2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/pagemap.h>
25 #include <linux/highmem.h>
26 #include <linux/time.h>
27 #include <linux/init.h>
28 #include <linux/string.h>
29 #include <linux/smp_lock.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/bit_spinlock.h>
37 #include <linux/version.h>
38 #include <linux/xattr.h>
39 #include <linux/posix_acl.h>
42 #include "transaction.h"
43 #include "btrfs_inode.h"
45 #include "print-tree.h"
47 #include "ordered-data.h"
49 struct btrfs_iget_args {
51 struct btrfs_root *root;
54 static struct inode_operations btrfs_dir_inode_operations;
55 static struct inode_operations btrfs_symlink_inode_operations;
56 static struct inode_operations btrfs_dir_ro_inode_operations;
57 static struct inode_operations btrfs_special_inode_operations;
58 static struct inode_operations btrfs_file_inode_operations;
59 static struct address_space_operations btrfs_aops;
60 static struct address_space_operations btrfs_symlink_aops;
61 static struct file_operations btrfs_dir_file_operations;
62 static struct extent_io_ops btrfs_extent_io_ops;
64 static struct kmem_cache *btrfs_inode_cachep;
65 struct kmem_cache *btrfs_trans_handle_cachep;
66 struct kmem_cache *btrfs_transaction_cachep;
67 struct kmem_cache *btrfs_bit_radix_cachep;
68 struct kmem_cache *btrfs_path_cachep;
71 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
72 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
73 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
74 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
75 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
76 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
77 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
78 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
81 static void btrfs_truncate(struct inode *inode);
83 int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
92 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
93 total = btrfs_super_total_bytes(&root->fs_info->super_copy);
94 used = btrfs_super_bytes_used(&root->fs_info->super_copy);
102 if (used + root->fs_info->delalloc_bytes + num_required > thresh)
104 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
108 static int cow_file_range(struct inode *inode, u64 start, u64 end)
110 struct btrfs_root *root = BTRFS_I(inode)->root;
111 struct btrfs_trans_handle *trans;
115 u64 blocksize = root->sectorsize;
117 struct btrfs_key ins;
118 struct extent_map *em;
119 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
122 trans = btrfs_join_transaction(root, 1);
124 btrfs_set_trans_block_group(trans, inode);
126 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
127 num_bytes = max(blocksize, num_bytes);
128 orig_num_bytes = num_bytes;
130 if (alloc_hint == EXTENT_MAP_INLINE)
133 BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy));
134 mutex_lock(&BTRFS_I(inode)->extent_mutex);
135 btrfs_drop_extent_cache(inode, start, start + num_bytes - 1);
136 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
138 while(num_bytes > 0) {
139 cur_alloc_size = min(num_bytes, root->fs_info->max_extent);
140 ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
141 root->sectorsize, 0, 0,
147 em = alloc_extent_map(GFP_NOFS);
149 em->len = ins.offset;
150 em->block_start = ins.objectid;
151 em->bdev = root->fs_info->fs_devices->latest_bdev;
152 mutex_lock(&BTRFS_I(inode)->extent_mutex);
153 set_bit(EXTENT_FLAG_PINNED, &em->flags);
155 spin_lock(&em_tree->lock);
156 ret = add_extent_mapping(em_tree, em);
157 spin_unlock(&em_tree->lock);
158 if (ret != -EEXIST) {
162 btrfs_drop_extent_cache(inode, start,
163 start + ins.offset - 1);
165 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
167 cur_alloc_size = ins.offset;
168 ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
171 if (num_bytes < cur_alloc_size) {
172 printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes,
176 num_bytes -= cur_alloc_size;
177 alloc_hint = ins.objectid + ins.offset;
178 start += cur_alloc_size;
181 btrfs_end_transaction(trans, root);
185 static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end)
193 struct btrfs_root *root = BTRFS_I(inode)->root;
194 struct btrfs_block_group_cache *block_group;
195 struct extent_buffer *leaf;
197 struct btrfs_path *path;
198 struct btrfs_file_extent_item *item;
201 struct btrfs_key found_key;
203 total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
204 path = btrfs_alloc_path();
207 ret = btrfs_lookup_file_extent(NULL, root, path,
208 inode->i_ino, start, 0);
210 btrfs_free_path(path);
216 if (path->slots[0] == 0)
221 leaf = path->nodes[0];
222 item = btrfs_item_ptr(leaf, path->slots[0],
223 struct btrfs_file_extent_item);
225 /* are we inside the extent that was found? */
226 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
227 found_type = btrfs_key_type(&found_key);
228 if (found_key.objectid != inode->i_ino ||
229 found_type != BTRFS_EXTENT_DATA_KEY)
232 found_type = btrfs_file_extent_type(leaf, item);
233 extent_start = found_key.offset;
234 if (found_type == BTRFS_FILE_EXTENT_REG) {
235 u64 extent_num_bytes;
237 extent_num_bytes = btrfs_file_extent_num_bytes(leaf, item);
238 extent_end = extent_start + extent_num_bytes;
241 if (loops && start != extent_start)
244 if (start < extent_start || start >= extent_end)
247 cow_end = min(end, extent_end - 1);
248 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
252 if (btrfs_count_snapshots_in_path(root, path, inode->i_ino,
258 * we may be called by the resizer, make sure we're inside
259 * the limits of the FS
261 block_group = btrfs_lookup_block_group(root->fs_info,
263 if (!block_group || block_group->ro)
272 btrfs_free_path(path);
275 btrfs_release_path(root, path);
280 cow_file_range(inode, start, end);
285 static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
287 struct btrfs_root *root = BTRFS_I(inode)->root;
290 if (btrfs_test_opt(root, NODATACOW) ||
291 btrfs_test_flag(inode, NODATACOW))
292 ret = run_delalloc_nocow(inode, start, end);
294 ret = cow_file_range(inode, start, end);
299 int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
300 unsigned long old, unsigned long bits)
303 if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
304 struct btrfs_root *root = BTRFS_I(inode)->root;
305 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
306 BTRFS_I(inode)->delalloc_bytes += end - start + 1;
307 root->fs_info->delalloc_bytes += end - start + 1;
308 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
313 int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end,
314 unsigned long old, unsigned long bits)
316 if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
317 struct btrfs_root *root = BTRFS_I(inode)->root;
320 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
321 if (end - start + 1 > root->fs_info->delalloc_bytes) {
322 printk("warning: delalloc account %Lu %Lu\n",
323 end - start + 1, root->fs_info->delalloc_bytes);
324 root->fs_info->delalloc_bytes = 0;
325 BTRFS_I(inode)->delalloc_bytes = 0;
327 root->fs_info->delalloc_bytes -= end - start + 1;
328 BTRFS_I(inode)->delalloc_bytes -= end - start + 1;
330 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
335 int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
336 size_t size, struct bio *bio)
338 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
339 struct btrfs_mapping_tree *map_tree;
340 u64 logical = bio->bi_sector << 9;
345 length = bio->bi_size;
346 map_tree = &root->fs_info->mapping_tree;
348 ret = btrfs_map_block(map_tree, READ, logical,
349 &map_length, NULL, 0);
351 if (map_length < length + size) {
357 int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
360 struct btrfs_root *root = BTRFS_I(inode)->root;
363 ret = btrfs_csum_one_bio(root, inode, bio);
366 return btrfs_map_bio(root, rw, bio, mirror_num, 1);
369 int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
372 struct btrfs_root *root = BTRFS_I(inode)->root;
375 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
378 if (!(rw & (1 << BIO_RW))) {
382 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
383 inode, rw, bio, mirror_num,
384 __btrfs_submit_bio_hook);
386 return btrfs_map_bio(root, rw, bio, mirror_num, 0);
389 static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
390 struct inode *inode, u64 file_offset,
391 struct list_head *list)
393 struct list_head *cur;
394 struct btrfs_ordered_sum *sum;
396 btrfs_set_trans_block_group(trans, inode);
397 list_for_each(cur, list) {
398 sum = list_entry(cur, struct btrfs_ordered_sum, list);
399 mutex_lock(&BTRFS_I(inode)->csum_mutex);
400 btrfs_csum_file_blocks(trans, BTRFS_I(inode)->root,
402 mutex_unlock(&BTRFS_I(inode)->csum_mutex);
407 struct btrfs_writepage_fixup {
409 struct btrfs_work work;
412 /* see btrfs_writepage_start_hook for details on why this is required */
413 void btrfs_writepage_fixup_worker(struct btrfs_work *work)
415 struct btrfs_writepage_fixup *fixup;
416 struct btrfs_ordered_extent *ordered;
422 fixup = container_of(work, struct btrfs_writepage_fixup, work);
426 if (!page->mapping || !PageDirty(page) || !PageChecked(page)) {
427 ClearPageChecked(page);
431 inode = page->mapping->host;
432 page_start = page_offset(page);
433 page_end = page_offset(page) + PAGE_CACHE_SIZE - 1;
435 lock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
437 /* already ordered? We're done */
438 if (test_range_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
439 EXTENT_ORDERED, 0)) {
443 ordered = btrfs_lookup_ordered_extent(inode, page_start);
445 unlock_extent(&BTRFS_I(inode)->io_tree, page_start,
448 btrfs_start_ordered_extent(inode, ordered, 1);
452 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, page_end,
454 ClearPageChecked(page);
456 unlock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
459 page_cache_release(page);
463 * There are a few paths in the higher layers of the kernel that directly
464 * set the page dirty bit without asking the filesystem if it is a
465 * good idea. This causes problems because we want to make sure COW
466 * properly happens and the data=ordered rules are followed.
468 * In our case any range that doesn't have the EXTENT_ORDERED bit set
469 * hasn't been properly setup for IO. We kick off an async process
470 * to fix it up. The async helper will wait for ordered extents, set
471 * the delalloc bit and make it safe to write the page.
473 int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
475 struct inode *inode = page->mapping->host;
476 struct btrfs_writepage_fixup *fixup;
477 struct btrfs_root *root = BTRFS_I(inode)->root;
480 ret = test_range_bit(&BTRFS_I(inode)->io_tree, start, end,
485 if (PageChecked(page))
488 fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
492 SetPageChecked(page);
493 page_cache_get(page);
494 fixup->work.func = btrfs_writepage_fixup_worker;
496 btrfs_queue_worker(&root->fs_info->fixup_workers, &fixup->work);
500 static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
502 struct btrfs_root *root = BTRFS_I(inode)->root;
503 struct btrfs_trans_handle *trans;
504 struct btrfs_ordered_extent *ordered_extent;
505 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
506 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
507 struct extent_map *em;
508 struct extent_map *em_orig;
512 struct list_head list;
513 struct btrfs_key ins;
517 ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1);
521 trans = btrfs_join_transaction(root, 1);
523 ordered_extent = btrfs_lookup_ordered_extent(inode, start);
524 BUG_ON(!ordered_extent);
526 lock_extent(io_tree, ordered_extent->file_offset,
527 ordered_extent->file_offset + ordered_extent->len - 1,
530 INIT_LIST_HEAD(&list);
532 ins.objectid = ordered_extent->start;
533 ins.offset = ordered_extent->len;
534 ins.type = BTRFS_EXTENT_ITEM_KEY;
536 ret = btrfs_alloc_reserved_extent(trans, root, root->root_key.objectid,
537 trans->transid, inode->i_ino,
538 ordered_extent->file_offset, &ins);
541 mutex_lock(&BTRFS_I(inode)->extent_mutex);
543 spin_lock(&em_tree->lock);
544 clear_start = ordered_extent->file_offset;
545 clear_end = ordered_extent->file_offset + ordered_extent->len;
546 em = lookup_extent_mapping(em_tree, clear_start,
547 ordered_extent->len);
549 while(em && clear_start < extent_map_end(em) && clear_end > em->start) {
550 clear_bit(EXTENT_FLAG_PINNED, &em->flags);
551 rb = rb_next(&em->rb_node);
554 em = rb_entry(rb, struct extent_map, rb_node);
556 free_extent_map(em_orig);
557 spin_unlock(&em_tree->lock);
559 ret = btrfs_drop_extents(trans, root, inode,
560 ordered_extent->file_offset,
561 ordered_extent->file_offset +
563 ordered_extent->file_offset, &alloc_hint);
565 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
566 ordered_extent->file_offset,
567 ordered_extent->start,
569 ordered_extent->len, 0);
572 btrfs_drop_extent_cache(inode, ordered_extent->file_offset,
573 ordered_extent->file_offset +
574 ordered_extent->len - 1);
575 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
577 inode->i_blocks += ordered_extent->len >> 9;
578 unlock_extent(io_tree, ordered_extent->file_offset,
579 ordered_extent->file_offset + ordered_extent->len - 1,
581 add_pending_csums(trans, inode, ordered_extent->file_offset,
582 &ordered_extent->list);
584 btrfs_ordered_update_i_size(inode, ordered_extent);
585 btrfs_remove_ordered_extent(inode, ordered_extent);
588 btrfs_put_ordered_extent(ordered_extent);
589 /* once for the tree */
590 btrfs_put_ordered_extent(ordered_extent);
592 btrfs_update_inode(trans, root, inode);
593 btrfs_end_transaction(trans, root);
597 int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
598 struct extent_state *state, int uptodate)
600 return btrfs_finish_ordered_io(page->mapping->host, start, end);
603 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
606 struct inode *inode = page->mapping->host;
607 struct btrfs_root *root = BTRFS_I(inode)->root;
608 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
609 struct btrfs_csum_item *item;
610 struct btrfs_path *path = NULL;
613 if (btrfs_test_opt(root, NODATASUM) ||
614 btrfs_test_flag(inode, NODATASUM))
618 * It is possible there is an ordered extent that has
619 * not yet finished for this range in the file. If so,
620 * that extent will have a csum cached, and it will insert
621 * the sum after all the blocks in the extent are fully
622 * on disk. So, look for an ordered extent and use the
623 * sum if found. We have to do this before looking in the
624 * btree because csum items are pre-inserted based on
625 * the file size. btrfs_lookup_csum might find an item
626 * that still hasn't been fully filled.
628 ret = btrfs_find_ordered_sum(inode, start, &csum);
633 path = btrfs_alloc_path();
634 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
637 /* a csum that isn't present is a preallocated region. */
638 if (ret == -ENOENT || ret == -EFBIG)
641 printk("no csum found for inode %lu start %Lu\n", inode->i_ino,
645 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
648 set_state_private(io_tree, start, csum);
651 btrfs_free_path(path);
655 struct io_failure_record {
663 int btrfs_io_failed_hook(struct bio *failed_bio,
664 struct page *page, u64 start, u64 end,
665 struct extent_state *state)
667 struct io_failure_record *failrec = NULL;
669 struct extent_map *em;
670 struct inode *inode = page->mapping->host;
671 struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
672 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
679 ret = get_state_private(failure_tree, start, &private);
681 failrec = kmalloc(sizeof(*failrec), GFP_NOFS);
684 failrec->start = start;
685 failrec->len = end - start + 1;
686 failrec->last_mirror = 0;
688 spin_lock(&em_tree->lock);
689 em = lookup_extent_mapping(em_tree, start, failrec->len);
690 if (em->start > start || em->start + em->len < start) {
694 spin_unlock(&em_tree->lock);
696 if (!em || IS_ERR(em)) {
700 logical = start - em->start;
701 logical = em->block_start + logical;
702 failrec->logical = logical;
704 set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
705 EXTENT_DIRTY, GFP_NOFS);
706 set_state_private(failure_tree, start,
707 (u64)(unsigned long)failrec);
709 failrec = (struct io_failure_record *)(unsigned long)private;
711 num_copies = btrfs_num_copies(
712 &BTRFS_I(inode)->root->fs_info->mapping_tree,
713 failrec->logical, failrec->len);
714 failrec->last_mirror++;
716 spin_lock_irq(&BTRFS_I(inode)->io_tree.lock);
717 state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
720 if (state && state->start != failrec->start)
722 spin_unlock_irq(&BTRFS_I(inode)->io_tree.lock);
724 if (!state || failrec->last_mirror > num_copies) {
725 set_state_private(failure_tree, failrec->start, 0);
726 clear_extent_bits(failure_tree, failrec->start,
727 failrec->start + failrec->len - 1,
728 EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
732 bio = bio_alloc(GFP_NOFS, 1);
733 bio->bi_private = state;
734 bio->bi_end_io = failed_bio->bi_end_io;
735 bio->bi_sector = failrec->logical >> 9;
736 bio->bi_bdev = failed_bio->bi_bdev;
738 bio_add_page(bio, page, failrec->len, start - page_offset(page));
739 if (failed_bio->bi_rw & (1 << BIO_RW))
744 BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
745 failrec->last_mirror);
749 int btrfs_clean_io_failures(struct inode *inode, u64 start)
753 struct io_failure_record *failure;
757 if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
758 (u64)-1, 1, EXTENT_DIRTY)) {
759 ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
760 start, &private_failure);
762 failure = (struct io_failure_record *)(unsigned long)
764 set_state_private(&BTRFS_I(inode)->io_failure_tree,
766 clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
768 failure->start + failure->len - 1,
769 EXTENT_DIRTY | EXTENT_LOCKED,
777 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
778 struct extent_state *state)
780 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
781 struct inode *inode = page->mapping->host;
782 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
784 u64 private = ~(u32)0;
786 struct btrfs_root *root = BTRFS_I(inode)->root;
790 if (btrfs_test_opt(root, NODATASUM) ||
791 btrfs_test_flag(inode, NODATASUM))
793 if (state && state->start == start) {
794 private = state->private;
797 ret = get_state_private(io_tree, start, &private);
799 local_irq_save(flags);
800 kaddr = kmap_atomic(page, KM_IRQ0);
804 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
805 btrfs_csum_final(csum, (char *)&csum);
806 if (csum != private) {
809 kunmap_atomic(kaddr, KM_IRQ0);
810 local_irq_restore(flags);
812 /* if the io failure tree for this inode is non-empty,
813 * check to see if we've recovered from a failed IO
815 btrfs_clean_io_failures(inode, start);
819 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
820 page->mapping->host->i_ino, (unsigned long long)start, csum,
822 memset(kaddr + offset, 1, end - start + 1);
823 flush_dcache_page(page);
824 kunmap_atomic(kaddr, KM_IRQ0);
825 local_irq_restore(flags);
832 * This creates an orphan entry for the given inode in case something goes
833 * wrong in the middle of an unlink/truncate.
835 int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode)
837 struct btrfs_root *root = BTRFS_I(inode)->root;
840 spin_lock(&root->orphan_lock);
842 /* already on the orphan list, we're good */
843 if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
844 spin_unlock(&root->orphan_lock);
848 list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);
850 spin_unlock(&root->orphan_lock);
853 * insert an orphan item to track this unlinked/truncated file
855 ret = btrfs_insert_orphan_item(trans, root, inode->i_ino);
861 * We have done the truncate/delete so we can go ahead and remove the orphan
862 * item for this particular inode.
864 int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode)
866 struct btrfs_root *root = BTRFS_I(inode)->root;
869 spin_lock(&root->orphan_lock);
871 if (list_empty(&BTRFS_I(inode)->i_orphan)) {
872 spin_unlock(&root->orphan_lock);
876 list_del_init(&BTRFS_I(inode)->i_orphan);
878 spin_unlock(&root->orphan_lock);
882 spin_unlock(&root->orphan_lock);
884 ret = btrfs_del_orphan_item(trans, root, inode->i_ino);
890 * this cleans up any orphans that may be left on the list from the last use
893 void btrfs_orphan_cleanup(struct btrfs_root *root)
895 struct btrfs_path *path;
896 struct extent_buffer *leaf;
897 struct btrfs_item *item;
898 struct btrfs_key key, found_key;
899 struct btrfs_trans_handle *trans;
901 int ret = 0, nr_unlink = 0, nr_truncate = 0;
903 /* don't do orphan cleanup if the fs is readonly. */
904 if (root->inode->i_sb->s_flags & MS_RDONLY)
907 path = btrfs_alloc_path();
912 key.objectid = BTRFS_ORPHAN_OBJECTID;
913 btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY);
914 key.offset = (u64)-1;
916 trans = btrfs_start_transaction(root, 1);
917 btrfs_set_trans_block_group(trans, root->inode);
920 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
922 printk(KERN_ERR "Error searching slot for orphan: %d"
928 * if ret == 0 means we found what we were searching for, which
929 * is weird, but possible, so only screw with path if we didnt
930 * find the key and see if we have stuff that matches
933 if (path->slots[0] == 0)
938 /* pull out the item */
939 leaf = path->nodes[0];
940 item = btrfs_item_nr(leaf, path->slots[0]);
941 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
943 /* make sure the item matches what we want */
944 if (found_key.objectid != BTRFS_ORPHAN_OBJECTID)
946 if (btrfs_key_type(&found_key) != BTRFS_ORPHAN_ITEM_KEY)
949 /* release the path since we're done with it */
950 btrfs_release_path(root, path);
953 * this is where we are basically btrfs_lookup, without the
954 * crossing root thing. we store the inode number in the
955 * offset of the orphan item.
957 inode = btrfs_iget_locked(root->inode->i_sb,
958 found_key.offset, root);
962 if (inode->i_state & I_NEW) {
963 BTRFS_I(inode)->root = root;
965 /* have to set the location manually */
966 BTRFS_I(inode)->location.objectid = inode->i_ino;
967 BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
968 BTRFS_I(inode)->location.offset = 0;
970 btrfs_read_locked_inode(inode);
971 unlock_new_inode(inode);
975 * add this inode to the orphan list so btrfs_orphan_del does
976 * the proper thing when we hit it
978 spin_lock(&root->orphan_lock);
979 list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);
980 spin_unlock(&root->orphan_lock);
983 * if this is a bad inode, means we actually succeeded in
984 * removing the inode, but not the orphan record, which means
985 * we need to manually delete the orphan since iput will just
988 if (is_bad_inode(inode)) {
989 btrfs_orphan_del(trans, inode);
994 /* if we have links, this was a truncate, lets do that */
995 if (inode->i_nlink) {
997 btrfs_truncate(inode);
1002 /* this will do delete_inode and everything for us */
1007 printk(KERN_INFO "btrfs: unlinked %d orphans\n", nr_unlink);
1009 printk(KERN_INFO "btrfs: truncated %d orphans\n", nr_truncate);
1011 btrfs_free_path(path);
1012 btrfs_end_transaction(trans, root);
1015 void btrfs_read_locked_inode(struct inode *inode)
1017 struct btrfs_path *path;
1018 struct extent_buffer *leaf;
1019 struct btrfs_inode_item *inode_item;
1020 struct btrfs_timespec *tspec;
1021 struct btrfs_root *root = BTRFS_I(inode)->root;
1022 struct btrfs_key location;
1023 u64 alloc_group_block;
1027 path = btrfs_alloc_path();
1029 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
1031 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
1035 leaf = path->nodes[0];
1036 inode_item = btrfs_item_ptr(leaf, path->slots[0],
1037 struct btrfs_inode_item);
1039 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
1040 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
1041 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
1042 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
1043 btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item));
1045 tspec = btrfs_inode_atime(inode_item);
1046 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
1047 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
1049 tspec = btrfs_inode_mtime(inode_item);
1050 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
1051 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
1053 tspec = btrfs_inode_ctime(inode_item);
1054 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
1055 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
1057 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
1058 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
1060 rdev = btrfs_inode_rdev(leaf, inode_item);
1062 BTRFS_I(inode)->index_cnt = (u64)-1;
1064 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
1065 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
1067 BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
1068 if (!BTRFS_I(inode)->block_group) {
1069 BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
1071 BTRFS_BLOCK_GROUP_METADATA, 0);
1073 btrfs_free_path(path);
1076 switch (inode->i_mode & S_IFMT) {
1078 inode->i_mapping->a_ops = &btrfs_aops;
1079 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
1080 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
1081 inode->i_fop = &btrfs_file_operations;
1082 inode->i_op = &btrfs_file_inode_operations;
1085 inode->i_fop = &btrfs_dir_file_operations;
1086 if (root == root->fs_info->tree_root)
1087 inode->i_op = &btrfs_dir_ro_inode_operations;
1089 inode->i_op = &btrfs_dir_inode_operations;
1092 inode->i_op = &btrfs_symlink_inode_operations;
1093 inode->i_mapping->a_ops = &btrfs_symlink_aops;
1094 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
1097 init_special_inode(inode, inode->i_mode, rdev);
1103 btrfs_free_path(path);
1104 make_bad_inode(inode);
1107 static void fill_inode_item(struct extent_buffer *leaf,
1108 struct btrfs_inode_item *item,
1109 struct inode *inode)
1111 btrfs_set_inode_uid(leaf, item, inode->i_uid);
1112 btrfs_set_inode_gid(leaf, item, inode->i_gid);
1113 btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size);
1114 btrfs_set_inode_mode(leaf, item, inode->i_mode);
1115 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
1117 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
1118 inode->i_atime.tv_sec);
1119 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
1120 inode->i_atime.tv_nsec);
1122 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
1123 inode->i_mtime.tv_sec);
1124 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
1125 inode->i_mtime.tv_nsec);
1127 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
1128 inode->i_ctime.tv_sec);
1129 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
1130 inode->i_ctime.tv_nsec);
1132 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
1133 btrfs_set_inode_generation(leaf, item, inode->i_generation);
1134 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
1135 btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
1136 btrfs_set_inode_block_group(leaf, item,
1137 BTRFS_I(inode)->block_group->key.objectid);
1140 int noinline btrfs_update_inode(struct btrfs_trans_handle *trans,
1141 struct btrfs_root *root,
1142 struct inode *inode)
1144 struct btrfs_inode_item *inode_item;
1145 struct btrfs_path *path;
1146 struct extent_buffer *leaf;
1149 path = btrfs_alloc_path();
1151 ret = btrfs_lookup_inode(trans, root, path,
1152 &BTRFS_I(inode)->location, 1);
1159 leaf = path->nodes[0];
1160 inode_item = btrfs_item_ptr(leaf, path->slots[0],
1161 struct btrfs_inode_item);
1163 fill_inode_item(leaf, inode_item, inode);
1164 btrfs_mark_buffer_dirty(leaf);
1165 btrfs_set_inode_last_trans(trans, inode);
1168 btrfs_free_path(path);
1173 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
1174 struct btrfs_root *root,
1176 struct dentry *dentry)
1178 struct btrfs_path *path;
1179 const char *name = dentry->d_name.name;
1180 int name_len = dentry->d_name.len;
1182 struct extent_buffer *leaf;
1183 struct btrfs_dir_item *di;
1184 struct btrfs_key key;
1187 path = btrfs_alloc_path();
1193 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
1194 name, name_len, -1);
1203 leaf = path->nodes[0];
1204 btrfs_dir_item_key_to_cpu(leaf, di, &key);
1205 ret = btrfs_delete_one_dir_name(trans, root, path, di);
1208 btrfs_release_path(root, path);
1210 ret = btrfs_del_inode_ref(trans, root, name, name_len,
1211 dentry->d_inode->i_ino,
1212 dentry->d_parent->d_inode->i_ino, &index);
1214 printk("failed to delete reference to %.*s, "
1215 "inode %lu parent %lu\n", name_len, name,
1216 dentry->d_inode->i_ino,
1217 dentry->d_parent->d_inode->i_ino);
1221 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
1222 index, name, name_len, -1);
1231 ret = btrfs_delete_one_dir_name(trans, root, path, di);
1232 btrfs_release_path(root, path);
1234 dentry->d_inode->i_ctime = dir->i_ctime;
1236 btrfs_free_path(path);
1238 btrfs_i_size_write(dir, dir->i_size - name_len * 2);
1239 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
1240 btrfs_update_inode(trans, root, dir);
1241 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1242 dentry->d_inode->i_nlink--;
1244 drop_nlink(dentry->d_inode);
1246 ret = btrfs_update_inode(trans, root, dentry->d_inode);
1247 dir->i_sb->s_dirt = 1;
1252 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
1254 struct btrfs_root *root;
1255 struct btrfs_trans_handle *trans;
1256 struct inode *inode = dentry->d_inode;
1258 unsigned long nr = 0;
1260 root = BTRFS_I(dir)->root;
1262 ret = btrfs_check_free_space(root, 1, 1);
1266 trans = btrfs_start_transaction(root, 1);
1268 btrfs_set_trans_block_group(trans, dir);
1269 ret = btrfs_unlink_trans(trans, root, dir, dentry);
1271 if (inode->i_nlink == 0)
1272 ret = btrfs_orphan_add(trans, inode);
1274 nr = trans->blocks_used;
1276 btrfs_end_transaction_throttle(trans, root);
1278 btrfs_btree_balance_dirty(root, nr);
1282 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
1284 struct inode *inode = dentry->d_inode;
1287 struct btrfs_root *root = BTRFS_I(dir)->root;
1288 struct btrfs_trans_handle *trans;
1289 unsigned long nr = 0;
1291 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
1295 ret = btrfs_check_free_space(root, 1, 1);
1299 trans = btrfs_start_transaction(root, 1);
1300 btrfs_set_trans_block_group(trans, dir);
1302 err = btrfs_orphan_add(trans, inode);
1306 /* now the directory is empty */
1307 err = btrfs_unlink_trans(trans, root, dir, dentry);
1309 btrfs_i_size_write(inode, 0);
1313 nr = trans->blocks_used;
1314 ret = btrfs_end_transaction_throttle(trans, root);
1316 btrfs_btree_balance_dirty(root, nr);
1324 * this can truncate away extent items, csum items and directory items.
1325 * It starts at a high offset and removes keys until it can't find
1326 * any higher than i_size.
1328 * csum items that cross the new i_size are truncated to the new size
1331 * min_type is the minimum key type to truncate down to. If set to 0, this
1332 * will kill all the items on this inode, including the INODE_ITEM_KEY.
1334 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
1335 struct btrfs_root *root,
1336 struct inode *inode,
1340 struct btrfs_path *path;
1341 struct btrfs_key key;
1342 struct btrfs_key found_key;
1344 struct extent_buffer *leaf;
1345 struct btrfs_file_extent_item *fi;
1346 u64 extent_start = 0;
1347 u64 extent_num_bytes = 0;
1353 int pending_del_nr = 0;
1354 int pending_del_slot = 0;
1355 int extent_type = -1;
1356 u64 mask = root->sectorsize - 1;
1358 btrfs_drop_extent_cache(inode, inode->i_size & (~mask), (u64)-1);
1359 path = btrfs_alloc_path();
1363 /* FIXME, add redo link to tree so we don't leak on crash */
1364 key.objectid = inode->i_ino;
1365 key.offset = (u64)-1;
1368 btrfs_init_path(path);
1370 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1375 BUG_ON(path->slots[0] == 0);
1381 leaf = path->nodes[0];
1382 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1383 found_type = btrfs_key_type(&found_key);
1385 if (found_key.objectid != inode->i_ino)
1388 if (found_type < min_type)
1391 item_end = found_key.offset;
1392 if (found_type == BTRFS_EXTENT_DATA_KEY) {
1393 fi = btrfs_item_ptr(leaf, path->slots[0],
1394 struct btrfs_file_extent_item);
1395 extent_type = btrfs_file_extent_type(leaf, fi);
1396 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
1398 btrfs_file_extent_num_bytes(leaf, fi);
1399 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1400 struct btrfs_item *item = btrfs_item_nr(leaf,
1402 item_end += btrfs_file_extent_inline_len(leaf,
1407 if (found_type == BTRFS_CSUM_ITEM_KEY) {
1408 ret = btrfs_csum_truncate(trans, root, path,
1412 if (item_end < inode->i_size) {
1413 if (found_type == BTRFS_DIR_ITEM_KEY) {
1414 found_type = BTRFS_INODE_ITEM_KEY;
1415 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
1416 found_type = BTRFS_CSUM_ITEM_KEY;
1417 } else if (found_type == BTRFS_EXTENT_DATA_KEY) {
1418 found_type = BTRFS_XATTR_ITEM_KEY;
1419 } else if (found_type == BTRFS_XATTR_ITEM_KEY) {
1420 found_type = BTRFS_INODE_REF_KEY;
1421 } else if (found_type) {
1426 btrfs_set_key_type(&key, found_type);
1429 if (found_key.offset >= inode->i_size)
1435 /* FIXME, shrink the extent if the ref count is only 1 */
1436 if (found_type != BTRFS_EXTENT_DATA_KEY)
1439 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
1441 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
1443 u64 orig_num_bytes =
1444 btrfs_file_extent_num_bytes(leaf, fi);
1445 extent_num_bytes = inode->i_size -
1446 found_key.offset + root->sectorsize - 1;
1447 extent_num_bytes = extent_num_bytes &
1448 ~((u64)root->sectorsize - 1);
1449 btrfs_set_file_extent_num_bytes(leaf, fi,
1451 num_dec = (orig_num_bytes -
1453 if (extent_start != 0)
1454 dec_i_blocks(inode, num_dec);
1455 btrfs_mark_buffer_dirty(leaf);
1458 btrfs_file_extent_disk_num_bytes(leaf,
1460 /* FIXME blocksize != 4096 */
1461 num_dec = btrfs_file_extent_num_bytes(leaf, fi);
1462 if (extent_start != 0) {
1464 dec_i_blocks(inode, num_dec);
1466 root_gen = btrfs_header_generation(leaf);
1467 root_owner = btrfs_header_owner(leaf);
1469 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1471 u32 newsize = inode->i_size - found_key.offset;
1472 dec_i_blocks(inode, item_end + 1 -
1473 found_key.offset - newsize);
1475 btrfs_file_extent_calc_inline_size(newsize);
1476 ret = btrfs_truncate_item(trans, root, path,
1480 dec_i_blocks(inode, item_end + 1 -
1486 if (!pending_del_nr) {
1487 /* no pending yet, add ourselves */
1488 pending_del_slot = path->slots[0];
1490 } else if (pending_del_nr &&
1491 path->slots[0] + 1 == pending_del_slot) {
1492 /* hop on the pending chunk */
1494 pending_del_slot = path->slots[0];
1496 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot);
1502 ret = btrfs_free_extent(trans, root, extent_start,
1505 root_gen, inode->i_ino,
1506 found_key.offset, 0);
1510 if (path->slots[0] == 0) {
1513 btrfs_release_path(root, path);
1518 if (pending_del_nr &&
1519 path->slots[0] + 1 != pending_del_slot) {
1520 struct btrfs_key debug;
1522 btrfs_item_key_to_cpu(path->nodes[0], &debug,
1524 ret = btrfs_del_items(trans, root, path,
1529 btrfs_release_path(root, path);
1535 if (pending_del_nr) {
1536 ret = btrfs_del_items(trans, root, path, pending_del_slot,
1539 btrfs_free_path(path);
1540 inode->i_sb->s_dirt = 1;
1545 * taken from block_truncate_page, but does cow as it zeros out
1546 * any bytes left in the last page in the file.
1548 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
1550 struct inode *inode = mapping->host;
1551 struct btrfs_root *root = BTRFS_I(inode)->root;
1552 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1553 struct btrfs_ordered_extent *ordered;
1555 u32 blocksize = root->sectorsize;
1556 pgoff_t index = from >> PAGE_CACHE_SHIFT;
1557 unsigned offset = from & (PAGE_CACHE_SIZE-1);
1563 if ((offset & (blocksize - 1)) == 0)
1568 page = grab_cache_page(mapping, index);
1572 page_start = page_offset(page);
1573 page_end = page_start + PAGE_CACHE_SIZE - 1;
1575 if (!PageUptodate(page)) {
1576 ret = btrfs_readpage(NULL, page);
1578 if (page->mapping != mapping) {
1580 page_cache_release(page);
1583 if (!PageUptodate(page)) {
1588 wait_on_page_writeback(page);
1590 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
1591 set_page_extent_mapped(page);
1593 ordered = btrfs_lookup_ordered_extent(inode, page_start);
1595 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
1597 page_cache_release(page);
1598 btrfs_start_ordered_extent(inode, ordered, 1);
1599 btrfs_put_ordered_extent(ordered);
1603 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
1604 page_end, GFP_NOFS);
1606 if (offset != PAGE_CACHE_SIZE) {
1608 memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
1609 flush_dcache_page(page);
1612 ClearPageChecked(page);
1613 set_page_dirty(page);
1614 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
1618 page_cache_release(page);
1623 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
1625 struct inode *inode = dentry->d_inode;
1628 err = inode_change_ok(inode, attr);
1632 if (S_ISREG(inode->i_mode) &&
1633 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
1634 struct btrfs_trans_handle *trans;
1635 struct btrfs_root *root = BTRFS_I(inode)->root;
1636 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1638 u64 mask = root->sectorsize - 1;
1639 u64 hole_start = (inode->i_size + mask) & ~mask;
1640 u64 block_end = (attr->ia_size + mask) & ~mask;
1644 if (attr->ia_size <= hole_start)
1647 err = btrfs_check_free_space(root, 1, 0);
1651 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1653 hole_size = block_end - hole_start;
1654 btrfs_wait_ordered_range(inode, hole_start, hole_size);
1655 lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1657 trans = btrfs_start_transaction(root, 1);
1658 btrfs_set_trans_block_group(trans, inode);
1659 mutex_lock(&BTRFS_I(inode)->extent_mutex);
1660 err = btrfs_drop_extents(trans, root, inode,
1661 hole_start, block_end, hole_start,
1664 if (alloc_hint != EXTENT_MAP_INLINE) {
1665 err = btrfs_insert_file_extent(trans, root,
1669 btrfs_drop_extent_cache(inode, hole_start,
1671 btrfs_check_file(root, inode);
1673 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
1674 btrfs_end_transaction(trans, root);
1675 unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1680 err = inode_setattr(inode, attr);
1682 if (!err && ((attr->ia_valid & ATTR_MODE)))
1683 err = btrfs_acl_chmod(inode);
1688 void btrfs_delete_inode(struct inode *inode)
1690 struct btrfs_trans_handle *trans;
1691 struct btrfs_root *root = BTRFS_I(inode)->root;
1695 truncate_inode_pages(&inode->i_data, 0);
1696 if (is_bad_inode(inode)) {
1697 btrfs_orphan_del(NULL, inode);
1700 btrfs_wait_ordered_range(inode, 0, (u64)-1);
1702 btrfs_i_size_write(inode, 0);
1703 trans = btrfs_start_transaction(root, 1);
1705 btrfs_set_trans_block_group(trans, inode);
1706 ret = btrfs_truncate_in_trans(trans, root, inode, 0);
1708 btrfs_orphan_del(NULL, inode);
1709 goto no_delete_lock;
1712 btrfs_orphan_del(trans, inode);
1714 nr = trans->blocks_used;
1717 btrfs_end_transaction(trans, root);
1718 btrfs_btree_balance_dirty(root, nr);
1722 nr = trans->blocks_used;
1723 btrfs_end_transaction(trans, root);
1724 btrfs_btree_balance_dirty(root, nr);
1730 * this returns the key found in the dir entry in the location pointer.
1731 * If no dir entries were found, location->objectid is 0.
1733 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
1734 struct btrfs_key *location)
1736 const char *name = dentry->d_name.name;
1737 int namelen = dentry->d_name.len;
1738 struct btrfs_dir_item *di;
1739 struct btrfs_path *path;
1740 struct btrfs_root *root = BTRFS_I(dir)->root;
1743 if (namelen == 1 && strcmp(name, ".") == 0) {
1744 location->objectid = dir->i_ino;
1745 location->type = BTRFS_INODE_ITEM_KEY;
1746 location->offset = 0;
1749 path = btrfs_alloc_path();
1752 if (namelen == 2 && strcmp(name, "..") == 0) {
1753 struct btrfs_key key;
1754 struct extent_buffer *leaf;
1758 key.objectid = dir->i_ino;
1759 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1761 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1765 leaf = path->nodes[0];
1766 slot = path->slots[0];
1767 nritems = btrfs_header_nritems(leaf);
1768 if (slot >= nritems)
1771 btrfs_item_key_to_cpu(leaf, &key, slot);
1772 if (key.objectid != dir->i_ino ||
1773 key.type != BTRFS_INODE_REF_KEY) {
1776 location->objectid = key.offset;
1777 location->type = BTRFS_INODE_ITEM_KEY;
1778 location->offset = 0;
1782 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
1786 if (!di || IS_ERR(di)) {
1789 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
1791 btrfs_free_path(path);
1794 location->objectid = 0;
1799 * when we hit a tree root in a directory, the btrfs part of the inode
1800 * needs to be changed to reflect the root directory of the tree root. This
1801 * is kind of like crossing a mount point.
1803 static int fixup_tree_root_location(struct btrfs_root *root,
1804 struct btrfs_key *location,
1805 struct btrfs_root **sub_root,
1806 struct dentry *dentry)
1808 struct btrfs_path *path;
1809 struct btrfs_root_item *ri;
1811 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
1813 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1816 path = btrfs_alloc_path();
1819 *sub_root = btrfs_read_fs_root(root->fs_info, location,
1820 dentry->d_name.name,
1821 dentry->d_name.len);
1822 if (IS_ERR(*sub_root))
1823 return PTR_ERR(*sub_root);
1825 ri = &(*sub_root)->root_item;
1826 location->objectid = btrfs_root_dirid(ri);
1827 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1828 location->offset = 0;
1830 btrfs_free_path(path);
1834 static int btrfs_init_locked_inode(struct inode *inode, void *p)
1836 struct btrfs_iget_args *args = p;
1837 inode->i_ino = args->ino;
1838 BTRFS_I(inode)->root = args->root;
1839 BTRFS_I(inode)->delalloc_bytes = 0;
1840 BTRFS_I(inode)->disk_i_size = 0;
1841 BTRFS_I(inode)->index_cnt = (u64)-1;
1842 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1843 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1844 inode->i_mapping, GFP_NOFS);
1845 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1846 inode->i_mapping, GFP_NOFS);
1847 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
1848 mutex_init(&BTRFS_I(inode)->csum_mutex);
1849 mutex_init(&BTRFS_I(inode)->extent_mutex);
1853 static int btrfs_find_actor(struct inode *inode, void *opaque)
1855 struct btrfs_iget_args *args = opaque;
1856 return (args->ino == inode->i_ino &&
1857 args->root == BTRFS_I(inode)->root);
1860 struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
1863 struct btrfs_iget_args args;
1864 args.ino = objectid;
1865 args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
1870 return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
1873 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
1874 struct btrfs_root *root)
1876 struct inode *inode;
1877 struct btrfs_iget_args args;
1878 args.ino = objectid;
1881 inode = iget5_locked(s, objectid, btrfs_find_actor,
1882 btrfs_init_locked_inode,
1887 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
1888 struct nameidata *nd)
1890 struct inode * inode;
1891 struct btrfs_inode *bi = BTRFS_I(dir);
1892 struct btrfs_root *root = bi->root;
1893 struct btrfs_root *sub_root = root;
1894 struct btrfs_key location;
1895 int ret, do_orphan = 0;
1897 if (dentry->d_name.len > BTRFS_NAME_LEN)
1898 return ERR_PTR(-ENAMETOOLONG);
1900 ret = btrfs_inode_by_name(dir, dentry, &location);
1903 return ERR_PTR(ret);
1906 if (location.objectid) {
1907 ret = fixup_tree_root_location(root, &location, &sub_root,
1910 return ERR_PTR(ret);
1912 return ERR_PTR(-ENOENT);
1914 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
1917 return ERR_PTR(-EACCES);
1918 if (inode->i_state & I_NEW) {
1919 /* the inode and parent dir are two different roots */
1920 if (sub_root != root) {
1922 sub_root->inode = inode;
1925 BTRFS_I(inode)->root = sub_root;
1926 memcpy(&BTRFS_I(inode)->location, &location,
1928 btrfs_read_locked_inode(inode);
1929 unlock_new_inode(inode);
1933 if (unlikely(do_orphan))
1934 btrfs_orphan_cleanup(sub_root);
1936 return d_splice_alias(inode, dentry);
1939 static unsigned char btrfs_filetype_table[] = {
1940 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1943 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1945 struct inode *inode = filp->f_dentry->d_inode;
1946 struct btrfs_root *root = BTRFS_I(inode)->root;
1947 struct btrfs_item *item;
1948 struct btrfs_dir_item *di;
1949 struct btrfs_key key;
1950 struct btrfs_key found_key;
1951 struct btrfs_path *path;
1954 struct extent_buffer *leaf;
1957 unsigned char d_type;
1962 int key_type = BTRFS_DIR_INDEX_KEY;
1967 /* FIXME, use a real flag for deciding about the key type */
1968 if (root->fs_info->tree_root == root)
1969 key_type = BTRFS_DIR_ITEM_KEY;
1971 /* special case for "." */
1972 if (filp->f_pos == 0) {
1973 over = filldir(dirent, ".", 1,
1981 key.objectid = inode->i_ino;
1982 path = btrfs_alloc_path();
1985 /* special case for .., just use the back ref */
1986 if (filp->f_pos == 1) {
1987 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1989 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1991 leaf = path->nodes[0];
1992 slot = path->slots[0];
1993 nritems = btrfs_header_nritems(leaf);
1994 if (slot >= nritems) {
1995 btrfs_release_path(root, path);
1996 goto read_dir_items;
1998 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1999 btrfs_release_path(root, path);
2000 if (found_key.objectid != key.objectid ||
2001 found_key.type != BTRFS_INODE_REF_KEY)
2002 goto read_dir_items;
2003 over = filldir(dirent, "..", 2,
2004 2, found_key.offset, DT_DIR);
2011 btrfs_set_key_type(&key, key_type);
2012 key.offset = filp->f_pos;
2014 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2019 leaf = path->nodes[0];
2020 nritems = btrfs_header_nritems(leaf);
2021 slot = path->slots[0];
2022 if (advance || slot >= nritems) {
2023 if (slot >= nritems -1) {
2024 ret = btrfs_next_leaf(root, path);
2027 leaf = path->nodes[0];
2028 nritems = btrfs_header_nritems(leaf);
2029 slot = path->slots[0];
2036 item = btrfs_item_nr(leaf, slot);
2037 btrfs_item_key_to_cpu(leaf, &found_key, slot);
2039 if (found_key.objectid != key.objectid)
2041 if (btrfs_key_type(&found_key) != key_type)
2043 if (found_key.offset < filp->f_pos)
2046 filp->f_pos = found_key.offset;
2048 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
2050 di_total = btrfs_item_size(leaf, item);
2051 while(di_cur < di_total) {
2052 struct btrfs_key location;
2054 name_len = btrfs_dir_name_len(leaf, di);
2055 if (name_len < 32) {
2056 name_ptr = tmp_name;
2058 name_ptr = kmalloc(name_len, GFP_NOFS);
2061 read_extent_buffer(leaf, name_ptr,
2062 (unsigned long)(di + 1), name_len);
2064 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
2065 btrfs_dir_item_key_to_cpu(leaf, di, &location);
2066 over = filldir(dirent, name_ptr, name_len,
2071 if (name_ptr != tmp_name)
2076 di_len = btrfs_dir_name_len(leaf, di) +
2077 btrfs_dir_data_len(leaf, di) +sizeof(*di);
2079 di = (struct btrfs_dir_item *)((char *)di + di_len);
2082 if (key_type == BTRFS_DIR_INDEX_KEY)
2083 filp->f_pos = INT_LIMIT(typeof(filp->f_pos));
2089 btrfs_free_path(path);
2093 int btrfs_write_inode(struct inode *inode, int wait)
2095 struct btrfs_root *root = BTRFS_I(inode)->root;
2096 struct btrfs_trans_handle *trans;
2100 trans = btrfs_join_transaction(root, 1);
2101 btrfs_set_trans_block_group(trans, inode);
2102 ret = btrfs_commit_transaction(trans, root);
2108 * This is somewhat expensive, updating the tree every time the
2109 * inode changes. But, it is most likely to find the inode in cache.
2110 * FIXME, needs more benchmarking...there are no reasons other than performance
2111 * to keep or drop this code.
2113 void btrfs_dirty_inode(struct inode *inode)
2115 struct btrfs_root *root = BTRFS_I(inode)->root;
2116 struct btrfs_trans_handle *trans;
2118 trans = btrfs_join_transaction(root, 1);
2119 btrfs_set_trans_block_group(trans, inode);
2120 btrfs_update_inode(trans, root, inode);
2121 btrfs_end_transaction(trans, root);
2124 static int btrfs_set_inode_index_count(struct inode *inode)
2126 struct btrfs_root *root = BTRFS_I(inode)->root;
2127 struct btrfs_key key, found_key;
2128 struct btrfs_path *path;
2129 struct extent_buffer *leaf;
2132 key.objectid = inode->i_ino;
2133 btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY);
2134 key.offset = (u64)-1;
2136 path = btrfs_alloc_path();
2140 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2143 /* FIXME: we should be able to handle this */
2149 * MAGIC NUMBER EXPLANATION:
2150 * since we search a directory based on f_pos we have to start at 2
2151 * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody
2152 * else has to start at 2
2154 if (path->slots[0] == 0) {
2155 BTRFS_I(inode)->index_cnt = 2;
2161 leaf = path->nodes[0];
2162 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2164 if (found_key.objectid != inode->i_ino ||
2165 btrfs_key_type(&found_key) != BTRFS_DIR_INDEX_KEY) {
2166 BTRFS_I(inode)->index_cnt = 2;
2170 BTRFS_I(inode)->index_cnt = found_key.offset + 1;
2172 btrfs_free_path(path);
2176 static int btrfs_set_inode_index(struct inode *dir, struct inode *inode)
2180 if (BTRFS_I(dir)->index_cnt == (u64)-1) {
2181 ret = btrfs_set_inode_index_count(dir);
2186 BTRFS_I(inode)->index = BTRFS_I(dir)->index_cnt;
2187 BTRFS_I(dir)->index_cnt++;
2192 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
2193 struct btrfs_root *root,
2195 const char *name, int name_len,
2198 struct btrfs_block_group_cache *group,
2201 struct inode *inode;
2202 struct btrfs_inode_item *inode_item;
2203 struct btrfs_block_group_cache *new_inode_group;
2204 struct btrfs_key *location;
2205 struct btrfs_path *path;
2206 struct btrfs_inode_ref *ref;
2207 struct btrfs_key key[2];
2213 path = btrfs_alloc_path();
2216 inode = new_inode(root->fs_info->sb);
2218 return ERR_PTR(-ENOMEM);
2221 ret = btrfs_set_inode_index(dir, inode);
2223 return ERR_PTR(ret);
2225 BTRFS_I(inode)->index = 0;
2228 * index_cnt is ignored for everything but a dir,
2229 * btrfs_get_inode_index_count has an explanation for the magic
2232 BTRFS_I(inode)->index_cnt = 2;
2234 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2235 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
2236 inode->i_mapping, GFP_NOFS);
2237 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
2238 inode->i_mapping, GFP_NOFS);
2239 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
2240 mutex_init(&BTRFS_I(inode)->csum_mutex);
2241 mutex_init(&BTRFS_I(inode)->extent_mutex);
2242 BTRFS_I(inode)->delalloc_bytes = 0;
2243 BTRFS_I(inode)->disk_i_size = 0;
2244 BTRFS_I(inode)->root = root;
2250 new_inode_group = btrfs_find_block_group(root, group, 0,
2251 BTRFS_BLOCK_GROUP_METADATA, owner);
2252 if (!new_inode_group) {
2253 printk("find_block group failed\n");
2254 new_inode_group = group;
2256 BTRFS_I(inode)->block_group = new_inode_group;
2257 BTRFS_I(inode)->flags = 0;
2259 key[0].objectid = objectid;
2260 btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
2263 key[1].objectid = objectid;
2264 btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
2265 key[1].offset = ref_objectid;
2267 sizes[0] = sizeof(struct btrfs_inode_item);
2268 sizes[1] = name_len + sizeof(*ref);
2270 ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
2274 if (objectid > root->highest_inode)
2275 root->highest_inode = objectid;
2277 inode->i_uid = current->fsuid;
2278 inode->i_gid = current->fsgid;
2279 inode->i_mode = mode;
2280 inode->i_ino = objectid;
2281 inode->i_blocks = 0;
2282 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2283 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
2284 struct btrfs_inode_item);
2285 fill_inode_item(path->nodes[0], inode_item, inode);
2287 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
2288 struct btrfs_inode_ref);
2289 btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
2290 btrfs_set_inode_ref_index(path->nodes[0], ref, BTRFS_I(inode)->index);
2291 ptr = (unsigned long)(ref + 1);
2292 write_extent_buffer(path->nodes[0], name, ptr, name_len);
2294 btrfs_mark_buffer_dirty(path->nodes[0]);
2295 btrfs_free_path(path);
2297 location = &BTRFS_I(inode)->location;
2298 location->objectid = objectid;
2299 location->offset = 0;
2300 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
2302 insert_inode_hash(inode);
2306 BTRFS_I(dir)->index_cnt--;
2307 btrfs_free_path(path);
2308 return ERR_PTR(ret);
2311 static inline u8 btrfs_inode_type(struct inode *inode)
2313 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
2316 static int btrfs_add_link(struct btrfs_trans_handle *trans,
2317 struct dentry *dentry, struct inode *inode,
2321 struct btrfs_key key;
2322 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
2323 struct inode *parent_inode = dentry->d_parent->d_inode;
2325 key.objectid = inode->i_ino;
2326 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
2329 ret = btrfs_insert_dir_item(trans, root,
2330 dentry->d_name.name, dentry->d_name.len,
2331 dentry->d_parent->d_inode->i_ino,
2332 &key, btrfs_inode_type(inode),
2333 BTRFS_I(inode)->index);
2336 ret = btrfs_insert_inode_ref(trans, root,
2337 dentry->d_name.name,
2340 parent_inode->i_ino,
2341 BTRFS_I(inode)->index);
2343 btrfs_i_size_write(parent_inode, parent_inode->i_size +
2344 dentry->d_name.len * 2);
2345 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
2346 ret = btrfs_update_inode(trans, root,
2347 dentry->d_parent->d_inode);
2352 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
2353 struct dentry *dentry, struct inode *inode,
2356 int err = btrfs_add_link(trans, dentry, inode, backref);
2358 d_instantiate(dentry, inode);
2366 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
2367 int mode, dev_t rdev)
2369 struct btrfs_trans_handle *trans;
2370 struct btrfs_root *root = BTRFS_I(dir)->root;
2371 struct inode *inode = NULL;
2375 unsigned long nr = 0;
2377 if (!new_valid_dev(rdev))
2380 err = btrfs_check_free_space(root, 1, 0);
2384 trans = btrfs_start_transaction(root, 1);
2385 btrfs_set_trans_block_group(trans, dir);
2387 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2393 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
2395 dentry->d_parent->d_inode->i_ino, objectid,
2396 BTRFS_I(dir)->block_group, mode);
2397 err = PTR_ERR(inode);
2401 err = btrfs_init_acl(inode, dir);
2407 btrfs_set_trans_block_group(trans, inode);
2408 err = btrfs_add_nondir(trans, dentry, inode, 0);
2412 inode->i_op = &btrfs_special_inode_operations;
2413 init_special_inode(inode, inode->i_mode, rdev);
2414 btrfs_update_inode(trans, root, inode);
2416 dir->i_sb->s_dirt = 1;
2417 btrfs_update_inode_block_group(trans, inode);
2418 btrfs_update_inode_block_group(trans, dir);
2420 nr = trans->blocks_used;
2421 btrfs_end_transaction_throttle(trans, root);
2424 inode_dec_link_count(inode);
2427 btrfs_btree_balance_dirty(root, nr);
2431 static int btrfs_create(struct inode *dir, struct dentry *dentry,
2432 int mode, struct nameidata *nd)
2434 struct btrfs_trans_handle *trans;
2435 struct btrfs_root *root = BTRFS_I(dir)->root;
2436 struct inode *inode = NULL;
2439 unsigned long nr = 0;
2442 err = btrfs_check_free_space(root, 1, 0);
2445 trans = btrfs_start_transaction(root, 1);
2446 btrfs_set_trans_block_group(trans, dir);
2448 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2454 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
2456 dentry->d_parent->d_inode->i_ino,
2457 objectid, BTRFS_I(dir)->block_group, mode);
2458 err = PTR_ERR(inode);
2462 err = btrfs_init_acl(inode, dir);
2468 btrfs_set_trans_block_group(trans, inode);
2469 err = btrfs_add_nondir(trans, dentry, inode, 0);
2473 inode->i_mapping->a_ops = &btrfs_aops;
2474 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
2475 inode->i_fop = &btrfs_file_operations;
2476 inode->i_op = &btrfs_file_inode_operations;
2477 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2478 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
2479 inode->i_mapping, GFP_NOFS);
2480 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
2481 inode->i_mapping, GFP_NOFS);
2482 mutex_init(&BTRFS_I(inode)->csum_mutex);
2483 mutex_init(&BTRFS_I(inode)->extent_mutex);
2484 BTRFS_I(inode)->delalloc_bytes = 0;
2485 BTRFS_I(inode)->disk_i_size = 0;
2486 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
2487 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
2489 dir->i_sb->s_dirt = 1;
2490 btrfs_update_inode_block_group(trans, inode);
2491 btrfs_update_inode_block_group(trans, dir);
2493 nr = trans->blocks_used;
2494 btrfs_end_transaction_throttle(trans, root);
2497 inode_dec_link_count(inode);
2500 btrfs_btree_balance_dirty(root, nr);
2504 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
2505 struct dentry *dentry)
2507 struct btrfs_trans_handle *trans;
2508 struct btrfs_root *root = BTRFS_I(dir)->root;
2509 struct inode *inode = old_dentry->d_inode;
2510 unsigned long nr = 0;
2514 if (inode->i_nlink == 0)
2517 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2522 err = btrfs_check_free_space(root, 1, 0);
2525 err = btrfs_set_inode_index(dir, inode);
2529 trans = btrfs_start_transaction(root, 1);
2531 btrfs_set_trans_block_group(trans, dir);
2532 atomic_inc(&inode->i_count);
2534 err = btrfs_add_nondir(trans, dentry, inode, 1);
2539 dir->i_sb->s_dirt = 1;
2540 btrfs_update_inode_block_group(trans, dir);
2541 err = btrfs_update_inode(trans, root, inode);
2546 nr = trans->blocks_used;
2547 btrfs_end_transaction_throttle(trans, root);
2550 inode_dec_link_count(inode);
2553 btrfs_btree_balance_dirty(root, nr);
2557 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
2559 struct inode *inode = NULL;
2560 struct btrfs_trans_handle *trans;
2561 struct btrfs_root *root = BTRFS_I(dir)->root;
2563 int drop_on_err = 0;
2565 unsigned long nr = 1;
2567 err = btrfs_check_free_space(root, 1, 0);
2571 trans = btrfs_start_transaction(root, 1);
2572 btrfs_set_trans_block_group(trans, dir);
2574 if (IS_ERR(trans)) {
2575 err = PTR_ERR(trans);
2579 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2585 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
2587 dentry->d_parent->d_inode->i_ino, objectid,
2588 BTRFS_I(dir)->block_group, S_IFDIR | mode);
2589 if (IS_ERR(inode)) {
2590 err = PTR_ERR(inode);
2596 err = btrfs_init_acl(inode, dir);
2600 inode->i_op = &btrfs_dir_inode_operations;
2601 inode->i_fop = &btrfs_dir_file_operations;
2602 btrfs_set_trans_block_group(trans, inode);
2604 btrfs_i_size_write(inode, 0);
2605 err = btrfs_update_inode(trans, root, inode);
2609 err = btrfs_add_link(trans, dentry, inode, 0);
2613 d_instantiate(dentry, inode);
2615 dir->i_sb->s_dirt = 1;
2616 btrfs_update_inode_block_group(trans, inode);
2617 btrfs_update_inode_block_group(trans, dir);
2620 nr = trans->blocks_used;
2621 btrfs_end_transaction_throttle(trans, root);
2626 btrfs_btree_balance_dirty(root, nr);
2630 static int merge_extent_mapping(struct extent_map_tree *em_tree,
2631 struct extent_map *existing,
2632 struct extent_map *em,
2633 u64 map_start, u64 map_len)
2637 BUG_ON(map_start < em->start || map_start >= extent_map_end(em));
2638 start_diff = map_start - em->start;
2639 em->start = map_start;
2641 if (em->block_start < EXTENT_MAP_LAST_BYTE)
2642 em->block_start += start_diff;
2643 return add_extent_mapping(em_tree, em);
2646 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
2647 size_t pg_offset, u64 start, u64 len,
2653 u64 extent_start = 0;
2655 u64 objectid = inode->i_ino;
2657 struct btrfs_path *path = NULL;
2658 struct btrfs_root *root = BTRFS_I(inode)->root;
2659 struct btrfs_file_extent_item *item;
2660 struct extent_buffer *leaf;
2661 struct btrfs_key found_key;
2662 struct extent_map *em = NULL;
2663 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2664 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2665 struct btrfs_trans_handle *trans = NULL;
2668 spin_lock(&em_tree->lock);
2669 em = lookup_extent_mapping(em_tree, start, len);
2671 em->bdev = root->fs_info->fs_devices->latest_bdev;
2672 spin_unlock(&em_tree->lock);
2675 if (em->start > start || em->start + em->len <= start)
2676 free_extent_map(em);
2677 else if (em->block_start == EXTENT_MAP_INLINE && page)
2678 free_extent_map(em);
2682 em = alloc_extent_map(GFP_NOFS);
2687 em->bdev = root->fs_info->fs_devices->latest_bdev;
2688 em->start = EXTENT_MAP_HOLE;
2692 path = btrfs_alloc_path();
2696 ret = btrfs_lookup_file_extent(trans, root, path,
2697 objectid, start, trans != NULL);
2704 if (path->slots[0] == 0)
2709 leaf = path->nodes[0];
2710 item = btrfs_item_ptr(leaf, path->slots[0],
2711 struct btrfs_file_extent_item);
2712 /* are we inside the extent that was found? */
2713 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2714 found_type = btrfs_key_type(&found_key);
2715 if (found_key.objectid != objectid ||
2716 found_type != BTRFS_EXTENT_DATA_KEY) {
2720 found_type = btrfs_file_extent_type(leaf, item);
2721 extent_start = found_key.offset;
2722 if (found_type == BTRFS_FILE_EXTENT_REG) {
2723 extent_end = extent_start +
2724 btrfs_file_extent_num_bytes(leaf, item);
2726 if (start < extent_start || start >= extent_end) {
2728 if (start < extent_start) {
2729 if (start + len <= extent_start)
2731 em->len = extent_end - extent_start;
2737 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
2739 em->start = extent_start;
2740 em->len = extent_end - extent_start;
2741 em->block_start = EXTENT_MAP_HOLE;
2744 bytenr += btrfs_file_extent_offset(leaf, item);
2745 em->block_start = bytenr;
2746 em->start = extent_start;
2747 em->len = extent_end - extent_start;
2749 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
2754 size_t extent_offset;
2757 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
2759 extent_end = (extent_start + size + root->sectorsize - 1) &
2760 ~((u64)root->sectorsize - 1);
2761 if (start < extent_start || start >= extent_end) {
2763 if (start < extent_start) {
2764 if (start + len <= extent_start)
2766 em->len = extent_end - extent_start;
2772 em->block_start = EXTENT_MAP_INLINE;
2775 em->start = extent_start;
2780 page_start = page_offset(page) + pg_offset;
2781 extent_offset = page_start - extent_start;
2782 copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
2783 size - extent_offset);
2784 em->start = extent_start + extent_offset;
2785 em->len = (copy_size + root->sectorsize - 1) &
2786 ~((u64)root->sectorsize - 1);
2788 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
2789 if (create == 0 && !PageUptodate(page)) {
2790 read_extent_buffer(leaf, map + pg_offset, ptr,
2792 flush_dcache_page(page);
2793 } else if (create && PageUptodate(page)) {
2796 free_extent_map(em);
2798 btrfs_release_path(root, path);
2799 trans = btrfs_join_transaction(root, 1);
2802 write_extent_buffer(leaf, map + pg_offset, ptr,
2804 btrfs_mark_buffer_dirty(leaf);
2807 set_extent_uptodate(io_tree, em->start,
2808 extent_map_end(em) - 1, GFP_NOFS);
2811 printk("unkknown found_type %d\n", found_type);
2818 em->block_start = EXTENT_MAP_HOLE;
2820 btrfs_release_path(root, path);
2821 if (em->start > start || extent_map_end(em) <= start) {
2822 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len);
2828 spin_lock(&em_tree->lock);
2829 ret = add_extent_mapping(em_tree, em);
2830 /* it is possible that someone inserted the extent into the tree
2831 * while we had the lock dropped. It is also possible that
2832 * an overlapping map exists in the tree
2834 if (ret == -EEXIST) {
2835 struct extent_map *existing;
2839 existing = lookup_extent_mapping(em_tree, start, len);
2840 if (existing && (existing->start > start ||
2841 existing->start + existing->len <= start)) {
2842 free_extent_map(existing);
2846 existing = lookup_extent_mapping(em_tree, em->start,
2849 err = merge_extent_mapping(em_tree, existing,
2852 free_extent_map(existing);
2854 free_extent_map(em);
2859 printk("failing to insert %Lu %Lu\n",
2861 free_extent_map(em);
2865 free_extent_map(em);
2870 spin_unlock(&em_tree->lock);
2873 btrfs_free_path(path);
2875 ret = btrfs_end_transaction(trans, root);
2881 free_extent_map(em);
2883 return ERR_PTR(err);
2888 #if 0 /* waiting for O_DIRECT reads */
2889 static int btrfs_get_block(struct inode *inode, sector_t iblock,
2890 struct buffer_head *bh_result, int create)
2892 struct extent_map *em;
2893 u64 start = (u64)iblock << inode->i_blkbits;
2894 struct btrfs_multi_bio *multi = NULL;
2895 struct btrfs_root *root = BTRFS_I(inode)->root;
2901 em = btrfs_get_extent(inode, NULL, 0, start, bh_result->b_size, 0);
2903 if (!em || IS_ERR(em))
2906 if (em->start > start || em->start + em->len <= start) {
2910 if (em->block_start == EXTENT_MAP_INLINE) {
2915 len = em->start + em->len - start;
2916 len = min_t(u64, len, INT_LIMIT(typeof(bh_result->b_size)));
2918 if (em->block_start == EXTENT_MAP_HOLE ||
2919 em->block_start == EXTENT_MAP_DELALLOC) {
2920 bh_result->b_size = len;
2924 logical = start - em->start;
2925 logical = em->block_start + logical;
2928 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
2929 logical, &map_length, &multi, 0);
2931 bh_result->b_blocknr = multi->stripes[0].physical >> inode->i_blkbits;
2932 bh_result->b_size = min(map_length, len);
2934 bh_result->b_bdev = multi->stripes[0].dev->bdev;
2935 set_buffer_mapped(bh_result);
2938 free_extent_map(em);
2943 static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
2944 const struct iovec *iov, loff_t offset,
2945 unsigned long nr_segs)
2949 struct file *file = iocb->ki_filp;
2950 struct inode *inode = file->f_mapping->host;
2955 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2956 offset, nr_segs, btrfs_get_block, NULL);
2960 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
2962 return extent_bmap(mapping, iblock, btrfs_get_extent);
2965 int btrfs_readpage(struct file *file, struct page *page)
2967 struct extent_io_tree *tree;
2968 tree = &BTRFS_I(page->mapping->host)->io_tree;
2969 return extent_read_full_page(tree, page, btrfs_get_extent);
2972 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
2974 struct extent_io_tree *tree;
2977 if (current->flags & PF_MEMALLOC) {
2978 redirty_page_for_writepage(wbc, page);
2982 tree = &BTRFS_I(page->mapping->host)->io_tree;
2983 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
2986 int btrfs_writepages(struct address_space *mapping,
2987 struct writeback_control *wbc)
2989 struct extent_io_tree *tree;
2990 tree = &BTRFS_I(mapping->host)->io_tree;
2991 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
2995 btrfs_readpages(struct file *file, struct address_space *mapping,
2996 struct list_head *pages, unsigned nr_pages)
2998 struct extent_io_tree *tree;
2999 tree = &BTRFS_I(mapping->host)->io_tree;
3000 return extent_readpages(tree, mapping, pages, nr_pages,
3003 static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
3005 struct extent_io_tree *tree;
3006 struct extent_map_tree *map;
3009 tree = &BTRFS_I(page->mapping->host)->io_tree;
3010 map = &BTRFS_I(page->mapping->host)->extent_tree;
3011 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
3013 ClearPagePrivate(page);
3014 set_page_private(page, 0);
3015 page_cache_release(page);
3020 static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
3022 return __btrfs_releasepage(page, gfp_flags);
3025 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
3027 struct extent_io_tree *tree;
3028 struct btrfs_ordered_extent *ordered;
3029 u64 page_start = page_offset(page);
3030 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
3032 wait_on_page_writeback(page);
3033 tree = &BTRFS_I(page->mapping->host)->io_tree;
3035 btrfs_releasepage(page, GFP_NOFS);
3039 lock_extent(tree, page_start, page_end, GFP_NOFS);
3040 ordered = btrfs_lookup_ordered_extent(page->mapping->host,
3044 * IO on this page will never be started, so we need
3045 * to account for any ordered extents now
3047 clear_extent_bit(tree, page_start, page_end,
3048 EXTENT_DIRTY | EXTENT_DELALLOC |
3049 EXTENT_LOCKED, 1, 0, GFP_NOFS);
3050 btrfs_finish_ordered_io(page->mapping->host,
3051 page_start, page_end);
3052 btrfs_put_ordered_extent(ordered);
3053 lock_extent(tree, page_start, page_end, GFP_NOFS);
3055 clear_extent_bit(tree, page_start, page_end,
3056 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
3059 __btrfs_releasepage(page, GFP_NOFS);
3061 ClearPageChecked(page);
3062 if (PagePrivate(page)) {
3063 ClearPagePrivate(page);
3064 set_page_private(page, 0);
3065 page_cache_release(page);
3070 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
3071 * called from a page fault handler when a page is first dirtied. Hence we must
3072 * be careful to check for EOF conditions here. We set the page up correctly
3073 * for a written page which means we get ENOSPC checking when writing into
3074 * holes and correct delalloc and unwritten extent mapping on filesystems that
3075 * support these features.
3077 * We are not allowed to take the i_mutex here so we have to play games to
3078 * protect against truncate races as the page could now be beyond EOF. Because
3079 * vmtruncate() writes the inode size before removing pages, once we have the
3080 * page lock we can determine safely if the page is beyond EOF. If it is not
3081 * beyond EOF, then the page is guaranteed safe against truncation until we
3084 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
3086 struct inode *inode = fdentry(vma->vm_file)->d_inode;
3087 struct btrfs_root *root = BTRFS_I(inode)->root;
3088 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
3089 struct btrfs_ordered_extent *ordered;
3091 unsigned long zero_start;
3097 ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
3104 size = i_size_read(inode);
3105 page_start = page_offset(page);
3106 page_end = page_start + PAGE_CACHE_SIZE - 1;
3108 if ((page->mapping != inode->i_mapping) ||
3109 (page_start >= size)) {
3110 /* page got truncated out from underneath us */
3113 wait_on_page_writeback(page);
3115 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
3116 set_page_extent_mapped(page);
3119 * we can't set the delalloc bits if there are pending ordered
3120 * extents. Drop our locks and wait for them to finish
3122 ordered = btrfs_lookup_ordered_extent(inode, page_start);
3124 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
3126 btrfs_start_ordered_extent(inode, ordered, 1);
3127 btrfs_put_ordered_extent(ordered);
3131 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
3132 page_end, GFP_NOFS);
3135 /* page is wholly or partially inside EOF */
3136 if (page_start + PAGE_CACHE_SIZE > size)
3137 zero_start = size & ~PAGE_CACHE_MASK;
3139 zero_start = PAGE_CACHE_SIZE;
3141 if (zero_start != PAGE_CACHE_SIZE) {
3143 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
3144 flush_dcache_page(page);
3147 ClearPageChecked(page);
3148 set_page_dirty(page);
3149 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
3157 static void btrfs_truncate(struct inode *inode)
3159 struct btrfs_root *root = BTRFS_I(inode)->root;
3161 struct btrfs_trans_handle *trans;
3163 u64 mask = root->sectorsize - 1;
3165 if (!S_ISREG(inode->i_mode))
3167 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
3170 btrfs_truncate_page(inode->i_mapping, inode->i_size);
3171 btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
3173 trans = btrfs_start_transaction(root, 1);
3174 btrfs_set_trans_block_group(trans, inode);
3175 btrfs_i_size_write(inode, inode->i_size);
3177 ret = btrfs_orphan_add(trans, inode);
3180 /* FIXME, add redo link to tree so we don't leak on crash */
3181 ret = btrfs_truncate_in_trans(trans, root, inode,
3182 BTRFS_EXTENT_DATA_KEY);
3183 btrfs_update_inode(trans, root, inode);
3185 ret = btrfs_orphan_del(trans, inode);
3189 nr = trans->blocks_used;
3190 ret = btrfs_end_transaction_throttle(trans, root);
3192 btrfs_btree_balance_dirty(root, nr);
3196 * Invalidate a single dcache entry at the root of the filesystem.
3197 * Needed after creation of snapshot or subvolume.
3199 void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name,
3202 struct dentry *alias, *entry;
3205 alias = d_find_alias(root->fs_info->sb->s_root->d_inode);
3209 /* change me if btrfs ever gets a d_hash operation */
3210 qstr.hash = full_name_hash(qstr.name, qstr.len);
3211 entry = d_lookup(alias, &qstr);
3214 d_invalidate(entry);
3220 int btrfs_create_subvol_root(struct btrfs_root *new_root,
3221 struct btrfs_trans_handle *trans, u64 new_dirid,
3222 struct btrfs_block_group_cache *block_group)
3224 struct inode *inode;
3226 inode = btrfs_new_inode(trans, new_root, NULL, "..", 2, new_dirid,
3227 new_dirid, block_group, S_IFDIR | 0700);
3229 return PTR_ERR(inode);
3230 inode->i_op = &btrfs_dir_inode_operations;
3231 inode->i_fop = &btrfs_dir_file_operations;
3232 new_root->inode = inode;
3235 btrfs_i_size_write(inode, 0);
3237 return btrfs_update_inode(trans, new_root, inode);
3240 unsigned long btrfs_force_ra(struct address_space *mapping,
3241 struct file_ra_state *ra, struct file *file,
3242 pgoff_t offset, pgoff_t last_index)
3244 pgoff_t req_size = last_index - offset + 1;
3246 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3247 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
3250 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
3251 return offset + req_size;
3255 struct inode *btrfs_alloc_inode(struct super_block *sb)
3257 struct btrfs_inode *ei;
3259 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
3263 btrfs_ordered_inode_tree_init(&ei->ordered_tree);
3264 ei->i_acl = BTRFS_ACL_NOT_CACHED;
3265 ei->i_default_acl = BTRFS_ACL_NOT_CACHED;
3266 INIT_LIST_HEAD(&ei->i_orphan);
3267 return &ei->vfs_inode;
3270 void btrfs_destroy_inode(struct inode *inode)
3272 struct btrfs_ordered_extent *ordered;
3273 WARN_ON(!list_empty(&inode->i_dentry));
3274 WARN_ON(inode->i_data.nrpages);
3276 if (BTRFS_I(inode)->i_acl &&
3277 BTRFS_I(inode)->i_acl != BTRFS_ACL_NOT_CACHED)
3278 posix_acl_release(BTRFS_I(inode)->i_acl);
3279 if (BTRFS_I(inode)->i_default_acl &&
3280 BTRFS_I(inode)->i_default_acl != BTRFS_ACL_NOT_CACHED)
3281 posix_acl_release(BTRFS_I(inode)->i_default_acl);
3283 spin_lock(&BTRFS_I(inode)->root->orphan_lock);
3284 if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
3285 printk(KERN_ERR "BTRFS: inode %lu: inode still on the orphan"
3286 " list\n", inode->i_ino);
3289 spin_unlock(&BTRFS_I(inode)->root->orphan_lock);
3292 ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
3296 printk("found ordered extent %Lu %Lu\n",
3297 ordered->file_offset, ordered->len);
3298 btrfs_remove_ordered_extent(inode, ordered);
3299 btrfs_put_ordered_extent(ordered);
3300 btrfs_put_ordered_extent(ordered);
3303 btrfs_drop_extent_cache(inode, 0, (u64)-1);
3304 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
3307 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3308 static void init_once(struct kmem_cache * cachep, void *foo)
3310 static void init_once(void * foo, struct kmem_cache * cachep,
3311 unsigned long flags)
3314 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
3316 inode_init_once(&ei->vfs_inode);
3319 void btrfs_destroy_cachep(void)
3321 if (btrfs_inode_cachep)
3322 kmem_cache_destroy(btrfs_inode_cachep);
3323 if (btrfs_trans_handle_cachep)
3324 kmem_cache_destroy(btrfs_trans_handle_cachep);
3325 if (btrfs_transaction_cachep)
3326 kmem_cache_destroy(btrfs_transaction_cachep);
3327 if (btrfs_bit_radix_cachep)
3328 kmem_cache_destroy(btrfs_bit_radix_cachep);
3329 if (btrfs_path_cachep)
3330 kmem_cache_destroy(btrfs_path_cachep);
3333 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
3334 unsigned long extra_flags,
3335 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3336 void (*ctor)(struct kmem_cache *, void *)
3338 void (*ctor)(void *, struct kmem_cache *,
3343 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
3344 SLAB_MEM_SPREAD | extra_flags), ctor
3345 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3351 int btrfs_init_cachep(void)
3353 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
3354 sizeof(struct btrfs_inode),
3356 if (!btrfs_inode_cachep)
3358 btrfs_trans_handle_cachep =
3359 btrfs_cache_create("btrfs_trans_handle_cache",
3360 sizeof(struct btrfs_trans_handle),
3362 if (!btrfs_trans_handle_cachep)
3364 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
3365 sizeof(struct btrfs_transaction),
3367 if (!btrfs_transaction_cachep)
3369 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
3370 sizeof(struct btrfs_path),
3372 if (!btrfs_path_cachep)
3374 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
3375 SLAB_DESTROY_BY_RCU, NULL);
3376 if (!btrfs_bit_radix_cachep)
3380 btrfs_destroy_cachep();
3384 static int btrfs_getattr(struct vfsmount *mnt,
3385 struct dentry *dentry, struct kstat *stat)
3387 struct inode *inode = dentry->d_inode;
3388 generic_fillattr(inode, stat);
3389 stat->blksize = PAGE_CACHE_SIZE;
3390 stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9);
3394 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
3395 struct inode * new_dir,struct dentry *new_dentry)
3397 struct btrfs_trans_handle *trans;
3398 struct btrfs_root *root = BTRFS_I(old_dir)->root;
3399 struct inode *new_inode = new_dentry->d_inode;
3400 struct inode *old_inode = old_dentry->d_inode;
3401 struct timespec ctime = CURRENT_TIME;
3404 if (S_ISDIR(old_inode->i_mode) && new_inode &&
3405 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
3409 ret = btrfs_check_free_space(root, 1, 0);
3413 trans = btrfs_start_transaction(root, 1);
3415 btrfs_set_trans_block_group(trans, new_dir);
3417 old_dentry->d_inode->i_nlink++;
3418 old_dir->i_ctime = old_dir->i_mtime = ctime;
3419 new_dir->i_ctime = new_dir->i_mtime = ctime;
3420 old_inode->i_ctime = ctime;
3422 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
3427 new_inode->i_ctime = CURRENT_TIME;
3428 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
3431 if (new_inode->i_nlink == 0) {
3432 ret = btrfs_orphan_add(trans, new_inode);
3437 ret = btrfs_set_inode_index(new_dir, old_inode);
3441 ret = btrfs_add_link(trans, new_dentry, old_inode, 1);
3446 btrfs_end_transaction(trans, root);
3451 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
3452 const char *symname)
3454 struct btrfs_trans_handle *trans;
3455 struct btrfs_root *root = BTRFS_I(dir)->root;
3456 struct btrfs_path *path;
3457 struct btrfs_key key;
3458 struct inode *inode = NULL;
3465 struct btrfs_file_extent_item *ei;
3466 struct extent_buffer *leaf;
3467 unsigned long nr = 0;
3469 name_len = strlen(symname) + 1;
3470 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
3471 return -ENAMETOOLONG;
3473 err = btrfs_check_free_space(root, 1, 0);
3477 trans = btrfs_start_transaction(root, 1);
3478 btrfs_set_trans_block_group(trans, dir);
3480 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
3486 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
3488 dentry->d_parent->d_inode->i_ino, objectid,
3489 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
3490 err = PTR_ERR(inode);
3494 err = btrfs_init_acl(inode, dir);
3500 btrfs_set_trans_block_group(trans, inode);
3501 err = btrfs_add_nondir(trans, dentry, inode, 0);
3505 inode->i_mapping->a_ops = &btrfs_aops;
3506 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3507 inode->i_fop = &btrfs_file_operations;
3508 inode->i_op = &btrfs_file_inode_operations;
3509 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
3510 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
3511 inode->i_mapping, GFP_NOFS);
3512 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
3513 inode->i_mapping, GFP_NOFS);
3514 mutex_init(&BTRFS_I(inode)->csum_mutex);
3515 mutex_init(&BTRFS_I(inode)->extent_mutex);
3516 BTRFS_I(inode)->delalloc_bytes = 0;
3517 BTRFS_I(inode)->disk_i_size = 0;
3518 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
3519 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
3521 dir->i_sb->s_dirt = 1;
3522 btrfs_update_inode_block_group(trans, inode);
3523 btrfs_update_inode_block_group(trans, dir);
3527 path = btrfs_alloc_path();
3529 key.objectid = inode->i_ino;
3531 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
3532 datasize = btrfs_file_extent_calc_inline_size(name_len);
3533 err = btrfs_insert_empty_item(trans, root, path, &key,
3539 leaf = path->nodes[0];
3540 ei = btrfs_item_ptr(leaf, path->slots[0],
3541 struct btrfs_file_extent_item);
3542 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
3543 btrfs_set_file_extent_type(leaf, ei,
3544 BTRFS_FILE_EXTENT_INLINE);
3545 ptr = btrfs_file_extent_inline_start(ei);
3546 write_extent_buffer(leaf, symname, ptr, name_len);
3547 btrfs_mark_buffer_dirty(leaf);
3548 btrfs_free_path(path);
3550 inode->i_op = &btrfs_symlink_inode_operations;
3551 inode->i_mapping->a_ops = &btrfs_symlink_aops;
3552 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3553 btrfs_i_size_write(inode, name_len - 1);
3554 err = btrfs_update_inode(trans, root, inode);
3559 nr = trans->blocks_used;
3560 btrfs_end_transaction_throttle(trans, root);
3563 inode_dec_link_count(inode);
3566 btrfs_btree_balance_dirty(root, nr);
3570 static int btrfs_set_page_dirty(struct page *page)
3572 return __set_page_dirty_nobuffers(page);
3575 static int btrfs_permission(struct inode *inode, int mask,
3576 struct nameidata *nd)
3578 if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
3580 return generic_permission(inode, mask, btrfs_check_acl);
3583 static struct inode_operations btrfs_dir_inode_operations = {
3584 .lookup = btrfs_lookup,
3585 .create = btrfs_create,
3586 .unlink = btrfs_unlink,
3588 .mkdir = btrfs_mkdir,
3589 .rmdir = btrfs_rmdir,
3590 .rename = btrfs_rename,
3591 .symlink = btrfs_symlink,
3592 .setattr = btrfs_setattr,
3593 .mknod = btrfs_mknod,
3594 .setxattr = generic_setxattr,
3595 .getxattr = generic_getxattr,
3596 .listxattr = btrfs_listxattr,
3597 .removexattr = generic_removexattr,
3598 .permission = btrfs_permission,
3600 static struct inode_operations btrfs_dir_ro_inode_operations = {
3601 .lookup = btrfs_lookup,
3602 .permission = btrfs_permission,
3604 static struct file_operations btrfs_dir_file_operations = {
3605 .llseek = generic_file_llseek,
3606 .read = generic_read_dir,
3607 .readdir = btrfs_readdir,
3608 .unlocked_ioctl = btrfs_ioctl,
3609 #ifdef CONFIG_COMPAT
3610 .compat_ioctl = btrfs_ioctl,
3612 .release = btrfs_release_file,
3615 static struct extent_io_ops btrfs_extent_io_ops = {
3616 .fill_delalloc = run_delalloc_range,
3617 .submit_bio_hook = btrfs_submit_bio_hook,
3618 .merge_bio_hook = btrfs_merge_bio_hook,
3619 .readpage_io_hook = btrfs_readpage_io_hook,
3620 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
3621 .writepage_end_io_hook = btrfs_writepage_end_io_hook,
3622 .writepage_start_hook = btrfs_writepage_start_hook,
3623 .readpage_io_failed_hook = btrfs_io_failed_hook,
3624 .set_bit_hook = btrfs_set_bit_hook,
3625 .clear_bit_hook = btrfs_clear_bit_hook,
3628 static struct address_space_operations btrfs_aops = {
3629 .readpage = btrfs_readpage,
3630 .writepage = btrfs_writepage,
3631 .writepages = btrfs_writepages,
3632 .readpages = btrfs_readpages,
3633 .sync_page = block_sync_page,
3635 .direct_IO = btrfs_direct_IO,
3636 .invalidatepage = btrfs_invalidatepage,
3637 .releasepage = btrfs_releasepage,
3638 .set_page_dirty = btrfs_set_page_dirty,
3641 static struct address_space_operations btrfs_symlink_aops = {
3642 .readpage = btrfs_readpage,
3643 .writepage = btrfs_writepage,
3644 .invalidatepage = btrfs_invalidatepage,
3645 .releasepage = btrfs_releasepage,
3648 static struct inode_operations btrfs_file_inode_operations = {
3649 .truncate = btrfs_truncate,
3650 .getattr = btrfs_getattr,
3651 .setattr = btrfs_setattr,
3652 .setxattr = generic_setxattr,
3653 .getxattr = generic_getxattr,
3654 .listxattr = btrfs_listxattr,
3655 .removexattr = generic_removexattr,
3656 .permission = btrfs_permission,
3658 static struct inode_operations btrfs_special_inode_operations = {
3659 .getattr = btrfs_getattr,
3660 .setattr = btrfs_setattr,
3661 .permission = btrfs_permission,
3662 .setxattr = generic_setxattr,
3663 .getxattr = generic_getxattr,
3664 .listxattr = btrfs_listxattr,
3665 .removexattr = generic_removexattr,
3667 static struct inode_operations btrfs_symlink_inode_operations = {
3668 .readlink = generic_readlink,
3669 .follow_link = page_follow_link_light,
3670 .put_link = page_put_link,
3671 .permission = btrfs_permission,