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
27 #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 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 tree = &BTRFS_I(page->mapping->host)->io_tree;
440 map = &BTRFS_I(page->mapping->host)->extent_tree;
441 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
443 invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE);
444 ClearPagePrivate(page);
445 set_page_private(page, 0);
446 page_cache_release(page);
451 static void btree_invalidatepage(struct page *page, unsigned long offset)
453 struct extent_io_tree *tree;
454 tree = &BTRFS_I(page->mapping->host)->io_tree;
455 extent_invalidatepage(tree, page, offset);
456 btree_releasepage(page, GFP_NOFS);
460 static int btree_writepage(struct page *page, struct writeback_control *wbc)
462 struct buffer_head *bh;
463 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
464 struct buffer_head *head;
465 if (!page_has_buffers(page)) {
466 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
467 (1 << BH_Dirty)|(1 << BH_Uptodate));
469 head = page_buffers(page);
472 if (buffer_dirty(bh))
473 csum_tree_block(root, bh, 0);
474 bh = bh->b_this_page;
475 } while (bh != head);
476 return block_write_full_page(page, btree_get_block, wbc);
480 static struct address_space_operations btree_aops = {
481 .readpage = btree_readpage,
482 .writepage = btree_writepage,
483 .writepages = btree_writepages,
484 .releasepage = btree_releasepage,
485 .invalidatepage = btree_invalidatepage,
486 .sync_page = block_sync_page,
489 int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
491 struct extent_buffer *buf = NULL;
492 struct inode *btree_inode = root->fs_info->btree_inode;
495 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
498 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
499 buf, 0, 0, btree_get_extent, 0);
500 free_extent_buffer(buf);
504 static int close_all_devices(struct btrfs_fs_info *fs_info)
506 struct list_head *list;
507 struct list_head *next;
508 struct btrfs_device *device;
510 list = &fs_info->fs_devices->devices;
511 list_for_each(next, list) {
512 device = list_entry(next, struct btrfs_device, dev_list);
513 if (device->bdev && device->bdev != fs_info->sb->s_bdev)
514 close_bdev_excl(device->bdev);
520 int btrfs_verify_block_csum(struct btrfs_root *root,
521 struct extent_buffer *buf)
523 return btrfs_buffer_uptodate(buf);
526 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
527 u64 bytenr, u32 blocksize)
529 struct inode *btree_inode = root->fs_info->btree_inode;
530 struct extent_buffer *eb;
531 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
532 bytenr, blocksize, GFP_NOFS);
536 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
537 u64 bytenr, u32 blocksize)
539 struct inode *btree_inode = root->fs_info->btree_inode;
540 struct extent_buffer *eb;
542 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
543 bytenr, blocksize, NULL, GFP_NOFS);
548 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
551 struct extent_buffer *buf = NULL;
552 struct inode *btree_inode = root->fs_info->btree_inode;
553 struct extent_io_tree *io_tree;
556 io_tree = &BTRFS_I(btree_inode)->io_tree;
558 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
562 ret = btree_read_extent_buffer_pages(root, buf, 0);
565 buf->flags |= EXTENT_UPTODATE;
571 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
572 struct extent_buffer *buf)
574 struct inode *btree_inode = root->fs_info->btree_inode;
575 if (btrfs_header_generation(buf) ==
576 root->fs_info->running_transaction->transid)
577 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
582 int wait_on_tree_block_writeback(struct btrfs_root *root,
583 struct extent_buffer *buf)
585 struct inode *btree_inode = root->fs_info->btree_inode;
586 wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree,
591 static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
592 u32 stripesize, struct btrfs_root *root,
593 struct btrfs_fs_info *fs_info,
598 root->commit_root = NULL;
599 root->sectorsize = sectorsize;
600 root->nodesize = nodesize;
601 root->leafsize = leafsize;
602 root->stripesize = stripesize;
604 root->track_dirty = 0;
606 root->fs_info = fs_info;
607 root->objectid = objectid;
608 root->last_trans = 0;
609 root->highest_inode = 0;
610 root->last_inode_alloc = 0;
614 INIT_LIST_HEAD(&root->dirty_list);
615 memset(&root->root_key, 0, sizeof(root->root_key));
616 memset(&root->root_item, 0, sizeof(root->root_item));
617 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
618 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
619 init_completion(&root->kobj_unregister);
620 root->defrag_running = 0;
621 root->defrag_level = 0;
622 root->root_key.objectid = objectid;
626 static int find_and_setup_root(struct btrfs_root *tree_root,
627 struct btrfs_fs_info *fs_info,
629 struct btrfs_root *root)
634 __setup_root(tree_root->nodesize, tree_root->leafsize,
635 tree_root->sectorsize, tree_root->stripesize,
636 root, fs_info, objectid);
637 ret = btrfs_find_last_root(tree_root, objectid,
638 &root->root_item, &root->root_key);
641 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
642 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
648 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
649 struct btrfs_key *location)
651 struct btrfs_root *root;
652 struct btrfs_root *tree_root = fs_info->tree_root;
653 struct btrfs_path *path;
654 struct extent_buffer *l;
659 root = kzalloc(sizeof(*root), GFP_NOFS);
661 return ERR_PTR(-ENOMEM);
662 if (location->offset == (u64)-1) {
663 ret = find_and_setup_root(tree_root, fs_info,
664 location->objectid, root);
672 __setup_root(tree_root->nodesize, tree_root->leafsize,
673 tree_root->sectorsize, tree_root->stripesize,
674 root, fs_info, location->objectid);
676 path = btrfs_alloc_path();
678 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
685 read_extent_buffer(l, &root->root_item,
686 btrfs_item_ptr_offset(l, path->slots[0]),
687 sizeof(root->root_item));
688 memcpy(&root->root_key, location, sizeof(*location));
691 btrfs_release_path(root, path);
692 btrfs_free_path(path);
697 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
698 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
703 ret = btrfs_find_highest_inode(root, &highest_inode);
705 root->highest_inode = highest_inode;
706 root->last_inode_alloc = highest_inode;
711 struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
714 struct btrfs_root *root;
716 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
717 return fs_info->tree_root;
718 if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
719 return fs_info->extent_root;
721 root = radix_tree_lookup(&fs_info->fs_roots_radix,
722 (unsigned long)root_objectid);
726 struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
727 struct btrfs_key *location)
729 struct btrfs_root *root;
732 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
733 return fs_info->tree_root;
734 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
735 return fs_info->extent_root;
737 root = radix_tree_lookup(&fs_info->fs_roots_radix,
738 (unsigned long)location->objectid);
742 root = btrfs_read_fs_root_no_radix(fs_info, location);
745 ret = radix_tree_insert(&fs_info->fs_roots_radix,
746 (unsigned long)root->root_key.objectid,
749 free_extent_buffer(root->node);
753 ret = btrfs_find_dead_roots(fs_info->tree_root,
754 root->root_key.objectid, root);
760 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
761 struct btrfs_key *location,
762 const char *name, int namelen)
764 struct btrfs_root *root;
767 root = btrfs_read_fs_root_no_name(fs_info, location);
774 ret = btrfs_set_root_name(root, name, namelen);
776 free_extent_buffer(root->node);
781 ret = btrfs_sysfs_add_root(root);
783 free_extent_buffer(root->node);
792 static int add_hasher(struct btrfs_fs_info *info, char *type) {
793 struct btrfs_hasher *hasher;
795 hasher = kmalloc(sizeof(*hasher), GFP_NOFS);
798 hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC);
799 if (!hasher->hash_tfm) {
803 spin_lock(&info->hash_lock);
804 list_add(&hasher->list, &info->hashers);
805 spin_unlock(&info->hash_lock);
810 static int btrfs_congested_fn(void *congested_data, int bdi_bits)
812 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
814 struct list_head *cur;
815 struct btrfs_device *device;
816 struct backing_dev_info *bdi;
818 list_for_each(cur, &info->fs_devices->devices) {
819 device = list_entry(cur, struct btrfs_device, dev_list);
820 bdi = blk_get_backing_dev_info(device->bdev);
821 if (bdi && bdi_congested(bdi, bdi_bits)) {
829 void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
831 struct list_head *cur;
832 struct btrfs_device *device;
833 struct btrfs_fs_info *info;
835 info = (struct btrfs_fs_info *)bdi->unplug_io_data;
836 list_for_each(cur, &info->fs_devices->devices) {
837 device = list_entry(cur, struct btrfs_device, dev_list);
838 bdi = blk_get_backing_dev_info(device->bdev);
839 if (bdi->unplug_io_fn) {
840 bdi->unplug_io_fn(bdi, page);
845 static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
848 bdi->ra_pages = default_backing_dev_info.ra_pages * 4;
850 bdi->capabilities = default_backing_dev_info.capabilities;
851 bdi->unplug_io_fn = btrfs_unplug_io_fn;
852 bdi->unplug_io_data = info;
853 bdi->congested_fn = btrfs_congested_fn;
854 bdi->congested_data = info;
858 static int bio_ready_for_csum(struct bio *bio)
864 struct extent_io_tree *io_tree = NULL;
865 struct btrfs_fs_info *info = NULL;
866 struct bio_vec *bvec;
870 bio_for_each_segment(bvec, bio, i) {
871 page = bvec->bv_page;
872 if (page->private == EXTENT_PAGE_PRIVATE) {
873 length += bvec->bv_len;
876 if (!page->private) {
877 length += bvec->bv_len;
880 length = bvec->bv_len;
881 buf_len = page->private >> 2;
882 start = page_offset(page) + bvec->bv_offset;
883 io_tree = &BTRFS_I(page->mapping->host)->io_tree;
884 info = BTRFS_I(page->mapping->host)->root->fs_info;
886 /* are we fully contained in this bio? */
887 if (buf_len <= length)
890 ret = extent_range_uptodate(io_tree, start + length,
891 start + buf_len - 1);
897 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
898 void btrfs_end_io_csum(void *p)
900 void btrfs_end_io_csum(struct work_struct *work)
903 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
904 struct btrfs_fs_info *fs_info = p;
906 struct btrfs_fs_info *fs_info = container_of(work,
907 struct btrfs_fs_info,
911 struct end_io_wq *end_io_wq;
913 struct list_head *next;
918 spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
919 if (list_empty(&fs_info->end_io_work_list)) {
920 spin_unlock_irqrestore(&fs_info->end_io_work_lock,
924 next = fs_info->end_io_work_list.next;
926 spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags);
928 end_io_wq = list_entry(next, struct end_io_wq, list);
930 bio = end_io_wq->bio;
931 if (end_io_wq->metadata && !bio_ready_for_csum(bio)) {
932 spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
933 was_empty = list_empty(&fs_info->end_io_work_list);
934 list_add_tail(&end_io_wq->list,
935 &fs_info->end_io_work_list);
936 spin_unlock_irqrestore(&fs_info->end_io_work_lock,
942 error = end_io_wq->error;
943 bio->bi_private = end_io_wq->private;
944 bio->bi_end_io = end_io_wq->end_io;
946 bio_endio(bio, error);
951 struct btrfs_root *open_ctree(struct super_block *sb,
952 struct btrfs_fs_devices *fs_devices)
959 struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
961 struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
963 struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
965 struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root),
967 struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root),
971 struct btrfs_super_block *disk_super;
973 if (!extent_root || !tree_root || !fs_info) {
977 end_io_workqueue = create_workqueue("btrfs-end-io");
978 BUG_ON(!end_io_workqueue);
980 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
981 INIT_LIST_HEAD(&fs_info->trans_list);
982 INIT_LIST_HEAD(&fs_info->dead_roots);
983 INIT_LIST_HEAD(&fs_info->hashers);
984 INIT_LIST_HEAD(&fs_info->end_io_work_list);
985 spin_lock_init(&fs_info->hash_lock);
986 spin_lock_init(&fs_info->end_io_work_lock);
987 spin_lock_init(&fs_info->delalloc_lock);
988 spin_lock_init(&fs_info->new_trans_lock);
990 init_completion(&fs_info->kobj_unregister);
991 sb_set_blocksize(sb, BTRFS_SUPER_INFO_SIZE);
992 fs_info->tree_root = tree_root;
993 fs_info->extent_root = extent_root;
994 fs_info->chunk_root = chunk_root;
995 fs_info->dev_root = dev_root;
996 fs_info->fs_devices = fs_devices;
997 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
998 INIT_LIST_HEAD(&fs_info->space_info);
999 btrfs_mapping_init(&fs_info->mapping_tree);
1001 fs_info->max_extent = (u64)-1;
1002 fs_info->max_inline = 8192 * 1024;
1003 setup_bdi(fs_info, &fs_info->bdi);
1004 fs_info->btree_inode = new_inode(sb);
1005 fs_info->btree_inode->i_ino = 1;
1006 fs_info->btree_inode->i_nlink = 1;
1007 fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
1008 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
1009 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
1011 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
1012 fs_info->btree_inode->i_mapping,
1014 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
1017 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
1019 extent_io_tree_init(&fs_info->free_space_cache,
1020 fs_info->btree_inode->i_mapping, GFP_NOFS);
1021 extent_io_tree_init(&fs_info->block_group_cache,
1022 fs_info->btree_inode->i_mapping, GFP_NOFS);
1023 extent_io_tree_init(&fs_info->pinned_extents,
1024 fs_info->btree_inode->i_mapping, GFP_NOFS);
1025 extent_io_tree_init(&fs_info->pending_del,
1026 fs_info->btree_inode->i_mapping, GFP_NOFS);
1027 extent_io_tree_init(&fs_info->extent_ins,
1028 fs_info->btree_inode->i_mapping, GFP_NOFS);
1029 fs_info->do_barriers = 1;
1031 INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum);
1032 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1033 INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info);
1035 INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
1037 BTRFS_I(fs_info->btree_inode)->root = tree_root;
1038 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
1039 sizeof(struct btrfs_key));
1040 insert_inode_hash(fs_info->btree_inode);
1041 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1043 mutex_init(&fs_info->trans_mutex);
1044 mutex_init(&fs_info->fs_mutex);
1047 ret = add_hasher(fs_info, "crc32c");
1049 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
1054 __setup_root(4096, 4096, 4096, 4096, tree_root,
1055 fs_info, BTRFS_ROOT_TREE_OBJECTID);
1057 fs_info->sb_buffer = read_tree_block(tree_root,
1058 BTRFS_SUPER_INFO_OFFSET,
1061 if (!fs_info->sb_buffer)
1064 read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0,
1065 sizeof(fs_info->super_copy));
1067 read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
1068 (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
1071 disk_super = &fs_info->super_copy;
1072 if (!btrfs_super_root(disk_super))
1073 goto fail_sb_buffer;
1075 if (btrfs_super_num_devices(disk_super) != fs_devices->num_devices) {
1076 printk("Btrfs: wanted %llu devices, but found %llu\n",
1077 (unsigned long long)btrfs_super_num_devices(disk_super),
1078 (unsigned long long)fs_devices->num_devices);
1079 goto fail_sb_buffer;
1081 nodesize = btrfs_super_nodesize(disk_super);
1082 leafsize = btrfs_super_leafsize(disk_super);
1083 sectorsize = btrfs_super_sectorsize(disk_super);
1084 stripesize = btrfs_super_stripesize(disk_super);
1085 tree_root->nodesize = nodesize;
1086 tree_root->leafsize = leafsize;
1087 tree_root->sectorsize = sectorsize;
1088 tree_root->stripesize = stripesize;
1089 sb_set_blocksize(sb, sectorsize);
1091 i_size_write(fs_info->btree_inode,
1092 btrfs_super_total_bytes(disk_super));
1094 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
1095 sizeof(disk_super->magic))) {
1096 printk("btrfs: valid FS not found on %s\n", sb->s_id);
1097 goto fail_sb_buffer;
1100 mutex_lock(&fs_info->fs_mutex);
1102 ret = btrfs_read_sys_array(tree_root);
1105 blocksize = btrfs_level_size(tree_root,
1106 btrfs_super_chunk_root_level(disk_super));
1108 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1109 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1111 chunk_root->node = read_tree_block(chunk_root,
1112 btrfs_super_chunk_root(disk_super),
1114 BUG_ON(!chunk_root->node);
1116 ret = btrfs_read_chunk_tree(chunk_root);
1119 blocksize = btrfs_level_size(tree_root,
1120 btrfs_super_root_level(disk_super));
1123 tree_root->node = read_tree_block(tree_root,
1124 btrfs_super_root(disk_super),
1126 if (!tree_root->node)
1127 goto fail_sb_buffer;
1130 ret = find_and_setup_root(tree_root, fs_info,
1131 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
1133 goto fail_tree_root;
1134 extent_root->track_dirty = 1;
1136 ret = find_and_setup_root(tree_root, fs_info,
1137 BTRFS_DEV_TREE_OBJECTID, dev_root);
1138 dev_root->track_dirty = 1;
1141 goto fail_extent_root;
1143 btrfs_read_block_groups(extent_root);
1145 fs_info->generation = btrfs_super_generation(disk_super) + 1;
1146 fs_info->data_alloc_profile = (u64)-1;
1147 fs_info->metadata_alloc_profile = (u64)-1;
1148 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
1150 mutex_unlock(&fs_info->fs_mutex);
1154 free_extent_buffer(extent_root->node);
1156 mutex_unlock(&fs_info->fs_mutex);
1157 free_extent_buffer(tree_root->node);
1159 free_extent_buffer(fs_info->sb_buffer);
1160 extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
1162 iput(fs_info->btree_inode);
1164 close_all_devices(fs_info);
1167 bdi_destroy(&fs_info->bdi);
1169 return ERR_PTR(err);
1172 static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
1174 char b[BDEVNAME_SIZE];
1177 set_buffer_uptodate(bh);
1179 if (!buffer_eopnotsupp(bh) && printk_ratelimit()) {
1180 printk(KERN_WARNING "lost page write due to "
1181 "I/O error on %s\n",
1182 bdevname(bh->b_bdev, b));
1184 set_buffer_write_io_error(bh);
1185 clear_buffer_uptodate(bh);
1191 int write_all_supers(struct btrfs_root *root)
1193 struct list_head *cur;
1194 struct list_head *head = &root->fs_info->fs_devices->devices;
1195 struct btrfs_device *dev;
1196 struct extent_buffer *sb;
1197 struct btrfs_dev_item *dev_item;
1198 struct buffer_head *bh;
1202 do_barriers = !btrfs_test_opt(root, NOBARRIER);
1204 sb = root->fs_info->sb_buffer;
1205 dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block,
1207 list_for_each(cur, head) {
1208 dev = list_entry(cur, struct btrfs_device, dev_list);
1209 btrfs_set_device_type(sb, dev_item, dev->type);
1210 btrfs_set_device_id(sb, dev_item, dev->devid);
1211 btrfs_set_device_total_bytes(sb, dev_item, dev->total_bytes);
1212 btrfs_set_device_bytes_used(sb, dev_item, dev->bytes_used);
1213 btrfs_set_device_io_align(sb, dev_item, dev->io_align);
1214 btrfs_set_device_io_width(sb, dev_item, dev->io_width);
1215 btrfs_set_device_sector_size(sb, dev_item, dev->sector_size);
1216 write_extent_buffer(sb, dev->uuid,
1217 (unsigned long)btrfs_device_uuid(dev_item),
1218 BTRFS_DEV_UUID_SIZE);
1220 btrfs_set_header_flag(sb, BTRFS_HEADER_FLAG_WRITTEN);
1221 csum_tree_block(root, sb, 0);
1223 bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET /
1224 root->fs_info->sb->s_blocksize,
1225 BTRFS_SUPER_INFO_SIZE);
1227 read_extent_buffer(sb, bh->b_data, 0, BTRFS_SUPER_INFO_SIZE);
1228 dev->pending_io = bh;
1231 set_buffer_uptodate(bh);
1233 bh->b_end_io = btrfs_end_buffer_write_sync;
1235 if (do_barriers && dev->barriers) {
1236 ret = submit_bh(WRITE_BARRIER, bh);
1237 if (ret == -EOPNOTSUPP) {
1238 printk("btrfs: disabling barriers on dev %s\n",
1240 set_buffer_uptodate(bh);
1244 ret = submit_bh(WRITE, bh);
1247 ret = submit_bh(WRITE, bh);
1252 list_for_each(cur, head) {
1253 dev = list_entry(cur, struct btrfs_device, dev_list);
1254 BUG_ON(!dev->pending_io);
1255 bh = dev->pending_io;
1257 if (!buffer_uptodate(dev->pending_io)) {
1258 if (do_barriers && dev->barriers) {
1259 printk("btrfs: disabling barriers on dev %s\n",
1261 set_buffer_uptodate(bh);
1265 ret = submit_bh(WRITE, bh);
1268 BUG_ON(!buffer_uptodate(bh));
1274 dev->pending_io = NULL;
1280 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
1285 ret = write_all_supers(root);
1287 if (!btrfs_test_opt(root, NOBARRIER))
1288 blkdev_issue_flush(sb->s_bdev, NULL);
1289 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, super);
1290 ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping,
1291 super->start, super->len);
1292 if (!btrfs_test_opt(root, NOBARRIER))
1293 blkdev_issue_flush(sb->s_bdev, NULL);
1298 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
1300 radix_tree_delete(&fs_info->fs_roots_radix,
1301 (unsigned long)root->root_key.objectid);
1303 btrfs_sysfs_del_root(root);
1307 free_extent_buffer(root->node);
1308 if (root->commit_root)
1309 free_extent_buffer(root->commit_root);
1316 static int del_fs_roots(struct btrfs_fs_info *fs_info)
1319 struct btrfs_root *gang[8];
1323 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
1328 for (i = 0; i < ret; i++)
1329 btrfs_free_fs_root(fs_info, gang[i]);
1334 int close_ctree(struct btrfs_root *root)
1337 struct btrfs_trans_handle *trans;
1338 struct btrfs_fs_info *fs_info = root->fs_info;
1340 fs_info->closing = 1;
1341 btrfs_transaction_flush_work(root);
1342 mutex_lock(&fs_info->fs_mutex);
1343 btrfs_defrag_dirty_roots(root->fs_info);
1344 trans = btrfs_start_transaction(root, 1);
1345 ret = btrfs_commit_transaction(trans, root);
1346 /* run commit again to drop the original snapshot */
1347 trans = btrfs_start_transaction(root, 1);
1348 btrfs_commit_transaction(trans, root);
1349 ret = btrfs_write_and_wait_transaction(NULL, root);
1351 write_ctree_super(NULL, root);
1352 mutex_unlock(&fs_info->fs_mutex);
1354 if (fs_info->delalloc_bytes) {
1355 printk("btrfs: at unmount delalloc count %Lu\n",
1356 fs_info->delalloc_bytes);
1358 if (fs_info->extent_root->node)
1359 free_extent_buffer(fs_info->extent_root->node);
1361 if (fs_info->tree_root->node)
1362 free_extent_buffer(fs_info->tree_root->node);
1364 if (root->fs_info->chunk_root->node);
1365 free_extent_buffer(root->fs_info->chunk_root->node);
1367 if (root->fs_info->dev_root->node);
1368 free_extent_buffer(root->fs_info->dev_root->node);
1370 free_extent_buffer(fs_info->sb_buffer);
1372 btrfs_free_block_groups(root->fs_info);
1373 del_fs_roots(fs_info);
1375 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
1377 extent_io_tree_empty_lru(&fs_info->free_space_cache);
1378 extent_io_tree_empty_lru(&fs_info->block_group_cache);
1379 extent_io_tree_empty_lru(&fs_info->pinned_extents);
1380 extent_io_tree_empty_lru(&fs_info->pending_del);
1381 extent_io_tree_empty_lru(&fs_info->extent_ins);
1382 extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
1384 truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
1385 flush_workqueue(end_io_workqueue);
1386 destroy_workqueue(end_io_workqueue);
1388 iput(fs_info->btree_inode);
1390 while(!list_empty(&fs_info->hashers)) {
1391 struct btrfs_hasher *hasher;
1392 hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher,
1394 list_del(&hasher->hashers);
1395 crypto_free_hash(&fs_info->hash_tfm);
1399 close_all_devices(fs_info);
1400 btrfs_mapping_tree_free(&fs_info->mapping_tree);
1401 bdi_destroy(&fs_info->bdi);
1403 kfree(fs_info->extent_root);
1404 kfree(fs_info->tree_root);
1405 kfree(fs_info->chunk_root);
1406 kfree(fs_info->dev_root);
1410 int btrfs_buffer_uptodate(struct extent_buffer *buf)
1412 struct inode *btree_inode = buf->first_page->mapping->host;
1413 return extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf);
1416 int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
1418 struct inode *btree_inode = buf->first_page->mapping->host;
1419 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
1423 void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
1425 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1426 u64 transid = btrfs_header_generation(buf);
1427 struct inode *btree_inode = root->fs_info->btree_inode;
1429 if (transid != root->fs_info->generation) {
1430 printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
1431 (unsigned long long)buf->start,
1432 transid, root->fs_info->generation);
1435 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf);
1438 void btrfs_throttle(struct btrfs_root *root)
1440 struct backing_dev_info *bdi;
1442 bdi = root->fs_info->sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
1443 if (root->fs_info->throttles && bdi_write_congested(bdi)) {
1444 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18)
1445 congestion_wait(WRITE, HZ/20);
1447 blk_congestion_wait(WRITE, HZ/20);
1452 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
1454 balance_dirty_pages_ratelimited_nr(
1455 root->fs_info->btree_inode->i_mapping, 1);
1458 void btrfs_set_buffer_defrag(struct extent_buffer *buf)
1460 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1461 struct inode *btree_inode = root->fs_info->btree_inode;
1462 set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
1463 buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS);
1466 void btrfs_set_buffer_defrag_done(struct extent_buffer *buf)
1468 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1469 struct inode *btree_inode = root->fs_info->btree_inode;
1470 set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
1471 buf->start + buf->len - 1, EXTENT_DEFRAG_DONE,
1475 int btrfs_buffer_defrag(struct extent_buffer *buf)
1477 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1478 struct inode *btree_inode = root->fs_info->btree_inode;
1479 return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
1480 buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0);
1483 int btrfs_buffer_defrag_done(struct extent_buffer *buf)
1485 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1486 struct inode *btree_inode = root->fs_info->btree_inode;
1487 return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
1488 buf->start, buf->start + buf->len - 1,
1489 EXTENT_DEFRAG_DONE, 0);
1492 int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf)
1494 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1495 struct inode *btree_inode = root->fs_info->btree_inode;
1496 return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
1497 buf->start, buf->start + buf->len - 1,
1498 EXTENT_DEFRAG_DONE, GFP_NOFS);
1501 int btrfs_clear_buffer_defrag(struct extent_buffer *buf)
1503 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1504 struct inode *btree_inode = root->fs_info->btree_inode;
1505 return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
1506 buf->start, buf->start + buf->len - 1,
1507 EXTENT_DEFRAG, GFP_NOFS);
1510 int btrfs_read_buffer(struct extent_buffer *buf)
1512 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1514 ret = btree_read_extent_buffer_pages(root, buf, 0);
1516 buf->flags |= EXTENT_UPTODATE;
1521 static struct extent_io_ops btree_extent_io_ops = {
1522 .writepage_io_hook = btree_writepage_io_hook,
1523 .readpage_end_io_hook = btree_readpage_end_io_hook,
1524 .submit_bio_hook = btree_submit_bio_hook,
1525 /* note we're sharing with inode.c for the merge bio hook */
1526 .merge_bio_hook = btrfs_merge_bio_hook,