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/scatterlist.h>
22 #include <linux/swap.h>
23 #include <linux/radix-tree.h>
24 #include <linux/writeback.h>
25 #include <linux/buffer_head.h> // for block_sync_page
26 #include <linux/workqueue.h>
30 #include "transaction.h"
31 #include "btrfs_inode.h"
33 #include "print-tree.h"
36 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
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));
48 static struct extent_io_ops btree_extent_io_ops;
49 static struct workqueue_struct *end_io_workqueue;
55 struct btrfs_fs_info *info;
58 struct list_head list;
61 struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
62 size_t page_offset, u64 start, u64 len,
65 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
66 struct extent_map *em;
70 spin_lock(&em_tree->lock);
71 em = lookup_extent_mapping(em_tree, start, len);
72 spin_unlock(&em_tree->lock);
76 em = alloc_extent_map(GFP_NOFS);
78 em = ERR_PTR(-ENOMEM);
82 em->len = i_size_read(inode);
84 em->bdev = inode->i_sb->s_bdev;
86 spin_lock(&em_tree->lock);
87 ret = add_extent_mapping(em_tree, em);
88 spin_unlock(&em_tree->lock);
101 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
103 return btrfs_crc32c(seed, data, len);
106 void btrfs_csum_final(u32 crc, char *result)
108 *(__le32 *)result = ~cpu_to_le32(crc);
111 static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
114 char result[BTRFS_CRC32_SIZE];
116 unsigned long cur_len;
117 unsigned long offset = BTRFS_CSUM_SIZE;
118 char *map_token = NULL;
120 unsigned long map_start;
121 unsigned long map_len;
125 len = buf->len - offset;
127 err = map_private_extent_buffer(buf, offset, 32,
129 &map_start, &map_len, KM_USER0);
131 printk("failed to map extent buffer! %lu\n",
135 cur_len = min(len, map_len - (offset - map_start));
136 crc = btrfs_csum_data(root, kaddr + offset - map_start,
140 unmap_extent_buffer(buf, map_token, KM_USER0);
142 btrfs_csum_final(crc, result);
145 int from_this_trans = 0;
147 if (root->fs_info->running_transaction &&
148 btrfs_header_generation(buf) ==
149 root->fs_info->running_transaction->transid)
152 /* FIXME, this is not good */
153 if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) {
156 memcpy(&found, result, BTRFS_CRC32_SIZE);
158 read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE);
159 printk("btrfs: %s checksum verify failed on %llu "
160 "wanted %X found %X from_this_trans %d "
162 root->fs_info->sb->s_id,
163 buf->start, val, found, from_this_trans,
164 btrfs_header_level(buf));
168 write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE);
173 static int btree_read_extent_buffer_pages(struct btrfs_root *root,
174 struct extent_buffer *eb,
177 struct extent_io_tree *io_tree;
182 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
184 ret = read_extent_buffer_pages(io_tree, eb, start, 1,
185 btree_get_extent, mirror_num);
188 printk("good read %Lu mirror %d total %d\n", eb->start, mirror_num, num_copies);
191 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
193 printk("failed to read %Lu mirror %d total %d\n", eb->start, mirror_num, num_copies);
194 if (num_copies == 1) {
195 printk("reading %Lu failed only one copy\n", eb->start);
199 if (mirror_num > num_copies) {
200 printk("bailing at mirror %d of %d\n", mirror_num, num_copies);
204 printk("read extent buffer page last\n");
208 int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
210 struct extent_io_tree *tree;
211 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
215 struct extent_buffer *eb;
218 tree = &BTRFS_I(page->mapping->host)->io_tree;
220 if (page->private == EXTENT_PAGE_PRIVATE)
224 len = page->private >> 2;
228 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
229 ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE);
231 btrfs_clear_buffer_defrag(eb);
232 found_start = btrfs_header_bytenr(eb);
233 if (found_start != start) {
234 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
235 start, found_start, len);
239 if (eb->first_page != page) {
240 printk("bad first page %lu %lu\n", eb->first_page->index,
245 if (!PageUptodate(page)) {
246 printk("csum not up to date page %lu\n", page->index);
250 found_level = btrfs_header_level(eb);
251 spin_lock(&root->fs_info->hash_lock);
252 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
253 spin_unlock(&root->fs_info->hash_lock);
254 csum_tree_block(root, eb, 0);
256 free_extent_buffer(eb);
261 static int btree_writepage_io_hook(struct page *page, u64 start, u64 end)
263 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
265 csum_dirty_buffer(root, page);
269 int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
270 struct extent_state *state)
272 struct extent_io_tree *tree;
276 struct extent_buffer *eb;
277 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
280 tree = &BTRFS_I(page->mapping->host)->io_tree;
281 if (page->private == EXTENT_PAGE_PRIVATE)
285 len = page->private >> 2;
289 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
291 btrfs_clear_buffer_defrag(eb);
292 found_start = btrfs_header_bytenr(eb);
293 if (found_start != start) {
294 printk("bad start on %Lu found %Lu\n", eb->start, found_start);
298 if (eb->first_page != page) {
299 printk("bad first page %lu %lu\n", eb->first_page->index,
305 found_level = btrfs_header_level(eb);
307 ret = csum_tree_block(root, eb, 1);
311 end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
312 end = eb->start + end - 1;
313 release_extent_buffer_tail_pages(eb);
315 free_extent_buffer(eb);
320 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
321 static void end_workqueue_bio(struct bio *bio, int err)
323 static int end_workqueue_bio(struct bio *bio,
324 unsigned int bytes_done, int err)
327 struct end_io_wq *end_io_wq = bio->bi_private;
328 struct btrfs_fs_info *fs_info;
331 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
336 fs_info = end_io_wq->info;
337 spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
338 end_io_wq->error = err;
339 list_add_tail(&end_io_wq->list, &fs_info->end_io_work_list);
340 spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags);
341 queue_work(end_io_workqueue, &fs_info->end_io_work);
343 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
348 int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
351 struct end_io_wq *end_io_wq;
352 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
356 end_io_wq->private = bio->bi_private;
357 end_io_wq->end_io = bio->bi_end_io;
358 end_io_wq->info = info;
359 end_io_wq->error = 0;
360 end_io_wq->bio = bio;
361 end_io_wq->metadata = metadata;
363 bio->bi_private = end_io_wq;
364 bio->bi_end_io = end_workqueue_bio;
368 static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
371 struct btrfs_root *root = BTRFS_I(inode)->root;
375 offset = bio->bi_sector << 9;
377 if (rw & (1 << BIO_RW)) {
378 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num);
381 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1);
384 if (offset == BTRFS_SUPER_INFO_OFFSET) {
385 bio->bi_bdev = root->fs_info->fs_devices->latest_bdev;
389 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num);
392 static int btree_writepage(struct page *page, struct writeback_control *wbc)
394 struct extent_io_tree *tree;
395 tree = &BTRFS_I(page->mapping->host)->io_tree;
396 return extent_write_full_page(tree, page, btree_get_extent, wbc);
399 static int btree_writepages(struct address_space *mapping,
400 struct writeback_control *wbc)
402 struct extent_io_tree *tree;
403 tree = &BTRFS_I(mapping->host)->io_tree;
404 if (wbc->sync_mode == WB_SYNC_NONE) {
407 unsigned long thresh = 96 * 1024 * 1024;
409 if (wbc->for_kupdate)
412 if (current_is_pdflush()) {
413 thresh = 96 * 1024 * 1024;
415 thresh = 8 * 1024 * 1024;
417 num_dirty = count_range_bits(tree, &start, (u64)-1,
418 thresh, EXTENT_DIRTY);
419 if (num_dirty < thresh) {
423 return extent_writepages(tree, mapping, btree_get_extent, wbc);
426 int btree_readpage(struct file *file, struct page *page)
428 struct extent_io_tree *tree;
429 tree = &BTRFS_I(page->mapping->host)->io_tree;
430 return extent_read_full_page(tree, page, btree_get_extent);
433 static int btree_releasepage(struct page *page, gfp_t gfp_flags)
435 struct extent_io_tree *tree;
436 struct extent_map_tree *map;
439 if (page_count(page) > 3) {
440 /* once for page->private, once for the caller, once
441 * once for the page cache
445 tree = &BTRFS_I(page->mapping->host)->io_tree;
446 map = &BTRFS_I(page->mapping->host)->extent_tree;
447 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
449 invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE);
450 ClearPagePrivate(page);
451 set_page_private(page, 0);
452 page_cache_release(page);
457 static void btree_invalidatepage(struct page *page, unsigned long offset)
459 struct extent_io_tree *tree;
460 tree = &BTRFS_I(page->mapping->host)->io_tree;
461 extent_invalidatepage(tree, page, offset);
462 btree_releasepage(page, GFP_NOFS);
466 static int btree_writepage(struct page *page, struct writeback_control *wbc)
468 struct buffer_head *bh;
469 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
470 struct buffer_head *head;
471 if (!page_has_buffers(page)) {
472 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
473 (1 << BH_Dirty)|(1 << BH_Uptodate));
475 head = page_buffers(page);
478 if (buffer_dirty(bh))
479 csum_tree_block(root, bh, 0);
480 bh = bh->b_this_page;
481 } while (bh != head);
482 return block_write_full_page(page, btree_get_block, wbc);
486 static struct address_space_operations btree_aops = {
487 .readpage = btree_readpage,
488 .writepage = btree_writepage,
489 .writepages = btree_writepages,
490 .releasepage = btree_releasepage,
491 .invalidatepage = btree_invalidatepage,
492 .sync_page = block_sync_page,
495 int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
497 struct extent_buffer *buf = NULL;
498 struct inode *btree_inode = root->fs_info->btree_inode;
501 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
504 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
505 buf, 0, 0, btree_get_extent, 0);
506 free_extent_buffer(buf);
510 static int close_all_devices(struct btrfs_fs_info *fs_info)
512 struct list_head *list;
513 struct list_head *next;
514 struct btrfs_device *device;
516 list = &fs_info->fs_devices->devices;
517 list_for_each(next, list) {
518 device = list_entry(next, struct btrfs_device, dev_list);
519 if (device->bdev && device->bdev != fs_info->sb->s_bdev)
520 close_bdev_excl(device->bdev);
526 int btrfs_verify_block_csum(struct btrfs_root *root,
527 struct extent_buffer *buf)
529 return btrfs_buffer_uptodate(buf);
532 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
533 u64 bytenr, u32 blocksize)
535 struct inode *btree_inode = root->fs_info->btree_inode;
536 struct extent_buffer *eb;
537 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
538 bytenr, blocksize, GFP_NOFS);
542 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
543 u64 bytenr, u32 blocksize)
545 struct inode *btree_inode = root->fs_info->btree_inode;
546 struct extent_buffer *eb;
548 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
549 bytenr, blocksize, NULL, GFP_NOFS);
554 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
557 struct extent_buffer *buf = NULL;
558 struct inode *btree_inode = root->fs_info->btree_inode;
559 struct extent_io_tree *io_tree;
562 io_tree = &BTRFS_I(btree_inode)->io_tree;
564 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
568 ret = btree_read_extent_buffer_pages(root, buf, 0);
571 buf->flags |= EXTENT_UPTODATE;
577 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
578 struct extent_buffer *buf)
580 struct inode *btree_inode = root->fs_info->btree_inode;
581 if (btrfs_header_generation(buf) ==
582 root->fs_info->running_transaction->transid)
583 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
588 int wait_on_tree_block_writeback(struct btrfs_root *root,
589 struct extent_buffer *buf)
591 struct inode *btree_inode = root->fs_info->btree_inode;
592 wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree,
597 static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
598 u32 stripesize, struct btrfs_root *root,
599 struct btrfs_fs_info *fs_info,
604 root->commit_root = NULL;
605 root->sectorsize = sectorsize;
606 root->nodesize = nodesize;
607 root->leafsize = leafsize;
608 root->stripesize = stripesize;
610 root->track_dirty = 0;
612 root->fs_info = fs_info;
613 root->objectid = objectid;
614 root->last_trans = 0;
615 root->highest_inode = 0;
616 root->last_inode_alloc = 0;
620 INIT_LIST_HEAD(&root->dirty_list);
621 memset(&root->root_key, 0, sizeof(root->root_key));
622 memset(&root->root_item, 0, sizeof(root->root_item));
623 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
624 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
625 init_completion(&root->kobj_unregister);
626 root->defrag_running = 0;
627 root->defrag_level = 0;
628 root->root_key.objectid = objectid;
632 static int find_and_setup_root(struct btrfs_root *tree_root,
633 struct btrfs_fs_info *fs_info,
635 struct btrfs_root *root)
640 __setup_root(tree_root->nodesize, tree_root->leafsize,
641 tree_root->sectorsize, tree_root->stripesize,
642 root, fs_info, objectid);
643 ret = btrfs_find_last_root(tree_root, objectid,
644 &root->root_item, &root->root_key);
647 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
648 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
654 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
655 struct btrfs_key *location)
657 struct btrfs_root *root;
658 struct btrfs_root *tree_root = fs_info->tree_root;
659 struct btrfs_path *path;
660 struct extent_buffer *l;
665 root = kzalloc(sizeof(*root), GFP_NOFS);
667 return ERR_PTR(-ENOMEM);
668 if (location->offset == (u64)-1) {
669 ret = find_and_setup_root(tree_root, fs_info,
670 location->objectid, root);
678 __setup_root(tree_root->nodesize, tree_root->leafsize,
679 tree_root->sectorsize, tree_root->stripesize,
680 root, fs_info, location->objectid);
682 path = btrfs_alloc_path();
684 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
691 read_extent_buffer(l, &root->root_item,
692 btrfs_item_ptr_offset(l, path->slots[0]),
693 sizeof(root->root_item));
694 memcpy(&root->root_key, location, sizeof(*location));
697 btrfs_release_path(root, path);
698 btrfs_free_path(path);
703 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
704 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
709 ret = btrfs_find_highest_inode(root, &highest_inode);
711 root->highest_inode = highest_inode;
712 root->last_inode_alloc = highest_inode;
717 struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
720 struct btrfs_root *root;
722 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
723 return fs_info->tree_root;
724 if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
725 return fs_info->extent_root;
727 root = radix_tree_lookup(&fs_info->fs_roots_radix,
728 (unsigned long)root_objectid);
732 struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
733 struct btrfs_key *location)
735 struct btrfs_root *root;
738 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
739 return fs_info->tree_root;
740 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
741 return fs_info->extent_root;
743 root = radix_tree_lookup(&fs_info->fs_roots_radix,
744 (unsigned long)location->objectid);
748 root = btrfs_read_fs_root_no_radix(fs_info, location);
751 ret = radix_tree_insert(&fs_info->fs_roots_radix,
752 (unsigned long)root->root_key.objectid,
755 free_extent_buffer(root->node);
759 ret = btrfs_find_dead_roots(fs_info->tree_root,
760 root->root_key.objectid, root);
766 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
767 struct btrfs_key *location,
768 const char *name, int namelen)
770 struct btrfs_root *root;
773 root = btrfs_read_fs_root_no_name(fs_info, location);
780 ret = btrfs_set_root_name(root, name, namelen);
782 free_extent_buffer(root->node);
787 ret = btrfs_sysfs_add_root(root);
789 free_extent_buffer(root->node);
798 static int add_hasher(struct btrfs_fs_info *info, char *type) {
799 struct btrfs_hasher *hasher;
801 hasher = kmalloc(sizeof(*hasher), GFP_NOFS);
804 hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC);
805 if (!hasher->hash_tfm) {
809 spin_lock(&info->hash_lock);
810 list_add(&hasher->list, &info->hashers);
811 spin_unlock(&info->hash_lock);
816 static int btrfs_congested_fn(void *congested_data, int bdi_bits)
818 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
820 struct list_head *cur;
821 struct btrfs_device *device;
822 struct backing_dev_info *bdi;
824 list_for_each(cur, &info->fs_devices->devices) {
825 device = list_entry(cur, struct btrfs_device, dev_list);
826 bdi = blk_get_backing_dev_info(device->bdev);
827 if (bdi && bdi_congested(bdi, bdi_bits)) {
835 void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
837 struct list_head *cur;
838 struct btrfs_device *device;
839 struct btrfs_fs_info *info;
841 info = (struct btrfs_fs_info *)bdi->unplug_io_data;
842 list_for_each(cur, &info->fs_devices->devices) {
843 device = list_entry(cur, struct btrfs_device, dev_list);
844 bdi = blk_get_backing_dev_info(device->bdev);
845 if (bdi->unplug_io_fn) {
846 bdi->unplug_io_fn(bdi, page);
851 static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
854 bdi->ra_pages = default_backing_dev_info.ra_pages * 4;
856 bdi->capabilities = default_backing_dev_info.capabilities;
857 bdi->unplug_io_fn = btrfs_unplug_io_fn;
858 bdi->unplug_io_data = info;
859 bdi->congested_fn = btrfs_congested_fn;
860 bdi->congested_data = info;
864 static int bio_ready_for_csum(struct bio *bio)
870 struct extent_io_tree *io_tree = NULL;
871 struct btrfs_fs_info *info = NULL;
872 struct bio_vec *bvec;
876 bio_for_each_segment(bvec, bio, i) {
877 page = bvec->bv_page;
878 if (page->private == EXTENT_PAGE_PRIVATE) {
879 length += bvec->bv_len;
882 if (!page->private) {
883 length += bvec->bv_len;
886 length = bvec->bv_len;
887 buf_len = page->private >> 2;
888 start = page_offset(page) + bvec->bv_offset;
889 io_tree = &BTRFS_I(page->mapping->host)->io_tree;
890 info = BTRFS_I(page->mapping->host)->root->fs_info;
892 /* are we fully contained in this bio? */
893 if (buf_len <= length)
896 ret = extent_range_uptodate(io_tree, start + length,
897 start + buf_len - 1);
903 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
904 void btrfs_end_io_csum(void *p)
906 void btrfs_end_io_csum(struct work_struct *work)
909 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
910 struct btrfs_fs_info *fs_info = p;
912 struct btrfs_fs_info *fs_info = container_of(work,
913 struct btrfs_fs_info,
917 struct end_io_wq *end_io_wq;
919 struct list_head *next;
924 spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
925 if (list_empty(&fs_info->end_io_work_list)) {
926 spin_unlock_irqrestore(&fs_info->end_io_work_lock,
930 next = fs_info->end_io_work_list.next;
932 spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags);
934 end_io_wq = list_entry(next, struct end_io_wq, list);
936 bio = end_io_wq->bio;
937 if (end_io_wq->metadata && !bio_ready_for_csum(bio)) {
938 spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
939 was_empty = list_empty(&fs_info->end_io_work_list);
940 list_add_tail(&end_io_wq->list,
941 &fs_info->end_io_work_list);
942 spin_unlock_irqrestore(&fs_info->end_io_work_lock,
948 error = end_io_wq->error;
949 bio->bi_private = end_io_wq->private;
950 bio->bi_end_io = end_io_wq->end_io;
952 bio_endio(bio, error);
957 struct btrfs_root *open_ctree(struct super_block *sb,
958 struct btrfs_fs_devices *fs_devices)
965 struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
967 struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
969 struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
971 struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root),
973 struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root),
977 struct btrfs_super_block *disk_super;
979 if (!extent_root || !tree_root || !fs_info) {
983 end_io_workqueue = create_workqueue("btrfs-end-io");
984 BUG_ON(!end_io_workqueue);
986 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
987 INIT_LIST_HEAD(&fs_info->trans_list);
988 INIT_LIST_HEAD(&fs_info->dead_roots);
989 INIT_LIST_HEAD(&fs_info->hashers);
990 INIT_LIST_HEAD(&fs_info->end_io_work_list);
991 spin_lock_init(&fs_info->hash_lock);
992 spin_lock_init(&fs_info->end_io_work_lock);
993 spin_lock_init(&fs_info->delalloc_lock);
994 spin_lock_init(&fs_info->new_trans_lock);
996 init_completion(&fs_info->kobj_unregister);
997 sb_set_blocksize(sb, BTRFS_SUPER_INFO_SIZE);
998 fs_info->tree_root = tree_root;
999 fs_info->extent_root = extent_root;
1000 fs_info->chunk_root = chunk_root;
1001 fs_info->dev_root = dev_root;
1002 fs_info->fs_devices = fs_devices;
1003 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
1004 INIT_LIST_HEAD(&fs_info->space_info);
1005 btrfs_mapping_init(&fs_info->mapping_tree);
1007 fs_info->max_extent = (u64)-1;
1008 fs_info->max_inline = 8192 * 1024;
1009 setup_bdi(fs_info, &fs_info->bdi);
1010 fs_info->btree_inode = new_inode(sb);
1011 fs_info->btree_inode->i_ino = 1;
1012 fs_info->btree_inode->i_nlink = 1;
1013 fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
1014 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
1015 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
1017 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
1018 fs_info->btree_inode->i_mapping,
1020 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
1023 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
1025 extent_io_tree_init(&fs_info->free_space_cache,
1026 fs_info->btree_inode->i_mapping, GFP_NOFS);
1027 extent_io_tree_init(&fs_info->block_group_cache,
1028 fs_info->btree_inode->i_mapping, GFP_NOFS);
1029 extent_io_tree_init(&fs_info->pinned_extents,
1030 fs_info->btree_inode->i_mapping, GFP_NOFS);
1031 extent_io_tree_init(&fs_info->pending_del,
1032 fs_info->btree_inode->i_mapping, GFP_NOFS);
1033 extent_io_tree_init(&fs_info->extent_ins,
1034 fs_info->btree_inode->i_mapping, GFP_NOFS);
1035 fs_info->do_barriers = 1;
1037 INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum);
1038 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1039 INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info);
1041 INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
1043 BTRFS_I(fs_info->btree_inode)->root = tree_root;
1044 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
1045 sizeof(struct btrfs_key));
1046 insert_inode_hash(fs_info->btree_inode);
1047 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1049 mutex_init(&fs_info->trans_mutex);
1050 mutex_init(&fs_info->fs_mutex);
1053 ret = add_hasher(fs_info, "crc32c");
1055 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
1060 __setup_root(4096, 4096, 4096, 4096, tree_root,
1061 fs_info, BTRFS_ROOT_TREE_OBJECTID);
1063 fs_info->sb_buffer = read_tree_block(tree_root,
1064 BTRFS_SUPER_INFO_OFFSET,
1067 if (!fs_info->sb_buffer)
1070 read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0,
1071 sizeof(fs_info->super_copy));
1073 read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
1074 (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
1077 disk_super = &fs_info->super_copy;
1078 if (!btrfs_super_root(disk_super))
1079 goto fail_sb_buffer;
1081 if (btrfs_super_num_devices(disk_super) != fs_devices->num_devices) {
1082 printk("Btrfs: wanted %llu devices, but found %llu\n",
1083 (unsigned long long)btrfs_super_num_devices(disk_super),
1084 (unsigned long long)fs_devices->num_devices);
1085 goto fail_sb_buffer;
1087 nodesize = btrfs_super_nodesize(disk_super);
1088 leafsize = btrfs_super_leafsize(disk_super);
1089 sectorsize = btrfs_super_sectorsize(disk_super);
1090 stripesize = btrfs_super_stripesize(disk_super);
1091 tree_root->nodesize = nodesize;
1092 tree_root->leafsize = leafsize;
1093 tree_root->sectorsize = sectorsize;
1094 tree_root->stripesize = stripesize;
1095 sb_set_blocksize(sb, sectorsize);
1097 i_size_write(fs_info->btree_inode,
1098 btrfs_super_total_bytes(disk_super));
1100 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
1101 sizeof(disk_super->magic))) {
1102 printk("btrfs: valid FS not found on %s\n", sb->s_id);
1103 goto fail_sb_buffer;
1106 mutex_lock(&fs_info->fs_mutex);
1108 ret = btrfs_read_sys_array(tree_root);
1111 blocksize = btrfs_level_size(tree_root,
1112 btrfs_super_chunk_root_level(disk_super));
1114 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1115 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1117 chunk_root->node = read_tree_block(chunk_root,
1118 btrfs_super_chunk_root(disk_super),
1120 BUG_ON(!chunk_root->node);
1122 ret = btrfs_read_chunk_tree(chunk_root);
1125 blocksize = btrfs_level_size(tree_root,
1126 btrfs_super_root_level(disk_super));
1129 tree_root->node = read_tree_block(tree_root,
1130 btrfs_super_root(disk_super),
1132 if (!tree_root->node)
1133 goto fail_sb_buffer;
1136 ret = find_and_setup_root(tree_root, fs_info,
1137 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
1139 goto fail_tree_root;
1140 extent_root->track_dirty = 1;
1142 ret = find_and_setup_root(tree_root, fs_info,
1143 BTRFS_DEV_TREE_OBJECTID, dev_root);
1144 dev_root->track_dirty = 1;
1147 goto fail_extent_root;
1149 btrfs_read_block_groups(extent_root);
1151 fs_info->generation = btrfs_super_generation(disk_super) + 1;
1152 fs_info->data_alloc_profile = (u64)-1;
1153 fs_info->metadata_alloc_profile = (u64)-1;
1154 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
1156 mutex_unlock(&fs_info->fs_mutex);
1160 free_extent_buffer(extent_root->node);
1162 mutex_unlock(&fs_info->fs_mutex);
1163 free_extent_buffer(tree_root->node);
1165 free_extent_buffer(fs_info->sb_buffer);
1166 extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
1168 iput(fs_info->btree_inode);
1170 close_all_devices(fs_info);
1173 bdi_destroy(&fs_info->bdi);
1175 return ERR_PTR(err);
1178 static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
1180 char b[BDEVNAME_SIZE];
1183 set_buffer_uptodate(bh);
1185 if (!buffer_eopnotsupp(bh) && printk_ratelimit()) {
1186 printk(KERN_WARNING "lost page write due to "
1187 "I/O error on %s\n",
1188 bdevname(bh->b_bdev, b));
1190 set_buffer_write_io_error(bh);
1191 clear_buffer_uptodate(bh);
1197 int write_all_supers(struct btrfs_root *root)
1199 struct list_head *cur;
1200 struct list_head *head = &root->fs_info->fs_devices->devices;
1201 struct btrfs_device *dev;
1202 struct extent_buffer *sb;
1203 struct btrfs_dev_item *dev_item;
1204 struct buffer_head *bh;
1208 do_barriers = !btrfs_test_opt(root, NOBARRIER);
1210 sb = root->fs_info->sb_buffer;
1211 dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block,
1213 list_for_each(cur, head) {
1214 dev = list_entry(cur, struct btrfs_device, dev_list);
1215 btrfs_set_device_type(sb, dev_item, dev->type);
1216 btrfs_set_device_id(sb, dev_item, dev->devid);
1217 btrfs_set_device_total_bytes(sb, dev_item, dev->total_bytes);
1218 btrfs_set_device_bytes_used(sb, dev_item, dev->bytes_used);
1219 btrfs_set_device_io_align(sb, dev_item, dev->io_align);
1220 btrfs_set_device_io_width(sb, dev_item, dev->io_width);
1221 btrfs_set_device_sector_size(sb, dev_item, dev->sector_size);
1222 write_extent_buffer(sb, dev->uuid,
1223 (unsigned long)btrfs_device_uuid(dev_item),
1224 BTRFS_DEV_UUID_SIZE);
1226 btrfs_set_header_flag(sb, BTRFS_HEADER_FLAG_WRITTEN);
1227 csum_tree_block(root, sb, 0);
1229 bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET /
1230 root->fs_info->sb->s_blocksize,
1231 BTRFS_SUPER_INFO_SIZE);
1233 read_extent_buffer(sb, bh->b_data, 0, BTRFS_SUPER_INFO_SIZE);
1234 dev->pending_io = bh;
1237 set_buffer_uptodate(bh);
1239 bh->b_end_io = btrfs_end_buffer_write_sync;
1241 if (do_barriers && dev->barriers) {
1242 ret = submit_bh(WRITE_BARRIER, bh);
1243 if (ret == -EOPNOTSUPP) {
1244 printk("btrfs: disabling barriers on dev %s\n",
1246 set_buffer_uptodate(bh);
1250 ret = submit_bh(WRITE, bh);
1253 ret = submit_bh(WRITE, bh);
1258 list_for_each(cur, head) {
1259 dev = list_entry(cur, struct btrfs_device, dev_list);
1260 BUG_ON(!dev->pending_io);
1261 bh = dev->pending_io;
1263 if (!buffer_uptodate(dev->pending_io)) {
1264 if (do_barriers && dev->barriers) {
1265 printk("btrfs: disabling barriers on dev %s\n",
1267 set_buffer_uptodate(bh);
1271 ret = submit_bh(WRITE, bh);
1274 BUG_ON(!buffer_uptodate(bh));
1280 dev->pending_io = NULL;
1286 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
1291 ret = write_all_supers(root);
1293 if (!btrfs_test_opt(root, NOBARRIER))
1294 blkdev_issue_flush(sb->s_bdev, NULL);
1295 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, super);
1296 ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping,
1297 super->start, super->len);
1298 if (!btrfs_test_opt(root, NOBARRIER))
1299 blkdev_issue_flush(sb->s_bdev, NULL);
1304 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
1306 radix_tree_delete(&fs_info->fs_roots_radix,
1307 (unsigned long)root->root_key.objectid);
1309 btrfs_sysfs_del_root(root);
1313 free_extent_buffer(root->node);
1314 if (root->commit_root)
1315 free_extent_buffer(root->commit_root);
1322 static int del_fs_roots(struct btrfs_fs_info *fs_info)
1325 struct btrfs_root *gang[8];
1329 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
1334 for (i = 0; i < ret; i++)
1335 btrfs_free_fs_root(fs_info, gang[i]);
1340 int close_ctree(struct btrfs_root *root)
1343 struct btrfs_trans_handle *trans;
1344 struct btrfs_fs_info *fs_info = root->fs_info;
1346 fs_info->closing = 1;
1347 btrfs_transaction_flush_work(root);
1348 mutex_lock(&fs_info->fs_mutex);
1349 btrfs_defrag_dirty_roots(root->fs_info);
1350 trans = btrfs_start_transaction(root, 1);
1351 ret = btrfs_commit_transaction(trans, root);
1352 /* run commit again to drop the original snapshot */
1353 trans = btrfs_start_transaction(root, 1);
1354 btrfs_commit_transaction(trans, root);
1355 ret = btrfs_write_and_wait_transaction(NULL, root);
1357 write_ctree_super(NULL, root);
1358 mutex_unlock(&fs_info->fs_mutex);
1360 if (fs_info->delalloc_bytes) {
1361 printk("btrfs: at unmount delalloc count %Lu\n",
1362 fs_info->delalloc_bytes);
1364 if (fs_info->extent_root->node)
1365 free_extent_buffer(fs_info->extent_root->node);
1367 if (fs_info->tree_root->node)
1368 free_extent_buffer(fs_info->tree_root->node);
1370 if (root->fs_info->chunk_root->node);
1371 free_extent_buffer(root->fs_info->chunk_root->node);
1373 if (root->fs_info->dev_root->node);
1374 free_extent_buffer(root->fs_info->dev_root->node);
1376 free_extent_buffer(fs_info->sb_buffer);
1378 btrfs_free_block_groups(root->fs_info);
1379 del_fs_roots(fs_info);
1381 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
1383 extent_io_tree_empty_lru(&fs_info->free_space_cache);
1384 extent_io_tree_empty_lru(&fs_info->block_group_cache);
1385 extent_io_tree_empty_lru(&fs_info->pinned_extents);
1386 extent_io_tree_empty_lru(&fs_info->pending_del);
1387 extent_io_tree_empty_lru(&fs_info->extent_ins);
1388 extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
1390 truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
1391 flush_workqueue(end_io_workqueue);
1392 destroy_workqueue(end_io_workqueue);
1394 iput(fs_info->btree_inode);
1396 while(!list_empty(&fs_info->hashers)) {
1397 struct btrfs_hasher *hasher;
1398 hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher,
1400 list_del(&hasher->hashers);
1401 crypto_free_hash(&fs_info->hash_tfm);
1405 close_all_devices(fs_info);
1406 btrfs_mapping_tree_free(&fs_info->mapping_tree);
1407 bdi_destroy(&fs_info->bdi);
1409 kfree(fs_info->extent_root);
1410 kfree(fs_info->tree_root);
1411 kfree(fs_info->chunk_root);
1412 kfree(fs_info->dev_root);
1416 int btrfs_buffer_uptodate(struct extent_buffer *buf)
1418 struct inode *btree_inode = buf->first_page->mapping->host;
1419 return extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf);
1422 int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
1424 struct inode *btree_inode = buf->first_page->mapping->host;
1425 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
1429 void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
1431 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1432 u64 transid = btrfs_header_generation(buf);
1433 struct inode *btree_inode = root->fs_info->btree_inode;
1435 if (transid != root->fs_info->generation) {
1436 printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
1437 (unsigned long long)buf->start,
1438 transid, root->fs_info->generation);
1441 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf);
1444 void btrfs_throttle(struct btrfs_root *root)
1446 struct backing_dev_info *bdi;
1448 bdi = root->fs_info->sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
1449 if (root->fs_info->throttles && bdi_write_congested(bdi)) {
1450 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18)
1451 congestion_wait(WRITE, HZ/20);
1453 blk_congestion_wait(WRITE, HZ/20);
1458 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
1460 balance_dirty_pages_ratelimited_nr(
1461 root->fs_info->btree_inode->i_mapping, 1);
1464 void btrfs_set_buffer_defrag(struct extent_buffer *buf)
1466 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1467 struct inode *btree_inode = root->fs_info->btree_inode;
1468 set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
1469 buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS);
1472 void btrfs_set_buffer_defrag_done(struct extent_buffer *buf)
1474 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1475 struct inode *btree_inode = root->fs_info->btree_inode;
1476 set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
1477 buf->start + buf->len - 1, EXTENT_DEFRAG_DONE,
1481 int btrfs_buffer_defrag(struct extent_buffer *buf)
1483 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1484 struct inode *btree_inode = root->fs_info->btree_inode;
1485 return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
1486 buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0);
1489 int btrfs_buffer_defrag_done(struct extent_buffer *buf)
1491 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1492 struct inode *btree_inode = root->fs_info->btree_inode;
1493 return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
1494 buf->start, buf->start + buf->len - 1,
1495 EXTENT_DEFRAG_DONE, 0);
1498 int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf)
1500 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1501 struct inode *btree_inode = root->fs_info->btree_inode;
1502 return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
1503 buf->start, buf->start + buf->len - 1,
1504 EXTENT_DEFRAG_DONE, GFP_NOFS);
1507 int btrfs_clear_buffer_defrag(struct extent_buffer *buf)
1509 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1510 struct inode *btree_inode = root->fs_info->btree_inode;
1511 return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
1512 buf->start, buf->start + buf->len - 1,
1513 EXTENT_DEFRAG, GFP_NOFS);
1516 int btrfs_read_buffer(struct extent_buffer *buf)
1518 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1520 ret = btree_read_extent_buffer_pages(root, buf, 0);
1522 buf->flags |= EXTENT_UPTODATE;
1527 static struct extent_io_ops btree_extent_io_ops = {
1528 .writepage_io_hook = btree_writepage_io_hook,
1529 .readpage_end_io_hook = btree_readpage_end_io_hook,
1530 .submit_bio_hook = btree_submit_bio_hook,
1531 /* note we're sharing with inode.c for the merge bio hook */
1532 .merge_bio_hook = btrfs_merge_bio_hook,