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 int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
407 struct extent_state *state, int uptodate)
409 struct inode *inode = page->mapping->host;
410 struct btrfs_root *root = BTRFS_I(inode)->root;
411 struct btrfs_trans_handle *trans;
412 struct btrfs_ordered_extent *ordered_extent;
413 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
415 struct list_head list;
416 struct btrfs_key ins;
419 ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1);
424 trans = btrfs_start_transaction(root, 1);
426 ordered_extent = btrfs_lookup_ordered_extent(inode, start);
427 BUG_ON(!ordered_extent);
429 lock_extent(io_tree, ordered_extent->file_offset,
430 ordered_extent->file_offset + ordered_extent->len - 1,
433 INIT_LIST_HEAD(&list);
435 ins.objectid = ordered_extent->start;
436 ins.offset = ordered_extent->len;
437 ins.type = BTRFS_EXTENT_ITEM_KEY;
438 ret = btrfs_alloc_reserved_extent(trans, root, root->root_key.objectid,
439 trans->transid, inode->i_ino,
440 ordered_extent->file_offset, &ins);
442 ret = btrfs_drop_extents(trans, root, inode,
443 ordered_extent->file_offset,
444 ordered_extent->file_offset +
446 ordered_extent->file_offset, &alloc_hint);
448 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
449 ordered_extent->file_offset,
450 ordered_extent->start,
452 ordered_extent->len, 0);
454 btrfs_drop_extent_cache(inode, ordered_extent->file_offset,
455 ordered_extent->file_offset +
456 ordered_extent->len - 1);
457 inode->i_blocks += ordered_extent->len >> 9;
458 unlock_extent(io_tree, ordered_extent->file_offset,
459 ordered_extent->file_offset + ordered_extent->len - 1,
461 add_pending_csums(trans, inode, ordered_extent->file_offset,
462 &ordered_extent->list);
464 btrfs_remove_ordered_extent(inode, ordered_extent);
466 btrfs_put_ordered_extent(ordered_extent);
467 /* once for the tree */
468 btrfs_put_ordered_extent(ordered_extent);
470 btrfs_update_inode(trans, root, inode);
471 btrfs_end_transaction(trans, root);
475 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
478 struct inode *inode = page->mapping->host;
479 struct btrfs_root *root = BTRFS_I(inode)->root;
480 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
481 struct btrfs_csum_item *item;
482 struct btrfs_path *path = NULL;
485 if (btrfs_test_opt(root, NODATASUM) ||
486 btrfs_test_flag(inode, NODATASUM))
489 path = btrfs_alloc_path();
490 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
493 /* a csum that isn't present is a preallocated region. */
494 if (ret == -ENOENT || ret == -EFBIG)
497 printk("no csum found for inode %lu start %Lu\n", inode->i_ino,
501 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
503 set_state_private(io_tree, start, csum);
506 btrfs_free_path(path);
510 struct io_failure_record {
518 int btrfs_io_failed_hook(struct bio *failed_bio,
519 struct page *page, u64 start, u64 end,
520 struct extent_state *state)
522 struct io_failure_record *failrec = NULL;
524 struct extent_map *em;
525 struct inode *inode = page->mapping->host;
526 struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
527 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
534 ret = get_state_private(failure_tree, start, &private);
536 failrec = kmalloc(sizeof(*failrec), GFP_NOFS);
539 failrec->start = start;
540 failrec->len = end - start + 1;
541 failrec->last_mirror = 0;
543 spin_lock(&em_tree->lock);
544 em = lookup_extent_mapping(em_tree, start, failrec->len);
545 if (em->start > start || em->start + em->len < start) {
549 spin_unlock(&em_tree->lock);
551 if (!em || IS_ERR(em)) {
555 logical = start - em->start;
556 logical = em->block_start + logical;
557 failrec->logical = logical;
559 set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
560 EXTENT_DIRTY, GFP_NOFS);
561 set_state_private(failure_tree, start,
562 (u64)(unsigned long)failrec);
564 failrec = (struct io_failure_record *)(unsigned long)private;
566 num_copies = btrfs_num_copies(
567 &BTRFS_I(inode)->root->fs_info->mapping_tree,
568 failrec->logical, failrec->len);
569 failrec->last_mirror++;
571 spin_lock_irq(&BTRFS_I(inode)->io_tree.lock);
572 state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
575 if (state && state->start != failrec->start)
577 spin_unlock_irq(&BTRFS_I(inode)->io_tree.lock);
579 if (!state || failrec->last_mirror > num_copies) {
580 set_state_private(failure_tree, failrec->start, 0);
581 clear_extent_bits(failure_tree, failrec->start,
582 failrec->start + failrec->len - 1,
583 EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
587 bio = bio_alloc(GFP_NOFS, 1);
588 bio->bi_private = state;
589 bio->bi_end_io = failed_bio->bi_end_io;
590 bio->bi_sector = failrec->logical >> 9;
591 bio->bi_bdev = failed_bio->bi_bdev;
593 bio_add_page(bio, page, failrec->len, start - page_offset(page));
594 if (failed_bio->bi_rw & (1 << BIO_RW))
599 BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
600 failrec->last_mirror);
604 int btrfs_clean_io_failures(struct inode *inode, u64 start)
608 struct io_failure_record *failure;
612 if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
613 (u64)-1, 1, EXTENT_DIRTY)) {
614 ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
615 start, &private_failure);
617 failure = (struct io_failure_record *)(unsigned long)
619 set_state_private(&BTRFS_I(inode)->io_failure_tree,
621 clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
623 failure->start + failure->len - 1,
624 EXTENT_DIRTY | EXTENT_LOCKED,
632 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
633 struct extent_state *state)
635 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
636 struct inode *inode = page->mapping->host;
637 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
639 u64 private = ~(u32)0;
641 struct btrfs_root *root = BTRFS_I(inode)->root;
645 if (btrfs_test_opt(root, NODATASUM) ||
646 btrfs_test_flag(inode, NODATASUM))
648 if (state && state->start == start) {
649 private = state->private;
652 ret = get_state_private(io_tree, start, &private);
654 local_irq_save(flags);
655 kaddr = kmap_atomic(page, KM_IRQ0);
659 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
660 btrfs_csum_final(csum, (char *)&csum);
661 if (csum != private) {
664 kunmap_atomic(kaddr, KM_IRQ0);
665 local_irq_restore(flags);
667 /* if the io failure tree for this inode is non-empty,
668 * check to see if we've recovered from a failed IO
670 btrfs_clean_io_failures(inode, start);
674 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
675 page->mapping->host->i_ino, (unsigned long long)start, csum,
677 memset(kaddr + offset, 1, end - start + 1);
678 flush_dcache_page(page);
679 kunmap_atomic(kaddr, KM_IRQ0);
680 local_irq_restore(flags);
686 void btrfs_read_locked_inode(struct inode *inode)
688 struct btrfs_path *path;
689 struct extent_buffer *leaf;
690 struct btrfs_inode_item *inode_item;
691 struct btrfs_timespec *tspec;
692 struct btrfs_root *root = BTRFS_I(inode)->root;
693 struct btrfs_key location;
694 u64 alloc_group_block;
698 path = btrfs_alloc_path();
700 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
702 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
706 leaf = path->nodes[0];
707 inode_item = btrfs_item_ptr(leaf, path->slots[0],
708 struct btrfs_inode_item);
710 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
711 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
712 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
713 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
714 inode->i_size = btrfs_inode_size(leaf, inode_item);
716 tspec = btrfs_inode_atime(inode_item);
717 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
718 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
720 tspec = btrfs_inode_mtime(inode_item);
721 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
722 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
724 tspec = btrfs_inode_ctime(inode_item);
725 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
726 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
728 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
729 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
731 rdev = btrfs_inode_rdev(leaf, inode_item);
733 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
734 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
736 BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
737 if (!BTRFS_I(inode)->block_group) {
738 BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
740 BTRFS_BLOCK_GROUP_METADATA, 0);
742 btrfs_free_path(path);
745 switch (inode->i_mode & S_IFMT) {
747 inode->i_mapping->a_ops = &btrfs_aops;
748 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
749 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
750 inode->i_fop = &btrfs_file_operations;
751 inode->i_op = &btrfs_file_inode_operations;
754 inode->i_fop = &btrfs_dir_file_operations;
755 if (root == root->fs_info->tree_root)
756 inode->i_op = &btrfs_dir_ro_inode_operations;
758 inode->i_op = &btrfs_dir_inode_operations;
761 inode->i_op = &btrfs_symlink_inode_operations;
762 inode->i_mapping->a_ops = &btrfs_symlink_aops;
763 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
766 init_special_inode(inode, inode->i_mode, rdev);
772 btrfs_free_path(path);
773 make_bad_inode(inode);
776 static void fill_inode_item(struct extent_buffer *leaf,
777 struct btrfs_inode_item *item,
780 btrfs_set_inode_uid(leaf, item, inode->i_uid);
781 btrfs_set_inode_gid(leaf, item, inode->i_gid);
782 btrfs_set_inode_size(leaf, item, inode->i_size);
783 btrfs_set_inode_mode(leaf, item, inode->i_mode);
784 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
786 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
787 inode->i_atime.tv_sec);
788 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
789 inode->i_atime.tv_nsec);
791 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
792 inode->i_mtime.tv_sec);
793 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
794 inode->i_mtime.tv_nsec);
796 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
797 inode->i_ctime.tv_sec);
798 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
799 inode->i_ctime.tv_nsec);
801 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
802 btrfs_set_inode_generation(leaf, item, inode->i_generation);
803 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
804 btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
805 btrfs_set_inode_block_group(leaf, item,
806 BTRFS_I(inode)->block_group->key.objectid);
809 int btrfs_update_inode(struct btrfs_trans_handle *trans,
810 struct btrfs_root *root,
813 struct btrfs_inode_item *inode_item;
814 struct btrfs_path *path;
815 struct extent_buffer *leaf;
818 path = btrfs_alloc_path();
820 ret = btrfs_lookup_inode(trans, root, path,
821 &BTRFS_I(inode)->location, 1);
828 leaf = path->nodes[0];
829 inode_item = btrfs_item_ptr(leaf, path->slots[0],
830 struct btrfs_inode_item);
832 fill_inode_item(leaf, inode_item, inode);
833 btrfs_mark_buffer_dirty(leaf);
834 btrfs_set_inode_last_trans(trans, inode);
837 btrfs_free_path(path);
842 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
843 struct btrfs_root *root,
845 struct dentry *dentry)
847 struct btrfs_path *path;
848 const char *name = dentry->d_name.name;
849 int name_len = dentry->d_name.len;
851 struct extent_buffer *leaf;
852 struct btrfs_dir_item *di;
853 struct btrfs_key key;
855 path = btrfs_alloc_path();
861 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
871 leaf = path->nodes[0];
872 btrfs_dir_item_key_to_cpu(leaf, di, &key);
873 ret = btrfs_delete_one_dir_name(trans, root, path, di);
876 btrfs_release_path(root, path);
878 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
879 key.objectid, name, name_len, -1);
888 ret = btrfs_delete_one_dir_name(trans, root, path, di);
889 btrfs_release_path(root, path);
891 dentry->d_inode->i_ctime = dir->i_ctime;
892 ret = btrfs_del_inode_ref(trans, root, name, name_len,
893 dentry->d_inode->i_ino,
894 dentry->d_parent->d_inode->i_ino);
896 printk("failed to delete reference to %.*s, "
897 "inode %lu parent %lu\n", name_len, name,
898 dentry->d_inode->i_ino,
899 dentry->d_parent->d_inode->i_ino);
902 btrfs_free_path(path);
904 dir->i_size -= name_len * 2;
905 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
906 btrfs_update_inode(trans, root, dir);
907 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
908 dentry->d_inode->i_nlink--;
910 drop_nlink(dentry->d_inode);
912 ret = btrfs_update_inode(trans, root, dentry->d_inode);
913 dir->i_sb->s_dirt = 1;
918 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
920 struct btrfs_root *root;
921 struct btrfs_trans_handle *trans;
923 unsigned long nr = 0;
925 root = BTRFS_I(dir)->root;
927 ret = btrfs_check_free_space(root, 1, 1);
931 trans = btrfs_start_transaction(root, 1);
933 btrfs_set_trans_block_group(trans, dir);
934 ret = btrfs_unlink_trans(trans, root, dir, dentry);
935 nr = trans->blocks_used;
937 btrfs_end_transaction_throttle(trans, root);
939 btrfs_btree_balance_dirty(root, nr);
943 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
945 struct inode *inode = dentry->d_inode;
948 struct btrfs_root *root = BTRFS_I(dir)->root;
949 struct btrfs_trans_handle *trans;
950 unsigned long nr = 0;
952 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
956 ret = btrfs_check_free_space(root, 1, 1);
960 trans = btrfs_start_transaction(root, 1);
961 btrfs_set_trans_block_group(trans, dir);
963 /* now the directory is empty */
964 err = btrfs_unlink_trans(trans, root, dir, dentry);
969 nr = trans->blocks_used;
970 ret = btrfs_end_transaction_throttle(trans, root);
972 btrfs_btree_balance_dirty(root, nr);
980 * this can truncate away extent items, csum items and directory items.
981 * It starts at a high offset and removes keys until it can't find
982 * any higher than i_size.
984 * csum items that cross the new i_size are truncated to the new size
987 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
988 struct btrfs_root *root,
993 struct btrfs_path *path;
994 struct btrfs_key key;
995 struct btrfs_key found_key;
997 struct extent_buffer *leaf;
998 struct btrfs_file_extent_item *fi;
999 u64 extent_start = 0;
1000 u64 extent_num_bytes = 0;
1006 int pending_del_nr = 0;
1007 int pending_del_slot = 0;
1008 int extent_type = -1;
1009 u64 mask = root->sectorsize - 1;
1011 btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
1012 btrfs_drop_extent_cache(inode, inode->i_size & (~mask), (u64)-1);
1013 path = btrfs_alloc_path();
1017 /* FIXME, add redo link to tree so we don't leak on crash */
1018 key.objectid = inode->i_ino;
1019 key.offset = (u64)-1;
1022 btrfs_init_path(path);
1024 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1029 BUG_ON(path->slots[0] == 0);
1035 leaf = path->nodes[0];
1036 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1037 found_type = btrfs_key_type(&found_key);
1039 if (found_key.objectid != inode->i_ino)
1042 if (found_type < min_type)
1045 item_end = found_key.offset;
1046 if (found_type == BTRFS_EXTENT_DATA_KEY) {
1047 fi = btrfs_item_ptr(leaf, path->slots[0],
1048 struct btrfs_file_extent_item);
1049 extent_type = btrfs_file_extent_type(leaf, fi);
1050 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
1052 btrfs_file_extent_num_bytes(leaf, fi);
1053 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1054 struct btrfs_item *item = btrfs_item_nr(leaf,
1056 item_end += btrfs_file_extent_inline_len(leaf,
1061 if (found_type == BTRFS_CSUM_ITEM_KEY) {
1062 ret = btrfs_csum_truncate(trans, root, path,
1066 if (item_end < inode->i_size) {
1067 if (found_type == BTRFS_DIR_ITEM_KEY) {
1068 found_type = BTRFS_INODE_ITEM_KEY;
1069 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
1070 found_type = BTRFS_CSUM_ITEM_KEY;
1071 } else if (found_type == BTRFS_EXTENT_DATA_KEY) {
1072 found_type = BTRFS_XATTR_ITEM_KEY;
1073 } else if (found_type == BTRFS_XATTR_ITEM_KEY) {
1074 found_type = BTRFS_INODE_REF_KEY;
1075 } else if (found_type) {
1080 btrfs_set_key_type(&key, found_type);
1083 if (found_key.offset >= inode->i_size)
1089 /* FIXME, shrink the extent if the ref count is only 1 */
1090 if (found_type != BTRFS_EXTENT_DATA_KEY)
1093 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
1095 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
1097 u64 orig_num_bytes =
1098 btrfs_file_extent_num_bytes(leaf, fi);
1099 extent_num_bytes = inode->i_size -
1100 found_key.offset + root->sectorsize - 1;
1101 extent_num_bytes = extent_num_bytes &
1102 ~((u64)root->sectorsize - 1);
1103 btrfs_set_file_extent_num_bytes(leaf, fi,
1105 num_dec = (orig_num_bytes -
1107 if (extent_start != 0)
1108 dec_i_blocks(inode, num_dec);
1109 btrfs_mark_buffer_dirty(leaf);
1112 btrfs_file_extent_disk_num_bytes(leaf,
1114 /* FIXME blocksize != 4096 */
1115 num_dec = btrfs_file_extent_num_bytes(leaf, fi);
1116 if (extent_start != 0) {
1118 dec_i_blocks(inode, num_dec);
1120 root_gen = btrfs_header_generation(leaf);
1121 root_owner = btrfs_header_owner(leaf);
1123 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1125 u32 newsize = inode->i_size - found_key.offset;
1126 dec_i_blocks(inode, item_end + 1 -
1127 found_key.offset - newsize);
1129 btrfs_file_extent_calc_inline_size(newsize);
1130 ret = btrfs_truncate_item(trans, root, path,
1134 dec_i_blocks(inode, item_end + 1 -
1140 if (!pending_del_nr) {
1141 /* no pending yet, add ourselves */
1142 pending_del_slot = path->slots[0];
1144 } else if (pending_del_nr &&
1145 path->slots[0] + 1 == pending_del_slot) {
1146 /* hop on the pending chunk */
1148 pending_del_slot = path->slots[0];
1150 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot);
1156 ret = btrfs_free_extent(trans, root, extent_start,
1159 root_gen, inode->i_ino,
1160 found_key.offset, 0);
1164 if (path->slots[0] == 0) {
1167 btrfs_release_path(root, path);
1172 if (pending_del_nr &&
1173 path->slots[0] + 1 != pending_del_slot) {
1174 struct btrfs_key debug;
1176 btrfs_item_key_to_cpu(path->nodes[0], &debug,
1178 ret = btrfs_del_items(trans, root, path,
1183 btrfs_release_path(root, path);
1189 if (pending_del_nr) {
1190 ret = btrfs_del_items(trans, root, path, pending_del_slot,
1193 btrfs_free_path(path);
1194 inode->i_sb->s_dirt = 1;
1199 * taken from block_truncate_page, but does cow as it zeros out
1200 * any bytes left in the last page in the file.
1202 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
1204 struct inode *inode = mapping->host;
1205 struct btrfs_root *root = BTRFS_I(inode)->root;
1206 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1207 struct btrfs_ordered_extent *ordered;
1209 u32 blocksize = root->sectorsize;
1210 pgoff_t index = from >> PAGE_CACHE_SHIFT;
1211 unsigned offset = from & (PAGE_CACHE_SIZE-1);
1217 if ((offset & (blocksize - 1)) == 0)
1222 page = grab_cache_page(mapping, index);
1226 page_start = page_offset(page);
1227 page_end = page_start + PAGE_CACHE_SIZE - 1;
1229 if (!PageUptodate(page)) {
1230 ret = btrfs_readpage(NULL, page);
1232 if (page->mapping != mapping) {
1234 page_cache_release(page);
1237 if (!PageUptodate(page)) {
1242 wait_on_page_writeback(page);
1244 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
1245 set_page_extent_mapped(page);
1247 ordered = btrfs_lookup_ordered_extent(inode, page_start);
1249 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
1251 page_cache_release(page);
1252 btrfs_wait_ordered_extent(inode, ordered);
1253 btrfs_put_ordered_extent(ordered);
1257 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
1258 page_end, GFP_NOFS);
1260 if (offset != PAGE_CACHE_SIZE) {
1262 memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
1263 flush_dcache_page(page);
1266 set_page_dirty(page);
1267 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
1270 page_cache_release(page);
1275 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
1277 struct inode *inode = dentry->d_inode;
1280 err = inode_change_ok(inode, attr);
1284 if (S_ISREG(inode->i_mode) &&
1285 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
1286 struct btrfs_trans_handle *trans;
1287 struct btrfs_root *root = BTRFS_I(inode)->root;
1288 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1290 u64 mask = root->sectorsize - 1;
1291 u64 hole_start = (inode->i_size + mask) & ~mask;
1292 u64 block_end = (attr->ia_size + mask) & ~mask;
1296 if (attr->ia_size <= hole_start)
1299 err = btrfs_check_free_space(root, 1, 0);
1303 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1305 hole_size = block_end - hole_start;
1306 btrfs_wait_ordered_range(inode, hole_start, hole_size);
1307 lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1309 trans = btrfs_start_transaction(root, 1);
1310 btrfs_set_trans_block_group(trans, inode);
1311 err = btrfs_drop_extents(trans, root, inode,
1312 hole_start, block_end, hole_start,
1315 if (alloc_hint != EXTENT_MAP_INLINE) {
1316 err = btrfs_insert_file_extent(trans, root,
1320 btrfs_drop_extent_cache(inode, hole_start,
1322 btrfs_check_file(root, inode);
1324 btrfs_end_transaction(trans, root);
1325 unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1330 err = inode_setattr(inode, attr);
1335 void btrfs_delete_inode(struct inode *inode)
1337 struct btrfs_trans_handle *trans;
1338 struct btrfs_root *root = BTRFS_I(inode)->root;
1342 btrfs_wait_ordered_range(inode, 0, (u64)-1);
1343 truncate_inode_pages(&inode->i_data, 0);
1344 if (is_bad_inode(inode)) {
1349 trans = btrfs_start_transaction(root, 1);
1351 btrfs_set_trans_block_group(trans, inode);
1352 ret = btrfs_truncate_in_trans(trans, root, inode, 0);
1354 goto no_delete_lock;
1356 nr = trans->blocks_used;
1359 btrfs_end_transaction(trans, root);
1360 btrfs_btree_balance_dirty(root, nr);
1364 nr = trans->blocks_used;
1365 btrfs_end_transaction(trans, root);
1366 btrfs_btree_balance_dirty(root, nr);
1372 * this returns the key found in the dir entry in the location pointer.
1373 * If no dir entries were found, location->objectid is 0.
1375 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
1376 struct btrfs_key *location)
1378 const char *name = dentry->d_name.name;
1379 int namelen = dentry->d_name.len;
1380 struct btrfs_dir_item *di;
1381 struct btrfs_path *path;
1382 struct btrfs_root *root = BTRFS_I(dir)->root;
1385 if (namelen == 1 && strcmp(name, ".") == 0) {
1386 location->objectid = dir->i_ino;
1387 location->type = BTRFS_INODE_ITEM_KEY;
1388 location->offset = 0;
1391 path = btrfs_alloc_path();
1394 if (namelen == 2 && strcmp(name, "..") == 0) {
1395 struct btrfs_key key;
1396 struct extent_buffer *leaf;
1400 key.objectid = dir->i_ino;
1401 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1403 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1407 leaf = path->nodes[0];
1408 slot = path->slots[0];
1409 nritems = btrfs_header_nritems(leaf);
1410 if (slot >= nritems)
1413 btrfs_item_key_to_cpu(leaf, &key, slot);
1414 if (key.objectid != dir->i_ino ||
1415 key.type != BTRFS_INODE_REF_KEY) {
1418 location->objectid = key.offset;
1419 location->type = BTRFS_INODE_ITEM_KEY;
1420 location->offset = 0;
1424 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
1428 if (!di || IS_ERR(di)) {
1431 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
1433 btrfs_free_path(path);
1436 location->objectid = 0;
1441 * when we hit a tree root in a directory, the btrfs part of the inode
1442 * needs to be changed to reflect the root directory of the tree root. This
1443 * is kind of like crossing a mount point.
1445 static int fixup_tree_root_location(struct btrfs_root *root,
1446 struct btrfs_key *location,
1447 struct btrfs_root **sub_root,
1448 struct dentry *dentry)
1450 struct btrfs_path *path;
1451 struct btrfs_root_item *ri;
1453 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
1455 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1458 path = btrfs_alloc_path();
1461 *sub_root = btrfs_read_fs_root(root->fs_info, location,
1462 dentry->d_name.name,
1463 dentry->d_name.len);
1464 if (IS_ERR(*sub_root))
1465 return PTR_ERR(*sub_root);
1467 ri = &(*sub_root)->root_item;
1468 location->objectid = btrfs_root_dirid(ri);
1469 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1470 location->offset = 0;
1472 btrfs_free_path(path);
1476 static int btrfs_init_locked_inode(struct inode *inode, void *p)
1478 struct btrfs_iget_args *args = p;
1479 inode->i_ino = args->ino;
1480 BTRFS_I(inode)->root = args->root;
1481 BTRFS_I(inode)->delalloc_bytes = 0;
1482 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1483 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1484 inode->i_mapping, GFP_NOFS);
1485 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1486 inode->i_mapping, GFP_NOFS);
1487 mutex_init(&BTRFS_I(inode)->csum_mutex);
1491 static int btrfs_find_actor(struct inode *inode, void *opaque)
1493 struct btrfs_iget_args *args = opaque;
1494 return (args->ino == inode->i_ino &&
1495 args->root == BTRFS_I(inode)->root);
1498 struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
1501 struct btrfs_iget_args args;
1502 args.ino = objectid;
1503 args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
1508 return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
1511 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
1512 struct btrfs_root *root)
1514 struct inode *inode;
1515 struct btrfs_iget_args args;
1516 args.ino = objectid;
1519 inode = iget5_locked(s, objectid, btrfs_find_actor,
1520 btrfs_init_locked_inode,
1525 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
1526 struct nameidata *nd)
1528 struct inode * inode;
1529 struct btrfs_inode *bi = BTRFS_I(dir);
1530 struct btrfs_root *root = bi->root;
1531 struct btrfs_root *sub_root = root;
1532 struct btrfs_key location;
1535 if (dentry->d_name.len > BTRFS_NAME_LEN)
1536 return ERR_PTR(-ENAMETOOLONG);
1538 ret = btrfs_inode_by_name(dir, dentry, &location);
1541 return ERR_PTR(ret);
1544 if (location.objectid) {
1545 ret = fixup_tree_root_location(root, &location, &sub_root,
1548 return ERR_PTR(ret);
1550 return ERR_PTR(-ENOENT);
1551 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
1554 return ERR_PTR(-EACCES);
1555 if (inode->i_state & I_NEW) {
1556 /* the inode and parent dir are two different roots */
1557 if (sub_root != root) {
1559 sub_root->inode = inode;
1561 BTRFS_I(inode)->root = sub_root;
1562 memcpy(&BTRFS_I(inode)->location, &location,
1564 btrfs_read_locked_inode(inode);
1565 unlock_new_inode(inode);
1568 return d_splice_alias(inode, dentry);
1571 static unsigned char btrfs_filetype_table[] = {
1572 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1575 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1577 struct inode *inode = filp->f_dentry->d_inode;
1578 struct btrfs_root *root = BTRFS_I(inode)->root;
1579 struct btrfs_item *item;
1580 struct btrfs_dir_item *di;
1581 struct btrfs_key key;
1582 struct btrfs_key found_key;
1583 struct btrfs_path *path;
1586 struct extent_buffer *leaf;
1589 unsigned char d_type;
1594 int key_type = BTRFS_DIR_INDEX_KEY;
1599 /* FIXME, use a real flag for deciding about the key type */
1600 if (root->fs_info->tree_root == root)
1601 key_type = BTRFS_DIR_ITEM_KEY;
1603 /* special case for "." */
1604 if (filp->f_pos == 0) {
1605 over = filldir(dirent, ".", 1,
1613 key.objectid = inode->i_ino;
1614 path = btrfs_alloc_path();
1617 /* special case for .., just use the back ref */
1618 if (filp->f_pos == 1) {
1619 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1621 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1623 leaf = path->nodes[0];
1624 slot = path->slots[0];
1625 nritems = btrfs_header_nritems(leaf);
1626 if (slot >= nritems) {
1627 btrfs_release_path(root, path);
1628 goto read_dir_items;
1630 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1631 btrfs_release_path(root, path);
1632 if (found_key.objectid != key.objectid ||
1633 found_key.type != BTRFS_INODE_REF_KEY)
1634 goto read_dir_items;
1635 over = filldir(dirent, "..", 2,
1636 2, found_key.offset, DT_DIR);
1643 btrfs_set_key_type(&key, key_type);
1644 key.offset = filp->f_pos;
1646 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1651 leaf = path->nodes[0];
1652 nritems = btrfs_header_nritems(leaf);
1653 slot = path->slots[0];
1654 if (advance || slot >= nritems) {
1655 if (slot >= nritems -1) {
1656 ret = btrfs_next_leaf(root, path);
1659 leaf = path->nodes[0];
1660 nritems = btrfs_header_nritems(leaf);
1661 slot = path->slots[0];
1668 item = btrfs_item_nr(leaf, slot);
1669 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1671 if (found_key.objectid != key.objectid)
1673 if (btrfs_key_type(&found_key) != key_type)
1675 if (found_key.offset < filp->f_pos)
1678 filp->f_pos = found_key.offset;
1680 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1682 di_total = btrfs_item_size(leaf, item);
1683 while(di_cur < di_total) {
1684 struct btrfs_key location;
1686 name_len = btrfs_dir_name_len(leaf, di);
1687 if (name_len < 32) {
1688 name_ptr = tmp_name;
1690 name_ptr = kmalloc(name_len, GFP_NOFS);
1693 read_extent_buffer(leaf, name_ptr,
1694 (unsigned long)(di + 1), name_len);
1696 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1697 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1698 over = filldir(dirent, name_ptr, name_len,
1703 if (name_ptr != tmp_name)
1708 di_len = btrfs_dir_name_len(leaf, di) +
1709 btrfs_dir_data_len(leaf, di) +sizeof(*di);
1711 di = (struct btrfs_dir_item *)((char *)di + di_len);
1714 if (key_type == BTRFS_DIR_INDEX_KEY)
1715 filp->f_pos = INT_LIMIT(typeof(filp->f_pos));
1721 btrfs_free_path(path);
1725 int btrfs_write_inode(struct inode *inode, int wait)
1727 struct btrfs_root *root = BTRFS_I(inode)->root;
1728 struct btrfs_trans_handle *trans;
1732 trans = btrfs_start_transaction(root, 1);
1733 btrfs_set_trans_block_group(trans, inode);
1734 ret = btrfs_commit_transaction(trans, root);
1740 * This is somewhat expensive, updating the tree every time the
1741 * inode changes. But, it is most likely to find the inode in cache.
1742 * FIXME, needs more benchmarking...there are no reasons other than performance
1743 * to keep or drop this code.
1745 void btrfs_dirty_inode(struct inode *inode)
1747 struct btrfs_root *root = BTRFS_I(inode)->root;
1748 struct btrfs_trans_handle *trans;
1750 trans = btrfs_start_transaction(root, 1);
1751 btrfs_set_trans_block_group(trans, inode);
1752 btrfs_update_inode(trans, root, inode);
1753 btrfs_end_transaction(trans, root);
1756 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1757 struct btrfs_root *root,
1758 const char *name, int name_len,
1761 struct btrfs_block_group_cache *group,
1764 struct inode *inode;
1765 struct btrfs_inode_item *inode_item;
1766 struct btrfs_block_group_cache *new_inode_group;
1767 struct btrfs_key *location;
1768 struct btrfs_path *path;
1769 struct btrfs_inode_ref *ref;
1770 struct btrfs_key key[2];
1776 path = btrfs_alloc_path();
1779 inode = new_inode(root->fs_info->sb);
1781 return ERR_PTR(-ENOMEM);
1783 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1784 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1785 inode->i_mapping, GFP_NOFS);
1786 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1787 inode->i_mapping, GFP_NOFS);
1788 mutex_init(&BTRFS_I(inode)->csum_mutex);
1789 BTRFS_I(inode)->delalloc_bytes = 0;
1790 BTRFS_I(inode)->root = root;
1796 new_inode_group = btrfs_find_block_group(root, group, 0,
1797 BTRFS_BLOCK_GROUP_METADATA, owner);
1798 if (!new_inode_group) {
1799 printk("find_block group failed\n");
1800 new_inode_group = group;
1802 BTRFS_I(inode)->block_group = new_inode_group;
1803 BTRFS_I(inode)->flags = 0;
1805 key[0].objectid = objectid;
1806 btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
1809 key[1].objectid = objectid;
1810 btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
1811 key[1].offset = ref_objectid;
1813 sizes[0] = sizeof(struct btrfs_inode_item);
1814 sizes[1] = name_len + sizeof(*ref);
1816 ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
1820 if (objectid > root->highest_inode)
1821 root->highest_inode = objectid;
1823 inode->i_uid = current->fsuid;
1824 inode->i_gid = current->fsgid;
1825 inode->i_mode = mode;
1826 inode->i_ino = objectid;
1827 inode->i_blocks = 0;
1828 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1829 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1830 struct btrfs_inode_item);
1831 fill_inode_item(path->nodes[0], inode_item, inode);
1833 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
1834 struct btrfs_inode_ref);
1835 btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
1836 ptr = (unsigned long)(ref + 1);
1837 write_extent_buffer(path->nodes[0], name, ptr, name_len);
1839 btrfs_mark_buffer_dirty(path->nodes[0]);
1840 btrfs_free_path(path);
1842 location = &BTRFS_I(inode)->location;
1843 location->objectid = objectid;
1844 location->offset = 0;
1845 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1847 insert_inode_hash(inode);
1850 btrfs_free_path(path);
1851 return ERR_PTR(ret);
1854 static inline u8 btrfs_inode_type(struct inode *inode)
1856 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1859 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1860 struct dentry *dentry, struct inode *inode,
1864 struct btrfs_key key;
1865 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1866 struct inode *parent_inode;
1868 key.objectid = inode->i_ino;
1869 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1872 ret = btrfs_insert_dir_item(trans, root,
1873 dentry->d_name.name, dentry->d_name.len,
1874 dentry->d_parent->d_inode->i_ino,
1875 &key, btrfs_inode_type(inode));
1878 ret = btrfs_insert_inode_ref(trans, root,
1879 dentry->d_name.name,
1882 dentry->d_parent->d_inode->i_ino);
1884 parent_inode = dentry->d_parent->d_inode;
1885 parent_inode->i_size += dentry->d_name.len * 2;
1886 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1887 ret = btrfs_update_inode(trans, root,
1888 dentry->d_parent->d_inode);
1893 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1894 struct dentry *dentry, struct inode *inode,
1897 int err = btrfs_add_link(trans, dentry, inode, backref);
1899 d_instantiate(dentry, inode);
1907 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1908 int mode, dev_t rdev)
1910 struct btrfs_trans_handle *trans;
1911 struct btrfs_root *root = BTRFS_I(dir)->root;
1912 struct inode *inode = NULL;
1916 unsigned long nr = 0;
1918 if (!new_valid_dev(rdev))
1921 err = btrfs_check_free_space(root, 1, 0);
1925 trans = btrfs_start_transaction(root, 1);
1926 btrfs_set_trans_block_group(trans, dir);
1928 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1934 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1936 dentry->d_parent->d_inode->i_ino, objectid,
1937 BTRFS_I(dir)->block_group, mode);
1938 err = PTR_ERR(inode);
1942 btrfs_set_trans_block_group(trans, inode);
1943 err = btrfs_add_nondir(trans, dentry, inode, 0);
1947 inode->i_op = &btrfs_special_inode_operations;
1948 init_special_inode(inode, inode->i_mode, rdev);
1949 btrfs_update_inode(trans, root, inode);
1951 dir->i_sb->s_dirt = 1;
1952 btrfs_update_inode_block_group(trans, inode);
1953 btrfs_update_inode_block_group(trans, dir);
1955 nr = trans->blocks_used;
1956 btrfs_end_transaction_throttle(trans, root);
1959 inode_dec_link_count(inode);
1962 btrfs_btree_balance_dirty(root, nr);
1966 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1967 int mode, struct nameidata *nd)
1969 struct btrfs_trans_handle *trans;
1970 struct btrfs_root *root = BTRFS_I(dir)->root;
1971 struct inode *inode = NULL;
1974 unsigned long nr = 0;
1977 err = btrfs_check_free_space(root, 1, 0);
1980 trans = btrfs_start_transaction(root, 1);
1981 btrfs_set_trans_block_group(trans, dir);
1983 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1989 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1991 dentry->d_parent->d_inode->i_ino,
1992 objectid, BTRFS_I(dir)->block_group, mode);
1993 err = PTR_ERR(inode);
1997 btrfs_set_trans_block_group(trans, inode);
1998 err = btrfs_add_nondir(trans, dentry, inode, 0);
2002 inode->i_mapping->a_ops = &btrfs_aops;
2003 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
2004 inode->i_fop = &btrfs_file_operations;
2005 inode->i_op = &btrfs_file_inode_operations;
2006 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2007 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
2008 inode->i_mapping, GFP_NOFS);
2009 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
2010 inode->i_mapping, GFP_NOFS);
2011 mutex_init(&BTRFS_I(inode)->csum_mutex);
2012 BTRFS_I(inode)->delalloc_bytes = 0;
2013 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
2015 dir->i_sb->s_dirt = 1;
2016 btrfs_update_inode_block_group(trans, inode);
2017 btrfs_update_inode_block_group(trans, dir);
2019 nr = trans->blocks_used;
2020 btrfs_end_transaction_throttle(trans, root);
2023 inode_dec_link_count(inode);
2026 btrfs_btree_balance_dirty(root, nr);
2030 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
2031 struct dentry *dentry)
2033 struct btrfs_trans_handle *trans;
2034 struct btrfs_root *root = BTRFS_I(dir)->root;
2035 struct inode *inode = old_dentry->d_inode;
2036 unsigned long nr = 0;
2040 if (inode->i_nlink == 0)
2043 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2048 err = btrfs_check_free_space(root, 1, 0);
2051 trans = btrfs_start_transaction(root, 1);
2053 btrfs_set_trans_block_group(trans, dir);
2054 atomic_inc(&inode->i_count);
2055 err = btrfs_add_nondir(trans, dentry, inode, 1);
2060 dir->i_sb->s_dirt = 1;
2061 btrfs_update_inode_block_group(trans, dir);
2062 err = btrfs_update_inode(trans, root, inode);
2067 nr = trans->blocks_used;
2068 btrfs_end_transaction_throttle(trans, root);
2071 inode_dec_link_count(inode);
2074 btrfs_btree_balance_dirty(root, nr);
2078 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
2080 struct inode *inode = NULL;
2081 struct btrfs_trans_handle *trans;
2082 struct btrfs_root *root = BTRFS_I(dir)->root;
2084 int drop_on_err = 0;
2086 unsigned long nr = 1;
2088 err = btrfs_check_free_space(root, 1, 0);
2092 trans = btrfs_start_transaction(root, 1);
2093 btrfs_set_trans_block_group(trans, dir);
2095 if (IS_ERR(trans)) {
2096 err = PTR_ERR(trans);
2100 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2106 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2108 dentry->d_parent->d_inode->i_ino, objectid,
2109 BTRFS_I(dir)->block_group, S_IFDIR | mode);
2110 if (IS_ERR(inode)) {
2111 err = PTR_ERR(inode);
2116 inode->i_op = &btrfs_dir_inode_operations;
2117 inode->i_fop = &btrfs_dir_file_operations;
2118 btrfs_set_trans_block_group(trans, inode);
2121 err = btrfs_update_inode(trans, root, inode);
2125 err = btrfs_add_link(trans, dentry, inode, 0);
2129 d_instantiate(dentry, inode);
2131 dir->i_sb->s_dirt = 1;
2132 btrfs_update_inode_block_group(trans, inode);
2133 btrfs_update_inode_block_group(trans, dir);
2136 nr = trans->blocks_used;
2137 btrfs_end_transaction_throttle(trans, root);
2142 btrfs_btree_balance_dirty(root, nr);
2146 static int merge_extent_mapping(struct extent_map_tree *em_tree,
2147 struct extent_map *existing,
2148 struct extent_map *em,
2149 u64 map_start, u64 map_len)
2153 BUG_ON(map_start < em->start || map_start >= extent_map_end(em));
2154 start_diff = map_start - em->start;
2155 em->start = map_start;
2157 if (em->block_start < EXTENT_MAP_LAST_BYTE)
2158 em->block_start += start_diff;
2159 return add_extent_mapping(em_tree, em);
2162 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
2163 size_t pg_offset, u64 start, u64 len,
2169 u64 extent_start = 0;
2171 u64 objectid = inode->i_ino;
2173 struct btrfs_path *path;
2174 struct btrfs_root *root = BTRFS_I(inode)->root;
2175 struct btrfs_file_extent_item *item;
2176 struct extent_buffer *leaf;
2177 struct btrfs_key found_key;
2178 struct extent_map *em = NULL;
2179 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2180 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2181 struct btrfs_trans_handle *trans = NULL;
2183 path = btrfs_alloc_path();
2187 spin_lock(&em_tree->lock);
2188 em = lookup_extent_mapping(em_tree, start, len);
2190 em->bdev = root->fs_info->fs_devices->latest_bdev;
2191 spin_unlock(&em_tree->lock);
2194 if (em->start > start || em->start + em->len <= start)
2195 free_extent_map(em);
2196 else if (em->block_start == EXTENT_MAP_INLINE && page)
2197 free_extent_map(em);
2201 em = alloc_extent_map(GFP_NOFS);
2206 em->bdev = root->fs_info->fs_devices->latest_bdev;
2207 em->start = EXTENT_MAP_HOLE;
2209 ret = btrfs_lookup_file_extent(trans, root, path,
2210 objectid, start, trans != NULL);
2217 if (path->slots[0] == 0)
2222 leaf = path->nodes[0];
2223 item = btrfs_item_ptr(leaf, path->slots[0],
2224 struct btrfs_file_extent_item);
2225 /* are we inside the extent that was found? */
2226 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2227 found_type = btrfs_key_type(&found_key);
2228 if (found_key.objectid != objectid ||
2229 found_type != BTRFS_EXTENT_DATA_KEY) {
2233 found_type = btrfs_file_extent_type(leaf, item);
2234 extent_start = found_key.offset;
2235 if (found_type == BTRFS_FILE_EXTENT_REG) {
2236 extent_end = extent_start +
2237 btrfs_file_extent_num_bytes(leaf, item);
2239 if (start < extent_start || start >= extent_end) {
2241 if (start < extent_start) {
2242 if (start + len <= extent_start)
2244 em->len = extent_end - extent_start;
2250 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
2252 em->start = extent_start;
2253 em->len = extent_end - extent_start;
2254 em->block_start = EXTENT_MAP_HOLE;
2257 bytenr += btrfs_file_extent_offset(leaf, item);
2258 em->block_start = bytenr;
2259 em->start = extent_start;
2260 em->len = extent_end - extent_start;
2262 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
2267 size_t extent_offset;
2270 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
2272 extent_end = (extent_start + size + root->sectorsize - 1) &
2273 ~((u64)root->sectorsize - 1);
2274 if (start < extent_start || start >= extent_end) {
2276 if (start < extent_start) {
2277 if (start + len <= extent_start)
2279 em->len = extent_end - extent_start;
2285 em->block_start = EXTENT_MAP_INLINE;
2288 em->start = extent_start;
2293 page_start = page_offset(page) + pg_offset;
2294 extent_offset = page_start - extent_start;
2295 copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
2296 size - extent_offset);
2297 em->start = extent_start + extent_offset;
2298 em->len = (copy_size + root->sectorsize - 1) &
2299 ~((u64)root->sectorsize - 1);
2301 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
2302 if (create == 0 && !PageUptodate(page)) {
2303 read_extent_buffer(leaf, map + pg_offset, ptr,
2305 flush_dcache_page(page);
2306 } else if (create && PageUptodate(page)) {
2309 free_extent_map(em);
2311 btrfs_release_path(root, path);
2312 trans = btrfs_start_transaction(root, 1);
2315 write_extent_buffer(leaf, map + pg_offset, ptr,
2317 btrfs_mark_buffer_dirty(leaf);
2320 set_extent_uptodate(io_tree, em->start,
2321 extent_map_end(em) - 1, GFP_NOFS);
2324 printk("unkknown found_type %d\n", found_type);
2331 em->block_start = EXTENT_MAP_HOLE;
2333 btrfs_release_path(root, path);
2334 if (em->start > start || extent_map_end(em) <= start) {
2335 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len);
2341 spin_lock(&em_tree->lock);
2342 ret = add_extent_mapping(em_tree, em);
2343 /* it is possible that someone inserted the extent into the tree
2344 * while we had the lock dropped. It is also possible that
2345 * an overlapping map exists in the tree
2347 if (ret == -EEXIST) {
2348 struct extent_map *existing;
2352 existing = lookup_extent_mapping(em_tree, start, len);
2353 if (existing && (existing->start > start ||
2354 existing->start + existing->len <= start)) {
2355 free_extent_map(existing);
2359 existing = lookup_extent_mapping(em_tree, em->start,
2362 err = merge_extent_mapping(em_tree, existing,
2365 free_extent_map(existing);
2367 free_extent_map(em);
2372 printk("failing to insert %Lu %Lu\n",
2374 free_extent_map(em);
2378 free_extent_map(em);
2383 spin_unlock(&em_tree->lock);
2385 btrfs_free_path(path);
2387 ret = btrfs_end_transaction(trans, root);
2393 free_extent_map(em);
2395 return ERR_PTR(err);
2400 #if 0 /* waiting for O_DIRECT reads */
2401 static int btrfs_get_block(struct inode *inode, sector_t iblock,
2402 struct buffer_head *bh_result, int create)
2404 struct extent_map *em;
2405 u64 start = (u64)iblock << inode->i_blkbits;
2406 struct btrfs_multi_bio *multi = NULL;
2407 struct btrfs_root *root = BTRFS_I(inode)->root;
2413 em = btrfs_get_extent(inode, NULL, 0, start, bh_result->b_size, 0);
2415 if (!em || IS_ERR(em))
2418 if (em->start > start || em->start + em->len <= start) {
2422 if (em->block_start == EXTENT_MAP_INLINE) {
2427 len = em->start + em->len - start;
2428 len = min_t(u64, len, INT_LIMIT(typeof(bh_result->b_size)));
2430 if (em->block_start == EXTENT_MAP_HOLE ||
2431 em->block_start == EXTENT_MAP_DELALLOC) {
2432 bh_result->b_size = len;
2436 logical = start - em->start;
2437 logical = em->block_start + logical;
2440 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
2441 logical, &map_length, &multi, 0);
2443 bh_result->b_blocknr = multi->stripes[0].physical >> inode->i_blkbits;
2444 bh_result->b_size = min(map_length, len);
2446 bh_result->b_bdev = multi->stripes[0].dev->bdev;
2447 set_buffer_mapped(bh_result);
2450 free_extent_map(em);
2455 static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
2456 const struct iovec *iov, loff_t offset,
2457 unsigned long nr_segs)
2461 struct file *file = iocb->ki_filp;
2462 struct inode *inode = file->f_mapping->host;
2467 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2468 offset, nr_segs, btrfs_get_block, NULL);
2472 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
2474 return extent_bmap(mapping, iblock, btrfs_get_extent);
2477 int btrfs_readpage(struct file *file, struct page *page)
2479 struct extent_io_tree *tree;
2480 tree = &BTRFS_I(page->mapping->host)->io_tree;
2481 return extent_read_full_page(tree, page, btrfs_get_extent);
2484 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
2486 struct extent_io_tree *tree;
2489 if (current->flags & PF_MEMALLOC) {
2490 redirty_page_for_writepage(wbc, page);
2494 tree = &BTRFS_I(page->mapping->host)->io_tree;
2495 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
2498 static int btrfs_writepages(struct address_space *mapping,
2499 struct writeback_control *wbc)
2501 struct extent_io_tree *tree;
2502 tree = &BTRFS_I(mapping->host)->io_tree;
2503 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
2507 btrfs_readpages(struct file *file, struct address_space *mapping,
2508 struct list_head *pages, unsigned nr_pages)
2510 struct extent_io_tree *tree;
2511 tree = &BTRFS_I(mapping->host)->io_tree;
2512 return extent_readpages(tree, mapping, pages, nr_pages,
2515 static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
2517 struct extent_io_tree *tree;
2518 struct extent_map_tree *map;
2521 tree = &BTRFS_I(page->mapping->host)->io_tree;
2522 map = &BTRFS_I(page->mapping->host)->extent_tree;
2523 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
2525 invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE);
2526 ClearPagePrivate(page);
2527 set_page_private(page, 0);
2528 page_cache_release(page);
2533 static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
2535 struct btrfs_ordered_extent *ordered;
2537 ordered = btrfs_lookup_ordered_extent(page->mapping->host,
2540 btrfs_put_ordered_extent(ordered);
2543 return __btrfs_releasepage(page, gfp_flags);
2546 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
2548 struct extent_io_tree *tree;
2549 struct btrfs_ordered_extent *ordered;
2550 u64 page_start = page_offset(page);
2551 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
2553 wait_on_page_writeback(page);
2554 tree = &BTRFS_I(page->mapping->host)->io_tree;
2556 btrfs_releasepage(page, GFP_NOFS);
2560 lock_extent(tree, page_start, page_end, GFP_NOFS);
2561 ordered = btrfs_lookup_ordered_extent(page->mapping->host,
2564 clear_extent_bit(tree, page_start, page_end,
2565 EXTENT_DIRTY | EXTENT_DELALLOC |
2566 EXTENT_LOCKED, 1, 0, GFP_NOFS);
2567 btrfs_writepage_end_io_hook(page, page_start,
2569 btrfs_put_ordered_extent(ordered);
2570 lock_extent(tree, page_start, page_end, GFP_NOFS);
2572 clear_extent_bit(tree, page_start, page_end,
2573 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
2576 __btrfs_releasepage(page, GFP_NOFS);
2578 if (PagePrivate(page)) {
2579 invalidate_extent_lru(tree, page_offset(page),
2581 ClearPagePrivate(page);
2582 set_page_private(page, 0);
2583 page_cache_release(page);
2588 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2589 * called from a page fault handler when a page is first dirtied. Hence we must
2590 * be careful to check for EOF conditions here. We set the page up correctly
2591 * for a written page which means we get ENOSPC checking when writing into
2592 * holes and correct delalloc and unwritten extent mapping on filesystems that
2593 * support these features.
2595 * We are not allowed to take the i_mutex here so we have to play games to
2596 * protect against truncate races as the page could now be beyond EOF. Because
2597 * vmtruncate() writes the inode size before removing pages, once we have the
2598 * page lock we can determine safely if the page is beyond EOF. If it is not
2599 * beyond EOF, then the page is guaranteed safe against truncation until we
2602 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
2604 struct inode *inode = fdentry(vma->vm_file)->d_inode;
2605 struct btrfs_root *root = BTRFS_I(inode)->root;
2606 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2607 struct btrfs_ordered_extent *ordered;
2609 unsigned long zero_start;
2615 ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
2622 size = i_size_read(inode);
2623 page_start = page_offset(page);
2624 page_end = page_start + PAGE_CACHE_SIZE - 1;
2626 if ((page->mapping != inode->i_mapping) ||
2627 (page_start >= size)) {
2628 /* page got truncated out from underneath us */
2631 wait_on_page_writeback(page);
2633 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
2634 set_page_extent_mapped(page);
2636 ordered = btrfs_lookup_ordered_extent(inode, page_start);
2638 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
2640 btrfs_wait_ordered_extent(inode, ordered);
2641 btrfs_put_ordered_extent(ordered);
2645 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
2646 page_end, GFP_NOFS);
2649 /* page is wholly or partially inside EOF */
2650 if (page_start + PAGE_CACHE_SIZE > size)
2651 zero_start = size & ~PAGE_CACHE_MASK;
2653 zero_start = PAGE_CACHE_SIZE;
2655 if (zero_start != PAGE_CACHE_SIZE) {
2657 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
2658 flush_dcache_page(page);
2661 set_page_dirty(page);
2662 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
2670 static void btrfs_truncate(struct inode *inode)
2672 struct btrfs_root *root = BTRFS_I(inode)->root;
2674 struct btrfs_trans_handle *trans;
2677 if (!S_ISREG(inode->i_mode))
2679 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
2682 btrfs_truncate_page(inode->i_mapping, inode->i_size);
2684 trans = btrfs_start_transaction(root, 1);
2685 btrfs_set_trans_block_group(trans, inode);
2687 /* FIXME, add redo link to tree so we don't leak on crash */
2688 ret = btrfs_truncate_in_trans(trans, root, inode,
2689 BTRFS_EXTENT_DATA_KEY);
2690 btrfs_update_inode(trans, root, inode);
2691 nr = trans->blocks_used;
2693 ret = btrfs_end_transaction_throttle(trans, root);
2695 btrfs_btree_balance_dirty(root, nr);
2699 * Invalidate a single dcache entry at the root of the filesystem.
2700 * Needed after creation of snapshot or subvolume.
2702 void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name,
2705 struct dentry *alias, *entry;
2708 alias = d_find_alias(root->fs_info->sb->s_root->d_inode);
2712 /* change me if btrfs ever gets a d_hash operation */
2713 qstr.hash = full_name_hash(qstr.name, qstr.len);
2714 entry = d_lookup(alias, &qstr);
2717 d_invalidate(entry);
2723 int btrfs_create_subvol_root(struct btrfs_root *new_root,
2724 struct btrfs_trans_handle *trans, u64 new_dirid,
2725 struct btrfs_block_group_cache *block_group)
2727 struct inode *inode;
2730 inode = btrfs_new_inode(trans, new_root, "..", 2, new_dirid,
2731 new_dirid, block_group, S_IFDIR | 0700);
2733 return PTR_ERR(inode);
2734 inode->i_op = &btrfs_dir_inode_operations;
2735 inode->i_fop = &btrfs_dir_file_operations;
2736 new_root->inode = inode;
2738 ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid,
2743 return btrfs_update_inode(trans, new_root, inode);
2746 unsigned long btrfs_force_ra(struct address_space *mapping,
2747 struct file_ra_state *ra, struct file *file,
2748 pgoff_t offset, pgoff_t last_index)
2750 pgoff_t req_size = last_index - offset + 1;
2752 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2753 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2756 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2757 return offset + req_size;
2761 struct inode *btrfs_alloc_inode(struct super_block *sb)
2763 struct btrfs_inode *ei;
2765 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2769 btrfs_ordered_inode_tree_init(&ei->ordered_tree);
2770 return &ei->vfs_inode;
2773 void btrfs_destroy_inode(struct inode *inode)
2775 struct btrfs_ordered_extent *ordered;
2776 WARN_ON(!list_empty(&inode->i_dentry));
2777 WARN_ON(inode->i_data.nrpages);
2780 ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
2784 printk("found ordered extent %Lu %Lu\n",
2785 ordered->file_offset, ordered->len);
2786 btrfs_remove_ordered_extent(inode, ordered);
2787 btrfs_put_ordered_extent(ordered);
2788 btrfs_put_ordered_extent(ordered);
2791 btrfs_drop_extent_cache(inode, 0, (u64)-1);
2792 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2795 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2796 static void init_once(struct kmem_cache * cachep, void *foo)
2798 static void init_once(void * foo, struct kmem_cache * cachep,
2799 unsigned long flags)
2802 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2804 inode_init_once(&ei->vfs_inode);
2807 void btrfs_destroy_cachep(void)
2809 if (btrfs_inode_cachep)
2810 kmem_cache_destroy(btrfs_inode_cachep);
2811 if (btrfs_trans_handle_cachep)
2812 kmem_cache_destroy(btrfs_trans_handle_cachep);
2813 if (btrfs_transaction_cachep)
2814 kmem_cache_destroy(btrfs_transaction_cachep);
2815 if (btrfs_bit_radix_cachep)
2816 kmem_cache_destroy(btrfs_bit_radix_cachep);
2817 if (btrfs_path_cachep)
2818 kmem_cache_destroy(btrfs_path_cachep);
2821 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2822 unsigned long extra_flags,
2823 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2824 void (*ctor)(struct kmem_cache *, void *)
2826 void (*ctor)(void *, struct kmem_cache *,
2831 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2832 SLAB_MEM_SPREAD | extra_flags), ctor
2833 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2839 int btrfs_init_cachep(void)
2841 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2842 sizeof(struct btrfs_inode),
2844 if (!btrfs_inode_cachep)
2846 btrfs_trans_handle_cachep =
2847 btrfs_cache_create("btrfs_trans_handle_cache",
2848 sizeof(struct btrfs_trans_handle),
2850 if (!btrfs_trans_handle_cachep)
2852 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
2853 sizeof(struct btrfs_transaction),
2855 if (!btrfs_transaction_cachep)
2857 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
2858 sizeof(struct btrfs_path),
2860 if (!btrfs_path_cachep)
2862 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
2863 SLAB_DESTROY_BY_RCU, NULL);
2864 if (!btrfs_bit_radix_cachep)
2868 btrfs_destroy_cachep();
2872 static int btrfs_getattr(struct vfsmount *mnt,
2873 struct dentry *dentry, struct kstat *stat)
2875 struct inode *inode = dentry->d_inode;
2876 generic_fillattr(inode, stat);
2877 stat->blksize = PAGE_CACHE_SIZE;
2878 stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9);
2882 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2883 struct inode * new_dir,struct dentry *new_dentry)
2885 struct btrfs_trans_handle *trans;
2886 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2887 struct inode *new_inode = new_dentry->d_inode;
2888 struct inode *old_inode = old_dentry->d_inode;
2889 struct timespec ctime = CURRENT_TIME;
2892 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2893 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2897 ret = btrfs_check_free_space(root, 1, 0);
2901 trans = btrfs_start_transaction(root, 1);
2903 btrfs_set_trans_block_group(trans, new_dir);
2905 old_dentry->d_inode->i_nlink++;
2906 old_dir->i_ctime = old_dir->i_mtime = ctime;
2907 new_dir->i_ctime = new_dir->i_mtime = ctime;
2908 old_inode->i_ctime = ctime;
2910 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2915 new_inode->i_ctime = CURRENT_TIME;
2916 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2920 ret = btrfs_add_link(trans, new_dentry, old_inode, 1);
2925 btrfs_end_transaction(trans, root);
2930 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2931 const char *symname)
2933 struct btrfs_trans_handle *trans;
2934 struct btrfs_root *root = BTRFS_I(dir)->root;
2935 struct btrfs_path *path;
2936 struct btrfs_key key;
2937 struct inode *inode = NULL;
2944 struct btrfs_file_extent_item *ei;
2945 struct extent_buffer *leaf;
2946 unsigned long nr = 0;
2948 name_len = strlen(symname) + 1;
2949 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2950 return -ENAMETOOLONG;
2952 err = btrfs_check_free_space(root, 1, 0);
2956 trans = btrfs_start_transaction(root, 1);
2957 btrfs_set_trans_block_group(trans, dir);
2959 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2965 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2967 dentry->d_parent->d_inode->i_ino, objectid,
2968 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2969 err = PTR_ERR(inode);
2973 btrfs_set_trans_block_group(trans, inode);
2974 err = btrfs_add_nondir(trans, dentry, inode, 0);
2978 inode->i_mapping->a_ops = &btrfs_aops;
2979 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
2980 inode->i_fop = &btrfs_file_operations;
2981 inode->i_op = &btrfs_file_inode_operations;
2982 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2983 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
2984 inode->i_mapping, GFP_NOFS);
2985 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
2986 inode->i_mapping, GFP_NOFS);
2987 mutex_init(&BTRFS_I(inode)->csum_mutex);
2988 BTRFS_I(inode)->delalloc_bytes = 0;
2989 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
2991 dir->i_sb->s_dirt = 1;
2992 btrfs_update_inode_block_group(trans, inode);
2993 btrfs_update_inode_block_group(trans, dir);
2997 path = btrfs_alloc_path();
2999 key.objectid = inode->i_ino;
3001 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
3002 datasize = btrfs_file_extent_calc_inline_size(name_len);
3003 err = btrfs_insert_empty_item(trans, root, path, &key,
3009 leaf = path->nodes[0];
3010 ei = btrfs_item_ptr(leaf, path->slots[0],
3011 struct btrfs_file_extent_item);
3012 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
3013 btrfs_set_file_extent_type(leaf, ei,
3014 BTRFS_FILE_EXTENT_INLINE);
3015 ptr = btrfs_file_extent_inline_start(ei);
3016 write_extent_buffer(leaf, symname, ptr, name_len);
3017 btrfs_mark_buffer_dirty(leaf);
3018 btrfs_free_path(path);
3020 inode->i_op = &btrfs_symlink_inode_operations;
3021 inode->i_mapping->a_ops = &btrfs_symlink_aops;
3022 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3023 inode->i_size = name_len - 1;
3024 err = btrfs_update_inode(trans, root, inode);
3029 nr = trans->blocks_used;
3030 btrfs_end_transaction_throttle(trans, root);
3033 inode_dec_link_count(inode);
3036 btrfs_btree_balance_dirty(root, nr);
3040 static int btrfs_set_page_dirty(struct page *page)
3042 struct inode *inode = page->mapping->host;
3043 u64 page_start = page_offset(page);
3044 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
3046 if (!test_range_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
3047 EXTENT_DELALLOC, 0)) {
3048 printk("inode %lu page %Lu not delalloc\n", inode->i_ino, page_offset(page));
3051 return __set_page_dirty_nobuffers(page);
3054 static int btrfs_permission(struct inode *inode, int mask,
3055 struct nameidata *nd)
3057 if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
3059 return generic_permission(inode, mask, NULL);
3062 static struct inode_operations btrfs_dir_inode_operations = {
3063 .lookup = btrfs_lookup,
3064 .create = btrfs_create,
3065 .unlink = btrfs_unlink,
3067 .mkdir = btrfs_mkdir,
3068 .rmdir = btrfs_rmdir,
3069 .rename = btrfs_rename,
3070 .symlink = btrfs_symlink,
3071 .setattr = btrfs_setattr,
3072 .mknod = btrfs_mknod,
3073 .setxattr = generic_setxattr,
3074 .getxattr = generic_getxattr,
3075 .listxattr = btrfs_listxattr,
3076 .removexattr = generic_removexattr,
3077 .permission = btrfs_permission,
3079 static struct inode_operations btrfs_dir_ro_inode_operations = {
3080 .lookup = btrfs_lookup,
3081 .permission = btrfs_permission,
3083 static struct file_operations btrfs_dir_file_operations = {
3084 .llseek = generic_file_llseek,
3085 .read = generic_read_dir,
3086 .readdir = btrfs_readdir,
3087 .unlocked_ioctl = btrfs_ioctl,
3088 #ifdef CONFIG_COMPAT
3089 .compat_ioctl = btrfs_ioctl,
3091 .release = btrfs_release_file,
3094 static struct extent_io_ops btrfs_extent_io_ops = {
3095 .fill_delalloc = run_delalloc_range,
3096 .submit_bio_hook = btrfs_submit_bio_hook,
3097 .merge_bio_hook = btrfs_merge_bio_hook,
3098 .readpage_io_hook = btrfs_readpage_io_hook,
3099 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
3100 .writepage_end_io_hook = btrfs_writepage_end_io_hook,
3101 .readpage_io_failed_hook = btrfs_io_failed_hook,
3102 .set_bit_hook = btrfs_set_bit_hook,
3103 .clear_bit_hook = btrfs_clear_bit_hook,
3106 static struct address_space_operations btrfs_aops = {
3107 .readpage = btrfs_readpage,
3108 .writepage = btrfs_writepage,
3109 .writepages = btrfs_writepages,
3110 .readpages = btrfs_readpages,
3111 .sync_page = block_sync_page,
3113 .direct_IO = btrfs_direct_IO,
3114 .invalidatepage = btrfs_invalidatepage,
3115 .releasepage = btrfs_releasepage,
3116 .set_page_dirty = btrfs_set_page_dirty,
3119 static struct address_space_operations btrfs_symlink_aops = {
3120 .readpage = btrfs_readpage,
3121 .writepage = btrfs_writepage,
3122 .invalidatepage = btrfs_invalidatepage,
3123 .releasepage = btrfs_releasepage,
3126 static struct inode_operations btrfs_file_inode_operations = {
3127 .truncate = btrfs_truncate,
3128 .getattr = btrfs_getattr,
3129 .setattr = btrfs_setattr,
3130 .setxattr = generic_setxattr,
3131 .getxattr = generic_getxattr,
3132 .listxattr = btrfs_listxattr,
3133 .removexattr = generic_removexattr,
3134 .permission = btrfs_permission,
3136 static struct inode_operations btrfs_special_inode_operations = {
3137 .getattr = btrfs_getattr,
3138 .setattr = btrfs_setattr,
3139 .permission = btrfs_permission,
3141 static struct inode_operations btrfs_symlink_inode_operations = {
3142 .readlink = generic_readlink,
3143 .follow_link = page_follow_link_light,
3144 .put_link = page_put_link,
3145 .permission = btrfs_permission,