710                         /*
3711                          * we want to start caching kthreads, but not too many
3712                          * right off the bat so we don't overwhelm the system,
3713                          * so only start them if there are less than 2 and we're
3714                          * in the initial allocation phase.
3715                          */
3716                         if (loop > LOOP_CACHING_NOWAIT ||
3717                             atomic_read(&space_info->caching_threads) < 2) {
3718                                 ret = cache_block_group(block_group);
3719                                 BUG_ON(ret);
3720                         }
3721                 }
3722
3723                 cached = block_group_cache_done(block_group);
3724                 if (unlikely(!cached)) {
3725                         found_uncached_bg = true;
3726
3727                         /* if we only want cached bgs, loop */
3728                         if (loop == LOOP_CACHED_ONLY)
3729                                 goto loop;
3730                 }
3731
3732                 if (unlikely(block_group->ro))
3733                         goto loop;
3734
3735                 if (last_ptr) {
3736                         /*
3737                          * the refill lock keeps out other
3738                          * people trying to start a new cluster
3739                          */
3740                         spin_lock(&last_ptr->refill_lock);
3741                         if (last_ptr->block_group &&
3742                             (last_ptr->block_group->ro ||
3743                             !block_group_bits(last_ptr->block_group, data))) {
3744                                 offset = 0;
3745                                 goto refill_cluster;
3746                         }
3747
3748                         offset = btrfs_alloc_from_cluster(block_group, last_ptr,
3749                                                  num_bytes, search_start);
3750                         if (offset) {
3751                                 /* we have a block, we're done */
3752                                 spin_unlock(&last_ptr->refill_lock);
3753                                 goto checks;
3754                         }
3755
3756                         spin_lock(&last_ptr->lock);
3757                         /*
3758                          * whoops, this cluster doesn't actually point to
3759                          * this block group.  Get a ref on the block
3760                          * group is does point to and try again
3761                          */
3762                         if (!last_ptr_loop && last_ptr->block_group &&
3763                             last_ptr->block_group != block_group) {
3764
3765                                 btrfs_put_block_group(block_group);
3766                                 block_group = last_ptr->block_group;
3767                                 atomic_inc(&block_group->count);
3768                                 spin_unlock(&last_ptr->lock);
3769                                 spin_unlock(&last_ptr->refill_lock);
3770
3771                                 last_ptr_loop = 1;
3772                                 search_start = block_group->key.objectid;
3773                                 /*
3774                                  * we know this block group is properly
3775                                  * in the list because
3776                                  * btrfs_remove_block_group, drops the
3777                                  * cluster before it removes the block
3778                                  * group from the list
3779                                  */
3780                                 goto have_block_group;
3781                         }
3782                         spin_unlock(&last_ptr->lock);
3783 refill_cluster:
3784                         /*
3785                          * this cluster didn't work out, free it and
3786                          * start over
3787                          */
3788                         btrfs_return_cluster_to_free_space(NULL, last_ptr);
3789
3790                         last_ptr_loop = 0;
3791
3792                         /* allocate a cluster in this block group */
3793                         ret = btrfs_find_space_cluster(trans, root,
3794                                                block_group, last_ptr,
3795                                                offset, num_bytes,
3796                                                empty_cluster + empty_size);
3797                         if (ret == 0) {
3798                                 /*
3799                                  * now pull our allocation out of this
3800                                  * cluster
3801                                  */
3802                                 offset = btrfs_alloc_from_cluster(block_group,
3803                                                   last_ptr, num_bytes,
3804                                                   search_start);
3805                                 if (offset) {
3806                                         /* we found one, proceed */
3807                                         spin_unlock(&last_ptr->refill_lock);
3808                                         goto checks;
3809                                 }
3810                         } else if (!cached && loop > LOOP_CACHING_NOWAIT) {
3811                                 spin_unlock(&last_ptr->refill_lock);
3812
3813                                 wait_block_group_cache_progress(block_group,
3814                                        num_bytes + empty_cluster + empty_size);
3815                                 goto have_block_group;
3816                         }
3817
3818                         /*
3819                          * at this point we either didn't find a cluster
3820                          * or we weren't able to allocate a block from our
3821                          * cluster.  Free the cluster we've been trying
3822                          * to use, and go to the next block group
3823                          */
3824                         if (loop < LOOP_NO_EMPTY_SIZE) {
3825                                 btrfs_return_cluster_to_free_space(NULL,
3826                                                                    last_ptr);
3827                                 spin_unlock(&last_ptr->refill_lock);
3828                                 goto loop;
3829                         }
3830                         spin_unlock(&last_ptr->refill_lock);
3831                 }
3832
3833                 offset = btrfs_find_space_for_alloc(block_group, search_start,
3834                                                     num_bytes, empty_size);
3835                 if (!offset && (cached || (!cached &&
3836                                            loop == LOOP_CACHING_NOWAIT))) {
3837                         goto loop;
3838                 } else if (!offset && (!cached &&
3839                                        loop > LOOP_CACHING_NOWAIT)) {
3840                         wait_block_group_cache_progress(block_group,
3841                                         num_bytes + empty_size);
3842                         goto have_block_group;
3843                 }
3844 checks:
3845                 search_start = stripe_align(root, offset);
3846                 /* move on to the next group */
3847                 if (search_start + num_bytes >= search_end) {
3848                         btrfs_add_free_space(block_group, offset, num_bytes);
3849                         goto loop;
3850                 }
3851
3852                 /* move on to the next group */
3853                 if (search_start + num_bytes >
3854                     block_group->key.objectid + block_group->key.offset) {
3855                         btrfs_add_free_space(block_group, offset, num_bytes);
3856                         goto loop;
3857                 }
3858
3859                 if (exclude_nr > 0 &&
3860                     (search_start + num_bytes > exclude_start &&
3861                      search_start < exclude_start + exclude_nr)) {
3862                         search_start = exclude_start + exclude_nr;
3863
3864                         btrfs_add_free_space(block_group, offset, num_bytes);
3865                         /*
3866                          * if search_start is still in this block group
3867                          * then we just re-search this block group
3868                          */
3869                         if (search_start >= block_group->key.objectid &&
3870                             search_start < (block_group->key.objectid +
3871                                             block_group->key.offset))
3872                                 goto have_block_group;
3873                         goto loop;
3874                 }
3875
3876                 ins->objectid = search_start;
3877                 ins->offset = num_bytes;
3878
3879                 if (offset < search_start)
3880                         btrfs_add_free_space(block_group, offset,
3881                                              search_start - offset);
3882                 BUG_ON(offset > search_start);
3883
3884                 /* we are all good, lets return */
3885                 break;
3886 loop:
3887                 btrfs_put_block_group(block_group);
3888         }
3889         up_read(&space_info->groups_sem);
3890
3891         /* LOOP_CACHED_ONLY, only search fully cached block groups
3892          * LOOP_CACHING_NOWAIT, search partially cached block groups, but
3893          *                      dont wait foR them to finish caching
3894          * LOOP_CACHING_WAIT, search everything, and wait if our bg is caching
3895          * LOOP_ALLOC_CHUNK, force a chunk allocation and try again
3896          * LOOP_NO_EMPTY_SIZE, set empty_size and empty_cluster to 0 and try
3897          *                      again
3898          */
3899         if (!ins->objectid && loop < LOOP_NO_EMPTY_SIZE &&
3900             (found_uncached_bg || empty_size || empty_cluster ||
3901              allowed_chunk_alloc)) {
3902                 if (found_uncached_bg) {
3903                         found_uncached_bg = false;
3904                         if (loop < LOOP_CACHING_WAIT) {
3905                                 loop++;
3906                                 goto search;
3907                         }
3908                 }
3909
3910                 if (loop == LOOP_ALLOC_CHUNK) {
3911                         empty_size = 0;
3912                         empty_cluster = 0;
3913                 }
3914
3915                 if (allowed_chunk_alloc) {
3916                         ret = do_chunk_alloc(trans, root, num_bytes +
3917                                              2 * 1024 * 1024, data, 1);
3918                         allowed_chunk_alloc = 0;
3919                 } else {
3920                         space_info->force_alloc = 1;
3921                 }
3922
3923                 if (loop < LOOP_NO_EMPTY_SIZE) {
3924                         loop++;
3925                         goto search;
3926                 }
3927                 ret = -ENOSPC;
3928         } else if (!ins->objectid) {
3929                 ret = -ENOSPC;
3930         }
3931
3932         /* we found what we needed */
3933         if (ins->objectid) {
3934                 if (!(data & BTRFS_BLOCK_GROUP_DATA))
3935                         trans->block_group = block_group->key.objectid;
3936
3937                 btrfs_put_block_group(block_group);
3938                 ret = 0;
3939         }
3940
3941         return ret;
3942 }
3943
3944 static void dump_space_info(struct btrfs_space_info *info, u64 bytes)
3945 {
3946         struct btrfs_block_group_cache *cache;
3947
3948         printk(KERN_INFO "space_info has %llu free, is %sfull\n",
3949                (unsigned long long)(info->total_bytes - info->bytes_used -
3950                                     info->bytes_pinned - info->bytes_reserved),
3951                (info->full) ? "" : "not ");
3952         printk(KERN_INFO "space_info total=%llu, pinned=%llu, delalloc=%llu,"
3953                " may_use=%llu, used=%llu\n",
3954                (unsigned long long)info->total_bytes,
3955                (unsigned long long)info->bytes_pinned,
3956                (unsigned long long)info->bytes_delalloc,
3957                (unsigned long long)info->bytes_may_use,
3958                (unsigned long long)info->bytes_used);
3959
3960         down_read(&info->groups_sem);
3961         list_for_each_entry(cache, &info->block_groups, list) {
3962                 spin_lock(&cache->lock);
3963                 printk(KERN_INFO "block group %llu has %llu bytes, %llu used "
3964                        "%llu pinned %llu reserved\n",
3965                        (unsigned long long)cache->key.objectid,
3966                        (unsigned long long)cache->key.offset,
3967                        (unsigned long long)btrfs_block_group_used(&cache->item),
3968                        (unsigned long long)cache->pinned,
3969                        (unsigned long long)cache->reserved);
3970                 btrfs_dump_free_space(cache, bytes);
3971                 spin_unlock(&cache->lock);
3972         }
3973         up_read(&info->groups_sem);
3974 }
3975
3976 static int __btrfs_reserve_extent(struct btrfs_trans_handle *trans,
3977                                   struct btrfs_root *root,
3978                                   u64 num_bytes, u64 min_alloc_size,
3979                                   u64 empty_size, u64 hint_byte,
3980                                   u64 search_end, struct btrfs_key *ins,
3981                                   u64 data)
3982 {
3983         int ret;
3984         u64 search_start = 0;
3985         struct btrfs_fs_info *info = root->fs_info;
3986
3987         data = btrfs_get_alloc_profile(root, data);
3988 again:
3989         /*
3990          * the only place that sets empty_size is btrfs_realloc_node, which
3991          * is not called recursively on allocations
3992          */
3993         if (empty_size || root->ref_cows) {
3994                 if (!(data & BTRFS_BLOCK_GROUP_METADATA)) {
3995                         ret = do_chunk_alloc(trans, root->fs_info->extent_root,
3996                                      2 * 1024 * 1024,
3997                                      BTRFS_BLOCK_GROUP_METADATA |
3998                                      (info->metadata_alloc_profile &
3999                                       info->avail_metadata_alloc_bits), 0);
4000                 }
4001                 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
4002                                      num_bytes + 2 * 1024 * 1024, data, 0);
4003         }
4004
4005         WARN_ON(num_bytes < root->sectorsize);
4006         ret = find_free_extent(trans, root, num_bytes, empty_size,
4007                                search_start, search_end, hint_byte, ins,
4008                                trans->alloc_exclude_start,
4009                                trans->alloc_exclude_nr, data);
4010
4011         if (ret == -ENOSPC && num_bytes > min_alloc_size) {
4012                 num_bytes = num_bytes >> 1;
4013                 num_bytes = num_bytes & ~(root->sectorsize - 1);
4014                 num_bytes = max(num_bytes, min_alloc_size);
4015                 do_chunk_alloc(trans, root->fs_info->extent_root,
4016                                num_bytes, data, 1);
4017                 goto again;
4018         }
4019         if (ret == -ENOSPC) {
4020                 struct btrfs_space_info *sinfo;
4021
4022                 sinfo = __find_space_info(root->fs_info, data);
4023                 printk(KERN_ERR "btrfs allocation failed flags %llu, "
4024                        "wanted %llu\n", (unsigned long long)data,
4025                        (unsigned long long)num_bytes);
4026                 dump_space_info(sinfo, num_bytes);
4027         }
4028
4029         return ret;
4030 }
4031
4032 int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
4033 {
4034         struct btrfs_block_group_cache *cache;
4035         int ret = 0;
4036
4037         cache = btrfs_lookup_block_group(root->fs_info, start);
4038         if (!cache) {
4039                 printk(KERN_ERR "Unable to find block group for %llu\n",
4040                        (unsigned long long)start);
4041                 return -ENOSPC;
4042         }
4043
4044         ret = btrfs_discard_extent(root, start, len);
4045
4046         btrfs_add_free_space(cache, start, len);
4047         btrfs_put_block_group(cache);
4048         update_reserved_extents(root, start, len, 0);
4049
4050         return ret;
4051 }
4052
4053 int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
4054                                   struct btrfs_root *root,
4055                                   u64 num_bytes, u64 min_alloc_size,
4056                                   u64 empty_size, u64 hint_byte,
4057                                   u64 search_end, struct btrfs_key *ins,
4058                                   u64 data)
4059 {
4060         int ret;
4061         ret = __btrfs_reserve_extent(trans, root, num_bytes, min_alloc_size,
4062                                      empty_size, hint_byte, search_end, ins,
4063                                      data);
4064         if (!ret)
4065                 update_reserved_extents(root, ins->objectid, ins->offset, 1);
4066
4067         return ret;
4068 }
4069
4070 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4071                                       struct btrfs_root *root,
4072                                       u64 parent, u64 root_objectid,
4073                                       u64 flags, u64 owner, u64 offset,
4074                                       struct btrfs_key *ins, int ref_mod)
4075 {
4076         int ret;
4077         struct btrfs_fs_info *fs_info = root->fs_info;
4078         struct btrfs_extent_item *extent_item;
4079         struct btrfs_extent_inline_ref *iref;
4080         struct btrfs_path *path;
4081         struct extent_buffer *leaf;
4082         int type;
4083         u32 size;
4084
4085         if (parent > 0)
4086                 type = BTRFS_SHARED_DATA_REF_KEY;
4087         else
4088                 type = BTRFS_EXTENT_DATA_REF_KEY;
4089
4090         size = sizeof(*extent_item) + btrfs_extent_inline_ref_size(type);
4091
4092         path = btrfs_alloc_path();
4093         BUG_ON(!path);
4094
4095         path->leave_spinning = 1;
4096         ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
4097                                       ins, size);
4098         BUG_ON(ret);
4099
4100         leaf = path->nodes[0];
4101         extent_item = btrfs_item_ptr(leaf, path->slots[0],
4102                                      struct btrfs_extent_item);
4103         btrfs_set_extent_refs(leaf, extent_item, ref_mod);
4104         btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4105         btrfs_set_extent_flags(leaf, extent_item,
4106                                flags | BTRFS_EXTENT_FLAG_DATA);
4107
4108         iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
4109         btrfs_set_extent_inline_ref_type(leaf, iref, type);
4110         if (parent > 0) {
4111                 struct btrfs_shared_data_ref *ref;
4112                 ref = (struct btrfs_shared_data_ref *)(iref + 1);
4113                 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
4114                 btrfs_set_shared_data_ref_count(leaf, ref, ref_mod);
4115         } else {
4116                 struct btrfs_extent_data_ref *ref;
4117                 ref = (struct btrfs_extent_data_ref *)(&iref->offset);
4118                 btrfs_set_extent_data_ref_root(leaf, ref, root_objectid);
4119                 btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
4120                 btrfs_set_extent_data_ref_offset(leaf, ref, offset);
4121                 btrfs_set_extent_data_ref_count(leaf, ref, ref_mod);
4122         }
4123
4124         btrfs_mark_buffer_dirty(path->nodes[0]);
4125         btrfs_free_path(path);
4126
4127         ret = update_block_group(trans, root, ins->objectid, ins->offset,
4128                                  1, 0);
4129         if (ret) {
4130                 printk(KERN_ERR "btrfs update block group failed for %llu "
4131                        "%llu\n", (unsigned long long)ins->objectid,
4132                        (unsigned long long)ins->offset);
4133                 BUG();
4134         }
4135         return ret;
4136 }
4137
4138 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
4139                                      struct btrfs_root *root,
4140                                      u64 parent, u64 root_objectid,
4141                                      u64 flags, struct btrfs_disk_key *key,
4142                                      int level, struct btrfs_key *ins)
4143 {
4144         int ret;
4145         struct btrfs_fs_info *fs_info = root->fs_info;
4146         struct btrfs_extent_item *extent_item;
4147         struct btrfs_tree_block_info *block_info;
4148         struct btrfs_extent_inline_ref *iref;
4149         struct btrfs_path *path;
4150         struct extent_buffer *leaf;
4151         u32 size = sizeof(*extent_item) + sizeof(*block_info) + sizeof(*iref);
4152
4153         path = btrfs_alloc_path();
4154         BUG_ON(!path);
4155
4156         path->leave_spinning = 1;
4157         ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
4158                                       ins, size);
4159         BUG_ON(ret);
4160
4161         leaf = path->nodes[0];
4162         extent_item = btrfs_item_ptr(leaf, path->slots[0],
4163                                      struct btrfs_extent_item);
4164         btrfs_set_extent_refs(leaf, extent_item, 1);
4165         btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4166         btrfs_set_extent_flags(leaf, extent_item,
4167                                flags | BTRFS_EXTENT_FLAG_TREE_BLOCK);
4168         block_info = (struct btrfs_tree_block_info *)(extent_item + 1);
4169
4170         btrfs_set_tree_block_key(leaf, block_info, key);
4171         btrfs_set_tree_block_level(leaf, block_info, level);
4172
4173         iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
4174         if (parent > 0) {
4175                 BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
4176                 btrfs_set_extent_inline_ref_type(leaf, iref,
4177                                                  BTRFS_SHARED_BLOCK_REF_KEY);
4178                 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
4179         } else {
4180                 btrfs_set_extent_inline_ref_type(leaf, iref,
4181                                                  BTRFS_TREE_BLOCK_REF_KEY);
4182                 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
4183         }
4184
4185         btrfs_mark_buffer_dirty(leaf);
4186         btrfs_free_path(path);
4187
4188         ret = update_block_group(trans, root, ins->objectid, ins->offset,
4189                                  1, 0);
4190         if (ret) {
4191                 printk(KERN_ERR "btrfs update block group failed for %llu "
4192                        "%llu\n", (unsigned long long)ins->objectid,
4193                        (unsigned long long)ins->offset);
4194                 BUG();
4195         }
4196         return ret;
4197 }
4198
4199 int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4200                                      struct btrfs_root *root,
4201                                      u64 root_objectid, u64 owner,
4202                                      u64 offset, struct btrfs_key *ins)
4203 {
4204         int ret;
4205
4206         BUG_ON(root_objectid == BTRFS_TREE_LOG_OBJECTID);
4207
4208         ret = btrfs_add_delayed_data_ref(trans, ins->objectid, ins->offset,
4209                                          0, root_objectid, owner, offset,
4210                                          BTRFS_ADD_DELAYED_EXTENT, NULL);
4211         return ret;
4212 }
4213
4214 /*
4215  * this is used by the tree logging recovery code.  It records that
4216  * an extent has been allocated and makes sure to clear the free
4217  * space cache bits as well
4218  */
4219 int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
4220                                    struct btrfs_root *root,
4221                                    u64 root_objectid, u64 owner, u64 offset,
4222                                    struct btrfs_key *ins)
4223 {
4224         int ret;
4225         struct btrfs_block_group_cache *block_group;
4226
4227         block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
4228         cache_block_group(block_group);
4229         wait_event(block_group->caching_q,
4230                    block_group_cache_done(block_group));
4231
4232         ret = btrfs_remove_free_space(block_group, ins->objectid,
4233                                       ins->offset);
4234         BUG_ON(ret);
4235         btrfs_put_block_group(block_group);
4236         ret = alloc_reserved_file_extent(trans, root, 0, root_objectid,
4237                                          0, owner, offset, ins, 1);
4238         return ret;
4239 }
4240
4241 /*
4242  * finds a free extent and does all the dirty work required for allocation
4243  * returns the key for the extent through ins, and a tree buffer for
4244  * the first block of the extent through buf.
4245  *
4246  * returns 0 if everything worked, non-zero otherwise.
4247  */
4248 static int alloc_tree_block(struct btrfs_trans_handle *trans,
4249                             struct btrfs_root *root,
4250                             u64 num_bytes, u64 parent, u64 root_objectid,
4251                             struct btrfs_disk_key *key, int level,
4252                             u64 empty_size, u64 hint_byte, u64 search_end,
4253                             struct btrfs_key *ins)
4254 {
4255         int ret;
4256         u64 flags = 0;
4257
4258         ret = __btrfs_reserve_extent(trans, root, num_bytes, num_bytes,
4259                                      empty_size, hint_byte, search_end,
4260                                      ins, 0);
4261         if (ret)
4262                 return ret;
4263
4264         if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
4265                 if (parent == 0)
4266                         parent = ins->objectid;
4267                 flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
4268         } else
4269                 BUG_ON(parent > 0);
4270
4271         update_reserved_extents(root, ins->objectid, ins->offset, 1);
4272         if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
4273                 struct btrfs_delayed_extent_op *extent_op;
4274                 extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
4275                 BUG_ON(!extent_op);
4276                 if (key)
4277                         memcpy(&extent_op->key, key, sizeof(extent_op->key));
4278                 else
4279                         memset(&extent_op->key, 0, sizeof(extent_op->key));
4280                 extent_op->flags_to_set = flags;
4281                 extent_op->update_key = 1;
4282                 extent_op->update_flags = 1;
4283                 extent_op->is_data = 0;
4284
4285                 ret = btrfs_add_delayed_tree_ref(trans, ins->objectid,
4286                                         ins->offset, parent, root_objectid,
4287                                         level, BTRFS_ADD_DELAYED_EXTENT,
4288                                         extent_op);
4289                 BUG_ON(ret);
4290         }
4291         return ret;
4292 }
4293
4294 struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
4295                                             struct btrfs_root *root,
4296                                             u64 bytenr, u32 blocksize,
4297                                             int level)
4298 {
4299         struct extent_buffer *buf;
4300
4301         buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
4302         if (!buf)
4303                 return ERR_PTR(-ENOMEM);
4304         btrfs_set_header_generation(buf, trans->transid);
4305         btrfs_set_buffer_lockdep_class(buf, level);
4306         btrfs_tree_lock(buf);
4307         clean_tree_block(trans, root, buf);
4308
4309         btrfs_set_lock_blocking(buf);
4310         btrfs_set_buffer_uptodate(buf);
4311
4312         if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
4313                 set_extent_dirty(&root->dirty_log_pages, buf->start,
4314                          buf->start + buf->len - 1, GFP_NOFS);
4315         } else {
4316                 set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
4317                          buf->start + buf->len - 1, GFP_NOFS);
4318         }
4319         trans->blocks_used++;
4320         /* this returns a buffer locked for blocking */
4321         return buf;
4322 }
4323
4324 /*
4325  * helper function to allocate a block for a given tree
4326  * returns the tree buffer or NULL.
4327  */
4328 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
4329                                         struct btrfs_root *root, u32 blocksize,
4330                                         u64 parent, u64 root_objectid,
4331                                         struct btrfs_disk_key *key, int level,
4332                                         u64 hint, u64 empty_size)
4333 {
4334         struct btrfs_key ins;
4335         int ret;
4336         struct extent_buffer *buf;
4337
4338         ret = alloc_tree_block(trans, root, blocksize, parent, root_objectid,
4339                                key, level, empty_size, hint, (u64)-1, &ins);
4340         if (ret) {
4341                 BUG_ON(ret > 0);
4342                 return ERR_PTR(ret);
4343         }
4344
4345         buf = btrfs_init_new_buffer(trans, root, ins.objectid,
4346                                     blocksize, level);
4347         return buf;
4348 }
4349
4350 #if 0
4351 int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans,
4352                         struct btrfs_root *root, struct extent_buffer *leaf)
4353 {
4354         u64 disk_bytenr;
4355         u64 num_bytes;
4356         struct btrfs_key key;
4357         struct btrfs_file_extent_item *fi;
4358         u32 nritems;
4359         int i;
4360         int ret;
4361
4362         BUG_ON(!btrfs_is_leaf(leaf));
4363         nritems = btrfs_header_nritems(leaf);
4364
4365         for (i = 0; i < nritems; i++) {
4366                 cond_resched();
4367                 btrfs_item_key_to_cpu(leaf, &key, i);
4368
4369                 /* only extents have references, skip everything else */
4370                 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
4371                         continue;
4372
4373                 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
4374
4375                 /* inline extents live in the btree, they don't have refs */
4376                 if (btrfs_file_extent_type(leaf, fi) ==
4377                     BTRFS_FILE_EXTENT_INLINE)
4378                         continue;
4379
4380                 disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
4381
4382                 /* holes don't have refs */
4383                 if (disk_bytenr == 0)
4384                         continue;
4385
4386                 num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
4387                 ret = btrfs_free_extent(trans, root, disk_bytenr, num_bytes,
4388                                         leaf->start, 0, key.objectid, 0);
4389                 BUG_ON(ret);
4390         }
4391         return 0;
4392 }
4393
4394 static noinline int cache_drop_leaf_ref(struct btrfs_trans_handle *trans,
4395                                         struct btrfs_root *root,
4396                                         struct btrfs_leaf_ref *ref)
4397 {
4398         int i;
4399         int ret;
4400         struct btrfs_extent_info *info;
4401         struct refsort *sorted;
4402
4403         if (ref->nritems == 0)
4404                 return 0;
4405
4406         sorted = kmalloc(sizeof(*sorted) * ref->nritems, GFP_NOFS);
4407         for (i = 0; i < ref->nritems; i++) {
4408                 sorted[i].bytenr = ref->extents[i].bytenr;
4409                 sorted[i].slot = i;
4410         }
4411         sort(sorted, ref->nritems, sizeof(struct refsort), refsort_cmp, NULL);
4412
4413         /*
4414          * the items in the ref were sorted when the ref was inserted
4415          * into the ref cache, so this is already in order
4416          */
4417         for (i = 0; i < ref->nritems; i++) {
4418                 info = ref->extents + sorted[i].slot;
4419                 ret = btrfs_free_extent(trans, root, info->bytenr,
4420                                           info->num_bytes, ref->bytenr,
4421                                           ref->owner, ref->generation,
4422                                           info->objectid, 0);
4423
4424                 atomic_inc(&root->fs_info->throttle_gen);
4425                 wake_up(&root->fs_info->transaction_throttle);
4426                 cond_resched();
4427
4428                 BUG_ON(ret);
4429                 info++;
4430         }
4431
4432         kfree(sorted);
4433         return 0;
4434 }
4435
4436
4437 static int drop_snap_lookup_refcount(struct btrfs_trans_handle *trans,
4438                                      struct btrfs_root *root, u64 start,
4439                                      u64 len, u32 *refs)
4440 {
4441         int ret;
4442
4443         ret = btrfs_lookup_extent_refs(trans, root, start, len, refs);
4444         BUG_ON(ret);
4445
4446 #if 0 /* some debugging code in case we see problems here */
4447         /* if the refs count is one, it won't get increased again.  But
4448          * if the ref count is > 1, someone may be decreasing it at
4449          * the same time we are.
4450          */
4451         if (*refs != 1) {
4452                 struct extent_buffer *eb = NULL;
4453                 eb = btrfs_find_create_tree_block(root, start, len);
4454                 if (eb)
4455                         btrfs_tree_lock(eb);
4456
4457                 mutex_lock(&root->fs_info->alloc_mutex);
4458                 ret = lookup_extent_ref(NULL, root, start, len, refs);
4459                 BUG_ON(ret);
4460                 mutex_unlock(&root->fs_info->alloc_mutex);
4461
4462                 if (eb) {
4463                         btrfs_tree_unlock(eb);
4464                         free_extent_buffer(eb);
4465                 }
4466                 if (*refs == 1) {
4467                         printk(KERN_ERR "btrfs block %llu went down to one "
4468                                "during drop_snap\n", (unsigned long long)start);
4469                 }
4470
4471         }
4472 #endif
4473
4474         cond_resched();
4475         return ret;
4476 }
4477
4478
4479 /*
4480  * this is used while deleting old snapshots, and it drops the refs
4481  * on a whole subtree starting from a level 1 node.
4482  *
4483  * The idea is to sort all the leaf pointers, and then drop the
4484  * ref on all the leaves in order.  Most of the time the leaves
4485  * will have ref cache entries, so no leaf IOs will be required to
4486  * find the extents they have references on.
4487  *
4488  * For each leaf, any references it has are also dropped in order
4489  *
4490  * This ends up dropping the references in something close to optimal
4491  * order for reading and modifying the extent allocation tree.
4492  */
4493 static noinline int drop_level_one_refs(struct btrfs_trans_handle *trans,
4494                                         struct btrfs_root *root,
4495                                         struct btrfs_path *path)
4496 {
4497         u64 bytenr;
4498         u64 root_owner;
4499         u64 root_gen;
4500         struct extent_buffer *eb = path->nodes[1];
4501         struct extent_buffer *leaf;
4502         struct btrfs_leaf_ref *ref;
4503         struct refsort *sorted = NULL;
4504         int nritems = btrfs_header_nritems(eb);
4505         int ret;
4506         int i;
4507         int refi = 0;
4508         int slot = path->slots[1];
4509         u32 blocksize = btrfs_level_size(root, 0);
4510         u32 refs;
4511
4512         if (nritems == 0)
4513                 goto out;
4514
4515         root_owner = btrfs_header_owner(eb);
4516         root_gen = btrfs_header_generation(eb);
4517         sorted = kmalloc(sizeof(*sorted) * nritems, GFP_NOFS);
4518
4519         /*
4520          * step one, sort all the leaf pointers so we don't scribble
4521          * randomly into the extent allocation tree
4522          */
4523         for (i = slot; i < nritems; i++) {
4524                 sorted[refi].bytenr = btrfs_node_blockptr(eb, i);
4525                 sorted[refi].slot = i;
4526                 refi++;
4527         }
4528
4529         /*
4530          * nritems won't be zero, but if we're picking up drop_snapshot
4531          * after a crash, slot might be > 0, so double check things
4532          * just in case.
4533          */
4534         if (refi == 0)
4535                 goto out;
4536
4537         sort(sorted, refi, sizeof(struct refsort), refsort_cmp, NULL);
4538
4539         /*
4540          * the first loop frees everything the leaves point to
4541          */
4542         for (i = 0; i < refi; i++) {
4543                 u64 ptr_gen;
4544
4545                 bytenr = sorted[i].bytenr;
4546
4547                 /*
4548                  * check the reference count on this leaf.  If it is > 1
4549                  * we just decrement it below and don't update any
4550                  * of the refs the leaf points to.
4551                  */
4552                 ret = drop_snap_lookup_refcount(trans, root, bytenr,
4553                                                 blocksize, &refs);
4554                 BUG_ON(ret);
4555                 if (refs != 1)
4556                         continue;
4557
4558                 ptr_gen = btrfs_node_ptr_generation(eb, sorted[i].slot);
4559
4560                 /*
4561                  * the leaf only had one reference, which means the
4562                  * only thing pointing to this leaf is the snapshot
4563                  * we're deleting.  It isn't possible for the reference
4564                  * count to increase again later
4565                  *
4566                  * The reference cache is checked for the leaf,
4567                  * and if found we'll be able to drop any refs held by
4568                  * the leaf without needing to read it in.
4569                  */
4570                 ref = btrfs_lookup_leaf_ref(root, bytenr);
4571                 if (ref && ref->generation != ptr_gen) {
4572                         btrfs_free_leaf_ref(root, ref);
4573                         ref = NULL;
4574                 }
4575                 if (ref) {
4576                         ret = cache_drop_leaf_ref(trans, root, ref);
4577                         BUG_ON(ret);
4578                         btrfs_remove_leaf_ref(root, ref);
4579                         btrfs_free_leaf_ref(root, ref);
4580                 } else {
4581                         /*
4582                          * the leaf wasn't in the reference cache, so
4583                          * we have to read it.
4584                          */
4585                         leaf = read_tree_block(root, bytenr, blocksize,
4586                                                ptr_gen);
4587                         ret = btrfs_drop_leaf_ref(trans, root, leaf);
4588                         BUG_ON(ret);
4589                         free_extent_buffer(leaf);
4590                 }
4591                 atomic_inc(&root->fs_info->throttle_gen);
4592                 wake_up(&root->fs_info->transaction_throttle);
4593                 cond_resched();
4594         }
4595
4596         /*
4597          * run through the loop again to free the refs on the leaves.
4598          * This is faster than doing it in the loop above because
4599          * the leaves are likely to be clustered together.  We end up
4600          * working in nice chunks on the extent allocation tree.
4601          */
4602         for (i = 0; i < refi; i++) {
4603                 bytenr = sorted[i].bytenr;
4604                 ret = btrfs_free_extent(trans, root, bytenr,
4605                                         blocksize, eb->start,
4606                                         root_owner, root_gen, 0, 1);
4607                 BUG_ON(ret);
4608
4609                 atomic_inc(&root->fs_info->throttle_gen);
4610                 wake_up(&root->fs_info->transaction_throttle);
4611                 cond_resched();
4612         }
4613 out:
4614         kfree(sorted);
4615
4616         /*
4617          * update the path to show we've processed the entire level 1
4618          * node.  This will get saved into the root's drop_snapshot_progress
4619          * field so these drops are not repeated again if this transaction
4620          * commits.
4621          */
4622         path->slots[1] = nritems;
4623         return 0;
4624 }
4625
4626 /*
4627  * helper function for drop_snapshot, this walks down the tree dropping ref
4628  * counts as it goes.
4629  */
4630 static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
4631                                    struct btrfs_root *root,
4632                                    struct btrfs_path *path, int *level)
4633 {
4634         u64 root_owner;
4635         u64 root_gen;
4636         u64 bytenr;
4637         u64 ptr_gen;
4638         struct extent_buffer *next;
4639         struct extent_buffer *cur;
4640         struct extent_buffer *parent;
4641         u32 blocksize;
4642         int ret;
4643         u32 refs;
4644
4645         WARN_ON(*level < 0);
4646         WARN_ON(*level >= BTRFS_MAX_LEVEL);
4647         ret = drop_snap_lookup_refcount(trans, root, path->nodes[*level]->start,
4648                                 path->nodes[*level]->len, &refs);
4649         BUG_ON(ret);
4650         if (refs > 1)
4651                 goto out;
4652
4653         /*
4654          * walk down to the last node level and free all the leaves
4655          */
4656         while (*level >= 0) {
4657                 WARN_ON(*level < 0);
4658                 WARN_ON(*level >= BTRFS_MAX_LEVEL);
4659                 cur = path->nodes[*level];
4660
4661                 if (btrfs_header_level(cur) != *level)
4662                         WARN_ON(1);
4663
4664                 if (path->slots[*level] >=
4665                     btrfs_header_nritems(cur))
4666                         break;
4667
4668                 /* the new code goes down to level 1 and does all the
4669                  * leaves pointed to that node in bulk.  So, this check
4670                  * for level 0 will always be false.
4671                  *
4672                  * But, the disk format allows the drop_snapshot_progress
4673                  * field in the root to leave things in a state where
4674                  * a leaf will need cleaning up here.  If someone crashes
4675                  * with the old code and then boots with the new code,
4676                  * we might find a leaf here.
4677                  */
4678                 if (*level == 0) {
4679                         ret = btrfs_drop_leaf_ref(trans, root, cur);
4680                         BUG_ON(ret);
4681                         break;
4682                 }
4683
4684                 /*
4685                  * once we get to level one, process the whole node
4686                  * at once, including everything below it.
4687                  */
4688                 if (*level == 1) {
4689                         ret = drop_level_one_refs(trans, root, path);
4690                         BUG_ON(ret);
4691                         break;
4692                 }
4693
4694                 bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
4695                 ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
4696                 blocksize = btrfs_level_size(root, *level - 1);
4697
4698                 ret = drop_snap_lookup_refcount(trans, root, bytenr,
4699                                                 blocksize, &refs);
4700                 BUG_ON(ret);
4701
4702                 /*
4703                  * if there is more than one reference, we don't need
4704                  * to read that node to drop any references it has.  We
4705                  * just drop the ref we hold on that node and move on to the
4706                  * next slot in this level.
4707                  */
4708                 if (refs != 1) {
4709                         parent = path->nodes[*level];
4710                         root_owner = btrfs_header_owner(parent);
4711                         root_gen = btrfs_header_generation(parent);
4712                         path->slots[*level]++;
4713
4714                         ret = btrfs_free_extent(trans, root, bytenr,
4715                                                 blocksize, parent->start,
4716                                                 root_owner, root_gen,
4717                                                 *level - 1, 1);
4718                         BUG_ON(ret);
4719
4720                         atomic_inc(&root->fs_info->throttle_gen);
4721                         wake_up(&root->fs_info->transaction_throttle);
4722                         cond_resched();
4723
4724                         continue;
4725                 }
4726
4727                 /*
4728                  * we need to keep freeing things in the next level down.
4729                  * read the block and loop around to process it
4730                  */
4731                 next = read_tree_block(root, bytenr, blocksize, ptr_gen);
4732                 WARN_ON(*level <= 0);
4733                 if (path->nodes[*level-1])
4734                         free_extent_buffer(path->nodes[*level-1]);
4735                 path->nodes[*level-1] = next;
4736                 *level = btrfs_header_level(next);
4737                 path->slots[*level] = 0;
4738                 cond_resched();
4739         }
4740 out:
4741         WARN_ON(*level < 0);
4742         WARN_ON(*level >= BTRFS_MAX_LEVEL);
4743
4744         if (path->nodes[*level] == root->node) {
4745                 parent = path->nodes[*level];
4746                 bytenr = path->nodes[*level]->start;
4747         } else {
4748                 parent = path->nodes[*level + 1];
4749                 bytenr = btrfs_node_blockptr(parent, path->slots[*level + 1]);
4750         }
4751
4752         blocksize = btrfs_level_size(root, *level);
4753         root_owner = btrfs_header_owner(parent);
4754         root_gen = btrfs_header_generation(parent);
4755
4756         /*
4757          * cleanup and free the reference on the last node
4758          * we processed
4759          */
4760         ret = btrfs_free_extent(trans, root, bytenr, blocksize,
4761                                   parent->start, root_owner, root_gen,
4762                                   *level, 1);
4763         free_extent_buffer(path->nodes[*level]);
4764         path->nodes[*level] = NULL;
4765
4766         *level += 1;
4767         BUG_ON(ret);
4768
4769         cond_resched();
4770         return 0;
4771 }
4772 #endif
4773
4774 struct walk_control {
4775         u64 refs[BTRFS_MAX_LEVEL];
4776         u64 flags[BTRFS_MAX_LEVEL];
4777         struct btrfs_key update_progress;
4778         int stage;
4779         int level;
4780         int shared_level;
4781         int update_ref;
4782         int keep_locks;
4783 };
4784
4785 #define DROP_REFERENCE  1
4786 #define UPDATE_BACKREF  2
4787
4788 /*
4789  * hepler to process tree block while walking down the tree.
4790  *
4791  * when wc->stage == DROP_REFERENCE, this function checks
4792  * reference count of the block. if the block is shared and
4793  * we need update back refs for the subtree rooted at the
4794  * block, this function changes wc->stage to UPDATE_BACKREF
4795  *
4796  * when wc->stage == UPDATE_BACKREF, this function updates
4797  * back refs for pointers in the block.
4798  *
4799  * NOTE: return value 1 means we should stop walking down.
4800  */
4801 static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
4802                                    struct btrfs_root *root,
4803                                    struct btrfs_path *path,
4804                                    struct walk_control *wc)
4805 {
4806         int level = wc->level;
4807         struct extent_buffer *eb = path->nodes[level];
4808         struct btrfs_key key;
4809         u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF;
4810         int ret;
4811
4812         if (wc->stage == UPDATE_BACKREF &&
4813             btrfs_header_owner(eb) != root->root_key.objectid)
4814                 return 1;
4815
4816         /*
4817          * when reference count of tree block is 1, it won't increase
4818          * again. once full backref flag is set, we never clear it.
4819          */
4820         if ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
4821             (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag))) {
4822                 BUG_ON(!path->locks[level]);
4823                 ret = btrfs_lookup_extent_info(trans, root,
4824                                                eb->start, eb->len,
4825                                                &wc->refs[level],
4826                                                &wc->flags[level]);
4827                 BUG_ON(ret);
4828                 BUG_ON(wc->refs[level] == 0);
4829         }
4830
4831         if (wc->stage == DROP_REFERENCE &&
4832             wc->update_ref && wc->refs[level] > 1) {
4833                 BUG_ON(eb == root->node);
4834                 BUG_ON(path->slots[level] > 0);
4835                 if (level == 0)
4836                         btrfs_item_key_to_cpu(eb, &key, path->slots[level]);
4837                 else
4838                         btrfs_node_key_to_cpu(eb, &key, path->slots[level]);
4839                 if (btrfs_header_owner(eb) == root->root_key.objectid &&
4840                     btrfs_comp_cpu_keys(&key, &wc->update_progress) >= 0) {
4841                         wc->stage = UPDATE_BACKREF;
4842                         wc->shared_level = level;
4843                 }
4844         }
4845
4846         if (wc->stage == DROP_REFERENCE) {
4847                 if (wc->refs[level] > 1)
4848                         return 1;
4849
4850                 if (path->locks[level] && !wc->keep_locks) {
4851                         btrfs_tree_unlock(eb);
4852                         path->locks[level] = 0;
4853                 }
4854                 return 0;
4855         }
4856
4857         /* wc->stage == UPDATE_BACKREF */
4858         if (!(wc->flags[level] & flag)) {
4859                 BUG_ON(!path->locks[level]);
4860                 ret = btrfs_inc_ref(trans, root, eb, 1);
4861                 BUG_ON(ret);
4862                 ret = btrfs_dec_ref(trans, root, eb, 0);
4863                 BUG_ON(ret);
4864                 ret = btrfs_set_disk_extent_flags(trans, root, eb->start,
4865                                                   eb->len, flag, 0);
4866                 BUG_ON(ret);
4867                 wc->flags[level] |= flag;
4868         }
4869
4870         /*
4871          * the block is shared by multiple trees, so it's not good to
4872          * keep the tree lock
4873          */
4874         if (path->locks[level] && level > 0) {
4875                 btrfs_tree_unlock(eb);
4876                 path->locks[level] = 0;
4877         }
4878         return 0;
4879 }
4880
4881 /*
4882  * hepler to process tree block while walking up the tree.
4883  *
4884  * when wc->stage == DROP_REFERENCE, this function drops
4885  * reference count on the block.
4886  *
4887  * when wc->stage == UPDATE_BACKREF, this function changes
4888  * wc->stage back to DROP_REFERENCE if we changed wc->stage
4889  * to UPDATE_BACKREF previously while processing the block.
4890  *
4891  * NOTE: return value 1 means we should stop walking up.
4892  */
4893 static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
4894                                  struct btrfs_root *root,
4895                                  struct btrfs_path *path,
4896                                  struct walk_control *wc)
4897 {
4898         int ret = 0;
4899         int level = wc->level;
4900         struct extent_buffer *eb = path->nodes[level];
4901         u64 parent = 0;
4902
4903         if (wc->stage == UPDATE_BACKREF) {
4904                 BUG_ON(wc->shared_level < level);
4905                 if (level < wc->shared_level)
4906                         goto out;
4907
4908                 BUG_ON(wc->refs[level] <= 1);
4909                 ret = find_next_key(path, level + 1, &wc->update_progress);
4910                 if (ret > 0)
4911                         wc->update_ref = 0;
4912
4913                 wc->stage = DROP_REFERENCE;
4914                 wc->shared_level = -1;
4915                 path->slots[level] = 0;
4916
4917                 /*
4918                  * check reference count again if the block isn't locked.
4919                  * we should start walking down the tree again if reference
4920                  * count is one.
4921                  */
4922                 if (!path->locks[level]) {
4923                         BUG_ON(level == 0);
4924                         btrfs_tree_lock(eb);
4925                         btrfs_set_lock_blocking(eb);
4926                         path->locks[level] = 1;
4927
4928                         ret = btrfs_lookup_extent_info(trans, root,
4929                                                        eb->start, eb->len,
4930                                                        &wc->refs[level],
4931                                                        &wc->flags[level]);
4932                         BUG_ON(ret);
4933                         BUG_ON(wc->refs[level] == 0);
4934                         if (wc->refs[level] == 1) {
4935                                 btrfs_tree_unlock(eb);
4936                                 path->locks[level] = 0;
4937                                 return 1;
4938                         }
4939                 } else {
4940                         BUG_ON(level != 0);
4941                 }
4942         }
4943
4944         /* wc->stage == DROP_REFERENCE */
4945         BUG_ON(wc->refs[level] > 1 && !path->locks[level]);
4946
4947         if (wc->refs[level] == 1) {
4948                 if (level == 0) {
4949                         if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
4950                                 ret = btrfs_dec_ref(trans, root, eb, 1);
4951                         else
4952                                 ret = btrfs_dec_ref(trans, root, eb, 0);
4953                         BUG_ON(ret);
4954                 }
4955                 /* make block locked assertion in clean_tree_block happy */
4956                 if (!path->locks[level] &&
4957                     btrfs_header_generation(eb) == trans->transid) {
4958                         btrfs_tree_lock(eb);
4959                         btrfs_set_lock_blocking(eb);
4960                         path->locks[level] = 1;
4961                 }
4962                 clean_tree_block(trans, root, eb);
4963         }
4964
4965         if (eb == root->node) {
4966                 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
4967                         parent = eb->start;
4968                 else
4969                         BUG_ON(root->root_key.objectid !=
4970                                btrfs_header_owner(eb));
4971         } else {
4972                 if (wc->flags[level + 1] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
4973                         parent = path->nodes[level + 1]->start;
4974                 else
4975                         BUG_ON(root->root_key.objectid !=
4976                                btrfs_header_owner(path->nodes[level + 1]));
4977         }
4978
4979         ret = btrfs_free_extent(trans, root, eb->start, eb->len, parent,
4980                                 root->root_key.objectid, level, 0);
4981         BUG_ON(ret);
4982 out:
4983         wc->refs[level] = 0;
4984         wc->flags[level] = 0;
4985         return ret;
4986 }
4987
4988 static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
4989                                    struct btrfs_root *root,
4990                                    struct btrfs_path *path,
4991                                    struct walk_control *wc)
4992 {
4993         struct extent_buffer *next;
4994         struct extent_buffer *cur;
4995         u64 bytenr;
4996         u64 ptr_gen;
4997         u32 blocksize;
4998         int level = wc->level;
4999         int ret;
5000
5001         while (level >= 0) {
5002                 cur = path->nodes[level];
5003                 BUG_ON(path->slots[level] >= btrfs_header_nritems(cur));
5004
5005                 ret = walk_down_proc(trans, root, path, wc);
5006                 if (ret > 0)
5007                         break;
5008
5009                 if (level == 0)
5010                         break;
5011
5012                 bytenr = btrfs_node_blockptr(cur, path->slots[level]);
5013                 blocksize = btrfs_level_size(root, level - 1);
5014                 ptr_gen = btrfs_node_ptr_generation(cur, path->slots[level]);
5015
5016                 next = read_tree_block(root, bytenr, blocksize, ptr_gen);
5017                 btrfs_tree_lock(next);
5018                 btrfs_set_lock_blocking(next);
5019
5020                 level--;
5021                 BUG_ON(level != btrfs_header_level(next));
5022                 path->nodes[level] = next;
5023                 path->slots[level] = 0;
5024                 path->locks[level] = 1;
5025                 wc->level = level;
5026         }
5027         return 0;
5028 }
5029
5030 static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
5031                                  struct btrfs_root *root,
5032                                  struct btrfs_path *path,
5033                                  struct walk_control *wc, int max_level)
5034 {
5035         int level = wc->level;
5036         int ret;
5037
5038         path->slots[level] = btrfs_header_nritems(path->nodes[level]);
5039         while (level < max_level && path->nodes[level]) {
5040                 wc->level = level;
5041                 if (path->slots[level] + 1 <
5042                     btrfs_header_nritems(path->nodes[level])) {
5043                         path->slots[level]++;
5044                         return 0;
5045                 } else {
5046                         ret = walk_up_proc(trans, root, path, wc);
5047                         if (ret > 0)
5048                                 return 0;
5049
5050                         if (path->locks[level]) {
5051                                 btrfs_tree_unlock(path->nodes[level]);
5052                                 path->locks[level] = 0;
5053                         }
5054                         free_extent_buffer(path->nodes[level]);
5055                         path->nodes[level] = NULL;
5056                         level++;
5057                 }
5058         }
5059         return 1;
5060 }
5061
5062 /*
5063  * drop a subvolume tree.
5064  *
5065  * this function traverses the tree freeing any blocks that only
5066  * referenced by the tree.
5067  *
5068  * when a shared tree block is found. this function decreases its
5069  * reference count by one. if update_ref is true, this function
5070  * also make sure backrefs for the shared block and all lower level
5071  * blocks are properly updated.
5072  */
5073 int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref)
5074 {
5075         struct btrfs_path *path;
5076         struct btrfs_trans_handle *trans;
5077         struct btrfs_root *tree_root = root->fs_info->tree_root;
5078         struct btrfs_root_item *root_item = &root->root_item;
5079         struct walk_control *wc;
5080         struct btrfs_key key;
5081         int err = 0;
5082         int ret;
5083         int level;
5084
5085         path = btrfs_alloc_path();
5086         BUG_ON(!path);
5087
5088         wc = kzalloc(sizeof(*wc), GFP_NOFS);
5089         BUG_ON(!wc);
5090
5091         trans = btrfs_start_transaction(tree_root, 1);
5092
5093         if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
5094                 level = btrfs_header_level(root->node);
5095                 path->nodes[level] = btrfs_lock_root_node(root);
5096                 btrfs_set_lock_blocking(path->nodes[level]);
5097                 path->slots[level] = 0;
5098                 path->locks[level] = 1;
5099                 memset(&wc->update_progress, 0,
5100                        sizeof(wc->update_progress));
5101         } else {
5102                 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
5103                 memcpy(&wc->update_progress, &key,
5104                        sizeof(wc->update_progress));
5105
5106                 level = root_item->drop_level;
5107                 BUG_ON(level == 0);
5108                 path->lowest_level = level;
5109                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5110                 path->lowest_level = 0;
5111                 if (ret < 0) {
5112                         err = ret;
5113                         goto out;
5114                 }
5115                 btrfs_node_key_to_cpu(path->nodes[level], &key,
5116                                       path->slots[level]);
5117                 WARN_ON(memcmp(&key, &wc->update_progress, sizeof(key)));
5118
5119                 /*
5120                  * unlock our path, this is safe because only this
5121                  * function is allowed to delete this snapshot
5122                  */
5123                 btrfs_unlock_up_safe(path, 0);
5124
5125                 level = btrfs_header_level(root->node);
5126                 while (1) {
5127                         btrfs_tree_lock(path->nodes[level]);
5128                         btrfs_set_lock_blocking(path->nodes[level]);
5129
5130                         ret = btrfs_lookup_extent_info(trans, root,
5131                                                 path->nodes[level]->start,
5132                                                 path->nodes[level]->len,
5133                                                 &wc->refs[level],
5134                                                 &wc->flags[level]);
5135                         BUG_ON(ret);
5136                         BUG_ON(wc->refs[level] == 0);
5137
5138                         if (level == root_item->drop_level)
5139                                 break;
5140
5141                         btrfs_tree_unlock(path->nodes[level]);
5142                         WARN_ON(wc->refs[level] != 1);
5143                         level--;
5144                 }
5145         }
5146
5147         wc->level = level;
5148         wc->shared_level = -1;
5149         wc->stage = DROP_REFERENCE;
5150         wc->update_ref = update_ref;
5151         wc->keep_locks = 0;
5152
5153         while (1) {
5154                 ret = walk_down_tree(trans, root, path, wc);
5155                 if (ret < 0) {
5156                         err = ret;
5157                         break;
5158                 }
5159
5160                 ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL);
5161                 if (ret < 0) {
5162                         err = ret;
5163                         break;
5164                 }
5165
5166                 if (ret > 0) {
5167                         BUG_ON(wc->stage != DROP_REFERENCE);
5168                         break;
5169                 }
5170
5171                 if (wc->stage == DROP_REFERENCE) {
5172                         level = wc->level;
5173                         btrfs_node_key(path->nodes[level],
5174                                        &root_item->drop_progress,
5175                                        path->slots[level]);
5176                         root_item->drop_level = level;
5177                 }
5178
5179                 BUG_ON(wc->level == 0);
5180                 if (trans->transaction->in_commit ||
5181                     trans->transaction->delayed_refs.flushing) {
5182                         ret = btrfs_update_root(trans, tree_root,
5183                                                 &root->root_key,
5184                                                 root_item);
5185                         BUG_ON(ret);
5186
5187                         btrfs_end_transaction(trans, tree_root);
5188                         trans = btrfs_start_transaction(tree_root, 1);
5189                 } else {
5190                         unsigned long update;
5191                         update = trans->delayed_ref_updates;
5192                         trans->delayed_ref_updates = 0;
5193                         if (update)
5194                                 btrfs_run_delayed_refs(trans, tree_root,
5195                                                        update);
5196                 }
5197         }
5198         btrfs_release_path(root, path);
5199         BUG_ON(err);
5200
5201         ret = btrfs_del_root(trans, tree_root, &root->root_key);
5202         BUG_ON(ret);
5203
5204         free_extent_buffer(root->node);
5205         free_extent_buffer(root->commit_root);
5206         kfree(root);
5207 out:
5208         btrfs_end_transaction(trans, tree_root);
5209         kfree(wc);
5210         btrfs_free_path(path);
5211         return err;
5212 }
5213
5214 /*
5215  * drop subtree rooted at tree block 'node'.
5216  *
5217  * NOTE: this function will unlock and release tree block 'node'
5218  */
5219 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
5220                         struct btrfs_root *root,
5221                         struct extent_buffer *node,
5222                         struct extent_buffer *parent)
5223 {
5224         struct btrfs_path *path;
5225         struct walk_control *wc;
5226         int level;
5227         int parent_level;
5228         int ret = 0;
5229         int wret;
5230
5231         BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
5232
5233         path = btrfs_alloc_path();
5234         BUG_ON(!path);
5235
5236         wc = kzalloc(sizeof(*wc), GFP_NOFS);
5237         BUG_ON(!wc);
5238
5239         btrfs_assert_tree_locked(parent);
5240         parent_level = btrfs_header_level(parent);
5241         extent_buffer_get(parent);
5242         path->nodes[parent_level] = parent;
5243         path->slots[parent_level] = btrfs_header_nritems(parent);
5244
5245         btrfs_assert_tree_locked(node);
5246         level = btrfs_header_level(node);
5247         path->nodes[level] = node;
5248         path->slots[level] = 0;
5249         path->locks[level] = 1;
5250
5251         wc->refs[parent_level] = 1;
5252         wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF;
5253         wc->level = level;
5254         wc->shared_level = -1;
5255         wc->stage = DROP_REFERENCE;
5256         wc->update_ref = 0;
5257         wc->keep_locks = 1;
5258
5259         while (1) {
5260                 wret = walk_down_tree(trans, root, path, wc);
5261                 if (wret < 0) {
5262                         ret = wret;
5263                         break;
5264                 }
5265
5266                 wret = walk_up_tree(trans, root, path, wc, parent_level);
5267                 if (wret < 0)
5268                         ret = wret;
5269                 if (wret != 0)
5270                         break;
5271         }
5272
5273         kfree(wc);
5274         btrfs_free_path(path);
5275         return ret;
5276 }
5277
5278 #if 0
5279 static unsigned long calc_ra(unsigned long start, unsigned long last,
5280                              unsigned long nr)
5281 {
5282         return min(last, start + nr - 1);
5283 }
5284
5285 static noinline int relocate_inode_pages(struct inode *inode, u64 start,
5286                                          u64 len)
5287 {
5288         u64 page_start;
5289         u64 page_end;
5290         unsigned long first_index;
5291         unsigned long last_index;
5292         unsigned long i;
5293         struct page *page;
5294         struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
5295         struct file_ra_state *ra;
5296         struct btrfs_ordered_extent *ordered;
5297         unsigned int total_read = 0;
5298         unsigned int total_dirty = 0;
5299         int ret = 0;
5300
5301         ra = kzalloc(sizeof(*ra), GFP_NOFS);
5302
5303         mutex_lock(&inode->i_mutex);
5304         first_index = start >> PAGE_CACHE_SHIFT;
5305         last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
5306
5307         /* make sure the dirty trick played by the caller work */
5308         ret = invalidate_inode_pages2_range(inode->i_mapping,
5309                                             first_index, last_index);
5310         if (ret)
5311                 goto out_unlock;
5312
5313         file_ra_state_init(ra, inode->i_mapping);
5314
5315         for (i = first_index ; i <= last_index; i++) {
5316                 if (total_read % ra->ra_pages == 0) {
5317                         btrfs_force_ra(inode->i_mapping, ra, NULL, i,
5318                                        calc_ra(i, last_index, ra->ra_pages));
5319                 }
5320                 total_read++;
5321 again:
5322                 if (((u64)i << PAGE_CACHE_SHIFT) > i_size_read(inode))
5323                         BUG_ON(1);
5324                 page = grab_cache_page(inode->i_mapping, i);
5325                 if (!page) {
5326                         ret = -ENOMEM;
5327                         goto out_unlock;
5328                 }
5329                 if (!PageUptodate(page)) {
5330                         btrfs_readpage(NULL, page);
5331                         lock_page(page);
5332                         if (!PageUptodate(page)) {
5333                                 unlock_page(page);
5334                                 page_cache_release(page);
5335                                 ret = -EIO;
5336                                 goto out_unlock;
5337                         }
5338                 }
5339                 wait_on_page_writeback(page);
5340
5341                 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
5342                 page_end = page_start + PAGE_CACHE_SIZE - 1;
5343                 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
5344
5345                 ordered = btrfs_lookup_ordered_extent(inode, page_start);
5346                 if (ordered) {
5347                         unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
5348                         unlock_page(page);
5349                         page_cache_release(page);
5350                         btrfs_start_ordered_extent(inode, ordered, 1);
5351                         btrfs_put_ordered_extent(ordered);
5352                         goto again;
5353                 }
5354                 set_page_extent_mapped(page);
5355
5356                 if (i == first_index)
5357                         set_extent_bits(io_tree, page_start, page_end,
5358                                         EXTENT_BOUNDARY, GFP_NOFS);
5359                 btrfs_set_extent_delalloc(inode, page_start, page_end);
5360
5361                 set_page_dirty(page);
5362                 total_dirty++;
5363
5364                 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
5365                 unlock_page(page);
5366                 page_cache_release(page);
5367         }
5368
5369 out_unlock:
5370         kfree(ra);
5371         mutex_unlock(&inode->i_mutex);
5372         balance_dirty_pages_ratelimited_nr(inode->i_mapping, total_dirty);
5373         return ret;
5374 }
5375
5376 static noinline int relocate_data_extent(struct inode *reloc_inode,
5377                                          struct btrfs_key *extent_key,
5378                                          u64 offset)
5379 {
5380         struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
5381         struct extent_map_tree *em_tree = &BTRFS_I(reloc_inode)->extent_tree;
5382         struct extent_map *em;
5383         u64 start = extent_key->objectid - offset;
5384         u64 end = start + extent_key->offset - 1;
5385
5386         em = alloc_extent_map(GFP_NOFS);
5387         BUG_ON(!em || IS_ERR(em));
5388
5389         em->start = start;
5390         em->len = extent_key->offset;
5391         em->block_len = extent_key->offset;
5392         em->block_start = extent_key->objectid;
5393         em->bdev = root->fs_info->fs_devices->latest_bdev;
5394         set_bit(EXTENT_FLAG_PINNED, &em->flags);
5395
5396         /* setup extent map to cheat btrfs_readpage */
5397         lock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
5398         while (1) {
5399                 int ret;
5400                 spin_lock(&em_tree->lock);
5401                 ret = add_extent_mapping(em_tree, em);
5402                 spin_unlock(&em_tree->lock);
5403                 if (ret != -EEXIST) {
5404                         free_extent_map(em);
5405                         break;
5406                 }
5407                 btrfs_drop_extent_cache(reloc_inode, start, end, 0);
5408         }
5409         unlock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
5410
5411         return relocate_inode_pages(reloc_inode, start, extent_key->offset);
5412 }
5413
5414 struct btrfs_ref_path {
5415         u64 extent_start;
5416         u64 nodes[BTRFS_MAX_LEVEL];
5417         u64 root_objectid;
5418         u64 root_generation;
5419         u64 owner_objectid;
5420         u32 num_refs;
5421         int lowest_level;
5422         int current_level;
5423         int shared_level;
5424
5425         struct btrfs_key node_keys[BTRFS_MAX_LEVEL];
5426         u64 new_nodes[BTRFS_MAX_LEVEL];
5427 };
5428
5429 struct disk_extent {
5430         u64 ram_bytes;
5431         u64 disk_bytenr;
5432         u64 disk_num_bytes;
5433         u64 offset;
5434         u64 num_bytes;
5435         u8 compression;
5436         u8 encryption;
5437         u16 other_encoding;
5438 };
5439
5440 static int is_cowonly_root(u64 root_objectid)
5441 {
5442         if (root_objectid == BTRFS_ROOT_TREE_OBJECTID ||
5443             root_objectid == BTRFS_EXTENT_TREE_OBJECTID ||
5444             root_objectid == BTRFS_CHUNK_TREE_OBJECTID ||
5445             root_objectid == BTRFS_DEV_TREE_OBJECTID ||
5446             root_objectid == BTRFS_TREE_LOG_OBJECTID ||
5447             root_objectid == BTRFS_CSUM_TREE_OBJECTID)
5448                 return 1;
5449         return 0;
5450 }
5451
5452 static noinline int __next_ref_path(struct btrfs_trans_handle *trans,
5453                                     struct btrfs_root *extent_root,
5454                                     struct btrfs_ref_path *ref_path,
5455                                     int first_time)
5456 {
5457         struct extent_buffer *leaf;
5458         struct btrfs_path *path;
5459         struct btrfs_extent_ref *ref;
5460         struct btrfs_key key;
5461         struct btrfs_key found_key;
5462         u64 bytenr;
5463         u32 nritems;
5464         int level;
5465         int ret = 1;
5466
5467         path = btrfs_alloc_path();
5468         if (!path)
5469                 return -ENOMEM;
5470
5471         if (first_time) {
5472                 ref_path->lowest_level = -1;
5473                 ref_path->current_level = -1;
5474                 ref_path->shared_level = -1;
5475                 goto walk_up;
5476         }
5477 walk_down:
5478         level = ref_path->current_level - 1;
5479         while (level >= -1) {
5480                 u64 parent;
5481                 if (level < ref_path->lowest_level)
5482                         break;
5483
5484                 if (level >= 0)
5485                         bytenr = ref_path->nodes[level];
5486                 else
5487                         bytenr = ref_path->extent_start;
5488                 BUG_ON(bytenr == 0);
5489
5490                 parent = ref_path->nodes[level + 1];
5491                 ref_path->nodes[level + 1] = 0;
5492                 ref_path->current_level = level;
5493                 BUG_ON(parent == 0);
5494
5495                 key.objectid = bytenr;
5496                 key.offset = parent + 1;
5497                 key.type = BTRFS_EXTENT_REF_KEY;
5498
5499                 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
5500                 if (ret < 0)
5501                         goto out;
5502                 BUG_ON(ret == 0);
5503
5504                 leaf = path->nodes[0];
5505                 nritems = btrfs_header_nritems(leaf);
5506                 if (path->slots[0] >= nritems) {
5507                         ret = btrfs_next_leaf(extent_root, path);
5508                         if (ret < 0)
5509                                 goto out;
5510                         if (ret > 0)
5511                                 goto next;
5512                         leaf = path->nodes[0];
5513                 }
5514
5515                 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5516                 if (found_key.objectid == bytenr &&
5517                     found_key.type == BTRFS_EXTENT_REF_KEY) {
5518                         if (level < ref_path->shared_level)
5519                                 ref_path->shared_level = level;
5520                         goto found;
5521                 }
5522 next:
5523                 level--;
5524                 btrfs_release_path(extent_root, path);
5525                 cond_resched();
5526         }
5527         /* reached lowest level */
5528         ret = 1;
5529         goto out;
5530 walk_up:
5531         level = ref_path->current_level;
5532         while (level < BTRFS_MAX_LEVEL - 1) {
5533                 u64 ref_objectid;
5534
5535                 if (level >= 0)
5536                         bytenr = ref_path->nodes[level];
5537                 else
5538                         bytenr = ref_path->extent_start;
5539
5540                 BUG_ON(bytenr == 0);
5541
5542                 key.objectid = bytenr;
5543                 key.offset = 0;
5544                 key.type = BTRFS_EXTENT_REF_KEY;
5545
5546                 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
5547                 if (ret < 0)
5548                         goto out;
5549
5550                 leaf = path->nodes[0];
5551                 nritems = btrfs_header_nritems(leaf);
5552                 if (path->slots[0] >= nritems) {
5553                         ret = btrfs_next_leaf(extent_root, path);
5554                         if (ret < 0)
5555                                 goto out;
5556                         if (ret > 0) {
5557                                 /* the extent was freed by someone */
5558                                 if (ref_path->lowest_level == level)
5559                                         goto out;
5560                                 btrfs_release_path(extent_root, path);
5561                                 goto walk_down;
5562                         }
5563                         leaf = path->nodes[0];
5564                 }
5565
5566                 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5567                 if (found_key.objectid != bytenr ||
5568                                 found_key.type != BTRFS_EXTENT_REF_KEY) {
5569                         /* the extent was freed by someone */
5570                         if (ref_path->lowest_level == level) {
5571                                 ret = 1;
5572                                 goto out;
5573                         }
5574                         btrfs_release_path(extent_root, path);
5575                         goto walk_down;
5576                 }
5577 found:
5578                 ref = btrfs_item_ptr(leaf, path->slots[0],
5579                                 struct btrfs_extent_ref);
5580                 ref_objectid = btrfs_ref_objectid(leaf, ref);
5581                 if (ref_objectid < BTRFS_FIRST_FREE_OBJECTID) {
5582                         if (first_time) {
5583                                 level = (int)ref_objectid;
5584                                 BUG_ON(level >= BTRFS_MAX_LEVEL);
5585                                 ref_path->lowest_level = level;
5586                                 ref_path->current_level = level;
5587                                 ref_path->nodes[level] = bytenr;
5588                         } else {
5589                                 WARN_ON(ref_objectid != level);
5590                         }
5591                 } else {
5592                         WARN_ON(level != -1);
5593                 }
5594                 first_time = 0;
5595
5596                 if (ref_path->lowest_level == level) {
5597                         ref_path->owner_objectid = ref_objectid;
5598                         ref_path->num_refs = btrfs_ref_num_refs(leaf, ref);
5599                 }
5600
5601                 /*
5602                  * the block is tree root or the block isn't in reference
5603                  * counted tree.
5604                  */
5605                 if (found_key.objectid == found_key.offset ||
5606                     is_cowonly_root(btrfs_ref_root(leaf, ref))) {
5607                         ref_path->root_objectid = btrfs_ref_root(leaf, ref);
5608                         ref_path->root_generation =
5609                                 btrfs_ref_generation(leaf, ref);
5610                         if (level < 0) {
5611                                 /* special reference from the tree log */
5612                                 ref_path->nodes[0] = found_key.offset;
5613                                 ref_path->current_level = 0;
5614                         }
5615                         ret = 0;
5616                         goto out;
5617                 }
5618
5619                 level++;
5620                 BUG_ON(ref_path->nodes[level] != 0);
5621                 ref_path->nodes[level] = found_key.offset;
5622                 ref_path->current_level = level;
5623
5624                 /*
5625                  * the reference was created in the running transaction,
5626                  * no need to continue walking up.
5627                  */
5628                 if (btrfs_ref_generation(leaf, ref) == trans->transid) {
5629                         ref_path->root_objectid = btrfs_ref_root(leaf, ref);
5630                         ref_path->root_generation =
5631                                 btrfs_ref_generation(leaf, ref);
5632                         ret = 0;
5633                         goto out;
5634                 }
5635
5636                 btrfs_release_path(extent_root, path);
5637                 cond_resched();
5638         }
5639         /* reached max tree level, but no tree root found. */
5640         BUG();
5641 out:
5642         btrfs_free_path(path);
5643         return ret;
5644 }
5645
5646 static int btrfs_first_ref_path(struct btrfs_trans_handle *trans,
5647                                 struct btrfs_root *extent_root,
5648                                 struct btrfs_ref_path *ref_path,
5649                                 u64 extent_start)
5650 {
5651         memset(ref_path, 0, sizeof(*ref_path));
5652         ref_path->extent_start = extent_start;
5653
5654         return __next_ref_path(trans, extent_root, ref_path, 1);
5655 }
5656
5657 static int btrfs_next_ref_path(struct btrfs_trans_handle *trans,
5658                                struct btrfs_root *extent_root,
5659                                struct btrfs_ref_path *ref_path)
5660 {
5661         return __next_ref_path(trans, extent_root, ref_path, 0);
5662 }
5663
5664 static noinline int get_new_locations(struct inode *reloc_inode,
5665                                       struct btrfs_key *extent_key,
5666                                       u64 offset, int no_fragment,
5667                                       struct disk_extent **extents,
5668                                       int *nr_extents)
5669 {
5670         struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
5671         struct btrfs_path *path;
5672         struct btrfs_file_extent_item *fi;
5673         struct extent_buffer *leaf;
5674         struct disk_extent *exts = *extents;
5675         struct btrfs_key found_key;
5676         u64 cur_pos;
5677         u64 last_byte;
5678         u32 nritems;
5679         int nr = 0;
5680         int max = *nr_extents;
5681         int ret;
5682
5683         WARN_ON(!no_fragment && *extents);
5684         if (!exts) {
5685                 max = 1;
5686                 exts = kmalloc(sizeof(*exts) * max, GFP_NOFS);
5687                 if (!exts)
5688                         return -ENOMEM;
5689         }
5690
5691         path = btrfs_alloc_path();
5692         BUG_ON(!path);
5693
5694         cur_pos = extent_key->objectid - offset;
5695         last_byte = extent_key->objectid + extent_key->offset;
5696         ret = btrfs_lookup_file_extent(NULL, root, path, reloc_inode->i_ino,
5697                                        cur_pos, 0);
5698         if (ret < 0)
5699                 goto out;
5700         if (ret > 0) {
5701                 ret = -ENOENT;
5702                 goto out;
5703         }
5704
5705         while (1) {
5706                 leaf = path->nodes[0];
5707                 nritems = btrfs_header_nritems(leaf);
5708                 if (path->slots[0] >= nritems) {
5709                         ret = btrfs_next_leaf(root, path);
5710                         if (ret < 0)
5711                                 goto out;
5712                         if (ret > 0)
5713                                 break;
5714                         leaf = path->nodes[0];
5715                 }
5716
5717                 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5718                 if (found_key.offset != cur_pos ||
5719                     found_key.type != BTRFS_EXTENT_DATA_KEY ||
5720                     found_key.objectid != reloc_inode->i_ino)
5721                         break;
5722
5723                 fi = btrfs_item_ptr(leaf, path->slots[0],
5724                                     struct btrfs_file_extent_item);
5725                 if (btrfs_file_extent_type(leaf, fi) !=
5726                     BTRFS_FILE_EXTENT_REG ||
5727                     btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
5728                         break;
5729
5730                 if (nr == max) {
5731                         struct disk_extent *old = exts;
5732                         max *= 2;
5733                         exts = kzalloc(sizeof(*exts) * max, GFP_NOFS);
5734                         memcpy(exts, old, sizeof(*exts) * nr);
5735                         if (old != *extents)
5736                                 kfree(old);
5737                 }
5738
5739                 exts[nr].disk_bytenr =
5740                         btrfs_file_extent_disk_bytenr(leaf, fi);
5741                 exts[nr].disk_num_bytes =
5742                         btrfs_file_extent_disk_num_bytes(leaf, fi);
5743                 exts[nr].offset = btrfs_file_extent_offset(leaf, fi);
5744                 exts[nr].num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
5745                 exts[nr].ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
5746                 exts[nr].compression = btrfs_file_extent_compression(leaf, fi);
5747                 exts[nr].encryption = btrfs_file_extent_encryption(leaf, fi);
5748                 exts[nr].other_encoding = btrfs_file_extent_other_encoding(leaf,
5749                                                                            fi);
5750                 BUG_ON(exts[nr].offset > 0);
5751                 BUG_ON(exts[nr].compression || exts[nr].encryption);
5752                 BUG_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes);
5753
5754                 cur_pos += exts[nr].num_bytes;
5755                 nr++;
5756
5757                 if (cur_pos + offset >= last_byte)
5758                         break;
5759
5760                 if (no_fragment) {
5761                         ret = 1;
5762                         goto out;
5763                 }
5764                 path->slots[0]++;
5765         }
5766
5767         BUG_ON(cur_pos + offset > last_byte);
5768         if (cur_pos + offset < last_byte) {
5769                 ret = -ENOENT;
5770                 goto out;
5771         }
5772         ret = 0;
5773 out:
5774         btrfs_free_path(path);
5775         if (ret) {
5776                 if (exts != *extents)
5777                         kfree(exts);
5778         } else {
5779                 *extents = exts;
5780                 *nr_extents = nr;
5781         }
5782         return ret;
5783 }
5784
5785 static noinline int replace_one_extent(struct btrfs_trans_handle *trans,
5786                                         struct btrfs_root *root,
5787                                         struct btrfs_path *path,
5788                                         struct btrfs_key *extent_key,
5789                                         struct btrfs_key *leaf_key,
5790                                         struct btrfs_ref_path *ref_path,
5791                                         struct disk_extent *new_extents,
5792                                         int nr_extents)
5793 {
5794         struct extent_buffer *leaf;
5795         struct btrfs_file_extent_item *fi;
5796         struct inode *inode = NULL;
5797         struct btrfs_key key;
5798         u64 lock_start = 0;
5799         u64 lock_end = 0;
5800         u64 num_bytes;
5801         u64 ext_offset;
5802         u64 search_end = (u64)-1;
5803         u32 nritems;
5804         int nr_scaned = 0;
5805         int extent_locked = 0;
5806         int extent_type;
5807         int ret;
5808
5809         memcpy(&key, leaf_key, sizeof(key));
5810         if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
5811                 if (key.objectid < ref_path->owner_objectid ||
5812                     (key.objectid == ref_path->owner_objectid &&
5813                      key.type < BTRFS_EXTENT_DATA_KEY)) {
5814                         key.objectid = ref_path->owner_objectid;
5815                         key.type = BTRFS_EXTENT_DATA_KEY;
5816                         key.offset = 0;
5817                 }
5818         }
5819
5820         while (1) {
5821                 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
5822                 if (ret < 0)
5823                         goto out;
5824
5825                 leaf = path->nodes[0];
5826                 nritems = btrfs_header_nritems(leaf);
5827 next:
5828                 if (extent_locked && ret > 0) {
5829                         /*
5830                          * the file extent item was modified by someone
5831                          * before the extent got locked.
5832                          */
5833                         unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
5834                                       lock_end, GFP_NOFS);
5835                         extent_locked = 0;
5836                 }
5837
5838                 if (path->slots[0] >= nritems) {
5839                         if (++nr_scaned > 2)
5840                                 break;
5841
5842                         BUG_ON(extent_locked);
5843                         ret = btrfs_next_leaf(root, path);
5844                         if (ret < 0)
5845                                 goto out;
5846                         if (ret > 0)
5847                                 break;
5848                         leaf = path->nodes[0];
5849                         nritems = btrfs_header_nritems(leaf);
5850                 }
5851
5852                 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
5853
5854                 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
5855                         if ((key.objectid > ref_path->owner_objectid) ||
5856                             (key.objectid == ref_path->owner_objectid &&
5857                              key.type > BTRFS_EXTENT_DATA_KEY) ||
5858                             key.offset >= search_end)
5859                                 break;
5860                 }
5861
5862                 if (inode && key.objectid != inode->i_ino) {
5863                         BUG_ON(extent_locked);
5864                         btrfs_release_path(root, path);
5865                         mutex_unlock(&inode->i_mutex);
5866                         iput(inode);
5867                         inode = NULL;
5868                         continue;
5869                 }
5870
5871                 if (key.type != BTRFS_EXTENT_DATA_KEY) {
5872                         path->slots[0]++;
5873                         ret = 1;
5874                         goto next;
5875                 }
5876                 fi = btrfs_item_ptr(leaf, path->slots[0],
5877                                     struct btrfs_file_extent_item);
5878                 extent_type = btrfs_file_extent_type(leaf, fi);
5879                 if ((extent_type != BTRFS_FILE_EXTENT_REG &&
5880                      extent_type != BTRFS_FILE_EXTENT_PREALLOC) ||
5881                     (btrfs_file_extent_disk_bytenr(leaf, fi) !=
5882                      extent_key->objectid)) {
5883                         path->slots[0]++;
5884                         ret = 1;
5885                         goto next;
5886                 }
5887
5888                 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
5889                 ext_offset = btrfs_file_extent_offset(leaf, fi);
5890
5891                 if (search_end == (u64)-1) {
5892                         search_end = key.offset - ext_offset +
5893                                 btrfs_file_extent_ram_bytes(leaf, fi);
5894                 }
5895
5896                 if (!extent_locked) {
5897                         lock_start = key.offset;
5898                         lock_end = lock_start + num_bytes - 1;
5899                 } else {
5900                         if (lock_start > key.offset ||
5901                             lock_end + 1 < key.offset + num_bytes) {
5902                                 unlock_extent(&BTRFS_I(inode)->io_tree,
5903                                               lock_start, lock_end, GFP_NOFS);
5904                                 extent_locked = 0;
5905                         }
5906                 }
5907
5908                 if (!inode) {
5909                         btrfs_release_path(root, path);
5910
5911                         inode = btrfs_iget_locked(root->fs_info->sb,
5912                                                   key.objectid, root);
5913                         if (inode->i_state & I_NEW) {
5914                                 BTRFS_I(inode)->root = root;
5915                                 BTRFS_I(inode)->location.objectid =
5916                                         key.objectid;
5917                                 BTRFS_I(inode)->location.type =
5918                                         BTRFS_INODE_ITEM_KEY;
5919                                 BTRFS_I(inode)->location.offset = 0;
5920                                 btrfs_read_locked_inode(inode);
5921                                 unlock_new_inode(inode);
5922                         }
5923                         /*
5924                          * some code call btrfs_commit_transaction while
5925                          * holding the i_mutex, so we can't use mutex_lock
5926                          * here.
5927                          */
5928                         if (is_bad_inode(inode) ||
5929                             !mutex_trylock(&inode->i_mutex)) {
5930                                 iput(inode);
5931                                 inode = NULL;
5932                                 key.offset = (u64)-1;
5933                                 goto skip;
5934                         }
5935                 }
5936
5937                 if (!extent_locked) {
5938                         struct btrfs_ordered_extent *ordered;
5939
5940                         btrfs_release_path(root, path);
5941
5942                         lock_extent(&BTRFS_I(inode)->io_tree, lock_start,
5943                                     lock_end, GFP_NOFS);
5944                         ordered = btrfs_lookup_first_ordered_extent(inode,
5945                                                                     lock_end);
5946                         if (ordered &&
5947                             ordered->file_offset <= lock_end &&
5948                             ordered->file_offset + ordered->len > lock_start) {
5949                                 unlock_extent(&BTRFS_I(inode)->io_tree,
5950                                               lock_start, lock_end, GFP_NOFS);
5951                                 btrfs_start_ordered_extent(inode, ordered, 1);
5952                                 btrfs_put_ordered_extent(ordered);
5953                                 key.offset += num_bytes;
5954                                 goto skip;
5955                         }
5956                         if (ordered)
5957                                 btrfs_put_ordered_extent(ordered);
5958
5959                         extent_locked = 1;
5960                         continue;
5961                 }
5962
5963                 if (nr_extents == 1) {
5964                         /* update extent pointer in place */
5965                         btrfs_set_file_extent_disk_bytenr(leaf, fi,
5966                                                 new_extents[0].disk_bytenr);
5967                         btrfs_set_file_extent_disk_num_bytes(leaf, fi,
5968                                                 new_extents[0].disk_num_bytes);
5969                         btrfs_mark_buffer_dirty(leaf);
5970
5971                         btrfs_drop_extent_cache(inode, key.offset,
5972                                                 key.offset + num_bytes - 1, 0);
5973
5974                         ret = btrfs_inc_extent_ref(trans, root,
5975                                                 new_extents[0].disk_bytenr,
5976                                                 new_extents[0].disk_num_bytes,
5977                                                 leaf->start,
5978                                                 root->root_key.objectid,
5979                                                 trans->transid,
5980                                                 key.objectid);
5981                         BUG_ON(ret);
5982
5983                         ret = btrfs_free_extent(trans, root,
5984                                                 extent_key->objectid,
5985                                                 extent_key->offset,
5986                                                 leaf->start,
5987                                                 btrfs_header_owner(leaf),
5988                                                 btrfs_header_generation(leaf),
5989                                                 key.objectid, 0);
5990                         BUG_ON(ret);
5991
5992                         btrfs_release_path(root, path);
5993                         key.offset += num_bytes;
5994                 } else {
5995                         BUG_ON(1);
5996 #if 0
5997                         u64 alloc_hint;
5998                         u64 extent_len;
5999                         int i;
6000                         /*
6001                          * drop old extent pointer at first, then insert the
6002                          * new pointers one bye one
6003                          */
6004                         btrfs_release_path(root, path);
6005                         ret = btrfs_drop_extents(trans, root, inode, key.offset,
6006                                                  key.offset + num_bytes,
6007                                                  key.offset, &alloc_hint);
6008                         BUG_ON(ret);
6009
6010                         for (i = 0; i < nr_extents; i++) {
6011                                 if (ext_offset >= new_extents[i].num_bytes) {
6012                                         ext_offset -= new_extents[i].num_bytes;
6013                                         continue;
6014                                 }
6015                                 extent_len = min(new_extents[i].num_bytes -
6016                                                  ext_offset, num_bytes);
6017
6018                                 ret = btrfs_insert_empty_item(trans, root,
6019                                                               path, &key,
6020                                                               sizeof(*fi));
6021                                 BUG_ON(ret);
6022
6023                                 leaf = path->nodes[0];
6024                                 fi = btrfs_item_ptr(leaf, path->slots[0],
6025                                                 struct btrfs_file_extent_item);
6026                                 btrfs_set_file_extent_generation(leaf, fi,
6027                                                         trans->transid);
6028                                 btrfs_set_file_extent_type(leaf, fi,
6029                                                         BTRFS_FILE_EXTENT_REG);
6030                                 btrfs_set_file_extent_disk_bytenr(leaf, fi,
6031                                                 new_extents[i].disk_bytenr);
6032                                 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
6033                                                 new_extents[i].disk_num_bytes);
6034                                 btrfs_set_file_extent_ram_bytes(leaf, fi,
6035                                                 new_extents[i].ram_bytes);
6036
6037                                 btrfs_set_file_extent_compression(leaf, fi,
6038                                                 new_extents[i].compression);
6039                                 btrfs_set_file_extent_encryption(leaf, fi,
6040                                                 new_extents[i].encryption);
6041                                 btrfs_set_file_extent_other_encoding(leaf, fi,
6042                                                 new_extents[i].other_encoding);
6043
6044                                 btrfs_set_file_extent_num_bytes(leaf, fi,
6045                                                         extent_len);
6046                                 ext_offset += new_extents[i].offset;
6047                                 btrfs_set_file_extent_offset(leaf, fi,
6048                                                         ext_offset);
6049                                 btrfs_mark_buffer_dirty(leaf);
6050
6051                                 btrfs_drop_extent_cache(inode, key.offset,
6052                                                 key.offset + extent_len - 1, 0);
6053
6054                                 ret = btrfs_inc_extent_ref(trans, root,
6055                                                 new_extents[i].disk_bytenr,
6056                                                 new_extents[i].disk_num_bytes,
6057                                                 leaf->start,
6058                                                 root->root_key.objectid,
6059                                                 trans->transid, key.objectid);
6060                                 BUG_ON(ret);
6061                                 btrfs_release_path(root, path);
6062
6063                                 inode_add_bytes(inode, extent_len);
6064
6065                                 ext_offset = 0;
6066                                 num_bytes -= extent_len;
6067                                 key.offset += extent_len;
6068
6069                                 if (num_bytes == 0)
6070                                         break;
6071                         }
6072                         BUG_ON(i >= nr_extents);
6073 #endif
6074                 }
6075
6076                 if (extent_locked) {
6077                         unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
6078                                       lock_end, GFP_NOFS);
6079                         extent_locked = 0;
6080                 }
6081 skip:
6082                 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS &&
6083                     key.offset >= search_end)
6084                         break;
6085
6086                 cond_resched();
6087         }
6088         ret = 0;
6089 out:
6090         btrfs_release_path(root, path);
6091         if (inode) {
6092                 mutex_unlock(&inode->i_mutex);
6093                 if (extent_locked) {
6094                         unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
6095                                       lock_end, GFP_NOFS);
6096                 }
6097                 iput(inode);
6098         }
6099         return ret;
6100 }
6101
6102 int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans,
6103                                struct btrfs_root *root,
6104                                struct extent_buffer *buf, u64 orig_start)
6105 {
6106         int level;
6107         int ret;
6108
6109         BUG_ON(btrfs_header_generation(buf) != trans->transid);
6110         BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
6111
6112         level = btrfs_header_level(buf);
6113         if (level == 0) {
6114                 struct btrfs_leaf_ref *ref;
6115                 struct btrfs_leaf_ref *orig_ref;
6116
6117                 orig_ref = btrfs_lookup_leaf_ref(root, orig_start);
6118                 if (!orig_ref)
6119                         return -ENOENT;
6120
6121                 ref = btrfs_alloc_leaf_ref(root, orig_ref->nritems);
6122                 if (!ref) {
6123                         btrfs_free_leaf_ref(root, orig_ref);
6124                         return -ENOMEM;
6125                 }
6126
6127                 ref->nritems = orig_ref->nritems;
6128                 memcpy(ref->extents, orig_ref->extents,
6129                         sizeof(ref->extents[0]) * ref->nritems);
6130
6131                 btrfs_free_leaf_ref(root, orig_ref);
6132
6133                 ref->root_gen = trans->transid;
6134                 ref->bytenr = buf->start;
6135                 ref->owner = btrfs_header_owner(buf);
6136                 ref->generation = btrfs_header_generation(buf);
6137
6138                 ret = btrfs_add_leaf_ref(root, ref, 0);
6139                 WARN_ON(ret);
6140                 btrfs_free_leaf_ref(root, ref);
6141         }
6142         return 0;
6143 }
6144
6145 static noinline int invalidate_extent_cache(struct btrfs_root *root,
6146                                         struct extent_buffer *leaf,
6147                                         struct btrfs_block_group_cache *group,
6148                                         struct btrfs_root *target_root)
6149 {
6150         struct btrfs_key key;
6151         struct inode *inode = NULL;
6152         struct btrfs_file_extent_item *fi;
6153         u64 num_bytes;
6154         u64 skip_objectid = 0;
6155         u32 nritems;
6156         u32 i;
6157
6158         nritems = btrfs_header_nritems(leaf);
6159         for (i = 0; i < nritems; i++) {
6160                 btrfs_item_key_to_cpu(leaf, &key, i);
6161                 if (key.objectid == skip_objectid ||
6162                     key.type != BTRFS_EXTENT_DATA_KEY)
6163                         continue;
6164                 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
6165                 if (btrfs_file_extent_type(leaf, fi) ==
6166                     BTRFS_FILE_EXTENT_INLINE)
6167                         continue;
6168                 if (btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
6169                         continue;
6170                 if (!inode || inode->i_ino != key.objectid) {
6171                         iput(inode);
6172                         inode = btrfs_ilookup(target_root->fs_info->sb,
6173                                               key.objectid, target_root, 1);
6174                 }
6175                 if (!inode) {
6176                         skip_objectid = key.objectid;
6177                         continue;
6178                 }
6179                 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
6180
6181                 lock_extent(&BTRFS_I(inode)->io_tree, key.offset,
6182                             key.offset + num_bytes - 1, GFP_NOFS);
6183                 btrfs_drop_extent_cache(inode, key.offset,
6184                                         key.offset + num_bytes - 1, 1);
6185                 unlock_extent(&BTRFS_I(inode)->io_tree, key.offset,
6186                               key.offset + num_bytes - 1, GFP_NOFS);
6187                 cond_resched();
6188         }
6189         iput(inode);
6190         return 0;
6191 }
6192
6193 static noinline int replace_extents_in_leaf(struct btrfs_trans_handle *trans,
6194                                         struct btrfs_root *root,
6195                                         struct extent_buffer *leaf,
6196                                         struct btrfs_block_group_cache *group,
6197                                         struct inode *reloc_inode)
6198 {
6199         struct btrfs_key key;
6200         struct btrfs_key extent_key;
6201         struct btrfs_file_extent_item *fi;
6202         struct btrfs_leaf_ref *ref;
6203         struct disk_extent *new_extent;
6204         u64 bytenr;
6205         u64 num_bytes;
6206         u32 nritems;
6207         u32 i;
6208         int ext_index;
6209         int nr_extent;
6210         int ret;
6211
6212         new_extent = kmalloc(sizeof(*new_extent), GFP_NOFS);
6213         BUG_ON(!new_extent);
6214
6215         ref = btrfs_lookup_leaf_ref(root, leaf->start);
6216         BUG_ON(!ref);
6217
6218         ext_index = -1;
6219         nritems = btrfs_header_nritems(leaf);
6220         for (i = 0; i < nritems; i++) {
6221                 btrfs_item_key_to_cpu(leaf, &key, i);
6222                 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
6223                         continue;
6224                 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
6225                 if (btrfs_file_extent_type(leaf, fi) ==
6226                     BTRFS_FILE_EXTENT_INLINE)
6227                         continue;
6228                 bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
6229                 num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
6230                 if (bytenr == 0)
6231                         continue;
6232
6233                 ext_index++;
6234                 if (bytenr >= group->key.objectid + group->key.offset ||
6235                     bytenr + num_bytes <= group->key.objectid)
6236                         continue;
6237
6238                 extent_key.objectid = bytenr;
6239                 extent_key.offset = num_bytes;
6240                 extent_key.type = BTRFS_EXTENT_ITEM_KEY;
6241                 nr_extent = 1;
6242                 ret = get_new_locations(reloc_inode, &extent_key,
6243                                         group->key.objectid, 1,
6244                                         &new_extent, &nr_extent);
6245                 if (ret > 0)
6246                         continue;
6247                 BUG_ON(ret < 0);
6248
6249                 BUG_ON(ref->extents[ext_index].bytenr != bytenr);
6250                 BUG_ON(ref->extents[ext_index].num_bytes != num_bytes);
6251                 ref->extents[ext_index].bytenr = new_extent->disk_bytenr;
6252                 ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes;
6253
6254                 btrfs_set_file_extent_disk_bytenr(leaf, fi,
6255                                                 new_extent->disk_bytenr);
6256                 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
6257                                                 new_extent->disk_num_bytes);
6258                 btrfs_mark_buffer_dirty(leaf);
6259
6260                 ret = btrfs_inc_extent_ref(trans, root,
6261                                         new_extent->disk_bytenr,
6262                                         new_extent->disk_num_bytes,
6263                                         leaf->start,
6264                                         root->root_key.objectid,
6265                                         trans->transid, key.objectid);
6266                 BUG_ON(ret);
6267
6268                 ret = btrfs_free_extent(trans, root,
6269                                         bytenr, num_bytes, leaf->start,
6270                                         btrfs_header_owner(leaf),
6271                                         btrfs_header_generation(leaf),
6272                                         key.objectid, 0);
6273                 BUG_ON(ret);
6274                 cond_resched();
6275         }
6276         kfree(new_extent);
6277         BUG_ON(ext_index + 1 != ref->nritems);
6278         btrfs_free_leaf_ref(root, ref);
6279         return 0;
6280 }
6281
6282 int btrfs_free_reloc_root(struct btrfs_trans_handle *trans,
6283                           struct btrfs_root *root)
6284 {
6285         struct btrfs_root *reloc_root;
6286         int ret;
6287
6288         if (root->reloc_root) {
6289                 reloc_root = root->reloc_root;
6290                 root->reloc_root = NULL;
6291                 list_add(&reloc_root->dead_list,
6292                          &root->fs_info->dead_reloc_roots);
6293
6294                 btrfs_set_root_bytenr(&reloc_root->root_item,
6295                                       reloc_root->node->start);
6296                 btrfs_set_root_level(&root->root_item,
6297                                      btrfs_header_level(reloc_root->node));
6298                 memset(&reloc_root->root_item.drop_progress, 0,
6299                         sizeof(struct btrfs_disk_key));
6300                 reloc_root->root_item.drop_level = 0;
6301
6302                 ret = btrfs_update_root(trans, root->fs_info->tree_root,
6303                                         &reloc_root->root_key,
6304                                         &reloc_root->root_item);
6305                 BUG_ON(ret);
6306         }
6307         return 0;
6308 }
6309
6310 int btrfs_drop_dead_reloc_roots(struct btrfs_root *root)
6311 {
6312         struct btrfs_trans_handle *trans;
6313         struct btrfs_root *reloc_root;
6314         struct btrfs_root *prev_root = NULL;
6315         struct list_head dead_roots;
6316         int ret;
6317         unsigned long nr;
6318
6319         INIT_LIST_HEAD(&dead_roots);
6320         list_splice_init(&root->fs_info->dead_reloc_roots, &dead_roots);
6321
6322         while (!list_empty(&dead_roots)) {
6323                 reloc_root = list_entry(dead_roots.prev,
6324                                         struct btrfs_root, dead_list);
6325                 list_del_init(&reloc_root->dead_list);
6326
6327                 BUG_ON(reloc_root->commit_root != NULL);
6328                 while (1) {
6329                         trans = btrfs_join_transaction(root, 1);
6330                         BUG_ON(!trans);
6331
6332                         mutex_lock(&root->fs_info->drop_mutex);
6333                         ret = btrfs_drop_snapshot(trans, reloc_root);
6334                         if (ret != -EAGAIN)
6335                                 break;
6336                         mutex_unlock(&root->fs_info->drop_mutex);
6337
6338                         nr = trans->blocks_used;
6339                         ret = btrfs_end_transaction(trans, root);
6340                         BUG_ON(ret);
6341                         btrfs_btree_balance_dirty(root, nr);
6342                 }
6343
6344                 free_extent_buffer(reloc_root->node);
6345
6346                 ret = btrfs_del_root(trans, root->fs_info->tree_root,
6347                                      &reloc_root->root_key);
6348                 BUG_ON(ret);
6349                 mutex_unlock(&root->fs_info->drop_mutex);
6350
6351                 nr = trans->blocks_used;
6352                 ret = btrfs_end_transaction(trans, root);
6353                 BUG_ON(ret);
6354                 btrfs_btree_balance_dirty(root, nr);
6355
6356                 kfree(prev_root);
6357                 prev_root = reloc_root;
6358         }
6359         if (prev_root) {
6360                 btrfs_remove_leaf_refs(prev_root, (u64)-1, 0);
6361                 kfree(prev_root);
6362         }
6363         return 0;
6364 }
6365
6366 int btrfs_add_dead_reloc_root(struct btrfs_root *root)
6367 {
6368         list_add(&root->dead_list, &root->fs_info->dead_reloc_roots);
6369         return 0;
6370 }
6371
6372 int btrfs_cleanup_reloc_trees(struct btrfs_root *root)
6373 {
6374         struct btrfs_root *reloc_root;
6375         struct btrfs_trans_handle *trans;
6376         struct btrfs_key location;
6377         int found;
6378         int ret;
6379
6380         mutex_lock(&root->fs_info->tree_reloc_mutex);
6381         ret = btrfs_find_dead_roots(root, BTRFS_TREE_RELOC_OBJECTID, NULL);
6382         BUG_ON(ret);
6383         found = !list_empty(&root->fs_info->dead_reloc_roots);
6384         mutex_unlock(&root->fs_info->tree_reloc_mutex);
6385
6386         if (found) {
6387                 trans = btrfs_start_transaction(root, 1);
6388                 BUG_ON(!trans);
6389                 ret = btrfs_commit_transaction(trans, root);
6390                 BUG_ON(ret);
6391         }
6392
6393         location.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
6394         location.offset = (u64)-1;
6395         location.type = BTRFS_ROOT_ITEM_KEY;
6396
6397         reloc_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
6398         BUG_ON(!reloc_root);
6399         btrfs_orphan_cleanup(reloc_root);
6400         return 0;
6401 }
6402
6403 static noinline int init_reloc_tree(struct btrfs_trans_handle *trans,
6404                                     struct btrfs_root *root)
6405 {
6406         struct btrfs_root *reloc_root;
6407         struct extent_buffer *eb;
6408         struct btrfs_root_item *root_item;
6409         struct btrfs_key root_key;
6410         int ret;
6411
6412         BUG_ON(!root->ref_cows);
6413         if (root->reloc_root)
6414                 return 0;
6415
6416         root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
6417         BUG_ON(!root_item);
6418
6419         ret = btrfs_copy_root(trans, root, root->commit_root,
6420                               &eb, BTRFS_TREE_RELOC_OBJECTID);
6421         BUG_ON(ret);
6422
6423         root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
6424         root_key.offset = root->root_key.objectid;
6425         root_key.type = BTRFS_ROOT_ITEM_KEY;
6426
6427         memcpy(root_item, &root->root_item, sizeof(root_item));
6428         btrfs_set_root_refs(root_item, 0);
6429         btrfs_set_root_bytenr(root_item, eb->start);
6430         btrfs_set_root_level(root_item, btrfs_header_level(eb));
6431         btrfs_set_root_generation(root_item, trans->transid);
6432
6433         btrfs_tree_unlock(eb);
6434         free_extent_buffer(eb);
6435
6436         ret = btrfs_insert_root(trans, root->fs_info->tree_root,
6437                                 &root_key, root_item);
6438         BUG_ON(ret);
6439         kfree(root_item);
6440
6441         reloc_root = btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
6442                                                  &root_key);
6443         BUG_ON(!reloc_root);
6444         reloc_root->last_trans = trans->transid;
6445         reloc_root->commit_root = NULL;
6446         reloc_root->ref_tree = &root->fs_info->reloc_ref_tree;
6447
6448         root->reloc_root = reloc_root;
6449         return 0;
6450 }
6451
6452 /*
6453  * Core function of space balance.
6454  *
6455  * The idea is using reloc trees to relocate tree blocks in reference
6456  * counted roots. There is one reloc tree for each subvol, and all
6457  * reloc trees share same root key objectid. Reloc trees are snapshots
6458  * of the latest committed roots of subvols (root->commit_root).
6459  *
6460  * To relocate a tree block referenced by a subvol, there are two steps.
6461  * COW the block through subvol's reloc tree, then update block pointer
6462  * in the subvol to point to the new block. Since all reloc trees share
6463  * same root key objectid, doing special handing for tree blocks owned
6464  * by them is easy. Once a tree block has been COWed in one reloc tree,
6465  * we can use the resulting new block directly when the same block is
6466  * required to COW again through other reloc trees. By this way, relocated
6467  * tree blocks are shared between reloc trees, so they are also shared
6468  * between subvols.
6469  */
6470 static noinline int relocate_one_path(struct btrfs_trans_handle *trans,
6471                                       struct btrfs_root *root,
6472                                       struct btrfs_path *path,
6473                                       struct btrfs_key *first_key,
6474                                       struct btrfs_ref_path *ref_path,
6475                                       struct btrfs_block_group_cache *group,
6476                                       struct inode *reloc_inode)
6477 {
6478         struct btrfs_root *reloc_root;
6479         struct extent_buffer *eb = NULL;
6480         struct btrfs_key *keys;
6481         u64 *nodes;
6482         int level;
6483         int shared_level;
6484         int lowest_level = 0;
6485         int ret;
6486
6487         if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
6488                 lowest_level = ref_path->owner_objectid;
6489
6490         if (!root->ref_cows) {
6491                 path->lowest_level = lowest_level;
6492                 ret = btrfs_search_slot(trans, root, first_key, path, 0, 1);
6493                 BUG_ON(ret < 0);
6494                 path->lowest_level = 0;
6495                 btrfs_release_path(root, path);
6496                 return 0;
6497         }
6498
6499         mutex_lock(&root->fs_info->tree_reloc_mutex);
6500         ret = init_reloc_tree(trans, root);
6501         BUG_ON(ret);
6502         reloc_root = root->reloc_root;
6503
6504         shared_level = ref_path->shared_level;
6505         ref_path->shared_level = BTRFS_MAX_LEVEL - 1;
6506
6507         keys = ref_path->node_keys;
6508         nodes = ref_path->new_nodes;
6509         memset(&keys[shared_level + 1], 0,
6510                sizeof(*keys) * (BTRFS_MAX_LEVEL - shared_level - 1));
6511         memset(&nodes[shared_level + 1], 0,
6512                sizeof(*nodes) * (BTRFS_MAX_LEVEL - shared_level - 1));
6513
6514         if (nodes[lowest_level] == 0) {
6515                 path->lowest_level = lowest_level;
6516                 ret = btrfs_search_slot(trans, reloc_root, first_key, path,
6517                                         0, 1);
6518                 BUG_ON(ret);
6519                 for (level = lowest_level; level < BTRFS_MAX_LEVEL; level++) {
6520                         eb = path->nodes[level];
6521                         if (!eb || eb == reloc_root->node)
6522                                 break;
6523                         nodes[level] = eb->start;
6524                         if (level == 0)
6525                                 btrfs_item_key_to_cpu(eb, &keys[level], 0);
6526                         else
6527                                 btrfs_node_key_to_cpu(eb, &keys[level], 0);
6528                 }
6529                 if (nodes[0] &&
6530                     ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
6531                         eb = path->nodes[0];
6532                         ret = replace_extents_in_leaf(trans, reloc_root, eb,
6533                                                       group, reloc_inode);
6534                         BUG_ON(ret);
6535                 }
6536                 btrfs_release_path(reloc_root, path);
6537         } else {
6538                 ret = btrfs_merge_path(trans, reloc_root, keys, nodes,
6539                                        lowest_level);
6540                 BUG_ON(ret);
6541         }
6542
6543         /*
6544          * replace tree blocks in the fs tree with tree blocks in
6545          * the reloc tree.
6546          */
6547         ret = btrfs_merge_path(trans, root, keys, nodes, lowest_level);
6548         BUG_ON(ret < 0);
6549
6550         if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
6551                 ret = btrfs_search_slot(trans, reloc_root, first_key, path,
6552                                         0, 0);
6553                 BUG_ON(ret);
6554                 extent_buffer_get(path->nodes[0]);
6555                 eb = path->nodes[0];
6556                 btrfs_release_path(reloc_root, path);
6557                 ret = invalidate_extent_cache(reloc_root, eb, group, root);
6558                 BUG_ON(ret);
6559                 free_extent_buffer(eb);
6560         }
6561
6562         mutex_unlock(&root->fs_info->tree_reloc_mutex);
6563         path->lowest_level = 0;
6564         return 0;
6565 }
6566
6567 static noinline int relocate_tree_block(struct btrfs_trans_handle *trans,
6568                                         struct btrfs_root *root,
6569                                         struct btrfs_path *path,
6570                                         struct btrfs_key *first_key,
6571                                         struct btrfs_ref_path *ref_path)
6572 {
6573         int ret;
6574
6575         ret = relocate_one_path(trans, root, path, first_key,
6576                                 ref_path, NULL, NULL);
6577         BUG_ON(ret);
6578
6579         return 0;
6580 }
6581
6582 static noinline int del_extent_zero(struct btrfs_trans_handle *trans,
6583                                     struct btrfs_root *extent_root,
6584                                     struct btrfs_path *path,
6585                                     struct btrfs_key *extent_key)
6586 {
6587         int ret;
6588
6589         ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1);
6590         if (ret)
6591                 goto out;
6592         ret = btrfs_del_item(trans, extent_root, path);
6593 out:
6594         btrfs_release_path(extent_root, path);
6595         return ret;
6596 }
6597
6598 static noinline struct btrfs_root *read_ref_root(struct btrfs_fs_info *fs_info,
6599                                                 struct btrfs_ref_path *ref_path)
6600 {
6601         struct btrfs_key root_key;
6602
6603         root_key.objectid = ref_path->root_objectid;
6604         root_key.type = BTRFS_ROOT_ITEM_KEY;
6605         if (is_cowonly_root(ref_path->root_objectid))
6606                 root_key.offset = 0;
6607         else
6608                 root_key.offset = (u64)-1;
6609
6610         return btrfs_read_fs_root_no_name(fs_info, &root_key);
6611 }
6612
6613 static noinline int relocate_one_extent(struct btrfs_root *extent_root,
6614                                         struct btrfs_path *path,
6615                                         struct btrfs_key *extent_key,
6616                                         struct btrfs_block_group_cache *group,
6617                                         struct inode *reloc_inode, int pass)
6618 {
6619         struct btrfs_trans_handle *trans;
6620         struct btrfs_root *found_root;
6621         struct btrfs_ref_path *ref_path = NULL;
6622         struct disk_extent *new_extents = NULL;
6623         int nr_extents = 0;
6624         int loops;
6625         int ret;
6626         int level;
6627         struct btrfs_key first_key;
6628         u64 prev_block = 0;
6629
6630
6631         trans = btrfs_start_transaction(extent_root, 1);
6632         BUG_ON(!trans);
6633
6634         if (extent_key->objectid == 0) {
6635                 ret = del_extent_zero(trans, extent_root, path, extent_key);
6636                 goto out;
6637         }
6638
6639         ref_path = kmalloc(sizeof(*ref_path), GFP_NOFS);
6640         if (!ref_path) {
6641                 ret = -ENOMEM;
6642                 goto out;
6643         }
6644
6645         for (loops = 0; ; loops++) {
6646                 if (loops == 0) {
6647                         ret = btrfs_first_ref_path(trans, extent_root, ref_path,
6648                                                    extent_key->objectid);
6649                 } else {
6650                         ret = btrfs_next_ref_path(trans, extent_root, ref_path);
6651                 }
6652                 if (ret < 0)
6653                         goto out;
6654                 if (ret > 0)
6655                         break;
6656
6657                 if (ref_path->root_objectid == BTRFS_TREE_LOG_OBJECTID ||
6658                     ref_path->root_objectid == BTRFS_TREE_RELOC_OBJECTID)
6659                         continue;
6660
6661                 found_root = read_ref_root(extent_root->fs_info, ref_path);
6662                 BUG_ON(!found_root);
6663                 /*
6664                  * for reference counted tree, only process reference paths
6665                  * rooted at the latest committed root.
6666                  */
6667                 if (found_root->ref_cows &&
6668                     ref_path->root_generation != found_root->root_key.offset)
6669                         continue;
6670
6671                 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
6672                         if (pass == 0) {
6673                                 /*
6674                                  * copy data extents to new locations
6675                                  */
6676                                 u64 group_start = group->key.objectid;
6677                                 ret = relocate_data_extent(reloc_inode,
6678                                                            extent_key,
6679                                                            group_start);
6680                                 if (ret < 0)
6681                                         goto out;
6682                                 break;
6683                         }
6684                         level = 0;
6685                 } else {
6686                         level = ref_path->owner_objectid;
6687                 }
6688
6689                 if (prev_block != ref_path->nodes[level]) {
6690                         struct extent_buffer *eb;
6691                         u64 block_start = ref_path->nodes[level];
6692                         u64 block_size = btrfs_level_size(found_root, level);
6693
6694                         eb = read_tree_block(found_root, block_start,
6695                                              block_size, 0);
6696                         btrfs_tree_lock(eb);
6697                         BUG_ON(level != btrfs_header_level(eb));
6698
6699                         if (level == 0)
6700                                 btrfs_item_key_to_cpu(eb, &first_key, 0);
6701                         else
6702                                 btrfs_node_key_to_cpu(eb, &first_key, 0);
6703
6704                         btrfs_tree_unlock(eb);
6705                         free_extent_buffer(eb);
6706                         prev_block = block_start;
6707                 }
6708
6709                 mutex_lock(&extent_root->fs_info->trans_mutex);
6710                 btrfs_record_root_in_trans(found_root);
6711                 mutex_unlock(&extent_root->fs_info->trans_mutex);
6712                 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
6713                         /*
6714                          * try to update data extent references while
6715                          * keeping metadata shared between snapshots.
6716                          */
6717                         if (pass == 1) {
6718                                 ret = relocate_one_path(trans, found_root,
6719                                                 path, &first_key, ref_path,
6720                                                 group, reloc_inode);
6721                                 if (ret < 0)
6722                                         goto out;
6723                                 continue;
6724                         }
6725                         /*
6726                          * use fallback method to process the remaining
6727                          * references.
6728                          */
6729                         if (!new_extents) {
6730                                 u64 group_start = group->key.objectid;
6731                                 new_extents = kmalloc(sizeof(*new_extents),
6732                                                       GFP_NOFS);
6733                                 nr_extents = 1;
6734                                 ret = get_new_locations(reloc_inode,
6735                                                         extent_key,
6736                                                         group_start, 1,
6737                                                         &new_extents,
6738                                                         &nr_extents);
6739                                 if (ret)
6740                                         goto out;
6741                         }
6742                         ret = replace_one_extent(trans, found_root,
6743                                                 path, extent_key,
6744                                                 &first_key, ref_path,
6745                                                 new_extents, nr_extents);
6746                 } else {
6747                         ret = relocate_tree_block(trans, found_root, path,
6748                                                   &first_key, ref_path);
6749                 }
6750                 if (ret < 0)
6751                         goto out;
6752         }
6753         ret = 0;
6754 out:
6755         btrfs_end_transaction(trans, extent_root);
6756         kfree(new_extents);
6757         kfree(ref_path);
6758         return ret;
6759 }
6760 #endif
6761
6762 static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
6763 {
6764         u64 num_devices;
6765         u64 stripped = BTRFS_BLOCK_GROUP_RAID0 |
6766                 BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
6767
6768         num_devices = root->fs_info->fs_devices->rw_devices;
6769         if (num_devices == 1) {
6770                 stripped |= BTRFS_BLOCK_GROUP_DUP;
6771                 stripped = flags & ~stripped;
6772
6773                 /* turn raid0 into single device chunks */
6774                 if (flags & BTRFS_BLOCK_GROUP_RAID0)
6775                         return stripped;
6776
6777                 /* turn mirroring into duplication */
6778                 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
6779                              BTRFS_BLOCK_GROUP_RAID10))
6780                         return stripped | BTRFS_BLOCK_GROUP_DUP;
6781                 return flags;
6782         } else {
6783                 /* they already had raid on here, just return */
6784                 if (flags & stripped)
6785                         return flags;
6786
6787                 stripped |= BTRFS_BLOCK_GROUP_DUP;
6788                 stripped = flags & ~stripped;
6789
6790                 /* switch duplicated blocks with raid1 */
6791                 if (flags & BTRFS_BLOCK_GROUP_DUP)
6792                         return stripped | BTRFS_BLOCK_GROUP_RAID1;
6793
6794                 /* turn single device chunks into raid0 */
6795                 return stripped | BTRFS_BLOCK_GROUP_RAID0;
6796         }
6797         return flags;
6798 }
6799
6800 static int __alloc_chunk_for_shrink(struct btrfs_root *root,
6801                      struct btrfs_block_group_cache *shrink_block_group,
6802                      int force)
6803 {
6804         struct btrfs_trans_handle *trans;
6805         u64 new_alloc_flags;
6806         u64 calc;
6807
6808         spin_lock(&shrink_block_group->lock);
6809         if (btrfs_block_group_used(&shrink_block_group->item) +
6810             shrink_block_group->reserved > 0) {
6811                 spin_unlock(&shrink_block_group->lock);
6812
6813                 trans = btrfs_start_transaction(root, 1);
6814                 spin_lock(&shrink_block_group->lock);
6815
6816                 new_alloc_flags = update_block_group_flags(root,
6817                                                    shrink_block_group->flags);
6818                 if (new_alloc_flags != shrink_block_group->flags) {
6819                         calc =
6820                              btrfs_block_group_used(&shrink_block_group->item);
6821                 } else {
6822                         calc = shrink_block_group->key.offset;
6823                 }
6824                 spin_unlock(&shrink_block_group->lock);
6825
6826                 do_chunk_alloc(trans, root->fs_info->extent_root,
6827                                calc + 2 * 1024 * 1024, new_alloc_flags, force);
6828
6829                 btrfs_end_transaction(trans, root);
6830         } else
6831                 spin_unlock(&shrink_block_group->lock);
6832         return 0;
6833 }
6834
6835
6836 int btrfs_prepare_block_group_relocation(struct btrfs_root *root,
6837                                          struct btrfs_block_group_cache *group)
6838
6839 {
6840         __alloc_chunk_for_shrink(root, group, 1);
6841         set_block_group_readonly(group);
6842         return 0;
6843 }
6844
6845 #if 0
6846 static int __insert_orphan_inode(struct btrfs_trans_handle *trans,
6847                                  struct btrfs_root *root,
6848                                  u64 objectid, u64 size)
6849 {
6850         struct btrfs_path *path;
6851         struct btrfs_inode_item *item;
6852         struct extent_buffer *leaf;
6853         int ret;
6854
6855         path = btrfs_alloc_path();
6856         if (!path)
6857                 return -ENOMEM;
6858
6859         path->leave_spinning = 1;
6860         ret = btrfs_insert_empty_inode(trans, root, path, objectid);
6861         if (ret)
6862                 goto out;
6863
6864         leaf = path->nodes[0];
6865         item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_inode_item);
6866         memset_extent_buffer(leaf, 0, (unsigned long)item, sizeof(*item));
6867         btrfs_set_inode_generation(leaf, item, 1);
6868         btrfs_set_inode_size(leaf, item, size);
6869         btrfs_set_inode_mode(leaf, item, S_IFREG | 0600);
6870         btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NOCOMPRESS);
6871         btrfs_mark_buffer_dirty(leaf);
6872         btrfs_release_path(root, path);
6873 out:
6874         btrfs_free_path(path);
6875         return ret;
6876 }
6877
6878 static noinline struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info,
6879                                         struct btrfs_block_group_cache *group)
6880 {
6881         struct inode *inode = NULL;
6882         struct btrfs_trans_handle *trans;
6883         struct btrfs_root *root;
6884         struct btrfs_key root_key;
6885         u64 objectid = BTRFS_FIRST_FREE_OBJECTID;
6886         int err = 0;
6887
6888         root_key.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
6889         root_key.type = BTRFS_ROOT_ITEM_KEY;
6890         root_key.offset = (u64)-1;
6891         root = btrfs_read_fs_root_no_name(fs_info, &root_key);
6892         if (IS_ERR(root))
6893                 return ERR_CAST(root);
6894
6895         trans = btrfs_start_transaction(root, 1);
6896         BUG_ON(!trans);
6897
6898         err = btrfs_find_free_objectid(trans, root, objectid, &objectid);
6899         if (err)
6900                 goto out;
6901
6902         err = __insert_orphan_inode(trans, root, objectid, group->key.offset);
6903         BUG_ON(err);
6904
6905         err = btrfs_insert_file_extent(trans, root, objectid, 0, 0, 0,
6906                                        group->key.offset, 0, group->key.offset,
6907                                        0, 0, 0);
6908         BUG_ON(err);
6909
6910         inode = btrfs_iget_locked(root->fs_info->sb, objectid, root);
6911         if (inode->i_state & I_NEW) {
6912                 BTRFS_I(inode)->root = root;
6913                 BTRFS_I(inode)->location.objectid = objectid;
6914                 BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
6915                 BTRFS_I(inode)->location.offset = 0;
6916                 btrfs_read_locked_inode(inode);
6917                 unlock_new_inode(inode);
6918                 BUG_ON(is_bad_inode(inode));
6919         } else {
6920                 BUG_ON(1);
6921         }
6922         BTRFS_I(inode)->index_cnt = group->key.objectid;
6923
6924         err = btrfs_orphan_add(trans, inode);
6925 out:
6926         btrfs_end_transaction(trans, root);
6927         if (err) {
6928                 if (inode)
6929                         iput(inode);
6930                 inode = ERR_PTR(err);
6931         }
6932         return inode;
6933 }
6934
6935 int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len)
6936 {
6937
6938         struct btrfs_ordered_sum *sums;
6939         struct btrfs_sector_sum *sector_sum;
6940         struct btrfs_ordered_extent *ordered;
6941         struct btrfs_root *root = BTRFS_I(inode)->root;
6942         struct list_head list;
6943         size_t offset;
6944         int ret;
6945         u64 disk_bytenr;
6946
6947         INIT_LIST_HEAD(&list);
6948
6949         ordered = btrfs_lookup_ordered_extent(inode, file_pos);
6950         BUG_ON(ordered->file_offset != file_pos || ordered->len != len);
6951
6952         disk_bytenr = file_pos + BTRFS_I(inode)->index_cnt;
6953         ret = btrfs_lookup_csums_range(root->fs_info->csum_root, disk_bytenr,
6954                                        disk_bytenr + len - 1, &list);
6955
6956         while (!list_empty(&list)) {
6957                 sums = list_entry(list.next, struct btrfs_ordered_sum, list);
6958                 list_del_init(&sums->list);
6959
6960                 sector_sum = sums->sums;
6961                 sums->bytenr = ordered->start;
6962
6963                 offset = 0;
6964                 while (offset < sums->len) {
6965                         sector_sum->bytenr += ordered->start - disk_bytenr;
6966                         sector_sum++;
6967                         offset += root->sectorsize;
6968                 }
6969
6970                 btrfs_add_ordered_sum(inode, ordered, sums);
6971         }
6972         btrfs_put_ordered_extent(ordered);
6973         return 0;
6974 }
6975
6976 int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start)
6977 {
6978         struct btrfs_trans_handle *trans;
6979         struct btrfs_path *path;
6980         struct btrfs_fs_info *info = root->fs_info;
6981         struct extent_buffer *leaf;
6982         struct inode *reloc_inode;
6983         struct btrfs_block_group_cache *block_group;
6984         struct btrfs_key key;
6985         u64 skipped;
6986         u64 cur_byte;
6987         u64 total_found;
6988         u32 nritems;
6989         int ret;
6990         int progress;
6991         int pass = 0;
6992
6993         root = root->fs_info->extent_root;
6994
6995         block_group = btrfs_lookup_block_group(info, group_start);
6996         BUG_ON(!block_group);
6997
6998         printk(KERN_INFO "btrfs relocating block group %llu flags %llu\n",
6999                (unsigned long long)block_group->key.objectid,
7000                (unsigned long long)block_group->flags);
7001
7002         path = btrfs_alloc_path();
7003         BUG_ON(!path);
7004
7005         reloc_inode = create_reloc_inode(info, block_group);
7006         BUG_ON(IS_ERR(reloc_inode));
7007
7008         __alloc_chunk_for_shrink(root, block_group, 1);
7009         set_block_group_readonly(block_group);
7010
7011         btrfs_start_delalloc_inodes(info->tree_root);
7012         btrfs_wait_ordered_extents(info->tree_root, 0);
7013 again:
7014         skipped = 0;
7015         total_found = 0;
7016         progress = 0;
7017         key.objectid = block_group->key.objectid;
7018         key.offset = 0;
7019         key.type = 0;
7020         cur_byte = key.objectid;
7021
7022         trans = btrfs_start_transaction(info->tree_root, 1);
7023         btrfs_commit_transaction(trans, info->tree_root);
7024
7025         mutex_lock(&root->fs_info->cleaner_mutex);
7026         btrfs_clean_old_snapshots(info->tree_root);
7027         btrfs_remove_leaf_refs(info->tree_root, (u64)-1, 1);
7028         mutex_unlock(&root->fs_info->cleaner_mutex);
7029
7030         trans = btrfs_start_transaction(info->tree_root, 1);
7031         btrfs_commit_transaction(trans, info->tree_root);
7032
7033         while (1) {
7034                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
7035                 if (ret < 0)
7036                         goto out;
7037 next:
7038                 leaf = path->nodes[0];
7039                 nritems = btrfs_header_nritems(leaf);
7040                 if (path->slots[0] >= nritems) {
7041                         ret = btrfs_next_leaf(root, path);
7042                         if (ret < 0)
7043                                 goto out;
7044                         if (ret == 1) {
7045                                 ret = 0;
7046                                 break;
7047                         }
7048                         leaf = path->nodes[0];
7049                         nritems = btrfs_header_nritems(leaf);
7050                 }
7051
7052                 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
7053
7054                 if (key.objectid >= block_group->key.objectid +
7055                     block_group->key.offset)
7056                         break;
7057
7058                 if (progress && need_resched()) {
7059                         btrfs_release_path(root, path);
7060                         cond_resched();
7061                         progress = 0;
7062                         continue;
7063                 }
7064                 progress = 1;
7065
7066                 if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY ||
7067                     key.objectid + key.offset <= cur_byte) {
7068                         path->slots[0]++;
7069                         goto next;
7070                 }
7071
7072                 total_found++;
7073                 cur_byte = key.objectid + key.offset;
7074                 btrfs_release_path(root, path);
7075
7076                 __alloc_chunk_for_shrink(root, block_group, 0);
7077                 ret = relocate_one_extent(root, path, &key, block_group,
7078                                           reloc_inode, pass);
7079                 BUG_ON(ret < 0);
7080                 if (ret > 0)
7081                         skipped++;
7082
7083                 key.objectid = cur_byte;
7084                 key.type = 0;
7085                 key.offset = 0;
7086         }
7087
7088         btrfs_release_path(root, path);
7089
7090         if (pass == 0) {
7091                 btrfs_wait_ordered_range(reloc_inode, 0, (u64)-1);
7092                 invalidate_mapping_pages(reloc_inode->i_mapping, 0, -1);
7093         }
7094
7095         if (total_found > 0) {
7096                 printk(KERN_INFO "btrfs found %llu extents in pass %d\n",
7097                        (unsigned long long)total_found, pass);
7098                 pass++;
7099                 if (total_found == skipped && pass > 2) {
7100                         iput(reloc_inode);
7101                         reloc_inode = create_reloc_inode(info, block_group);
7102                         pass = 0;
7103                 }
7104                 goto again;
7105         }
7106
7107         /* delete reloc_inode */
7108         iput(reloc_inode);
7109
7110         /* unpin extents in this range */
7111         trans = btrfs_start_transaction(info->tree_root, 1);
7112         btrfs_commit_transaction(trans, info->tree_root);
7113
7114         spin_lock(&block_group->lock);
7115         WARN_ON(block_group->pinned > 0);
7116         WARN_ON(block_group->reserved > 0);
7117         WARN_ON(btrfs_block_group_used(&block_group->item) > 0);
7118         spin_unlock(&block_group->lock);
7119         btrfs_put_block_group(block_group);
7120         ret = 0;
7121 out:
7122         btrfs_free_path(path);
7123         return ret;
7124 }
7125 #endif
7126
7127 static int find_first_block_group(struct btrfs_root *root,
7128                 struct btrfs_path *path, struct btrfs_key *key)
7129 {
7130         int ret = 0;
7131         struct btrfs_key found_key;
7132         struct extent_buffer *leaf;
7133         int slot;
7134
7135         ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
7136         if (ret < 0)
7137                 goto out;
7138
7139         while (1) {
7140                 slot = path->slots[0];
7141                 leaf = path->nodes[0];
7142                 if (slot >= btrfs_header_nritems(leaf)) {
7143                         ret = btrfs_next_leaf(root, path);
7144                         if (ret == 0)
7145                                 continue;
7146                         if (ret < 0)
7147                                 goto out;
7148                         break;
7149                 }
7150                 btrfs_item_key_to_cpu(leaf, &found_key, slot);
7151
7152                 if (found_key.objectid >= key->objectid &&
7153                     found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
7154                         ret = 0;
7155                         goto out;
7156                 }
7157                 path->slots[0]++;
7158         }
7159         ret = -ENOENT;
7160 out:
7161         return ret;
7162 }
7163
7164 int btrfs_free_block_groups(struct btrfs_fs_info *info)
7165 {
7166         struct btrfs_block_group_cache *block_group;
7167         struct btrfs_space_info *space_info;
7168         struct rb_node *n;
7169
7170         spin_lock(&info->block_group_cache_lock);
7171         while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
7172                 block_group = rb_entry(n, struct btrfs_block_group_cache,
7173                                        cache_node);
7174                 rb_erase(&block_group->cache_node,
7175                          &info->block_group_cache_tree);
7176                 spin_unlock(&info->block_group_cache_lock);
7177
7178                 down_write(&block_group->space_info->groups_sem);
7179                 list_del(&block_group->list);
7180                 up_write(&block_group->space_info->groups_sem);
7181
7182                 if (block_group->cached == BTRFS_CACHE_STARTED)
7183                         wait_event(block_group->caching_q,
7184                                    block_group_cache_done(block_group));
7185
7186                 btrfs_remove_free_space_cache(block_group);
7187
7188                 WARN_ON(atomic_read(&block_group->count) != 1);
7189                 kfree(block_group);
7190
7191                 spin_lock(&info->block_group_cache_lock);
7192         }
7193         spin_unlock(&info->block_group_cache_lock);
7194
7195         /* now that all the block groups are freed, go through and
7196          * free all the space_info structs.  This is only called during
7197          * the final stages of unmount, and so we know nobody is
7198          * using them.  We call synchronize_rcu() once before we start,
7199          * just to be on the safe side.
7200          */
7201         synchronize_rcu();
7202
7203         while(!list_empty(&info->space_info)) {
7204                 space_info = list_entry(info->space_info.next,
7205                                         struct btrfs_space_info,
7206                                         list);
7207
7208                 list_del(&space_info->list);
7209                 kfree(space_info);
7210         }
7211         return 0;
7212 }
7213
7214 int btrfs_read_block_groups(struct btrfs_root *root)
7215 {
7216         struct btrfs_path *path;
7217         int ret;
7218         struct btrfs_block_group_cache *cache;
7219         struct btrfs_fs_info *info = root->fs_info;
7220         struct btrfs_space_info *space_info;
7221         struct btrfs_key key;
7222         struct btrfs_key found_key;
7223         struct extent_buffer *leaf;
7224
7225         root = info->extent_root;
7226         key.objectid = 0;
7227         key.offset = 0;
7228         btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
7229         path = btrfs_alloc_path();
7230         if (!path)
7231                 return -ENOMEM;
7232
7233         while (1) {
7234                 ret = find_first_block_group(root, path, &key);
7235                 if (ret > 0) {
7236                         ret = 0;
7237                         goto error;
7238                 }
7239                 if (ret != 0)
7240                         goto error;
7241
7242                 leaf = path->nodes[0];
7243                 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
7244                 cache = kzalloc(sizeof(*cache), GFP_NOFS);
7245                 if (!cache) {
7246                         ret = -ENOMEM;
7247                         break;
7248                 }
7249
7250                 atomic_set(&cache->count, 1);
7251                 spin_lock_init(&cache->lock);
7252                 spin_lock_init(&cache->tree_lock);
7253                 cache->fs_info = info;
7254                 init_waitqueue_head(&cache->caching_q);
7255                 INIT_LIST_HEAD(&cache->list);
7256                 INIT_LIST_HEAD(&cache->cluster_list);
7257
7258                 /*
7259                  * we only want to have 32k of ram per block group for keeping
7260                  * track of free space, and if we pass 1/2 of that we want to
7261                  * start converting things over to using bitmaps
7262                  */
7263                 cache->extents_thresh = ((1024 * 32) / 2) /
7264                         sizeof(struct btrfs_free_space);
7265
7266                 read_extent_buffer(leaf, &cache->item,
7267                                    btrfs_item_ptr_offset(leaf, path->slots[0]),
7268                                    sizeof(cache->item));
7269                 memcpy(&cache->key, &found_key, sizeof(found_key));
7270
7271                 key.objectid = found_key.objectid + found_key.offset;
7272                 btrfs_release_path(root, path);
7273                 cache->flags = btrfs_block_group_flags(&cache->item);
7274                 cache->sectorsize = root->sectorsize;
7275
7276                 remove_sb_from_cache(root, cache);
7277
7278                 /*
7279                  * check for two cases, either we are full, and therefore
7280                  * don't need to bother with the caching work since we won't
7281                  * find any space, or we are empty, and we can just add all
7282                  * the space in and be done with it.  This saves us _alot_ of
7283                  * time, particularly in the full case.
7284                  */
7285                 if (found_key.offset == btrfs_block_group_used(&cache->item)) {
7286                         cache->cached = BTRFS_CACHE_FINISHED;
7287                 } else if (btrfs_block_group_used(&cache->item) == 0) {
7288                         cache->cached = BTRFS_CACHE_FINISHED;
7289                         add_new_free_space(cache, root->fs_info,
7290                                            found_key.objectid,
7291                                            found_key.objectid +
7292                                            found_key.offset);
7293                 }
7294
7295                 ret = update_space_info(info, cache->flags, found_key.offset,
7296                                         btrfs_block_group_used(&cache->item),
7297                                         &space_info);
7298                 BUG_ON(ret);
7299                 cache->space_info = space_info;
7300                 down_write(&space_info->groups_sem);
7301                 list_add_tail(&cache->list, &space_info->block_groups);
7302                 up_write(&space_info->groups_sem);
7303
7304                 ret = btrfs_add_block_group_cache(root->fs_info, cache);
7305                 BUG_ON(ret);
7306
7307                 set_avail_alloc_bits(root->fs_info, cache->flags);
7308                 if (btrfs_chunk_readonly(root, cache->key.objectid))
7309                         set_block_group_readonly(cache);
7310         }
7311         ret = 0;
7312 error:
7313         btrfs_free_path(path);
7314         return ret;
7315 }
7316
7317 int btrfs_make_block_group(struct btrfs_trans_handle *trans,
7318                            struct btrfs_root *root, u64 bytes_used,
7319                            u64 type, u64 chunk_objectid, u64 chunk_offset,
7320                            u64 size)
7321 {
7322         int ret;
7323         struct btrfs_root *extent_root;
7324         struct btrfs_block_group_cache *cache;
7325
7326         extent_root = root->fs_info->extent_root;
7327
7328         root->fs_info->last_trans_log_full_commit = trans->transid;
7329
7330         cache = kzalloc(sizeof(*cache), GFP_NOFS);
7331         if (!cache)
7332                 return -ENOMEM;
7333
7334         cache->key.objectid = chunk_offset;
7335         cache->key.offset = size;
7336         cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
7337         cache->sectorsize = root->sectorsize;
7338
7339         /*
7340          * we only want to have 32k of ram per block group for keeping track
7341          * of free space, and if we pass 1/2 of that we want to start
7342          * converting things over to using bitmaps
7343          */
7344         cache->extents_thresh = ((1024 * 32) / 2) /
7345                 sizeof(struct btrfs_free_space);
7346         atomic_set(&cache->count, 1);
7347         spin_lock_init(&cache->lock);
7348         spin_lock_init(&cache->tree_lock);
7349         init_waitqueue_head(&cache->caching_q);
7350         INIT_LIST_HEAD(&cache->list);
7351         INIT_LIST_HEAD(&cache->cluster_list);
7352
7353         btrfs_set_block_group_used(&cache->item, bytes_used);
7354         btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
7355         cache->flags = type;
7356         btrfs_set_block_group_flags(&cache->item, type);
7357
7358         cache->cached = BTRFS_CACHE_FINISHED;
7359         remove_sb_from_cache(root, cache);
7360
7361         add_new_free_space(cache, root->fs_info, chunk_offset,
7362                            chunk_offset + size);
7363
7364         ret = update_space_info(root->fs_info, cache->flags, size, bytes_used,
7365                                 &cache->space_info);
7366         BUG_ON(ret);
7367         down_write(&cache->space_info->groups_sem);
7368         list_add_tail(&cache->list, &cache->space_info->block_groups);
7369         up_write(&cache->space_info->groups_sem);
7370
7371         ret = btrfs_add_block_group_cache(root->fs_info, cache);
7372         BUG_ON(ret);
7373
7374         ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
7375                                 sizeof(cache->item));
7376         BUG_ON(ret);
7377
7378         set_avail_alloc_bits(extent_root->fs_info, type);
7379
7380         return 0;
7381 }
7382
7383 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
7384                              struct btrfs_root *root, u64 group_start)
7385 {
7386         struct btrfs_path *path;
7387         struct btrfs_block_group_cache *block_group;
7388         struct btrfs_free_cluster *cluster;
7389         struct btrfs_key key;
7390         int ret;
7391
7392         root = root->fs_info->extent_root;
7393
7394         block_group = btrfs_lookup_block_group(root->fs_info, group_start);
7395         BUG_ON(!block_group);
7396         BUG_ON(!block_group->ro);
7397
7398         memcpy(&key, &block_group->key, sizeof(key));
7399
7400         /* make sure this block group isn't part of an allocation cluster */
7401         cluster = &root->fs_info->data_alloc_cluster;
7402         spin_lock(&cluster->refill_lock);
7403         btrfs_return_cluster_to_free_space(block_group, cluster);
7404         spin_unlock(&cluster->refill_lock);
7405
7406         /*
7407          * make sure this block group isn't part of a metadata
7408          * allocation cluster
7409          */
7410         cluster = &root->fs_info->meta_alloc_cluster;
7411         spin_lock(&cluster->refill_lock);
7412         btrfs_return_cluster_to_free_space(block_group, cluster);
7413         spin_unlock(&cluster->refill_lock);
7414
7415         path = btrfs_alloc_path();
7416         BUG_ON(!path);
7417
7418         spin_lock(&root->fs_info->block_group_cache_lock);
7419         rb_erase(&block_group->cache_node,
7420                  &root->fs_info->block_group_cache_tree);
7421         spin_unlock(&root->fs_info->block_group_cache_lock);
7422
7423         down_write(&block_group->space_info->groups_sem);
7424         /*
7425          * we must use list_del_init so people can check to see if they
7426          * are still on the list after taking the semaphore
7427          */
7428         list_del_init(&block_group->list);
7429         up_write(&block_group->space_info->groups_sem);
7430
7431         if (block_group->cached == BTRFS_CACHE_STARTED)
7432                 wait_event(block_group->caching_q,
7433                            block_group_cache_done(block_group));
7434
7435         btrfs_remove_free_space_cache(block_group);
7436
7437         spin_lock(&block_group->space_info->lock);
7438         block_group->space_info->total_bytes -= block_group->key.offset;
7439         block_group->space_info->bytes_readonly -= block_group->key.offset;
7440         spin_unlock(&block_group->space_info->lock);
7441
7442         btrfs_clear_space_info_full(root->fs_info);
7443
7444         btrfs_put_block_group(block_group);
7445         btrfs_put_block_group(block_group);
7446
7447         ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
7448         if (ret > 0)
7449                 ret = -EIO;
7450         if (ret < 0)
7451                 goto out;
7452
7453         ret = btrfs_del_item(trans, root, path);
7454 out:
7455         btrfs_free_path(path);
7456         return ret;
7457 }