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
  19 #include <linux/sched.h>
  20 #include "ctree.h"
  21 #include "disk-io.h"
  22 #include "print-tree.h"
  23 #include "transaction.h"
  24
  25 #define BLOCK_GROUP_DATA EXTENT_WRITEBACK
  26 #define BLOCK_GROUP_METADATA EXTENT_UPTODATE
  27 #define BLOCK_GROUP_DIRTY EXTENT_DIRTY
  28
  29 static int finish_current_insert(struct btrfs_trans_handle *trans, struct
  30                                  btrfs_root *extent_root);
  31 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
  32                                btrfs_root *extent_root);
  33
  34 static int cache_block_group(struct btrfs_root *root,
  35                              struct btrfs_block_group_cache *block_group)
  36 {
  37         struct btrfs_path *path;
  38         int ret;
  39         struct btrfs_key key;
  40         struct extent_buffer *leaf;
  41         struct extent_map_tree *free_space_cache;
  42         int slot;
  43         u64 last = 0;
  44         u64 hole_size;
  45         u64 first_free;
  46         int found = 0;
  47
  48         if (!block_group)
  49                 return 0;
  50
  51         root = root->fs_info->extent_root;
  52         free_space_cache = &root->fs_info->free_space_cache;
  53
  54         if (block_group->cached)
  55                 return 0;
  56
  57         path = btrfs_alloc_path();
  58         if (!path)
  59                 return -ENOMEM;
  60
  61         path->reada = 2;
  62         first_free = block_group->key.objectid;
  63         key.objectid = block_group->key.objectid;
  64         key.offset = 0;
  65
  66         btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
  67         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
  68
  69         if (ret < 0)
  70                 return ret;
  71
  72         if (ret && path->slots[0] > 0)
  73                 path->slots[0]--;
  74
  75         while(1) {
  76                 leaf = path->nodes[0];
  77                 slot = path->slots[0];
  78                 if (slot >= btrfs_header_nritems(leaf)) {
  79                         ret = btrfs_next_leaf(root, path);
  80                         if (ret < 0)
  81                                 goto err;
  82                         if (ret == 0) {
  83                                 continue;
  84                         } else {
  85                                 break;
  86                         }
  87                 }
  88
  89                 btrfs_item_key_to_cpu(leaf, &key, slot);
  90                 if (key.objectid < block_group->key.objectid) {
  91                         if (key.objectid + key.offset > first_free)
  92                                 first_free = key.objectid + key.offset;
  93                         goto next;
  94                 }
  95
  96                 if (key.objectid >= block_group->key.objectid +
  97                     block_group->key.offset) {
  98                         break;
  99                 }
 100
 101                 if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
 102                         if (!found) {
 103                                 last = first_free;
 104                                 found = 1;
 105                         }
 106                         if (key.objectid > last) {
 107                                 hole_size = key.objectid - last;
 108                                 set_extent_dirty(free_space_cache, last,
 109                                                  last + hole_size - 1,
 110                                                  GFP_NOFS);
 111                         }
 112                         last = key.objectid + key.offset;
 113                 }
 114 next:
 115                 path->slots[0]++;
 116         }
 117
 118         if (!found)
 119                 last = first_free;
 120         if (block_group->key.objectid +
 121             block_group->key.offset > last) {
 122                 hole_size = block_group->key.objectid +
 123                         block_group->key.offset - last;
 124                 set_extent_dirty(free_space_cache, last,
 125                                  last + hole_size - 1, GFP_NOFS);
 126         }
 127         block_group->cached = 1;
 128 err:
 129         btrfs_free_path(path);
 130         return 0;
 131 }
 132
 133 struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 134                                                          btrfs_fs_info *info,
 135                                                          u64 bytenr)
 136 {
 137         struct extent_map_tree *block_group_cache;
 138         struct btrfs_block_group_cache *block_group = NULL;
 139         u64 ptr;
 140         u64 start;
 141         u64 end;
 142         int ret;
 143
 144         block_group_cache = &info->block_group_cache;
 145         ret = find_first_extent_bit(block_group_cache,
 146                                     bytenr, &start, &end,
 147                                     BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA);
 148         if (ret) {
 149                 return NULL;
 150         }
 151         ret = get_state_private(block_group_cache, start, &ptr);
 152         if (ret)
 153                 return NULL;
 154
 155         block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr;
 156         if (block_group->key.objectid <= bytenr && bytenr <
 157             block_group->key.objectid + block_group->key.offset)
 158                 return block_group;
 159         return NULL;
 160 }
 161 static u64 find_search_start(struct btrfs_root *root,
 162                              struct btrfs_block_group_cache **cache_ret,
 163                              u64 search_start, int num,
 164                              int data, int full_scan)
 165 {
 166         int ret;
 167         struct btrfs_block_group_cache *cache = *cache_ret;
 168         u64 last;
 169         u64 start = 0;
 170         u64 end = 0;
 171         u64 cache_miss = 0;
 172         int wrapped = 0;
 173
 174         if (!cache) {
 175                 goto out;
 176         }
 177 again:
 178         ret = cache_block_group(root, cache);
 179         if (ret)
 180                 goto out;
 181
 182         last = max(search_start, cache->key.objectid);
 183
 184         while(1) {
 185                 ret = find_first_extent_bit(&root->fs_info->free_space_cache,
 186                                             last, &start, &end, EXTENT_DIRTY);
 187                 if (ret) {
 188                         if (!cache_miss)
 189                                 cache_miss = last;
 190                         goto new_group;
 191                 }
 192
 193                 start = max(last, start);
 194                 last = end + 1;
 195                 if (last - start < num) {
 196                         if (last == cache->key.objectid + cache->key.offset)
 197                                 cache_miss = start;
 198                         continue;
 199                 }
 200                 if (data != BTRFS_BLOCK_GROUP_MIXED &&
 201                     start + num > cache->key.objectid + cache->key.offset)
 202                         goto new_group;
 203                 return start;
 204         }
 205 out:
 206         cache = btrfs_lookup_block_group(root->fs_info, search_start);
 207         if (!cache) {
 208                 printk("Unable to find block group for %Lu\n",
 209                        search_start);
 210                 WARN_ON(1);
 211                 return search_start;
 212         }
 213         return search_start;
 214
 215 new_group:
 216         last = cache->key.objectid + cache->key.offset;
 217 wrapped:
 218         cache = btrfs_lookup_block_group(root->fs_info, last);
 219         if (!cache) {
 220 no_cache:
 221                 if (!wrapped) {
 222                         wrapped = 1;
 223                         last = search_start;
 224                         data = BTRFS_BLOCK_GROUP_MIXED;
 225                         goto wrapped;
 226                 }
 227                 goto out;
 228         }
 229         if (cache_miss && !cache->cached) {
 230                 cache_block_group(root, cache);
 231                 last = cache_miss;
 232                 cache = btrfs_lookup_block_group(root->fs_info, last);
 233         }
 234         cache = btrfs_find_block_group(root, cache, last, data, 0);
 235         if (!cache)
 236                 goto no_cache;
 237         *cache_ret = cache;
 238         cache_miss = 0;
 239         goto again;
 240 }
 241
 242 static u64 div_factor(u64 num, int factor)
 243 {
 244         if (factor == 10)
 245                 return num;
 246         num *= factor;
 247         do_div(num, 10);
 248         return num;
 249 }
 250
 251 struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
 252                                                  struct btrfs_block_group_cache
 253                                                  *hint, u64 search_start,
 254                                                  int data, int owner)
 255 {
 256         struct btrfs_block_group_cache *cache;
 257         struct extent_map_tree *block_group_cache;
 258         struct btrfs_block_group_cache *found_group = NULL;
 259         struct btrfs_fs_info *info = root->fs_info;
 260         u64 used;
 261         u64 last = 0;
 262         u64 hint_last;
 263         u64 start;
 264         u64 end;
 265         u64 free_check;
 266         u64 ptr;
 267         int bit;
 268         int ret;
 269         int full_search = 0;
 270         int factor = 8;
 271         int data_swap = 0;
 272
 273         block_group_cache = &info->block_group_cache;
 274
 275         if (!owner)
 276                 factor = 8;
 277
 278         if (data == BTRFS_BLOCK_GROUP_MIXED) {
 279                 bit = BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA;
 280                 factor = 10;
 281         } else if (data)
 282                 bit = BLOCK_GROUP_DATA;
 283         else
 284                 bit = BLOCK_GROUP_METADATA;
 285
 286         if (search_start) {
 287                 struct btrfs_block_group_cache *shint;
 288                 shint = btrfs_lookup_block_group(info, search_start);
 289                 if (shint && (shint->data == data ||
 290                               shint->data == BTRFS_BLOCK_GROUP_MIXED)) {
 291                         used = btrfs_block_group_used(&shint->item);
 292                         if (used + shint->pinned <
 293                             div_factor(shint->key.offset, factor)) {
 294                                 return shint;
 295                         }
 296                 }
 297         }
 298         if (hint && (hint->data == data ||
 299                      hint->data == BTRFS_BLOCK_GROUP_MIXED)) {
 300                 used = btrfs_block_group_used(&hint->item);
 301                 if (used + hint->pinned <
 302                     div_factor(hint->key.offset, factor)) {
 303                         return hint;
 304                 }
 305                 last = hint->key.objectid + hint->key.offset;
 306                 hint_last = last;
 307         } else {
 308                 if (hint)
 309                         hint_last = max(hint->key.objectid, search_start);
 310                 else
 311                         hint_last = search_start;
 312
 313                 last = hint_last;
 314         }
 315 again:
 316         while(1) {
 317                 ret = find_first_extent_bit(block_group_cache, last,
 318                                             &start, &end, bit);
 319                 if (ret)
 320                         break;
 321
 322                 ret = get_state_private(block_group_cache, start, &ptr);
 323                 if (ret)
 324                         break;
 325
 326                 cache = (struct btrfs_block_group_cache *)(unsigned long)ptr;
 327                 last = cache->key.objectid + cache->key.offset;
 328                 used = btrfs_block_group_used(&cache->item);
 329
 330                 if (full_search)
 331                         free_check = cache->key.offset;
 332                 else
 333                         free_check = div_factor(cache->key.offset, factor);
 334                 if (used + cache->pinned < free_check) {
 335                         found_group = cache;
 336                         goto found;
 337                 }
 338                 cond_resched();
 339         }
 340         if (!full_search) {
 341                 last = search_start;
 342                 full_search = 1;
 343                 goto again;
 344         }
 345         if (!data_swap) {
 346                 data_swap = 1;
 347                 bit = BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA;
 348                 last = search_start;
 349                 goto again;
 350         }
 351 found:
 352         return found_group;
 353 }
 354
 355 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 356                                 struct btrfs_root *root,
 357                                 u64 bytenr, u64 num_bytes)
 358 {
 359         struct btrfs_path *path;
 360         int ret;
 361         struct btrfs_key key;
 362         struct extent_buffer *l;
 363         struct btrfs_extent_item *item;
 364         u32 refs;
 365
 366         WARN_ON(num_bytes < root->sectorsize);
 367         path = btrfs_alloc_path();
 368         if (!path)
 369                 return -ENOMEM;
 370
 371         key.objectid = bytenr;
 372         btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 373         key.offset = num_bytes;
 374         ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
 375                                 0, 1);
 376         if (ret < 0)
 377                 return ret;
 378         if (ret != 0) {
 379                 BUG();
 380         }
 381         BUG_ON(ret != 0);
 382         l = path->nodes[0];
 383         item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
 384         refs = btrfs_extent_refs(l, item);
 385         btrfs_set_extent_refs(l, item, refs + 1);
 386         btrfs_mark_buffer_dirty(path->nodes[0]);
 387
 388         btrfs_release_path(root->fs_info->extent_root, path);
 389         btrfs_free_path(path);
 390         finish_current_insert(trans, root->fs_info->extent_root);
 391         del_pending_extents(trans, root->fs_info->extent_root);
 392         return 0;
 393 }
 394
 395 int btrfs_extent_post_op(struct btrfs_trans_handle *trans,
 396                          struct btrfs_root *root)
 397 {
 398         finish_current_insert(trans, root->fs_info->extent_root);
 399         del_pending_extents(trans, root->fs_info->extent_root);
 400         return 0;
 401 }
 402
 403 static int lookup_extent_ref(struct btrfs_trans_handle *trans,
 404                              struct btrfs_root *root, u64 bytenr,
 405                              u64 num_bytes, u32 *refs)
 406 {
 407         struct btrfs_path *path;
 408         int ret;
 409         struct btrfs_key key;
 410         struct extent_buffer *l;
 411         struct btrfs_extent_item *item;
 412
 413         WARN_ON(num_bytes < root->sectorsize);
 414         path = btrfs_alloc_path();
 415         key.objectid = bytenr;
 416         key.offset = num_bytes;
 417         btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 418         ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
 419                                 0, 0);
 420         if (ret < 0)
 421                 goto out;
 422         if (ret != 0) {
 423                 btrfs_print_leaf(root, path->nodes[0]);
 424                 printk("failed to find block number %Lu\n", bytenr);
 425                 BUG();
 426         }
 427         l = path->nodes[0];
 428         item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
 429         *refs = btrfs_extent_refs(l, item);
 430 out:
 431         btrfs_free_path(path);
 432         return 0;
 433 }
 434
 435 int btrfs_inc_root_ref(struct btrfs_trans_handle *trans,
 436                        struct btrfs_root *root)
 437 {
 438         return btrfs_inc_extent_ref(trans, root, root->node->start,
 439                                     root->node->len);
 440 }
 441
 442 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 443                   struct extent_buffer *buf)
 444 {
 445         u64 bytenr;
 446         u32 nritems;
 447         struct btrfs_key key;
 448         struct btrfs_file_extent_item *fi;
 449         int i;
 450         int level;
 451         int ret;
 452         int faili;
 453         int err;
 454
 455         if (!root->ref_cows)
 456                 return 0;
 457
 458         level = btrfs_header_level(buf);
 459         nritems = btrfs_header_nritems(buf);
 460         for (i = 0; i < nritems; i++) {
 461                 if (level == 0) {
 462                         u64 disk_bytenr;
 463                         btrfs_item_key_to_cpu(buf, &key, i);
 464                         if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 465                                 continue;
 466                         fi = btrfs_item_ptr(buf, i,
 467                                             struct btrfs_file_extent_item);
 468                         if (btrfs_file_extent_type(buf, fi) ==
 469                             BTRFS_FILE_EXTENT_INLINE)
 470                                 continue;
 471                         disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
 472                         if (disk_bytenr == 0)
 473                                 continue;
 474                         ret = btrfs_inc_extent_ref(trans, root, disk_bytenr,
 475                                     btrfs_file_extent_disk_num_bytes(buf, fi));
 476                         if (ret) {
 477                                 faili = i;
 478                                 goto fail;
 479                         }
 480                 } else {
 481                         bytenr = btrfs_node_blockptr(buf, i);
 482                         ret = btrfs_inc_extent_ref(trans, root, bytenr,
 483                                            btrfs_level_size(root, level - 1));
 484                         if (ret) {
 485                                 faili = i;
 486                                 goto fail;
 487                         }
 488                 }
 489         }
 490         return 0;
 491 fail:
 492         WARN_ON(1);
 493         for (i =0; i < faili; i++) {
 494                 if (level == 0) {
 495                         u64 disk_bytenr;
 496                         btrfs_item_key_to_cpu(buf, &key, i);
 497                         if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 498                                 continue;
 499                         fi = btrfs_item_ptr(buf, i,
 500                                             struct btrfs_file_extent_item);
 501                         if (btrfs_file_extent_type(buf, fi) ==
 502                             BTRFS_FILE_EXTENT_INLINE)
 503                                 continue;
 504                         disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
 505                         if (disk_bytenr == 0)
 506                                 continue;
 507                         err = btrfs_free_extent(trans, root, disk_bytenr,
 508                                     btrfs_file_extent_disk_num_bytes(buf,
 509                                                                       fi), 0);
 510                         BUG_ON(err);
 511                 } else {
 512                         bytenr = btrfs_node_blockptr(buf, i);
 513                         err = btrfs_free_extent(trans, root, bytenr,
 514                                         btrfs_level_size(root, level - 1), 0);
 515                         BUG_ON(err);
 516                 }
 517         }
 518         return ret;
 519 }
 520
 521 static int write_one_cache_group(struct btrfs_trans_handle *trans,
 522                                  struct btrfs_root *root,
 523                                  struct btrfs_path *path,
 524                                  struct btrfs_block_group_cache *cache)
 525 {
 526         int ret;
 527         int pending_ret;
 528         struct btrfs_root *extent_root = root->fs_info->extent_root;
 529         unsigned long bi;
 530         struct extent_buffer *leaf;
 531
 532         ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
 533         if (ret < 0)
 534                 goto fail;
 535         BUG_ON(ret);
 536
 537         leaf = path->nodes[0];
 538         bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
 539         write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
 540         btrfs_mark_buffer_dirty(leaf);
 541         btrfs_release_path(extent_root, path);
 542 fail:
 543         finish_current_insert(trans, extent_root);
 544         pending_ret = del_pending_extents(trans, extent_root);
 545         if (ret)
 546                 return ret;
 547         if (pending_ret)
 548                 return pending_ret;
 549         return 0;
 550
 551 }
 552
 553 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 554                                    struct btrfs_root *root)
 555 {
 556         struct extent_map_tree *block_group_cache;
 557         struct btrfs_block_group_cache *cache;
 558         int ret;
 559         int err = 0;
 560         int werr = 0;
 561         struct btrfs_path *path;
 562         u64 last = 0;
 563         u64 start;
 564         u64 end;
 565         u64 ptr;
 566
 567         block_group_cache = &root->fs_info->block_group_cache;
 568         path = btrfs_alloc_path();
 569         if (!path)
 570                 return -ENOMEM;
 571
 572         while(1) {
 573                 ret = find_first_extent_bit(block_group_cache, last,
 574                                             &start, &end, BLOCK_GROUP_DIRTY);
 575                 if (ret)
 576                         break;
 577
 578                 last = end + 1;
 579                 ret = get_state_private(block_group_cache, start, &ptr);
 580                 if (ret)
 581                         break;
 582
 583                 cache = (struct btrfs_block_group_cache *)(unsigned long)ptr;
 584                 err = write_one_cache_group(trans, root,
 585                                             path, cache);
 586                 /*
 587                  * if we fail to write the cache group, we want
 588                  * to keep it marked dirty in hopes that a later
 589                  * write will work
 590                  */
 591                 if (err) {
 592                         werr = err;
 593                         continue;
 594                 }
 595                 clear_extent_bits(block_group_cache, start, end,
 596                                   BLOCK_GROUP_DIRTY, GFP_NOFS);
 597         }
 598         btrfs_free_path(path);
 599         return werr;
 600 }
 601
 602 static int update_block_group(struct btrfs_trans_handle *trans,
 603                               struct btrfs_root *root,
 604                               u64 bytenr, u64 num_bytes, int alloc,
 605                               int mark_free, int data)
 606 {
 607         struct btrfs_block_group_cache *cache;
 608         struct btrfs_fs_info *info = root->fs_info;
 609         u64 total = num_bytes;
 610         u64 old_val;
 611         u64 byte_in_group;
 612         u64 start;
 613         u64 end;
 614
 615         while(total) {
 616                 cache = btrfs_lookup_block_group(info, bytenr);
 617                 if (!cache) {
 618                         return -1;
 619                 }
 620                 byte_in_group = bytenr - cache->key.objectid;
 621                 WARN_ON(byte_in_group > cache->key.offset);
 622                 start = cache->key.objectid;
 623                 end = start + cache->key.offset - 1;
 624                 set_extent_bits(&info->block_group_cache, start, end,
 625                                 BLOCK_GROUP_DIRTY, GFP_NOFS);
 626
 627                 old_val = btrfs_block_group_used(&cache->item);
 628                 num_bytes = min(total, cache->key.offset - byte_in_group);
 629                 if (alloc) {
 630                         if (cache->data != data &&
 631                             old_val < (cache->key.offset >> 1)) {
 632                                 int bit_to_clear;
 633                                 int bit_to_set;
 634                                 cache->data = data;
 635                                 if (data) {
 636                                         bit_to_clear = BLOCK_GROUP_METADATA;
 637                                         bit_to_set = BLOCK_GROUP_DATA;
 638                                         cache->item.flags &=
 639                                                 ~BTRFS_BLOCK_GROUP_MIXED;
 640                                         cache->item.flags |=
 641                                                 BTRFS_BLOCK_GROUP_DATA;
 642                                 } else {
 643                                         bit_to_clear = BLOCK_GROUP_DATA;
 644                                         bit_to_set = BLOCK_GROUP_METADATA;
 645                                         cache->item.flags &=
 646                                                 ~BTRFS_BLOCK_GROUP_MIXED;
 647                                         cache->item.flags &=
 648                                                 ~BTRFS_BLOCK_GROUP_DATA;
 649                                 }
 650                                 clear_extent_bits(&info->block_group_cache,
 651                                                   start, end, bit_to_clear,
 652                                                   GFP_NOFS);
 653                                 set_extent_bits(&info->block_group_cache,
 654                                                 start, end, bit_to_set,
 655                                                 GFP_NOFS);
 656                         } else if (cache->data != data &&
 657                                    cache->data != BTRFS_BLOCK_GROUP_MIXED) {
 658                                 cache->data = BTRFS_BLOCK_GROUP_MIXED;
 659                                 set_extent_bits(&info->block_group_cache,
 660                                                 start, end,
 661                                                 BLOCK_GROUP_DATA |
 662                                                 BLOCK_GROUP_METADATA,
 663                                                 GFP_NOFS);
 664                         }
 665                         old_val += num_bytes;
 666                 } else {
 667                         old_val -= num_bytes;
 668                         if (mark_free) {
 669                                 set_extent_dirty(&info->free_space_cache,
 670                                                  bytenr, bytenr + num_bytes - 1,
 671                                                  GFP_NOFS);
 672                         }
 673                 }
 674                 btrfs_set_block_group_used(&cache->item, old_val);
 675                 total -= num_bytes;
 676                 bytenr += num_bytes;
 677         }
 678         return 0;
 679 }
 680 static int update_pinned_extents(struct btrfs_root *root,
 681                                 u64 bytenr, u64 num, int pin)
 682 {
 683         u64 len;
 684         struct btrfs_block_group_cache *cache;
 685         struct btrfs_fs_info *fs_info = root->fs_info;
 686
 687         if (pin) {
 688                 set_extent_dirty(&fs_info->pinned_extents,
 689                                 bytenr, bytenr + num - 1, GFP_NOFS);
 690         } else {
 691                 clear_extent_dirty(&fs_info->pinned_extents,
 692                                 bytenr, bytenr + num - 1, GFP_NOFS);
 693         }
 694         while (num > 0) {
 695                 cache = btrfs_lookup_block_group(fs_info, bytenr);
 696                 WARN_ON(!cache);
 697                 len = min(num, cache->key.offset -
 698                           (bytenr - cache->key.objectid));
 699                 if (pin) {
 700                         cache->pinned += len;
 701                         fs_info->total_pinned += len;
 702                 } else {
 703                         cache->pinned -= len;
 704                         fs_info->total_pinned -= len;
 705                 }
 706                 bytenr += len;
 707                 num -= len;
 708         }
 709         return 0;
 710 }
 711
 712 int btrfs_copy_pinned(struct btrfs_root *root, struct extent_map_tree *copy)
 713 {
 714         u64 last = 0;
 715         u64 start;
 716         u64 end;
 717         struct extent_map_tree *pinned_extents = &root->fs_info->pinned_extents;
 718         int ret;
 719
 720         while(1) {
 721                 ret = find_first_extent_bit(pinned_extents, last,
 722                                             &start, &end, EXTENT_DIRTY);
 723                 if (ret)
 724                         break;
 725                 set_extent_dirty(copy, start, end, GFP_NOFS);
 726                 last = end + 1;
 727         }
 728         return 0;
 729 }
 730
 731 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
 732                                struct btrfs_root *root,
 733                                struct extent_map_tree *unpin)
 734 {
 735         u64 start;
 736         u64 end;
 737         int ret;
 738         struct extent_map_tree *free_space_cache;
 739         free_space_cache = &root->fs_info->free_space_cache;
 740
 741         while(1) {
 742                 ret = find_first_extent_bit(unpin, 0, &start, &end,
 743                                             EXTENT_DIRTY);
 744                 if (ret)
 745                         break;
 746                 update_pinned_extents(root, start, end + 1 - start, 0);
 747                 clear_extent_dirty(unpin, start, end, GFP_NOFS);
 748                 set_extent_dirty(free_space_cache, start, end, GFP_NOFS);
 749         }
 750         return 0;
 751 }
 752
 753 static int finish_current_insert(struct btrfs_trans_handle *trans, struct
 754                                  btrfs_root *extent_root)
 755 {
 756         struct btrfs_key ins;
 757         struct btrfs_extent_item extent_item;
 758         int ret;
 759         int err = 0;
 760         u64 start;
 761         u64 end;
 762         struct btrfs_fs_info *info = extent_root->fs_info;
 763
 764         btrfs_set_stack_extent_refs(&extent_item, 1);
 765         btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY);
 766         btrfs_set_stack_extent_owner(&extent_item,
 767                                      extent_root->root_key.objectid);
 768
 769         while(1) {
 770                 ret = find_first_extent_bit(&info->extent_ins, 0, &start,
 771                                             &end, EXTENT_LOCKED);
 772                 if (ret)
 773                         break;
 774
 775                 ins.objectid = start;
 776                 ins.offset = end + 1 - start;
 777                 err = btrfs_insert_item(trans, extent_root, &ins,
 778                                         &extent_item, sizeof(extent_item));
 779                 clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED,
 780                                   GFP_NOFS);
 781         }
 782         return 0;
 783 }
 784
 785 static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes,
 786                           int pending)
 787 {
 788         int err = 0;
 789         struct extent_buffer *buf;
 790
 791         if (!pending) {
 792                 buf = btrfs_find_tree_block(root, bytenr, num_bytes);
 793                 if (buf) {
 794                         if (btrfs_buffer_uptodate(buf)) {
 795                                 u64 transid =
 796                                     root->fs_info->running_transaction->transid;
 797                                 if (btrfs_header_generation(buf) == transid) {
 798                                         free_extent_buffer(buf);
 799                                         return 1;
 800                                 }
 801                         }
 802                         free_extent_buffer(buf);
 803                 }
 804                 update_pinned_extents(root, bytenr, num_bytes, 1);
 805         } else {
 806                 set_extent_bits(&root->fs_info->pending_del,
 807                                 bytenr, bytenr + num_bytes - 1,
 808                                 EXTENT_LOCKED, GFP_NOFS);
 809         }
 810         BUG_ON(err < 0);
 811         return 0;
 812 }
 813
 814 /*
 815  * remove an extent from the root, returns 0 on success
 816  */
 817 static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 818                          *root, u64 bytenr, u64 num_bytes, int pin,
 819                          int mark_free)
 820 {
 821         struct btrfs_path *path;
 822         struct btrfs_key key;
 823         struct btrfs_fs_info *info = root->fs_info;
 824         struct btrfs_root *extent_root = info->extent_root;
 825         struct extent_buffer *leaf;
 826         int ret;
 827         struct btrfs_extent_item *ei;
 828         u32 refs;
 829
 830         key.objectid = bytenr;
 831         btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 832         key.offset = num_bytes;
 833
 834         path = btrfs_alloc_path();
 835         if (!path)
 836                 return -ENOMEM;
 837
 838         ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1);
 839         if (ret < 0)
 840                 return ret;
 841         BUG_ON(ret);
 842
 843         leaf = path->nodes[0];
 844         ei = btrfs_item_ptr(leaf, path->slots[0],
 845                             struct btrfs_extent_item);
 846         refs = btrfs_extent_refs(leaf, ei);
 847         BUG_ON(refs == 0);
 848         refs -= 1;
 849         btrfs_set_extent_refs(leaf, ei, refs);
 850         btrfs_mark_buffer_dirty(leaf);
 851
 852         if (refs == 0) {
 853                 u64 super_used;
 854                 u64 root_used;
 855
 856                 if (pin) {
 857                         ret = pin_down_bytes(root, bytenr, num_bytes, 0);
 858                         if (ret > 0)
 859                                 mark_free = 1;
 860                         BUG_ON(ret < 0);
 861                 }
 862
 863                 /* block accounting for super block */
 864                 super_used = btrfs_super_bytes_used(&info->super_copy);
 865                 btrfs_set_super_bytes_used(&info->super_copy,
 866                                            super_used - num_bytes);
 867
 868                 /* block accounting for root item */
 869                 root_used = btrfs_root_used(&root->root_item);
 870                 btrfs_set_root_used(&root->root_item,
 871                                            root_used - num_bytes);
 872
 873                 ret = btrfs_del_item(trans, extent_root, path);
 874                 if (ret) {
 875                         return ret;
 876                 }
 877                 ret = update_block_group(trans, root, bytenr, num_bytes, 0,
 878                                          mark_free, 0);
 879                 BUG_ON(ret);
 880         }
 881         btrfs_free_path(path);
 882         finish_current_insert(trans, extent_root);
 883         return ret;
 884 }
 885
 886 /*
 887  * find all the blocks marked as pending in the radix tree and remove
 888  * them from the extent map
 889  */
 890 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
 891                                btrfs_root *extent_root)
 892 {
 893         int ret;
 894         int err = 0;
 895         u64 start;
 896         u64 end;
 897         struct extent_map_tree *pending_del;
 898         struct extent_map_tree *pinned_extents;
 899
 900         pending_del = &extent_root->fs_info->pending_del;
 901         pinned_extents = &extent_root->fs_info->pinned_extents;
 902
 903         while(1) {
 904                 ret = find_first_extent_bit(pending_del, 0, &start, &end,
 905                                             EXTENT_LOCKED);
 906                 if (ret)
 907                         break;
 908                 update_pinned_extents(extent_root, start, end + 1 - start, 1);
 909                 clear_extent_bits(pending_del, start, end, EXTENT_LOCKED,
 910                                   GFP_NOFS);
 911                 ret = __free_extent(trans, extent_root,
 912                                      start, end + 1 - start, 0, 0);
 913                 if (ret)
 914                         err = ret;
 915         }
 916         return err;
 917 }
 918
 919 /*
 920  * remove an extent from the root, returns 0 on success
 921  */
 922 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 923                       *root, u64 bytenr, u64 num_bytes, int pin)
 924 {
 925         struct btrfs_root *extent_root = root->fs_info->extent_root;
 926         int pending_ret;
 927         int ret;
 928
 929         WARN_ON(num_bytes < root->sectorsize);
 930         if (root == extent_root) {
 931                 pin_down_bytes(root, bytenr, num_bytes, 1);
 932                 return 0;
 933         }
 934         ret = __free_extent(trans, root, bytenr, num_bytes, pin, pin == 0);
 935         pending_ret = del_pending_extents(trans, root->fs_info->extent_root);
 936         return ret ? ret : pending_ret;
 937 }
 938
 939 static u64 stripe_align(struct btrfs_root *root, u64 val)
 940 {
 941         u64 mask = ((u64)root->stripesize - 1);
 942         u64 ret = (val + mask) & ~mask;
 943         return ret;
 944 }
 945
 946 /*
 947  * walks the btree of allocated extents and find a hole of a given size.
 948  * The key ins is changed to record the hole:
 949  * ins->objectid == block start
 950  * ins->flags = BTRFS_EXTENT_ITEM_KEY
 951  * ins->offset == number of blocks
 952  * Any available blocks before search_start are skipped.
 953  */
 954 static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 955                             *orig_root, u64 num_bytes, u64 empty_size,
 956                             u64 search_start, u64 search_end, u64 hint_byte,
 957                             struct btrfs_key *ins, u64 exclude_start,
 958                             u64 exclude_nr, int data)
 959 {
 960         struct btrfs_path *path;
 961         struct btrfs_key key;
 962         u64 hole_size = 0;
 963         u64 aligned;
 964         int ret;
 965         int slot = 0;
 966         u64 last_byte = 0;
 967         u64 orig_search_start = search_start;
 968         int start_found;
 969         struct extent_buffer *l;
 970         struct btrfs_root * root = orig_root->fs_info->extent_root;
 971         struct btrfs_fs_info *info = root->fs_info;
 972         u64 total_needed = num_bytes;
 973         int level;
 974         struct btrfs_block_group_cache *block_group;
 975         int full_scan = 0;
 976         int wrapped = 0;
 977         u64 cached_start;
 978
 979         WARN_ON(num_bytes < root->sectorsize);
 980         btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
 981
 982         level = btrfs_header_level(root->node);
 983
 984         if (num_bytes >= 32 * 1024 * 1024 && hint_byte) {
 985                 data = BTRFS_BLOCK_GROUP_MIXED;
 986         }
 987
 988         if (search_end == (u64)-1)
 989                 search_end = btrfs_super_total_bytes(&info->super_copy);
 990         if (hint_byte) {
 991                 block_group = btrfs_lookup_block_group(info, hint_byte);
 992                 if (!block_group)
 993                         hint_byte = search_start;
 994                 block_group = btrfs_find_block_group(root, block_group,
 995                                                      hint_byte, data, 1);
 996         } else {
 997                 block_group = btrfs_find_block_group(root,
 998                                                      trans->block_group,
 999                                                      search_start, data, 1);
1000         }
1001
1002         total_needed += empty_size;
1003         path = btrfs_alloc_path();
1004 check_failed:
1005         search_start = find_search_start(root, &block_group, search_start,
1006                                          total_needed, data, full_scan);
1007         search_start = stripe_align(root, search_start);
1008         cached_start = search_start;
1009         btrfs_init_path(path);
1010         ins->objectid = search_start;
1011         ins->offset = 0;
1012         start_found = 0;
1013         path->reada = 2;
1014
1015         ret = btrfs_search_slot(trans, root, ins, path, 0, 0);
1016         if (ret < 0)
1017                 goto error;
1018
1019         if (path->slots[0] > 0) {
1020                 path->slots[0]--;
1021         }
1022
1023         l = path->nodes[0];
1024         btrfs_item_key_to_cpu(l, &key, path->slots[0]);
1025
1026         /*
1027          * a rare case, go back one key if we hit a block group item
1028          * instead of an extent item
1029          */
1030         if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY &&
1031             key.objectid + key.offset >= search_start) {
1032                 ins->objectid = key.objectid;
1033                 ins->offset = key.offset - 1;
1034                 btrfs_release_path(root, path);
1035                 ret = btrfs_search_slot(trans, root, ins, path, 0, 0);
1036                 if (ret < 0)
1037                         goto error;
1038
1039                 if (path->slots[0] > 0) {
1040                         path->slots[0]--;
1041                 }
1042         }
1043
1044         while (1) {
1045                 l = path->nodes[0];
1046                 slot = path->slots[0];
1047                 if (slot >= btrfs_header_nritems(l)) {
1048                         ret = btrfs_next_leaf(root, path);
1049                         if (ret == 0)
1050                                 continue;
1051                         if (ret < 0)
1052                                 goto error;
1053
1054                         search_start = max(search_start,
1055                                            block_group->key.objectid);
1056                         if (!start_found) {
1057                                 aligned = stripe_align(root, search_start);
1058                                 ins->objectid = aligned;
1059                                 if (aligned >= search_end) {
1060                                         ret = -ENOSPC;
1061                                         goto error;
1062                                 }
1063                                 ins->offset = search_end - aligned;
1064                                 start_found = 1;
1065                                 goto check_pending;
1066                         }
1067                         ins->objectid = stripe_align(root,
1068                                                      last_byte > search_start ?
1069                                                      last_byte : search_start);
1070                         if (search_end <= ins->objectid) {
1071                                 ret = -ENOSPC;
1072                                 goto error;
1073                         }
1074                         ins->offset = search_end - ins->objectid;
1075                         BUG_ON(ins->objectid >= search_end);
1076                         goto check_pending;
1077                 }
1078                 btrfs_item_key_to_cpu(l, &key, slot);
1079
1080                 if (key.objectid >= search_start && key.objectid > last_byte &&
1081                     start_found) {
1082                         if (last_byte < search_start)
1083                                 last_byte = search_start;
1084                         aligned = stripe_align(root, last_byte);
1085                         hole_size = key.objectid - aligned;
1086                         if (key.objectid > aligned && hole_size >= num_bytes) {
1087                                 ins->objectid = aligned;
1088                                 ins->offset = hole_size;
1089                                 goto check_pending;
1090                         }
1091                 }
1092                 if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY) {
1093                         if (!start_found) {
1094                                 last_byte = key.objectid;
1095                                 start_found = 1;
1096                         }
1097                         goto next;
1098                 }
1099
1100
1101                 start_found = 1;
1102                 last_byte = key.objectid + key.offset;
1103
1104                 if (!full_scan && data != BTRFS_BLOCK_GROUP_MIXED &&
1105                     last_byte >= block_group->key.objectid +
1106                     block_group->key.offset) {
1107                         btrfs_release_path(root, path);
1108                         search_start = block_group->key.objectid +
1109                                 block_group->key.offset;
1110                         goto new_group;
1111                 }
1112 next:
1113                 path->slots[0]++;
1114                 cond_resched();
1115         }
1116 check_pending:
1117         /* we have to make sure we didn't find an extent that has already
1118          * been allocated by the map tree or the original allocation
1119          */
1120         btrfs_release_path(root, path);
1121         BUG_ON(ins->objectid < search_start);
1122
1123         if (ins->objectid + num_bytes >= search_end)
1124                 goto enospc;
1125         if (!full_scan && data != BTRFS_BLOCK_GROUP_MIXED &&
1126             ins->objectid + num_bytes > block_group->
1127             key.objectid + block_group->key.offset) {
1128                 search_start = block_group->key.objectid +
1129                         block_group->key.offset;
1130                 goto new_group;
1131         }
1132         if (test_range_bit(&info->extent_ins, ins->objectid,
1133                            ins->objectid + num_bytes -1, EXTENT_LOCKED, 0)) {
1134                 search_start = ins->objectid + num_bytes;
1135                 goto new_group;
1136         }
1137         if (test_range_bit(&info->pinned_extents, ins->objectid,
1138                            ins->objectid + num_bytes -1, EXTENT_DIRTY, 0)) {
1139                 search_start = ins->objectid + num_bytes;
1140                 goto new_group;
1141         }
1142         if (exclude_nr > 0 && (ins->objectid + num_bytes > exclude_start &&
1143             ins->objectid < exclude_start + exclude_nr)) {
1144                 search_start = exclude_start + exclude_nr;
1145                 goto new_group;
1146         }
1147         if (!data) {
1148                 block_group = btrfs_lookup_block_group(info, ins->objectid);
1149                 if (block_group)
1150                         trans->block_group = block_group;
1151         }
1152         ins->offset = num_bytes;
1153         btrfs_free_path(path);
1154         return 0;
1155
1156 new_group:
1157         if (search_start + num_bytes >= search_end) {
1158 enospc:
1159                 search_start = orig_search_start;
1160                 if (full_scan) {
1161                         ret = -ENOSPC;
1162                         goto error;
1163                 }
1164                 if (wrapped) {
1165                         if (!full_scan)
1166                                 total_needed -= empty_size;
1167                         full_scan = 1;
1168                         data = BTRFS_BLOCK_GROUP_MIXED;
1169                 } else
1170                         wrapped = 1;
1171         }
1172         block_group = btrfs_lookup_block_group(info, search_start);
1173         cond_resched();
1174         block_group = btrfs_find_block_group(root, block_group,
1175                                              search_start, data, 0);
1176         goto check_failed;
1177
1178 error:
1179         btrfs_release_path(root, path);
1180         btrfs_free_path(path);
1181         return ret;
1182 }
1183 /*
1184  * finds a free extent and does all the dirty work required for allocation
1185  * returns the key for the extent through ins, and a tree buffer for
1186  * the first block of the extent through buf.
1187  *
1188  * returns 0 if everything worked, non-zero otherwise.
1189  */
1190 int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
1191                        struct btrfs_root *root, u64 owner,
1192                        u64 num_bytes, u64 empty_size, u64 hint_byte,
1193                        u64 search_end, struct btrfs_key *ins, int data)
1194 {
1195         int ret;
1196         int pending_ret;
1197         u64 super_used, root_used;
1198         u64 search_start = 0;
1199         struct btrfs_fs_info *info = root->fs_info;
1200         struct btrfs_root *extent_root = info->extent_root;
1201         struct btrfs_extent_item extent_item;
1202
1203         btrfs_set_stack_extent_refs(&extent_item, 1);
1204         btrfs_set_stack_extent_owner(&extent_item, owner);
1205
1206         WARN_ON(num_bytes < root->sectorsize);
1207         ret = find_free_extent(trans, root, num_bytes, empty_size,
1208                                search_start, search_end, hint_byte, ins,
1209                                trans->alloc_exclude_start,
1210                                trans->alloc_exclude_nr, data);
1211         BUG_ON(ret);
1212         if (ret)
1213                 return ret;
1214
1215         /* block accounting for super block */
1216         super_used = btrfs_super_bytes_used(&info->super_copy);
1217         btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes);
1218
1219         /* block accounting for root item */
1220         root_used = btrfs_root_used(&root->root_item);
1221         btrfs_set_root_used(&root->root_item, root_used + num_bytes);
1222
1223         clear_extent_dirty(&root->fs_info->free_space_cache,
1224                            ins->objectid, ins->objectid + ins->offset - 1,
1225                            GFP_NOFS);
1226
1227         if (root == extent_root) {
1228                 set_extent_bits(&root->fs_info->extent_ins, ins->objectid,
1229                                 ins->objectid + ins->offset - 1,
1230                                 EXTENT_LOCKED, GFP_NOFS);
1231                 WARN_ON(data == 1);
1232                 goto update_block;
1233         }
1234
1235         WARN_ON(trans->alloc_exclude_nr);
1236         trans->alloc_exclude_start = ins->objectid;
1237         trans->alloc_exclude_nr = ins->offset;
1238         ret = btrfs_insert_item(trans, extent_root, ins, &extent_item,
1239                                 sizeof(extent_item));
1240
1241         trans->alloc_exclude_start = 0;
1242         trans->alloc_exclude_nr = 0;
1243
1244         BUG_ON(ret);
1245         finish_current_insert(trans, extent_root);
1246         pending_ret = del_pending_extents(trans, extent_root);
1247
1248         if (ret) {
1249                 return ret;
1250         }
1251         if (pending_ret) {
1252                 return pending_ret;
1253         }
1254
1255 update_block:
1256         ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0,
1257                                  data);
1258         BUG_ON(ret);
1259         return 0;
1260 }
1261
1262 /*
1263  * helper function to allocate a block for a given tree
1264  * returns the tree buffer or NULL.
1265  */
1266 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
1267                                              struct btrfs_root *root,
1268                                              u32 blocksize, u64 hint,
1269                                              u64 empty_size)
1270 {
1271         struct btrfs_key ins;
1272         int ret;
1273         struct extent_buffer *buf;
1274
1275         ret = btrfs_alloc_extent(trans, root, root->root_key.objectid,
1276                                  blocksize, empty_size, hint,
1277                                  (u64)-1, &ins, 0);
1278         if (ret) {
1279                 BUG_ON(ret > 0);
1280                 return ERR_PTR(ret);
1281         }
1282         buf = btrfs_find_create_tree_block(root, ins.objectid, blocksize);
1283         if (!buf) {
1284                 btrfs_free_extent(trans, root, ins.objectid, blocksize, 0);
1285                 return ERR_PTR(-ENOMEM);
1286         }
1287         btrfs_set_buffer_uptodate(buf);
1288         set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
1289                          buf->start + buf->len - 1, GFP_NOFS);
1290         set_extent_bits(&BTRFS_I(root->fs_info->btree_inode)->extent_tree,
1291                         buf->start, buf->start + buf->len - 1,
1292                         EXTENT_CSUM, GFP_NOFS);
1293         buf->flags |= EXTENT_CSUM;
1294         btrfs_set_buffer_defrag(buf);
1295         trans->blocks_used++;
1296         return buf;
1297 }
1298
1299 static int drop_leaf_ref(struct btrfs_trans_handle *trans,
1300                          struct btrfs_root *root, struct extent_buffer *leaf)
1301 {
1302         struct btrfs_key key;
1303         struct btrfs_file_extent_item *fi;
1304         int i;
1305         int nritems;
1306         int ret;
1307
1308         BUG_ON(!btrfs_is_leaf(leaf));
1309         nritems = btrfs_header_nritems(leaf);
1310         for (i = 0; i < nritems; i++) {
1311                 u64 disk_bytenr;
1312
1313                 btrfs_item_key_to_cpu(leaf, &key, i);
1314                 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
1315                         continue;
1316                 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
1317                 if (btrfs_file_extent_type(leaf, fi) ==
1318                     BTRFS_FILE_EXTENT_INLINE)
1319                         continue;
1320                 /*
1321                  * FIXME make sure to insert a trans record that
1322                  * repeats the snapshot del on crash
1323                  */
1324                 disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
1325                 if (disk_bytenr == 0)
1326                         continue;
1327                 ret = btrfs_free_extent(trans, root, disk_bytenr,
1328                                 btrfs_file_extent_disk_num_bytes(leaf, fi), 0);
1329                 BUG_ON(ret);
1330         }
1331         return 0;
1332 }
1333
1334 static void reada_walk_down(struct btrfs_root *root,
1335                             struct extent_buffer *node)
1336 {
1337         int i;
1338         u32 nritems;
1339         u64 bytenr;
1340         int ret;
1341         u32 refs;
1342         int level;
1343         u32 blocksize;
1344
1345         nritems = btrfs_header_nritems(node);
1346         level = btrfs_header_level(node);
1347         for (i = 0; i < nritems; i++) {
1348                 bytenr = btrfs_node_blockptr(node, i);
1349                 blocksize = btrfs_level_size(root, level - 1);
1350                 ret = lookup_extent_ref(NULL, root, bytenr, blocksize, &refs);
1351                 BUG_ON(ret);
1352                 if (refs != 1)
1353                         continue;
1354                 mutex_unlock(&root->fs_info->fs_mutex);
1355                 ret = readahead_tree_block(root, bytenr, blocksize);
1356                 cond_resched();
1357                 mutex_lock(&root->fs_info->fs_mutex);
1358                 if (ret)
1359                         break;
1360         }
1361 }
1362
1363 /*
1364  * helper function for drop_snapshot, this walks down the tree dropping ref
1365  * counts as it goes.
1366  */
1367 static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
1368                           *root, struct btrfs_path *path, int *level)
1369 {
1370         struct extent_buffer *next;
1371         struct extent_buffer *cur;
1372         u64 bytenr;
1373         u32 blocksize;
1374         int ret;
1375         u32 refs;
1376
1377         WARN_ON(*level < 0);
1378         WARN_ON(*level >= BTRFS_MAX_LEVEL);
1379         ret = lookup_extent_ref(trans, root,
1380                                 path->nodes[*level]->start,
1381                                 path->nodes[*level]->len, &refs);
1382         BUG_ON(ret);
1383         if (refs > 1)
1384                 goto out;
1385
1386         /*
1387          * walk down to the last node level and free all the leaves
1388          */
1389         while(*level >= 0) {
1390                 WARN_ON(*level < 0);
1391                 WARN_ON(*level >= BTRFS_MAX_LEVEL);
1392                 cur = path->nodes[*level];
1393
1394                 if (*level > 0 && path->slots[*level] == 0)
1395                         reada_walk_down(root, cur);
1396
1397                 if (btrfs_header_level(cur) != *level)
1398                         WARN_ON(1);
1399
1400                 if (path->slots[*level] >=
1401                     btrfs_header_nritems(cur))
1402                         break;
1403                 if (*level == 0) {
1404                         ret = drop_leaf_ref(trans, root, cur);
1405                         BUG_ON(ret);
1406                         break;
1407                 }
1408                 bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
1409                 blocksize = btrfs_level_size(root, *level - 1);
1410                 ret = lookup_extent_ref(trans, root, bytenr, blocksize, &refs);
1411                 BUG_ON(ret);
1412                 if (refs != 1) {
1413                         path->slots[*level]++;
1414                         ret = btrfs_free_extent(trans, root, bytenr,
1415                                                 blocksize, 1);
1416                         BUG_ON(ret);
1417                         continue;
1418                 }
1419                 next = btrfs_find_tree_block(root, bytenr, blocksize);
1420                 if (!next || !btrfs_buffer_uptodate(next)) {
1421                         free_extent_buffer(next);
1422                         mutex_unlock(&root->fs_info->fs_mutex);
1423                         next = read_tree_block(root, bytenr, blocksize);
1424                         mutex_lock(&root->fs_info->fs_mutex);
1425
1426                         /* we dropped the lock, check one more time */
1427                         ret = lookup_extent_ref(trans, root, bytenr,
1428                                                 blocksize, &refs);
1429                         BUG_ON(ret);
1430                         if (refs != 1) {
1431                                 path->slots[*level]++;
1432                                 free_extent_buffer(next);
1433                                 ret = btrfs_free_extent(trans, root,
1434                                                         bytenr, blocksize, 1);
1435                                 BUG_ON(ret);
1436                                 continue;
1437                         }
1438                 }
1439                 WARN_ON(*level <= 0);
1440                 if (path->nodes[*level-1])
1441                         free_extent_buffer(path->nodes[*level-1]);
1442                 path->nodes[*level-1] = next;
1443                 *level = btrfs_header_level(next);
1444                 path->slots[*level] = 0;
1445         }
1446 out:
1447         WARN_ON(*level < 0);
1448         WARN_ON(*level >= BTRFS_MAX_LEVEL);
1449         ret = btrfs_free_extent(trans, root, path->nodes[*level]->start,
1450                                 path->nodes[*level]->len, 1);
1451         free_extent_buffer(path->nodes[*level]);
1452         path->nodes[*level] = NULL;
1453         *level += 1;
1454         BUG_ON(ret);
1455         return 0;
1456 }
1457
1458 /*
1459  * helper for dropping snapshots.  This walks back up the tree in the path
1460  * to find the first node higher up where we haven't yet gone through
1461  * all the slots
1462  */
1463 static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root
1464                         *root, struct btrfs_path *path, int *level)
1465 {
1466         int i;
1467         int slot;
1468         int ret;
1469         struct btrfs_root_item *root_item = &root->root_item;
1470
1471         for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
1472                 slot = path->slots[i];
1473                 if (slot < btrfs_header_nritems(path->nodes[i]) - 1) {
1474                         struct extent_buffer *node;
1475                         struct btrfs_disk_key disk_key;
1476                         node = path->nodes[i];
1477                         path->slots[i]++;
1478                         *level = i;
1479                         WARN_ON(*level == 0);
1480                         btrfs_node_key(node, &disk_key, path->slots[i]);
1481                         memcpy(&root_item->drop_progress,
1482                                &disk_key, sizeof(disk_key));
1483                         root_item->drop_level = i;
1484                         return 0;
1485                 } else {
1486                         ret = btrfs_free_extent(trans, root,
1487                                                 path->nodes[*level]->start,
1488                                                 path->nodes[*level]->len, 1);
1489                         BUG_ON(ret);
1490                         free_extent_buffer(path->nodes[*level]);
1491                         path->nodes[*level] = NULL;
1492                         *level = i + 1;
1493                 }
1494         }
1495         return 1;
1496 }
1497
1498 /*
1499  * drop the reference count on the tree rooted at 'snap'.  This traverses
1500  * the tree freeing any blocks that have a ref count of zero after being
1501  * decremented.
1502  */
1503 int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
1504                         *root)
1505 {
1506         int ret = 0;
1507         int wret;
1508         int level;
1509         struct btrfs_path *path;
1510         int i;
1511         int orig_level;
1512         struct btrfs_root_item *root_item = &root->root_item;
1513
1514         path = btrfs_alloc_path();
1515         BUG_ON(!path);
1516
1517         level = btrfs_header_level(root->node);
1518         orig_level = level;
1519         if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
1520                 path->nodes[level] = root->node;
1521                 extent_buffer_get(root->node);
1522                 path->slots[level] = 0;
1523         } else {
1524                 struct btrfs_key key;
1525                 struct btrfs_disk_key found_key;
1526                 struct extent_buffer *node;
1527
1528                 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
1529                 level = root_item->drop_level;
1530                 path->lowest_level = level;
1531                 wret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1532                 if (wret < 0) {
1533                         ret = wret;
1534                         goto out;
1535                 }
1536                 node = path->nodes[level];
1537                 btrfs_node_key(node, &found_key, path->slots[level]);
1538                 WARN_ON(memcmp(&found_key, &root_item->drop_progress,
1539                                sizeof(found_key)));
1540         }
1541         while(1) {
1542                 wret = walk_down_tree(trans, root, path, &level);
1543                 if (wret > 0)
1544                         break;
1545                 if (wret < 0)
1546                         ret = wret;
1547
1548                 wret = walk_up_tree(trans, root, path, &level);
1549                 if (wret > 0)
1550                         break;
1551                 if (wret < 0)
1552                         ret = wret;
1553                 ret = -EAGAIN;
1554                 break;
1555         }
1556         for (i = 0; i <= orig_level; i++) {
1557                 if (path->nodes[i]) {
1558                         free_extent_buffer(path->nodes[i]);
1559                         path->nodes[i] = NULL;
1560                 }
1561         }
1562 out:
1563         btrfs_free_path(path);
1564         return ret;
1565 }
1566
1567 int btrfs_free_block_groups(struct btrfs_fs_info *info)
1568 {
1569         u64 start;
1570         u64 end;
1571         u64 ptr;
1572         int ret;
1573         while(1) {
1574                 ret = find_first_extent_bit(&info->block_group_cache, 0,
1575                                             &start, &end, (unsigned int)-1);
1576                 if (ret)
1577                         break;
1578                 ret = get_state_private(&info->block_group_cache, start, &ptr);
1579                 if (!ret)
1580                         kfree((void *)(unsigned long)ptr);
1581                 clear_extent_bits(&info->block_group_cache, start,
1582                                   end, (unsigned int)-1, GFP_NOFS);
1583         }
1584         while(1) {
1585                 ret = find_first_extent_bit(&info->free_space_cache, 0,
1586                                             &start, &end, EXTENT_DIRTY);
1587                 if (ret)
1588                         break;
1589                 clear_extent_dirty(&info->free_space_cache, start,
1590                                    end, GFP_NOFS);
1591         }
1592         return 0;
1593 }
1594
1595 int btrfs_read_block_groups(struct btrfs_root *root)
1596 {
1597         struct btrfs_path *path;
1598         int ret;
1599         int err = 0;
1600         int bit;
1601         struct btrfs_block_group_cache *cache;
1602         struct btrfs_fs_info *info = root->fs_info;
1603         struct extent_map_tree *block_group_cache;
1604         struct btrfs_key key;
1605         struct btrfs_key found_key;
1606         struct extent_buffer *leaf;
1607
1608         block_group_cache = &info->block_group_cache;
1609
1610         root = info->extent_root;
1611         key.objectid = 0;
1612         key.offset = BTRFS_BLOCK_GROUP_SIZE;
1613         btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
1614
1615         path = btrfs_alloc_path();
1616         if (!path)
1617                 return -ENOMEM;
1618
1619         while(1) {
1620                 ret = btrfs_search_slot(NULL, info->extent_root,
1621                                         &key, path, 0, 0);
1622                 if (ret != 0) {
1623                         err = ret;
1624                         break;
1625                 }
1626                 leaf = path->nodes[0];
1627                 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1628                 cache = kmalloc(sizeof(*cache), GFP_NOFS);
1629                 if (!cache) {
1630                         err = -1;
1631                         break;
1632                 }
1633
1634                 read_extent_buffer(leaf, &cache->item,
1635                                    btrfs_item_ptr_offset(leaf, path->slots[0]),
1636                                    sizeof(cache->item));
1637                 memcpy(&cache->key, &found_key, sizeof(found_key));
1638                 cache->cached = 0;
1639                 cache->pinned = 0;
1640                 key.objectid = found_key.objectid + found_key.offset;
1641                 btrfs_release_path(root, path);
1642
1643                 if (cache->item.flags & BTRFS_BLOCK_GROUP_MIXED) {
1644                         bit = BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA;
1645                         cache->data = BTRFS_BLOCK_GROUP_MIXED;
1646                 } else if (cache->item.flags & BTRFS_BLOCK_GROUP_DATA) {
1647                         bit = BLOCK_GROUP_DATA;
1648                         cache->data = BTRFS_BLOCK_GROUP_DATA;
1649                 } else {
1650                         bit = BLOCK_GROUP_METADATA;
1651                         cache->data = 0;
1652                 }
1653
1654                 /* use EXTENT_LOCKED to prevent merging */
1655                 set_extent_bits(block_group_cache, found_key.objectid,
1656                                 found_key.objectid + found_key.offset - 1,
1657                                 bit | EXTENT_LOCKED, GFP_NOFS);
1658                 set_state_private(block_group_cache, found_key.objectid,
1659                                   (unsigned long)cache);
1660
1661                 if (key.objectid >=
1662                     btrfs_super_total_bytes(&info->super_copy))
1663                         break;
1664         }
1665
1666         btrfs_free_path(path);
1667         return 0;
1668 }