2 * Copyright (C) 2007 Oracle. All rights reserved.
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.
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.
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.
19 #include <linux/version.h>
21 #include <linux/blkdev.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 <linux/kthread.h>
29 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20)
30 # include <linux/freezer.h>
32 # include <linux/sched.h>
37 #include "transaction.h"
38 #include "btrfs_inode.h"
40 #include "print-tree.h"
41 #include "async-thread.h"
43 #include "ref-cache.h"
46 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
48 if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) {
49 printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n",
50 (unsigned long long)extent_buffer_blocknr(buf),
51 (unsigned long long)btrfs_header_blocknr(buf));
58 static struct extent_io_ops btree_extent_io_ops;
59 static void end_workqueue_fn(struct btrfs_work *work);
65 struct btrfs_fs_info *info;
68 struct list_head list;
69 struct btrfs_work work;
72 struct async_submit_bio {
75 struct list_head list;
76 extent_submit_bio_hook_t *submit_bio_hook;
79 struct btrfs_work work;
82 struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
83 size_t page_offset, u64 start, u64 len,
86 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
87 struct extent_map *em;
90 spin_lock(&em_tree->lock);
91 em = lookup_extent_mapping(em_tree, start, len);
94 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
95 spin_unlock(&em_tree->lock);
98 spin_unlock(&em_tree->lock);
100 em = alloc_extent_map(GFP_NOFS);
102 em = ERR_PTR(-ENOMEM);
108 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
110 spin_lock(&em_tree->lock);
111 ret = add_extent_mapping(em_tree, em);
112 if (ret == -EEXIST) {
113 u64 failed_start = em->start;
114 u64 failed_len = em->len;
116 printk("failed to insert %Lu %Lu -> %Lu into tree\n",
117 em->start, em->len, em->block_start);
119 em = lookup_extent_mapping(em_tree, start, len);
121 printk("after failing, found %Lu %Lu %Lu\n",
122 em->start, em->len, em->block_start);
125 em = lookup_extent_mapping(em_tree, failed_start,
128 printk("double failure lookup gives us "
129 "%Lu %Lu -> %Lu\n", em->start,
130 em->len, em->block_start);
139 spin_unlock(&em_tree->lock);
147 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
149 return btrfs_crc32c(seed, data, len);
152 void btrfs_csum_final(u32 crc, char *result)
154 *(__le32 *)result = ~cpu_to_le32(crc);
157 static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
160 char result[BTRFS_CRC32_SIZE];
162 unsigned long cur_len;
163 unsigned long offset = BTRFS_CSUM_SIZE;
164 char *map_token = NULL;
166 unsigned long map_start;
167 unsigned long map_len;
171 len = buf->len - offset;
173 err = map_private_extent_buffer(buf, offset, 32,
175 &map_start, &map_len, KM_USER0);
177 printk("failed to map extent buffer! %lu\n",
181 cur_len = min(len, map_len - (offset - map_start));
182 crc = btrfs_csum_data(root, kaddr + offset - map_start,
186 unmap_extent_buffer(buf, map_token, KM_USER0);
188 btrfs_csum_final(crc, result);
191 int from_this_trans = 0;
193 if (root->fs_info->running_transaction &&
194 btrfs_header_generation(buf) ==
195 root->fs_info->running_transaction->transid)
198 /* FIXME, this is not good */
199 if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) {
202 memcpy(&found, result, BTRFS_CRC32_SIZE);
204 read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE);
205 printk("btrfs: %s checksum verify failed on %llu "
206 "wanted %X found %X from_this_trans %d "
208 root->fs_info->sb->s_id,
209 buf->start, val, found, from_this_trans,
210 btrfs_header_level(buf));
214 write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE);
219 static int verify_parent_transid(struct extent_io_tree *io_tree,
220 struct extent_buffer *eb, u64 parent_transid)
224 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
227 lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS);
228 if (extent_buffer_uptodate(io_tree, eb) &&
229 btrfs_header_generation(eb) == parent_transid) {
233 printk("parent transid verify failed on %llu wanted %llu found %llu\n",
234 (unsigned long long)eb->start,
235 (unsigned long long)parent_transid,
236 (unsigned long long)btrfs_header_generation(eb));
239 clear_extent_buffer_uptodate(io_tree, eb);
240 unlock_extent(io_tree, eb->start, eb->start + eb->len - 1,
246 static int btree_read_extent_buffer_pages(struct btrfs_root *root,
247 struct extent_buffer *eb,
248 u64 start, u64 parent_transid)
250 struct extent_io_tree *io_tree;
255 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
257 ret = read_extent_buffer_pages(io_tree, eb, start, 1,
258 btree_get_extent, mirror_num);
260 !verify_parent_transid(io_tree, eb, parent_transid))
263 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
269 if (mirror_num > num_copies)
275 int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
277 struct extent_io_tree *tree;
278 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
282 struct extent_buffer *eb;
285 tree = &BTRFS_I(page->mapping->host)->io_tree;
287 if (page->private == EXTENT_PAGE_PRIVATE)
291 len = page->private >> 2;
295 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
296 ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
297 btrfs_header_generation(eb));
299 found_start = btrfs_header_bytenr(eb);
300 if (found_start != start) {
301 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
302 start, found_start, len);
306 if (eb->first_page != page) {
307 printk("bad first page %lu %lu\n", eb->first_page->index,
312 if (!PageUptodate(page)) {
313 printk("csum not up to date page %lu\n", page->index);
317 found_level = btrfs_header_level(eb);
318 spin_lock(&root->fs_info->hash_lock);
319 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
320 spin_unlock(&root->fs_info->hash_lock);
321 csum_tree_block(root, eb, 0);
323 free_extent_buffer(eb);
328 static int btree_writepage_io_hook(struct page *page, u64 start, u64 end)
330 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
332 csum_dirty_buffer(root, page);
336 int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
337 struct extent_state *state)
339 struct extent_io_tree *tree;
343 struct extent_buffer *eb;
344 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
347 tree = &BTRFS_I(page->mapping->host)->io_tree;
348 if (page->private == EXTENT_PAGE_PRIVATE)
352 len = page->private >> 2;
356 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
358 found_start = btrfs_header_bytenr(eb);
359 if (found_start != start) {
363 if (eb->first_page != page) {
364 printk("bad first page %lu %lu\n", eb->first_page->index,
370 if (memcmp_extent_buffer(eb, root->fs_info->fsid,
371 (unsigned long)btrfs_header_fsid(eb),
373 printk("bad fsid on block %Lu\n", eb->start);
377 found_level = btrfs_header_level(eb);
379 ret = csum_tree_block(root, eb, 1);
383 end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
384 end = eb->start + end - 1;
386 free_extent_buffer(eb);
391 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
392 static void end_workqueue_bio(struct bio *bio, int err)
394 static int end_workqueue_bio(struct bio *bio,
395 unsigned int bytes_done, int err)
398 struct end_io_wq *end_io_wq = bio->bi_private;
399 struct btrfs_fs_info *fs_info;
401 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
406 fs_info = end_io_wq->info;
407 end_io_wq->error = err;
408 end_io_wq->work.func = end_workqueue_fn;
409 end_io_wq->work.flags = 0;
410 if (bio->bi_rw & (1 << BIO_RW))
411 btrfs_queue_worker(&fs_info->endio_write_workers,
414 btrfs_queue_worker(&fs_info->endio_workers, &end_io_wq->work);
416 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
421 int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
424 struct end_io_wq *end_io_wq;
425 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
429 end_io_wq->private = bio->bi_private;
430 end_io_wq->end_io = bio->bi_end_io;
431 end_io_wq->info = info;
432 end_io_wq->error = 0;
433 end_io_wq->bio = bio;
434 end_io_wq->metadata = metadata;
436 bio->bi_private = end_io_wq;
437 bio->bi_end_io = end_workqueue_bio;
441 static void run_one_async_submit(struct btrfs_work *work)
443 struct btrfs_fs_info *fs_info;
444 struct async_submit_bio *async;
446 async = container_of(work, struct async_submit_bio, work);
447 fs_info = BTRFS_I(async->inode)->root->fs_info;
448 atomic_dec(&fs_info->nr_async_submits);
449 async->submit_bio_hook(async->inode, async->rw, async->bio,
454 int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
455 int rw, struct bio *bio, int mirror_num,
456 extent_submit_bio_hook_t *submit_bio_hook)
458 struct async_submit_bio *async;
460 async = kmalloc(sizeof(*async), GFP_NOFS);
464 async->inode = inode;
467 async->mirror_num = mirror_num;
468 async->submit_bio_hook = submit_bio_hook;
469 async->work.func = run_one_async_submit;
470 async->work.flags = 0;
471 atomic_inc(&fs_info->nr_async_submits);
472 btrfs_queue_worker(&fs_info->workers, &async->work);
476 static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
479 struct btrfs_root *root = BTRFS_I(inode)->root;
483 offset = bio->bi_sector << 9;
486 * when we're called for a write, we're already in the async
487 * submission context. Just jump ingo btrfs_map_bio
489 if (rw & (1 << BIO_RW)) {
490 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
495 * called for a read, do the setup so that checksum validation
496 * can happen in the async kernel threads
498 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1);
501 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
504 static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
508 * kthread helpers are used to submit writes so that checksumming
509 * can happen in parallel across all CPUs
511 if (!(rw & (1 << BIO_RW))) {
512 return __btree_submit_bio_hook(inode, rw, bio, mirror_num);
514 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
515 inode, rw, bio, mirror_num,
516 __btree_submit_bio_hook);
519 static int btree_writepage(struct page *page, struct writeback_control *wbc)
521 struct extent_io_tree *tree;
522 tree = &BTRFS_I(page->mapping->host)->io_tree;
523 return extent_write_full_page(tree, page, btree_get_extent, wbc);
526 static int btree_writepages(struct address_space *mapping,
527 struct writeback_control *wbc)
529 struct extent_io_tree *tree;
530 tree = &BTRFS_I(mapping->host)->io_tree;
531 if (wbc->sync_mode == WB_SYNC_NONE) {
534 unsigned long thresh = 96 * 1024 * 1024;
536 if (wbc->for_kupdate)
539 if (current_is_pdflush()) {
540 thresh = 96 * 1024 * 1024;
542 thresh = 8 * 1024 * 1024;
544 num_dirty = count_range_bits(tree, &start, (u64)-1,
545 thresh, EXTENT_DIRTY);
546 if (num_dirty < thresh) {
550 return extent_writepages(tree, mapping, btree_get_extent, wbc);
553 int btree_readpage(struct file *file, struct page *page)
555 struct extent_io_tree *tree;
556 tree = &BTRFS_I(page->mapping->host)->io_tree;
557 return extent_read_full_page(tree, page, btree_get_extent);
560 static int btree_releasepage(struct page *page, gfp_t gfp_flags)
562 struct extent_io_tree *tree;
563 struct extent_map_tree *map;
566 tree = &BTRFS_I(page->mapping->host)->io_tree;
567 map = &BTRFS_I(page->mapping->host)->extent_tree;
569 ret = try_release_extent_state(map, tree, page, gfp_flags);
574 ret = try_release_extent_buffer(tree, page);
576 ClearPagePrivate(page);
577 set_page_private(page, 0);
578 page_cache_release(page);
584 static void btree_invalidatepage(struct page *page, unsigned long offset)
586 struct extent_io_tree *tree;
587 tree = &BTRFS_I(page->mapping->host)->io_tree;
588 extent_invalidatepage(tree, page, offset);
589 btree_releasepage(page, GFP_NOFS);
590 if (PagePrivate(page)) {
591 printk("warning page private not zero on page %Lu\n",
593 ClearPagePrivate(page);
594 set_page_private(page, 0);
595 page_cache_release(page);
600 static int btree_writepage(struct page *page, struct writeback_control *wbc)
602 struct buffer_head *bh;
603 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
604 struct buffer_head *head;
605 if (!page_has_buffers(page)) {
606 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
607 (1 << BH_Dirty)|(1 << BH_Uptodate));
609 head = page_buffers(page);
612 if (buffer_dirty(bh))
613 csum_tree_block(root, bh, 0);
614 bh = bh->b_this_page;
615 } while (bh != head);
616 return block_write_full_page(page, btree_get_block, wbc);
620 static struct address_space_operations btree_aops = {
621 .readpage = btree_readpage,
622 .writepage = btree_writepage,
623 .writepages = btree_writepages,
624 .releasepage = btree_releasepage,
625 .invalidatepage = btree_invalidatepage,
626 .sync_page = block_sync_page,
629 int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
632 struct extent_buffer *buf = NULL;
633 struct inode *btree_inode = root->fs_info->btree_inode;
636 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
639 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
640 buf, 0, 0, btree_get_extent, 0);
641 free_extent_buffer(buf);
645 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
646 u64 bytenr, u32 blocksize)
648 struct inode *btree_inode = root->fs_info->btree_inode;
649 struct extent_buffer *eb;
650 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
651 bytenr, blocksize, GFP_NOFS);
655 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
656 u64 bytenr, u32 blocksize)
658 struct inode *btree_inode = root->fs_info->btree_inode;
659 struct extent_buffer *eb;
661 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
662 bytenr, blocksize, NULL, GFP_NOFS);
667 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
668 u32 blocksize, u64 parent_transid)
670 struct extent_buffer *buf = NULL;
671 struct inode *btree_inode = root->fs_info->btree_inode;
672 struct extent_io_tree *io_tree;
675 io_tree = &BTRFS_I(btree_inode)->io_tree;
677 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
681 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
684 buf->flags |= EXTENT_UPTODATE;
690 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
691 struct extent_buffer *buf)
693 struct inode *btree_inode = root->fs_info->btree_inode;
694 if (btrfs_header_generation(buf) ==
695 root->fs_info->running_transaction->transid) {
696 WARN_ON(!btrfs_tree_locked(buf));
697 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
703 int wait_on_tree_block_writeback(struct btrfs_root *root,
704 struct extent_buffer *buf)
706 struct inode *btree_inode = root->fs_info->btree_inode;
707 wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree,
712 static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
713 u32 stripesize, struct btrfs_root *root,
714 struct btrfs_fs_info *fs_info,
719 root->commit_root = NULL;
720 root->ref_tree = NULL;
721 root->sectorsize = sectorsize;
722 root->nodesize = nodesize;
723 root->leafsize = leafsize;
724 root->stripesize = stripesize;
726 root->track_dirty = 0;
728 root->fs_info = fs_info;
729 root->objectid = objectid;
730 root->last_trans = 0;
731 root->highest_inode = 0;
732 root->last_inode_alloc = 0;
736 INIT_LIST_HEAD(&root->dirty_list);
737 INIT_LIST_HEAD(&root->orphan_list);
738 spin_lock_init(&root->node_lock);
739 spin_lock_init(&root->orphan_lock);
740 mutex_init(&root->objectid_mutex);
742 btrfs_leaf_ref_tree_init(&root->ref_tree_struct);
743 root->ref_tree = &root->ref_tree_struct;
745 memset(&root->root_key, 0, sizeof(root->root_key));
746 memset(&root->root_item, 0, sizeof(root->root_item));
747 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
748 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
749 root->defrag_trans_start = fs_info->generation;
750 init_completion(&root->kobj_unregister);
751 root->defrag_running = 0;
752 root->defrag_level = 0;
753 root->root_key.objectid = objectid;
757 static int find_and_setup_root(struct btrfs_root *tree_root,
758 struct btrfs_fs_info *fs_info,
760 struct btrfs_root *root)
765 __setup_root(tree_root->nodesize, tree_root->leafsize,
766 tree_root->sectorsize, tree_root->stripesize,
767 root, fs_info, objectid);
768 ret = btrfs_find_last_root(tree_root, objectid,
769 &root->root_item, &root->root_key);
772 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
773 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
779 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
780 struct btrfs_key *location)
782 struct btrfs_root *root;
783 struct btrfs_root *tree_root = fs_info->tree_root;
784 struct btrfs_path *path;
785 struct extent_buffer *l;
790 root = kzalloc(sizeof(*root), GFP_NOFS);
792 return ERR_PTR(-ENOMEM);
793 if (location->offset == (u64)-1) {
794 ret = find_and_setup_root(tree_root, fs_info,
795 location->objectid, root);
803 __setup_root(tree_root->nodesize, tree_root->leafsize,
804 tree_root->sectorsize, tree_root->stripesize,
805 root, fs_info, location->objectid);
807 path = btrfs_alloc_path();
809 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
816 read_extent_buffer(l, &root->root_item,
817 btrfs_item_ptr_offset(l, path->slots[0]),
818 sizeof(root->root_item));
819 memcpy(&root->root_key, location, sizeof(*location));
822 btrfs_release_path(root, path);
823 btrfs_free_path(path);
828 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
829 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
834 ret = btrfs_find_highest_inode(root, &highest_inode);
836 root->highest_inode = highest_inode;
837 root->last_inode_alloc = highest_inode;
842 struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
845 struct btrfs_root *root;
847 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
848 return fs_info->tree_root;
849 if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
850 return fs_info->extent_root;
852 root = radix_tree_lookup(&fs_info->fs_roots_radix,
853 (unsigned long)root_objectid);
857 struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
858 struct btrfs_key *location)
860 struct btrfs_root *root;
863 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
864 return fs_info->tree_root;
865 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
866 return fs_info->extent_root;
867 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
868 return fs_info->chunk_root;
869 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
870 return fs_info->dev_root;
872 root = radix_tree_lookup(&fs_info->fs_roots_radix,
873 (unsigned long)location->objectid);
877 root = btrfs_read_fs_root_no_radix(fs_info, location);
880 ret = radix_tree_insert(&fs_info->fs_roots_radix,
881 (unsigned long)root->root_key.objectid,
884 free_extent_buffer(root->node);
888 ret = btrfs_find_dead_roots(fs_info->tree_root,
889 root->root_key.objectid, root);
895 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
896 struct btrfs_key *location,
897 const char *name, int namelen)
899 struct btrfs_root *root;
902 root = btrfs_read_fs_root_no_name(fs_info, location);
909 ret = btrfs_set_root_name(root, name, namelen);
911 free_extent_buffer(root->node);
916 ret = btrfs_sysfs_add_root(root);
918 free_extent_buffer(root->node);
927 static int add_hasher(struct btrfs_fs_info *info, char *type) {
928 struct btrfs_hasher *hasher;
930 hasher = kmalloc(sizeof(*hasher), GFP_NOFS);
933 hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC);
934 if (!hasher->hash_tfm) {
938 spin_lock(&info->hash_lock);
939 list_add(&hasher->list, &info->hashers);
940 spin_unlock(&info->hash_lock);
945 static int btrfs_congested_fn(void *congested_data, int bdi_bits)
947 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
949 int limit = 256 * info->fs_devices->open_devices;
950 struct list_head *cur;
951 struct btrfs_device *device;
952 struct backing_dev_info *bdi;
954 if ((bdi_bits & (1 << BDI_write_congested)) &&
955 atomic_read(&info->nr_async_submits) > limit) {
959 list_for_each(cur, &info->fs_devices->devices) {
960 device = list_entry(cur, struct btrfs_device, dev_list);
963 bdi = blk_get_backing_dev_info(device->bdev);
964 if (bdi && bdi_congested(bdi, bdi_bits)) {
973 * this unplugs every device on the box, and it is only used when page
976 static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
978 struct list_head *cur;
979 struct btrfs_device *device;
980 struct btrfs_fs_info *info;
982 info = (struct btrfs_fs_info *)bdi->unplug_io_data;
983 list_for_each(cur, &info->fs_devices->devices) {
984 device = list_entry(cur, struct btrfs_device, dev_list);
985 bdi = blk_get_backing_dev_info(device->bdev);
986 if (bdi->unplug_io_fn) {
987 bdi->unplug_io_fn(bdi, page);
992 void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
995 struct extent_map_tree *em_tree;
996 struct extent_map *em;
997 struct address_space *mapping;
1000 /* the generic O_DIRECT read code does this */
1002 __unplug_io_fn(bdi, page);
1007 * page->mapping may change at any time. Get a consistent copy
1008 * and use that for everything below
1011 mapping = page->mapping;
1015 inode = mapping->host;
1016 offset = page_offset(page);
1018 em_tree = &BTRFS_I(inode)->extent_tree;
1019 spin_lock(&em_tree->lock);
1020 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
1021 spin_unlock(&em_tree->lock);
1023 __unplug_io_fn(bdi, page);
1027 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1028 free_extent_map(em);
1029 __unplug_io_fn(bdi, page);
1032 offset = offset - em->start;
1033 btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree,
1034 em->block_start + offset, page);
1035 free_extent_map(em);
1038 static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1040 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
1043 bdi->ra_pages = default_backing_dev_info.ra_pages;
1045 bdi->capabilities = default_backing_dev_info.capabilities;
1046 bdi->unplug_io_fn = btrfs_unplug_io_fn;
1047 bdi->unplug_io_data = info;
1048 bdi->congested_fn = btrfs_congested_fn;
1049 bdi->congested_data = info;
1053 static int bio_ready_for_csum(struct bio *bio)
1059 struct extent_io_tree *io_tree = NULL;
1060 struct btrfs_fs_info *info = NULL;
1061 struct bio_vec *bvec;
1065 bio_for_each_segment(bvec, bio, i) {
1066 page = bvec->bv_page;
1067 if (page->private == EXTENT_PAGE_PRIVATE) {
1068 length += bvec->bv_len;
1071 if (!page->private) {
1072 length += bvec->bv_len;
1075 length = bvec->bv_len;
1076 buf_len = page->private >> 2;
1077 start = page_offset(page) + bvec->bv_offset;
1078 io_tree = &BTRFS_I(page->mapping->host)->io_tree;
1079 info = BTRFS_I(page->mapping->host)->root->fs_info;
1081 /* are we fully contained in this bio? */
1082 if (buf_len <= length)
1085 ret = extent_range_uptodate(io_tree, start + length,
1086 start + buf_len - 1);
1093 * called by the kthread helper functions to finally call the bio end_io
1094 * functions. This is where read checksum verification actually happens
1096 static void end_workqueue_fn(struct btrfs_work *work)
1099 struct end_io_wq *end_io_wq;
1100 struct btrfs_fs_info *fs_info;
1103 end_io_wq = container_of(work, struct end_io_wq, work);
1104 bio = end_io_wq->bio;
1105 fs_info = end_io_wq->info;
1107 /* metadata bios are special because the whole tree block must
1108 * be checksummed at once. This makes sure the entire block is in
1109 * ram and up to date before trying to verify things. For
1110 * blocksize <= pagesize, it is basically a noop
1112 if (end_io_wq->metadata && !bio_ready_for_csum(bio)) {
1113 btrfs_queue_worker(&fs_info->endio_workers,
1117 error = end_io_wq->error;
1118 bio->bi_private = end_io_wq->private;
1119 bio->bi_end_io = end_io_wq->end_io;
1121 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
1122 bio_endio(bio, bio->bi_size, error);
1124 bio_endio(bio, error);
1128 static int cleaner_kthread(void *arg)
1130 struct btrfs_root *root = arg;
1134 if (root->fs_info->closing)
1137 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1138 mutex_lock(&root->fs_info->cleaner_mutex);
1139 btrfs_clean_old_snapshots(root);
1140 mutex_unlock(&root->fs_info->cleaner_mutex);
1142 if (freezing(current)) {
1146 if (root->fs_info->closing)
1148 set_current_state(TASK_INTERRUPTIBLE);
1150 __set_current_state(TASK_RUNNING);
1152 } while (!kthread_should_stop());
1156 static int transaction_kthread(void *arg)
1158 struct btrfs_root *root = arg;
1159 struct btrfs_trans_handle *trans;
1160 struct btrfs_transaction *cur;
1162 unsigned long delay;
1167 if (root->fs_info->closing)
1171 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1172 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1174 printk("btrfs: total reference cache size %Lu\n",
1175 root->fs_info->total_ref_cache_size);
1177 mutex_lock(&root->fs_info->trans_mutex);
1178 cur = root->fs_info->running_transaction;
1180 mutex_unlock(&root->fs_info->trans_mutex);
1184 now = get_seconds();
1185 if (now < cur->start_time || now - cur->start_time < 30) {
1186 mutex_unlock(&root->fs_info->trans_mutex);
1190 mutex_unlock(&root->fs_info->trans_mutex);
1191 trans = btrfs_start_transaction(root, 1);
1192 ret = btrfs_commit_transaction(trans, root);
1194 wake_up_process(root->fs_info->cleaner_kthread);
1195 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1197 if (freezing(current)) {
1200 if (root->fs_info->closing)
1202 set_current_state(TASK_INTERRUPTIBLE);
1203 schedule_timeout(delay);
1204 __set_current_state(TASK_RUNNING);
1206 } while (!kthread_should_stop());
1210 struct btrfs_root *open_ctree(struct super_block *sb,
1211 struct btrfs_fs_devices *fs_devices,
1219 struct buffer_head *bh;
1220 struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
1222 struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
1224 struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
1226 struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root),
1228 struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root),
1233 struct btrfs_super_block *disk_super;
1235 if (!extent_root || !tree_root || !fs_info) {
1239 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
1240 INIT_LIST_HEAD(&fs_info->trans_list);
1241 INIT_LIST_HEAD(&fs_info->dead_roots);
1242 INIT_LIST_HEAD(&fs_info->hashers);
1243 spin_lock_init(&fs_info->hash_lock);
1244 spin_lock_init(&fs_info->delalloc_lock);
1245 spin_lock_init(&fs_info->new_trans_lock);
1246 spin_lock_init(&fs_info->ref_cache_lock);
1248 init_completion(&fs_info->kobj_unregister);
1249 fs_info->tree_root = tree_root;
1250 fs_info->extent_root = extent_root;
1251 fs_info->chunk_root = chunk_root;
1252 fs_info->dev_root = dev_root;
1253 fs_info->fs_devices = fs_devices;
1254 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
1255 INIT_LIST_HEAD(&fs_info->space_info);
1256 btrfs_mapping_init(&fs_info->mapping_tree);
1257 atomic_set(&fs_info->nr_async_submits, 0);
1258 atomic_set(&fs_info->throttles, 0);
1260 fs_info->max_extent = (u64)-1;
1261 fs_info->max_inline = 8192 * 1024;
1262 setup_bdi(fs_info, &fs_info->bdi);
1263 fs_info->btree_inode = new_inode(sb);
1264 fs_info->btree_inode->i_ino = 1;
1265 fs_info->btree_inode->i_nlink = 1;
1266 fs_info->thread_pool_size = min(num_online_cpus() + 2, 8);
1268 INIT_LIST_HEAD(&fs_info->ordered_extents);
1269 spin_lock_init(&fs_info->ordered_extent_lock);
1271 sb->s_blocksize = 4096;
1272 sb->s_blocksize_bits = blksize_bits(4096);
1275 * we set the i_size on the btree inode to the max possible int.
1276 * the real end of the address space is determined by all of
1277 * the devices in the system
1279 fs_info->btree_inode->i_size = OFFSET_MAX;
1280 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
1281 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
1283 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
1284 fs_info->btree_inode->i_mapping,
1286 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
1289 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
1291 extent_io_tree_init(&fs_info->free_space_cache,
1292 fs_info->btree_inode->i_mapping, GFP_NOFS);
1293 extent_io_tree_init(&fs_info->block_group_cache,
1294 fs_info->btree_inode->i_mapping, GFP_NOFS);
1295 extent_io_tree_init(&fs_info->pinned_extents,
1296 fs_info->btree_inode->i_mapping, GFP_NOFS);
1297 extent_io_tree_init(&fs_info->pending_del,
1298 fs_info->btree_inode->i_mapping, GFP_NOFS);
1299 extent_io_tree_init(&fs_info->extent_ins,
1300 fs_info->btree_inode->i_mapping, GFP_NOFS);
1301 fs_info->do_barriers = 1;
1303 BTRFS_I(fs_info->btree_inode)->root = tree_root;
1304 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
1305 sizeof(struct btrfs_key));
1306 insert_inode_hash(fs_info->btree_inode);
1307 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1309 mutex_init(&fs_info->trans_mutex);
1310 mutex_init(&fs_info->drop_mutex);
1311 mutex_init(&fs_info->alloc_mutex);
1312 mutex_init(&fs_info->chunk_mutex);
1313 mutex_init(&fs_info->transaction_kthread_mutex);
1314 mutex_init(&fs_info->cleaner_mutex);
1315 mutex_init(&fs_info->volume_mutex);
1316 init_waitqueue_head(&fs_info->transaction_throttle);
1317 init_waitqueue_head(&fs_info->transaction_wait);
1320 ret = add_hasher(fs_info, "crc32c");
1322 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
1327 __setup_root(4096, 4096, 4096, 4096, tree_root,
1328 fs_info, BTRFS_ROOT_TREE_OBJECTID);
1331 bh = __bread(fs_devices->latest_bdev,
1332 BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
1336 memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy));
1339 memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE);
1341 disk_super = &fs_info->super_copy;
1342 if (!btrfs_super_root(disk_super))
1343 goto fail_sb_buffer;
1345 err = btrfs_parse_options(tree_root, options);
1347 goto fail_sb_buffer;
1350 * we need to start all the end_io workers up front because the
1351 * queue work function gets called at interrupt time, and so it
1352 * cannot dynamically grow.
1354 btrfs_init_workers(&fs_info->workers, fs_info->thread_pool_size);
1355 btrfs_init_workers(&fs_info->submit_workers, fs_info->thread_pool_size);
1356 btrfs_init_workers(&fs_info->fixup_workers, 1);
1357 btrfs_init_workers(&fs_info->endio_workers, fs_info->thread_pool_size);
1358 btrfs_init_workers(&fs_info->endio_write_workers,
1359 fs_info->thread_pool_size);
1360 btrfs_start_workers(&fs_info->workers, 1);
1361 btrfs_start_workers(&fs_info->submit_workers, 1);
1362 btrfs_start_workers(&fs_info->fixup_workers, 1);
1363 btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size);
1364 btrfs_start_workers(&fs_info->endio_write_workers,
1365 fs_info->thread_pool_size);
1368 if (btrfs_super_num_devices(disk_super) > fs_devices->open_devices) {
1369 printk("Btrfs: wanted %llu devices, but found %llu\n",
1370 (unsigned long long)btrfs_super_num_devices(disk_super),
1371 (unsigned long long)fs_devices->open_devices);
1372 if (btrfs_test_opt(tree_root, DEGRADED))
1373 printk("continuing in degraded mode\n");
1375 goto fail_sb_buffer;
1379 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
1381 nodesize = btrfs_super_nodesize(disk_super);
1382 leafsize = btrfs_super_leafsize(disk_super);
1383 sectorsize = btrfs_super_sectorsize(disk_super);
1384 stripesize = btrfs_super_stripesize(disk_super);
1385 tree_root->nodesize = nodesize;
1386 tree_root->leafsize = leafsize;
1387 tree_root->sectorsize = sectorsize;
1388 tree_root->stripesize = stripesize;
1390 sb->s_blocksize = sectorsize;
1391 sb->s_blocksize_bits = blksize_bits(sectorsize);
1393 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
1394 sizeof(disk_super->magic))) {
1395 printk("btrfs: valid FS not found on %s\n", sb->s_id);
1396 goto fail_sb_buffer;
1399 mutex_lock(&fs_info->chunk_mutex);
1400 ret = btrfs_read_sys_array(tree_root);
1401 mutex_unlock(&fs_info->chunk_mutex);
1403 printk("btrfs: failed to read the system array on %s\n",
1405 goto fail_sys_array;
1408 blocksize = btrfs_level_size(tree_root,
1409 btrfs_super_chunk_root_level(disk_super));
1411 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1412 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1414 chunk_root->node = read_tree_block(chunk_root,
1415 btrfs_super_chunk_root(disk_super),
1417 BUG_ON(!chunk_root->node);
1419 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
1420 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
1423 mutex_lock(&fs_info->chunk_mutex);
1424 ret = btrfs_read_chunk_tree(chunk_root);
1425 mutex_unlock(&fs_info->chunk_mutex);
1428 btrfs_close_extra_devices(fs_devices);
1430 blocksize = btrfs_level_size(tree_root,
1431 btrfs_super_root_level(disk_super));
1434 tree_root->node = read_tree_block(tree_root,
1435 btrfs_super_root(disk_super),
1437 if (!tree_root->node)
1438 goto fail_sb_buffer;
1441 ret = find_and_setup_root(tree_root, fs_info,
1442 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
1444 goto fail_tree_root;
1445 extent_root->track_dirty = 1;
1447 ret = find_and_setup_root(tree_root, fs_info,
1448 BTRFS_DEV_TREE_OBJECTID, dev_root);
1449 dev_root->track_dirty = 1;
1452 goto fail_extent_root;
1454 btrfs_read_block_groups(extent_root);
1456 fs_info->generation = btrfs_super_generation(disk_super) + 1;
1457 fs_info->data_alloc_profile = (u64)-1;
1458 fs_info->metadata_alloc_profile = (u64)-1;
1459 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
1460 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
1462 if (!fs_info->cleaner_kthread)
1463 goto fail_extent_root;
1465 fs_info->transaction_kthread = kthread_run(transaction_kthread,
1467 "btrfs-transaction");
1468 if (!fs_info->transaction_kthread)
1475 kthread_stop(fs_info->cleaner_kthread);
1477 free_extent_buffer(extent_root->node);
1479 free_extent_buffer(tree_root->node);
1482 btrfs_stop_workers(&fs_info->fixup_workers);
1483 btrfs_stop_workers(&fs_info->workers);
1484 btrfs_stop_workers(&fs_info->endio_workers);
1485 btrfs_stop_workers(&fs_info->endio_write_workers);
1486 btrfs_stop_workers(&fs_info->submit_workers);
1488 iput(fs_info->btree_inode);
1490 btrfs_close_devices(fs_info->fs_devices);
1491 btrfs_mapping_tree_free(&fs_info->mapping_tree);
1495 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
1496 bdi_destroy(&fs_info->bdi);
1499 return ERR_PTR(err);
1502 static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
1504 char b[BDEVNAME_SIZE];
1507 set_buffer_uptodate(bh);
1509 if (!buffer_eopnotsupp(bh) && printk_ratelimit()) {
1510 printk(KERN_WARNING "lost page write due to "
1511 "I/O error on %s\n",
1512 bdevname(bh->b_bdev, b));
1514 /* note, we dont' set_buffer_write_io_error because we have
1515 * our own ways of dealing with the IO errors
1517 clear_buffer_uptodate(bh);
1523 int write_all_supers(struct btrfs_root *root)
1525 struct list_head *cur;
1526 struct list_head *head = &root->fs_info->fs_devices->devices;
1527 struct btrfs_device *dev;
1528 struct btrfs_super_block *sb;
1529 struct btrfs_dev_item *dev_item;
1530 struct buffer_head *bh;
1534 int total_errors = 0;
1538 max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
1539 do_barriers = !btrfs_test_opt(root, NOBARRIER);
1541 sb = &root->fs_info->super_for_commit;
1542 dev_item = &sb->dev_item;
1543 list_for_each(cur, head) {
1544 dev = list_entry(cur, struct btrfs_device, dev_list);
1549 if (!dev->in_fs_metadata)
1552 btrfs_set_stack_device_type(dev_item, dev->type);
1553 btrfs_set_stack_device_id(dev_item, dev->devid);
1554 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
1555 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
1556 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
1557 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
1558 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
1559 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
1560 flags = btrfs_super_flags(sb);
1561 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
1565 crc = btrfs_csum_data(root, (char *)sb + BTRFS_CSUM_SIZE, crc,
1566 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1567 btrfs_csum_final(crc, sb->csum);
1569 bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / 4096,
1570 BTRFS_SUPER_INFO_SIZE);
1572 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
1573 dev->pending_io = bh;
1576 set_buffer_uptodate(bh);
1578 bh->b_end_io = btrfs_end_buffer_write_sync;
1580 if (do_barriers && dev->barriers) {
1581 ret = submit_bh(WRITE_BARRIER, bh);
1582 if (ret == -EOPNOTSUPP) {
1583 printk("btrfs: disabling barriers on dev %s\n",
1585 set_buffer_uptodate(bh);
1589 ret = submit_bh(WRITE, bh);
1592 ret = submit_bh(WRITE, bh);
1597 if (total_errors > max_errors) {
1598 printk("btrfs: %d errors while writing supers\n", total_errors);
1603 list_for_each(cur, head) {
1604 dev = list_entry(cur, struct btrfs_device, dev_list);
1607 if (!dev->in_fs_metadata)
1610 BUG_ON(!dev->pending_io);
1611 bh = dev->pending_io;
1613 if (!buffer_uptodate(dev->pending_io)) {
1614 if (do_barriers && dev->barriers) {
1615 printk("btrfs: disabling barriers on dev %s\n",
1617 set_buffer_uptodate(bh);
1621 ret = submit_bh(WRITE, bh);
1624 if (!buffer_uptodate(bh))
1631 dev->pending_io = NULL;
1634 if (total_errors > max_errors) {
1635 printk("btrfs: %d errors while writing supers\n", total_errors);
1641 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
1646 ret = write_all_supers(root);
1650 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
1652 radix_tree_delete(&fs_info->fs_roots_radix,
1653 (unsigned long)root->root_key.objectid);
1655 btrfs_sysfs_del_root(root);
1659 free_extent_buffer(root->node);
1660 if (root->commit_root)
1661 free_extent_buffer(root->commit_root);
1668 static int del_fs_roots(struct btrfs_fs_info *fs_info)
1671 struct btrfs_root *gang[8];
1675 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
1680 for (i = 0; i < ret; i++)
1681 btrfs_free_fs_root(fs_info, gang[i]);
1686 int close_ctree(struct btrfs_root *root)
1689 struct btrfs_trans_handle *trans;
1690 struct btrfs_fs_info *fs_info = root->fs_info;
1692 fs_info->closing = 1;
1695 kthread_stop(root->fs_info->transaction_kthread);
1696 kthread_stop(root->fs_info->cleaner_kthread);
1698 btrfs_clean_old_snapshots(root);
1699 trans = btrfs_start_transaction(root, 1);
1700 ret = btrfs_commit_transaction(trans, root);
1701 /* run commit again to drop the original snapshot */
1702 trans = btrfs_start_transaction(root, 1);
1703 btrfs_commit_transaction(trans, root);
1704 ret = btrfs_write_and_wait_transaction(NULL, root);
1707 write_ctree_super(NULL, root);
1709 if (fs_info->delalloc_bytes) {
1710 printk("btrfs: at unmount delalloc count %Lu\n",
1711 fs_info->delalloc_bytes);
1713 if (fs_info->total_ref_cache_size) {
1714 printk("btrfs: at umount reference cache size %Lu\n",
1715 fs_info->total_ref_cache_size);
1718 if (fs_info->extent_root->node)
1719 free_extent_buffer(fs_info->extent_root->node);
1721 if (fs_info->tree_root->node)
1722 free_extent_buffer(fs_info->tree_root->node);
1724 if (root->fs_info->chunk_root->node);
1725 free_extent_buffer(root->fs_info->chunk_root->node);
1727 if (root->fs_info->dev_root->node);
1728 free_extent_buffer(root->fs_info->dev_root->node);
1730 btrfs_free_block_groups(root->fs_info);
1731 del_fs_roots(fs_info);
1733 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
1735 truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
1737 btrfs_stop_workers(&fs_info->fixup_workers);
1738 btrfs_stop_workers(&fs_info->workers);
1739 btrfs_stop_workers(&fs_info->endio_workers);
1740 btrfs_stop_workers(&fs_info->endio_write_workers);
1741 btrfs_stop_workers(&fs_info->submit_workers);
1743 iput(fs_info->btree_inode);
1745 while(!list_empty(&fs_info->hashers)) {
1746 struct btrfs_hasher *hasher;
1747 hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher,
1749 list_del(&hasher->hashers);
1750 crypto_free_hash(&fs_info->hash_tfm);
1754 btrfs_close_devices(fs_info->fs_devices);
1755 btrfs_mapping_tree_free(&fs_info->mapping_tree);
1757 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
1758 bdi_destroy(&fs_info->bdi);
1761 kfree(fs_info->extent_root);
1762 kfree(fs_info->tree_root);
1763 kfree(fs_info->chunk_root);
1764 kfree(fs_info->dev_root);
1768 int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
1771 struct inode *btree_inode = buf->first_page->mapping->host;
1773 ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf);
1777 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
1782 int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
1784 struct inode *btree_inode = buf->first_page->mapping->host;
1785 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
1789 void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
1791 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1792 u64 transid = btrfs_header_generation(buf);
1793 struct inode *btree_inode = root->fs_info->btree_inode;
1795 WARN_ON(!btrfs_tree_locked(buf));
1796 if (transid != root->fs_info->generation) {
1797 printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
1798 (unsigned long long)buf->start,
1799 transid, root->fs_info->generation);
1802 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf);
1805 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
1808 * looks as though older kernels can get into trouble with
1809 * this code, they end up stuck in balance_dirty_pages forever
1811 struct extent_io_tree *tree;
1814 unsigned long thresh = 16 * 1024 * 1024;
1815 tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
1817 if (current_is_pdflush())
1820 num_dirty = count_range_bits(tree, &start, (u64)-1,
1821 thresh, EXTENT_DIRTY);
1822 if (num_dirty > thresh) {
1823 balance_dirty_pages_ratelimited_nr(
1824 root->fs_info->btree_inode->i_mapping, 1);
1829 int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
1831 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1833 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
1835 buf->flags |= EXTENT_UPTODATE;
1840 static struct extent_io_ops btree_extent_io_ops = {
1841 .writepage_io_hook = btree_writepage_io_hook,
1842 .readpage_end_io_hook = btree_readpage_end_io_hook,
1843 .submit_bio_hook = btree_submit_bio_hook,
1844 /* note we're sharing with inode.c for the merge bio hook */
1845 .merge_bio_hook = btrfs_merge_bio_hook,