fs/btrfs/disk-io.c

   1 /*
   2  * Copyright (C) 2007 Oracle.  All rights reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU General Public
   6  * License v2 as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public
  14  * License along with this program; if not, write to the
  15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  16  * Boston, MA 021110-1307, USA.
  17  */
  18
  19 #include <linux/fs.h>
  20 #include <linux/blkdev.h>
  21 #include <linux/crc32c.h>
  22 #include <linux/scatterlist.h>
  23 #include <linux/swap.h>
  24 #include <linux/radix-tree.h>
  25 #include <linux/writeback.h>
  26 #include <linux/buffer_head.h> // for block_sync_page
  27 #include <linux/workqueue.h>
  28 #include "ctree.h"
  29 #include "disk-io.h"
  30 #include "transaction.h"
  31 #include "btrfs_inode.h"
  32 #include "volumes.h"
  33 #include "print-tree.h"
  34
  35 #if 0
  36 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
  37 {
  38         if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) {
  39                 printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n",
  40                        (unsigned long long)extent_buffer_blocknr(buf),
  41                        (unsigned long long)btrfs_header_blocknr(buf));
  42                 return 1;
  43         }
  44         return 0;
  45 }
  46 #endif
  47
  48 static struct extent_io_ops btree_extent_io_ops;
  49 static struct workqueue_struct *end_io_workqueue;
  50
  51 struct end_io_wq {
  52         struct bio *bio;
  53         bio_end_io_t *end_io;
  54         void *private;
  55         struct btrfs_fs_info *info;
  56         int error;
  57         struct list_head list;
  58 };
  59
  60 struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
  61                                     size_t page_offset, u64 start, u64 len,
  62                                     int create)
  63 {
  64         struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
  65         struct extent_map *em;
  66         int ret;
  67
  68 again:
  69         spin_lock(&em_tree->lock);
  70         em = lookup_extent_mapping(em_tree, start, len);
  71         spin_unlock(&em_tree->lock);
  72         if (em) {
  73                 goto out;
  74         }
  75         em = alloc_extent_map(GFP_NOFS);
  76         if (!em) {
  77                 em = ERR_PTR(-ENOMEM);
  78                 goto out;
  79         }
  80         em->start = 0;
  81         em->len = i_size_read(inode);
  82         em->block_start = 0;
  83         em->bdev = inode->i_sb->s_bdev;
  84
  85         spin_lock(&em_tree->lock);
  86         ret = add_extent_mapping(em_tree, em);
  87         spin_unlock(&em_tree->lock);
  88
  89         if (ret == -EEXIST) {
  90                 free_extent_map(em);
  91                 em = NULL;
  92                 goto again;
  93         } else if (ret) {
  94                 em = ERR_PTR(ret);
  95         }
  96 out:
  97         return em;
  98 }
  99
 100 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
 101 {
 102         return crc32c(seed, data, len);
 103 }
 104
 105 void btrfs_csum_final(u32 crc, char *result)
 106 {
 107         *(__le32 *)result = ~cpu_to_le32(crc);
 108 }
 109
 110 static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 111                            int verify)
 112 {
 113         char result[BTRFS_CRC32_SIZE];
 114         unsigned long len;
 115         unsigned long cur_len;
 116         unsigned long offset = BTRFS_CSUM_SIZE;
 117         char *map_token = NULL;
 118         char *kaddr;
 119         unsigned long map_start;
 120         unsigned long map_len;
 121         int err;
 122         u32 crc = ~(u32)0;
 123
 124         len = buf->len - offset;
 125         while(len > 0) {
 126                 err = map_private_extent_buffer(buf, offset, 32,
 127                                         &map_token, &kaddr,
 128                                         &map_start, &map_len, KM_USER0);
 129                 if (err) {
 130                         printk("failed to map extent buffer! %lu\n",
 131                                offset);
 132                         return 1;
 133                 }
 134                 cur_len = min(len, map_len - (offset - map_start));
 135                 crc = btrfs_csum_data(root, kaddr + offset - map_start,
 136                                       crc, cur_len);
 137                 len -= cur_len;
 138                 offset += cur_len;
 139                 unmap_extent_buffer(buf, map_token, KM_USER0);
 140         }
 141         btrfs_csum_final(crc, result);
 142
 143         if (verify) {
 144                 int from_this_trans = 0;
 145
 146                 if (root->fs_info->running_transaction &&
 147                     btrfs_header_generation(buf) ==
 148                     root->fs_info->running_transaction->transid)
 149                         from_this_trans = 1;
 150
 151                 /* FIXME, this is not good */
 152                 if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) {
 153                         u32 val;
 154                         u32 found = 0;
 155                         memcpy(&found, result, BTRFS_CRC32_SIZE);
 156
 157                         read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE);
 158                         WARN_ON(1);
 159                         printk("btrfs: %s checksum verify failed on %llu "
 160                                "wanted %X found %X from_this_trans %d "
 161                                "level %d\n",
 162                                root->fs_info->sb->s_id,
 163                                buf->start, val, found, from_this_trans,
 164                                btrfs_header_level(buf));
 165                         return 1;
 166                 }
 167         } else {
 168                 write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE);
 169         }
 170         return 0;
 171 }
 172
 173
 174 int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
 175 {
 176         struct extent_io_tree *tree;
 177         u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
 178         u64 found_start;
 179         int found_level;
 180         unsigned long len;
 181         struct extent_buffer *eb;
 182         tree = &BTRFS_I(page->mapping->host)->io_tree;
 183
 184         if (page->private == EXTENT_PAGE_PRIVATE)
 185                 goto out;
 186         if (!page->private)
 187                 goto out;
 188         len = page->private >> 2;
 189         if (len == 0) {
 190                 WARN_ON(1);
 191         }
 192         eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
 193         read_extent_buffer_pages(tree, eb, start + PAGE_CACHE_SIZE, 1,
 194                                  btree_get_extent);
 195         btrfs_clear_buffer_defrag(eb);
 196         found_start = btrfs_header_bytenr(eb);
 197         if (found_start != start) {
 198                 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
 199                        start, found_start, len);
 200                 WARN_ON(1);
 201                 goto err;
 202         }
 203         if (eb->first_page != page) {
 204                 printk("bad first page %lu %lu\n", eb->first_page->index,
 205                        page->index);
 206                 WARN_ON(1);
 207                 goto err;
 208         }
 209         if (!PageUptodate(page)) {
 210                 printk("csum not up to date page %lu\n", page->index);
 211                 WARN_ON(1);
 212                 goto err;
 213         }
 214         found_level = btrfs_header_level(eb);
 215         spin_lock(&root->fs_info->hash_lock);
 216         btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
 217         spin_unlock(&root->fs_info->hash_lock);
 218         csum_tree_block(root, eb, 0);
 219 err:
 220         free_extent_buffer(eb);
 221 out:
 222         return 0;
 223 }
 224
 225 static int btree_writepage_io_hook(struct page *page, u64 start, u64 end)
 226 {
 227         struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
 228
 229         csum_dirty_buffer(root, page);
 230         return 0;
 231 }
 232
 233 int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 234                                struct extent_state *state)
 235 {
 236         struct extent_io_tree *tree;
 237         u64 found_start;
 238         int found_level;
 239         unsigned long len;
 240         struct extent_buffer *eb;
 241         struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
 242         int ret;
 243
 244         tree = &BTRFS_I(page->mapping->host)->io_tree;
 245         if (page->private == EXTENT_PAGE_PRIVATE)
 246                 goto out;
 247         if (!page->private)
 248                 goto out;
 249         len = page->private >> 2;
 250         if (len == 0) {
 251                 WARN_ON(1);
 252         }
 253         eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
 254         read_extent_buffer_pages(tree, eb, start + PAGE_CACHE_SIZE, 1,
 255                                  btree_get_extent);
 256         btrfs_clear_buffer_defrag(eb);
 257         found_start = btrfs_header_bytenr(eb);
 258         if (found_start != start) {
 259                 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
 260                        start, found_start, len);
 261                 WARN_ON(1);
 262                 goto err;
 263         }
 264         if (eb->first_page != page) {
 265                 printk("bad first page %lu %lu\n", eb->first_page->index,
 266                        page->index);
 267                 WARN_ON(1);
 268                 goto err;
 269         }
 270         found_level = btrfs_header_level(eb);
 271
 272         ret = csum_tree_block(root, eb, 1);
 273
 274         end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
 275         end = eb->start + end - 1;
 276         release_extent_buffer_tail_pages(eb);
 277 err:
 278         free_extent_buffer(eb);
 279 out:
 280         return 0;
 281 }
 282
 283 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
 284 static void end_workqueue_bio(struct bio *bio, int err)
 285 #else
 286 static int end_workqueue_bio(struct bio *bio,
 287                                    unsigned int bytes_done, int err)
 288 #endif
 289 {
 290         struct end_io_wq *end_io_wq = bio->bi_private;
 291         struct btrfs_fs_info *fs_info;
 292         unsigned long flags;
 293
 294 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
 295         if (bio->bi_size)
 296                 return 1;
 297 #endif
 298
 299         fs_info = end_io_wq->info;
 300         spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
 301         end_io_wq->error = err;
 302         list_add_tail(&end_io_wq->list, &fs_info->end_io_work_list);
 303         spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags);
 304         queue_work(end_io_workqueue, &fs_info->end_io_work);
 305
 306 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
 307         return 0;
 308 #endif
 309 }
 310
 311 static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio)
 312 {
 313         struct btrfs_root *root = BTRFS_I(inode)->root;
 314         struct end_io_wq *end_io_wq;
 315         u64 offset;
 316         offset = bio->bi_sector << 9;
 317
 318         if (rw & (1 << BIO_RW)) {
 319                 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio);
 320         }
 321
 322         end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
 323         if (!end_io_wq)
 324                 return -ENOMEM;
 325
 326         end_io_wq->private = bio->bi_private;
 327         end_io_wq->end_io = bio->bi_end_io;
 328         end_io_wq->info = root->fs_info;
 329         end_io_wq->error = 0;
 330         end_io_wq->bio = bio;
 331
 332         bio->bi_private = end_io_wq;
 333         bio->bi_end_io = end_workqueue_bio;
 334
 335         if (offset == BTRFS_SUPER_INFO_OFFSET) {
 336                 bio->bi_bdev = root->fs_info->sb->s_bdev;
 337                 submit_bio(rw, bio);
 338                 return 0;
 339         }
 340         return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio);
 341 }
 342
 343 static int btree_writepage(struct page *page, struct writeback_control *wbc)
 344 {
 345         struct extent_io_tree *tree;
 346         tree = &BTRFS_I(page->mapping->host)->io_tree;
 347         return extent_write_full_page(tree, page, btree_get_extent, wbc);
 348 }
 349
 350 static int btree_writepages(struct address_space *mapping,
 351                             struct writeback_control *wbc)
 352 {
 353         struct extent_io_tree *tree;
 354         tree = &BTRFS_I(mapping->host)->io_tree;
 355         if (wbc->sync_mode == WB_SYNC_NONE) {
 356                 u64 num_dirty;
 357                 u64 start = 0;
 358                 unsigned long thresh = 96 * 1024 * 1024;
 359
 360                 if (wbc->for_kupdate)
 361                         return 0;
 362
 363                 if (current_is_pdflush()) {
 364                         thresh = 96 * 1024 * 1024;
 365                 } else {
 366                         thresh = 8 * 1024 * 1024;
 367                 }
 368                 num_dirty = count_range_bits(tree, &start, (u64)-1,
 369                                              thresh, EXTENT_DIRTY);
 370                 if (num_dirty < thresh) {
 371                         return 0;
 372                 }
 373         }
 374         return extent_writepages(tree, mapping, btree_get_extent, wbc);
 375 }
 376
 377 int btree_readpage(struct file *file, struct page *page)
 378 {
 379         struct extent_io_tree *tree;
 380         tree = &BTRFS_I(page->mapping->host)->io_tree;
 381         return extent_read_full_page(tree, page, btree_get_extent);
 382 }
 383
 384 static int btree_releasepage(struct page *page, gfp_t gfp_flags)
 385 {
 386         struct extent_io_tree *tree;
 387         struct extent_map_tree *map;
 388         int ret;
 389
 390         tree = &BTRFS_I(page->mapping->host)->io_tree;
 391         map = &BTRFS_I(page->mapping->host)->extent_tree;
 392         ret = try_release_extent_mapping(map, tree, page, gfp_flags);
 393         if (ret == 1) {
 394                 invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE);
 395                 ClearPagePrivate(page);
 396                 set_page_private(page, 0);
 397                 page_cache_release(page);
 398         }
 399         return ret;
 400 }
 401
 402 static void btree_invalidatepage(struct page *page, unsigned long offset)
 403 {
 404         struct extent_io_tree *tree;
 405         tree = &BTRFS_I(page->mapping->host)->io_tree;
 406         extent_invalidatepage(tree, page, offset);
 407         btree_releasepage(page, GFP_NOFS);
 408 }
 409
 410 #if 0
 411 static int btree_writepage(struct page *page, struct writeback_control *wbc)
 412 {
 413         struct buffer_head *bh;
 414         struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
 415         struct buffer_head *head;
 416         if (!page_has_buffers(page)) {
 417                 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
 418                                         (1 << BH_Dirty)|(1 << BH_Uptodate));
 419         }
 420         head = page_buffers(page);
 421         bh = head;
 422         do {
 423                 if (buffer_dirty(bh))
 424                         csum_tree_block(root, bh, 0);
 425                 bh = bh->b_this_page;
 426         } while (bh != head);
 427         return block_write_full_page(page, btree_get_block, wbc);
 428 }
 429 #endif
 430
 431 static struct address_space_operations btree_aops = {
 432         .readpage       = btree_readpage,
 433         .writepage      = btree_writepage,
 434         .writepages     = btree_writepages,
 435         .releasepage    = btree_releasepage,
 436         .invalidatepage = btree_invalidatepage,
 437         .sync_page      = block_sync_page,
 438 };
 439
 440 int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
 441 {
 442         struct extent_buffer *buf = NULL;
 443         struct inode *btree_inode = root->fs_info->btree_inode;
 444         int ret = 0;
 445
 446         buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
 447         if (!buf)
 448                 return 0;
 449         read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
 450                                  buf, 0, 0, btree_get_extent);
 451         free_extent_buffer(buf);
 452         return ret;
 453 }
 454
 455 static int close_all_devices(struct btrfs_fs_info *fs_info)
 456 {
 457         struct list_head *list;
 458         struct list_head *next;
 459         struct btrfs_device *device;
 460
 461         list = &fs_info->fs_devices->devices;
 462         list_for_each(next, list) {
 463                 device = list_entry(next, struct btrfs_device, dev_list);
 464                 if (device->bdev && device->bdev != fs_info->sb->s_bdev)
 465                         close_bdev_excl(device->bdev);
 466                 device->bdev = NULL;
 467         }
 468         return 0;
 469 }
 470
 471 int btrfs_verify_block_csum(struct btrfs_root *root,
 472                             struct extent_buffer *buf)
 473 {
 474         return btrfs_buffer_uptodate(buf);
 475 }
 476
 477 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
 478                                             u64 bytenr, u32 blocksize)
 479 {
 480         struct inode *btree_inode = root->fs_info->btree_inode;
 481         struct extent_buffer *eb;
 482         eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
 483                                 bytenr, blocksize, GFP_NOFS);
 484         return eb;
 485 }
 486
 487 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
 488                                                  u64 bytenr, u32 blocksize)
 489 {
 490         struct inode *btree_inode = root->fs_info->btree_inode;
 491         struct extent_buffer *eb;
 492
 493         eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
 494                                  bytenr, blocksize, NULL, GFP_NOFS);
 495         return eb;
 496 }
 497
 498
 499 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
 500                                       u32 blocksize)
 501 {
 502         struct extent_buffer *buf = NULL;
 503         struct inode *btree_inode = root->fs_info->btree_inode;
 504         struct extent_io_tree *io_tree;
 505         int ret;
 506
 507         io_tree = &BTRFS_I(btree_inode)->io_tree;
 508
 509         buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
 510         if (!buf)
 511                 return NULL;
 512
 513         ret = read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, buf, 0,
 514                                        1, btree_get_extent);
 515
 516         if (ret == 0) {
 517                 buf->flags |= EXTENT_UPTODATE;
 518         }
 519         return buf;
 520
 521 }
 522
 523 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 524                      struct extent_buffer *buf)
 525 {
 526         struct inode *btree_inode = root->fs_info->btree_inode;
 527         if (btrfs_header_generation(buf) ==
 528             root->fs_info->running_transaction->transid)
 529                 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
 530                                           buf);
 531         return 0;
 532 }
 533
 534 int wait_on_tree_block_writeback(struct btrfs_root *root,
 535                                  struct extent_buffer *buf)
 536 {
 537         struct inode *btree_inode = root->fs_info->btree_inode;
 538         wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree,
 539                                         buf);
 540         return 0;
 541 }
 542
 543 static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 544                         u32 stripesize, struct btrfs_root *root,
 545                         struct btrfs_fs_info *fs_info,
 546                         u64 objectid)
 547 {
 548         root->node = NULL;
 549         root->inode = NULL;
 550         root->commit_root = NULL;
 551         root->sectorsize = sectorsize;
 552         root->nodesize = nodesize;
 553         root->leafsize = leafsize;
 554         root->stripesize = stripesize;
 555         root->ref_cows = 0;
 556         root->track_dirty = 0;
 557
 558         root->fs_info = fs_info;
 559         root->objectid = objectid;
 560         root->last_trans = 0;
 561         root->highest_inode = 0;
 562         root->last_inode_alloc = 0;
 563         root->name = NULL;
 564         root->in_sysfs = 0;
 565
 566         INIT_LIST_HEAD(&root->dirty_list);
 567         memset(&root->root_key, 0, sizeof(root->root_key));
 568         memset(&root->root_item, 0, sizeof(root->root_item));
 569         memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
 570         memset(&root->root_kobj, 0, sizeof(root->root_kobj));
 571         init_completion(&root->kobj_unregister);
 572         root->defrag_running = 0;
 573         root->defrag_level = 0;
 574         root->root_key.objectid = objectid;
 575         return 0;
 576 }
 577
 578 static int find_and_setup_root(struct btrfs_root *tree_root,
 579                                struct btrfs_fs_info *fs_info,
 580                                u64 objectid,
 581                                struct btrfs_root *root)
 582 {
 583         int ret;
 584         u32 blocksize;
 585
 586         __setup_root(tree_root->nodesize, tree_root->leafsize,
 587                      tree_root->sectorsize, tree_root->stripesize,
 588                      root, fs_info, objectid);
 589         ret = btrfs_find_last_root(tree_root, objectid,
 590                                    &root->root_item, &root->root_key);
 591         BUG_ON(ret);
 592
 593         blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
 594         root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
 595                                      blocksize);
 596         BUG_ON(!root->node);
 597         return 0;
 598 }
 599
 600 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 601                                                struct btrfs_key *location)
 602 {
 603         struct btrfs_root *root;
 604         struct btrfs_root *tree_root = fs_info->tree_root;
 605         struct btrfs_path *path;
 606         struct extent_buffer *l;
 607         u64 highest_inode;
 608         u32 blocksize;
 609         int ret = 0;
 610
 611         root = kzalloc(sizeof(*root), GFP_NOFS);
 612         if (!root)
 613                 return ERR_PTR(-ENOMEM);
 614         if (location->offset == (u64)-1) {
 615                 ret = find_and_setup_root(tree_root, fs_info,
 616                                           location->objectid, root);
 617                 if (ret) {
 618                         kfree(root);
 619                         return ERR_PTR(ret);
 620                 }
 621                 goto insert;
 622         }
 623
 624         __setup_root(tree_root->nodesize, tree_root->leafsize,
 625                      tree_root->sectorsize, tree_root->stripesize,
 626                      root, fs_info, location->objectid);
 627
 628         path = btrfs_alloc_path();
 629         BUG_ON(!path);
 630         ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
 631         if (ret != 0) {
 632                 if (ret > 0)
 633                         ret = -ENOENT;
 634                 goto out;
 635         }
 636         l = path->nodes[0];
 637         read_extent_buffer(l, &root->root_item,
 638                btrfs_item_ptr_offset(l, path->slots[0]),
 639                sizeof(root->root_item));
 640         memcpy(&root->root_key, location, sizeof(*location));
 641         ret = 0;
 642 out:
 643         btrfs_release_path(root, path);
 644         btrfs_free_path(path);
 645         if (ret) {
 646                 kfree(root);
 647                 return ERR_PTR(ret);
 648         }
 649         blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
 650         root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
 651                                      blocksize);
 652         BUG_ON(!root->node);
 653 insert:
 654         root->ref_cows = 1;
 655         ret = btrfs_find_highest_inode(root, &highest_inode);
 656         if (ret == 0) {
 657                 root->highest_inode = highest_inode;
 658                 root->last_inode_alloc = highest_inode;
 659         }
 660         return root;
 661 }
 662
 663 struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
 664                                         u64 root_objectid)
 665 {
 666         struct btrfs_root *root;
 667
 668         if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
 669                 return fs_info->tree_root;
 670         if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
 671                 return fs_info->extent_root;
 672
 673         root = radix_tree_lookup(&fs_info->fs_roots_radix,
 674                                  (unsigned long)root_objectid);
 675         return root;
 676 }
 677
 678 struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
 679                                               struct btrfs_key *location)
 680 {
 681         struct btrfs_root *root;
 682         int ret;
 683
 684         if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
 685                 return fs_info->tree_root;
 686         if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
 687                 return fs_info->extent_root;
 688
 689         root = radix_tree_lookup(&fs_info->fs_roots_radix,
 690                                  (unsigned long)location->objectid);
 691         if (root)
 692                 return root;
 693
 694         root = btrfs_read_fs_root_no_radix(fs_info, location);
 695         if (IS_ERR(root))
 696                 return root;
 697         ret = radix_tree_insert(&fs_info->fs_roots_radix,
 698                                 (unsigned long)root->root_key.objectid,
 699                                 root);
 700         if (ret) {
 701                 free_extent_buffer(root->node);
 702                 kfree(root);
 703                 return ERR_PTR(ret);
 704         }
 705         ret = btrfs_find_dead_roots(fs_info->tree_root,
 706                                     root->root_key.objectid, root);
 707         BUG_ON(ret);
 708
 709         return root;
 710 }
 711
 712 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 713                                       struct btrfs_key *location,
 714                                       const char *name, int namelen)
 715 {
 716         struct btrfs_root *root;
 717         int ret;
 718
 719         root = btrfs_read_fs_root_no_name(fs_info, location);
 720         if (!root)
 721                 return NULL;
 722
 723         if (root->in_sysfs)
 724                 return root;
 725
 726         ret = btrfs_set_root_name(root, name, namelen);
 727         if (ret) {
 728                 free_extent_buffer(root->node);
 729                 kfree(root);
 730                 return ERR_PTR(ret);
 731         }
 732
 733         ret = btrfs_sysfs_add_root(root);
 734         if (ret) {
 735                 free_extent_buffer(root->node);
 736                 kfree(root->name);
 737                 kfree(root);
 738                 return ERR_PTR(ret);
 739         }
 740         root->in_sysfs = 1;
 741         return root;
 742 }
 743 #if 0
 744 static int add_hasher(struct btrfs_fs_info *info, char *type) {
 745         struct btrfs_hasher *hasher;
 746
 747         hasher = kmalloc(sizeof(*hasher), GFP_NOFS);
 748         if (!hasher)
 749                 return -ENOMEM;
 750         hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC);
 751         if (!hasher->hash_tfm) {
 752                 kfree(hasher);
 753                 return -EINVAL;
 754         }
 755         spin_lock(&info->hash_lock);
 756         list_add(&hasher->list, &info->hashers);
 757         spin_unlock(&info->hash_lock);
 758         return 0;
 759 }
 760 #endif
 761
 762 static int btrfs_congested_fn(void *congested_data, int bdi_bits)
 763 {
 764         struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
 765         int ret = 0;
 766         struct list_head *cur;
 767         struct btrfs_device *device;
 768         struct backing_dev_info *bdi;
 769
 770         list_for_each(cur, &info->fs_devices->devices) {
 771                 device = list_entry(cur, struct btrfs_device, dev_list);
 772                 bdi = blk_get_backing_dev_info(device->bdev);
 773                 if (bdi && bdi_congested(bdi, bdi_bits)) {
 774                         ret = 1;
 775                         break;
 776                 }
 777         }
 778         return ret;
 779 }
 780
 781 void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
 782 {
 783         struct list_head *cur;
 784         struct btrfs_device *device;
 785         struct btrfs_fs_info *info;
 786
 787         info = (struct btrfs_fs_info *)bdi->unplug_io_data;
 788         list_for_each(cur, &info->fs_devices->devices) {
 789                 device = list_entry(cur, struct btrfs_device, dev_list);
 790                 bdi = blk_get_backing_dev_info(device->bdev);
 791                 if (bdi->unplug_io_fn) {
 792                         bdi->unplug_io_fn(bdi, page);
 793                 }
 794         }
 795 }
 796
 797 static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
 798 {
 799         bdi_init(bdi);
 800         bdi->ra_pages   = default_backing_dev_info.ra_pages * 4;
 801         bdi->state              = 0;
 802         bdi->capabilities       = default_backing_dev_info.capabilities;
 803         bdi->unplug_io_fn       = btrfs_unplug_io_fn;
 804         bdi->unplug_io_data     = info;
 805         bdi->congested_fn       = btrfs_congested_fn;
 806         bdi->congested_data     = info;
 807         return 0;
 808 }
 809
 810 static int bio_ready_for_csum(struct bio *bio)
 811 {
 812         u64 length = 0;
 813         u64 buf_len = 0;
 814         u64 start = 0;
 815         struct page *page;
 816         struct extent_io_tree *io_tree = NULL;
 817         struct btrfs_fs_info *info = NULL;
 818         struct bio_vec *bvec;
 819         int i;
 820         int ret;
 821
 822         bio_for_each_segment(bvec, bio, i) {
 823                 page = bvec->bv_page;
 824                 if (page->private == EXTENT_PAGE_PRIVATE) {
 825                         length += bvec->bv_len;
 826                         continue;
 827                 }
 828                 if (!page->private) {
 829                         length += bvec->bv_len;
 830                         continue;
 831                 }
 832                 length = bvec->bv_len;
 833                 buf_len = page->private >> 2;
 834                 start = page_offset(page) + bvec->bv_offset;
 835                 io_tree = &BTRFS_I(page->mapping->host)->io_tree;
 836                 info = BTRFS_I(page->mapping->host)->root->fs_info;
 837         }
 838         /* are we fully contained in this bio? */
 839         if (buf_len <= length)
 840                 return 1;
 841
 842         ret = extent_range_uptodate(io_tree, start + length,
 843                                     start + buf_len - 1);
 844         if (ret == 1)
 845                 return ret;
 846         return ret;
 847 }
 848
 849 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
 850 void btrfs_end_io_csum(void *p)
 851 #else
 852 void btrfs_end_io_csum(struct work_struct *work)
 853 #endif
 854 {
 855 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
 856         struct btrfs_fs_info *fs_info = p;
 857 #else
 858         struct btrfs_fs_info *fs_info = container_of(work,
 859                                                      struct btrfs_fs_info,
 860                                                      end_io_work);
 861 #endif
 862         unsigned long flags;
 863         struct end_io_wq *end_io_wq;
 864         struct bio *bio;
 865         struct list_head *next;
 866         int error;
 867         int was_empty;
 868
 869         while(1) {
 870                 spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
 871                 if (list_empty(&fs_info->end_io_work_list)) {
 872                         spin_unlock_irqrestore(&fs_info->end_io_work_lock,
 873                                                flags);
 874                         return;
 875                 }
 876                 next = fs_info->end_io_work_list.next;
 877                 list_del(next);
 878                 spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags);
 879
 880                 end_io_wq = list_entry(next, struct end_io_wq, list);
 881
 882                 bio = end_io_wq->bio;
 883                 if (!bio_ready_for_csum(bio)) {
 884                         spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
 885                         was_empty = list_empty(&fs_info->end_io_work_list);
 886                         list_add_tail(&end_io_wq->list,
 887                                       &fs_info->end_io_work_list);
 888                         spin_unlock_irqrestore(&fs_info->end_io_work_lock,
 889                                                flags);
 890                         if (was_empty)
 891                                 return;
 892                         continue;
 893                 }
 894                 error = end_io_wq->error;
 895                 bio->bi_private = end_io_wq->private;
 896                 bio->bi_end_io = end_io_wq->end_io;
 897                 kfree(end_io_wq);
 898                 bio_endio(bio, error);
 899         }
 900 }
 901
 902
 903 struct btrfs_root *open_ctree(struct super_block *sb,
 904                               struct btrfs_fs_devices *fs_devices)
 905 {
 906         u32 sectorsize;
 907         u32 nodesize;
 908         u32 leafsize;
 909         u32 blocksize;
 910         u32 stripesize;
 911         struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
 912                                                  GFP_NOFS);
 913         struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
 914                                                GFP_NOFS);
 915         struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
 916                                                 GFP_NOFS);
 917         struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root),
 918                                                 GFP_NOFS);
 919         struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root),
 920                                               GFP_NOFS);
 921         int ret;
 922         int err = -EINVAL;
 923         struct btrfs_super_block *disk_super;
 924
 925         if (!extent_root || !tree_root || !fs_info) {
 926                 err = -ENOMEM;
 927                 goto fail;
 928         }
 929         end_io_workqueue = create_workqueue("btrfs-end-io");
 930         BUG_ON(!end_io_workqueue);
 931
 932         INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
 933         INIT_LIST_HEAD(&fs_info->trans_list);
 934         INIT_LIST_HEAD(&fs_info->dead_roots);
 935         INIT_LIST_HEAD(&fs_info->hashers);
 936         INIT_LIST_HEAD(&fs_info->end_io_work_list);
 937         spin_lock_init(&fs_info->hash_lock);
 938         spin_lock_init(&fs_info->end_io_work_lock);
 939         spin_lock_init(&fs_info->delalloc_lock);
 940         spin_lock_init(&fs_info->new_trans_lock);
 941
 942         init_completion(&fs_info->kobj_unregister);
 943         sb_set_blocksize(sb, 4096);
 944         fs_info->tree_root = tree_root;
 945         fs_info->extent_root = extent_root;
 946         fs_info->chunk_root = chunk_root;
 947         fs_info->dev_root = dev_root;
 948         fs_info->fs_devices = fs_devices;
 949         INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
 950         INIT_LIST_HEAD(&fs_info->space_info);
 951         btrfs_mapping_init(&fs_info->mapping_tree);
 952         fs_info->sb = sb;
 953         fs_info->max_extent = (u64)-1;
 954         fs_info->max_inline = 8192 * 1024;
 955         setup_bdi(fs_info, &fs_info->bdi);
 956         fs_info->btree_inode = new_inode(sb);
 957         fs_info->btree_inode->i_ino = 1;
 958         fs_info->btree_inode->i_nlink = 1;
 959         fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
 960         fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
 961         fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
 962
 963         extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
 964                              fs_info->btree_inode->i_mapping,
 965                              GFP_NOFS);
 966         extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
 967                              GFP_NOFS);
 968
 969         BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
 970
 971         extent_io_tree_init(&fs_info->free_space_cache,
 972                              fs_info->btree_inode->i_mapping, GFP_NOFS);
 973         extent_io_tree_init(&fs_info->block_group_cache,
 974                              fs_info->btree_inode->i_mapping, GFP_NOFS);
 975         extent_io_tree_init(&fs_info->pinned_extents,
 976                              fs_info->btree_inode->i_mapping, GFP_NOFS);
 977         extent_io_tree_init(&fs_info->pending_del,
 978                              fs_info->btree_inode->i_mapping, GFP_NOFS);
 979         extent_io_tree_init(&fs_info->extent_ins,
 980                              fs_info->btree_inode->i_mapping, GFP_NOFS);
 981         fs_info->do_barriers = 1;
 982
 983         INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum);
 984 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
 985         INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info);
 986 #else
 987         INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
 988 #endif
 989         BTRFS_I(fs_info->btree_inode)->root = tree_root;
 990         memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
 991                sizeof(struct btrfs_key));
 992         insert_inode_hash(fs_info->btree_inode);
 993         mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
 994
 995         mutex_init(&fs_info->trans_mutex);
 996         mutex_init(&fs_info->fs_mutex);
 997
 998 #if 0
 999         ret = add_hasher(fs_info, "crc32c");
1000         if (ret) {
1001                 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
1002                 err = -ENOMEM;
1003                 goto fail_iput;
1004         }
1005 #endif
1006         __setup_root(4096, 4096, 4096, 4096, tree_root,
1007                      fs_info, BTRFS_ROOT_TREE_OBJECTID);
1008
1009         fs_info->sb_buffer = read_tree_block(tree_root,
1010                                              BTRFS_SUPER_INFO_OFFSET,
1011                                              4096);
1012
1013         if (!fs_info->sb_buffer)
1014                 goto fail_iput;
1015
1016         read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0,
1017                            sizeof(fs_info->super_copy));
1018
1019         read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
1020                            (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
1021                            BTRFS_FSID_SIZE);
1022
1023         disk_super = &fs_info->super_copy;
1024         if (!btrfs_super_root(disk_super))
1025                 goto fail_sb_buffer;
1026
1027         if (btrfs_super_num_devices(disk_super) != fs_devices->num_devices) {
1028                 printk("Btrfs: wanted %llu devices, but found %llu\n",
1029                        (unsigned long long)btrfs_super_num_devices(disk_super),
1030                        (unsigned long long)fs_devices->num_devices);
1031                 goto fail_sb_buffer;
1032         }
1033         nodesize = btrfs_super_nodesize(disk_super);
1034         leafsize = btrfs_super_leafsize(disk_super);
1035         sectorsize = btrfs_super_sectorsize(disk_super);
1036         stripesize = btrfs_super_stripesize(disk_super);
1037         tree_root->nodesize = nodesize;
1038         tree_root->leafsize = leafsize;
1039         tree_root->sectorsize = sectorsize;
1040         tree_root->stripesize = stripesize;
1041         sb_set_blocksize(sb, sectorsize);
1042
1043         i_size_write(fs_info->btree_inode,
1044                      btrfs_super_total_bytes(disk_super));
1045
1046         if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
1047                     sizeof(disk_super->magic))) {
1048                 printk("btrfs: valid FS not found on %s\n", sb->s_id);
1049                 goto fail_sb_buffer;
1050         }
1051
1052         mutex_lock(&fs_info->fs_mutex);
1053
1054         ret = btrfs_read_sys_array(tree_root);
1055         BUG_ON(ret);
1056
1057         blocksize = btrfs_level_size(tree_root,
1058                                      btrfs_super_chunk_root_level(disk_super));
1059
1060         __setup_root(nodesize, leafsize, sectorsize, stripesize,
1061                      chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1062
1063         chunk_root->node = read_tree_block(chunk_root,
1064                                            btrfs_super_chunk_root(disk_super),
1065                                            blocksize);
1066         BUG_ON(!chunk_root->node);
1067
1068         ret = btrfs_read_chunk_tree(chunk_root);
1069         BUG_ON(ret);
1070
1071         blocksize = btrfs_level_size(tree_root,
1072                                      btrfs_super_root_level(disk_super));
1073
1074
1075         tree_root->node = read_tree_block(tree_root,
1076                                           btrfs_super_root(disk_super),
1077                                           blocksize);
1078         if (!tree_root->node)
1079                 goto fail_sb_buffer;
1080
1081
1082         ret = find_and_setup_root(tree_root, fs_info,
1083                                   BTRFS_EXTENT_TREE_OBJECTID, extent_root);
1084         if (ret)
1085                 goto fail_tree_root;
1086         extent_root->track_dirty = 1;
1087
1088         ret = find_and_setup_root(tree_root, fs_info,
1089                                   BTRFS_DEV_TREE_OBJECTID, dev_root);
1090         dev_root->track_dirty = 1;
1091
1092         if (ret)
1093                 goto fail_extent_root;
1094
1095         btrfs_read_block_groups(extent_root);
1096
1097         fs_info->generation = btrfs_super_generation(disk_super) + 1;
1098         fs_info->data_alloc_profile = (u64)-1;
1099         fs_info->metadata_alloc_profile = (u64)-1;
1100         fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
1101
1102         mutex_unlock(&fs_info->fs_mutex);
1103         return tree_root;
1104
1105 fail_extent_root:
1106         free_extent_buffer(extent_root->node);
1107 fail_tree_root:
1108         mutex_unlock(&fs_info->fs_mutex);
1109         free_extent_buffer(tree_root->node);
1110 fail_sb_buffer:
1111         free_extent_buffer(fs_info->sb_buffer);
1112         extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
1113 fail_iput:
1114         iput(fs_info->btree_inode);
1115 fail:
1116         close_all_devices(fs_info);
1117         kfree(extent_root);
1118         kfree(tree_root);
1119         bdi_destroy(&fs_info->bdi);
1120         kfree(fs_info);
1121         return ERR_PTR(err);
1122 }
1123
1124 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
1125                       *root)
1126 {
1127         int ret;
1128         struct extent_buffer *super = root->fs_info->sb_buffer;
1129         struct inode *btree_inode = root->fs_info->btree_inode;
1130         struct super_block *sb = root->fs_info->sb;
1131
1132         if (!btrfs_test_opt(root, NOBARRIER))
1133                 blkdev_issue_flush(sb->s_bdev, NULL);
1134         set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, super);
1135         ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping,
1136                                      super->start, super->len);
1137         if (!btrfs_test_opt(root, NOBARRIER))
1138                 blkdev_issue_flush(sb->s_bdev, NULL);
1139         return ret;
1140 }
1141
1142 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
1143 {
1144         radix_tree_delete(&fs_info->fs_roots_radix,
1145                           (unsigned long)root->root_key.objectid);
1146         if (root->in_sysfs)
1147                 btrfs_sysfs_del_root(root);
1148         if (root->inode)
1149                 iput(root->inode);
1150         if (root->node)
1151                 free_extent_buffer(root->node);
1152         if (root->commit_root)
1153                 free_extent_buffer(root->commit_root);
1154         if (root->name)
1155                 kfree(root->name);
1156         kfree(root);
1157         return 0;
1158 }
1159
1160 static int del_fs_roots(struct btrfs_fs_info *fs_info)
1161 {
1162         int ret;
1163         struct btrfs_root *gang[8];
1164         int i;
1165
1166         while(1) {
1167                 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
1168                                              (void **)gang, 0,
1169                                              ARRAY_SIZE(gang));
1170                 if (!ret)
1171                         break;
1172                 for (i = 0; i < ret; i++)
1173                         btrfs_free_fs_root(fs_info, gang[i]);
1174         }
1175         return 0;
1176 }
1177
1178 int close_ctree(struct btrfs_root *root)
1179 {
1180         int ret;
1181         struct btrfs_trans_handle *trans;
1182         struct btrfs_fs_info *fs_info = root->fs_info;
1183
1184         fs_info->closing = 1;
1185         btrfs_transaction_flush_work(root);
1186         mutex_lock(&fs_info->fs_mutex);
1187         btrfs_defrag_dirty_roots(root->fs_info);
1188         trans = btrfs_start_transaction(root, 1);
1189         ret = btrfs_commit_transaction(trans, root);
1190         /* run commit again to  drop the original snapshot */
1191         trans = btrfs_start_transaction(root, 1);
1192         btrfs_commit_transaction(trans, root);
1193         ret = btrfs_write_and_wait_transaction(NULL, root);
1194         BUG_ON(ret);
1195         write_ctree_super(NULL, root);
1196         mutex_unlock(&fs_info->fs_mutex);
1197
1198         if (fs_info->delalloc_bytes) {
1199                 printk("btrfs: at unmount delalloc count %Lu\n",
1200                        fs_info->delalloc_bytes);
1201         }
1202         if (fs_info->extent_root->node)
1203                 free_extent_buffer(fs_info->extent_root->node);
1204
1205         if (fs_info->tree_root->node)
1206                 free_extent_buffer(fs_info->tree_root->node);
1207
1208         if (root->fs_info->chunk_root->node);
1209                 free_extent_buffer(root->fs_info->chunk_root->node);
1210
1211         if (root->fs_info->dev_root->node);
1212                 free_extent_buffer(root->fs_info->dev_root->node);
1213
1214         free_extent_buffer(fs_info->sb_buffer);
1215
1216         btrfs_free_block_groups(root->fs_info);
1217         del_fs_roots(fs_info);
1218
1219         filemap_write_and_wait(fs_info->btree_inode->i_mapping);
1220
1221         extent_io_tree_empty_lru(&fs_info->free_space_cache);
1222         extent_io_tree_empty_lru(&fs_info->block_group_cache);
1223         extent_io_tree_empty_lru(&fs_info->pinned_extents);
1224         extent_io_tree_empty_lru(&fs_info->pending_del);
1225         extent_io_tree_empty_lru(&fs_info->extent_ins);
1226         extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
1227
1228         truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
1229         flush_workqueue(end_io_workqueue);
1230         destroy_workqueue(end_io_workqueue);
1231
1232         iput(fs_info->btree_inode);
1233 #if 0
1234         while(!list_empty(&fs_info->hashers)) {
1235                 struct btrfs_hasher *hasher;
1236                 hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher,
1237                                     hashers);
1238                 list_del(&hasher->hashers);
1239                 crypto_free_hash(&fs_info->hash_tfm);
1240                 kfree(hasher);
1241         }
1242 #endif
1243         close_all_devices(fs_info);
1244         btrfs_mapping_tree_free(&fs_info->mapping_tree);
1245         bdi_destroy(&fs_info->bdi);
1246
1247         kfree(fs_info->extent_root);
1248         kfree(fs_info->tree_root);
1249         kfree(fs_info->chunk_root);
1250         kfree(fs_info->dev_root);
1251         return 0;
1252 }
1253
1254 int btrfs_buffer_uptodate(struct extent_buffer *buf)
1255 {
1256         struct inode *btree_inode = buf->first_page->mapping->host;
1257         return extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf);
1258 }
1259
1260 int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
1261 {
1262         struct inode *btree_inode = buf->first_page->mapping->host;
1263         return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
1264                                           buf);
1265 }
1266
1267 void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
1268 {
1269         struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1270         u64 transid = btrfs_header_generation(buf);
1271         struct inode *btree_inode = root->fs_info->btree_inode;
1272
1273         if (transid != root->fs_info->generation) {
1274                 printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
1275                         (unsigned long long)buf->start,
1276                         transid, root->fs_info->generation);
1277                 WARN_ON(1);
1278         }
1279         set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf);
1280 }
1281
1282 void btrfs_throttle(struct btrfs_root *root)
1283 {
1284         struct backing_dev_info *bdi;
1285
1286         bdi = root->fs_info->sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
1287         if (root->fs_info->throttles && bdi_write_congested(bdi)) {
1288 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18)
1289                 congestion_wait(WRITE, HZ/20);
1290 #else
1291                 blk_congestion_wait(WRITE, HZ/20);
1292 #endif
1293         }
1294 }
1295
1296 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
1297 {
1298         balance_dirty_pages_ratelimited_nr(
1299                                    root->fs_info->btree_inode->i_mapping, 1);
1300 }
1301
1302 void btrfs_set_buffer_defrag(struct extent_buffer *buf)
1303 {
1304         struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1305         struct inode *btree_inode = root->fs_info->btree_inode;
1306         set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
1307                         buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS);
1308 }
1309
1310 void btrfs_set_buffer_defrag_done(struct extent_buffer *buf)
1311 {
1312         struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1313         struct inode *btree_inode = root->fs_info->btree_inode;
1314         set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
1315                         buf->start + buf->len - 1, EXTENT_DEFRAG_DONE,
1316                         GFP_NOFS);
1317 }
1318
1319 int btrfs_buffer_defrag(struct extent_buffer *buf)
1320 {
1321         struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1322         struct inode *btree_inode = root->fs_info->btree_inode;
1323         return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
1324                      buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0);
1325 }
1326
1327 int btrfs_buffer_defrag_done(struct extent_buffer *buf)
1328 {
1329         struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1330         struct inode *btree_inode = root->fs_info->btree_inode;
1331         return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
1332                      buf->start, buf->start + buf->len - 1,
1333                      EXTENT_DEFRAG_DONE, 0);
1334 }
1335
1336 int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf)
1337 {
1338         struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1339         struct inode *btree_inode = root->fs_info->btree_inode;
1340         return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
1341                      buf->start, buf->start + buf->len - 1,
1342                      EXTENT_DEFRAG_DONE, GFP_NOFS);
1343 }
1344
1345 int btrfs_clear_buffer_defrag(struct extent_buffer *buf)
1346 {
1347         struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1348         struct inode *btree_inode = root->fs_info->btree_inode;
1349         return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
1350                      buf->start, buf->start + buf->len - 1,
1351                      EXTENT_DEFRAG, GFP_NOFS);
1352 }
1353
1354 int btrfs_read_buffer(struct extent_buffer *buf)
1355 {
1356         struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1357         struct inode *btree_inode = root->fs_info->btree_inode;
1358         int ret;
1359         ret = read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
1360                                         buf, 0, 1, btree_get_extent);
1361         if (ret == 0) {
1362                 buf->flags |= EXTENT_UPTODATE;
1363         }
1364         return ret;
1365 }
1366
1367 static struct extent_io_ops btree_extent_io_ops = {
1368         .writepage_io_hook = btree_writepage_io_hook,
1369         .readpage_end_io_hook = btree_readpage_end_io_hook,
1370         .submit_bio_hook = btree_submit_bio_hook,
1371         /* note we're sharing with inode.c for the merge bio hook */
1372         .merge_bio_hook = btrfs_merge_bio_hook,
1373 };