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>
41 #include "transaction.h"
42 #include "btrfs_inode.h"
44 #include "print-tree.h"
46 #include "ordered-data.h"
48 struct btrfs_iget_args {
50 struct btrfs_root *root;
53 static struct inode_operations btrfs_dir_inode_operations;
54 static struct inode_operations btrfs_symlink_inode_operations;
55 static struct inode_operations btrfs_dir_ro_inode_operations;
56 static struct inode_operations btrfs_special_inode_operations;
57 static struct inode_operations btrfs_file_inode_operations;
58 static struct address_space_operations btrfs_aops;
59 static struct address_space_operations btrfs_symlink_aops;
60 static struct file_operations btrfs_dir_file_operations;
61 static struct extent_io_ops btrfs_extent_io_ops;
63 static struct kmem_cache *btrfs_inode_cachep;
64 struct kmem_cache *btrfs_trans_handle_cachep;
65 struct kmem_cache *btrfs_transaction_cachep;
66 struct kmem_cache *btrfs_bit_radix_cachep;
67 struct kmem_cache *btrfs_path_cachep;
70 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
71 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
72 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
73 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
74 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
75 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
76 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
77 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
80 int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
89 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
90 total = btrfs_super_total_bytes(&root->fs_info->super_copy);
91 used = btrfs_super_bytes_used(&root->fs_info->super_copy);
99 if (used + root->fs_info->delalloc_bytes + num_required > thresh)
101 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
105 static int cow_file_range(struct inode *inode, u64 start, u64 end)
107 struct btrfs_root *root = BTRFS_I(inode)->root;
108 struct btrfs_trans_handle *trans;
112 u64 blocksize = root->sectorsize;
114 struct btrfs_key ins;
115 struct extent_map *em;
116 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
119 trans = btrfs_start_transaction(root, 1);
121 btrfs_set_trans_block_group(trans, inode);
123 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
124 num_bytes = max(blocksize, num_bytes);
125 orig_num_bytes = num_bytes;
127 if (alloc_hint == EXTENT_MAP_INLINE)
130 BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy));
131 btrfs_drop_extent_cache(inode, start, start + num_bytes - 1);
133 while(num_bytes > 0) {
134 cur_alloc_size = min(num_bytes, root->fs_info->max_extent);
135 ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
136 root->sectorsize, 0, 0,
142 em = alloc_extent_map(GFP_NOFS);
144 em->len = ins.offset;
145 em->block_start = ins.objectid;
146 em->bdev = root->fs_info->fs_devices->latest_bdev;
148 spin_lock(&em_tree->lock);
149 ret = add_extent_mapping(em_tree, em);
150 spin_unlock(&em_tree->lock);
151 if (ret != -EEXIST) {
155 btrfs_drop_extent_cache(inode, start,
156 start + ins.offset - 1);
159 cur_alloc_size = ins.offset;
160 ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
163 if (num_bytes < cur_alloc_size) {
164 printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes,
168 num_bytes -= cur_alloc_size;
169 alloc_hint = ins.objectid + ins.offset;
170 start += cur_alloc_size;
173 btrfs_end_transaction(trans, root);
177 static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end)
185 struct btrfs_root *root = BTRFS_I(inode)->root;
186 struct btrfs_block_group_cache *block_group;
187 struct extent_buffer *leaf;
189 struct btrfs_path *path;
190 struct btrfs_file_extent_item *item;
193 struct btrfs_key found_key;
195 total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
196 path = btrfs_alloc_path();
199 ret = btrfs_lookup_file_extent(NULL, root, path,
200 inode->i_ino, start, 0);
202 btrfs_free_path(path);
208 if (path->slots[0] == 0)
213 leaf = path->nodes[0];
214 item = btrfs_item_ptr(leaf, path->slots[0],
215 struct btrfs_file_extent_item);
217 /* are we inside the extent that was found? */
218 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
219 found_type = btrfs_key_type(&found_key);
220 if (found_key.objectid != inode->i_ino ||
221 found_type != BTRFS_EXTENT_DATA_KEY)
224 found_type = btrfs_file_extent_type(leaf, item);
225 extent_start = found_key.offset;
226 if (found_type == BTRFS_FILE_EXTENT_REG) {
227 u64 extent_num_bytes;
229 extent_num_bytes = btrfs_file_extent_num_bytes(leaf, item);
230 extent_end = extent_start + extent_num_bytes;
233 if (loops && start != extent_start)
236 if (start < extent_start || start >= extent_end)
239 cow_end = min(end, extent_end - 1);
240 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
244 if (btrfs_count_snapshots_in_path(root, path, inode->i_ino,
250 * we may be called by the resizer, make sure we're inside
251 * the limits of the FS
253 block_group = btrfs_lookup_block_group(root->fs_info,
255 if (!block_group || block_group->ro)
264 btrfs_free_path(path);
267 btrfs_release_path(root, path);
272 cow_file_range(inode, start, end);
277 static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
279 struct btrfs_root *root = BTRFS_I(inode)->root;
282 if (btrfs_test_opt(root, NODATACOW) ||
283 btrfs_test_flag(inode, NODATACOW))
284 ret = run_delalloc_nocow(inode, start, end);
286 ret = cow_file_range(inode, start, end);
291 int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
292 unsigned long old, unsigned long bits)
295 if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
296 struct btrfs_root *root = BTRFS_I(inode)->root;
297 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
298 BTRFS_I(inode)->delalloc_bytes += end - start + 1;
299 root->fs_info->delalloc_bytes += end - start + 1;
300 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
305 int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end,
306 unsigned long old, unsigned long bits)
308 if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
309 struct btrfs_root *root = BTRFS_I(inode)->root;
312 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
313 if (end - start + 1 > root->fs_info->delalloc_bytes) {
314 printk("warning: delalloc account %Lu %Lu\n",
315 end - start + 1, root->fs_info->delalloc_bytes);
316 root->fs_info->delalloc_bytes = 0;
317 BTRFS_I(inode)->delalloc_bytes = 0;
319 root->fs_info->delalloc_bytes -= end - start + 1;
320 BTRFS_I(inode)->delalloc_bytes -= end - start + 1;
322 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
327 int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
328 size_t size, struct bio *bio)
330 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
331 struct btrfs_mapping_tree *map_tree;
332 u64 logical = bio->bi_sector << 9;
337 length = bio->bi_size;
338 map_tree = &root->fs_info->mapping_tree;
340 ret = btrfs_map_block(map_tree, READ, logical,
341 &map_length, NULL, 0);
343 if (map_length < length + size) {
349 int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
352 struct btrfs_root *root = BTRFS_I(inode)->root;
354 struct btrfs_ordered_sum *sums;
356 ret = btrfs_csum_one_bio(root, bio, &sums);
359 ret = btrfs_add_ordered_sum(inode, sums);
362 return btrfs_map_bio(root, rw, bio, mirror_num, 1);
365 int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
368 struct btrfs_root *root = BTRFS_I(inode)->root;
371 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
374 if (!(rw & (1 << BIO_RW))) {
378 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
379 inode, rw, bio, mirror_num,
380 __btrfs_submit_bio_hook);
382 return btrfs_map_bio(root, rw, bio, mirror_num, 0);
385 static int add_pending_csums(struct btrfs_trans_handle *trans,
386 struct inode *inode, u64 file_offset,
387 struct list_head *list)
389 struct list_head *cur;
390 struct btrfs_ordered_sum *sum;
392 btrfs_set_trans_block_group(trans, inode);
393 while(!list_empty(list)) {
395 sum = list_entry(cur, struct btrfs_ordered_sum, list);
396 mutex_lock(&BTRFS_I(inode)->csum_mutex);
397 btrfs_csum_file_blocks(trans, BTRFS_I(inode)->root,
399 mutex_unlock(&BTRFS_I(inode)->csum_mutex);
400 list_del(&sum->list);
406 struct btrfs_writepage_fixup {
408 struct btrfs_work work;
411 /* see btrfs_writepage_start_hook for details on why this is required */
412 void btrfs_writepage_fixup_worker(struct btrfs_work *work)
414 struct btrfs_writepage_fixup *fixup;
415 struct btrfs_ordered_extent *ordered;
421 fixup = container_of(work, struct btrfs_writepage_fixup, work);
425 if (!page->mapping || !PageDirty(page) || !PageChecked(page)) {
426 ClearPageChecked(page);
430 inode = page->mapping->host;
431 page_start = page_offset(page);
432 page_end = page_offset(page) + PAGE_CACHE_SIZE - 1;
434 lock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
435 ordered = btrfs_lookup_ordered_extent(inode, page_start);
439 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, page_end,
441 ClearPageChecked(page);
443 unlock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
446 page_cache_release(page);
450 * There are a few paths in the higher layers of the kernel that directly
451 * set the page dirty bit without asking the filesystem if it is a
452 * good idea. This causes problems because we want to make sure COW
453 * properly happens and the data=ordered rules are followed.
455 * In our case any range that doesn't have the EXTENT_ORDERED bit set
456 * hasn't been properly setup for IO. We kick off an async process
457 * to fix it up. The async helper will wait for ordered extents, set
458 * the delalloc bit and make it safe to write the page.
460 int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
462 struct inode *inode = page->mapping->host;
463 struct btrfs_writepage_fixup *fixup;
464 struct btrfs_root *root = BTRFS_I(inode)->root;
467 ret = test_range_bit(&BTRFS_I(inode)->io_tree, start, end,
472 if (PageChecked(page))
475 fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
478 printk("queueing worker to fixup page %lu %Lu\n", inode->i_ino, page_offset(page));
479 SetPageChecked(page);
480 page_cache_get(page);
481 fixup->work.func = btrfs_writepage_fixup_worker;
483 btrfs_queue_worker(&root->fs_info->fixup_workers, &fixup->work);
487 int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
488 struct extent_state *state, int uptodate)
490 struct inode *inode = page->mapping->host;
491 struct btrfs_root *root = BTRFS_I(inode)->root;
492 struct btrfs_trans_handle *trans;
493 struct btrfs_ordered_extent *ordered_extent;
494 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
496 struct list_head list;
497 struct btrfs_key ins;
500 ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1);
505 trans = btrfs_start_transaction(root, 1);
507 ordered_extent = btrfs_lookup_ordered_extent(inode, start);
508 BUG_ON(!ordered_extent);
510 lock_extent(io_tree, ordered_extent->file_offset,
511 ordered_extent->file_offset + ordered_extent->len - 1,
514 INIT_LIST_HEAD(&list);
516 ins.objectid = ordered_extent->start;
517 ins.offset = ordered_extent->len;
518 ins.type = BTRFS_EXTENT_ITEM_KEY;
519 ret = btrfs_alloc_reserved_extent(trans, root, root->root_key.objectid,
520 trans->transid, inode->i_ino,
521 ordered_extent->file_offset, &ins);
523 ret = btrfs_drop_extents(trans, root, inode,
524 ordered_extent->file_offset,
525 ordered_extent->file_offset +
527 ordered_extent->file_offset, &alloc_hint);
529 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
530 ordered_extent->file_offset,
531 ordered_extent->start,
533 ordered_extent->len, 0);
535 btrfs_drop_extent_cache(inode, ordered_extent->file_offset,
536 ordered_extent->file_offset +
537 ordered_extent->len - 1);
538 inode->i_blocks += ordered_extent->len >> 9;
539 unlock_extent(io_tree, ordered_extent->file_offset,
540 ordered_extent->file_offset + ordered_extent->len - 1,
542 add_pending_csums(trans, inode, ordered_extent->file_offset,
543 &ordered_extent->list);
545 btrfs_remove_ordered_extent(inode, ordered_extent);
547 btrfs_put_ordered_extent(ordered_extent);
548 /* once for the tree */
549 btrfs_put_ordered_extent(ordered_extent);
551 btrfs_update_inode(trans, root, inode);
552 btrfs_end_transaction(trans, root);
556 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
559 struct inode *inode = page->mapping->host;
560 struct btrfs_root *root = BTRFS_I(inode)->root;
561 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
562 struct btrfs_csum_item *item;
563 struct btrfs_path *path = NULL;
566 if (btrfs_test_opt(root, NODATASUM) ||
567 btrfs_test_flag(inode, NODATASUM))
570 path = btrfs_alloc_path();
571 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
574 /* a csum that isn't present is a preallocated region. */
575 if (ret == -ENOENT || ret == -EFBIG)
578 printk("no csum found for inode %lu start %Lu\n", inode->i_ino,
582 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
584 set_state_private(io_tree, start, csum);
587 btrfs_free_path(path);
591 struct io_failure_record {
599 int btrfs_io_failed_hook(struct bio *failed_bio,
600 struct page *page, u64 start, u64 end,
601 struct extent_state *state)
603 struct io_failure_record *failrec = NULL;
605 struct extent_map *em;
606 struct inode *inode = page->mapping->host;
607 struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
608 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
615 ret = get_state_private(failure_tree, start, &private);
617 failrec = kmalloc(sizeof(*failrec), GFP_NOFS);
620 failrec->start = start;
621 failrec->len = end - start + 1;
622 failrec->last_mirror = 0;
624 spin_lock(&em_tree->lock);
625 em = lookup_extent_mapping(em_tree, start, failrec->len);
626 if (em->start > start || em->start + em->len < start) {
630 spin_unlock(&em_tree->lock);
632 if (!em || IS_ERR(em)) {
636 logical = start - em->start;
637 logical = em->block_start + logical;
638 failrec->logical = logical;
640 set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
641 EXTENT_DIRTY, GFP_NOFS);
642 set_state_private(failure_tree, start,
643 (u64)(unsigned long)failrec);
645 failrec = (struct io_failure_record *)(unsigned long)private;
647 num_copies = btrfs_num_copies(
648 &BTRFS_I(inode)->root->fs_info->mapping_tree,
649 failrec->logical, failrec->len);
650 failrec->last_mirror++;
652 spin_lock_irq(&BTRFS_I(inode)->io_tree.lock);
653 state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
656 if (state && state->start != failrec->start)
658 spin_unlock_irq(&BTRFS_I(inode)->io_tree.lock);
660 if (!state || failrec->last_mirror > num_copies) {
661 set_state_private(failure_tree, failrec->start, 0);
662 clear_extent_bits(failure_tree, failrec->start,
663 failrec->start + failrec->len - 1,
664 EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
668 bio = bio_alloc(GFP_NOFS, 1);
669 bio->bi_private = state;
670 bio->bi_end_io = failed_bio->bi_end_io;
671 bio->bi_sector = failrec->logical >> 9;
672 bio->bi_bdev = failed_bio->bi_bdev;
674 bio_add_page(bio, page, failrec->len, start - page_offset(page));
675 if (failed_bio->bi_rw & (1 << BIO_RW))
680 BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
681 failrec->last_mirror);
685 int btrfs_clean_io_failures(struct inode *inode, u64 start)
689 struct io_failure_record *failure;
693 if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
694 (u64)-1, 1, EXTENT_DIRTY)) {
695 ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
696 start, &private_failure);
698 failure = (struct io_failure_record *)(unsigned long)
700 set_state_private(&BTRFS_I(inode)->io_failure_tree,
702 clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
704 failure->start + failure->len - 1,
705 EXTENT_DIRTY | EXTENT_LOCKED,
713 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
714 struct extent_state *state)
716 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
717 struct inode *inode = page->mapping->host;
718 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
720 u64 private = ~(u32)0;
722 struct btrfs_root *root = BTRFS_I(inode)->root;
726 if (btrfs_test_opt(root, NODATASUM) ||
727 btrfs_test_flag(inode, NODATASUM))
729 if (state && state->start == start) {
730 private = state->private;
733 ret = get_state_private(io_tree, start, &private);
735 local_irq_save(flags);
736 kaddr = kmap_atomic(page, KM_IRQ0);
740 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
741 btrfs_csum_final(csum, (char *)&csum);
742 if (csum != private) {
745 kunmap_atomic(kaddr, KM_IRQ0);
746 local_irq_restore(flags);
748 /* if the io failure tree for this inode is non-empty,
749 * check to see if we've recovered from a failed IO
751 btrfs_clean_io_failures(inode, start);
755 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
756 page->mapping->host->i_ino, (unsigned long long)start, csum,
758 memset(kaddr + offset, 1, end - start + 1);
759 flush_dcache_page(page);
760 kunmap_atomic(kaddr, KM_IRQ0);
761 local_irq_restore(flags);
767 void btrfs_read_locked_inode(struct inode *inode)
769 struct btrfs_path *path;
770 struct extent_buffer *leaf;
771 struct btrfs_inode_item *inode_item;
772 struct btrfs_timespec *tspec;
773 struct btrfs_root *root = BTRFS_I(inode)->root;
774 struct btrfs_key location;
775 u64 alloc_group_block;
779 path = btrfs_alloc_path();
781 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
783 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
787 leaf = path->nodes[0];
788 inode_item = btrfs_item_ptr(leaf, path->slots[0],
789 struct btrfs_inode_item);
791 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
792 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
793 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
794 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
795 inode->i_size = btrfs_inode_size(leaf, inode_item);
797 tspec = btrfs_inode_atime(inode_item);
798 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
799 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
801 tspec = btrfs_inode_mtime(inode_item);
802 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
803 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
805 tspec = btrfs_inode_ctime(inode_item);
806 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
807 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
809 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
810 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
812 rdev = btrfs_inode_rdev(leaf, inode_item);
814 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
815 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
817 BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
818 if (!BTRFS_I(inode)->block_group) {
819 BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
821 BTRFS_BLOCK_GROUP_METADATA, 0);
823 btrfs_free_path(path);
826 switch (inode->i_mode & S_IFMT) {
828 inode->i_mapping->a_ops = &btrfs_aops;
829 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
830 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
831 inode->i_fop = &btrfs_file_operations;
832 inode->i_op = &btrfs_file_inode_operations;
835 inode->i_fop = &btrfs_dir_file_operations;
836 if (root == root->fs_info->tree_root)
837 inode->i_op = &btrfs_dir_ro_inode_operations;
839 inode->i_op = &btrfs_dir_inode_operations;
842 inode->i_op = &btrfs_symlink_inode_operations;
843 inode->i_mapping->a_ops = &btrfs_symlink_aops;
844 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
847 init_special_inode(inode, inode->i_mode, rdev);
853 btrfs_free_path(path);
854 make_bad_inode(inode);
857 static void fill_inode_item(struct extent_buffer *leaf,
858 struct btrfs_inode_item *item,
861 btrfs_set_inode_uid(leaf, item, inode->i_uid);
862 btrfs_set_inode_gid(leaf, item, inode->i_gid);
863 btrfs_set_inode_size(leaf, item, inode->i_size);
864 btrfs_set_inode_mode(leaf, item, inode->i_mode);
865 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
867 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
868 inode->i_atime.tv_sec);
869 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
870 inode->i_atime.tv_nsec);
872 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
873 inode->i_mtime.tv_sec);
874 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
875 inode->i_mtime.tv_nsec);
877 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
878 inode->i_ctime.tv_sec);
879 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
880 inode->i_ctime.tv_nsec);
882 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
883 btrfs_set_inode_generation(leaf, item, inode->i_generation);
884 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
885 btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
886 btrfs_set_inode_block_group(leaf, item,
887 BTRFS_I(inode)->block_group->key.objectid);
890 int btrfs_update_inode(struct btrfs_trans_handle *trans,
891 struct btrfs_root *root,
894 struct btrfs_inode_item *inode_item;
895 struct btrfs_path *path;
896 struct extent_buffer *leaf;
899 path = btrfs_alloc_path();
901 ret = btrfs_lookup_inode(trans, root, path,
902 &BTRFS_I(inode)->location, 1);
909 leaf = path->nodes[0];
910 inode_item = btrfs_item_ptr(leaf, path->slots[0],
911 struct btrfs_inode_item);
913 fill_inode_item(leaf, inode_item, inode);
914 btrfs_mark_buffer_dirty(leaf);
915 btrfs_set_inode_last_trans(trans, inode);
918 btrfs_free_path(path);
923 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
924 struct btrfs_root *root,
926 struct dentry *dentry)
928 struct btrfs_path *path;
929 const char *name = dentry->d_name.name;
930 int name_len = dentry->d_name.len;
932 struct extent_buffer *leaf;
933 struct btrfs_dir_item *di;
934 struct btrfs_key key;
936 path = btrfs_alloc_path();
942 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
952 leaf = path->nodes[0];
953 btrfs_dir_item_key_to_cpu(leaf, di, &key);
954 ret = btrfs_delete_one_dir_name(trans, root, path, di);
957 btrfs_release_path(root, path);
959 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
960 key.objectid, name, name_len, -1);
969 ret = btrfs_delete_one_dir_name(trans, root, path, di);
970 btrfs_release_path(root, path);
972 dentry->d_inode->i_ctime = dir->i_ctime;
973 ret = btrfs_del_inode_ref(trans, root, name, name_len,
974 dentry->d_inode->i_ino,
975 dentry->d_parent->d_inode->i_ino);
977 printk("failed to delete reference to %.*s, "
978 "inode %lu parent %lu\n", name_len, name,
979 dentry->d_inode->i_ino,
980 dentry->d_parent->d_inode->i_ino);
983 btrfs_free_path(path);
985 dir->i_size -= name_len * 2;
986 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
987 btrfs_update_inode(trans, root, dir);
988 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
989 dentry->d_inode->i_nlink--;
991 drop_nlink(dentry->d_inode);
993 ret = btrfs_update_inode(trans, root, dentry->d_inode);
994 dir->i_sb->s_dirt = 1;
999 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
1001 struct btrfs_root *root;
1002 struct btrfs_trans_handle *trans;
1004 unsigned long nr = 0;
1006 root = BTRFS_I(dir)->root;
1008 ret = btrfs_check_free_space(root, 1, 1);
1012 trans = btrfs_start_transaction(root, 1);
1014 btrfs_set_trans_block_group(trans, dir);
1015 ret = btrfs_unlink_trans(trans, root, dir, dentry);
1016 nr = trans->blocks_used;
1018 btrfs_end_transaction_throttle(trans, root);
1020 btrfs_btree_balance_dirty(root, nr);
1024 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
1026 struct inode *inode = dentry->d_inode;
1029 struct btrfs_root *root = BTRFS_I(dir)->root;
1030 struct btrfs_trans_handle *trans;
1031 unsigned long nr = 0;
1033 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
1037 ret = btrfs_check_free_space(root, 1, 1);
1041 trans = btrfs_start_transaction(root, 1);
1042 btrfs_set_trans_block_group(trans, dir);
1044 /* now the directory is empty */
1045 err = btrfs_unlink_trans(trans, root, dir, dentry);
1050 nr = trans->blocks_used;
1051 ret = btrfs_end_transaction_throttle(trans, root);
1053 btrfs_btree_balance_dirty(root, nr);
1061 * this can truncate away extent items, csum items and directory items.
1062 * It starts at a high offset and removes keys until it can't find
1063 * any higher than i_size.
1065 * csum items that cross the new i_size are truncated to the new size
1068 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
1069 struct btrfs_root *root,
1070 struct inode *inode,
1074 struct btrfs_path *path;
1075 struct btrfs_key key;
1076 struct btrfs_key found_key;
1078 struct extent_buffer *leaf;
1079 struct btrfs_file_extent_item *fi;
1080 u64 extent_start = 0;
1081 u64 extent_num_bytes = 0;
1087 int pending_del_nr = 0;
1088 int pending_del_slot = 0;
1089 int extent_type = -1;
1090 u64 mask = root->sectorsize - 1;
1092 btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
1093 btrfs_drop_extent_cache(inode, inode->i_size & (~mask), (u64)-1);
1094 path = btrfs_alloc_path();
1098 /* FIXME, add redo link to tree so we don't leak on crash */
1099 key.objectid = inode->i_ino;
1100 key.offset = (u64)-1;
1103 btrfs_init_path(path);
1105 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1110 BUG_ON(path->slots[0] == 0);
1116 leaf = path->nodes[0];
1117 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1118 found_type = btrfs_key_type(&found_key);
1120 if (found_key.objectid != inode->i_ino)
1123 if (found_type < min_type)
1126 item_end = found_key.offset;
1127 if (found_type == BTRFS_EXTENT_DATA_KEY) {
1128 fi = btrfs_item_ptr(leaf, path->slots[0],
1129 struct btrfs_file_extent_item);
1130 extent_type = btrfs_file_extent_type(leaf, fi);
1131 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
1133 btrfs_file_extent_num_bytes(leaf, fi);
1134 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1135 struct btrfs_item *item = btrfs_item_nr(leaf,
1137 item_end += btrfs_file_extent_inline_len(leaf,
1142 if (found_type == BTRFS_CSUM_ITEM_KEY) {
1143 ret = btrfs_csum_truncate(trans, root, path,
1147 if (item_end < inode->i_size) {
1148 if (found_type == BTRFS_DIR_ITEM_KEY) {
1149 found_type = BTRFS_INODE_ITEM_KEY;
1150 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
1151 found_type = BTRFS_CSUM_ITEM_KEY;
1152 } else if (found_type == BTRFS_EXTENT_DATA_KEY) {
1153 found_type = BTRFS_XATTR_ITEM_KEY;
1154 } else if (found_type == BTRFS_XATTR_ITEM_KEY) {
1155 found_type = BTRFS_INODE_REF_KEY;
1156 } else if (found_type) {
1161 btrfs_set_key_type(&key, found_type);
1164 if (found_key.offset >= inode->i_size)
1170 /* FIXME, shrink the extent if the ref count is only 1 */
1171 if (found_type != BTRFS_EXTENT_DATA_KEY)
1174 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
1176 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
1178 u64 orig_num_bytes =
1179 btrfs_file_extent_num_bytes(leaf, fi);
1180 extent_num_bytes = inode->i_size -
1181 found_key.offset + root->sectorsize - 1;
1182 extent_num_bytes = extent_num_bytes &
1183 ~((u64)root->sectorsize - 1);
1184 btrfs_set_file_extent_num_bytes(leaf, fi,
1186 num_dec = (orig_num_bytes -
1188 if (extent_start != 0)
1189 dec_i_blocks(inode, num_dec);
1190 btrfs_mark_buffer_dirty(leaf);
1193 btrfs_file_extent_disk_num_bytes(leaf,
1195 /* FIXME blocksize != 4096 */
1196 num_dec = btrfs_file_extent_num_bytes(leaf, fi);
1197 if (extent_start != 0) {
1199 dec_i_blocks(inode, num_dec);
1201 root_gen = btrfs_header_generation(leaf);
1202 root_owner = btrfs_header_owner(leaf);
1204 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1206 u32 newsize = inode->i_size - found_key.offset;
1207 dec_i_blocks(inode, item_end + 1 -
1208 found_key.offset - newsize);
1210 btrfs_file_extent_calc_inline_size(newsize);
1211 ret = btrfs_truncate_item(trans, root, path,
1215 dec_i_blocks(inode, item_end + 1 -
1221 if (!pending_del_nr) {
1222 /* no pending yet, add ourselves */
1223 pending_del_slot = path->slots[0];
1225 } else if (pending_del_nr &&
1226 path->slots[0] + 1 == pending_del_slot) {
1227 /* hop on the pending chunk */
1229 pending_del_slot = path->slots[0];
1231 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot);
1237 ret = btrfs_free_extent(trans, root, extent_start,
1240 root_gen, inode->i_ino,
1241 found_key.offset, 0);
1245 if (path->slots[0] == 0) {
1248 btrfs_release_path(root, path);
1253 if (pending_del_nr &&
1254 path->slots[0] + 1 != pending_del_slot) {
1255 struct btrfs_key debug;
1257 btrfs_item_key_to_cpu(path->nodes[0], &debug,
1259 ret = btrfs_del_items(trans, root, path,
1264 btrfs_release_path(root, path);
1270 if (pending_del_nr) {
1271 ret = btrfs_del_items(trans, root, path, pending_del_slot,
1274 btrfs_free_path(path);
1275 inode->i_sb->s_dirt = 1;
1280 * taken from block_truncate_page, but does cow as it zeros out
1281 * any bytes left in the last page in the file.
1283 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
1285 struct inode *inode = mapping->host;
1286 struct btrfs_root *root = BTRFS_I(inode)->root;
1287 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1288 struct btrfs_ordered_extent *ordered;
1290 u32 blocksize = root->sectorsize;
1291 pgoff_t index = from >> PAGE_CACHE_SHIFT;
1292 unsigned offset = from & (PAGE_CACHE_SIZE-1);
1298 if ((offset & (blocksize - 1)) == 0)
1303 page = grab_cache_page(mapping, index);
1307 page_start = page_offset(page);
1308 page_end = page_start + PAGE_CACHE_SIZE - 1;
1310 if (!PageUptodate(page)) {
1311 ret = btrfs_readpage(NULL, page);
1313 if (page->mapping != mapping) {
1315 page_cache_release(page);
1318 if (!PageUptodate(page)) {
1323 wait_on_page_writeback(page);
1325 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
1326 set_page_extent_mapped(page);
1328 ordered = btrfs_lookup_ordered_extent(inode, page_start);
1330 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
1332 page_cache_release(page);
1333 btrfs_wait_ordered_extent(inode, ordered);
1334 btrfs_put_ordered_extent(ordered);
1338 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
1339 page_end, GFP_NOFS);
1341 if (offset != PAGE_CACHE_SIZE) {
1343 memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
1344 flush_dcache_page(page);
1347 ClearPageChecked(page);
1348 set_page_dirty(page);
1349 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
1352 page_cache_release(page);
1357 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
1359 struct inode *inode = dentry->d_inode;
1362 err = inode_change_ok(inode, attr);
1366 if (S_ISREG(inode->i_mode) &&
1367 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
1368 struct btrfs_trans_handle *trans;
1369 struct btrfs_root *root = BTRFS_I(inode)->root;
1370 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1372 u64 mask = root->sectorsize - 1;
1373 u64 hole_start = (inode->i_size + mask) & ~mask;
1374 u64 block_end = (attr->ia_size + mask) & ~mask;
1378 if (attr->ia_size <= hole_start)
1381 err = btrfs_check_free_space(root, 1, 0);
1385 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1387 hole_size = block_end - hole_start;
1388 btrfs_wait_ordered_range(inode, hole_start, hole_size);
1389 lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1391 trans = btrfs_start_transaction(root, 1);
1392 btrfs_set_trans_block_group(trans, inode);
1393 err = btrfs_drop_extents(trans, root, inode,
1394 hole_start, block_end, hole_start,
1397 if (alloc_hint != EXTENT_MAP_INLINE) {
1398 err = btrfs_insert_file_extent(trans, root,
1402 btrfs_drop_extent_cache(inode, hole_start,
1404 btrfs_check_file(root, inode);
1406 btrfs_end_transaction(trans, root);
1407 unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1412 err = inode_setattr(inode, attr);
1417 void btrfs_delete_inode(struct inode *inode)
1419 struct btrfs_trans_handle *trans;
1420 struct btrfs_root *root = BTRFS_I(inode)->root;
1424 btrfs_wait_ordered_range(inode, 0, (u64)-1);
1425 truncate_inode_pages(&inode->i_data, 0);
1426 if (is_bad_inode(inode)) {
1431 trans = btrfs_start_transaction(root, 1);
1433 btrfs_set_trans_block_group(trans, inode);
1434 ret = btrfs_truncate_in_trans(trans, root, inode, 0);
1436 goto no_delete_lock;
1438 nr = trans->blocks_used;
1441 btrfs_end_transaction(trans, root);
1442 btrfs_btree_balance_dirty(root, nr);
1446 nr = trans->blocks_used;
1447 btrfs_end_transaction(trans, root);
1448 btrfs_btree_balance_dirty(root, nr);
1454 * this returns the key found in the dir entry in the location pointer.
1455 * If no dir entries were found, location->objectid is 0.
1457 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
1458 struct btrfs_key *location)
1460 const char *name = dentry->d_name.name;
1461 int namelen = dentry->d_name.len;
1462 struct btrfs_dir_item *di;
1463 struct btrfs_path *path;
1464 struct btrfs_root *root = BTRFS_I(dir)->root;
1467 if (namelen == 1 && strcmp(name, ".") == 0) {
1468 location->objectid = dir->i_ino;
1469 location->type = BTRFS_INODE_ITEM_KEY;
1470 location->offset = 0;
1473 path = btrfs_alloc_path();
1476 if (namelen == 2 && strcmp(name, "..") == 0) {
1477 struct btrfs_key key;
1478 struct extent_buffer *leaf;
1482 key.objectid = dir->i_ino;
1483 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1485 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1489 leaf = path->nodes[0];
1490 slot = path->slots[0];
1491 nritems = btrfs_header_nritems(leaf);
1492 if (slot >= nritems)
1495 btrfs_item_key_to_cpu(leaf, &key, slot);
1496 if (key.objectid != dir->i_ino ||
1497 key.type != BTRFS_INODE_REF_KEY) {
1500 location->objectid = key.offset;
1501 location->type = BTRFS_INODE_ITEM_KEY;
1502 location->offset = 0;
1506 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
1510 if (!di || IS_ERR(di)) {
1513 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
1515 btrfs_free_path(path);
1518 location->objectid = 0;
1523 * when we hit a tree root in a directory, the btrfs part of the inode
1524 * needs to be changed to reflect the root directory of the tree root. This
1525 * is kind of like crossing a mount point.
1527 static int fixup_tree_root_location(struct btrfs_root *root,
1528 struct btrfs_key *location,
1529 struct btrfs_root **sub_root,
1530 struct dentry *dentry)
1532 struct btrfs_path *path;
1533 struct btrfs_root_item *ri;
1535 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
1537 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1540 path = btrfs_alloc_path();
1543 *sub_root = btrfs_read_fs_root(root->fs_info, location,
1544 dentry->d_name.name,
1545 dentry->d_name.len);
1546 if (IS_ERR(*sub_root))
1547 return PTR_ERR(*sub_root);
1549 ri = &(*sub_root)->root_item;
1550 location->objectid = btrfs_root_dirid(ri);
1551 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1552 location->offset = 0;
1554 btrfs_free_path(path);
1558 static int btrfs_init_locked_inode(struct inode *inode, void *p)
1560 struct btrfs_iget_args *args = p;
1561 inode->i_ino = args->ino;
1562 BTRFS_I(inode)->root = args->root;
1563 BTRFS_I(inode)->delalloc_bytes = 0;
1564 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1565 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1566 inode->i_mapping, GFP_NOFS);
1567 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1568 inode->i_mapping, GFP_NOFS);
1569 mutex_init(&BTRFS_I(inode)->csum_mutex);
1573 static int btrfs_find_actor(struct inode *inode, void *opaque)
1575 struct btrfs_iget_args *args = opaque;
1576 return (args->ino == inode->i_ino &&
1577 args->root == BTRFS_I(inode)->root);
1580 struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
1583 struct btrfs_iget_args args;
1584 args.ino = objectid;
1585 args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
1590 return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
1593 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
1594 struct btrfs_root *root)
1596 struct inode *inode;
1597 struct btrfs_iget_args args;
1598 args.ino = objectid;
1601 inode = iget5_locked(s, objectid, btrfs_find_actor,
1602 btrfs_init_locked_inode,
1607 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
1608 struct nameidata *nd)
1610 struct inode * inode;
1611 struct btrfs_inode *bi = BTRFS_I(dir);
1612 struct btrfs_root *root = bi->root;
1613 struct btrfs_root *sub_root = root;
1614 struct btrfs_key location;
1617 if (dentry->d_name.len > BTRFS_NAME_LEN)
1618 return ERR_PTR(-ENAMETOOLONG);
1620 ret = btrfs_inode_by_name(dir, dentry, &location);
1623 return ERR_PTR(ret);
1626 if (location.objectid) {
1627 ret = fixup_tree_root_location(root, &location, &sub_root,
1630 return ERR_PTR(ret);
1632 return ERR_PTR(-ENOENT);
1633 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
1636 return ERR_PTR(-EACCES);
1637 if (inode->i_state & I_NEW) {
1638 /* the inode and parent dir are two different roots */
1639 if (sub_root != root) {
1641 sub_root->inode = inode;
1643 BTRFS_I(inode)->root = sub_root;
1644 memcpy(&BTRFS_I(inode)->location, &location,
1646 btrfs_read_locked_inode(inode);
1647 unlock_new_inode(inode);
1650 return d_splice_alias(inode, dentry);
1653 static unsigned char btrfs_filetype_table[] = {
1654 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1657 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1659 struct inode *inode = filp->f_dentry->d_inode;
1660 struct btrfs_root *root = BTRFS_I(inode)->root;
1661 struct btrfs_item *item;
1662 struct btrfs_dir_item *di;
1663 struct btrfs_key key;
1664 struct btrfs_key found_key;
1665 struct btrfs_path *path;
1668 struct extent_buffer *leaf;
1671 unsigned char d_type;
1676 int key_type = BTRFS_DIR_INDEX_KEY;
1681 /* FIXME, use a real flag for deciding about the key type */
1682 if (root->fs_info->tree_root == root)
1683 key_type = BTRFS_DIR_ITEM_KEY;
1685 /* special case for "." */
1686 if (filp->f_pos == 0) {
1687 over = filldir(dirent, ".", 1,
1695 key.objectid = inode->i_ino;
1696 path = btrfs_alloc_path();
1699 /* special case for .., just use the back ref */
1700 if (filp->f_pos == 1) {
1701 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1703 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1705 leaf = path->nodes[0];
1706 slot = path->slots[0];
1707 nritems = btrfs_header_nritems(leaf);
1708 if (slot >= nritems) {
1709 btrfs_release_path(root, path);
1710 goto read_dir_items;
1712 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1713 btrfs_release_path(root, path);
1714 if (found_key.objectid != key.objectid ||
1715 found_key.type != BTRFS_INODE_REF_KEY)
1716 goto read_dir_items;
1717 over = filldir(dirent, "..", 2,
1718 2, found_key.offset, DT_DIR);
1725 btrfs_set_key_type(&key, key_type);
1726 key.offset = filp->f_pos;
1728 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1733 leaf = path->nodes[0];
1734 nritems = btrfs_header_nritems(leaf);
1735 slot = path->slots[0];
1736 if (advance || slot >= nritems) {
1737 if (slot >= nritems -1) {
1738 ret = btrfs_next_leaf(root, path);
1741 leaf = path->nodes[0];
1742 nritems = btrfs_header_nritems(leaf);
1743 slot = path->slots[0];
1750 item = btrfs_item_nr(leaf, slot);
1751 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1753 if (found_key.objectid != key.objectid)
1755 if (btrfs_key_type(&found_key) != key_type)
1757 if (found_key.offset < filp->f_pos)
1760 filp->f_pos = found_key.offset;
1762 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1764 di_total = btrfs_item_size(leaf, item);
1765 while(di_cur < di_total) {
1766 struct btrfs_key location;
1768 name_len = btrfs_dir_name_len(leaf, di);
1769 if (name_len < 32) {
1770 name_ptr = tmp_name;
1772 name_ptr = kmalloc(name_len, GFP_NOFS);
1775 read_extent_buffer(leaf, name_ptr,
1776 (unsigned long)(di + 1), name_len);
1778 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1779 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1780 over = filldir(dirent, name_ptr, name_len,
1785 if (name_ptr != tmp_name)
1790 di_len = btrfs_dir_name_len(leaf, di) +
1791 btrfs_dir_data_len(leaf, di) +sizeof(*di);
1793 di = (struct btrfs_dir_item *)((char *)di + di_len);
1796 if (key_type == BTRFS_DIR_INDEX_KEY)
1797 filp->f_pos = INT_LIMIT(typeof(filp->f_pos));
1803 btrfs_free_path(path);
1807 int btrfs_write_inode(struct inode *inode, int wait)
1809 struct btrfs_root *root = BTRFS_I(inode)->root;
1810 struct btrfs_trans_handle *trans;
1814 trans = btrfs_start_transaction(root, 1);
1815 btrfs_set_trans_block_group(trans, inode);
1816 ret = btrfs_commit_transaction(trans, root);
1822 * This is somewhat expensive, updating the tree every time the
1823 * inode changes. But, it is most likely to find the inode in cache.
1824 * FIXME, needs more benchmarking...there are no reasons other than performance
1825 * to keep or drop this code.
1827 void btrfs_dirty_inode(struct inode *inode)
1829 struct btrfs_root *root = BTRFS_I(inode)->root;
1830 struct btrfs_trans_handle *trans;
1832 trans = btrfs_start_transaction(root, 1);
1833 btrfs_set_trans_block_group(trans, inode);
1834 btrfs_update_inode(trans, root, inode);
1835 btrfs_end_transaction(trans, root);
1838 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1839 struct btrfs_root *root,
1840 const char *name, int name_len,
1843 struct btrfs_block_group_cache *group,
1846 struct inode *inode;
1847 struct btrfs_inode_item *inode_item;
1848 struct btrfs_block_group_cache *new_inode_group;
1849 struct btrfs_key *location;
1850 struct btrfs_path *path;
1851 struct btrfs_inode_ref *ref;
1852 struct btrfs_key key[2];
1858 path = btrfs_alloc_path();
1861 inode = new_inode(root->fs_info->sb);
1863 return ERR_PTR(-ENOMEM);
1865 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1866 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1867 inode->i_mapping, GFP_NOFS);
1868 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1869 inode->i_mapping, GFP_NOFS);
1870 mutex_init(&BTRFS_I(inode)->csum_mutex);
1871 BTRFS_I(inode)->delalloc_bytes = 0;
1872 BTRFS_I(inode)->root = root;
1878 new_inode_group = btrfs_find_block_group(root, group, 0,
1879 BTRFS_BLOCK_GROUP_METADATA, owner);
1880 if (!new_inode_group) {
1881 printk("find_block group failed\n");
1882 new_inode_group = group;
1884 BTRFS_I(inode)->block_group = new_inode_group;
1885 BTRFS_I(inode)->flags = 0;
1887 key[0].objectid = objectid;
1888 btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
1891 key[1].objectid = objectid;
1892 btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
1893 key[1].offset = ref_objectid;
1895 sizes[0] = sizeof(struct btrfs_inode_item);
1896 sizes[1] = name_len + sizeof(*ref);
1898 ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
1902 if (objectid > root->highest_inode)
1903 root->highest_inode = objectid;
1905 inode->i_uid = current->fsuid;
1906 inode->i_gid = current->fsgid;
1907 inode->i_mode = mode;
1908 inode->i_ino = objectid;
1909 inode->i_blocks = 0;
1910 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1911 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1912 struct btrfs_inode_item);
1913 fill_inode_item(path->nodes[0], inode_item, inode);
1915 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
1916 struct btrfs_inode_ref);
1917 btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
1918 ptr = (unsigned long)(ref + 1);
1919 write_extent_buffer(path->nodes[0], name, ptr, name_len);
1921 btrfs_mark_buffer_dirty(path->nodes[0]);
1922 btrfs_free_path(path);
1924 location = &BTRFS_I(inode)->location;
1925 location->objectid = objectid;
1926 location->offset = 0;
1927 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1929 insert_inode_hash(inode);
1932 btrfs_free_path(path);
1933 return ERR_PTR(ret);
1936 static inline u8 btrfs_inode_type(struct inode *inode)
1938 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1941 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1942 struct dentry *dentry, struct inode *inode,
1946 struct btrfs_key key;
1947 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1948 struct inode *parent_inode;
1950 key.objectid = inode->i_ino;
1951 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1954 ret = btrfs_insert_dir_item(trans, root,
1955 dentry->d_name.name, dentry->d_name.len,
1956 dentry->d_parent->d_inode->i_ino,
1957 &key, btrfs_inode_type(inode));
1960 ret = btrfs_insert_inode_ref(trans, root,
1961 dentry->d_name.name,
1964 dentry->d_parent->d_inode->i_ino);
1966 parent_inode = dentry->d_parent->d_inode;
1967 parent_inode->i_size += dentry->d_name.len * 2;
1968 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1969 ret = btrfs_update_inode(trans, root,
1970 dentry->d_parent->d_inode);
1975 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1976 struct dentry *dentry, struct inode *inode,
1979 int err = btrfs_add_link(trans, dentry, inode, backref);
1981 d_instantiate(dentry, inode);
1989 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1990 int mode, dev_t rdev)
1992 struct btrfs_trans_handle *trans;
1993 struct btrfs_root *root = BTRFS_I(dir)->root;
1994 struct inode *inode = NULL;
1998 unsigned long nr = 0;
2000 if (!new_valid_dev(rdev))
2003 err = btrfs_check_free_space(root, 1, 0);
2007 trans = btrfs_start_transaction(root, 1);
2008 btrfs_set_trans_block_group(trans, dir);
2010 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2016 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2018 dentry->d_parent->d_inode->i_ino, objectid,
2019 BTRFS_I(dir)->block_group, mode);
2020 err = PTR_ERR(inode);
2024 btrfs_set_trans_block_group(trans, inode);
2025 err = btrfs_add_nondir(trans, dentry, inode, 0);
2029 inode->i_op = &btrfs_special_inode_operations;
2030 init_special_inode(inode, inode->i_mode, rdev);
2031 btrfs_update_inode(trans, root, inode);
2033 dir->i_sb->s_dirt = 1;
2034 btrfs_update_inode_block_group(trans, inode);
2035 btrfs_update_inode_block_group(trans, dir);
2037 nr = trans->blocks_used;
2038 btrfs_end_transaction_throttle(trans, root);
2041 inode_dec_link_count(inode);
2044 btrfs_btree_balance_dirty(root, nr);
2048 static int btrfs_create(struct inode *dir, struct dentry *dentry,
2049 int mode, struct nameidata *nd)
2051 struct btrfs_trans_handle *trans;
2052 struct btrfs_root *root = BTRFS_I(dir)->root;
2053 struct inode *inode = NULL;
2056 unsigned long nr = 0;
2059 err = btrfs_check_free_space(root, 1, 0);
2062 trans = btrfs_start_transaction(root, 1);
2063 btrfs_set_trans_block_group(trans, dir);
2065 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2071 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2073 dentry->d_parent->d_inode->i_ino,
2074 objectid, BTRFS_I(dir)->block_group, mode);
2075 err = PTR_ERR(inode);
2079 btrfs_set_trans_block_group(trans, inode);
2080 err = btrfs_add_nondir(trans, dentry, inode, 0);
2084 inode->i_mapping->a_ops = &btrfs_aops;
2085 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
2086 inode->i_fop = &btrfs_file_operations;
2087 inode->i_op = &btrfs_file_inode_operations;
2088 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2089 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
2090 inode->i_mapping, GFP_NOFS);
2091 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
2092 inode->i_mapping, GFP_NOFS);
2093 mutex_init(&BTRFS_I(inode)->csum_mutex);
2094 BTRFS_I(inode)->delalloc_bytes = 0;
2095 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
2097 dir->i_sb->s_dirt = 1;
2098 btrfs_update_inode_block_group(trans, inode);
2099 btrfs_update_inode_block_group(trans, dir);
2101 nr = trans->blocks_used;
2102 btrfs_end_transaction_throttle(trans, root);
2105 inode_dec_link_count(inode);
2108 btrfs_btree_balance_dirty(root, nr);
2112 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
2113 struct dentry *dentry)
2115 struct btrfs_trans_handle *trans;
2116 struct btrfs_root *root = BTRFS_I(dir)->root;
2117 struct inode *inode = old_dentry->d_inode;
2118 unsigned long nr = 0;
2122 if (inode->i_nlink == 0)
2125 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2130 err = btrfs_check_free_space(root, 1, 0);
2133 trans = btrfs_start_transaction(root, 1);
2135 btrfs_set_trans_block_group(trans, dir);
2136 atomic_inc(&inode->i_count);
2137 err = btrfs_add_nondir(trans, dentry, inode, 1);
2142 dir->i_sb->s_dirt = 1;
2143 btrfs_update_inode_block_group(trans, dir);
2144 err = btrfs_update_inode(trans, root, inode);
2149 nr = trans->blocks_used;
2150 btrfs_end_transaction_throttle(trans, root);
2153 inode_dec_link_count(inode);
2156 btrfs_btree_balance_dirty(root, nr);
2160 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
2162 struct inode *inode = NULL;
2163 struct btrfs_trans_handle *trans;
2164 struct btrfs_root *root = BTRFS_I(dir)->root;
2166 int drop_on_err = 0;
2168 unsigned long nr = 1;
2170 err = btrfs_check_free_space(root, 1, 0);
2174 trans = btrfs_start_transaction(root, 1);
2175 btrfs_set_trans_block_group(trans, dir);
2177 if (IS_ERR(trans)) {
2178 err = PTR_ERR(trans);
2182 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2188 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2190 dentry->d_parent->d_inode->i_ino, objectid,
2191 BTRFS_I(dir)->block_group, S_IFDIR | mode);
2192 if (IS_ERR(inode)) {
2193 err = PTR_ERR(inode);
2198 inode->i_op = &btrfs_dir_inode_operations;
2199 inode->i_fop = &btrfs_dir_file_operations;
2200 btrfs_set_trans_block_group(trans, inode);
2203 err = btrfs_update_inode(trans, root, inode);
2207 err = btrfs_add_link(trans, dentry, inode, 0);
2211 d_instantiate(dentry, inode);
2213 dir->i_sb->s_dirt = 1;
2214 btrfs_update_inode_block_group(trans, inode);
2215 btrfs_update_inode_block_group(trans, dir);
2218 nr = trans->blocks_used;
2219 btrfs_end_transaction_throttle(trans, root);
2224 btrfs_btree_balance_dirty(root, nr);
2228 static int merge_extent_mapping(struct extent_map_tree *em_tree,
2229 struct extent_map *existing,
2230 struct extent_map *em,
2231 u64 map_start, u64 map_len)
2235 BUG_ON(map_start < em->start || map_start >= extent_map_end(em));
2236 start_diff = map_start - em->start;
2237 em->start = map_start;
2239 if (em->block_start < EXTENT_MAP_LAST_BYTE)
2240 em->block_start += start_diff;
2241 return add_extent_mapping(em_tree, em);
2244 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
2245 size_t pg_offset, u64 start, u64 len,
2251 u64 extent_start = 0;
2253 u64 objectid = inode->i_ino;
2255 struct btrfs_path *path;
2256 struct btrfs_root *root = BTRFS_I(inode)->root;
2257 struct btrfs_file_extent_item *item;
2258 struct extent_buffer *leaf;
2259 struct btrfs_key found_key;
2260 struct extent_map *em = NULL;
2261 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2262 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2263 struct btrfs_trans_handle *trans = NULL;
2265 path = btrfs_alloc_path();
2269 spin_lock(&em_tree->lock);
2270 em = lookup_extent_mapping(em_tree, start, len);
2272 em->bdev = root->fs_info->fs_devices->latest_bdev;
2273 spin_unlock(&em_tree->lock);
2276 if (em->start > start || em->start + em->len <= start)
2277 free_extent_map(em);
2278 else if (em->block_start == EXTENT_MAP_INLINE && page)
2279 free_extent_map(em);
2283 em = alloc_extent_map(GFP_NOFS);
2288 em->bdev = root->fs_info->fs_devices->latest_bdev;
2289 em->start = EXTENT_MAP_HOLE;
2291 ret = btrfs_lookup_file_extent(trans, root, path,
2292 objectid, start, trans != NULL);
2299 if (path->slots[0] == 0)
2304 leaf = path->nodes[0];
2305 item = btrfs_item_ptr(leaf, path->slots[0],
2306 struct btrfs_file_extent_item);
2307 /* are we inside the extent that was found? */
2308 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2309 found_type = btrfs_key_type(&found_key);
2310 if (found_key.objectid != objectid ||
2311 found_type != BTRFS_EXTENT_DATA_KEY) {
2315 found_type = btrfs_file_extent_type(leaf, item);
2316 extent_start = found_key.offset;
2317 if (found_type == BTRFS_FILE_EXTENT_REG) {
2318 extent_end = extent_start +
2319 btrfs_file_extent_num_bytes(leaf, item);
2321 if (start < extent_start || start >= extent_end) {
2323 if (start < extent_start) {
2324 if (start + len <= extent_start)
2326 em->len = extent_end - extent_start;
2332 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
2334 em->start = extent_start;
2335 em->len = extent_end - extent_start;
2336 em->block_start = EXTENT_MAP_HOLE;
2339 bytenr += btrfs_file_extent_offset(leaf, item);
2340 em->block_start = bytenr;
2341 em->start = extent_start;
2342 em->len = extent_end - extent_start;
2344 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
2349 size_t extent_offset;
2352 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
2354 extent_end = (extent_start + size + root->sectorsize - 1) &
2355 ~((u64)root->sectorsize - 1);
2356 if (start < extent_start || start >= extent_end) {
2358 if (start < extent_start) {
2359 if (start + len <= extent_start)
2361 em->len = extent_end - extent_start;
2367 em->block_start = EXTENT_MAP_INLINE;
2370 em->start = extent_start;
2375 page_start = page_offset(page) + pg_offset;
2376 extent_offset = page_start - extent_start;
2377 copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
2378 size - extent_offset);
2379 em->start = extent_start + extent_offset;
2380 em->len = (copy_size + root->sectorsize - 1) &
2381 ~((u64)root->sectorsize - 1);
2383 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
2384 if (create == 0 && !PageUptodate(page)) {
2385 read_extent_buffer(leaf, map + pg_offset, ptr,
2387 flush_dcache_page(page);
2388 } else if (create && PageUptodate(page)) {
2391 free_extent_map(em);
2393 btrfs_release_path(root, path);
2394 trans = btrfs_start_transaction(root, 1);
2397 write_extent_buffer(leaf, map + pg_offset, ptr,
2399 btrfs_mark_buffer_dirty(leaf);
2402 set_extent_uptodate(io_tree, em->start,
2403 extent_map_end(em) - 1, GFP_NOFS);
2406 printk("unkknown found_type %d\n", found_type);
2413 em->block_start = EXTENT_MAP_HOLE;
2415 btrfs_release_path(root, path);
2416 if (em->start > start || extent_map_end(em) <= start) {
2417 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len);
2423 spin_lock(&em_tree->lock);
2424 ret = add_extent_mapping(em_tree, em);
2425 /* it is possible that someone inserted the extent into the tree
2426 * while we had the lock dropped. It is also possible that
2427 * an overlapping map exists in the tree
2429 if (ret == -EEXIST) {
2430 struct extent_map *existing;
2434 existing = lookup_extent_mapping(em_tree, start, len);
2435 if (existing && (existing->start > start ||
2436 existing->start + existing->len <= start)) {
2437 free_extent_map(existing);
2441 existing = lookup_extent_mapping(em_tree, em->start,
2444 err = merge_extent_mapping(em_tree, existing,
2447 free_extent_map(existing);
2449 free_extent_map(em);
2454 printk("failing to insert %Lu %Lu\n",
2456 free_extent_map(em);
2460 free_extent_map(em);
2465 spin_unlock(&em_tree->lock);
2467 btrfs_free_path(path);
2469 ret = btrfs_end_transaction(trans, root);
2475 free_extent_map(em);
2477 return ERR_PTR(err);
2482 #if 0 /* waiting for O_DIRECT reads */
2483 static int btrfs_get_block(struct inode *inode, sector_t iblock,
2484 struct buffer_head *bh_result, int create)
2486 struct extent_map *em;
2487 u64 start = (u64)iblock << inode->i_blkbits;
2488 struct btrfs_multi_bio *multi = NULL;
2489 struct btrfs_root *root = BTRFS_I(inode)->root;
2495 em = btrfs_get_extent(inode, NULL, 0, start, bh_result->b_size, 0);
2497 if (!em || IS_ERR(em))
2500 if (em->start > start || em->start + em->len <= start) {
2504 if (em->block_start == EXTENT_MAP_INLINE) {
2509 len = em->start + em->len - start;
2510 len = min_t(u64, len, INT_LIMIT(typeof(bh_result->b_size)));
2512 if (em->block_start == EXTENT_MAP_HOLE ||
2513 em->block_start == EXTENT_MAP_DELALLOC) {
2514 bh_result->b_size = len;
2518 logical = start - em->start;
2519 logical = em->block_start + logical;
2522 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
2523 logical, &map_length, &multi, 0);
2525 bh_result->b_blocknr = multi->stripes[0].physical >> inode->i_blkbits;
2526 bh_result->b_size = min(map_length, len);
2528 bh_result->b_bdev = multi->stripes[0].dev->bdev;
2529 set_buffer_mapped(bh_result);
2532 free_extent_map(em);
2537 static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
2538 const struct iovec *iov, loff_t offset,
2539 unsigned long nr_segs)
2543 struct file *file = iocb->ki_filp;
2544 struct inode *inode = file->f_mapping->host;
2549 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2550 offset, nr_segs, btrfs_get_block, NULL);
2554 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
2556 return extent_bmap(mapping, iblock, btrfs_get_extent);
2559 int btrfs_readpage(struct file *file, struct page *page)
2561 struct extent_io_tree *tree;
2562 tree = &BTRFS_I(page->mapping->host)->io_tree;
2563 return extent_read_full_page(tree, page, btrfs_get_extent);
2566 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
2568 struct extent_io_tree *tree;
2571 if (current->flags & PF_MEMALLOC) {
2572 redirty_page_for_writepage(wbc, page);
2576 tree = &BTRFS_I(page->mapping->host)->io_tree;
2577 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
2580 static int btrfs_writepages(struct address_space *mapping,
2581 struct writeback_control *wbc)
2583 struct extent_io_tree *tree;
2584 tree = &BTRFS_I(mapping->host)->io_tree;
2585 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
2589 btrfs_readpages(struct file *file, struct address_space *mapping,
2590 struct list_head *pages, unsigned nr_pages)
2592 struct extent_io_tree *tree;
2593 tree = &BTRFS_I(mapping->host)->io_tree;
2594 return extent_readpages(tree, mapping, pages, nr_pages,
2597 static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
2599 struct extent_io_tree *tree;
2600 struct extent_map_tree *map;
2603 tree = &BTRFS_I(page->mapping->host)->io_tree;
2604 map = &BTRFS_I(page->mapping->host)->extent_tree;
2605 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
2607 invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE);
2608 ClearPagePrivate(page);
2609 set_page_private(page, 0);
2610 page_cache_release(page);
2615 static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
2617 struct btrfs_ordered_extent *ordered;
2619 ordered = btrfs_lookup_ordered_extent(page->mapping->host,
2622 btrfs_put_ordered_extent(ordered);
2625 return __btrfs_releasepage(page, gfp_flags);
2628 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
2630 struct extent_io_tree *tree;
2631 struct btrfs_ordered_extent *ordered;
2632 u64 page_start = page_offset(page);
2633 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
2635 wait_on_page_writeback(page);
2636 tree = &BTRFS_I(page->mapping->host)->io_tree;
2638 btrfs_releasepage(page, GFP_NOFS);
2642 lock_extent(tree, page_start, page_end, GFP_NOFS);
2643 ordered = btrfs_lookup_ordered_extent(page->mapping->host,
2646 clear_extent_bit(tree, page_start, page_end,
2647 EXTENT_DIRTY | EXTENT_DELALLOC |
2648 EXTENT_LOCKED, 1, 0, GFP_NOFS);
2649 btrfs_writepage_end_io_hook(page, page_start,
2651 btrfs_put_ordered_extent(ordered);
2652 lock_extent(tree, page_start, page_end, GFP_NOFS);
2654 clear_extent_bit(tree, page_start, page_end,
2655 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
2658 __btrfs_releasepage(page, GFP_NOFS);
2660 if (PagePrivate(page)) {
2661 invalidate_extent_lru(tree, page_offset(page),
2663 ClearPagePrivate(page);
2664 set_page_private(page, 0);
2665 page_cache_release(page);
2670 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2671 * called from a page fault handler when a page is first dirtied. Hence we must
2672 * be careful to check for EOF conditions here. We set the page up correctly
2673 * for a written page which means we get ENOSPC checking when writing into
2674 * holes and correct delalloc and unwritten extent mapping on filesystems that
2675 * support these features.
2677 * We are not allowed to take the i_mutex here so we have to play games to
2678 * protect against truncate races as the page could now be beyond EOF. Because
2679 * vmtruncate() writes the inode size before removing pages, once we have the
2680 * page lock we can determine safely if the page is beyond EOF. If it is not
2681 * beyond EOF, then the page is guaranteed safe against truncation until we
2684 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
2686 struct inode *inode = fdentry(vma->vm_file)->d_inode;
2687 struct btrfs_root *root = BTRFS_I(inode)->root;
2688 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2689 struct btrfs_ordered_extent *ordered;
2691 unsigned long zero_start;
2697 ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
2704 size = i_size_read(inode);
2705 page_start = page_offset(page);
2706 page_end = page_start + PAGE_CACHE_SIZE - 1;
2708 if ((page->mapping != inode->i_mapping) ||
2709 (page_start >= size)) {
2710 /* page got truncated out from underneath us */
2713 wait_on_page_writeback(page);
2715 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
2716 set_page_extent_mapped(page);
2718 ordered = btrfs_lookup_ordered_extent(inode, page_start);
2720 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
2722 btrfs_wait_ordered_extent(inode, ordered);
2723 btrfs_put_ordered_extent(ordered);
2727 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
2728 page_end, GFP_NOFS);
2731 /* page is wholly or partially inside EOF */
2732 if (page_start + PAGE_CACHE_SIZE > size)
2733 zero_start = size & ~PAGE_CACHE_MASK;
2735 zero_start = PAGE_CACHE_SIZE;
2737 if (zero_start != PAGE_CACHE_SIZE) {
2739 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
2740 flush_dcache_page(page);
2743 ClearPageChecked(page);
2744 set_page_dirty(page);
2745 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
2753 static void btrfs_truncate(struct inode *inode)
2755 struct btrfs_root *root = BTRFS_I(inode)->root;
2757 struct btrfs_trans_handle *trans;
2760 if (!S_ISREG(inode->i_mode))
2762 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
2765 btrfs_truncate_page(inode->i_mapping, inode->i_size);
2767 trans = btrfs_start_transaction(root, 1);
2768 btrfs_set_trans_block_group(trans, inode);
2770 /* FIXME, add redo link to tree so we don't leak on crash */
2771 ret = btrfs_truncate_in_trans(trans, root, inode,
2772 BTRFS_EXTENT_DATA_KEY);
2773 btrfs_update_inode(trans, root, inode);
2774 nr = trans->blocks_used;
2776 ret = btrfs_end_transaction_throttle(trans, root);
2778 btrfs_btree_balance_dirty(root, nr);
2782 * Invalidate a single dcache entry at the root of the filesystem.
2783 * Needed after creation of snapshot or subvolume.
2785 void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name,
2788 struct dentry *alias, *entry;
2791 alias = d_find_alias(root->fs_info->sb->s_root->d_inode);
2795 /* change me if btrfs ever gets a d_hash operation */
2796 qstr.hash = full_name_hash(qstr.name, qstr.len);
2797 entry = d_lookup(alias, &qstr);
2800 d_invalidate(entry);
2806 int btrfs_create_subvol_root(struct btrfs_root *new_root,
2807 struct btrfs_trans_handle *trans, u64 new_dirid,
2808 struct btrfs_block_group_cache *block_group)
2810 struct inode *inode;
2813 inode = btrfs_new_inode(trans, new_root, "..", 2, new_dirid,
2814 new_dirid, block_group, S_IFDIR | 0700);
2816 return PTR_ERR(inode);
2817 inode->i_op = &btrfs_dir_inode_operations;
2818 inode->i_fop = &btrfs_dir_file_operations;
2819 new_root->inode = inode;
2821 ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid,
2826 return btrfs_update_inode(trans, new_root, inode);
2829 unsigned long btrfs_force_ra(struct address_space *mapping,
2830 struct file_ra_state *ra, struct file *file,
2831 pgoff_t offset, pgoff_t last_index)
2833 pgoff_t req_size = last_index - offset + 1;
2835 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2836 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2839 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2840 return offset + req_size;
2844 struct inode *btrfs_alloc_inode(struct super_block *sb)
2846 struct btrfs_inode *ei;
2848 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2852 btrfs_ordered_inode_tree_init(&ei->ordered_tree);
2853 return &ei->vfs_inode;
2856 void btrfs_destroy_inode(struct inode *inode)
2858 struct btrfs_ordered_extent *ordered;
2859 WARN_ON(!list_empty(&inode->i_dentry));
2860 WARN_ON(inode->i_data.nrpages);
2863 ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
2867 printk("found ordered extent %Lu %Lu\n",
2868 ordered->file_offset, ordered->len);
2869 btrfs_remove_ordered_extent(inode, ordered);
2870 btrfs_put_ordered_extent(ordered);
2871 btrfs_put_ordered_extent(ordered);
2874 btrfs_drop_extent_cache(inode, 0, (u64)-1);
2875 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2878 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2879 static void init_once(struct kmem_cache * cachep, void *foo)
2881 static void init_once(void * foo, struct kmem_cache * cachep,
2882 unsigned long flags)
2885 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2887 inode_init_once(&ei->vfs_inode);
2890 void btrfs_destroy_cachep(void)
2892 if (btrfs_inode_cachep)
2893 kmem_cache_destroy(btrfs_inode_cachep);
2894 if (btrfs_trans_handle_cachep)
2895 kmem_cache_destroy(btrfs_trans_handle_cachep);
2896 if (btrfs_transaction_cachep)
2897 kmem_cache_destroy(btrfs_transaction_cachep);
2898 if (btrfs_bit_radix_cachep)
2899 kmem_cache_destroy(btrfs_bit_radix_cachep);
2900 if (btrfs_path_cachep)
2901 kmem_cache_destroy(btrfs_path_cachep);
2904 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2905 unsigned long extra_flags,
2906 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2907 void (*ctor)(struct kmem_cache *, void *)
2909 void (*ctor)(void *, struct kmem_cache *,
2914 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2915 SLAB_MEM_SPREAD | extra_flags), ctor
2916 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2922 int btrfs_init_cachep(void)
2924 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2925 sizeof(struct btrfs_inode),
2927 if (!btrfs_inode_cachep)
2929 btrfs_trans_handle_cachep =
2930 btrfs_cache_create("btrfs_trans_handle_cache",
2931 sizeof(struct btrfs_trans_handle),
2933 if (!btrfs_trans_handle_cachep)
2935 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
2936 sizeof(struct btrfs_transaction),
2938 if (!btrfs_transaction_cachep)
2940 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
2941 sizeof(struct btrfs_path),
2943 if (!btrfs_path_cachep)
2945 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
2946 SLAB_DESTROY_BY_RCU, NULL);
2947 if (!btrfs_bit_radix_cachep)
2951 btrfs_destroy_cachep();
2955 static int btrfs_getattr(struct vfsmount *mnt,
2956 struct dentry *dentry, struct kstat *stat)
2958 struct inode *inode = dentry->d_inode;
2959 generic_fillattr(inode, stat);
2960 stat->blksize = PAGE_CACHE_SIZE;
2961 stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9);
2965 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2966 struct inode * new_dir,struct dentry *new_dentry)
2968 struct btrfs_trans_handle *trans;
2969 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2970 struct inode *new_inode = new_dentry->d_inode;
2971 struct inode *old_inode = old_dentry->d_inode;
2972 struct timespec ctime = CURRENT_TIME;
2975 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2976 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2980 ret = btrfs_check_free_space(root, 1, 0);
2984 trans = btrfs_start_transaction(root, 1);
2986 btrfs_set_trans_block_group(trans, new_dir);
2988 old_dentry->d_inode->i_nlink++;
2989 old_dir->i_ctime = old_dir->i_mtime = ctime;
2990 new_dir->i_ctime = new_dir->i_mtime = ctime;
2991 old_inode->i_ctime = ctime;
2993 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2998 new_inode->i_ctime = CURRENT_TIME;
2999 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
3003 ret = btrfs_add_link(trans, new_dentry, old_inode, 1);
3008 btrfs_end_transaction(trans, root);
3013 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
3014 const char *symname)
3016 struct btrfs_trans_handle *trans;
3017 struct btrfs_root *root = BTRFS_I(dir)->root;
3018 struct btrfs_path *path;
3019 struct btrfs_key key;
3020 struct inode *inode = NULL;
3027 struct btrfs_file_extent_item *ei;
3028 struct extent_buffer *leaf;
3029 unsigned long nr = 0;
3031 name_len = strlen(symname) + 1;
3032 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
3033 return -ENAMETOOLONG;
3035 err = btrfs_check_free_space(root, 1, 0);
3039 trans = btrfs_start_transaction(root, 1);
3040 btrfs_set_trans_block_group(trans, dir);
3042 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
3048 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
3050 dentry->d_parent->d_inode->i_ino, objectid,
3051 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
3052 err = PTR_ERR(inode);
3056 btrfs_set_trans_block_group(trans, inode);
3057 err = btrfs_add_nondir(trans, dentry, inode, 0);
3061 inode->i_mapping->a_ops = &btrfs_aops;
3062 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3063 inode->i_fop = &btrfs_file_operations;
3064 inode->i_op = &btrfs_file_inode_operations;
3065 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
3066 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
3067 inode->i_mapping, GFP_NOFS);
3068 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
3069 inode->i_mapping, GFP_NOFS);
3070 mutex_init(&BTRFS_I(inode)->csum_mutex);
3071 BTRFS_I(inode)->delalloc_bytes = 0;
3072 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
3074 dir->i_sb->s_dirt = 1;
3075 btrfs_update_inode_block_group(trans, inode);
3076 btrfs_update_inode_block_group(trans, dir);
3080 path = btrfs_alloc_path();
3082 key.objectid = inode->i_ino;
3084 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
3085 datasize = btrfs_file_extent_calc_inline_size(name_len);
3086 err = btrfs_insert_empty_item(trans, root, path, &key,
3092 leaf = path->nodes[0];
3093 ei = btrfs_item_ptr(leaf, path->slots[0],
3094 struct btrfs_file_extent_item);
3095 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
3096 btrfs_set_file_extent_type(leaf, ei,
3097 BTRFS_FILE_EXTENT_INLINE);
3098 ptr = btrfs_file_extent_inline_start(ei);
3099 write_extent_buffer(leaf, symname, ptr, name_len);
3100 btrfs_mark_buffer_dirty(leaf);
3101 btrfs_free_path(path);
3103 inode->i_op = &btrfs_symlink_inode_operations;
3104 inode->i_mapping->a_ops = &btrfs_symlink_aops;
3105 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3106 inode->i_size = name_len - 1;
3107 err = btrfs_update_inode(trans, root, inode);
3112 nr = trans->blocks_used;
3113 btrfs_end_transaction_throttle(trans, root);
3116 inode_dec_link_count(inode);
3119 btrfs_btree_balance_dirty(root, nr);
3123 static int btrfs_set_page_dirty(struct page *page)
3125 return __set_page_dirty_nobuffers(page);
3128 static int btrfs_permission(struct inode *inode, int mask,
3129 struct nameidata *nd)
3131 if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
3133 return generic_permission(inode, mask, NULL);
3136 static struct inode_operations btrfs_dir_inode_operations = {
3137 .lookup = btrfs_lookup,
3138 .create = btrfs_create,
3139 .unlink = btrfs_unlink,
3141 .mkdir = btrfs_mkdir,
3142 .rmdir = btrfs_rmdir,
3143 .rename = btrfs_rename,
3144 .symlink = btrfs_symlink,
3145 .setattr = btrfs_setattr,
3146 .mknod = btrfs_mknod,
3147 .setxattr = generic_setxattr,
3148 .getxattr = generic_getxattr,
3149 .listxattr = btrfs_listxattr,
3150 .removexattr = generic_removexattr,
3151 .permission = btrfs_permission,
3153 static struct inode_operations btrfs_dir_ro_inode_operations = {
3154 .lookup = btrfs_lookup,
3155 .permission = btrfs_permission,
3157 static struct file_operations btrfs_dir_file_operations = {
3158 .llseek = generic_file_llseek,
3159 .read = generic_read_dir,
3160 .readdir = btrfs_readdir,
3161 .unlocked_ioctl = btrfs_ioctl,
3162 #ifdef CONFIG_COMPAT
3163 .compat_ioctl = btrfs_ioctl,
3165 .release = btrfs_release_file,
3168 static struct extent_io_ops btrfs_extent_io_ops = {
3169 .fill_delalloc = run_delalloc_range,
3170 .submit_bio_hook = btrfs_submit_bio_hook,
3171 .merge_bio_hook = btrfs_merge_bio_hook,
3172 .readpage_io_hook = btrfs_readpage_io_hook,
3173 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
3174 .writepage_end_io_hook = btrfs_writepage_end_io_hook,
3175 .writepage_start_hook = btrfs_writepage_start_hook,
3176 .readpage_io_failed_hook = btrfs_io_failed_hook,
3177 .set_bit_hook = btrfs_set_bit_hook,
3178 .clear_bit_hook = btrfs_clear_bit_hook,
3181 static struct address_space_operations btrfs_aops = {
3182 .readpage = btrfs_readpage,
3183 .writepage = btrfs_writepage,
3184 .writepages = btrfs_writepages,
3185 .readpages = btrfs_readpages,
3186 .sync_page = block_sync_page,
3188 .direct_IO = btrfs_direct_IO,
3189 .invalidatepage = btrfs_invalidatepage,
3190 .releasepage = btrfs_releasepage,
3191 .set_page_dirty = btrfs_set_page_dirty,
3194 static struct address_space_operations btrfs_symlink_aops = {
3195 .readpage = btrfs_readpage,
3196 .writepage = btrfs_writepage,
3197 .invalidatepage = btrfs_invalidatepage,
3198 .releasepage = btrfs_releasepage,
3201 static struct inode_operations btrfs_file_inode_operations = {
3202 .truncate = btrfs_truncate,
3203 .getattr = btrfs_getattr,
3204 .setattr = btrfs_setattr,
3205 .setxattr = generic_setxattr,
3206 .getxattr = generic_getxattr,
3207 .listxattr = btrfs_listxattr,
3208 .removexattr = generic_removexattr,
3209 .permission = btrfs_permission,
3211 static struct inode_operations btrfs_special_inode_operations = {
3212 .getattr = btrfs_getattr,
3213 .setattr = btrfs_setattr,
3214 .permission = btrfs_permission,
3216 static struct inode_operations btrfs_symlink_inode_operations = {
3217 .readlink = generic_readlink,
3218 .follow_link = page_follow_link_light,
3219 .put_link = page_put_link,
3220 .permission = btrfs_permission,