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.
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
29 #include "transaction.h"
30 #include "btrfs_inode.h"
32 #include "print-tree.h"
35 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
37 if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) {
38 printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n",
39 (unsigned long long)extent_buffer_blocknr(buf),
40 (unsigned long long)btrfs_header_blocknr(buf));
47 static struct extent_io_ops btree_extent_io_ops;
49 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
50 u64 bytenr, u32 blocksize)
52 struct inode *btree_inode = root->fs_info->btree_inode;
53 struct extent_buffer *eb;
54 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
55 bytenr, blocksize, GFP_NOFS);
59 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
60 u64 bytenr, u32 blocksize)
62 struct inode *btree_inode = root->fs_info->btree_inode;
63 struct extent_buffer *eb;
65 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
66 bytenr, blocksize, NULL, GFP_NOFS);
70 struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
71 size_t page_offset, u64 start, u64 len,
74 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
75 struct extent_map *em;
79 spin_lock(&em_tree->lock);
80 em = lookup_extent_mapping(em_tree, start, len);
81 spin_unlock(&em_tree->lock);
85 em = alloc_extent_map(GFP_NOFS);
87 em = ERR_PTR(-ENOMEM);
91 em->len = i_size_read(inode);
93 em->bdev = inode->i_sb->s_bdev;
95 spin_lock(&em_tree->lock);
96 ret = add_extent_mapping(em_tree, em);
97 spin_unlock(&em_tree->lock);
110 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
112 return crc32c(seed, data, len);
115 void btrfs_csum_final(u32 crc, char *result)
117 *(__le32 *)result = ~cpu_to_le32(crc);
120 static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
123 char result[BTRFS_CRC32_SIZE];
125 unsigned long cur_len;
126 unsigned long offset = BTRFS_CSUM_SIZE;
127 char *map_token = NULL;
129 unsigned long map_start;
130 unsigned long map_len;
134 len = buf->len - offset;
136 err = map_private_extent_buffer(buf, offset, 32,
138 &map_start, &map_len, KM_USER0);
140 printk("failed to map extent buffer! %lu\n",
144 cur_len = min(len, map_len - (offset - map_start));
145 crc = btrfs_csum_data(root, kaddr + offset - map_start,
149 unmap_extent_buffer(buf, map_token, KM_USER0);
151 btrfs_csum_final(crc, result);
154 int from_this_trans = 0;
156 if (root->fs_info->running_transaction &&
157 btrfs_header_generation(buf) ==
158 root->fs_info->running_transaction->transid)
161 /* FIXME, this is not good */
162 if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) {
165 memcpy(&found, result, BTRFS_CRC32_SIZE);
167 read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE);
169 printk("btrfs: %s checksum verify failed on %llu "
170 "wanted %X found %X from_this_trans %d "
172 root->fs_info->sb->s_id,
173 buf->start, val, found, from_this_trans,
174 btrfs_header_level(buf));
178 write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE);
184 int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
186 struct extent_io_tree *tree;
187 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
191 struct extent_buffer *eb;
192 tree = &BTRFS_I(page->mapping->host)->io_tree;
194 if (page->private == EXTENT_PAGE_PRIVATE)
198 len = page->private >> 2;
202 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
203 read_extent_buffer_pages(tree, eb, start + PAGE_CACHE_SIZE, 1,
205 btrfs_clear_buffer_defrag(eb);
206 found_start = btrfs_header_bytenr(eb);
207 if (found_start != start) {
208 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
209 start, found_start, len);
213 if (eb->first_page != page) {
214 printk("bad first page %lu %lu\n", eb->first_page->index,
219 if (!PageUptodate(page)) {
220 printk("csum not up to date page %lu\n", page->index);
224 found_level = btrfs_header_level(eb);
225 spin_lock(&root->fs_info->hash_lock);
226 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
227 spin_unlock(&root->fs_info->hash_lock);
228 csum_tree_block(root, eb, 0);
230 free_extent_buffer(eb);
235 static int btree_writepage_io_hook(struct page *page, u64 start, u64 end)
237 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
239 csum_dirty_buffer(root, page);
243 static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio)
245 struct btrfs_root *root = BTRFS_I(inode)->root;
247 offset = bio->bi_sector << 9;
248 if (offset == BTRFS_SUPER_INFO_OFFSET) {
249 bio->bi_bdev = root->fs_info->sb->s_bdev;
253 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio);
256 static int btree_writepage(struct page *page, struct writeback_control *wbc)
258 struct extent_io_tree *tree;
259 tree = &BTRFS_I(page->mapping->host)->io_tree;
260 return extent_write_full_page(tree, page, btree_get_extent, wbc);
263 static int btree_writepages(struct address_space *mapping,
264 struct writeback_control *wbc)
266 struct extent_io_tree *tree;
267 tree = &BTRFS_I(mapping->host)->io_tree;
268 if (wbc->sync_mode == WB_SYNC_NONE) {
271 unsigned long thresh = 96 * 1024 * 1024;
273 if (wbc->for_kupdate)
276 if (current_is_pdflush()) {
277 thresh = 96 * 1024 * 1024;
279 thresh = 8 * 1024 * 1024;
281 num_dirty = count_range_bits(tree, &start, (u64)-1,
282 thresh, EXTENT_DIRTY);
283 if (num_dirty < thresh) {
287 return extent_writepages(tree, mapping, btree_get_extent, wbc);
290 int btree_readpage(struct file *file, struct page *page)
292 struct extent_io_tree *tree;
293 tree = &BTRFS_I(page->mapping->host)->io_tree;
294 return extent_read_full_page(tree, page, btree_get_extent);
297 static int btree_releasepage(struct page *page, gfp_t gfp_flags)
299 struct extent_io_tree *tree;
300 struct extent_map_tree *map;
303 tree = &BTRFS_I(page->mapping->host)->io_tree;
304 map = &BTRFS_I(page->mapping->host)->extent_tree;
305 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
307 ClearPagePrivate(page);
308 set_page_private(page, 0);
309 page_cache_release(page);
314 static void btree_invalidatepage(struct page *page, unsigned long offset)
316 struct extent_io_tree *tree;
317 tree = &BTRFS_I(page->mapping->host)->io_tree;
318 extent_invalidatepage(tree, page, offset);
319 btree_releasepage(page, GFP_NOFS);
323 static int btree_writepage(struct page *page, struct writeback_control *wbc)
325 struct buffer_head *bh;
326 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
327 struct buffer_head *head;
328 if (!page_has_buffers(page)) {
329 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
330 (1 << BH_Dirty)|(1 << BH_Uptodate));
332 head = page_buffers(page);
335 if (buffer_dirty(bh))
336 csum_tree_block(root, bh, 0);
337 bh = bh->b_this_page;
338 } while (bh != head);
339 return block_write_full_page(page, btree_get_block, wbc);
343 static struct address_space_operations btree_aops = {
344 .readpage = btree_readpage,
345 .writepage = btree_writepage,
346 .writepages = btree_writepages,
347 .releasepage = btree_releasepage,
348 .invalidatepage = btree_invalidatepage,
349 .sync_page = block_sync_page,
352 int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
354 struct extent_buffer *buf = NULL;
355 struct inode *btree_inode = root->fs_info->btree_inode;
358 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
361 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
362 buf, 0, 0, btree_get_extent);
363 free_extent_buffer(buf);
367 static int close_all_devices(struct btrfs_fs_info *fs_info)
369 struct list_head *list;
370 struct list_head *next;
371 struct btrfs_device *device;
373 list = &fs_info->fs_devices->devices;
374 list_for_each(next, list) {
375 device = list_entry(next, struct btrfs_device, dev_list);
376 if (device->bdev && device->bdev != fs_info->sb->s_bdev)
377 close_bdev_excl(device->bdev);
383 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
386 struct extent_buffer *buf = NULL;
387 struct inode *btree_inode = root->fs_info->btree_inode;
388 struct extent_io_tree *io_tree;
392 io_tree = &BTRFS_I(btree_inode)->io_tree;
394 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
397 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, buf, 0, 1,
400 if (buf->flags & EXTENT_CSUM)
403 end = buf->start + PAGE_CACHE_SIZE - 1;
404 if (test_range_bit(io_tree, buf->start, end, EXTENT_CSUM, 1)) {
405 buf->flags |= EXTENT_CSUM;
409 lock_extent(io_tree, buf->start, end, GFP_NOFS);
411 if (test_range_bit(io_tree, buf->start, end, EXTENT_CSUM, 1)) {
412 buf->flags |= EXTENT_CSUM;
416 ret = csum_tree_block(root, buf, 1);
417 set_extent_bits(io_tree, buf->start, end, EXTENT_CSUM, GFP_NOFS);
418 buf->flags |= EXTENT_CSUM;
421 unlock_extent(io_tree, buf->start, end, GFP_NOFS);
425 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
426 struct extent_buffer *buf)
428 struct inode *btree_inode = root->fs_info->btree_inode;
429 if (btrfs_header_generation(buf) ==
430 root->fs_info->running_transaction->transid)
431 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
436 int wait_on_tree_block_writeback(struct btrfs_root *root,
437 struct extent_buffer *buf)
439 struct inode *btree_inode = root->fs_info->btree_inode;
440 wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree,
445 static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
446 u32 stripesize, struct btrfs_root *root,
447 struct btrfs_fs_info *fs_info,
452 root->commit_root = NULL;
453 root->sectorsize = sectorsize;
454 root->nodesize = nodesize;
455 root->leafsize = leafsize;
456 root->stripesize = stripesize;
458 root->track_dirty = 0;
460 root->fs_info = fs_info;
461 root->objectid = objectid;
462 root->last_trans = 0;
463 root->highest_inode = 0;
464 root->last_inode_alloc = 0;
468 INIT_LIST_HEAD(&root->dirty_list);
469 memset(&root->root_key, 0, sizeof(root->root_key));
470 memset(&root->root_item, 0, sizeof(root->root_item));
471 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
472 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
473 init_completion(&root->kobj_unregister);
474 root->defrag_running = 0;
475 root->defrag_level = 0;
476 root->root_key.objectid = objectid;
480 static int find_and_setup_root(struct btrfs_root *tree_root,
481 struct btrfs_fs_info *fs_info,
483 struct btrfs_root *root)
488 __setup_root(tree_root->nodesize, tree_root->leafsize,
489 tree_root->sectorsize, tree_root->stripesize,
490 root, fs_info, objectid);
491 ret = btrfs_find_last_root(tree_root, objectid,
492 &root->root_item, &root->root_key);
495 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
496 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
502 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
503 struct btrfs_key *location)
505 struct btrfs_root *root;
506 struct btrfs_root *tree_root = fs_info->tree_root;
507 struct btrfs_path *path;
508 struct extent_buffer *l;
513 root = kzalloc(sizeof(*root), GFP_NOFS);
515 return ERR_PTR(-ENOMEM);
516 if (location->offset == (u64)-1) {
517 ret = find_and_setup_root(tree_root, fs_info,
518 location->objectid, root);
526 __setup_root(tree_root->nodesize, tree_root->leafsize,
527 tree_root->sectorsize, tree_root->stripesize,
528 root, fs_info, location->objectid);
530 path = btrfs_alloc_path();
532 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
539 read_extent_buffer(l, &root->root_item,
540 btrfs_item_ptr_offset(l, path->slots[0]),
541 sizeof(root->root_item));
542 memcpy(&root->root_key, location, sizeof(*location));
545 btrfs_release_path(root, path);
546 btrfs_free_path(path);
551 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
552 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
557 ret = btrfs_find_highest_inode(root, &highest_inode);
559 root->highest_inode = highest_inode;
560 root->last_inode_alloc = highest_inode;
565 struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
568 struct btrfs_root *root;
570 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
571 return fs_info->tree_root;
572 if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
573 return fs_info->extent_root;
575 root = radix_tree_lookup(&fs_info->fs_roots_radix,
576 (unsigned long)root_objectid);
580 struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
581 struct btrfs_key *location)
583 struct btrfs_root *root;
586 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
587 return fs_info->tree_root;
588 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
589 return fs_info->extent_root;
591 root = radix_tree_lookup(&fs_info->fs_roots_radix,
592 (unsigned long)location->objectid);
596 root = btrfs_read_fs_root_no_radix(fs_info, location);
599 ret = radix_tree_insert(&fs_info->fs_roots_radix,
600 (unsigned long)root->root_key.objectid,
603 free_extent_buffer(root->node);
607 ret = btrfs_find_dead_roots(fs_info->tree_root,
608 root->root_key.objectid, root);
614 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
615 struct btrfs_key *location,
616 const char *name, int namelen)
618 struct btrfs_root *root;
621 root = btrfs_read_fs_root_no_name(fs_info, location);
628 ret = btrfs_set_root_name(root, name, namelen);
630 free_extent_buffer(root->node);
635 ret = btrfs_sysfs_add_root(root);
637 free_extent_buffer(root->node);
646 static int add_hasher(struct btrfs_fs_info *info, char *type) {
647 struct btrfs_hasher *hasher;
649 hasher = kmalloc(sizeof(*hasher), GFP_NOFS);
652 hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC);
653 if (!hasher->hash_tfm) {
657 spin_lock(&info->hash_lock);
658 list_add(&hasher->list, &info->hashers);
659 spin_unlock(&info->hash_lock);
664 static int btrfs_congested_fn(void *congested_data, int bdi_bits)
666 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
668 struct list_head *cur;
669 struct btrfs_device *device;
670 struct backing_dev_info *bdi;
672 list_for_each(cur, &info->fs_devices->devices) {
673 device = list_entry(cur, struct btrfs_device, dev_list);
674 bdi = blk_get_backing_dev_info(device->bdev);
675 if (bdi && bdi_congested(bdi, bdi_bits)) {
683 void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
685 struct list_head *cur;
686 struct btrfs_device *device;
687 struct btrfs_fs_info *info;
689 info = (struct btrfs_fs_info *)bdi->unplug_io_data;
690 list_for_each(cur, &info->fs_devices->devices) {
691 device = list_entry(cur, struct btrfs_device, dev_list);
692 bdi = blk_get_backing_dev_info(device->bdev);
693 if (bdi->unplug_io_fn) {
694 bdi->unplug_io_fn(bdi, page);
699 static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
702 bdi->ra_pages = default_backing_dev_info.ra_pages * 4;
704 bdi->capabilities = default_backing_dev_info.capabilities;
705 bdi->unplug_io_fn = btrfs_unplug_io_fn;
706 bdi->unplug_io_data = info;
707 bdi->congested_fn = btrfs_congested_fn;
708 bdi->congested_data = info;
712 struct btrfs_root *open_ctree(struct super_block *sb,
713 struct btrfs_fs_devices *fs_devices)
720 struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
722 struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
724 struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
726 struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root),
728 struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root),
732 struct btrfs_super_block *disk_super;
733 if (!extent_root || !tree_root || !fs_info) {
737 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
738 INIT_LIST_HEAD(&fs_info->trans_list);
739 INIT_LIST_HEAD(&fs_info->dead_roots);
740 INIT_LIST_HEAD(&fs_info->hashers);
741 spin_lock_init(&fs_info->hash_lock);
742 spin_lock_init(&fs_info->delalloc_lock);
743 spin_lock_init(&fs_info->new_trans_lock);
745 memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj));
746 init_completion(&fs_info->kobj_unregister);
747 sb_set_blocksize(sb, 4096);
748 fs_info->running_transaction = NULL;
749 fs_info->last_trans_committed = 0;
750 fs_info->tree_root = tree_root;
751 fs_info->extent_root = extent_root;
752 fs_info->chunk_root = chunk_root;
753 fs_info->dev_root = dev_root;
754 fs_info->fs_devices = fs_devices;
755 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
756 INIT_LIST_HEAD(&fs_info->space_info);
757 btrfs_mapping_init(&fs_info->mapping_tree);
759 fs_info->throttles = 0;
760 fs_info->mount_opt = 0;
761 fs_info->max_extent = (u64)-1;
762 fs_info->max_inline = 8192 * 1024;
763 fs_info->delalloc_bytes = 0;
764 setup_bdi(fs_info, &fs_info->bdi);
765 fs_info->btree_inode = new_inode(sb);
766 fs_info->btree_inode->i_ino = 1;
767 fs_info->btree_inode->i_nlink = 1;
768 fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
769 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
770 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
772 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
773 fs_info->btree_inode->i_mapping,
775 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
778 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
780 extent_io_tree_init(&fs_info->free_space_cache,
781 fs_info->btree_inode->i_mapping, GFP_NOFS);
782 extent_io_tree_init(&fs_info->block_group_cache,
783 fs_info->btree_inode->i_mapping, GFP_NOFS);
784 extent_io_tree_init(&fs_info->pinned_extents,
785 fs_info->btree_inode->i_mapping, GFP_NOFS);
786 extent_io_tree_init(&fs_info->pending_del,
787 fs_info->btree_inode->i_mapping, GFP_NOFS);
788 extent_io_tree_init(&fs_info->extent_ins,
789 fs_info->btree_inode->i_mapping, GFP_NOFS);
790 fs_info->do_barriers = 1;
791 fs_info->closing = 0;
792 fs_info->total_pinned = 0;
793 fs_info->last_alloc = 0;
794 fs_info->last_data_alloc = 0;
795 fs_info->extra_alloc_bits = 0;
796 fs_info->extra_data_alloc_bits = 0;
798 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
799 INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info);
801 INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
803 BTRFS_I(fs_info->btree_inode)->root = tree_root;
804 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
805 sizeof(struct btrfs_key));
806 insert_inode_hash(fs_info->btree_inode);
807 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
809 mutex_init(&fs_info->trans_mutex);
810 mutex_init(&fs_info->fs_mutex);
813 ret = add_hasher(fs_info, "crc32c");
815 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
820 __setup_root(4096, 4096, 4096, 4096, tree_root,
821 fs_info, BTRFS_ROOT_TREE_OBJECTID);
823 fs_info->sb_buffer = read_tree_block(tree_root,
824 BTRFS_SUPER_INFO_OFFSET,
827 if (!fs_info->sb_buffer)
830 read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0,
831 sizeof(fs_info->super_copy));
833 read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
834 (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
837 disk_super = &fs_info->super_copy;
838 if (!btrfs_super_root(disk_super))
841 if (btrfs_super_num_devices(disk_super) != fs_devices->num_devices) {
842 printk("Btrfs: wanted %llu devices, but found %llu\n",
843 (unsigned long long)btrfs_super_num_devices(disk_super),
844 (unsigned long long)fs_devices->num_devices);
847 nodesize = btrfs_super_nodesize(disk_super);
848 leafsize = btrfs_super_leafsize(disk_super);
849 sectorsize = btrfs_super_sectorsize(disk_super);
850 stripesize = btrfs_super_stripesize(disk_super);
851 tree_root->nodesize = nodesize;
852 tree_root->leafsize = leafsize;
853 tree_root->sectorsize = sectorsize;
854 tree_root->stripesize = stripesize;
855 sb_set_blocksize(sb, sectorsize);
857 i_size_write(fs_info->btree_inode,
858 btrfs_super_total_bytes(disk_super));
860 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
861 sizeof(disk_super->magic))) {
862 printk("btrfs: valid FS not found on %s\n", sb->s_id);
866 mutex_lock(&fs_info->fs_mutex);
868 ret = btrfs_read_sys_array(tree_root);
871 blocksize = btrfs_level_size(tree_root,
872 btrfs_super_chunk_root_level(disk_super));
874 __setup_root(nodesize, leafsize, sectorsize, stripesize,
875 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
877 chunk_root->node = read_tree_block(chunk_root,
878 btrfs_super_chunk_root(disk_super),
880 BUG_ON(!chunk_root->node);
882 ret = btrfs_read_chunk_tree(chunk_root);
885 blocksize = btrfs_level_size(tree_root,
886 btrfs_super_root_level(disk_super));
889 tree_root->node = read_tree_block(tree_root,
890 btrfs_super_root(disk_super),
892 if (!tree_root->node)
896 ret = find_and_setup_root(tree_root, fs_info,
897 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
900 extent_root->track_dirty = 1;
902 ret = find_and_setup_root(tree_root, fs_info,
903 BTRFS_DEV_TREE_OBJECTID, dev_root);
904 dev_root->track_dirty = 1;
907 goto fail_extent_root;
909 btrfs_read_block_groups(extent_root);
911 fs_info->generation = btrfs_super_generation(disk_super) + 1;
912 mutex_unlock(&fs_info->fs_mutex);
916 free_extent_buffer(extent_root->node);
918 mutex_unlock(&fs_info->fs_mutex);
919 free_extent_buffer(tree_root->node);
921 free_extent_buffer(fs_info->sb_buffer);
922 extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
924 iput(fs_info->btree_inode);
926 close_all_devices(fs_info);
929 bdi_destroy(&fs_info->bdi);
934 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
938 struct extent_buffer *super = root->fs_info->sb_buffer;
939 struct inode *btree_inode = root->fs_info->btree_inode;
940 struct super_block *sb = root->fs_info->sb;
942 if (!btrfs_test_opt(root, NOBARRIER))
943 blkdev_issue_flush(sb->s_bdev, NULL);
944 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, super);
945 ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping,
946 super->start, super->len);
947 if (!btrfs_test_opt(root, NOBARRIER))
948 blkdev_issue_flush(sb->s_bdev, NULL);
952 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
954 radix_tree_delete(&fs_info->fs_roots_radix,
955 (unsigned long)root->root_key.objectid);
957 btrfs_sysfs_del_root(root);
961 free_extent_buffer(root->node);
962 if (root->commit_root)
963 free_extent_buffer(root->commit_root);
970 static int del_fs_roots(struct btrfs_fs_info *fs_info)
973 struct btrfs_root *gang[8];
977 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
982 for (i = 0; i < ret; i++)
983 btrfs_free_fs_root(fs_info, gang[i]);
988 int close_ctree(struct btrfs_root *root)
991 struct btrfs_trans_handle *trans;
992 struct btrfs_fs_info *fs_info = root->fs_info;
994 fs_info->closing = 1;
995 btrfs_transaction_flush_work(root);
996 mutex_lock(&fs_info->fs_mutex);
997 btrfs_defrag_dirty_roots(root->fs_info);
998 trans = btrfs_start_transaction(root, 1);
999 ret = btrfs_commit_transaction(trans, root);
1000 /* run commit again to drop the original snapshot */
1001 trans = btrfs_start_transaction(root, 1);
1002 btrfs_commit_transaction(trans, root);
1003 ret = btrfs_write_and_wait_transaction(NULL, root);
1005 write_ctree_super(NULL, root);
1006 mutex_unlock(&fs_info->fs_mutex);
1008 if (fs_info->delalloc_bytes) {
1009 printk("btrfs: at unmount delalloc count %Lu\n",
1010 fs_info->delalloc_bytes);
1012 if (fs_info->extent_root->node)
1013 free_extent_buffer(fs_info->extent_root->node);
1015 if (fs_info->tree_root->node)
1016 free_extent_buffer(fs_info->tree_root->node);
1018 if (root->fs_info->chunk_root->node);
1019 free_extent_buffer(root->fs_info->chunk_root->node);
1021 if (root->fs_info->dev_root->node);
1022 free_extent_buffer(root->fs_info->dev_root->node);
1024 free_extent_buffer(fs_info->sb_buffer);
1026 btrfs_free_block_groups(root->fs_info);
1027 del_fs_roots(fs_info);
1029 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
1031 extent_io_tree_empty_lru(&fs_info->free_space_cache);
1032 extent_io_tree_empty_lru(&fs_info->block_group_cache);
1033 extent_io_tree_empty_lru(&fs_info->pinned_extents);
1034 extent_io_tree_empty_lru(&fs_info->pending_del);
1035 extent_io_tree_empty_lru(&fs_info->extent_ins);
1036 extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
1038 truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
1040 iput(fs_info->btree_inode);
1042 while(!list_empty(&fs_info->hashers)) {
1043 struct btrfs_hasher *hasher;
1044 hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher,
1046 list_del(&hasher->hashers);
1047 crypto_free_hash(&fs_info->hash_tfm);
1051 close_all_devices(fs_info);
1052 btrfs_mapping_tree_free(&fs_info->mapping_tree);
1053 bdi_destroy(&fs_info->bdi);
1055 kfree(fs_info->extent_root);
1056 kfree(fs_info->tree_root);
1057 kfree(fs_info->chunk_root);
1058 kfree(fs_info->dev_root);
1062 int btrfs_buffer_uptodate(struct extent_buffer *buf)
1064 struct inode *btree_inode = buf->first_page->mapping->host;
1065 return extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf);
1068 int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
1070 struct inode *btree_inode = buf->first_page->mapping->host;
1071 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
1075 void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
1077 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1078 u64 transid = btrfs_header_generation(buf);
1079 struct inode *btree_inode = root->fs_info->btree_inode;
1081 if (transid != root->fs_info->generation) {
1082 printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
1083 (unsigned long long)buf->start,
1084 transid, root->fs_info->generation);
1087 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf);
1090 void btrfs_throttle(struct btrfs_root *root)
1092 struct backing_dev_info *bdi;
1094 bdi = root->fs_info->sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
1095 if (root->fs_info->throttles && bdi_write_congested(bdi)) {
1096 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18)
1097 congestion_wait(WRITE, HZ/20);
1099 blk_congestion_wait(WRITE, HZ/20);
1104 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
1106 balance_dirty_pages_ratelimited_nr(
1107 root->fs_info->btree_inode->i_mapping, 1);
1110 void btrfs_set_buffer_defrag(struct extent_buffer *buf)
1112 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1113 struct inode *btree_inode = root->fs_info->btree_inode;
1114 set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
1115 buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS);
1118 void btrfs_set_buffer_defrag_done(struct extent_buffer *buf)
1120 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1121 struct inode *btree_inode = root->fs_info->btree_inode;
1122 set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
1123 buf->start + buf->len - 1, EXTENT_DEFRAG_DONE,
1127 int btrfs_buffer_defrag(struct extent_buffer *buf)
1129 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1130 struct inode *btree_inode = root->fs_info->btree_inode;
1131 return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
1132 buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0);
1135 int btrfs_buffer_defrag_done(struct extent_buffer *buf)
1137 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1138 struct inode *btree_inode = root->fs_info->btree_inode;
1139 return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
1140 buf->start, buf->start + buf->len - 1,
1141 EXTENT_DEFRAG_DONE, 0);
1144 int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf)
1146 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1147 struct inode *btree_inode = root->fs_info->btree_inode;
1148 return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
1149 buf->start, buf->start + buf->len - 1,
1150 EXTENT_DEFRAG_DONE, GFP_NOFS);
1153 int btrfs_clear_buffer_defrag(struct extent_buffer *buf)
1155 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1156 struct inode *btree_inode = root->fs_info->btree_inode;
1157 return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
1158 buf->start, buf->start + buf->len - 1,
1159 EXTENT_DEFRAG, GFP_NOFS);
1162 int btrfs_read_buffer(struct extent_buffer *buf)
1164 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1165 struct inode *btree_inode = root->fs_info->btree_inode;
1166 return read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
1167 buf, 0, 1, btree_get_extent);
1170 static struct extent_io_ops btree_extent_io_ops = {
1171 .writepage_io_hook = btree_writepage_io_hook,
1172 .submit_bio_hook = btree_submit_bio_hook,
1173 /* note we're sharing with inode.c for the merge bio hook */
1174 .merge_bio_hook = btrfs_merge_bio_hook,