fs/btrfs/free-space-cache.c

   1 /*
   2  * Copyright (C) 2008 Red Hat.  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/pagemap.h>
  20 #include <linux/sched.h>
  21 #include <linux/slab.h>
  22 #include <linux/math64.h>
  23 #include <linux/ratelimit.h>
  24 #include "ctree.h"
  25 #include "free-space-cache.h"
  26 #include "transaction.h"
  27 #include "disk-io.h"
  28 #include "extent_io.h"
  29 #include "inode-map.h"
  30 #include "volumes.h"
  31
  32 #define BITS_PER_BITMAP         (PAGE_SIZE * 8UL)
  33 #define MAX_CACHE_BYTES_PER_GIG SZ_32K
  34
  35 struct btrfs_trim_range {
  36         u64 start;
  37         u64 bytes;
  38         struct list_head list;
  39 };
  40
  41 static int link_free_space(struct btrfs_free_space_ctl *ctl,
  42                            struct btrfs_free_space *info);
  43 static void unlink_free_space(struct btrfs_free_space_ctl *ctl,
  44                               struct btrfs_free_space *info);
  45
  46 static struct inode *__lookup_free_space_inode(struct btrfs_root *root,
  47                                                struct btrfs_path *path,
  48                                                u64 offset)
  49 {
  50         struct btrfs_key key;
  51         struct btrfs_key location;
  52         struct btrfs_disk_key disk_key;
  53         struct btrfs_free_space_header *header;
  54         struct extent_buffer *leaf;
  55         struct inode *inode = NULL;
  56         int ret;
  57
  58         key.objectid = BTRFS_FREE_SPACE_OBJECTID;
  59         key.offset = offset;
  60         key.type = 0;
  61
  62         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
  63         if (ret < 0)
  64                 return ERR_PTR(ret);
  65         if (ret > 0) {
  66                 btrfs_release_path(path);
  67                 return ERR_PTR(-ENOENT);
  68         }
  69
  70         leaf = path->nodes[0];
  71         header = btrfs_item_ptr(leaf, path->slots[0],
  72                                 struct btrfs_free_space_header);
  73         btrfs_free_space_key(leaf, header, &disk_key);
  74         btrfs_disk_key_to_cpu(&location, &disk_key);
  75         btrfs_release_path(path);
  76
  77         inode = btrfs_iget(root->fs_info->sb, &location, root, NULL);
  78         if (!inode)
  79                 return ERR_PTR(-ENOENT);
  80         if (IS_ERR(inode))
  81                 return inode;
  82         if (is_bad_inode(inode)) {
  83                 iput(inode);
  84                 return ERR_PTR(-ENOENT);
  85         }
  86
  87         mapping_set_gfp_mask(inode->i_mapping,
  88                         mapping_gfp_constraint(inode->i_mapping,
  89                         ~(__GFP_FS | __GFP_HIGHMEM)));
  90
  91         return inode;
  92 }
  93
  94 struct inode *lookup_free_space_inode(struct btrfs_root *root,
  95                                       struct btrfs_block_group_cache
  96                                       *block_group, struct btrfs_path *path)
  97 {
  98         struct inode *inode = NULL;
  99         u32 flags = BTRFS_INODE_NODATASUM | BTRFS_INODE_NODATACOW;
 100
 101         spin_lock(&block_group->lock);
 102         if (block_group->inode)
 103                 inode = igrab(block_group->inode);
 104         spin_unlock(&block_group->lock);
 105         if (inode)
 106                 return inode;
 107
 108         inode = __lookup_free_space_inode(root, path,
 109                                           block_group->key.objectid);
 110         if (IS_ERR(inode))
 111                 return inode;
 112
 113         spin_lock(&block_group->lock);
 114         if (!((BTRFS_I(inode)->flags & flags) == flags)) {
 115                 btrfs_info(root->fs_info,
 116                         "Old style space inode found, converting.");
 117                 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM |
 118                         BTRFS_INODE_NODATACOW;
 119                 block_group->disk_cache_state = BTRFS_DC_CLEAR;
 120         }
 121
 122         if (!block_group->iref) {
 123                 block_group->inode = igrab(inode);
 124                 block_group->iref = 1;
 125         }
 126         spin_unlock(&block_group->lock);
 127
 128         return inode;
 129 }
 130
 131 static int __create_free_space_inode(struct btrfs_root *root,
 132                                      struct btrfs_trans_handle *trans,
 133                                      struct btrfs_path *path,
 134                                      u64 ino, u64 offset)
 135 {
 136         struct btrfs_key key;
 137         struct btrfs_disk_key disk_key;
 138         struct btrfs_free_space_header *header;
 139         struct btrfs_inode_item *inode_item;
 140         struct extent_buffer *leaf;
 141         u64 flags = BTRFS_INODE_NOCOMPRESS | BTRFS_INODE_PREALLOC;
 142         int ret;
 143
 144         ret = btrfs_insert_empty_inode(trans, root, path, ino);
 145         if (ret)
 146                 return ret;
 147
 148         /* We inline crc's for the free disk space cache */
 149         if (ino != BTRFS_FREE_INO_OBJECTID)
 150                 flags |= BTRFS_INODE_NODATASUM | BTRFS_INODE_NODATACOW;
 151
 152         leaf = path->nodes[0];
 153         inode_item = btrfs_item_ptr(leaf, path->slots[0],
 154                                     struct btrfs_inode_item);
 155         btrfs_item_key(leaf, &disk_key, path->slots[0]);
 156         memset_extent_buffer(leaf, 0, (unsigned long)inode_item,
 157                              sizeof(*inode_item));
 158         btrfs_set_inode_generation(leaf, inode_item, trans->transid);
 159         btrfs_set_inode_size(leaf, inode_item, 0);
 160         btrfs_set_inode_nbytes(leaf, inode_item, 0);
 161         btrfs_set_inode_uid(leaf, inode_item, 0);
 162         btrfs_set_inode_gid(leaf, inode_item, 0);
 163         btrfs_set_inode_mode(leaf, inode_item, S_IFREG | 0600);
 164         btrfs_set_inode_flags(leaf, inode_item, flags);
 165         btrfs_set_inode_nlink(leaf, inode_item, 1);
 166         btrfs_set_inode_transid(leaf, inode_item, trans->transid);
 167         btrfs_set_inode_block_group(leaf, inode_item, offset);
 168         btrfs_mark_buffer_dirty(leaf);
 169         btrfs_release_path(path);
 170
 171         key.objectid = BTRFS_FREE_SPACE_OBJECTID;
 172         key.offset = offset;
 173         key.type = 0;
 174         ret = btrfs_insert_empty_item(trans, root, path, &key,
 175                                       sizeof(struct btrfs_free_space_header));
 176         if (ret < 0) {
 177                 btrfs_release_path(path);
 178                 return ret;
 179         }
 180
 181         leaf = path->nodes[0];
 182         header = btrfs_item_ptr(leaf, path->slots[0],
 183                                 struct btrfs_free_space_header);
 184         memset_extent_buffer(leaf, 0, (unsigned long)header, sizeof(*header));
 185         btrfs_set_free_space_key(leaf, header, &disk_key);
 186         btrfs_mark_buffer_dirty(leaf);
 187         btrfs_release_path(path);
 188
 189         return 0;
 190 }
 191
 192 int create_free_space_inode(struct btrfs_root *root,
 193                             struct btrfs_trans_handle *trans,
 194                             struct btrfs_block_group_cache *block_group,
 195                             struct btrfs_path *path)
 196 {
 197         int ret;
 198         u64 ino;
 199
 200         ret = btrfs_find_free_objectid(root, &ino);
 201         if (ret < 0)
 202                 return ret;
 203
 204         return __create_free_space_inode(root, trans, path, ino,
 205                                          block_group->key.objectid);
 206 }
 207
 208 int btrfs_check_trunc_cache_free_space(struct btrfs_root *root,
 209                                        struct btrfs_block_rsv *rsv)
 210 {
 211         u64 needed_bytes;
 212         int ret;
 213
 214         /* 1 for slack space, 1 for updating the inode */
 215         needed_bytes = btrfs_calc_trunc_metadata_size(root, 1) +
 216                 btrfs_calc_trans_metadata_size(root, 1);
 217
 218         spin_lock(&rsv->lock);
 219         if (rsv->reserved < needed_bytes)
 220                 ret = -ENOSPC;
 221         else
 222                 ret = 0;
 223         spin_unlock(&rsv->lock);
 224         return ret;
 225 }
 226
 227 int btrfs_truncate_free_space_cache(struct btrfs_root *root,
 228                                     struct btrfs_trans_handle *trans,
 229                                     struct btrfs_block_group_cache *block_group,
 230                                     struct inode *inode)
 231 {
 232         int ret = 0;
 233         struct btrfs_path *path = btrfs_alloc_path();
 234         bool locked = false;
 235
 236         if (!path) {
 237                 ret = -ENOMEM;
 238                 goto fail;
 239         }
 240
 241         if (block_group) {
 242                 locked = true;
 243                 mutex_lock(&trans->transaction->cache_write_mutex);
 244                 if (!list_empty(&block_group->io_list)) {
 245                         list_del_init(&block_group->io_list);
 246
 247                         btrfs_wait_cache_io(root, trans, block_group,
 248                                             &block_group->io_ctl, path,
 249                                             block_group->key.objectid);
 250                         btrfs_put_block_group(block_group);
 251                 }
 252
 253                 /*
 254                  * now that we've truncated the cache away, its no longer
 255                  * setup or written
 256                  */
 257                 spin_lock(&block_group->lock);
 258                 block_group->disk_cache_state = BTRFS_DC_CLEAR;
 259                 spin_unlock(&block_group->lock);
 260         }
 261         btrfs_free_path(path);
 262
 263         btrfs_i_size_write(inode, 0);
 264         truncate_pagecache(inode, 0);
 265
 266         /*
 267          * We don't need an orphan item because truncating the free space cache
 268          * will never be split across transactions.
 269          * We don't need to check for -EAGAIN because we're a free space
 270          * cache inode
 271          */
 272         ret = btrfs_truncate_inode_items(trans, root, inode,
 273                                          0, BTRFS_EXTENT_DATA_KEY);
 274         if (ret)
 275                 goto fail;
 276
 277         ret = btrfs_update_inode(trans, root, inode);
 278
 279 fail:
 280         if (locked)
 281                 mutex_unlock(&trans->transaction->cache_write_mutex);
 282         if (ret)
 283                 btrfs_abort_transaction(trans, ret);
 284
 285         return ret;
 286 }
 287
 288 static int readahead_cache(struct inode *inode)
 289 {
 290         struct file_ra_state *ra;
 291         unsigned long last_index;
 292
 293         ra = kzalloc(sizeof(*ra), GFP_NOFS);
 294         if (!ra)
 295                 return -ENOMEM;
 296
 297         file_ra_state_init(ra, inode->i_mapping);
 298         last_index = (i_size_read(inode) - 1) >> PAGE_SHIFT;
 299
 300         page_cache_sync_readahead(inode->i_mapping, ra, NULL, 0, last_index);
 301
 302         kfree(ra);
 303
 304         return 0;
 305 }
 306
 307 static int io_ctl_init(struct btrfs_io_ctl *io_ctl, struct inode *inode,
 308                        struct btrfs_root *root, int write)
 309 {
 310         int num_pages;
 311         int check_crcs = 0;
 312
 313         num_pages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
 314
 315         if (btrfs_ino(inode) != BTRFS_FREE_INO_OBJECTID)
 316                 check_crcs = 1;
 317
 318         /* Make sure we can fit our crcs into the first page */
 319         if (write && check_crcs &&
 320             (num_pages * sizeof(u32)) >= PAGE_SIZE)
 321                 return -ENOSPC;
 322
 323         memset(io_ctl, 0, sizeof(struct btrfs_io_ctl));
 324
 325         io_ctl->pages = kcalloc(num_pages, sizeof(struct page *), GFP_NOFS);
 326         if (!io_ctl->pages)
 327                 return -ENOMEM;
 328
 329         io_ctl->num_pages = num_pages;
 330         io_ctl->root = root;
 331         io_ctl->check_crcs = check_crcs;
 332         io_ctl->inode = inode;
 333
 334         return 0;
 335 }
 336
 337 static void io_ctl_free(struct btrfs_io_ctl *io_ctl)
 338 {
 339         kfree(io_ctl->pages);
 340         io_ctl->pages = NULL;
 341 }
 342
 343 static void io_ctl_unmap_page(struct btrfs_io_ctl *io_ctl)
 344 {
 345         if (io_ctl->cur) {
 346                 io_ctl->cur = NULL;
 347                 io_ctl->orig = NULL;
 348         }
 349 }
 350
 351 static void io_ctl_map_page(struct btrfs_io_ctl *io_ctl, int clear)
 352 {
 353         ASSERT(io_ctl->index < io_ctl->num_pages);
 354         io_ctl->page = io_ctl->pages[io_ctl->index++];
 355         io_ctl->cur = page_address(io_ctl->page);
 356         io_ctl->orig = io_ctl->cur;
 357         io_ctl->size = PAGE_SIZE;
 358         if (clear)
 359                 memset(io_ctl->cur, 0, PAGE_SIZE);
 360 }
 361
 362 static void io_ctl_drop_pages(struct btrfs_io_ctl *io_ctl)
 363 {
 364         int i;
 365
 366         io_ctl_unmap_page(io_ctl);
 367
 368         for (i = 0; i < io_ctl->num_pages; i++) {
 369                 if (io_ctl->pages[i]) {
 370                         ClearPageChecked(io_ctl->pages[i]);
 371                         unlock_page(io_ctl->pages[i]);
 372                         put_page(io_ctl->pages[i]);
 373                 }
 374         }
 375 }
 376
 377 static int io_ctl_prepare_pages(struct btrfs_io_ctl *io_ctl, struct inode *inode,
 378                                 int uptodate)
 379 {
 380         struct page *page;
 381         gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
 382         int i;
 383
 384         for (i = 0; i < io_ctl->num_pages; i++) {
 385                 page = find_or_create_page(inode->i_mapping, i, mask);
 386                 if (!page) {
 387                         io_ctl_drop_pages(io_ctl);
 388                         return -ENOMEM;
 389                 }
 390                 io_ctl->pages[i] = page;
 391                 if (uptodate && !PageUptodate(page)) {
 392                         btrfs_readpage(NULL, page);
 393                         lock_page(page);
 394                         if (!PageUptodate(page)) {
 395                                 btrfs_err(BTRFS_I(inode)->root->fs_info,
 396                                            "error reading free space cache");
 397                                 io_ctl_drop_pages(io_ctl);
 398                                 return -EIO;
 399                         }
 400                 }
 401         }
 402
 403         for (i = 0; i < io_ctl->num_pages; i++) {
 404                 clear_page_dirty_for_io(io_ctl->pages[i]);
 405                 set_page_extent_mapped(io_ctl->pages[i]);
 406         }
 407
 408         return 0;
 409 }
 410
 411 static void io_ctl_set_generation(struct btrfs_io_ctl *io_ctl, u64 generation)
 412 {
 413         __le64 *val;
 414
 415         io_ctl_map_page(io_ctl, 1);
 416
 417         /*
 418          * Skip the csum areas.  If we don't check crcs then we just have a
 419          * 64bit chunk at the front of the first page.
 420          */
 421         if (io_ctl->check_crcs) {
 422                 io_ctl->cur += (sizeof(u32) * io_ctl->num_pages);
 423                 io_ctl->size -= sizeof(u64) + (sizeof(u32) * io_ctl->num_pages);
 424         } else {
 425                 io_ctl->cur += sizeof(u64);
 426                 io_ctl->size -= sizeof(u64) * 2;
 427         }
 428
 429         val = io_ctl->cur;
 430         *val = cpu_to_le64(generation);
 431         io_ctl->cur += sizeof(u64);
 432 }
 433
 434 static int io_ctl_check_generation(struct btrfs_io_ctl *io_ctl, u64 generation)
 435 {
 436         __le64 *gen;
 437
 438         /*
 439          * Skip the crc area.  If we don't check crcs then we just have a 64bit
 440          * chunk at the front of the first page.
 441          */
 442         if (io_ctl->check_crcs) {
 443                 io_ctl->cur += sizeof(u32) * io_ctl->num_pages;
 444                 io_ctl->size -= sizeof(u64) +
 445                         (sizeof(u32) * io_ctl->num_pages);
 446         } else {
 447                 io_ctl->cur += sizeof(u64);
 448                 io_ctl->size -= sizeof(u64) * 2;
 449         }
 450
 451         gen = io_ctl->cur;
 452         if (le64_to_cpu(*gen) != generation) {
 453                 btrfs_err_rl(io_ctl->root->fs_info,
 454                         "space cache generation (%llu) does not match inode (%llu)",
 455                                 *gen, generation);
 456                 io_ctl_unmap_page(io_ctl);
 457                 return -EIO;
 458         }
 459         io_ctl->cur += sizeof(u64);
 460         return 0;
 461 }
 462
 463 static void io_ctl_set_crc(struct btrfs_io_ctl *io_ctl, int index)
 464 {
 465         u32 *tmp;
 466         u32 crc = ~(u32)0;
 467         unsigned offset = 0;
 468
 469         if (!io_ctl->check_crcs) {
 470                 io_ctl_unmap_page(io_ctl);
 471                 return;
 472         }
 473
 474         if (index == 0)
 475                 offset = sizeof(u32) * io_ctl->num_pages;
 476
 477         crc = btrfs_csum_data(io_ctl->orig + offset, crc,
 478                               PAGE_SIZE - offset);
 479         btrfs_csum_final(crc, (char *)&crc);
 480         io_ctl_unmap_page(io_ctl);
 481         tmp = page_address(io_ctl->pages[0]);
 482         tmp += index;
 483         *tmp = crc;
 484 }
 485
 486 static int io_ctl_check_crc(struct btrfs_io_ctl *io_ctl, int index)
 487 {
 488         u32 *tmp, val;
 489         u32 crc = ~(u32)0;
 490         unsigned offset = 0;
 491
 492         if (!io_ctl->check_crcs) {
 493                 io_ctl_map_page(io_ctl, 0);
 494                 return 0;
 495         }
 496
 497         if (index == 0)
 498                 offset = sizeof(u32) * io_ctl->num_pages;
 499
 500         tmp = page_address(io_ctl->pages[0]);
 501         tmp += index;
 502         val = *tmp;
 503
 504         io_ctl_map_page(io_ctl, 0);
 505         crc = btrfs_csum_data(io_ctl->orig + offset, crc,
 506                               PAGE_SIZE - offset);
 507         btrfs_csum_final(crc, (char *)&crc);
 508         if (val != crc) {
 509                 btrfs_err_rl(io_ctl->root->fs_info,
 510                         "csum mismatch on free space cache");
 511                 io_ctl_unmap_page(io_ctl);
 512                 return -EIO;
 513         }
 514
 515         return 0;
 516 }
 517
 518 static int io_ctl_add_entry(struct btrfs_io_ctl *io_ctl, u64 offset, u64 bytes,
 519                             void *bitmap)
 520 {
 521         struct btrfs_free_space_entry *entry;
 522
 523         if (!io_ctl->cur)
 524                 return -ENOSPC;
 525
 526         entry = io_ctl->cur;
 527         entry->offset = cpu_to_le64(offset);
 528         entry->bytes = cpu_to_le64(bytes);
 529         entry->type = (bitmap) ? BTRFS_FREE_SPACE_BITMAP :
 530                 BTRFS_FREE_SPACE_EXTENT;
 531         io_ctl->cur += sizeof(struct btrfs_free_space_entry);
 532         io_ctl->size -= sizeof(struct btrfs_free_space_entry);
 533
 534         if (io_ctl->size >= sizeof(struct btrfs_free_space_entry))
 535                 return 0;
 536
 537         io_ctl_set_crc(io_ctl, io_ctl->index - 1);
 538
 539         /* No more pages to map */
 540         if (io_ctl->index >= io_ctl->num_pages)
 541                 return 0;
 542
 543         /* map the next page */
 544         io_ctl_map_page(io_ctl, 1);
 545         return 0;
 546 }
 547
 548 static int io_ctl_add_bitmap(struct btrfs_io_ctl *io_ctl, void *bitmap)
 549 {
 550         if (!io_ctl->cur)
 551                 return -ENOSPC;
 552
 553         /*
 554          * If we aren't at the start of the current page, unmap this one and
 555          * map the next one if there is any left.
 556          */
 557         if (io_ctl->cur != io_ctl->orig) {
 558                 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
 559                 if (io_ctl->index >= io_ctl->num_pages)
 560                         return -ENOSPC;
 561                 io_ctl_map_page(io_ctl, 0);
 562         }
 563
 564         memcpy(io_ctl->cur, bitmap, PAGE_SIZE);
 565         io_ctl_set_crc(io_ctl, io_ctl->index - 1);
 566         if (io_ctl->index < io_ctl->num_pages)
 567                 io_ctl_map_page(io_ctl, 0);
 568         return 0;
 569 }
 570
 571 static void io_ctl_zero_remaining_pages(struct btrfs_io_ctl *io_ctl)
 572 {
 573         /*
 574          * If we're not on the boundary we know we've modified the page and we
 575          * need to crc the page.
 576          */
 577         if (io_ctl->cur != io_ctl->orig)
 578                 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
 579         else
 580                 io_ctl_unmap_page(io_ctl);
 581
 582         while (io_ctl->index < io_ctl->num_pages) {
 583                 io_ctl_map_page(io_ctl, 1);
 584                 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
 585         }
 586 }
 587
 588 static int io_ctl_read_entry(struct btrfs_io_ctl *io_ctl,
 589                             struct btrfs_free_space *entry, u8 *type)
 590 {
 591         struct btrfs_free_space_entry *e;
 592         int ret;
 593
 594         if (!io_ctl->cur) {
 595                 ret = io_ctl_check_crc(io_ctl, io_ctl->index);
 596                 if (ret)
 597                         return ret;
 598         }
 599
 600         e = io_ctl->cur;
 601         entry->offset = le64_to_cpu(e->offset);
 602         entry->bytes = le64_to_cpu(e->bytes);
 603         *type = e->type;
 604         io_ctl->cur += sizeof(struct btrfs_free_space_entry);
 605         io_ctl->size -= sizeof(struct btrfs_free_space_entry);
 606
 607         if (io_ctl->size >= sizeof(struct btrfs_free_space_entry))
 608                 return 0;
 609
 610         io_ctl_unmap_page(io_ctl);
 611
 612         return 0;
 613 }
 614
 615 static int io_ctl_read_bitmap(struct btrfs_io_ctl *io_ctl,
 616                               struct btrfs_free_space *entry)
 617 {
 618         int ret;
 619
 620         ret = io_ctl_check_crc(io_ctl, io_ctl->index);
 621         if (ret)
 622                 return ret;
 623
 624         memcpy(entry->bitmap, io_ctl->cur, PAGE_SIZE);
 625         io_ctl_unmap_page(io_ctl);
 626
 627         return 0;
 628 }
 629
 630 /*
 631  * Since we attach pinned extents after the fact we can have contiguous sections
 632  * of free space that are split up in entries.  This poses a problem with the
 633  * tree logging stuff since it could have allocated across what appears to be 2
 634  * entries since we would have merged the entries when adding the pinned extents
 635  * back to the free space cache.  So run through the space cache that we just
 636  * loaded and merge contiguous entries.  This will make the log replay stuff not
 637  * blow up and it will make for nicer allocator behavior.
 638  */
 639 static void merge_space_tree(struct btrfs_free_space_ctl *ctl)
 640 {
 641         struct btrfs_free_space *e, *prev = NULL;
 642         struct rb_node *n;
 643
 644 again:
 645         spin_lock(&ctl->tree_lock);
 646         for (n = rb_first(&ctl->free_space_offset); n; n = rb_next(n)) {
 647                 e = rb_entry(n, struct btrfs_free_space, offset_index);
 648                 if (!prev)
 649                         goto next;
 650                 if (e->bitmap || prev->bitmap)
 651                         goto next;
 652                 if (prev->offset + prev->bytes == e->offset) {
 653                         unlink_free_space(ctl, prev);
 654                         unlink_free_space(ctl, e);
 655                         prev->bytes += e->bytes;
 656                         kmem_cache_free(btrfs_free_space_cachep, e);
 657                         link_free_space(ctl, prev);
 658                         prev = NULL;
 659                         spin_unlock(&ctl->tree_lock);
 660                         goto again;
 661                 }
 662 next:
 663                 prev = e;
 664         }
 665         spin_unlock(&ctl->tree_lock);
 666 }
 667
 668 static int __load_free_space_cache(struct btrfs_root *root, struct inode *inode,
 669                                    struct btrfs_free_space_ctl *ctl,
 670                                    struct btrfs_path *path, u64 offset)
 671 {
 672         struct btrfs_free_space_header *header;
 673         struct extent_buffer *leaf;
 674         struct btrfs_io_ctl io_ctl;
 675         struct btrfs_key key;
 676         struct btrfs_free_space *e, *n;
 677         LIST_HEAD(bitmaps);
 678         u64 num_entries;
 679         u64 num_bitmaps;
 680         u64 generation;
 681         u8 type;
 682         int ret = 0;
 683
 684         /* Nothing in the space cache, goodbye */
 685         if (!i_size_read(inode))
 686                 return 0;
 687
 688         key.objectid = BTRFS_FREE_SPACE_OBJECTID;
 689         key.offset = offset;
 690         key.type = 0;
 691
 692         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 693         if (ret < 0)
 694                 return 0;
 695         else if (ret > 0) {
 696                 btrfs_release_path(path);
 697                 return 0;
 698         }
 699
 700         ret = -1;
 701
 702         leaf = path->nodes[0];
 703         header = btrfs_item_ptr(leaf, path->slots[0],
 704                                 struct btrfs_free_space_header);
 705         num_entries = btrfs_free_space_entries(leaf, header);
 706         num_bitmaps = btrfs_free_space_bitmaps(leaf, header);
 707         generation = btrfs_free_space_generation(leaf, header);
 708         btrfs_release_path(path);
 709
 710         if (!BTRFS_I(inode)->generation) {
 711                 btrfs_info(root->fs_info,
 712                            "The free space cache file (%llu) is invalid. skip it\n",
 713                            offset);
 714                 return 0;
 715         }
 716
 717         if (BTRFS_I(inode)->generation != generation) {
 718                 btrfs_err(root->fs_info,
 719                         "free space inode generation (%llu) did not match free space cache generation (%llu)",
 720                         BTRFS_I(inode)->generation, generation);
 721                 return 0;
 722         }
 723
 724         if (!num_entries)
 725                 return 0;
 726
 727         ret = io_ctl_init(&io_ctl, inode, root, 0);
 728         if (ret)
 729                 return ret;
 730
 731         ret = readahead_cache(inode);
 732         if (ret)
 733                 goto out;
 734
 735         ret = io_ctl_prepare_pages(&io_ctl, inode, 1);
 736         if (ret)
 737                 goto out;
 738
 739         ret = io_ctl_check_crc(&io_ctl, 0);
 740         if (ret)
 741                 goto free_cache;
 742
 743         ret = io_ctl_check_generation(&io_ctl, generation);
 744         if (ret)
 745                 goto free_cache;
 746
 747         while (num_entries) {
 748                 e = kmem_cache_zalloc(btrfs_free_space_cachep,
 749                                       GFP_NOFS);
 750                 if (!e)
 751                         goto free_cache;
 752
 753                 ret = io_ctl_read_entry(&io_ctl, e, &type);
 754                 if (ret) {
 755                         kmem_cache_free(btrfs_free_space_cachep, e);
 756                         goto free_cache;
 757                 }
 758
 759                 if (!e->bytes) {
 760                         kmem_cache_free(btrfs_free_space_cachep, e);
 761                         goto free_cache;
 762                 }
 763
 764                 if (type == BTRFS_FREE_SPACE_EXTENT) {
 765                         spin_lock(&ctl->tree_lock);
 766                         ret = link_free_space(ctl, e);
 767                         spin_unlock(&ctl->tree_lock);
 768                         if (ret) {
 769                                 btrfs_err(root->fs_info,
 770                                         "Duplicate entries in free space cache, dumping");
 771                                 kmem_cache_free(btrfs_free_space_cachep, e);
 772                                 goto free_cache;
 773                         }
 774                 } else {
 775                         ASSERT(num_bitmaps);
 776                         num_bitmaps--;
 777                         e->bitmap = kzalloc(PAGE_SIZE, GFP_NOFS);
 778                         if (!e->bitmap) {
 779                                 kmem_cache_free(
 780                                         btrfs_free_space_cachep, e);
 781                                 goto free_cache;
 782                         }
 783                         spin_lock(&ctl->tree_lock);
 784                         ret = link_free_space(ctl, e);
 785                         ctl->total_bitmaps++;
 786                         ctl->op->recalc_thresholds(ctl);
 787                         spin_unlock(&ctl->tree_lock);
 788                         if (ret) {
 789                                 btrfs_err(root->fs_info,
 790                                         "Duplicate entries in free space cache, dumping");
 791                                 kmem_cache_free(btrfs_free_space_cachep, e);
 792                                 goto free_cache;
 793                         }
 794                         list_add_tail(&e->list, &bitmaps);
 795                 }
 796
 797                 num_entries--;
 798         }
 799
 800         io_ctl_unmap_page(&io_ctl);
 801
 802         /*
 803          * We add the bitmaps at the end of the entries in order that
 804          * the bitmap entries are added to the cache.
 805          */
 806         list_for_each_entry_safe(e, n, &bitmaps, list) {
 807                 list_del_init(&e->list);
 808                 ret = io_ctl_read_bitmap(&io_ctl, e);
 809                 if (ret)
 810                         goto free_cache;
 811         }
 812
 813         io_ctl_drop_pages(&io_ctl);
 814         merge_space_tree(ctl);
 815         ret = 1;
 816 out:
 817         io_ctl_free(&io_ctl);
 818         return ret;
 819 free_cache:
 820         io_ctl_drop_pages(&io_ctl);
 821         __btrfs_remove_free_space_cache(ctl);
 822         goto out;
 823 }
 824
 825 int load_free_space_cache(struct btrfs_fs_info *fs_info,
 826                           struct btrfs_block_group_cache *block_group)
 827 {
 828         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
 829         struct btrfs_root *root = fs_info->tree_root;
 830         struct inode *inode;
 831         struct btrfs_path *path;
 832         int ret = 0;
 833         bool matched;
 834         u64 used = btrfs_block_group_used(&block_group->item);
 835
 836         /*
 837          * If this block group has been marked to be cleared for one reason or
 838          * another then we can't trust the on disk cache, so just return.
 839          */
 840         spin_lock(&block_group->lock);
 841         if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
 842                 spin_unlock(&block_group->lock);
 843                 return 0;
 844         }
 845         spin_unlock(&block_group->lock);
 846
 847         path = btrfs_alloc_path();
 848         if (!path)
 849                 return 0;
 850         path->search_commit_root = 1;
 851         path->skip_locking = 1;
 852
 853         inode = lookup_free_space_inode(root, block_group, path);
 854         if (IS_ERR(inode)) {
 855                 btrfs_free_path(path);
 856                 return 0;
 857         }
 858
 859         /* We may have converted the inode and made the cache invalid. */
 860         spin_lock(&block_group->lock);
 861         if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
 862                 spin_unlock(&block_group->lock);
 863                 btrfs_free_path(path);
 864                 goto out;
 865         }
 866         spin_unlock(&block_group->lock);
 867
 868         ret = __load_free_space_cache(fs_info->tree_root, inode, ctl,
 869                                       path, block_group->key.objectid);
 870         btrfs_free_path(path);
 871         if (ret <= 0)
 872                 goto out;
 873
 874         spin_lock(&ctl->tree_lock);
 875         matched = (ctl->free_space == (block_group->key.offset - used -
 876                                        block_group->bytes_super));
 877         spin_unlock(&ctl->tree_lock);
 878
 879         if (!matched) {
 880                 __btrfs_remove_free_space_cache(ctl);
 881                 btrfs_warn(fs_info,
 882                            "block group %llu has wrong amount of free space",
 883                            block_group->key.objectid);
 884                 ret = -1;
 885         }
 886 out:
 887         if (ret < 0) {
 888                 /* This cache is bogus, make sure it gets cleared */
 889                 spin_lock(&block_group->lock);
 890                 block_group->disk_cache_state = BTRFS_DC_CLEAR;
 891                 spin_unlock(&block_group->lock);
 892                 ret = 0;
 893
 894                 btrfs_warn(fs_info,
 895                            "failed to load free space cache for block group %llu, rebuilding it now",
 896                            block_group->key.objectid);
 897         }
 898
 899         iput(inode);
 900         return ret;
 901 }
 902
 903 static noinline_for_stack
 904 int write_cache_extent_entries(struct btrfs_io_ctl *io_ctl,
 905                               struct btrfs_free_space_ctl *ctl,
 906                               struct btrfs_block_group_cache *block_group,
 907                               int *entries, int *bitmaps,
 908                               struct list_head *bitmap_list)
 909 {
 910         int ret;
 911         struct btrfs_free_cluster *cluster = NULL;
 912         struct btrfs_free_cluster *cluster_locked = NULL;
 913         struct rb_node *node = rb_first(&ctl->free_space_offset);
 914         struct btrfs_trim_range *trim_entry;
 915
 916         /* Get the cluster for this block_group if it exists */
 917         if (block_group && !list_empty(&block_group->cluster_list)) {
 918                 cluster = list_entry(block_group->cluster_list.next,
 919                                      struct btrfs_free_cluster,
 920                                      block_group_list);
 921         }
 922
 923         if (!node && cluster) {
 924                 cluster_locked = cluster;
 925                 spin_lock(&cluster_locked->lock);
 926                 node = rb_first(&cluster->root);
 927                 cluster = NULL;
 928         }
 929
 930         /* Write out the extent entries */
 931         while (node) {
 932                 struct btrfs_free_space *e;
 933
 934                 e = rb_entry(node, struct btrfs_free_space, offset_index);
 935                 *entries += 1;
 936
 937                 ret = io_ctl_add_entry(io_ctl, e->offset, e->bytes,
 938                                        e->bitmap);
 939                 if (ret)
 940                         goto fail;
 941
 942                 if (e->bitmap) {
 943                         list_add_tail(&e->list, bitmap_list);
 944                         *bitmaps += 1;
 945                 }
 946                 node = rb_next(node);
 947                 if (!node && cluster) {
 948                         node = rb_first(&cluster->root);
 949                         cluster_locked = cluster;
 950                         spin_lock(&cluster_locked->lock);
 951                         cluster = NULL;
 952                 }
 953         }
 954         if (cluster_locked) {
 955                 spin_unlock(&cluster_locked->lock);
 956                 cluster_locked = NULL;
 957         }
 958
 959         /*
 960          * Make sure we don't miss any range that was removed from our rbtree
 961          * because trimming is running. Otherwise after a umount+mount (or crash
 962          * after committing the transaction) we would leak free space and get
 963          * an inconsistent free space cache report from fsck.
 964          */
 965         list_for_each_entry(trim_entry, &ctl->trimming_ranges, list) {
 966                 ret = io_ctl_add_entry(io_ctl, trim_entry->start,
 967                                        trim_entry->bytes, NULL);
 968                 if (ret)
 969                         goto fail;
 970                 *entries += 1;
 971         }
 972
 973         return 0;
 974 fail:
 975         if (cluster_locked)
 976                 spin_unlock(&cluster_locked->lock);
 977         return -ENOSPC;
 978 }
 979
 980 static noinline_for_stack int
 981 update_cache_item(struct btrfs_trans_handle *trans,
 982                   struct btrfs_root *root,
 983                   struct inode *inode,
 984                   struct btrfs_path *path, u64 offset,
 985                   int entries, int bitmaps)
 986 {
 987         struct btrfs_key key;
 988         struct btrfs_free_space_header *header;
 989         struct extent_buffer *leaf;
 990         int ret;
 991
 992         key.objectid = BTRFS_FREE_SPACE_OBJECTID;
 993         key.offset = offset;
 994         key.type = 0;
 995
 996         ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
 997         if (ret < 0) {
 998                 clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, inode->i_size - 1,
 999                                  EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, NULL,
1000                                  GFP_NOFS);
1001                 goto fail;
1002         }
1003         leaf = path->nodes[0];
1004         if (ret > 0) {
1005                 struct btrfs_key found_key;
1006                 ASSERT(path->slots[0]);
1007                 path->slots[0]--;
1008                 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1009                 if (found_key.objectid != BTRFS_FREE_SPACE_OBJECTID ||
1010                     found_key.offset != offset) {
1011                         clear_extent_bit(&BTRFS_I(inode)->io_tree, 0,
1012                                          inode->i_size - 1,
1013                                          EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0,
1014                                          NULL, GFP_NOFS);
1015                         btrfs_release_path(path);
1016                         goto fail;
1017                 }
1018         }
1019
1020         BTRFS_I(inode)->generation = trans->transid;
1021         header = btrfs_item_ptr(leaf, path->slots[0],
1022                                 struct btrfs_free_space_header);
1023         btrfs_set_free_space_entries(leaf, header, entries);
1024         btrfs_set_free_space_bitmaps(leaf, header, bitmaps);
1025         btrfs_set_free_space_generation(leaf, header, trans->transid);
1026         btrfs_mark_buffer_dirty(leaf);
1027         btrfs_release_path(path);
1028
1029         return 0;
1030
1031 fail:
1032         return -1;
1033 }
1034
1035 static noinline_for_stack int
1036 write_pinned_extent_entries(struct btrfs_root *root,
1037                             struct btrfs_block_group_cache *block_group,
1038                             struct btrfs_io_ctl *io_ctl,
1039                             int *entries)
1040 {
1041         u64 start, extent_start, extent_end, len;
1042         struct extent_io_tree *unpin = NULL;
1043         int ret;
1044
1045         if (!block_group)
1046                 return 0;
1047
1048         /*
1049          * We want to add any pinned extents to our free space cache
1050          * so we don't leak the space
1051          *
1052          * We shouldn't have switched the pinned extents yet so this is the
1053          * right one
1054          */
1055         unpin = root->fs_info->pinned_extents;
1056
1057         start = block_group->key.objectid;
1058
1059         while (start < block_group->key.objectid + block_group->key.offset) {
1060                 ret = find_first_extent_bit(unpin, start,
1061                                             &extent_start, &extent_end,
1062                                             EXTENT_DIRTY, NULL);
1063                 if (ret)
1064                         return 0;
1065
1066                 /* This pinned extent is out of our range */
1067                 if (extent_start >= block_group->key.objectid +
1068                     block_group->key.offset)
1069                         return 0;
1070
1071                 extent_start = max(extent_start, start);
1072                 extent_end = min(block_group->key.objectid +
1073                                  block_group->key.offset, extent_end + 1);
1074                 len = extent_end - extent_start;
1075
1076                 *entries += 1;
1077                 ret = io_ctl_add_entry(io_ctl, extent_start, len, NULL);
1078                 if (ret)
1079                         return -ENOSPC;
1080
1081                 start = extent_end;
1082         }
1083
1084         return 0;
1085 }
1086
1087 static noinline_for_stack int
1088 write_bitmap_entries(struct btrfs_io_ctl *io_ctl, struct list_head *bitmap_list)
1089 {
1090         struct btrfs_free_space *entry, *next;
1091         int ret;
1092
1093         /* Write out the bitmaps */
1094         list_for_each_entry_safe(entry, next, bitmap_list, list) {
1095                 ret = io_ctl_add_bitmap(io_ctl, entry->bitmap);
1096                 if (ret)
1097                         return -ENOSPC;
1098                 list_del_init(&entry->list);
1099         }
1100
1101         return 0;
1102 }
1103
1104 static int flush_dirty_cache(struct inode *inode)
1105 {
1106         int ret;
1107
1108         ret = btrfs_wait_ordered_range(inode, 0, (u64)-1);
1109         if (ret)
1110                 clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, inode->i_size - 1,
1111                                  EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, NULL,
1112                                  GFP_NOFS);
1113
1114         return ret;
1115 }
1116
1117 static void noinline_for_stack
1118 cleanup_bitmap_list(struct list_head *bitmap_list)
1119 {
1120         struct btrfs_free_space *entry, *next;
1121
1122         list_for_each_entry_safe(entry, next, bitmap_list, list)
1123                 list_del_init(&entry->list);
1124 }
1125
1126 static void noinline_for_stack
1127 cleanup_write_cache_enospc(struct inode *inode,
1128                            struct btrfs_io_ctl *io_ctl,
1129                            struct extent_state **cached_state,
1130                            struct list_head *bitmap_list)
1131 {
1132         io_ctl_drop_pages(io_ctl);
1133         unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
1134                              i_size_read(inode) - 1, cached_state,
1135                              GFP_NOFS);
1136 }
1137
1138 int btrfs_wait_cache_io(struct btrfs_root *root,
1139                         struct btrfs_trans_handle *trans,
1140                         struct btrfs_block_group_cache *block_group,
1141                         struct btrfs_io_ctl *io_ctl,
1142                         struct btrfs_path *path, u64 offset)
1143 {
1144         int ret;
1145         struct inode *inode = io_ctl->inode;
1146
1147         if (!inode)
1148                 return 0;
1149
1150         if (block_group)
1151                 root = root->fs_info->tree_root;
1152
1153         /* Flush the dirty pages in the cache file. */
1154         ret = flush_dirty_cache(inode);
1155         if (ret)
1156                 goto out;
1157
1158         /* Update the cache item to tell everyone this cache file is valid. */
1159         ret = update_cache_item(trans, root, inode, path, offset,
1160                                 io_ctl->entries, io_ctl->bitmaps);
1161 out:
1162         io_ctl_free(io_ctl);
1163         if (ret) {
1164                 invalidate_inode_pages2(inode->i_mapping);
1165                 BTRFS_I(inode)->generation = 0;
1166                 if (block_group) {
1167 #ifdef DEBUG
1168                         btrfs_err(root->fs_info,
1169                                 "failed to write free space cache for block group %llu",
1170                                 block_group->key.objectid);
1171 #endif
1172                 }
1173         }
1174         btrfs_update_inode(trans, root, inode);
1175
1176         if (block_group) {
1177                 /* the dirty list is protected by the dirty_bgs_lock */
1178                 spin_lock(&trans->transaction->dirty_bgs_lock);
1179
1180                 /* the disk_cache_state is protected by the block group lock */
1181                 spin_lock(&block_group->lock);
1182
1183                 /*
1184                  * only mark this as written if we didn't get put back on
1185                  * the dirty list while waiting for IO.   Otherwise our
1186                  * cache state won't be right, and we won't get written again
1187                  */
1188                 if (!ret && list_empty(&block_group->dirty_list))
1189                         block_group->disk_cache_state = BTRFS_DC_WRITTEN;
1190                 else if (ret)
1191                         block_group->disk_cache_state = BTRFS_DC_ERROR;
1192
1193                 spin_unlock(&block_group->lock);
1194                 spin_unlock(&trans->transaction->dirty_bgs_lock);
1195                 io_ctl->inode = NULL;
1196                 iput(inode);
1197         }
1198
1199         return ret;
1200
1201 }
1202
1203 /**
1204  * __btrfs_write_out_cache - write out cached info to an inode
1205  * @root - the root the inode belongs to
1206  * @ctl - the free space cache we are going to write out
1207  * @block_group - the block_group for this cache if it belongs to a block_group
1208  * @trans - the trans handle
1209  * @path - the path to use
1210  * @offset - the offset for the key we'll insert
1211  *
1212  * This function writes out a free space cache struct to disk for quick recovery
1213  * on mount.  This will return 0 if it was successful in writing the cache out,
1214  * or an errno if it was not.
1215  */
1216 static int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode,
1217                                    struct btrfs_free_space_ctl *ctl,
1218                                    struct btrfs_block_group_cache *block_group,
1219                                    struct btrfs_io_ctl *io_ctl,
1220                                    struct btrfs_trans_handle *trans,
1221                                    struct btrfs_path *path, u64 offset)
1222 {
1223         struct extent_state *cached_state = NULL;
1224         LIST_HEAD(bitmap_list);
1225         int entries = 0;
1226         int bitmaps = 0;
1227         int ret;
1228         int must_iput = 0;
1229
1230         if (!i_size_read(inode))
1231                 return -EIO;
1232
1233         WARN_ON(io_ctl->pages);
1234         ret = io_ctl_init(io_ctl, inode, root, 1);
1235         if (ret)
1236                 return ret;
1237
1238         if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA)) {
1239                 down_write(&block_group->data_rwsem);
1240                 spin_lock(&block_group->lock);
1241                 if (block_group->delalloc_bytes) {
1242                         block_group->disk_cache_state = BTRFS_DC_WRITTEN;
1243                         spin_unlock(&block_group->lock);
1244                         up_write(&block_group->data_rwsem);
1245                         BTRFS_I(inode)->generation = 0;
1246                         ret = 0;
1247                         must_iput = 1;
1248                         goto out;
1249                 }
1250                 spin_unlock(&block_group->lock);
1251         }
1252
1253         /* Lock all pages first so we can lock the extent safely. */
1254         ret = io_ctl_prepare_pages(io_ctl, inode, 0);
1255         if (ret)
1256                 goto out;
1257
1258         lock_extent_bits(&BTRFS_I(inode)->io_tree, 0, i_size_read(inode) - 1,
1259                          &cached_state);
1260
1261         io_ctl_set_generation(io_ctl, trans->transid);
1262
1263         mutex_lock(&ctl->cache_writeout_mutex);
1264         /* Write out the extent entries in the free space cache */
1265         spin_lock(&ctl->tree_lock);
1266         ret = write_cache_extent_entries(io_ctl, ctl,
1267                                          block_group, &entries, &bitmaps,
1268                                          &bitmap_list);
1269         if (ret)
1270                 goto out_nospc_locked;
1271
1272         /*
1273          * Some spaces that are freed in the current transaction are pinned,
1274          * they will be added into free space cache after the transaction is
1275          * committed, we shouldn't lose them.
1276          *
1277          * If this changes while we are working we'll get added back to
1278          * the dirty list and redo it.  No locking needed
1279          */
1280         ret = write_pinned_extent_entries(root, block_group, io_ctl, &entries);
1281         if (ret)
1282                 goto out_nospc_locked;
1283
1284         /*
1285          * At last, we write out all the bitmaps and keep cache_writeout_mutex
1286          * locked while doing it because a concurrent trim can be manipulating
1287          * or freeing the bitmap.
1288          */
1289         ret = write_bitmap_entries(io_ctl, &bitmap_list);
1290         spin_unlock(&ctl->tree_lock);
1291         mutex_unlock(&ctl->cache_writeout_mutex);
1292         if (ret)
1293                 goto out_nospc;
1294
1295         /* Zero out the rest of the pages just to make sure */
1296         io_ctl_zero_remaining_pages(io_ctl);
1297
1298         /* Everything is written out, now we dirty the pages in the file. */
1299         ret = btrfs_dirty_pages(root, inode, io_ctl->pages, io_ctl->num_pages,
1300                                 0, i_size_read(inode), &cached_state);
1301         if (ret)
1302                 goto out_nospc;
1303
1304         if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA))
1305                 up_write(&block_group->data_rwsem);
1306         /*
1307          * Release the pages and unlock the extent, we will flush
1308          * them out later
1309          */
1310         io_ctl_drop_pages(io_ctl);
1311
1312         unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
1313                              i_size_read(inode) - 1, &cached_state, GFP_NOFS);
1314
1315         /*
1316          * at this point the pages are under IO and we're happy,
1317          * The caller is responsible for waiting on them and updating the
1318          * the cache and the inode
1319          */
1320         io_ctl->entries = entries;
1321         io_ctl->bitmaps = bitmaps;
1322
1323         ret = btrfs_fdatawrite_range(inode, 0, (u64)-1);
1324         if (ret)
1325                 goto out;
1326
1327         return 0;
1328
1329 out:
1330         io_ctl->inode = NULL;
1331         io_ctl_free(io_ctl);
1332         if (ret) {
1333                 invalidate_inode_pages2(inode->i_mapping);
1334                 BTRFS_I(inode)->generation = 0;
1335         }
1336         btrfs_update_inode(trans, root, inode);
1337         if (must_iput)
1338                 iput(inode);
1339         return ret;
1340
1341 out_nospc_locked:
1342         cleanup_bitmap_list(&bitmap_list);
1343         spin_unlock(&ctl->tree_lock);
1344         mutex_unlock(&ctl->cache_writeout_mutex);
1345
1346 out_nospc:
1347         cleanup_write_cache_enospc(inode, io_ctl, &cached_state, &bitmap_list);
1348
1349         if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA))
1350                 up_write(&block_group->data_rwsem);
1351
1352         goto out;
1353 }
1354
1355 int btrfs_write_out_cache(struct btrfs_root *root,
1356                           struct btrfs_trans_handle *trans,
1357                           struct btrfs_block_group_cache *block_group,
1358                           struct btrfs_path *path)
1359 {
1360         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
1361         struct inode *inode;
1362         int ret = 0;
1363
1364         root = root->fs_info->tree_root;
1365
1366         spin_lock(&block_group->lock);
1367         if (block_group->disk_cache_state < BTRFS_DC_SETUP) {
1368                 spin_unlock(&block_group->lock);
1369                 return 0;
1370         }
1371         spin_unlock(&block_group->lock);
1372
1373         inode = lookup_free_space_inode(root, block_group, path);
1374         if (IS_ERR(inode))
1375                 return 0;
1376
1377         ret = __btrfs_write_out_cache(root, inode, ctl, block_group,
1378                                       &block_group->io_ctl, trans,
1379                                       path, block_group->key.objectid);
1380         if (ret) {
1381 #ifdef DEBUG
1382                 btrfs_err(root->fs_info,
1383                         "failed to write free space cache for block group %llu",
1384                         block_group->key.objectid);
1385 #endif
1386                 spin_lock(&block_group->lock);
1387                 block_group->disk_cache_state = BTRFS_DC_ERROR;
1388                 spin_unlock(&block_group->lock);
1389
1390                 block_group->io_ctl.inode = NULL;
1391                 iput(inode);
1392         }
1393
1394         /*
1395          * if ret == 0 the caller is expected to call btrfs_wait_cache_io
1396          * to wait for IO and put the inode
1397          */
1398
1399         return ret;
1400 }
1401
1402 static inline unsigned long offset_to_bit(u64 bitmap_start, u32 unit,
1403                                           u64 offset)
1404 {
1405         ASSERT(offset >= bitmap_start);
1406         offset -= bitmap_start;
1407         return (unsigned long)(div_u64(offset, unit));
1408 }
1409
1410 static inline unsigned long bytes_to_bits(u64 bytes, u32 unit)
1411 {
1412         return (unsigned long)(div_u64(bytes, unit));
1413 }
1414
1415 static inline u64 offset_to_bitmap(struct btrfs_free_space_ctl *ctl,
1416                                    u64 offset)
1417 {
1418         u64 bitmap_start;
1419         u64 bytes_per_bitmap;
1420
1421         bytes_per_bitmap = BITS_PER_BITMAP * ctl->unit;
1422         bitmap_start = offset - ctl->start;
1423         bitmap_start = div64_u64(bitmap_start, bytes_per_bitmap);
1424         bitmap_start *= bytes_per_bitmap;
1425         bitmap_start += ctl->start;
1426
1427         return bitmap_start;
1428 }
1429
1430 static int tree_insert_offset(struct rb_root *root, u64 offset,
1431                               struct rb_node *node, int bitmap)
1432 {
1433         struct rb_node **p = &root->rb_node;
1434         struct rb_node *parent = NULL;
1435         struct btrfs_free_space *info;
1436
1437         while (*p) {
1438                 parent = *p;
1439                 info = rb_entry(parent, struct btrfs_free_space, offset_index);
1440
1441                 if (offset < info->offset) {
1442                         p = &(*p)->rb_left;
1443                 } else if (offset > info->offset) {
1444                         p = &(*p)->rb_right;
1445                 } else {
1446                         /*
1447                          * we could have a bitmap entry and an extent entry
1448                          * share the same offset.  If this is the case, we want
1449                          * the extent entry to always be found first if we do a
1450                          * linear search through the tree, since we want to have
1451                          * the quickest allocation time, and allocating from an
1452                          * extent is faster than allocating from a bitmap.  So
1453                          * if we're inserting a bitmap and we find an entry at
1454                          * this offset, we want to go right, or after this entry
1455                          * logically.  If we are inserting an extent and we've
1456                          * found a bitmap, we want to go left, or before
1457                          * logically.
1458                          */
1459                         if (bitmap) {
1460                                 if (info->bitmap) {
1461                                         WARN_ON_ONCE(1);
1462                                         return -EEXIST;
1463                                 }
1464                                 p = &(*p)->rb_right;
1465                         } else {
1466                                 if (!info->bitmap) {
1467                                         WARN_ON_ONCE(1);
1468                                         return -EEXIST;
1469                                 }
1470                                 p = &(*p)->rb_left;
1471                         }
1472                 }
1473         }
1474
1475         rb_link_node(node, parent, p);
1476         rb_insert_color(node, root);
1477
1478         return 0;
1479 }
1480
1481 /*
1482  * searches the tree for the given offset.
1483  *
1484  * fuzzy - If this is set, then we are trying to make an allocation, and we just
1485  * want a section that has at least bytes size and comes at or after the given
1486  * offset.
1487  */
1488 static struct btrfs_free_space *
1489 tree_search_offset(struct btrfs_free_space_ctl *ctl,
1490                    u64 offset, int bitmap_only, int fuzzy)
1491 {
1492         struct rb_node *n = ctl->free_space_offset.rb_node;
1493         struct btrfs_free_space *entry, *prev = NULL;
1494
1495         /* find entry that is closest to the 'offset' */
1496         while (1) {
1497                 if (!n) {
1498                         entry = NULL;
1499                         break;
1500                 }
1501
1502                 entry = rb_entry(n, struct btrfs_free_space, offset_index);
1503                 prev = entry;
1504
1505                 if (offset < entry->offset)
1506                         n = n->rb_left;
1507                 else if (offset > entry->offset)
1508                         n = n->rb_right;
1509                 else
1510                         break;
1511         }
1512
1513         if (bitmap_only) {
1514                 if (!entry)
1515                         return NULL;
1516                 if (entry->bitmap)
1517                         return entry;
1518
1519                 /*
1520                  * bitmap entry and extent entry may share same offset,
1521                  * in that case, bitmap entry comes after extent entry.
1522                  */
1523                 n = rb_next(n);
1524                 if (!n)
1525                         return NULL;
1526                 entry = rb_entry(n, struct btrfs_free_space, offset_index);
1527                 if (entry->offset != offset)
1528                         return NULL;
1529
1530                 WARN_ON(!entry->bitmap);
1531                 return entry;
1532         } else if (entry) {
1533                 if (entry->bitmap) {
1534                         /*
1535                          * if previous extent entry covers the offset,
1536                          * we should return it instead of the bitmap entry
1537                          */
1538                         n = rb_prev(&entry->offset_index);
1539                         if (n) {
1540                                 prev = rb_entry(n, struct btrfs_free_space,
1541                                                 offset_index);
1542                                 if (!prev->bitmap &&
1543                                     prev->offset + prev->bytes > offset)
1544                                         entry = prev;
1545                         }
1546                 }
1547                 return entry;
1548         }
1549
1550         if (!prev)
1551                 return NULL;
1552
1553         /* find last entry before the 'offset' */
1554         entry = prev;
1555         if (entry->offset > offset) {
1556                 n = rb_prev(&entry->offset_index);
1557                 if (n) {
1558                         entry = rb_entry(n, struct btrfs_free_space,
1559                                         offset_index);
1560                         ASSERT(entry->offset <= offset);
1561                 } else {
1562                         if (fuzzy)
1563                                 return entry;
1564                         else
1565                                 return NULL;
1566                 }
1567         }
1568
1569         if (entry->bitmap) {
1570                 n = rb_prev(&entry->offset_index);
1571                 if (n) {
1572                         prev = rb_entry(n, struct btrfs_free_space,
1573                                         offset_index);
1574                         if (!prev->bitmap &&
1575                             prev->offset + prev->bytes > offset)
1576                                 return prev;
1577                 }
1578                 if (entry->offset + BITS_PER_BITMAP * ctl->unit > offset)
1579                         return entry;
1580         } else if (entry->offset + entry->bytes > offset)
1581                 return entry;
1582
1583         if (!fuzzy)
1584                 return NULL;
1585
1586         while (1) {
1587                 if (entry->bitmap) {
1588                         if (entry->offset + BITS_PER_BITMAP *
1589                             ctl->unit > offset)
1590                                 break;
1591                 } else {
1592                         if (entry->offset + entry->bytes > offset)
1593                                 break;
1594                 }
1595
1596                 n = rb_next(&entry->offset_index);
1597                 if (!n)
1598                         return NULL;
1599                 entry = rb_entry(n, struct btrfs_free_space, offset_index);
1600         }
1601         return entry;
1602 }
1603
1604 static inline void
1605 __unlink_free_space(struct btrfs_free_space_ctl *ctl,
1606                     struct btrfs_free_space *info)
1607 {
1608         rb_erase(&info->offset_index, &ctl->free_space_offset);
1609         ctl->free_extents--;
1610 }
1611
1612 static void unlink_free_space(struct btrfs_free_space_ctl *ctl,
1613                               struct btrfs_free_space *info)
1614 {
1615         __unlink_free_space(ctl, info);
1616         ctl->free_space -= info->bytes;
1617 }
1618
1619 static int link_free_space(struct btrfs_free_space_ctl *ctl,
1620                            struct btrfs_free_space *info)
1621 {
1622         int ret = 0;
1623
1624         ASSERT(info->bytes || info->bitmap);
1625         ret = tree_insert_offset(&ctl->free_space_offset, info->offset,
1626                                  &info->offset_index, (info->bitmap != NULL));
1627         if (ret)
1628                 return ret;
1629
1630         ctl->free_space += info->bytes;
1631         ctl->free_extents++;
1632         return ret;
1633 }
1634
1635 static void recalculate_thresholds(struct btrfs_free_space_ctl *ctl)
1636 {
1637         struct btrfs_block_group_cache *block_group = ctl->private;
1638         u64 max_bytes;
1639         u64 bitmap_bytes;
1640         u64 extent_bytes;
1641         u64 size = block_group->key.offset;
1642         u64 bytes_per_bg = BITS_PER_BITMAP * ctl->unit;
1643         u64 max_bitmaps = div64_u64(size + bytes_per_bg - 1, bytes_per_bg);
1644
1645         max_bitmaps = max_t(u64, max_bitmaps, 1);
1646
1647         ASSERT(ctl->total_bitmaps <= max_bitmaps);
1648
1649         /*
1650          * The goal is to keep the total amount of memory used per 1gb of space
1651          * at or below 32k, so we need to adjust how much memory we allow to be
1652          * used by extent based free space tracking
1653          */
1654         if (size < SZ_1G)
1655                 max_bytes = MAX_CACHE_BYTES_PER_GIG;
1656         else
1657                 max_bytes = MAX_CACHE_BYTES_PER_GIG * div_u64(size, SZ_1G);
1658
1659         /*
1660          * we want to account for 1 more bitmap than what we have so we can make
1661          * sure we don't go over our overall goal of MAX_CACHE_BYTES_PER_GIG as
1662          * we add more bitmaps.
1663          */
1664         bitmap_bytes = (ctl->total_bitmaps + 1) * ctl->unit;
1665
1666         if (bitmap_bytes >= max_bytes) {
1667                 ctl->extents_thresh = 0;
1668                 return;
1669         }
1670
1671         /*
1672          * we want the extent entry threshold to always be at most 1/2 the max
1673          * bytes we can have, or whatever is less than that.
1674          */
1675         extent_bytes = max_bytes - bitmap_bytes;
1676         extent_bytes = min_t(u64, extent_bytes, max_bytes >> 1);
1677
1678         ctl->extents_thresh =
1679                 div_u64(extent_bytes, sizeof(struct btrfs_free_space));
1680 }
1681
1682 static inline void __bitmap_clear_bits(struct btrfs_free_space_ctl *ctl,
1683                                        struct btrfs_free_space *info,
1684                                        u64 offset, u64 bytes)
1685 {
1686         unsigned long start, count;
1687
1688         start = offset_to_bit(info->offset, ctl->unit, offset);
1689         count = bytes_to_bits(bytes, ctl->unit);
1690         ASSERT(start + count <= BITS_PER_BITMAP);
1691
1692         bitmap_clear(info->bitmap, start, count);
1693
1694         info->bytes -= bytes;
1695 }
1696
1697 static void bitmap_clear_bits(struct btrfs_free_space_ctl *ctl,
1698                               struct btrfs_free_space *info, u64 offset,
1699                               u64 bytes)
1700 {
1701         __bitmap_clear_bits(ctl, info, offset, bytes);
1702         ctl->free_space -= bytes;
1703 }
1704
1705 static void bitmap_set_bits(struct btrfs_free_space_ctl *ctl,
1706                             struct btrfs_free_space *info, u64 offset,
1707                             u64 bytes)
1708 {
1709         unsigned long start, count;
1710
1711         start = offset_to_bit(info->offset, ctl->unit, offset);
1712         count = bytes_to_bits(bytes, ctl->unit);
1713         ASSERT(start + count <= BITS_PER_BITMAP);
1714
1715         bitmap_set(info->bitmap, start, count);
1716
1717         info->bytes += bytes;
1718         ctl->free_space += bytes;
1719 }
1720
1721 /*
1722  * If we can not find suitable extent, we will use bytes to record
1723  * the size of the max extent.
1724  */
1725 static int search_bitmap(struct btrfs_free_space_ctl *ctl,
1726                          struct btrfs_free_space *bitmap_info, u64 *offset,
1727                          u64 *bytes, bool for_alloc)
1728 {
1729         unsigned long found_bits = 0;
1730         unsigned long max_bits = 0;
1731         unsigned long bits, i;
1732         unsigned long next_zero;
1733         unsigned long extent_bits;
1734
1735         /*
1736          * Skip searching the bitmap if we don't have a contiguous section that
1737          * is large enough for this allocation.
1738          */
1739         if (for_alloc &&
1740             bitmap_info->max_extent_size &&
1741             bitmap_info->max_extent_size < *bytes) {
1742                 *bytes = bitmap_info->max_extent_size;
1743                 return -1;
1744         }
1745
1746         i = offset_to_bit(bitmap_info->offset, ctl->unit,
1747                           max_t(u64, *offset, bitmap_info->offset));
1748         bits = bytes_to_bits(*bytes, ctl->unit);
1749
1750         for_each_set_bit_from(i, bitmap_info->bitmap, BITS_PER_BITMAP) {
1751                 if (for_alloc && bits == 1) {
1752                         found_bits = 1;
1753                         break;
1754                 }
1755                 next_zero = find_next_zero_bit(bitmap_info->bitmap,
1756                                                BITS_PER_BITMAP, i);
1757                 extent_bits = next_zero - i;
1758                 if (extent_bits >= bits) {
1759                         found_bits = extent_bits;
1760                         break;
1761                 } else if (extent_bits > max_bits) {
1762                         max_bits = extent_bits;
1763                 }
1764                 i = next_zero;
1765         }
1766
1767         if (found_bits) {
1768                 *offset = (u64)(i * ctl->unit) + bitmap_info->offset;
1769                 *bytes = (u64)(found_bits) * ctl->unit;
1770                 return 0;
1771         }
1772
1773         *bytes = (u64)(max_bits) * ctl->unit;
1774         bitmap_info->max_extent_size = *bytes;
1775         return -1;
1776 }
1777
1778 /* Cache the size of the max extent in bytes */
1779 static struct btrfs_free_space *
1780 find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes,
1781                 unsigned long align, u64 *max_extent_size)
1782 {
1783         struct btrfs_free_space *entry;
1784         struct rb_node *node;
1785         u64 tmp;
1786         u64 align_off;
1787         int ret;
1788
1789         if (!ctl->free_space_offset.rb_node)
1790                 goto out;
1791
1792         entry = tree_search_offset(ctl, offset_to_bitmap(ctl, *offset), 0, 1);
1793         if (!entry)
1794                 goto out;
1795
1796         for (node = &entry->offset_index; node; node = rb_next(node)) {
1797                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
1798                 if (entry->bytes < *bytes) {
1799                         if (entry->bytes > *max_extent_size)
1800                                 *max_extent_size = entry->bytes;
1801                         continue;
1802                 }
1803
1804                 /* make sure the space returned is big enough
1805                  * to match our requested alignment
1806                  */
1807                 if (*bytes >= align) {
1808                         tmp = entry->offset - ctl->start + align - 1;
1809                         tmp = div64_u64(tmp, align);
1810                         tmp = tmp * align + ctl->start;
1811                         align_off = tmp - entry->offset;
1812                 } else {
1813                         align_off = 0;
1814                         tmp = entry->offset;
1815                 }
1816
1817                 if (entry->bytes < *bytes + align_off) {
1818                         if (entry->bytes > *max_extent_size)
1819                                 *max_extent_size = entry->bytes;
1820                         continue;
1821                 }
1822
1823                 if (entry->bitmap) {
1824                         u64 size = *bytes;
1825
1826                         ret = search_bitmap(ctl, entry, &tmp, &size, true);
1827                         if (!ret) {
1828                                 *offset = tmp;
1829                                 *bytes = size;
1830                                 return entry;
1831                         } else if (size > *max_extent_size) {
1832                                 *max_extent_size = size;
1833                         }
1834                         continue;
1835                 }
1836
1837                 *offset = tmp;
1838                 *bytes = entry->bytes - align_off;
1839                 return entry;
1840         }
1841 out:
1842         return NULL;
1843 }
1844
1845 static void add_new_bitmap(struct btrfs_free_space_ctl *ctl,
1846                            struct btrfs_free_space *info, u64 offset)
1847 {
1848         info->offset = offset_to_bitmap(ctl, offset);
1849         info->bytes = 0;
1850         INIT_LIST_HEAD(&info->list);
1851         link_free_space(ctl, info);
1852         ctl->total_bitmaps++;
1853
1854         ctl->op->recalc_thresholds(ctl);
1855 }
1856
1857 static void free_bitmap(struct btrfs_free_space_ctl *ctl,
1858                         struct btrfs_free_space *bitmap_info)
1859 {
1860         unlink_free_space(ctl, bitmap_info);
1861         kfree(bitmap_info->bitmap);
1862         kmem_cache_free(btrfs_free_space_cachep, bitmap_info);
1863         ctl->total_bitmaps--;
1864         ctl->op->recalc_thresholds(ctl);
1865 }
1866
1867 static noinline int remove_from_bitmap(struct btrfs_free_space_ctl *ctl,
1868                               struct btrfs_free_space *bitmap_info,
1869                               u64 *offset, u64 *bytes)
1870 {
1871         u64 end;
1872         u64 search_start, search_bytes;
1873         int ret;
1874
1875 again:
1876         end = bitmap_info->offset + (u64)(BITS_PER_BITMAP * ctl->unit) - 1;
1877
1878         /*
1879          * We need to search for bits in this bitmap.  We could only cover some
1880          * of the extent in this bitmap thanks to how we add space, so we need
1881          * to search for as much as it as we can and clear that amount, and then
1882          * go searching for the next bit.
1883          */
1884         search_start = *offset;
1885         search_bytes = ctl->unit;
1886         search_bytes = min(search_bytes, end - search_start + 1);
1887         ret = search_bitmap(ctl, bitmap_info, &search_start, &search_bytes,
1888                             false);
1889         if (ret < 0 || search_start != *offset)
1890                 return -EINVAL;
1891
1892         /* We may have found more bits than what we need */
1893         search_bytes = min(search_bytes, *bytes);
1894
1895         /* Cannot clear past the end of the bitmap */
1896         search_bytes = min(search_bytes, end - search_start + 1);
1897
1898         bitmap_clear_bits(ctl, bitmap_info, search_start, search_bytes);
1899         *offset += search_bytes;
1900         *bytes -= search_bytes;
1901
1902         if (*bytes) {
1903                 struct rb_node *next = rb_next(&bitmap_info->offset_index);
1904                 if (!bitmap_info->bytes)
1905                         free_bitmap(ctl, bitmap_info);
1906
1907                 /*
1908                  * no entry after this bitmap, but we still have bytes to
1909                  * remove, so something has gone wrong.
1910                  */
1911                 if (!next)
1912                         return -EINVAL;
1913
1914                 bitmap_info = rb_entry(next, struct btrfs_free_space,
1915                                        offset_index);
1916
1917                 /*
1918                  * if the next entry isn't a bitmap we need to return to let the
1919                  * extent stuff do its work.
1920                  */
1921                 if (!bitmap_info->bitmap)
1922                         return -EAGAIN;
1923
1924                 /*
1925                  * Ok the next item is a bitmap, but it may not actually hold
1926                  * the information for the rest of this free space stuff, so
1927                  * look for it, and if we don't find it return so we can try
1928                  * everything over again.
1929                  */
1930                 search_start = *offset;
1931                 search_bytes = ctl->unit;
1932                 ret = search_bitmap(ctl, bitmap_info, &search_start,
1933                                     &search_bytes, false);
1934                 if (ret < 0 || search_start != *offset)
1935                         return -EAGAIN;
1936
1937                 goto again;
1938         } else if (!bitmap_info->bytes)
1939                 free_bitmap(ctl, bitmap_info);
1940
1941         return 0;
1942 }
1943
1944 static u64 add_bytes_to_bitmap(struct btrfs_free_space_ctl *ctl,
1945                                struct btrfs_free_space *info, u64 offset,
1946                                u64 bytes)
1947 {
1948         u64 bytes_to_set = 0;
1949         u64 end;
1950
1951         end = info->offset + (u64)(BITS_PER_BITMAP * ctl->unit);
1952
1953         bytes_to_set = min(end - offset, bytes);
1954
1955         bitmap_set_bits(ctl, info, offset, bytes_to_set);
1956
1957         /*
1958          * We set some bytes, we have no idea what the max extent size is
1959          * anymore.
1960          */
1961         info->max_extent_size = 0;
1962
1963         return bytes_to_set;
1964
1965 }
1966
1967 static bool use_bitmap(struct btrfs_free_space_ctl *ctl,
1968                       struct btrfs_free_space *info)
1969 {
1970         struct btrfs_block_group_cache *block_group = ctl->private;
1971         bool forced = false;
1972
1973 #ifdef CONFIG_BTRFS_DEBUG
1974         if (btrfs_should_fragment_free_space(block_group->fs_info->extent_root,
1975                                              block_group))
1976                 forced = true;
1977 #endif
1978
1979         /*
1980          * If we are below the extents threshold then we can add this as an
1981          * extent, and don't have to deal with the bitmap
1982          */
1983         if (!forced && ctl->free_extents < ctl->extents_thresh) {
1984                 /*
1985                  * If this block group has some small extents we don't want to
1986                  * use up all of our free slots in the cache with them, we want
1987                  * to reserve them to larger extents, however if we have plenty
1988                  * of cache left then go ahead an dadd them, no sense in adding
1989                  * the overhead of a bitmap if we don't have to.
1990                  */
1991                 if (info->bytes <= block_group->sectorsize * 4) {
1992                         if (ctl->free_extents * 2 <= ctl->extents_thresh)
1993                                 return false;
1994                 } else {
1995                         return false;
1996                 }
1997         }
1998
1999         /*
2000          * The original block groups from mkfs can be really small, like 8
2001          * megabytes, so don't bother with a bitmap for those entries.  However
2002          * some block groups can be smaller than what a bitmap would cover but
2003          * are still large enough that they could overflow the 32k memory limit,
2004          * so allow those block groups to still be allowed to have a bitmap
2005          * entry.
2006          */
2007         if (((BITS_PER_BITMAP * ctl->unit) >> 1) > block_group->key.offset)
2008                 return false;
2009
2010         return true;
2011 }
2012
2013 static const struct btrfs_free_space_op free_space_op = {
2014         .recalc_thresholds      = recalculate_thresholds,
2015         .use_bitmap             = use_bitmap,
2016 };
2017
2018 static int insert_into_bitmap(struct btrfs_free_space_ctl *ctl,
2019                               struct btrfs_free_space *info)
2020 {
2021         struct btrfs_free_space *bitmap_info;
2022         struct btrfs_block_group_cache *block_group = NULL;
2023         int added = 0;
2024         u64 bytes, offset, bytes_added;
2025         int ret;
2026
2027         bytes = info->bytes;
2028         offset = info->offset;
2029
2030         if (!ctl->op->use_bitmap(ctl, info))
2031                 return 0;
2032
2033         if (ctl->op == &free_space_op)
2034                 block_group = ctl->private;
2035 again:
2036         /*
2037          * Since we link bitmaps right into the cluster we need to see if we
2038          * have a cluster here, and if so and it has our bitmap we need to add
2039          * the free space to that bitmap.
2040          */
2041         if (block_group && !list_empty(&block_group->cluster_list)) {
2042                 struct btrfs_free_cluster *cluster;
2043                 struct rb_node *node;
2044                 struct btrfs_free_space *entry;
2045
2046                 cluster = list_entry(block_group->cluster_list.next,
2047                                      struct btrfs_free_cluster,
2048                                      block_group_list);
2049                 spin_lock(&cluster->lock);
2050                 node = rb_first(&cluster->root);
2051                 if (!node) {
2052                         spin_unlock(&cluster->lock);
2053                         goto no_cluster_bitmap;
2054                 }
2055
2056                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2057                 if (!entry->bitmap) {
2058                         spin_unlock(&cluster->lock);
2059                         goto no_cluster_bitmap;
2060                 }
2061
2062                 if (entry->offset == offset_to_bitmap(ctl, offset)) {
2063                         bytes_added = add_bytes_to_bitmap(ctl, entry,
2064                                                           offset, bytes);
2065                         bytes -= bytes_added;
2066                         offset += bytes_added;
2067                 }
2068                 spin_unlock(&cluster->lock);
2069                 if (!bytes) {
2070                         ret = 1;
2071                         goto out;
2072                 }
2073         }
2074
2075 no_cluster_bitmap:
2076         bitmap_info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
2077                                          1, 0);
2078         if (!bitmap_info) {
2079                 ASSERT(added == 0);
2080                 goto new_bitmap;
2081         }
2082
2083         bytes_added = add_bytes_to_bitmap(ctl, bitmap_info, offset, bytes);
2084         bytes -= bytes_added;
2085         offset += bytes_added;
2086         added = 0;
2087
2088         if (!bytes) {
2089                 ret = 1;
2090                 goto out;
2091         } else
2092                 goto again;
2093
2094 new_bitmap:
2095         if (info && info->bitmap) {
2096                 add_new_bitmap(ctl, info, offset);
2097                 added = 1;
2098                 info = NULL;
2099                 goto again;
2100         } else {
2101                 spin_unlock(&ctl->tree_lock);
2102
2103                 /* no pre-allocated info, allocate a new one */
2104                 if (!info) {
2105                         info = kmem_cache_zalloc(btrfs_free_space_cachep,
2106                                                  GFP_NOFS);
2107                         if (!info) {
2108                                 spin_lock(&ctl->tree_lock);
2109                                 ret = -ENOMEM;
2110                                 goto out;
2111                         }
2112                 }
2113
2114                 /* allocate the bitmap */
2115                 info->bitmap = kzalloc(PAGE_SIZE, GFP_NOFS);
2116                 spin_lock(&ctl->tree_lock);
2117                 if (!info->bitmap) {
2118                         ret = -ENOMEM;
2119                         goto out;
2120                 }
2121                 goto again;
2122         }
2123
2124 out:
2125         if (info) {
2126                 if (info->bitmap)
2127                         kfree(info->bitmap);
2128                 kmem_cache_free(btrfs_free_space_cachep, info);
2129         }
2130
2131         return ret;
2132 }
2133
2134 static bool try_merge_free_space(struct btrfs_free_space_ctl *ctl,
2135                           struct btrfs_free_space *info, bool update_stat)
2136 {
2137         struct btrfs_free_space *left_info;
2138         struct btrfs_free_space *right_info;
2139         bool merged = false;
2140         u64 offset = info->offset;
2141         u64 bytes = info->bytes;
2142
2143         /*
2144          * first we want to see if there is free space adjacent to the range we
2145          * are adding, if there is remove that struct and add a new one to
2146          * cover the entire range
2147          */
2148         right_info = tree_search_offset(ctl, offset + bytes, 0, 0);
2149         if (right_info && rb_prev(&right_info->offset_index))
2150                 left_info = rb_entry(rb_prev(&right_info->offset_index),
2151                                      struct btrfs_free_space, offset_index);
2152         else
2153                 left_info = tree_search_offset(ctl, offset - 1, 0, 0);
2154
2155         if (right_info && !right_info->bitmap) {
2156                 if (update_stat)
2157                         unlink_free_space(ctl, right_info);
2158                 else
2159                         __unlink_free_space(ctl, right_info);
2160                 info->bytes += right_info->bytes;
2161                 kmem_cache_free(btrfs_free_space_cachep, right_info);
2162                 merged = true;
2163         }
2164
2165         if (left_info && !left_info->bitmap &&
2166             left_info->offset + left_info->bytes == offset) {
2167                 if (update_stat)
2168                         unlink_free_space(ctl, left_info);
2169                 else
2170                         __unlink_free_space(ctl, left_info);
2171                 info->offset = left_info->offset;
2172                 info->bytes += left_info->bytes;
2173                 kmem_cache_free(btrfs_free_space_cachep, left_info);
2174                 merged = true;
2175         }
2176
2177         return merged;
2178 }
2179
2180 static bool steal_from_bitmap_to_end(struct btrfs_free_space_ctl *ctl,
2181                                      struct btrfs_free_space *info,
2182                                      bool update_stat)
2183 {
2184         struct btrfs_free_space *bitmap;
2185         unsigned long i;
2186         unsigned long j;
2187         const u64 end = info->offset + info->bytes;
2188         const u64 bitmap_offset = offset_to_bitmap(ctl, end);
2189         u64 bytes;
2190
2191         bitmap = tree_search_offset(ctl, bitmap_offset, 1, 0);
2192         if (!bitmap)
2193                 return false;
2194
2195         i = offset_to_bit(bitmap->offset, ctl->unit, end);
2196         j = find_next_zero_bit(bitmap->bitmap, BITS_PER_BITMAP, i);
2197         if (j == i)
2198                 return false;
2199         bytes = (j - i) * ctl->unit;
2200         info->bytes += bytes;
2201
2202         if (update_stat)
2203                 bitmap_clear_bits(ctl, bitmap, end, bytes);
2204         else
2205                 __bitmap_clear_bits(ctl, bitmap, end, bytes);
2206
2207         if (!bitmap->bytes)
2208                 free_bitmap(ctl, bitmap);
2209
2210         return true;
2211 }
2212
2213 static bool steal_from_bitmap_to_front(struct btrfs_free_space_ctl *ctl,
2214                                        struct btrfs_free_space *info,
2215                                        bool update_stat)
2216 {
2217         struct btrfs_free_space *bitmap;
2218         u64 bitmap_offset;
2219         unsigned long i;
2220         unsigned long j;
2221         unsigned long prev_j;
2222         u64 bytes;
2223
2224         bitmap_offset = offset_to_bitmap(ctl, info->offset);
2225         /* If we're on a boundary, try the previous logical bitmap. */
2226         if (bitmap_offset == info->offset) {
2227                 if (info->offset == 0)
2228                         return false;
2229                 bitmap_offset = offset_to_bitmap(ctl, info->offset - 1);
2230         }
2231
2232         bitmap = tree_search_offset(ctl, bitmap_offset, 1, 0);
2233         if (!bitmap)
2234                 return false;
2235
2236         i = offset_to_bit(bitmap->offset, ctl->unit, info->offset) - 1;
2237         j = 0;
2238         prev_j = (unsigned long)-1;
2239         for_each_clear_bit_from(j, bitmap->bitmap, BITS_PER_BITMAP) {
2240                 if (j > i)
2241                         break;
2242                 prev_j = j;
2243         }
2244         if (prev_j == i)
2245                 return false;
2246
2247         if (prev_j == (unsigned long)-1)
2248                 bytes = (i + 1) * ctl->unit;
2249         else
2250                 bytes = (i - prev_j) * ctl->unit;
2251
2252         info->offset -= bytes;
2253         info->bytes += bytes;
2254
2255         if (update_stat)
2256                 bitmap_clear_bits(ctl, bitmap, info->offset, bytes);
2257         else
2258                 __bitmap_clear_bits(ctl, bitmap, info->offset, bytes);
2259
2260         if (!bitmap->bytes)
2261                 free_bitmap(ctl, bitmap);
2262
2263         return true;
2264 }
2265
2266 /*
2267  * We prefer always to allocate from extent entries, both for clustered and
2268  * non-clustered allocation requests. So when attempting to add a new extent
2269  * entry, try to see if there's adjacent free space in bitmap entries, and if
2270  * there is, migrate that space from the bitmaps to the extent.
2271  * Like this we get better chances of satisfying space allocation requests
2272  * because we attempt to satisfy them based on a single cache entry, and never
2273  * on 2 or more entries - even if the entries represent a contiguous free space
2274  * region (e.g. 1 extent entry + 1 bitmap entry starting where the extent entry
2275  * ends).
2276  */
2277 static void steal_from_bitmap(struct btrfs_free_space_ctl *ctl,
2278                               struct btrfs_free_space *info,
2279                               bool update_stat)
2280 {
2281         /*
2282          * Only work with disconnected entries, as we can change their offset,
2283          * and must be extent entries.
2284          */
2285         ASSERT(!info->bitmap);
2286         ASSERT(RB_EMPTY_NODE(&info->offset_index));
2287
2288         if (ctl->total_bitmaps > 0) {
2289                 bool stole_end;
2290                 bool stole_front = false;
2291
2292                 stole_end = steal_from_bitmap_to_end(ctl, info, update_stat);
2293                 if (ctl->total_bitmaps > 0)
2294                         stole_front = steal_from_bitmap_to_front(ctl, info,
2295                                                                  update_stat);
2296
2297                 if (stole_end || stole_front)
2298                         try_merge_free_space(ctl, info, update_stat);
2299         }
2300 }
2301
2302 int __btrfs_add_free_space(struct btrfs_fs_info *fs_info,
2303                            struct btrfs_free_space_ctl *ctl,
2304                            u64 offset, u64 bytes)
2305 {
2306         struct btrfs_free_space *info;
2307         int ret = 0;
2308
2309         info = kmem_cache_zalloc(btrfs_free_space_cachep, GFP_NOFS);
2310         if (!info)
2311                 return -ENOMEM;
2312
2313         info->offset = offset;
2314         info->bytes = bytes;
2315         RB_CLEAR_NODE(&info->offset_index);
2316
2317         spin_lock(&ctl->tree_lock);
2318
2319         if (try_merge_free_space(ctl, info, true))
2320                 goto link;
2321
2322         /*
2323          * There was no extent directly to the left or right of this new
2324          * extent then we know we're going to have to allocate a new extent, so
2325          * before we do that see if we need to drop this into a bitmap
2326          */
2327         ret = insert_into_bitmap(ctl, info);
2328         if (ret < 0) {
2329                 goto out;
2330         } else if (ret) {
2331                 ret = 0;
2332                 goto out;
2333         }
2334 link:
2335         /*
2336          * Only steal free space from adjacent bitmaps if we're sure we're not
2337          * going to add the new free space to existing bitmap entries - because
2338          * that would mean unnecessary work that would be reverted. Therefore
2339          * attempt to steal space from bitmaps if we're adding an extent entry.
2340          */
2341         steal_from_bitmap(ctl, info, true);
2342
2343         ret = link_free_space(ctl, info);
2344         if (ret)
2345                 kmem_cache_free(btrfs_free_space_cachep, info);
2346 out:
2347         spin_unlock(&ctl->tree_lock);
2348
2349         if (ret) {
2350                 btrfs_crit(fs_info, "unable to add free space :%d", ret);
2351                 ASSERT(ret != -EEXIST);
2352         }
2353
2354         return ret;
2355 }
2356
2357 int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
2358                             u64 offset, u64 bytes)
2359 {
2360         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2361         struct btrfs_free_space *info;
2362         int ret;
2363         bool re_search = false;
2364
2365         spin_lock(&ctl->tree_lock);
2366
2367 again:
2368         ret = 0;
2369         if (!bytes)
2370                 goto out_lock;
2371
2372         info = tree_search_offset(ctl, offset, 0, 0);
2373         if (!info) {
2374                 /*
2375                  * oops didn't find an extent that matched the space we wanted
2376                  * to remove, look for a bitmap instead
2377                  */
2378                 info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
2379                                           1, 0);
2380                 if (!info) {
2381                         /*
2382                          * If we found a partial bit of our free space in a
2383                          * bitmap but then couldn't find the other part this may
2384                          * be a problem, so WARN about it.
2385                          */
2386                         WARN_ON(re_search);
2387                         goto out_lock;
2388                 }
2389         }
2390
2391         re_search = false;
2392         if (!info->bitmap) {
2393                 unlink_free_space(ctl, info);
2394                 if (offset == info->offset) {
2395                         u64 to_free = min(bytes, info->bytes);
2396
2397                         info->bytes -= to_free;
2398                         info->offset += to_free;
2399                         if (info->bytes) {
2400                                 ret = link_free_space(ctl, info);
2401                                 WARN_ON(ret);
2402                         } else {
2403                                 kmem_cache_free(btrfs_free_space_cachep, info);
2404                         }
2405
2406                         offset += to_free;
2407                         bytes -= to_free;
2408                         goto again;
2409                 } else {
2410                         u64 old_end = info->bytes + info->offset;
2411
2412                         info->bytes = offset - info->offset;
2413                         ret = link_free_space(ctl, info);
2414                         WARN_ON(ret);
2415                         if (ret)
2416                                 goto out_lock;
2417
2418                         /* Not enough bytes in this entry to satisfy us */
2419                         if (old_end < offset + bytes) {
2420                                 bytes -= old_end - offset;
2421                                 offset = old_end;
2422                                 goto again;
2423                         } else if (old_end == offset + bytes) {
2424                                 /* all done */
2425                                 goto out_lock;
2426                         }
2427                         spin_unlock(&ctl->tree_lock);
2428
2429                         ret = btrfs_add_free_space(block_group, offset + bytes,
2430                                                    old_end - (offset + bytes));
2431                         WARN_ON(ret);
2432                         goto out;
2433                 }
2434         }
2435
2436         ret = remove_from_bitmap(ctl, info, &offset, &bytes);
2437         if (ret == -EAGAIN) {
2438                 re_search = true;
2439                 goto again;
2440         }
2441 out_lock:
2442         spin_unlock(&ctl->tree_lock);
2443 out:
2444         return ret;
2445 }
2446
2447 void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
2448                            u64 bytes)
2449 {
2450         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2451         struct btrfs_free_space *info;
2452         struct rb_node *n;
2453         int count = 0;
2454
2455         for (n = rb_first(&ctl->free_space_offset); n; n = rb_next(n)) {
2456                 info = rb_entry(n, struct btrfs_free_space, offset_index);
2457                 if (info->bytes >= bytes && !block_group->ro)
2458                         count++;
2459                 btrfs_crit(block_group->fs_info,
2460                            "entry offset %llu, bytes %llu, bitmap %s",
2461                            info->offset, info->bytes,
2462                        (info->bitmap) ? "yes" : "no");
2463         }
2464         btrfs_info(block_group->fs_info, "block group has cluster?: %s",
2465                list_empty(&block_group->cluster_list) ? "no" : "yes");
2466         btrfs_info(block_group->fs_info,
2467                    "%d blocks of free space at or bigger than bytes is", count);
2468 }
2469
2470 void btrfs_init_free_space_ctl(struct btrfs_block_group_cache *block_group)
2471 {
2472         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2473
2474         spin_lock_init(&ctl->tree_lock);
2475         ctl->unit = block_group->sectorsize;
2476         ctl->start = block_group->key.objectid;
2477         ctl->private = block_group;
2478         ctl->op = &free_space_op;
2479         INIT_LIST_HEAD(&ctl->trimming_ranges);
2480         mutex_init(&ctl->cache_writeout_mutex);
2481
2482         /*
2483          * we only want to have 32k of ram per block group for keeping
2484          * track of free space, and if we pass 1/2 of that we want to
2485          * start converting things over to using bitmaps
2486          */
2487         ctl->extents_thresh = (SZ_32K / 2) / sizeof(struct btrfs_free_space);
2488 }
2489
2490 /*
2491  * for a given cluster, put all of its extents back into the free
2492  * space cache.  If the block group passed doesn't match the block group
2493  * pointed to by the cluster, someone else raced in and freed the
2494  * cluster already.  In that case, we just return without changing anything
2495  */
2496 static int
2497 __btrfs_return_cluster_to_free_space(
2498                              struct btrfs_block_group_cache *block_group,
2499                              struct btrfs_free_cluster *cluster)
2500 {
2501         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2502         struct btrfs_free_space *entry;
2503         struct rb_node *node;
2504
2505         spin_lock(&cluster->lock);
2506         if (cluster->block_group != block_group)
2507                 goto out;
2508
2509         cluster->block_group = NULL;
2510         cluster->window_start = 0;
2511         list_del_init(&cluster->block_group_list);
2512
2513         node = rb_first(&cluster->root);
2514         while (node) {
2515                 bool bitmap;
2516
2517                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2518                 node = rb_next(&entry->offset_index);
2519                 rb_erase(&entry->offset_index, &cluster->root);
2520                 RB_CLEAR_NODE(&entry->offset_index);
2521
2522                 bitmap = (entry->bitmap != NULL);
2523                 if (!bitmap) {
2524                         try_merge_free_space(ctl, entry, false);
2525                         steal_from_bitmap(ctl, entry, false);
2526                 }
2527                 tree_insert_offset(&ctl->free_space_offset,
2528                                    entry->offset, &entry->offset_index, bitmap);
2529         }
2530         cluster->root = RB_ROOT;
2531
2532 out:
2533         spin_unlock(&cluster->lock);
2534         btrfs_put_block_group(block_group);
2535         return 0;
2536 }
2537
2538 static void __btrfs_remove_free_space_cache_locked(
2539                                 struct btrfs_free_space_ctl *ctl)
2540 {
2541         struct btrfs_free_space *info;
2542         struct rb_node *node;
2543
2544         while ((node = rb_last(&ctl->free_space_offset)) != NULL) {
2545                 info = rb_entry(node, struct btrfs_free_space, offset_index);
2546                 if (!info->bitmap) {
2547                         unlink_free_space(ctl, info);
2548                         kmem_cache_free(btrfs_free_space_cachep, info);
2549                 } else {
2550                         free_bitmap(ctl, info);
2551                 }
2552
2553                 cond_resched_lock(&ctl->tree_lock);
2554         }
2555 }
2556
2557 void __btrfs_remove_free_space_cache(struct btrfs_free_space_ctl *ctl)
2558 {
2559         spin_lock(&ctl->tree_lock);
2560         __btrfs_remove_free_space_cache_locked(ctl);
2561         spin_unlock(&ctl->tree_lock);
2562 }
2563
2564 void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group)
2565 {
2566         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2567         struct btrfs_free_cluster *cluster;
2568         struct list_head *head;
2569
2570         spin_lock(&ctl->tree_lock);
2571         while ((head = block_group->cluster_list.next) !=
2572                &block_group->cluster_list) {
2573                 cluster = list_entry(head, struct btrfs_free_cluster,
2574                                      block_group_list);
2575
2576                 WARN_ON(cluster->block_group != block_group);
2577                 __btrfs_return_cluster_to_free_space(block_group, cluster);
2578
2579                 cond_resched_lock(&ctl->tree_lock);
2580         }
2581         __btrfs_remove_free_space_cache_locked(ctl);
2582         spin_unlock(&ctl->tree_lock);
2583
2584 }
2585
2586 u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
2587                                u64 offset, u64 bytes, u64 empty_size,
2588                                u64 *max_extent_size)
2589 {
2590         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2591         struct btrfs_free_space *entry = NULL;
2592         u64 bytes_search = bytes + empty_size;
2593         u64 ret = 0;
2594         u64 align_gap = 0;
2595         u64 align_gap_len = 0;
2596
2597         spin_lock(&ctl->tree_lock);
2598         entry = find_free_space(ctl, &offset, &bytes_search,
2599                                 block_group->full_stripe_len, max_extent_size);
2600         if (!entry)
2601                 goto out;
2602
2603         ret = offset;
2604         if (entry->bitmap) {
2605                 bitmap_clear_bits(ctl, entry, offset, bytes);
2606                 if (!entry->bytes)
2607                         free_bitmap(ctl, entry);
2608         } else {
2609                 unlink_free_space(ctl, entry);
2610                 align_gap_len = offset - entry->offset;
2611                 align_gap = entry->offset;
2612
2613                 entry->offset = offset + bytes;
2614                 WARN_ON(entry->bytes < bytes + align_gap_len);
2615
2616                 entry->bytes -= bytes + align_gap_len;
2617                 if (!entry->bytes)
2618                         kmem_cache_free(btrfs_free_space_cachep, entry);
2619                 else
2620                         link_free_space(ctl, entry);
2621         }
2622 out:
2623         spin_unlock(&ctl->tree_lock);
2624
2625         if (align_gap_len)
2626                 __btrfs_add_free_space(block_group->fs_info, ctl,
2627                                        align_gap, align_gap_len);
2628         return ret;
2629 }
2630
2631 /*
2632  * given a cluster, put all of its extents back into the free space
2633  * cache.  If a block group is passed, this function will only free
2634  * a cluster that belongs to the passed block group.
2635  *
2636  * Otherwise, it'll get a reference on the block group pointed to by the
2637  * cluster and remove the cluster from it.
2638  */
2639 int btrfs_return_cluster_to_free_space(
2640                                struct btrfs_block_group_cache *block_group,
2641                                struct btrfs_free_cluster *cluster)
2642 {
2643         struct btrfs_free_space_ctl *ctl;
2644         int ret;
2645
2646         /* first, get a safe pointer to the block group */
2647         spin_lock(&cluster->lock);
2648         if (!block_group) {
2649                 block_group = cluster->block_group;
2650                 if (!block_group) {
2651                         spin_unlock(&cluster->lock);
2652                         return 0;
2653                 }
2654         } else if (cluster->block_group != block_group) {
2655                 /* someone else has already freed it don't redo their work */
2656                 spin_unlock(&cluster->lock);
2657                 return 0;
2658         }
2659         atomic_inc(&block_group->count);
2660         spin_unlock(&cluster->lock);
2661
2662         ctl = block_group->free_space_ctl;
2663
2664         /* now return any extents the cluster had on it */
2665         spin_lock(&ctl->tree_lock);
2666         ret = __btrfs_return_cluster_to_free_space(block_group, cluster);
2667         spin_unlock(&ctl->tree_lock);
2668
2669         /* finally drop our ref */
2670         btrfs_put_block_group(block_group);
2671         return ret;
2672 }
2673
2674 static u64 btrfs_alloc_from_bitmap(struct btrfs_block_group_cache *block_group,
2675                                    struct btrfs_free_cluster *cluster,
2676                                    struct btrfs_free_space *entry,
2677                                    u64 bytes, u64 min_start,
2678                                    u64 *max_extent_size)
2679 {
2680         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2681         int err;
2682         u64 search_start = cluster->window_start;
2683         u64 search_bytes = bytes;
2684         u64 ret = 0;
2685
2686         search_start = min_start;
2687         search_bytes = bytes;
2688
2689         err = search_bitmap(ctl, entry, &search_start, &search_bytes, true);
2690         if (err) {
2691                 if (search_bytes > *max_extent_size)
2692                         *max_extent_size = search_bytes;
2693                 return 0;
2694         }
2695
2696         ret = search_start;
2697         __bitmap_clear_bits(ctl, entry, ret, bytes);
2698
2699         return ret;
2700 }
2701
2702 /*
2703  * given a cluster, try to allocate 'bytes' from it, returns 0
2704  * if it couldn't find anything suitably large, or a logical disk offset
2705  * if things worked out
2706  */
2707 u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
2708                              struct btrfs_free_cluster *cluster, u64 bytes,
2709                              u64 min_start, u64 *max_extent_size)
2710 {
2711         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2712         struct btrfs_free_space *entry = NULL;
2713         struct rb_node *node;
2714         u64 ret = 0;
2715
2716         spin_lock(&cluster->lock);
2717         if (bytes > cluster->max_size)
2718                 goto out;
2719
2720         if (cluster->block_group != block_group)
2721                 goto out;
2722
2723         node = rb_first(&cluster->root);
2724         if (!node)
2725                 goto out;
2726
2727         entry = rb_entry(node, struct btrfs_free_space, offset_index);
2728         while (1) {
2729                 if (entry->bytes < bytes && entry->bytes > *max_extent_size)
2730                         *max_extent_size = entry->bytes;
2731
2732                 if (entry->bytes < bytes ||
2733                     (!entry->bitmap && entry->offset < min_start)) {
2734                         node = rb_next(&entry->offset_index);
2735                         if (!node)
2736                                 break;
2737                         entry = rb_entry(node, struct btrfs_free_space,
2738                                          offset_index);
2739                         continue;
2740                 }
2741
2742                 if (entry->bitmap) {
2743                         ret = btrfs_alloc_from_bitmap(block_group,
2744                                                       cluster, entry, bytes,
2745                                                       cluster->window_start,
2746                                                       max_extent_size);
2747                         if (ret == 0) {
2748                                 node = rb_next(&entry->offset_index);
2749                                 if (!node)
2750                                         break;
2751                                 entry = rb_entry(node, struct btrfs_free_space,
2752                                                  offset_index);
2753                                 continue;
2754                         }
2755                         cluster->window_start += bytes;
2756                 } else {
2757                         ret = entry->offset;
2758
2759                         entry->offset += bytes;
2760                         entry->bytes -= bytes;
2761                 }
2762
2763                 if (entry->bytes == 0)
2764                         rb_erase(&entry->offset_index, &cluster->root);
2765                 break;
2766         }
2767 out:
2768         spin_unlock(&cluster->lock);
2769
2770         if (!ret)
2771                 return 0;
2772
2773         spin_lock(&ctl->tree_lock);
2774
2775         ctl->free_space -= bytes;
2776         if (entry->bytes == 0) {
2777                 ctl->free_extents--;
2778                 if (entry->bitmap) {
2779                         kfree(entry->bitmap);
2780                         ctl->total_bitmaps--;
2781                         ctl->op->recalc_thresholds(ctl);
2782                 }
2783                 kmem_cache_free(btrfs_free_space_cachep, entry);
2784         }
2785
2786         spin_unlock(&ctl->tree_lock);
2787
2788         return ret;
2789 }
2790
2791 static int btrfs_bitmap_cluster(struct btrfs_block_group_cache *block_group,
2792                                 struct btrfs_free_space *entry,
2793                                 struct btrfs_free_cluster *cluster,
2794                                 u64 offset, u64 bytes,
2795                                 u64 cont1_bytes, u64 min_bytes)
2796 {
2797         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2798         unsigned long next_zero;
2799         unsigned long i;
2800         unsigned long want_bits;
2801         unsigned long min_bits;
2802         unsigned long found_bits;
2803         unsigned long max_bits = 0;
2804         unsigned long start = 0;
2805         unsigned long total_found = 0;
2806         int ret;
2807
2808         i = offset_to_bit(entry->offset, ctl->unit,
2809                           max_t(u64, offset, entry->offset));
2810         want_bits = bytes_to_bits(bytes, ctl->unit);
2811         min_bits = bytes_to_bits(min_bytes, ctl->unit);
2812
2813         /*
2814          * Don't bother looking for a cluster in this bitmap if it's heavily
2815          * fragmented.
2816          */
2817         if (entry->max_extent_size &&
2818             entry->max_extent_size < cont1_bytes)
2819                 return -ENOSPC;
2820 again:
2821         found_bits = 0;
2822         for_each_set_bit_from(i, entry->bitmap, BITS_PER_BITMAP) {
2823                 next_zero = find_next_zero_bit(entry->bitmap,
2824                                                BITS_PER_BITMAP, i);
2825                 if (next_zero - i >= min_bits) {
2826                         found_bits = next_zero - i;
2827                         if (found_bits > max_bits)
2828                                 max_bits = found_bits;
2829                         break;
2830                 }
2831                 if (next_zero - i > max_bits)
2832                         max_bits = next_zero - i;
2833                 i = next_zero;
2834         }
2835
2836         if (!found_bits) {
2837                 entry->max_extent_size = (u64)max_bits * ctl->unit;
2838                 return -ENOSPC;
2839         }
2840
2841         if (!total_found) {
2842                 start = i;
2843                 cluster->max_size = 0;
2844         }
2845
2846         total_found += found_bits;
2847
2848         if (cluster->max_size < found_bits * ctl->unit)
2849                 cluster->max_size = found_bits * ctl->unit;
2850
2851         if (total_found < want_bits || cluster->max_size < cont1_bytes) {
2852                 i = next_zero + 1;
2853                 goto again;
2854         }
2855
2856         cluster->window_start = start * ctl->unit + entry->offset;
2857         rb_erase(&entry->offset_index, &ctl->free_space_offset);
2858         ret = tree_insert_offset(&cluster->root, entry->offset,
2859                                  &entry->offset_index, 1);
2860         ASSERT(!ret); /* -EEXIST; Logic error */
2861
2862         trace_btrfs_setup_cluster(block_group, cluster,
2863                                   total_found * ctl->unit, 1);
2864         return 0;
2865 }
2866
2867 /*
2868  * This searches the block group for just extents to fill the cluster with.
2869  * Try to find a cluster with at least bytes total bytes, at least one
2870  * extent of cont1_bytes, and other clusters of at least min_bytes.
2871  */
2872 static noinline int
2873 setup_cluster_no_bitmap(struct btrfs_block_group_cache *block_group,
2874                         struct btrfs_free_cluster *cluster,
2875                         struct list_head *bitmaps, u64 offset, u64 bytes,
2876                         u64 cont1_bytes, u64 min_bytes)
2877 {
2878         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2879         struct btrfs_free_space *first = NULL;
2880         struct btrfs_free_space *entry = NULL;
2881         struct btrfs_free_space *last;
2882         struct rb_node *node;
2883         u64 window_free;
2884         u64 max_extent;
2885         u64 total_size = 0;
2886
2887         entry = tree_search_offset(ctl, offset, 0, 1);
2888         if (!entry)
2889                 return -ENOSPC;
2890
2891         /*
2892          * We don't want bitmaps, so just move along until we find a normal
2893          * extent entry.
2894          */
2895         while (entry->bitmap || entry->bytes < min_bytes) {
2896                 if (entry->bitmap && list_empty(&entry->list))
2897                         list_add_tail(&entry->list, bitmaps);
2898                 node = rb_next(&entry->offset_index);
2899                 if (!node)
2900                         return -ENOSPC;
2901                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2902         }
2903
2904         window_free = entry->bytes;
2905         max_extent = entry->bytes;
2906         first = entry;
2907         last = entry;
2908
2909         for (node = rb_next(&entry->offset_index); node;
2910              node = rb_next(&entry->offset_index)) {
2911                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2912
2913                 if (entry->bitmap) {
2914                         if (list_empty(&entry->list))
2915                                 list_add_tail(&entry->list, bitmaps);
2916                         continue;
2917                 }
2918
2919                 if (entry->bytes < min_bytes)
2920                         continue;
2921
2922                 last = entry;
2923                 window_free += entry->bytes;
2924                 if (entry->bytes > max_extent)
2925                         max_extent = entry->bytes;
2926         }
2927
2928         if (window_free < bytes || max_extent < cont1_bytes)
2929                 return -ENOSPC;
2930
2931         cluster->window_start = first->offset;
2932
2933         node = &first->offset_index;
2934
2935         /*
2936          * now we've found our entries, pull them out of the free space
2937          * cache and put them into the cluster rbtree
2938          */
2939         do {
2940                 int ret;
2941
2942                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2943                 node = rb_next(&entry->offset_index);
2944                 if (entry->bitmap || entry->bytes < min_bytes)
2945                         continue;
2946
2947                 rb_erase(&entry->offset_index, &ctl->free_space_offset);
2948                 ret = tree_insert_offset(&cluster->root, entry->offset,
2949                                          &entry->offset_index, 0);
2950                 total_size += entry->bytes;
2951                 ASSERT(!ret); /* -EEXIST; Logic error */
2952         } while (node && entry != last);
2953
2954         cluster->max_size = max_extent;
2955         trace_btrfs_setup_cluster(block_group, cluster, total_size, 0);
2956         return 0;
2957 }
2958
2959 /*
2960  * This specifically looks for bitmaps that may work in the cluster, we assume
2961  * that we have already failed to find extents that will work.
2962  */
2963 static noinline int
2964 setup_cluster_bitmap(struct btrfs_block_group_cache *block_group,
2965                      struct btrfs_free_cluster *cluster,
2966                      struct list_head *bitmaps, u64 offset, u64 bytes,
2967                      u64 cont1_bytes, u64 min_bytes)
2968 {
2969         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2970         struct btrfs_free_space *entry = NULL;
2971         int ret = -ENOSPC;
2972         u64 bitmap_offset = offset_to_bitmap(ctl, offset);
2973
2974         if (ctl->total_bitmaps == 0)
2975                 return -ENOSPC;
2976
2977         /*
2978          * The bitmap that covers offset won't be in the list unless offset
2979          * is just its start offset.
2980          */
2981         if (!list_empty(bitmaps))
2982                 entry = list_first_entry(bitmaps, struct btrfs_free_space, list);
2983
2984         if (!entry || entry->offset != bitmap_offset) {
2985                 entry = tree_search_offset(ctl, bitmap_offset, 1, 0);
2986                 if (entry && list_empty(&entry->list))
2987                         list_add(&entry->list, bitmaps);
2988         }
2989
2990         list_for_each_entry(entry, bitmaps, list) {
2991                 if (entry->bytes < bytes)
2992                         continue;
2993                 ret = btrfs_bitmap_cluster(block_group, entry, cluster, offset,
2994                                            bytes, cont1_bytes, min_bytes);
2995                 if (!ret)
2996                         return 0;
2997         }
2998
2999         /*
3000          * The bitmaps list has all the bitmaps that record free space
3001          * starting after offset, so no more search is required.
3002          */
3003         return -ENOSPC;
3004 }
3005
3006 /*
3007  * here we try to find a cluster of blocks in a block group.  The goal
3008  * is to find at least bytes+empty_size.
3009  * We might not find them all in one contiguous area.
3010  *
3011  * returns zero and sets up cluster if things worked out, otherwise
3012  * it returns -enospc
3013  */
3014 int btrfs_find_space_cluster(struct btrfs_root *root,
3015                              struct btrfs_block_group_cache *block_group,
3016                              struct btrfs_free_cluster *cluster,
3017                              u64 offset, u64 bytes, u64 empty_size)
3018 {
3019         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3020         struct btrfs_free_space *entry, *tmp;
3021         LIST_HEAD(bitmaps);
3022         u64 min_bytes;
3023         u64 cont1_bytes;
3024         int ret;
3025
3026         /*
3027          * Choose the minimum extent size we'll require for this
3028          * cluster.  For SSD_SPREAD, don't allow any fragmentation.
3029          * For metadata, allow allocates with smaller extents.  For
3030          * data, keep it dense.
3031          */
3032         if (btrfs_test_opt(root->fs_info, SSD_SPREAD)) {
3033                 cont1_bytes = min_bytes = bytes + empty_size;
3034         } else if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) {
3035                 cont1_bytes = bytes;
3036                 min_bytes = block_group->sectorsize;
3037         } else {
3038                 cont1_bytes = max(bytes, (bytes + empty_size) >> 2);
3039                 min_bytes = block_group->sectorsize;
3040         }
3041
3042         spin_lock(&ctl->tree_lock);
3043
3044         /*
3045          * If we know we don't have enough space to make a cluster don't even
3046          * bother doing all the work to try and find one.
3047          */
3048         if (ctl->free_space < bytes) {
3049                 spin_unlock(&ctl->tree_lock);
3050                 return -ENOSPC;
3051         }
3052
3053         spin_lock(&cluster->lock);
3054
3055         /* someone already found a cluster, hooray */
3056         if (cluster->block_group) {
3057                 ret = 0;
3058                 goto out;
3059         }
3060
3061         trace_btrfs_find_cluster(block_group, offset, bytes, empty_size,
3062                                  min_bytes);
3063
3064         ret = setup_cluster_no_bitmap(block_group, cluster, &bitmaps, offset,
3065                                       bytes + empty_size,
3066                                       cont1_bytes, min_bytes);
3067         if (ret)
3068                 ret = setup_cluster_bitmap(block_group, cluster, &bitmaps,
3069                                            offset, bytes + empty_size,
3070                                            cont1_bytes, min_bytes);
3071
3072         /* Clear our temporary list */
3073         list_for_each_entry_safe(entry, tmp, &bitmaps, list)
3074                 list_del_init(&entry->list);
3075
3076         if (!ret) {
3077                 atomic_inc(&block_group->count);
3078                 list_add_tail(&cluster->block_group_list,
3079                               &block_group->cluster_list);
3080                 cluster->block_group = block_group;
3081         } else {
3082                 trace_btrfs_failed_cluster_setup(block_group);
3083         }
3084 out:
3085         spin_unlock(&cluster->lock);
3086         spin_unlock(&ctl->tree_lock);
3087
3088         return ret;
3089 }
3090
3091 /*
3092  * simple code to zero out a cluster
3093  */
3094 void btrfs_init_free_cluster(struct btrfs_free_cluster *cluster)
3095 {
3096         spin_lock_init(&cluster->lock);
3097         spin_lock_init(&cluster->refill_lock);
3098         cluster->root = RB_ROOT;
3099         cluster->max_size = 0;
3100         cluster->fragmented = false;
3101         INIT_LIST_HEAD(&cluster->block_group_list);
3102         cluster->block_group = NULL;
3103 }
3104
3105 static int do_trimming(struct btrfs_block_group_cache *block_group,
3106                        u64 *total_trimmed, u64 start, u64 bytes,
3107                        u64 reserved_start, u64 reserved_bytes,
3108                        struct btrfs_trim_range *trim_entry)
3109 {
3110         struct btrfs_space_info *space_info = block_group->space_info;
3111         struct btrfs_fs_info *fs_info = block_group->fs_info;
3112         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3113         int ret;
3114         int update = 0;
3115         u64 trimmed = 0;
3116
3117         spin_lock(&space_info->lock);
3118         spin_lock(&block_group->lock);
3119         if (!block_group->ro) {
3120                 block_group->reserved += reserved_bytes;
3121                 space_info->bytes_reserved += reserved_bytes;
3122                 update = 1;
3123         }
3124         spin_unlock(&block_group->lock);
3125         spin_unlock(&space_info->lock);
3126
3127         ret = btrfs_discard_extent(fs_info->extent_root,
3128                                    start, bytes, &trimmed);
3129         if (!ret)
3130                 *total_trimmed += trimmed;
3131
3132         mutex_lock(&ctl->cache_writeout_mutex);
3133         btrfs_add_free_space(block_group, reserved_start, reserved_bytes);
3134         list_del(&trim_entry->list);
3135         mutex_unlock(&ctl->cache_writeout_mutex);
3136
3137         if (update) {
3138                 spin_lock(&space_info->lock);
3139                 spin_lock(&block_group->lock);
3140                 if (block_group->ro)
3141                         space_info->bytes_readonly += reserved_bytes;
3142                 block_group->reserved -= reserved_bytes;
3143                 space_info->bytes_reserved -= reserved_bytes;
3144                 spin_unlock(&space_info->lock);
3145                 spin_unlock(&block_group->lock);
3146         }
3147
3148         return ret;
3149 }
3150
3151 static int trim_no_bitmap(struct btrfs_block_group_cache *block_group,
3152                           u64 *total_trimmed, u64 start, u64 end, u64 minlen)
3153 {
3154         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3155         struct btrfs_free_space *entry;
3156         struct rb_node *node;
3157         int ret = 0;
3158         u64 extent_start;
3159         u64 extent_bytes;
3160         u64 bytes;
3161
3162         while (start < end) {
3163                 struct btrfs_trim_range trim_entry;
3164
3165                 mutex_lock(&ctl->cache_writeout_mutex);
3166                 spin_lock(&ctl->tree_lock);
3167
3168                 if (ctl->free_space < minlen) {
3169                         spin_unlock(&ctl->tree_lock);
3170                         mutex_unlock(&ctl->cache_writeout_mutex);
3171                         break;
3172                 }
3173
3174                 entry = tree_search_offset(ctl, start, 0, 1);
3175                 if (!entry) {
3176                         spin_unlock(&ctl->tree_lock);
3177                         mutex_unlock(&ctl->cache_writeout_mutex);
3178                         break;
3179                 }
3180
3181                 /* skip bitmaps */
3182                 while (entry->bitmap) {
3183                         node = rb_next(&entry->offset_index);
3184                         if (!node) {
3185                                 spin_unlock(&ctl->tree_lock);
3186                                 mutex_unlock(&ctl->cache_writeout_mutex);
3187                                 goto out;
3188                         }
3189                         entry = rb_entry(node, struct btrfs_free_space,
3190                                          offset_index);
3191                 }
3192
3193                 if (entry->offset >= end) {
3194                         spin_unlock(&ctl->tree_lock);
3195                         mutex_unlock(&ctl->cache_writeout_mutex);
3196                         break;
3197                 }
3198
3199                 extent_start = entry->offset;
3200                 extent_bytes = entry->bytes;
3201                 start = max(start, extent_start);
3202                 bytes = min(extent_start + extent_bytes, end) - start;
3203                 if (bytes < minlen) {
3204                         spin_unlock(&ctl->tree_lock);
3205                         mutex_unlock(&ctl->cache_writeout_mutex);
3206                         goto next;
3207                 }
3208
3209                 unlink_free_space(ctl, entry);
3210                 kmem_cache_free(btrfs_free_space_cachep, entry);
3211
3212                 spin_unlock(&ctl->tree_lock);
3213                 trim_entry.start = extent_start;
3214                 trim_entry.bytes = extent_bytes;
3215                 list_add_tail(&trim_entry.list, &ctl->trimming_ranges);
3216                 mutex_unlock(&ctl->cache_writeout_mutex);
3217
3218                 ret = do_trimming(block_group, total_trimmed, start, bytes,
3219                                   extent_start, extent_bytes, &trim_entry);
3220                 if (ret)
3221                         break;
3222 next:
3223                 start += bytes;
3224
3225                 if (fatal_signal_pending(current)) {
3226                         ret = -ERESTARTSYS;
3227                         break;
3228                 }
3229
3230                 cond_resched();
3231         }
3232 out:
3233         return ret;
3234 }
3235
3236 static int trim_bitmaps(struct btrfs_block_group_cache *block_group,
3237                         u64 *total_trimmed, u64 start, u64 end, u64 minlen)
3238 {
3239         struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3240         struct btrfs_free_space *entry;
3241         int ret = 0;
3242         int ret2;
3243         u64 bytes;
3244         u64 offset = offset_to_bitmap(ctl, start);
3245
3246         while (offset < end) {
3247                 bool next_bitmap = false;
3248                 struct btrfs_trim_range trim_entry;
3249
3250                 mutex_lock(&ctl->cache_writeout_mutex);
3251                 spin_lock(&ctl->tree_lock);
3252
3253                 if (ctl->free_space < minlen) {
3254                         spin_unlock(&ctl->tree_lock);
3255                         mutex_unlock(&ctl->cache_writeout_mutex);
3256                         break;
3257                 }
3258
3259                 entry = tree_search_offset(ctl, offset, 1, 0);
3260                 if (!entry) {
3261                         spin_unlock(&ctl->tree_lock);
3262                         mutex_unlock(&ctl->cache_writeout_mutex);
3263                         next_bitmap = true;
3264                         goto next;
3265                 }
3266
3267                 bytes = minlen;
3268                 ret2 = search_bitmap(ctl, entry, &start, &bytes, false);
3269                 if (ret2 || start >= end) {
3270                         spin_unlock(&ctl->tree_lock);
3271                         mutex_unlock(&ctl->cache_writeout_mutex);
3272                         next_bitmap = true;
3273                         goto next;
3274                 }
3275
3276                 bytes = min(bytes, end - start);
3277                 if (bytes < minlen) {
3278                         spin_unlock(&ctl->tree_lock);
3279                         mutex_unlock(&ctl->cache_writeout_mutex);
3280                         goto next;
3281                 }
3282
3283                 bitmap_clear_bits(ctl, entry, start, bytes);
3284                 if (entry->bytes == 0)
3285                         free_bitmap(ctl, entry);
3286
3287                 spin_unlock(&ctl->tree_lock);
3288                 trim_entry.start = start;
3289                 trim_entry.bytes = bytes;
3290                 list_add_tail(&trim_entry.list, &ctl->trimming_ranges);
3291                 mutex_unlock(&ctl->cache_writeout_mutex);
3292
3293                 ret = do_trimming(block_group, total_trimmed, start, bytes,
3294                                   start, bytes, &trim_entry);
3295                 if (ret)
3296                         break;
3297 next:
3298                 if (next_bitmap) {
3299                         offset += BITS_PER_BITMAP * ctl->unit;
3300                 } else {
3301                         start += bytes;
3302                         if (start >= offset + BITS_PER_BITMAP * ctl->unit)
3303                                 offset += BITS_PER_BITMAP * ctl->unit;
3304                 }
3305
3306                 if (fatal_signal_pending(current)) {
3307                         ret = -ERESTARTSYS;
3308                         break;
3309                 }
3310
3311                 cond_resched();
3312         }
3313
3314         return ret;
3315 }
3316
3317 void btrfs_get_block_group_trimming(struct btrfs_block_group_cache *cache)
3318 {
3319         atomic_inc(&cache->trimming);
3320 }
3321
3322 void btrfs_put_block_group_trimming(struct btrfs_block_group_cache *block_group)
3323 {
3324         struct extent_map_tree *em_tree;
3325         struct extent_map *em;
3326         bool cleanup;
3327
3328         spin_lock(&block_group->lock);
3329         cleanup = (atomic_dec_and_test(&block_group->trimming) &&
3330                    block_group->removed);
3331         spin_unlock(&block_group->lock);
3332
3333         if (cleanup) {
3334                 lock_chunks(block_group->fs_info->chunk_root);
3335                 em_tree = &block_group->fs_info->mapping_tree.map_tree;
3336                 write_lock(&em_tree->lock);
3337                 em = lookup_extent_mapping(em_tree, block_group->key.objectid,
3338                                            1);
3339                 BUG_ON(!em); /* logic error, can't happen */
3340                 /*
3341                  * remove_extent_mapping() will delete us from the pinned_chunks
3342                  * list, which is protected by the chunk mutex.
3343                  */
3344                 remove_extent_mapping(em_tree, em);
3345                 write_unlock(&em_tree->lock);
3346                 unlock_chunks(block_group->fs_info->chunk_root);
3347
3348                 /* once for us and once for the tree */
3349                 free_extent_map(em);
3350                 free_extent_map(em);
3351
3352                 /*
3353                  * We've left one free space entry and other tasks trimming
3354                  * this block group have left 1 entry each one. Free them.
3355                  */
3356                 __btrfs_remove_free_space_cache(block_group->free_space_ctl);
3357         }
3358 }
3359
3360 int btrfs_trim_block_group(struct btrfs_block_group_cache *block_group,
3361                            u64 *trimmed, u64 start, u64 end, u64 minlen)
3362 {
3363         int ret;
3364
3365         *trimmed = 0;
3366
3367         spin_lock(&block_group->lock);
3368         if (block_group->removed) {
3369                 spin_unlock(&block_group->lock);
3370                 return 0;
3371         }
3372         btrfs_get_block_group_trimming(block_group);
3373         spin_unlock(&block_group->lock);
3374
3375         ret = trim_no_bitmap(block_group, trimmed, start, end, minlen);
3376         if (ret)
3377                 goto out;
3378
3379         ret = trim_bitmaps(block_group, trimmed, start, end, minlen);
3380 out:
3381         btrfs_put_block_group_trimming(block_group);
3382         return ret;
3383 }
3384
3385 /*
3386  * Find the left-most item in the cache tree, and then return the
3387  * smallest inode number in the item.
3388  *
3389  * Note: the returned inode number may not be the smallest one in
3390  * the tree, if the left-most item is a bitmap.
3391  */
3392 u64 btrfs_find_ino_for_alloc(struct btrfs_root *fs_root)
3393 {
3394         struct btrfs_free_space_ctl *ctl = fs_root->free_ino_ctl;
3395         struct btrfs_free_space *entry = NULL;
3396         u64 ino = 0;
3397
3398         spin_lock(&ctl->tree_lock);
3399
3400         if (RB_EMPTY_ROOT(&ctl->free_space_offset))
3401                 goto out;
3402
3403         entry = rb_entry(rb_first(&ctl->free_space_offset),
3404                          struct btrfs_free_space, offset_index);
3405
3406         if (!entry->bitmap) {
3407                 ino = entry->offset;
3408
3409                 unlink_free_space(ctl, entry);
3410                 entry->offset++;
3411                 entry->bytes--;
3412                 if (!entry->bytes)
3413                         kmem_cache_free(btrfs_free_space_cachep, entry);
3414                 else
3415                         link_free_space(ctl, entry);
3416         } else {
3417                 u64 offset = 0;
3418                 u64 count = 1;
3419                 int ret;
3420
3421                 ret = search_bitmap(ctl, entry, &offset, &count, true);
3422                 /* Logic error; Should be empty if it can't find anything */
3423                 ASSERT(!ret);
3424
3425                 ino = offset;
3426                 bitmap_clear_bits(ctl, entry, offset, 1);
3427                 if (entry->bytes == 0)
3428                         free_bitmap(ctl, entry);
3429         }
3430 out:
3431         spin_unlock(&ctl->tree_lock);
3432
3433         return ino;
3434 }
3435
3436 struct inode *lookup_free_ino_inode(struct btrfs_root *root,
3437                                     struct btrfs_path *path)
3438 {
3439         struct inode *inode = NULL;
3440
3441         spin_lock(&root->ino_cache_lock);
3442         if (root->ino_cache_inode)
3443                 inode = igrab(root->ino_cache_inode);
3444         spin_unlock(&root->ino_cache_lock);
3445         if (inode)
3446                 return inode;
3447
3448         inode = __lookup_free_space_inode(root, path, 0);
3449         if (IS_ERR(inode))
3450                 return inode;
3451
3452         spin_lock(&root->ino_cache_lock);
3453         if (!btrfs_fs_closing(root->fs_info))
3454                 root->ino_cache_inode = igrab(inode);
3455         spin_unlock(&root->ino_cache_lock);
3456
3457         return inode;
3458 }
3459
3460 int create_free_ino_inode(struct btrfs_root *root,
3461                           struct btrfs_trans_handle *trans,
3462                           struct btrfs_path *path)
3463 {
3464         return __create_free_space_inode(root, trans, path,
3465                                          BTRFS_FREE_INO_OBJECTID, 0);
3466 }
3467
3468 int load_free_ino_cache(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
3469 {
3470         struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
3471         struct btrfs_path *path;
3472         struct inode *inode;
3473         int ret = 0;
3474         u64 root_gen = btrfs_root_generation(&root->root_item);
3475
3476         if (!btrfs_test_opt(root->fs_info, INODE_MAP_CACHE))
3477                 return 0;
3478
3479         /*
3480          * If we're unmounting then just return, since this does a search on the
3481          * normal root and not the commit root and we could deadlock.
3482          */
3483         if (btrfs_fs_closing(fs_info))
3484                 return 0;
3485
3486         path = btrfs_alloc_path();
3487         if (!path)
3488                 return 0;
3489
3490         inode = lookup_free_ino_inode(root, path);
3491         if (IS_ERR(inode))
3492                 goto out;
3493
3494         if (root_gen != BTRFS_I(inode)->generation)
3495                 goto out_put;
3496
3497         ret = __load_free_space_cache(root, inode, ctl, path, 0);
3498
3499         if (ret < 0)
3500                 btrfs_err(fs_info,
3501                         "failed to load free ino cache for root %llu",
3502                         root->root_key.objectid);
3503 out_put:
3504         iput(inode);
3505 out:
3506         btrfs_free_path(path);
3507         return ret;
3508 }
3509
3510 int btrfs_write_out_ino_cache(struct btrfs_root *root,
3511                               struct btrfs_trans_handle *trans,
3512                               struct btrfs_path *path,
3513                               struct inode *inode)
3514 {
3515         struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
3516         int ret;
3517         struct btrfs_io_ctl io_ctl;
3518         bool release_metadata = true;
3519
3520         if (!btrfs_test_opt(root->fs_info, INODE_MAP_CACHE))
3521                 return 0;
3522
3523         memset(&io_ctl, 0, sizeof(io_ctl));
3524         ret = __btrfs_write_out_cache(root, inode, ctl, NULL, &io_ctl,
3525                                       trans, path, 0);
3526         if (!ret) {
3527                 /*
3528                  * At this point writepages() didn't error out, so our metadata
3529                  * reservation is released when the writeback finishes, at
3530                  * inode.c:btrfs_finish_ordered_io(), regardless of it finishing
3531                  * with or without an error.
3532                  */
3533                 release_metadata = false;
3534                 ret = btrfs_wait_cache_io(root, trans, NULL, &io_ctl, path, 0);
3535         }
3536
3537         if (ret) {
3538                 if (release_metadata)
3539                         btrfs_delalloc_release_metadata(inode, inode->i_size);
3540 #ifdef DEBUG
3541                 btrfs_err(root->fs_info,
3542                         "failed to write free ino cache for root %llu",
3543                         root->root_key.objectid);
3544 #endif
3545         }
3546
3547         return ret;
3548 }
3549
3550 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3551 /*
3552  * Use this if you need to make a bitmap or extent entry specifically, it
3553  * doesn't do any of the merging that add_free_space does, this acts a lot like
3554  * how the free space cache loading stuff works, so you can get really weird
3555  * configurations.
3556  */
3557 int test_add_free_space_entry(struct btrfs_block_group_cache *cache,
3558                               u64 offset, u64 bytes, bool bitmap)
3559 {
3560         struct btrfs_free_space_ctl *ctl = cache->free_space_ctl;
3561         struct btrfs_free_space *info = NULL, *bitmap_info;
3562         void *map = NULL;
3563         u64 bytes_added;
3564         int ret;
3565
3566 again:
3567         if (!info) {
3568                 info = kmem_cache_zalloc(btrfs_free_space_cachep, GFP_NOFS);
3569                 if (!info)
3570                         return -ENOMEM;
3571         }
3572
3573         if (!bitmap) {
3574                 spin_lock(&ctl->tree_lock);
3575                 info->offset = offset;
3576                 info->bytes = bytes;
3577                 info->max_extent_size = 0;
3578                 ret = link_free_space(ctl, info);
3579                 spin_unlock(&ctl->tree_lock);
3580                 if (ret)
3581                         kmem_cache_free(btrfs_free_space_cachep, info);
3582                 return ret;
3583         }
3584
3585         if (!map) {
3586                 map = kzalloc(PAGE_SIZE, GFP_NOFS);
3587                 if (!map) {
3588                         kmem_cache_free(btrfs_free_space_cachep, info);
3589                         return -ENOMEM;
3590                 }
3591         }
3592
3593         spin_lock(&ctl->tree_lock);
3594         bitmap_info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
3595                                          1, 0);
3596         if (!bitmap_info) {
3597                 info->bitmap = map;
3598                 map = NULL;
3599                 add_new_bitmap(ctl, info, offset);
3600                 bitmap_info = info;
3601                 info = NULL;
3602         }
3603
3604         bytes_added = add_bytes_to_bitmap(ctl, bitmap_info, offset, bytes);
3605
3606         bytes -= bytes_added;
3607         offset += bytes_added;
3608         spin_unlock(&ctl->tree_lock);
3609
3610         if (bytes)
3611                 goto again;
3612
3613         if (info)
3614                 kmem_cache_free(btrfs_free_space_cachep, info);
3615         if (map)
3616                 kfree(map);
3617         return 0;
3618 }
3619
3620 /*
3621  * Checks to see if the given range is in the free space cache.  This is really
3622  * just used to check the absence of space, so if there is free space in the
3623  * range at all we will return 1.
3624  */
3625 int test_check_exists(struct btrfs_block_group_cache *cache,
3626                       u64 offset, u64 bytes)
3627 {
3628         struct btrfs_free_space_ctl *ctl = cache->free_space_ctl;
3629         struct btrfs_free_space *info;
3630         int ret = 0;
3631
3632         spin_lock(&ctl->tree_lock);
3633         info = tree_search_offset(ctl, offset, 0, 0);
3634         if (!info) {
3635                 info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
3636                                           1, 0);
3637                 if (!info)
3638                         goto out;
3639         }
3640
3641 have_info:
3642         if (info->bitmap) {
3643                 u64 bit_off, bit_bytes;
3644                 struct rb_node *n;
3645                 struct btrfs_free_space *tmp;
3646
3647                 bit_off = offset;
3648                 bit_bytes = ctl->unit;
3649                 ret = search_bitmap(ctl, info, &bit_off, &bit_bytes, false);
3650                 if (!ret) {
3651                         if (bit_off == offset) {
3652                                 ret = 1;
3653                                 goto out;
3654                         } else if (bit_off > offset &&
3655                                    offset + bytes > bit_off) {
3656                                 ret = 1;
3657                                 goto out;
3658                         }
3659                 }
3660
3661                 n = rb_prev(&info->offset_index);
3662                 while (n) {
3663                         tmp = rb_entry(n, struct btrfs_free_space,
3664                                        offset_index);
3665                         if (tmp->offset + tmp->bytes < offset)
3666                                 break;
3667                         if (offset + bytes < tmp->offset) {
3668                                 n = rb_prev(&tmp->offset_index);
3669                                 continue;
3670                         }
3671                         info = tmp;
3672                         goto have_info;
3673                 }
3674
3675                 n = rb_next(&info->offset_index);
3676                 while (n) {
3677                         tmp = rb_entry(n, struct btrfs_free_space,
3678                                        offset_index);
3679                         if (offset + bytes < tmp->offset)
3680                                 break;
3681                         if (tmp->offset + tmp->bytes < offset) {
3682                                 n = rb_next(&tmp->offset_index);
3683                                 continue;
3684                         }
3685                         info = tmp;
3686                         goto have_info;
3687                 }
3688
3689                 ret = 0;
3690                 goto out;
3691         }
3692
3693         if (info->offset == offset) {
3694                 ret = 1;
3695                 goto out;
3696         }
3697
3698         if (offset > info->offset && offset < info->offset + info->bytes)
3699                 ret = 1;
3700 out:
3701         spin_unlock(&ctl->tree_lock);
3702         return ret;
3703 }
3704 #endif /* CONFIG_BTRFS_FS_RUN_SANITY_TESTS */