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/bio.h>
20 #include <linux/buffer_head.h>
22 #include <linux/pagemap.h>
23 #include <linux/highmem.h>
24 #include <linux/time.h>
25 #include <linux/init.h>
26 #include <linux/string.h>
27 #include <linux/smp_lock.h>
28 #include <linux/backing-dev.h>
29 #include <linux/mpage.h>
30 #include <linux/swap.h>
31 #include <linux/writeback.h>
32 #include <linux/statfs.h>
33 #include <linux/compat.h>
34 #include <linux/bit_spinlock.h>
35 #include <linux/version.h>
36 #include <linux/xattr.h>
39 #include "transaction.h"
40 #include "btrfs_inode.h"
42 #include "print-tree.h"
45 struct btrfs_iget_args {
47 struct btrfs_root *root;
50 static struct inode_operations btrfs_dir_inode_operations;
51 static struct inode_operations btrfs_symlink_inode_operations;
52 static struct inode_operations btrfs_dir_ro_inode_operations;
53 static struct inode_operations btrfs_special_inode_operations;
54 static struct inode_operations btrfs_file_inode_operations;
55 static struct address_space_operations btrfs_aops;
56 static struct address_space_operations btrfs_symlink_aops;
57 static struct file_operations btrfs_dir_file_operations;
58 static struct extent_io_ops btrfs_extent_io_ops;
60 static struct kmem_cache *btrfs_inode_cachep;
61 struct kmem_cache *btrfs_trans_handle_cachep;
62 struct kmem_cache *btrfs_transaction_cachep;
63 struct kmem_cache *btrfs_bit_radix_cachep;
64 struct kmem_cache *btrfs_path_cachep;
67 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
68 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
69 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
70 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
71 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
72 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
73 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
74 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
77 int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
80 u64 total = btrfs_super_total_bytes(&root->fs_info->super_copy);
81 u64 used = btrfs_super_bytes_used(&root->fs_info->super_copy);
92 spin_lock(&root->fs_info->delalloc_lock);
93 if (used + root->fs_info->delalloc_bytes + num_required > thresh)
95 spin_unlock(&root->fs_info->delalloc_lock);
99 static int cow_file_range(struct inode *inode, u64 start, u64 end)
101 struct btrfs_root *root = BTRFS_I(inode)->root;
102 struct btrfs_trans_handle *trans;
106 u64 blocksize = root->sectorsize;
107 u64 orig_start = start;
109 struct btrfs_key ins;
112 trans = btrfs_start_transaction(root, 1);
114 btrfs_set_trans_block_group(trans, inode);
116 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
117 num_bytes = max(blocksize, num_bytes);
118 ret = btrfs_drop_extents(trans, root, inode,
119 start, start + num_bytes, start, &alloc_hint);
120 orig_num_bytes = num_bytes;
122 if (alloc_hint == EXTENT_MAP_INLINE)
125 while(num_bytes > 0) {
126 cur_alloc_size = min(num_bytes, root->fs_info->max_extent);
127 ret = btrfs_alloc_extent(trans, root, cur_alloc_size,
129 root->root_key.objectid,
131 inode->i_ino, start, 0,
132 alloc_hint, (u64)-1, &ins, 1);
137 cur_alloc_size = ins.offset;
138 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
139 start, ins.objectid, ins.offset,
141 inode->i_blocks += ins.offset >> 9;
142 btrfs_check_file(root, inode);
143 num_bytes -= cur_alloc_size;
144 alloc_hint = ins.objectid + ins.offset;
145 start += cur_alloc_size;
147 btrfs_drop_extent_cache(inode, orig_start,
148 orig_start + orig_num_bytes - 1);
149 btrfs_add_ordered_inode(inode);
150 btrfs_update_inode(trans, root, inode);
152 btrfs_end_transaction(trans, root);
156 static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end)
164 struct btrfs_root *root = BTRFS_I(inode)->root;
165 struct extent_buffer *leaf;
167 struct btrfs_path *path;
168 struct btrfs_file_extent_item *item;
171 struct btrfs_key found_key;
173 total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
174 path = btrfs_alloc_path();
177 ret = btrfs_lookup_file_extent(NULL, root, path,
178 inode->i_ino, start, 0);
180 btrfs_free_path(path);
186 if (path->slots[0] == 0)
191 leaf = path->nodes[0];
192 item = btrfs_item_ptr(leaf, path->slots[0],
193 struct btrfs_file_extent_item);
195 /* are we inside the extent that was found? */
196 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
197 found_type = btrfs_key_type(&found_key);
198 if (found_key.objectid != inode->i_ino ||
199 found_type != BTRFS_EXTENT_DATA_KEY) {
203 found_type = btrfs_file_extent_type(leaf, item);
204 extent_start = found_key.offset;
205 if (found_type == BTRFS_FILE_EXTENT_REG) {
206 u64 extent_num_bytes;
208 extent_num_bytes = btrfs_file_extent_num_bytes(leaf, item);
209 extent_end = extent_start + extent_num_bytes;
212 if (loops && start != extent_start)
215 if (start < extent_start || start >= extent_end)
218 cow_end = min(end, extent_end - 1);
219 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
224 * we may be called by the resizer, make sure we're inside
225 * the limits of the FS
227 if (bytenr + extent_num_bytes > total_fs_bytes)
230 if (btrfs_count_snapshots_in_path(root, path, bytenr) != 1) {
240 btrfs_free_path(path);
243 btrfs_release_path(root, path);
248 cow_file_range(inode, start, cow_end);
253 static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
255 struct btrfs_root *root = BTRFS_I(inode)->root;
257 mutex_lock(&root->fs_info->fs_mutex);
258 if (btrfs_test_opt(root, NODATACOW) ||
259 btrfs_test_flag(inode, NODATACOW))
260 ret = run_delalloc_nocow(inode, start, end);
262 ret = cow_file_range(inode, start, end);
264 mutex_unlock(&root->fs_info->fs_mutex);
268 int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
269 unsigned long old, unsigned long bits)
271 if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
272 struct btrfs_root *root = BTRFS_I(inode)->root;
273 spin_lock(&root->fs_info->delalloc_lock);
274 BTRFS_I(inode)->delalloc_bytes += end - start + 1;
275 root->fs_info->delalloc_bytes += end - start + 1;
276 spin_unlock(&root->fs_info->delalloc_lock);
281 int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end,
282 unsigned long old, unsigned long bits)
284 if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
285 struct btrfs_root *root = BTRFS_I(inode)->root;
286 spin_lock(&root->fs_info->delalloc_lock);
287 if (end - start + 1 > root->fs_info->delalloc_bytes) {
288 printk("warning: delalloc account %Lu %Lu\n",
289 end - start + 1, root->fs_info->delalloc_bytes);
290 root->fs_info->delalloc_bytes = 0;
291 BTRFS_I(inode)->delalloc_bytes = 0;
293 root->fs_info->delalloc_bytes -= end - start + 1;
294 BTRFS_I(inode)->delalloc_bytes -= end - start + 1;
296 spin_unlock(&root->fs_info->delalloc_lock);
301 int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
302 size_t size, struct bio *bio)
304 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
305 struct btrfs_mapping_tree *map_tree;
306 u64 logical = bio->bi_sector << 9;
309 struct bio_vec *bvec;
313 bio_for_each_segment(bvec, bio, i) {
314 length += bvec->bv_len;
316 map_tree = &root->fs_info->mapping_tree;
318 ret = btrfs_map_block(map_tree, READ, logical,
319 &map_length, NULL, 0);
321 if (map_length < length + size) {
327 int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
330 struct btrfs_root *root = BTRFS_I(inode)->root;
331 struct btrfs_trans_handle *trans;
335 ret = btrfs_csum_one_bio(root, bio, &sums);
338 mutex_lock(&root->fs_info->fs_mutex);
339 trans = btrfs_start_transaction(root, 1);
341 btrfs_set_trans_block_group(trans, inode);
342 btrfs_csum_file_blocks(trans, root, inode, bio, sums);
344 ret = btrfs_end_transaction(trans, root);
346 mutex_unlock(&root->fs_info->fs_mutex);
350 return btrfs_map_bio(root, rw, bio, mirror_num);
353 int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
356 struct btrfs_root *root = BTRFS_I(inode)->root;
359 if (!(rw & (1 << BIO_RW))) {
360 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
365 if (btrfs_test_opt(root, NODATASUM) ||
366 btrfs_test_flag(inode, NODATASUM)) {
370 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
371 inode, rw, bio, mirror_num,
372 __btrfs_submit_bio_hook);
374 return btrfs_map_bio(root, rw, bio, mirror_num);
377 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
380 struct inode *inode = page->mapping->host;
381 struct btrfs_root *root = BTRFS_I(inode)->root;
382 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
383 struct btrfs_csum_item *item;
384 struct btrfs_path *path = NULL;
387 if (btrfs_test_opt(root, NODATASUM) ||
388 btrfs_test_flag(inode, NODATASUM))
391 mutex_lock(&root->fs_info->fs_mutex);
392 path = btrfs_alloc_path();
393 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
396 /* a csum that isn't present is a preallocated region. */
397 if (ret == -ENOENT || ret == -EFBIG)
400 printk("no csum found for inode %lu start %Lu\n", inode->i_ino, start);
403 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
405 set_state_private(io_tree, start, csum);
408 btrfs_free_path(path);
409 mutex_unlock(&root->fs_info->fs_mutex);
413 struct io_failure_record {
421 int btrfs_readpage_io_failed_hook(struct bio *failed_bio,
422 struct page *page, u64 start, u64 end,
423 struct extent_state *state)
425 struct io_failure_record *failrec = NULL;
427 struct extent_map *em;
428 struct inode *inode = page->mapping->host;
429 struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
435 ret = get_state_private(failure_tree, start, &private);
437 size_t pg_offset = start - page_offset(page);
438 failrec = kmalloc(sizeof(*failrec), GFP_NOFS);
441 failrec->start = start;
442 failrec->len = end - start + 1;
443 failrec->last_mirror = 0;
445 em = btrfs_get_extent(inode, NULL, pg_offset, start,
448 if (!em || IS_ERR(em)) {
452 logical = start - em->start;
453 logical = em->block_start + logical;
454 failrec->logical = logical;
456 set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
457 EXTENT_DIRTY, GFP_NOFS);
458 set_state_private(failure_tree, start,
459 (u64)(unsigned long)failrec);
461 failrec = (struct io_failure_record *)(unsigned long)private;
463 num_copies = btrfs_num_copies(
464 &BTRFS_I(inode)->root->fs_info->mapping_tree,
465 failrec->logical, failrec->len);
466 failrec->last_mirror++;
468 spin_lock_irq(&BTRFS_I(inode)->io_tree.lock);
469 state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
472 if (state && state->start != failrec->start)
474 spin_unlock_irq(&BTRFS_I(inode)->io_tree.lock);
476 if (!state || failrec->last_mirror > num_copies) {
477 set_state_private(failure_tree, failrec->start, 0);
478 clear_extent_bits(failure_tree, failrec->start,
479 failrec->start + failrec->len - 1,
480 EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
484 bio = bio_alloc(GFP_NOFS, 1);
485 bio->bi_private = state;
486 bio->bi_end_io = failed_bio->bi_end_io;
487 bio->bi_sector = failrec->logical >> 9;
488 bio->bi_bdev = failed_bio->bi_bdev;
489 bio_add_page(bio, page, failrec->len, start - page_offset(page));
490 btrfs_submit_bio_hook(inode, READ, bio, failrec->last_mirror);
494 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
495 struct extent_state *state)
497 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
498 struct inode *inode = page->mapping->host;
499 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
501 u64 private = ~(u32)0;
503 struct btrfs_root *root = BTRFS_I(inode)->root;
507 if (btrfs_test_opt(root, NODATASUM) ||
508 btrfs_test_flag(inode, NODATASUM))
510 if (state && state->start == start) {
511 private = state->private;
514 ret = get_state_private(io_tree, start, &private);
516 local_irq_save(flags);
517 kaddr = kmap_atomic(page, KM_IRQ0);
521 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
522 btrfs_csum_final(csum, (char *)&csum);
523 if (csum != private) {
526 kunmap_atomic(kaddr, KM_IRQ0);
527 local_irq_restore(flags);
529 /* if the io failure tree for this inode is non-empty,
530 * check to see if we've recovered from a failed IO
533 if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
534 (u64)-1, 1, EXTENT_DIRTY)) {
536 struct io_failure_record *failure;
537 ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
538 start, &private_failure);
540 failure = (struct io_failure_record *)(unsigned long)
542 set_state_private(&BTRFS_I(inode)->io_failure_tree,
544 clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
546 failure->start + failure->len - 1,
547 EXTENT_DIRTY | EXTENT_LOCKED,
555 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
556 page->mapping->host->i_ino, (unsigned long long)start, csum,
558 memset(kaddr + offset, 1, end - start + 1);
559 flush_dcache_page(page);
560 kunmap_atomic(kaddr, KM_IRQ0);
561 local_irq_restore(flags);
565 void btrfs_read_locked_inode(struct inode *inode)
567 struct btrfs_path *path;
568 struct extent_buffer *leaf;
569 struct btrfs_inode_item *inode_item;
570 struct btrfs_timespec *tspec;
571 struct btrfs_root *root = BTRFS_I(inode)->root;
572 struct btrfs_key location;
573 u64 alloc_group_block;
577 path = btrfs_alloc_path();
579 mutex_lock(&root->fs_info->fs_mutex);
580 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
582 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
586 leaf = path->nodes[0];
587 inode_item = btrfs_item_ptr(leaf, path->slots[0],
588 struct btrfs_inode_item);
590 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
591 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
592 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
593 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
594 inode->i_size = btrfs_inode_size(leaf, inode_item);
596 tspec = btrfs_inode_atime(inode_item);
597 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
598 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
600 tspec = btrfs_inode_mtime(inode_item);
601 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
602 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
604 tspec = btrfs_inode_ctime(inode_item);
605 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
606 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
608 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
609 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
611 rdev = btrfs_inode_rdev(leaf, inode_item);
613 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
614 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
616 BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
617 if (!BTRFS_I(inode)->block_group) {
618 BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
620 BTRFS_BLOCK_GROUP_METADATA, 0);
622 btrfs_free_path(path);
625 mutex_unlock(&root->fs_info->fs_mutex);
627 switch (inode->i_mode & S_IFMT) {
629 inode->i_mapping->a_ops = &btrfs_aops;
630 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
631 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
632 inode->i_fop = &btrfs_file_operations;
633 inode->i_op = &btrfs_file_inode_operations;
636 inode->i_fop = &btrfs_dir_file_operations;
637 if (root == root->fs_info->tree_root)
638 inode->i_op = &btrfs_dir_ro_inode_operations;
640 inode->i_op = &btrfs_dir_inode_operations;
643 inode->i_op = &btrfs_symlink_inode_operations;
644 inode->i_mapping->a_ops = &btrfs_symlink_aops;
645 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
648 init_special_inode(inode, inode->i_mode, rdev);
654 btrfs_release_path(root, path);
655 btrfs_free_path(path);
656 mutex_unlock(&root->fs_info->fs_mutex);
657 make_bad_inode(inode);
660 static void fill_inode_item(struct extent_buffer *leaf,
661 struct btrfs_inode_item *item,
664 btrfs_set_inode_uid(leaf, item, inode->i_uid);
665 btrfs_set_inode_gid(leaf, item, inode->i_gid);
666 btrfs_set_inode_size(leaf, item, inode->i_size);
667 btrfs_set_inode_mode(leaf, item, inode->i_mode);
668 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
670 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
671 inode->i_atime.tv_sec);
672 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
673 inode->i_atime.tv_nsec);
675 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
676 inode->i_mtime.tv_sec);
677 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
678 inode->i_mtime.tv_nsec);
680 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
681 inode->i_ctime.tv_sec);
682 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
683 inode->i_ctime.tv_nsec);
685 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
686 btrfs_set_inode_generation(leaf, item, inode->i_generation);
687 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
688 btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
689 btrfs_set_inode_block_group(leaf, item,
690 BTRFS_I(inode)->block_group->key.objectid);
693 int btrfs_update_inode(struct btrfs_trans_handle *trans,
694 struct btrfs_root *root,
697 struct btrfs_inode_item *inode_item;
698 struct btrfs_path *path;
699 struct extent_buffer *leaf;
702 path = btrfs_alloc_path();
704 ret = btrfs_lookup_inode(trans, root, path,
705 &BTRFS_I(inode)->location, 1);
712 leaf = path->nodes[0];
713 inode_item = btrfs_item_ptr(leaf, path->slots[0],
714 struct btrfs_inode_item);
716 fill_inode_item(leaf, inode_item, inode);
717 btrfs_mark_buffer_dirty(leaf);
718 btrfs_set_inode_last_trans(trans, inode);
721 btrfs_release_path(root, path);
722 btrfs_free_path(path);
727 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
728 struct btrfs_root *root,
730 struct dentry *dentry)
732 struct btrfs_path *path;
733 const char *name = dentry->d_name.name;
734 int name_len = dentry->d_name.len;
736 struct extent_buffer *leaf;
737 struct btrfs_dir_item *di;
738 struct btrfs_key key;
740 path = btrfs_alloc_path();
746 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
756 leaf = path->nodes[0];
757 btrfs_dir_item_key_to_cpu(leaf, di, &key);
758 ret = btrfs_delete_one_dir_name(trans, root, path, di);
761 btrfs_release_path(root, path);
763 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
764 key.objectid, name, name_len, -1);
773 ret = btrfs_delete_one_dir_name(trans, root, path, di);
775 dentry->d_inode->i_ctime = dir->i_ctime;
776 ret = btrfs_del_inode_ref(trans, root, name, name_len,
777 dentry->d_inode->i_ino,
778 dentry->d_parent->d_inode->i_ino);
780 printk("failed to delete reference to %.*s, "
781 "inode %lu parent %lu\n", name_len, name,
782 dentry->d_inode->i_ino,
783 dentry->d_parent->d_inode->i_ino);
786 btrfs_free_path(path);
788 dir->i_size -= name_len * 2;
789 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
790 btrfs_update_inode(trans, root, dir);
791 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
792 dentry->d_inode->i_nlink--;
794 drop_nlink(dentry->d_inode);
796 ret = btrfs_update_inode(trans, root, dentry->d_inode);
797 dir->i_sb->s_dirt = 1;
802 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
804 struct btrfs_root *root;
805 struct btrfs_trans_handle *trans;
806 struct inode *inode = dentry->d_inode;
808 unsigned long nr = 0;
810 root = BTRFS_I(dir)->root;
811 mutex_lock(&root->fs_info->fs_mutex);
813 ret = btrfs_check_free_space(root, 1, 1);
817 trans = btrfs_start_transaction(root, 1);
819 btrfs_set_trans_block_group(trans, dir);
820 ret = btrfs_unlink_trans(trans, root, dir, dentry);
821 nr = trans->blocks_used;
823 if (inode->i_nlink == 0) {
825 /* if the inode isn't linked anywhere,
826 * we don't need to worry about
829 found = btrfs_del_ordered_inode(inode);
831 atomic_dec(&inode->i_count);
835 btrfs_end_transaction(trans, root);
837 mutex_unlock(&root->fs_info->fs_mutex);
838 btrfs_btree_balance_dirty(root, nr);
839 btrfs_throttle(root);
843 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
845 struct inode *inode = dentry->d_inode;
848 struct btrfs_root *root = BTRFS_I(dir)->root;
849 struct btrfs_trans_handle *trans;
850 unsigned long nr = 0;
852 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
855 mutex_lock(&root->fs_info->fs_mutex);
856 ret = btrfs_check_free_space(root, 1, 1);
860 trans = btrfs_start_transaction(root, 1);
861 btrfs_set_trans_block_group(trans, dir);
863 /* now the directory is empty */
864 err = btrfs_unlink_trans(trans, root, dir, dentry);
869 nr = trans->blocks_used;
870 ret = btrfs_end_transaction(trans, root);
872 mutex_unlock(&root->fs_info->fs_mutex);
873 btrfs_btree_balance_dirty(root, nr);
874 btrfs_throttle(root);
882 * this can truncate away extent items, csum items and directory items.
883 * It starts at a high offset and removes keys until it can't find
884 * any higher than i_size.
886 * csum items that cross the new i_size are truncated to the new size
889 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
890 struct btrfs_root *root,
895 struct btrfs_path *path;
896 struct btrfs_key key;
897 struct btrfs_key found_key;
899 struct extent_buffer *leaf;
900 struct btrfs_file_extent_item *fi;
901 u64 extent_start = 0;
902 u64 extent_num_bytes = 0;
908 int pending_del_nr = 0;
909 int pending_del_slot = 0;
910 int extent_type = -1;
912 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
913 path = btrfs_alloc_path();
917 /* FIXME, add redo link to tree so we don't leak on crash */
918 key.objectid = inode->i_ino;
919 key.offset = (u64)-1;
922 btrfs_init_path(path);
924 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
929 BUG_ON(path->slots[0] == 0);
935 leaf = path->nodes[0];
936 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
937 found_type = btrfs_key_type(&found_key);
939 if (found_key.objectid != inode->i_ino)
942 if (found_type < min_type)
945 item_end = found_key.offset;
946 if (found_type == BTRFS_EXTENT_DATA_KEY) {
947 fi = btrfs_item_ptr(leaf, path->slots[0],
948 struct btrfs_file_extent_item);
949 extent_type = btrfs_file_extent_type(leaf, fi);
950 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
952 btrfs_file_extent_num_bytes(leaf, fi);
953 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
954 struct btrfs_item *item = btrfs_item_nr(leaf,
956 item_end += btrfs_file_extent_inline_len(leaf,
961 if (found_type == BTRFS_CSUM_ITEM_KEY) {
962 ret = btrfs_csum_truncate(trans, root, path,
966 if (item_end < inode->i_size) {
967 if (found_type == BTRFS_DIR_ITEM_KEY) {
968 found_type = BTRFS_INODE_ITEM_KEY;
969 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
970 found_type = BTRFS_CSUM_ITEM_KEY;
971 } else if (found_type == BTRFS_EXTENT_DATA_KEY) {
972 found_type = BTRFS_XATTR_ITEM_KEY;
973 } else if (found_type == BTRFS_XATTR_ITEM_KEY) {
974 found_type = BTRFS_INODE_REF_KEY;
975 } else if (found_type) {
980 btrfs_set_key_type(&key, found_type);
983 if (found_key.offset >= inode->i_size)
989 /* FIXME, shrink the extent if the ref count is only 1 */
990 if (found_type != BTRFS_EXTENT_DATA_KEY)
993 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
995 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
998 btrfs_file_extent_num_bytes(leaf, fi);
999 extent_num_bytes = inode->i_size -
1000 found_key.offset + root->sectorsize - 1;
1001 extent_num_bytes = extent_num_bytes &
1002 ~((u64)root->sectorsize - 1);
1003 btrfs_set_file_extent_num_bytes(leaf, fi,
1005 num_dec = (orig_num_bytes -
1007 if (extent_start != 0)
1008 dec_i_blocks(inode, num_dec);
1009 btrfs_mark_buffer_dirty(leaf);
1012 btrfs_file_extent_disk_num_bytes(leaf,
1014 /* FIXME blocksize != 4096 */
1015 num_dec = btrfs_file_extent_num_bytes(leaf, fi);
1016 if (extent_start != 0) {
1018 dec_i_blocks(inode, num_dec);
1020 root_gen = btrfs_header_generation(leaf);
1021 root_owner = btrfs_header_owner(leaf);
1023 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1025 u32 newsize = inode->i_size - found_key.offset;
1026 dec_i_blocks(inode, item_end + 1 -
1027 found_key.offset - newsize);
1029 btrfs_file_extent_calc_inline_size(newsize);
1030 ret = btrfs_truncate_item(trans, root, path,
1034 dec_i_blocks(inode, item_end + 1 -
1040 if (!pending_del_nr) {
1041 /* no pending yet, add ourselves */
1042 pending_del_slot = path->slots[0];
1044 } else if (pending_del_nr &&
1045 path->slots[0] + 1 == pending_del_slot) {
1046 /* hop on the pending chunk */
1048 pending_del_slot = path->slots[0];
1050 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot);
1056 ret = btrfs_free_extent(trans, root, extent_start,
1059 root_gen, inode->i_ino,
1060 found_key.offset, 0);
1064 if (path->slots[0] == 0) {
1067 btrfs_release_path(root, path);
1072 if (pending_del_nr &&
1073 path->slots[0] + 1 != pending_del_slot) {
1074 struct btrfs_key debug;
1076 btrfs_item_key_to_cpu(path->nodes[0], &debug,
1078 ret = btrfs_del_items(trans, root, path,
1083 btrfs_release_path(root, path);
1089 if (pending_del_nr) {
1090 ret = btrfs_del_items(trans, root, path, pending_del_slot,
1093 btrfs_release_path(root, path);
1094 btrfs_free_path(path);
1095 inode->i_sb->s_dirt = 1;
1099 static int btrfs_cow_one_page(struct inode *inode, struct page *page,
1103 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1104 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
1105 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
1108 WARN_ON(!PageLocked(page));
1109 set_page_extent_mapped(page);
1111 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
1112 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
1113 page_end, GFP_NOFS);
1115 if (zero_start != PAGE_CACHE_SIZE) {
1117 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
1118 flush_dcache_page(page);
1121 set_page_dirty(page);
1122 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
1128 * taken from block_truncate_page, but does cow as it zeros out
1129 * any bytes left in the last page in the file.
1131 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
1133 struct inode *inode = mapping->host;
1134 struct btrfs_root *root = BTRFS_I(inode)->root;
1135 u32 blocksize = root->sectorsize;
1136 pgoff_t index = from >> PAGE_CACHE_SHIFT;
1137 unsigned offset = from & (PAGE_CACHE_SIZE-1);
1142 if ((offset & (blocksize - 1)) == 0)
1146 page = grab_cache_page(mapping, index);
1149 if (!PageUptodate(page)) {
1150 ret = btrfs_readpage(NULL, page);
1152 if (!PageUptodate(page)) {
1157 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
1159 ret = btrfs_cow_one_page(inode, page, offset);
1162 page_cache_release(page);
1167 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
1169 struct inode *inode = dentry->d_inode;
1172 err = inode_change_ok(inode, attr);
1176 if (S_ISREG(inode->i_mode) &&
1177 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
1178 struct btrfs_trans_handle *trans;
1179 struct btrfs_root *root = BTRFS_I(inode)->root;
1180 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1182 u64 mask = root->sectorsize - 1;
1183 u64 hole_start = (inode->i_size + mask) & ~mask;
1184 u64 block_end = (attr->ia_size + mask) & ~mask;
1188 if (attr->ia_size <= hole_start)
1191 mutex_lock(&root->fs_info->fs_mutex);
1192 err = btrfs_check_free_space(root, 1, 0);
1193 mutex_unlock(&root->fs_info->fs_mutex);
1197 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1199 lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1200 hole_size = block_end - hole_start;
1202 mutex_lock(&root->fs_info->fs_mutex);
1203 trans = btrfs_start_transaction(root, 1);
1204 btrfs_set_trans_block_group(trans, inode);
1205 err = btrfs_drop_extents(trans, root, inode,
1206 hole_start, block_end, hole_start,
1209 if (alloc_hint != EXTENT_MAP_INLINE) {
1210 err = btrfs_insert_file_extent(trans, root,
1214 btrfs_drop_extent_cache(inode, hole_start,
1216 btrfs_check_file(root, inode);
1218 btrfs_end_transaction(trans, root);
1219 mutex_unlock(&root->fs_info->fs_mutex);
1220 unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1225 err = inode_setattr(inode, attr);
1230 void btrfs_put_inode(struct inode *inode)
1234 if (!BTRFS_I(inode)->ordered_trans) {
1238 if (mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY) ||
1239 mapping_tagged(inode->i_mapping, PAGECACHE_TAG_WRITEBACK))
1242 ret = btrfs_del_ordered_inode(inode);
1244 atomic_dec(&inode->i_count);
1248 void btrfs_delete_inode(struct inode *inode)
1250 struct btrfs_trans_handle *trans;
1251 struct btrfs_root *root = BTRFS_I(inode)->root;
1255 truncate_inode_pages(&inode->i_data, 0);
1256 if (is_bad_inode(inode)) {
1261 mutex_lock(&root->fs_info->fs_mutex);
1262 trans = btrfs_start_transaction(root, 1);
1264 btrfs_set_trans_block_group(trans, inode);
1265 ret = btrfs_truncate_in_trans(trans, root, inode, 0);
1267 goto no_delete_lock;
1269 nr = trans->blocks_used;
1272 btrfs_end_transaction(trans, root);
1273 mutex_unlock(&root->fs_info->fs_mutex);
1274 btrfs_btree_balance_dirty(root, nr);
1275 btrfs_throttle(root);
1279 nr = trans->blocks_used;
1280 btrfs_end_transaction(trans, root);
1281 mutex_unlock(&root->fs_info->fs_mutex);
1282 btrfs_btree_balance_dirty(root, nr);
1283 btrfs_throttle(root);
1289 * this returns the key found in the dir entry in the location pointer.
1290 * If no dir entries were found, location->objectid is 0.
1292 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
1293 struct btrfs_key *location)
1295 const char *name = dentry->d_name.name;
1296 int namelen = dentry->d_name.len;
1297 struct btrfs_dir_item *di;
1298 struct btrfs_path *path;
1299 struct btrfs_root *root = BTRFS_I(dir)->root;
1302 if (namelen == 1 && strcmp(name, ".") == 0) {
1303 location->objectid = dir->i_ino;
1304 location->type = BTRFS_INODE_ITEM_KEY;
1305 location->offset = 0;
1308 path = btrfs_alloc_path();
1311 if (namelen == 2 && strcmp(name, "..") == 0) {
1312 struct btrfs_key key;
1313 struct extent_buffer *leaf;
1317 key.objectid = dir->i_ino;
1318 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1320 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1324 leaf = path->nodes[0];
1325 slot = path->slots[0];
1326 nritems = btrfs_header_nritems(leaf);
1327 if (slot >= nritems)
1330 btrfs_item_key_to_cpu(leaf, &key, slot);
1331 if (key.objectid != dir->i_ino ||
1332 key.type != BTRFS_INODE_REF_KEY) {
1335 location->objectid = key.offset;
1336 location->type = BTRFS_INODE_ITEM_KEY;
1337 location->offset = 0;
1341 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
1345 if (!di || IS_ERR(di)) {
1348 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
1350 btrfs_free_path(path);
1353 location->objectid = 0;
1358 * when we hit a tree root in a directory, the btrfs part of the inode
1359 * needs to be changed to reflect the root directory of the tree root. This
1360 * is kind of like crossing a mount point.
1362 static int fixup_tree_root_location(struct btrfs_root *root,
1363 struct btrfs_key *location,
1364 struct btrfs_root **sub_root,
1365 struct dentry *dentry)
1367 struct btrfs_path *path;
1368 struct btrfs_root_item *ri;
1370 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
1372 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1375 path = btrfs_alloc_path();
1377 mutex_lock(&root->fs_info->fs_mutex);
1379 *sub_root = btrfs_read_fs_root(root->fs_info, location,
1380 dentry->d_name.name,
1381 dentry->d_name.len);
1382 if (IS_ERR(*sub_root))
1383 return PTR_ERR(*sub_root);
1385 ri = &(*sub_root)->root_item;
1386 location->objectid = btrfs_root_dirid(ri);
1387 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1388 location->offset = 0;
1390 btrfs_free_path(path);
1391 mutex_unlock(&root->fs_info->fs_mutex);
1395 static int btrfs_init_locked_inode(struct inode *inode, void *p)
1397 struct btrfs_iget_args *args = p;
1398 inode->i_ino = args->ino;
1399 BTRFS_I(inode)->root = args->root;
1400 BTRFS_I(inode)->delalloc_bytes = 0;
1401 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1402 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1403 inode->i_mapping, GFP_NOFS);
1404 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1405 inode->i_mapping, GFP_NOFS);
1409 static int btrfs_find_actor(struct inode *inode, void *opaque)
1411 struct btrfs_iget_args *args = opaque;
1412 return (args->ino == inode->i_ino &&
1413 args->root == BTRFS_I(inode)->root);
1416 struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
1419 struct btrfs_iget_args args;
1420 args.ino = objectid;
1421 args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
1426 return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
1429 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
1430 struct btrfs_root *root)
1432 struct inode *inode;
1433 struct btrfs_iget_args args;
1434 args.ino = objectid;
1437 inode = iget5_locked(s, objectid, btrfs_find_actor,
1438 btrfs_init_locked_inode,
1443 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
1444 struct nameidata *nd)
1446 struct inode * inode;
1447 struct btrfs_inode *bi = BTRFS_I(dir);
1448 struct btrfs_root *root = bi->root;
1449 struct btrfs_root *sub_root = root;
1450 struct btrfs_key location;
1453 if (dentry->d_name.len > BTRFS_NAME_LEN)
1454 return ERR_PTR(-ENAMETOOLONG);
1456 mutex_lock(&root->fs_info->fs_mutex);
1457 ret = btrfs_inode_by_name(dir, dentry, &location);
1458 mutex_unlock(&root->fs_info->fs_mutex);
1461 return ERR_PTR(ret);
1464 if (location.objectid) {
1465 ret = fixup_tree_root_location(root, &location, &sub_root,
1468 return ERR_PTR(ret);
1470 return ERR_PTR(-ENOENT);
1471 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
1474 return ERR_PTR(-EACCES);
1475 if (inode->i_state & I_NEW) {
1476 /* the inode and parent dir are two different roots */
1477 if (sub_root != root) {
1479 sub_root->inode = inode;
1481 BTRFS_I(inode)->root = sub_root;
1482 memcpy(&BTRFS_I(inode)->location, &location,
1484 btrfs_read_locked_inode(inode);
1485 unlock_new_inode(inode);
1488 return d_splice_alias(inode, dentry);
1491 static unsigned char btrfs_filetype_table[] = {
1492 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1495 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1497 struct inode *inode = filp->f_dentry->d_inode;
1498 struct btrfs_root *root = BTRFS_I(inode)->root;
1499 struct btrfs_item *item;
1500 struct btrfs_dir_item *di;
1501 struct btrfs_key key;
1502 struct btrfs_key found_key;
1503 struct btrfs_path *path;
1506 struct extent_buffer *leaf;
1509 unsigned char d_type;
1514 int key_type = BTRFS_DIR_INDEX_KEY;
1519 /* FIXME, use a real flag for deciding about the key type */
1520 if (root->fs_info->tree_root == root)
1521 key_type = BTRFS_DIR_ITEM_KEY;
1523 /* special case for "." */
1524 if (filp->f_pos == 0) {
1525 over = filldir(dirent, ".", 1,
1533 mutex_lock(&root->fs_info->fs_mutex);
1534 key.objectid = inode->i_ino;
1535 path = btrfs_alloc_path();
1538 /* special case for .., just use the back ref */
1539 if (filp->f_pos == 1) {
1540 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1542 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1544 leaf = path->nodes[0];
1545 slot = path->slots[0];
1546 nritems = btrfs_header_nritems(leaf);
1547 if (slot >= nritems) {
1548 btrfs_release_path(root, path);
1549 goto read_dir_items;
1551 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1552 btrfs_release_path(root, path);
1553 if (found_key.objectid != key.objectid ||
1554 found_key.type != BTRFS_INODE_REF_KEY)
1555 goto read_dir_items;
1556 over = filldir(dirent, "..", 2,
1557 2, found_key.offset, DT_DIR);
1564 btrfs_set_key_type(&key, key_type);
1565 key.offset = filp->f_pos;
1567 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1572 leaf = path->nodes[0];
1573 nritems = btrfs_header_nritems(leaf);
1574 slot = path->slots[0];
1575 if (advance || slot >= nritems) {
1576 if (slot >= nritems -1) {
1577 ret = btrfs_next_leaf(root, path);
1580 leaf = path->nodes[0];
1581 nritems = btrfs_header_nritems(leaf);
1582 slot = path->slots[0];
1589 item = btrfs_item_nr(leaf, slot);
1590 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1592 if (found_key.objectid != key.objectid)
1594 if (btrfs_key_type(&found_key) != key_type)
1596 if (found_key.offset < filp->f_pos)
1599 filp->f_pos = found_key.offset;
1601 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1603 di_total = btrfs_item_size(leaf, item);
1604 while(di_cur < di_total) {
1605 struct btrfs_key location;
1607 name_len = btrfs_dir_name_len(leaf, di);
1608 if (name_len < 32) {
1609 name_ptr = tmp_name;
1611 name_ptr = kmalloc(name_len, GFP_NOFS);
1614 read_extent_buffer(leaf, name_ptr,
1615 (unsigned long)(di + 1), name_len);
1617 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1618 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1619 over = filldir(dirent, name_ptr, name_len,
1624 if (name_ptr != tmp_name)
1629 di_len = btrfs_dir_name_len(leaf, di) +
1630 btrfs_dir_data_len(leaf, di) +sizeof(*di);
1632 di = (struct btrfs_dir_item *)((char *)di + di_len);
1635 if (key_type == BTRFS_DIR_INDEX_KEY)
1636 filp->f_pos = INT_LIMIT(typeof(filp->f_pos));
1642 btrfs_release_path(root, path);
1643 btrfs_free_path(path);
1644 mutex_unlock(&root->fs_info->fs_mutex);
1648 int btrfs_write_inode(struct inode *inode, int wait)
1650 struct btrfs_root *root = BTRFS_I(inode)->root;
1651 struct btrfs_trans_handle *trans;
1655 mutex_lock(&root->fs_info->fs_mutex);
1656 trans = btrfs_start_transaction(root, 1);
1657 btrfs_set_trans_block_group(trans, inode);
1658 ret = btrfs_commit_transaction(trans, root);
1659 mutex_unlock(&root->fs_info->fs_mutex);
1665 * This is somewhat expensive, updating the tree every time the
1666 * inode changes. But, it is most likely to find the inode in cache.
1667 * FIXME, needs more benchmarking...there are no reasons other than performance
1668 * to keep or drop this code.
1670 void btrfs_dirty_inode(struct inode *inode)
1672 struct btrfs_root *root = BTRFS_I(inode)->root;
1673 struct btrfs_trans_handle *trans;
1675 mutex_lock(&root->fs_info->fs_mutex);
1676 trans = btrfs_start_transaction(root, 1);
1677 btrfs_set_trans_block_group(trans, inode);
1678 btrfs_update_inode(trans, root, inode);
1679 btrfs_end_transaction(trans, root);
1680 mutex_unlock(&root->fs_info->fs_mutex);
1683 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1684 struct btrfs_root *root,
1685 const char *name, int name_len,
1688 struct btrfs_block_group_cache *group,
1691 struct inode *inode;
1692 struct btrfs_inode_item *inode_item;
1693 struct btrfs_block_group_cache *new_inode_group;
1694 struct btrfs_key *location;
1695 struct btrfs_path *path;
1696 struct btrfs_inode_ref *ref;
1697 struct btrfs_key key[2];
1703 path = btrfs_alloc_path();
1706 inode = new_inode(root->fs_info->sb);
1708 return ERR_PTR(-ENOMEM);
1710 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1711 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1712 inode->i_mapping, GFP_NOFS);
1713 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1714 inode->i_mapping, GFP_NOFS);
1715 BTRFS_I(inode)->delalloc_bytes = 0;
1716 BTRFS_I(inode)->root = root;
1722 new_inode_group = btrfs_find_block_group(root, group, 0,
1723 BTRFS_BLOCK_GROUP_METADATA, owner);
1724 if (!new_inode_group) {
1725 printk("find_block group failed\n");
1726 new_inode_group = group;
1728 BTRFS_I(inode)->block_group = new_inode_group;
1729 BTRFS_I(inode)->flags = 0;
1731 key[0].objectid = objectid;
1732 btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
1735 key[1].objectid = objectid;
1736 btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
1737 key[1].offset = ref_objectid;
1739 sizes[0] = sizeof(struct btrfs_inode_item);
1740 sizes[1] = name_len + sizeof(*ref);
1742 ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
1746 if (objectid > root->highest_inode)
1747 root->highest_inode = objectid;
1749 inode->i_uid = current->fsuid;
1750 inode->i_gid = current->fsgid;
1751 inode->i_mode = mode;
1752 inode->i_ino = objectid;
1753 inode->i_blocks = 0;
1754 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1755 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1756 struct btrfs_inode_item);
1757 fill_inode_item(path->nodes[0], inode_item, inode);
1759 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
1760 struct btrfs_inode_ref);
1761 btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
1762 ptr = (unsigned long)(ref + 1);
1763 write_extent_buffer(path->nodes[0], name, ptr, name_len);
1765 btrfs_mark_buffer_dirty(path->nodes[0]);
1766 btrfs_free_path(path);
1768 location = &BTRFS_I(inode)->location;
1769 location->objectid = objectid;
1770 location->offset = 0;
1771 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1773 insert_inode_hash(inode);
1776 btrfs_free_path(path);
1777 return ERR_PTR(ret);
1780 static inline u8 btrfs_inode_type(struct inode *inode)
1782 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1785 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1786 struct dentry *dentry, struct inode *inode,
1790 struct btrfs_key key;
1791 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1792 struct inode *parent_inode;
1794 key.objectid = inode->i_ino;
1795 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1798 ret = btrfs_insert_dir_item(trans, root,
1799 dentry->d_name.name, dentry->d_name.len,
1800 dentry->d_parent->d_inode->i_ino,
1801 &key, btrfs_inode_type(inode));
1804 ret = btrfs_insert_inode_ref(trans, root,
1805 dentry->d_name.name,
1808 dentry->d_parent->d_inode->i_ino);
1810 parent_inode = dentry->d_parent->d_inode;
1811 parent_inode->i_size += dentry->d_name.len * 2;
1812 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1813 ret = btrfs_update_inode(trans, root,
1814 dentry->d_parent->d_inode);
1819 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1820 struct dentry *dentry, struct inode *inode,
1823 int err = btrfs_add_link(trans, dentry, inode, backref);
1825 d_instantiate(dentry, inode);
1833 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1834 int mode, dev_t rdev)
1836 struct btrfs_trans_handle *trans;
1837 struct btrfs_root *root = BTRFS_I(dir)->root;
1838 struct inode *inode = NULL;
1842 unsigned long nr = 0;
1844 if (!new_valid_dev(rdev))
1847 mutex_lock(&root->fs_info->fs_mutex);
1848 err = btrfs_check_free_space(root, 1, 0);
1852 trans = btrfs_start_transaction(root, 1);
1853 btrfs_set_trans_block_group(trans, dir);
1855 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1861 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1863 dentry->d_parent->d_inode->i_ino, objectid,
1864 BTRFS_I(dir)->block_group, mode);
1865 err = PTR_ERR(inode);
1869 btrfs_set_trans_block_group(trans, inode);
1870 err = btrfs_add_nondir(trans, dentry, inode, 0);
1874 inode->i_op = &btrfs_special_inode_operations;
1875 init_special_inode(inode, inode->i_mode, rdev);
1876 btrfs_update_inode(trans, root, inode);
1878 dir->i_sb->s_dirt = 1;
1879 btrfs_update_inode_block_group(trans, inode);
1880 btrfs_update_inode_block_group(trans, dir);
1882 nr = trans->blocks_used;
1883 btrfs_end_transaction(trans, root);
1885 mutex_unlock(&root->fs_info->fs_mutex);
1888 inode_dec_link_count(inode);
1891 btrfs_btree_balance_dirty(root, nr);
1892 btrfs_throttle(root);
1896 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1897 int mode, struct nameidata *nd)
1899 struct btrfs_trans_handle *trans;
1900 struct btrfs_root *root = BTRFS_I(dir)->root;
1901 struct inode *inode = NULL;
1904 unsigned long nr = 0;
1907 mutex_lock(&root->fs_info->fs_mutex);
1908 err = btrfs_check_free_space(root, 1, 0);
1911 trans = btrfs_start_transaction(root, 1);
1912 btrfs_set_trans_block_group(trans, dir);
1914 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1920 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1922 dentry->d_parent->d_inode->i_ino,
1923 objectid, BTRFS_I(dir)->block_group, mode);
1924 err = PTR_ERR(inode);
1928 btrfs_set_trans_block_group(trans, inode);
1929 err = btrfs_add_nondir(trans, dentry, inode, 0);
1933 inode->i_mapping->a_ops = &btrfs_aops;
1934 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
1935 inode->i_fop = &btrfs_file_operations;
1936 inode->i_op = &btrfs_file_inode_operations;
1937 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1938 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1939 inode->i_mapping, GFP_NOFS);
1940 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1941 inode->i_mapping, GFP_NOFS);
1942 BTRFS_I(inode)->delalloc_bytes = 0;
1943 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
1945 dir->i_sb->s_dirt = 1;
1946 btrfs_update_inode_block_group(trans, inode);
1947 btrfs_update_inode_block_group(trans, dir);
1949 nr = trans->blocks_used;
1950 btrfs_end_transaction(trans, root);
1952 mutex_unlock(&root->fs_info->fs_mutex);
1955 inode_dec_link_count(inode);
1958 btrfs_btree_balance_dirty(root, nr);
1959 btrfs_throttle(root);
1963 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1964 struct dentry *dentry)
1966 struct btrfs_trans_handle *trans;
1967 struct btrfs_root *root = BTRFS_I(dir)->root;
1968 struct inode *inode = old_dentry->d_inode;
1969 unsigned long nr = 0;
1973 if (inode->i_nlink == 0)
1976 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1981 mutex_lock(&root->fs_info->fs_mutex);
1982 err = btrfs_check_free_space(root, 1, 0);
1985 trans = btrfs_start_transaction(root, 1);
1987 btrfs_set_trans_block_group(trans, dir);
1988 atomic_inc(&inode->i_count);
1989 err = btrfs_add_nondir(trans, dentry, inode, 1);
1994 dir->i_sb->s_dirt = 1;
1995 btrfs_update_inode_block_group(trans, dir);
1996 err = btrfs_update_inode(trans, root, inode);
2001 nr = trans->blocks_used;
2002 btrfs_end_transaction(trans, root);
2004 mutex_unlock(&root->fs_info->fs_mutex);
2007 inode_dec_link_count(inode);
2010 btrfs_btree_balance_dirty(root, nr);
2011 btrfs_throttle(root);
2015 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
2017 struct inode *inode;
2018 struct btrfs_trans_handle *trans;
2019 struct btrfs_root *root = BTRFS_I(dir)->root;
2021 int drop_on_err = 0;
2023 unsigned long nr = 1;
2025 mutex_lock(&root->fs_info->fs_mutex);
2026 err = btrfs_check_free_space(root, 1, 0);
2030 trans = btrfs_start_transaction(root, 1);
2031 btrfs_set_trans_block_group(trans, dir);
2033 if (IS_ERR(trans)) {
2034 err = PTR_ERR(trans);
2038 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2044 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2046 dentry->d_parent->d_inode->i_ino, objectid,
2047 BTRFS_I(dir)->block_group, S_IFDIR | mode);
2048 if (IS_ERR(inode)) {
2049 err = PTR_ERR(inode);
2054 inode->i_op = &btrfs_dir_inode_operations;
2055 inode->i_fop = &btrfs_dir_file_operations;
2056 btrfs_set_trans_block_group(trans, inode);
2059 err = btrfs_update_inode(trans, root, inode);
2063 err = btrfs_add_link(trans, dentry, inode, 0);
2067 d_instantiate(dentry, inode);
2069 dir->i_sb->s_dirt = 1;
2070 btrfs_update_inode_block_group(trans, inode);
2071 btrfs_update_inode_block_group(trans, dir);
2074 nr = trans->blocks_used;
2075 btrfs_end_transaction(trans, root);
2078 mutex_unlock(&root->fs_info->fs_mutex);
2081 btrfs_btree_balance_dirty(root, nr);
2082 btrfs_throttle(root);
2086 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
2087 size_t pg_offset, u64 start, u64 len,
2093 u64 extent_start = 0;
2095 u64 objectid = inode->i_ino;
2097 struct btrfs_path *path;
2098 struct btrfs_root *root = BTRFS_I(inode)->root;
2099 struct btrfs_file_extent_item *item;
2100 struct extent_buffer *leaf;
2101 struct btrfs_key found_key;
2102 struct extent_map *em = NULL;
2103 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2104 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2105 struct btrfs_trans_handle *trans = NULL;
2107 path = btrfs_alloc_path();
2109 mutex_lock(&root->fs_info->fs_mutex);
2112 spin_lock(&em_tree->lock);
2113 em = lookup_extent_mapping(em_tree, start, len);
2114 spin_unlock(&em_tree->lock);
2117 if (em->start > start) {
2118 printk("get_extent lookup [%Lu %Lu] em [%Lu %Lu]\n",
2119 start, len, em->start, em->len);
2122 if (em->block_start == EXTENT_MAP_INLINE && page)
2123 free_extent_map(em);
2127 em = alloc_extent_map(GFP_NOFS);
2133 em->start = EXTENT_MAP_HOLE;
2135 em->bdev = inode->i_sb->s_bdev;
2136 ret = btrfs_lookup_file_extent(trans, root, path,
2137 objectid, start, trans != NULL);
2144 if (path->slots[0] == 0)
2149 leaf = path->nodes[0];
2150 item = btrfs_item_ptr(leaf, path->slots[0],
2151 struct btrfs_file_extent_item);
2152 /* are we inside the extent that was found? */
2153 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2154 found_type = btrfs_key_type(&found_key);
2155 if (found_key.objectid != objectid ||
2156 found_type != BTRFS_EXTENT_DATA_KEY) {
2160 found_type = btrfs_file_extent_type(leaf, item);
2161 extent_start = found_key.offset;
2162 if (found_type == BTRFS_FILE_EXTENT_REG) {
2163 extent_end = extent_start +
2164 btrfs_file_extent_num_bytes(leaf, item);
2166 if (start < extent_start || start >= extent_end) {
2168 if (start < extent_start) {
2169 if (start + len <= extent_start)
2171 em->len = extent_end - extent_start;
2177 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
2179 em->start = extent_start;
2180 em->len = extent_end - extent_start;
2181 em->block_start = EXTENT_MAP_HOLE;
2184 bytenr += btrfs_file_extent_offset(leaf, item);
2185 em->block_start = bytenr;
2186 em->start = extent_start;
2187 em->len = extent_end - extent_start;
2189 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
2194 size_t extent_offset;
2197 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
2199 extent_end = (extent_start + size + root->sectorsize - 1) &
2200 ~((u64)root->sectorsize - 1);
2201 if (start < extent_start || start >= extent_end) {
2203 if (start < extent_start) {
2204 if (start + len <= extent_start)
2206 em->len = extent_end - extent_start;
2212 em->block_start = EXTENT_MAP_INLINE;
2215 em->start = extent_start;
2220 page_start = page_offset(page) + pg_offset;
2221 extent_offset = page_start - extent_start;
2222 copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
2223 size - extent_offset);
2224 em->start = extent_start + extent_offset;
2225 em->len = (copy_size + root->sectorsize - 1) &
2226 ~((u64)root->sectorsize - 1);
2228 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
2229 if (create == 0 && !PageUptodate(page)) {
2230 read_extent_buffer(leaf, map + pg_offset, ptr,
2232 flush_dcache_page(page);
2233 } else if (create && PageUptodate(page)) {
2236 free_extent_map(em);
2238 btrfs_release_path(root, path);
2239 trans = btrfs_start_transaction(root, 1);
2242 write_extent_buffer(leaf, map + pg_offset, ptr,
2244 btrfs_mark_buffer_dirty(leaf);
2247 set_extent_uptodate(io_tree, em->start,
2248 extent_map_end(em) - 1, GFP_NOFS);
2251 printk("unkknown found_type %d\n", found_type);
2258 em->block_start = EXTENT_MAP_HOLE;
2260 btrfs_release_path(root, path);
2261 if (em->start > start || extent_map_end(em) <= start) {
2262 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len);
2268 spin_lock(&em_tree->lock);
2269 ret = add_extent_mapping(em_tree, em);
2270 if (ret == -EEXIST) {
2271 free_extent_map(em);
2272 em = lookup_extent_mapping(em_tree, start, len);
2275 printk("failing to insert %Lu %Lu\n", start, len);
2278 spin_unlock(&em_tree->lock);
2280 btrfs_free_path(path);
2282 ret = btrfs_end_transaction(trans, root);
2286 mutex_unlock(&root->fs_info->fs_mutex);
2288 free_extent_map(em);
2290 return ERR_PTR(err);
2295 static int btrfs_get_block(struct inode *inode, sector_t iblock,
2296 struct buffer_head *bh_result, int create)
2298 struct extent_map *em;
2299 u64 start = (u64)iblock << inode->i_blkbits;
2300 struct btrfs_multi_bio *multi = NULL;
2301 struct btrfs_root *root = BTRFS_I(inode)->root;
2307 em = btrfs_get_extent(inode, NULL, 0, start, bh_result->b_size, 0);
2309 if (!em || IS_ERR(em))
2312 if (em->start > start || em->start + em->len <= start)
2315 if (em->block_start == EXTENT_MAP_INLINE) {
2320 if (em->block_start == EXTENT_MAP_HOLE ||
2321 em->block_start == EXTENT_MAP_DELALLOC) {
2325 len = em->start + em->len - start;
2326 len = min_t(u64, len, INT_LIMIT(typeof(bh_result->b_size)));
2328 logical = start - em->start;
2329 logical = em->block_start + logical;
2332 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
2333 logical, &map_length, &multi, 0);
2335 bh_result->b_blocknr = multi->stripes[0].physical >> inode->i_blkbits;
2336 bh_result->b_size = min(map_length, len);
2337 bh_result->b_bdev = multi->stripes[0].dev->bdev;
2338 set_buffer_mapped(bh_result);
2341 free_extent_map(em);
2345 static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
2346 const struct iovec *iov, loff_t offset,
2347 unsigned long nr_segs)
2349 struct file *file = iocb->ki_filp;
2350 struct inode *inode = file->f_mapping->host;
2355 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2356 offset, nr_segs, btrfs_get_block, NULL);
2359 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
2361 return extent_bmap(mapping, iblock, btrfs_get_extent);
2364 int btrfs_readpage(struct file *file, struct page *page)
2366 struct extent_io_tree *tree;
2367 tree = &BTRFS_I(page->mapping->host)->io_tree;
2368 return extent_read_full_page(tree, page, btrfs_get_extent);
2371 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
2373 struct extent_io_tree *tree;
2376 if (current->flags & PF_MEMALLOC) {
2377 redirty_page_for_writepage(wbc, page);
2381 tree = &BTRFS_I(page->mapping->host)->io_tree;
2382 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
2385 static int btrfs_writepages(struct address_space *mapping,
2386 struct writeback_control *wbc)
2388 struct extent_io_tree *tree;
2389 tree = &BTRFS_I(mapping->host)->io_tree;
2390 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
2394 btrfs_readpages(struct file *file, struct address_space *mapping,
2395 struct list_head *pages, unsigned nr_pages)
2397 struct extent_io_tree *tree;
2398 tree = &BTRFS_I(mapping->host)->io_tree;
2399 return extent_readpages(tree, mapping, pages, nr_pages,
2403 static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
2405 struct extent_io_tree *tree;
2406 struct extent_map_tree *map;
2409 tree = &BTRFS_I(page->mapping->host)->io_tree;
2410 map = &BTRFS_I(page->mapping->host)->extent_tree;
2411 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
2413 ClearPagePrivate(page);
2414 set_page_private(page, 0);
2415 page_cache_release(page);
2420 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
2422 struct extent_io_tree *tree;
2424 tree = &BTRFS_I(page->mapping->host)->io_tree;
2425 extent_invalidatepage(tree, page, offset);
2426 btrfs_releasepage(page, GFP_NOFS);
2430 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2431 * called from a page fault handler when a page is first dirtied. Hence we must
2432 * be careful to check for EOF conditions here. We set the page up correctly
2433 * for a written page which means we get ENOSPC checking when writing into
2434 * holes and correct delalloc and unwritten extent mapping on filesystems that
2435 * support these features.
2437 * We are not allowed to take the i_mutex here so we have to play games to
2438 * protect against truncate races as the page could now be beyond EOF. Because
2439 * vmtruncate() writes the inode size before removing pages, once we have the
2440 * page lock we can determine safely if the page is beyond EOF. If it is not
2441 * beyond EOF, then the page is guaranteed safe against truncation until we
2444 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
2446 struct inode *inode = fdentry(vma->vm_file)->d_inode;
2447 struct btrfs_root *root = BTRFS_I(inode)->root;
2453 mutex_lock(&root->fs_info->fs_mutex);
2454 ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
2455 mutex_unlock(&root->fs_info->fs_mutex);
2462 wait_on_page_writeback(page);
2463 size = i_size_read(inode);
2464 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2466 if ((page->mapping != inode->i_mapping) ||
2467 (page_start > size)) {
2468 /* page got truncated out from underneath us */
2472 /* page is wholly or partially inside EOF */
2473 if (page_start + PAGE_CACHE_SIZE > size)
2474 end = size & ~PAGE_CACHE_MASK;
2476 end = PAGE_CACHE_SIZE;
2478 ret = btrfs_cow_one_page(inode, page, end);
2486 static void btrfs_truncate(struct inode *inode)
2488 struct btrfs_root *root = BTRFS_I(inode)->root;
2490 struct btrfs_trans_handle *trans;
2493 if (!S_ISREG(inode->i_mode))
2495 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
2498 btrfs_truncate_page(inode->i_mapping, inode->i_size);
2500 mutex_lock(&root->fs_info->fs_mutex);
2501 trans = btrfs_start_transaction(root, 1);
2502 btrfs_set_trans_block_group(trans, inode);
2504 /* FIXME, add redo link to tree so we don't leak on crash */
2505 ret = btrfs_truncate_in_trans(trans, root, inode,
2506 BTRFS_EXTENT_DATA_KEY);
2507 btrfs_update_inode(trans, root, inode);
2508 nr = trans->blocks_used;
2510 ret = btrfs_end_transaction(trans, root);
2512 mutex_unlock(&root->fs_info->fs_mutex);
2513 btrfs_btree_balance_dirty(root, nr);
2514 btrfs_throttle(root);
2517 static int noinline create_subvol(struct btrfs_root *root, char *name,
2520 struct btrfs_trans_handle *trans;
2521 struct btrfs_key key;
2522 struct btrfs_root_item root_item;
2523 struct btrfs_inode_item *inode_item;
2524 struct extent_buffer *leaf;
2525 struct btrfs_root *new_root = root;
2526 struct inode *inode;
2531 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
2532 unsigned long nr = 1;
2534 mutex_lock(&root->fs_info->fs_mutex);
2535 ret = btrfs_check_free_space(root, 1, 0);
2539 trans = btrfs_start_transaction(root, 1);
2542 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
2547 leaf = __btrfs_alloc_free_block(trans, root, root->leafsize,
2548 objectid, trans->transid, 0, 0,
2551 return PTR_ERR(leaf);
2553 btrfs_set_header_nritems(leaf, 0);
2554 btrfs_set_header_level(leaf, 0);
2555 btrfs_set_header_bytenr(leaf, leaf->start);
2556 btrfs_set_header_generation(leaf, trans->transid);
2557 btrfs_set_header_owner(leaf, objectid);
2559 write_extent_buffer(leaf, root->fs_info->fsid,
2560 (unsigned long)btrfs_header_fsid(leaf),
2562 btrfs_mark_buffer_dirty(leaf);
2564 inode_item = &root_item.inode;
2565 memset(inode_item, 0, sizeof(*inode_item));
2566 inode_item->generation = cpu_to_le64(1);
2567 inode_item->size = cpu_to_le64(3);
2568 inode_item->nlink = cpu_to_le32(1);
2569 inode_item->nblocks = cpu_to_le64(1);
2570 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
2572 btrfs_set_root_bytenr(&root_item, leaf->start);
2573 btrfs_set_root_level(&root_item, 0);
2574 btrfs_set_root_refs(&root_item, 1);
2575 btrfs_set_root_used(&root_item, 0);
2577 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
2578 root_item.drop_level = 0;
2580 free_extent_buffer(leaf);
2583 btrfs_set_root_dirid(&root_item, new_dirid);
2585 key.objectid = objectid;
2587 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
2588 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
2594 * insert the directory item
2596 key.offset = (u64)-1;
2597 dir = root->fs_info->sb->s_root->d_inode;
2598 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
2599 name, namelen, dir->i_ino, &key,
2604 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
2605 name, namelen, objectid,
2606 root->fs_info->sb->s_root->d_inode->i_ino);
2610 ret = btrfs_commit_transaction(trans, root);
2614 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
2617 trans = btrfs_start_transaction(new_root, 1);
2620 inode = btrfs_new_inode(trans, new_root, "..", 2, new_dirid,
2622 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
2625 inode->i_op = &btrfs_dir_inode_operations;
2626 inode->i_fop = &btrfs_dir_file_operations;
2627 new_root->inode = inode;
2629 ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid,
2633 ret = btrfs_update_inode(trans, new_root, inode);
2637 nr = trans->blocks_used;
2638 err = btrfs_commit_transaction(trans, new_root);
2642 mutex_unlock(&root->fs_info->fs_mutex);
2643 btrfs_btree_balance_dirty(root, nr);
2644 btrfs_throttle(root);
2648 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
2650 struct btrfs_pending_snapshot *pending_snapshot;
2651 struct btrfs_trans_handle *trans;
2654 unsigned long nr = 0;
2656 if (!root->ref_cows)
2659 mutex_lock(&root->fs_info->fs_mutex);
2660 ret = btrfs_check_free_space(root, 1, 0);
2664 pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS);
2665 if (!pending_snapshot) {
2669 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
2670 if (!pending_snapshot->name) {
2672 kfree(pending_snapshot);
2675 memcpy(pending_snapshot->name, name, namelen);
2676 pending_snapshot->name[namelen] = '\0';
2677 trans = btrfs_start_transaction(root, 1);
2679 pending_snapshot->root = root;
2680 list_add(&pending_snapshot->list,
2681 &trans->transaction->pending_snapshots);
2682 ret = btrfs_update_inode(trans, root, root->inode);
2683 err = btrfs_commit_transaction(trans, root);
2686 mutex_unlock(&root->fs_info->fs_mutex);
2687 btrfs_btree_balance_dirty(root, nr);
2688 btrfs_throttle(root);
2692 unsigned long btrfs_force_ra(struct address_space *mapping,
2693 struct file_ra_state *ra, struct file *file,
2694 pgoff_t offset, pgoff_t last_index)
2698 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2699 req_size = last_index - offset + 1;
2700 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2703 req_size = min(last_index - offset + 1, (pgoff_t)128);
2704 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2705 return offset + req_size;
2709 int btrfs_defrag_file(struct file *file) {
2710 struct inode *inode = fdentry(file)->d_inode;
2711 struct btrfs_root *root = BTRFS_I(inode)->root;
2712 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2714 unsigned long last_index;
2715 unsigned long ra_index = 0;
2721 mutex_lock(&root->fs_info->fs_mutex);
2722 ret = btrfs_check_free_space(root, inode->i_size, 0);
2723 mutex_unlock(&root->fs_info->fs_mutex);
2727 mutex_lock(&inode->i_mutex);
2728 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
2729 for (i = 0; i <= last_index; i++) {
2730 if (i == ra_index) {
2731 ra_index = btrfs_force_ra(inode->i_mapping,
2733 file, ra_index, last_index);
2735 page = grab_cache_page(inode->i_mapping, i);
2738 if (!PageUptodate(page)) {
2739 btrfs_readpage(NULL, page);
2741 if (!PageUptodate(page)) {
2743 page_cache_release(page);
2747 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2748 page_end = page_start + PAGE_CACHE_SIZE - 1;
2750 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
2751 set_extent_delalloc(io_tree, page_start,
2752 page_end, GFP_NOFS);
2754 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
2755 set_page_dirty(page);
2757 page_cache_release(page);
2758 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
2762 mutex_unlock(&inode->i_mutex);
2766 static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
2770 struct btrfs_ioctl_vol_args *vol_args;
2771 struct btrfs_trans_handle *trans;
2777 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
2782 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
2786 namelen = strlen(vol_args->name);
2787 if (namelen > BTRFS_VOL_NAME_MAX) {
2792 sizestr = vol_args->name;
2793 if (!strcmp(sizestr, "max"))
2794 new_size = root->fs_info->sb->s_bdev->bd_inode->i_size;
2796 if (sizestr[0] == '-') {
2799 } else if (sizestr[0] == '+') {
2803 new_size = btrfs_parse_size(sizestr);
2804 if (new_size == 0) {
2810 mutex_lock(&root->fs_info->fs_mutex);
2811 old_size = btrfs_super_total_bytes(&root->fs_info->super_copy);
2814 if (new_size > old_size) {
2818 new_size = old_size - new_size;
2819 } else if (mod > 0) {
2820 new_size = old_size + new_size;
2823 if (new_size < 256 * 1024 * 1024) {
2827 if (new_size > root->fs_info->sb->s_bdev->bd_inode->i_size) {
2832 do_div(new_size, root->sectorsize);
2833 new_size *= root->sectorsize;
2835 printk("new size is %Lu\n", new_size);
2836 if (new_size > old_size) {
2837 trans = btrfs_start_transaction(root, 1);
2838 ret = btrfs_grow_extent_tree(trans, root, new_size);
2839 btrfs_commit_transaction(trans, root);
2841 ret = btrfs_shrink_extent_tree(root, new_size);
2845 mutex_unlock(&root->fs_info->fs_mutex);
2851 static int noinline btrfs_ioctl_snap_create(struct btrfs_root *root,
2854 struct btrfs_ioctl_vol_args *vol_args;
2855 struct btrfs_dir_item *di;
2856 struct btrfs_path *path;
2861 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
2866 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
2871 namelen = strlen(vol_args->name);
2872 if (namelen > BTRFS_VOL_NAME_MAX) {
2876 if (strchr(vol_args->name, '/')) {
2881 path = btrfs_alloc_path();
2887 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2888 mutex_lock(&root->fs_info->fs_mutex);
2889 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2891 vol_args->name, namelen, 0);
2892 mutex_unlock(&root->fs_info->fs_mutex);
2893 btrfs_free_path(path);
2895 if (di && !IS_ERR(di)) {
2905 if (root == root->fs_info->tree_root)
2906 ret = create_subvol(root, vol_args->name, namelen);
2908 ret = create_snapshot(root, vol_args->name, namelen);
2914 static int btrfs_ioctl_defrag(struct file *file)
2916 struct inode *inode = fdentry(file)->d_inode;
2917 struct btrfs_root *root = BTRFS_I(inode)->root;
2919 switch (inode->i_mode & S_IFMT) {
2921 mutex_lock(&root->fs_info->fs_mutex);
2922 btrfs_defrag_root(root, 0);
2923 btrfs_defrag_root(root->fs_info->extent_root, 0);
2924 mutex_unlock(&root->fs_info->fs_mutex);
2927 btrfs_defrag_file(file);
2934 long btrfs_ioctl(struct file *file, unsigned int
2935 cmd, unsigned long arg)
2937 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2940 case BTRFS_IOC_SNAP_CREATE:
2941 return btrfs_ioctl_snap_create(root, (void __user *)arg);
2942 case BTRFS_IOC_DEFRAG:
2943 return btrfs_ioctl_defrag(file);
2944 case BTRFS_IOC_RESIZE:
2945 return btrfs_ioctl_resize(root, (void __user *)arg);
2952 * Called inside transaction, so use GFP_NOFS
2954 struct inode *btrfs_alloc_inode(struct super_block *sb)
2956 struct btrfs_inode *ei;
2958 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2962 ei->ordered_trans = 0;
2963 return &ei->vfs_inode;
2966 void btrfs_destroy_inode(struct inode *inode)
2968 WARN_ON(!list_empty(&inode->i_dentry));
2969 WARN_ON(inode->i_data.nrpages);
2971 btrfs_drop_extent_cache(inode, 0, (u64)-1);
2972 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2975 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2976 static void init_once(struct kmem_cache * cachep, void *foo)
2978 static void init_once(void * foo, struct kmem_cache * cachep,
2979 unsigned long flags)
2982 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2984 inode_init_once(&ei->vfs_inode);
2987 void btrfs_destroy_cachep(void)
2989 if (btrfs_inode_cachep)
2990 kmem_cache_destroy(btrfs_inode_cachep);
2991 if (btrfs_trans_handle_cachep)
2992 kmem_cache_destroy(btrfs_trans_handle_cachep);
2993 if (btrfs_transaction_cachep)
2994 kmem_cache_destroy(btrfs_transaction_cachep);
2995 if (btrfs_bit_radix_cachep)
2996 kmem_cache_destroy(btrfs_bit_radix_cachep);
2997 if (btrfs_path_cachep)
2998 kmem_cache_destroy(btrfs_path_cachep);
3001 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
3002 unsigned long extra_flags,
3003 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3004 void (*ctor)(struct kmem_cache *, void *)
3006 void (*ctor)(void *, struct kmem_cache *,
3011 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
3012 SLAB_MEM_SPREAD | extra_flags), ctor
3013 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3019 int btrfs_init_cachep(void)
3021 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
3022 sizeof(struct btrfs_inode),
3024 if (!btrfs_inode_cachep)
3026 btrfs_trans_handle_cachep =
3027 btrfs_cache_create("btrfs_trans_handle_cache",
3028 sizeof(struct btrfs_trans_handle),
3030 if (!btrfs_trans_handle_cachep)
3032 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
3033 sizeof(struct btrfs_transaction),
3035 if (!btrfs_transaction_cachep)
3037 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
3038 sizeof(struct btrfs_path),
3040 if (!btrfs_path_cachep)
3042 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
3043 SLAB_DESTROY_BY_RCU, NULL);
3044 if (!btrfs_bit_radix_cachep)
3048 btrfs_destroy_cachep();
3052 static int btrfs_getattr(struct vfsmount *mnt,
3053 struct dentry *dentry, struct kstat *stat)
3055 struct inode *inode = dentry->d_inode;
3056 generic_fillattr(inode, stat);
3057 stat->blksize = PAGE_CACHE_SIZE;
3058 stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9);
3062 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
3063 struct inode * new_dir,struct dentry *new_dentry)
3065 struct btrfs_trans_handle *trans;
3066 struct btrfs_root *root = BTRFS_I(old_dir)->root;
3067 struct inode *new_inode = new_dentry->d_inode;
3068 struct inode *old_inode = old_dentry->d_inode;
3069 struct timespec ctime = CURRENT_TIME;
3070 struct btrfs_path *path;
3073 if (S_ISDIR(old_inode->i_mode) && new_inode &&
3074 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
3078 mutex_lock(&root->fs_info->fs_mutex);
3079 ret = btrfs_check_free_space(root, 1, 0);
3083 trans = btrfs_start_transaction(root, 1);
3085 btrfs_set_trans_block_group(trans, new_dir);
3086 path = btrfs_alloc_path();
3092 old_dentry->d_inode->i_nlink++;
3093 old_dir->i_ctime = old_dir->i_mtime = ctime;
3094 new_dir->i_ctime = new_dir->i_mtime = ctime;
3095 old_inode->i_ctime = ctime;
3097 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
3102 new_inode->i_ctime = CURRENT_TIME;
3103 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
3107 ret = btrfs_add_link(trans, new_dentry, old_inode, 1);
3112 btrfs_free_path(path);
3113 btrfs_end_transaction(trans, root);
3115 mutex_unlock(&root->fs_info->fs_mutex);
3119 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
3120 const char *symname)
3122 struct btrfs_trans_handle *trans;
3123 struct btrfs_root *root = BTRFS_I(dir)->root;
3124 struct btrfs_path *path;
3125 struct btrfs_key key;
3126 struct inode *inode = NULL;
3133 struct btrfs_file_extent_item *ei;
3134 struct extent_buffer *leaf;
3135 unsigned long nr = 0;
3137 name_len = strlen(symname) + 1;
3138 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
3139 return -ENAMETOOLONG;
3141 mutex_lock(&root->fs_info->fs_mutex);
3142 err = btrfs_check_free_space(root, 1, 0);
3146 trans = btrfs_start_transaction(root, 1);
3147 btrfs_set_trans_block_group(trans, dir);
3149 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
3155 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
3157 dentry->d_parent->d_inode->i_ino, objectid,
3158 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
3159 err = PTR_ERR(inode);
3163 btrfs_set_trans_block_group(trans, inode);
3164 err = btrfs_add_nondir(trans, dentry, inode, 0);
3168 inode->i_mapping->a_ops = &btrfs_aops;
3169 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3170 inode->i_fop = &btrfs_file_operations;
3171 inode->i_op = &btrfs_file_inode_operations;
3172 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
3173 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
3174 inode->i_mapping, GFP_NOFS);
3175 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
3176 inode->i_mapping, GFP_NOFS);
3177 BTRFS_I(inode)->delalloc_bytes = 0;
3178 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
3180 dir->i_sb->s_dirt = 1;
3181 btrfs_update_inode_block_group(trans, inode);
3182 btrfs_update_inode_block_group(trans, dir);
3186 path = btrfs_alloc_path();
3188 key.objectid = inode->i_ino;
3190 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
3191 datasize = btrfs_file_extent_calc_inline_size(name_len);
3192 err = btrfs_insert_empty_item(trans, root, path, &key,
3198 leaf = path->nodes[0];
3199 ei = btrfs_item_ptr(leaf, path->slots[0],
3200 struct btrfs_file_extent_item);
3201 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
3202 btrfs_set_file_extent_type(leaf, ei,
3203 BTRFS_FILE_EXTENT_INLINE);
3204 ptr = btrfs_file_extent_inline_start(ei);
3205 write_extent_buffer(leaf, symname, ptr, name_len);
3206 btrfs_mark_buffer_dirty(leaf);
3207 btrfs_free_path(path);
3209 inode->i_op = &btrfs_symlink_inode_operations;
3210 inode->i_mapping->a_ops = &btrfs_symlink_aops;
3211 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3212 inode->i_size = name_len - 1;
3213 err = btrfs_update_inode(trans, root, inode);
3218 nr = trans->blocks_used;
3219 btrfs_end_transaction(trans, root);
3221 mutex_unlock(&root->fs_info->fs_mutex);
3223 inode_dec_link_count(inode);
3226 btrfs_btree_balance_dirty(root, nr);
3227 btrfs_throttle(root);
3231 static int btrfs_permission(struct inode *inode, int mask,
3232 struct nameidata *nd)
3234 if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
3236 return generic_permission(inode, mask, NULL);
3239 static struct inode_operations btrfs_dir_inode_operations = {
3240 .lookup = btrfs_lookup,
3241 .create = btrfs_create,
3242 .unlink = btrfs_unlink,
3244 .mkdir = btrfs_mkdir,
3245 .rmdir = btrfs_rmdir,
3246 .rename = btrfs_rename,
3247 .symlink = btrfs_symlink,
3248 .setattr = btrfs_setattr,
3249 .mknod = btrfs_mknod,
3250 .setxattr = generic_setxattr,
3251 .getxattr = generic_getxattr,
3252 .listxattr = btrfs_listxattr,
3253 .removexattr = generic_removexattr,
3254 .permission = btrfs_permission,
3256 static struct inode_operations btrfs_dir_ro_inode_operations = {
3257 .lookup = btrfs_lookup,
3258 .permission = btrfs_permission,
3260 static struct file_operations btrfs_dir_file_operations = {
3261 .llseek = generic_file_llseek,
3262 .read = generic_read_dir,
3263 .readdir = btrfs_readdir,
3264 .unlocked_ioctl = btrfs_ioctl,
3265 #ifdef CONFIG_COMPAT
3266 .compat_ioctl = btrfs_ioctl,
3270 static struct extent_io_ops btrfs_extent_io_ops = {
3271 .fill_delalloc = run_delalloc_range,
3272 .submit_bio_hook = btrfs_submit_bio_hook,
3273 .merge_bio_hook = btrfs_merge_bio_hook,
3274 .readpage_io_hook = btrfs_readpage_io_hook,
3275 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
3276 .readpage_io_failed_hook = btrfs_readpage_io_failed_hook,
3277 .set_bit_hook = btrfs_set_bit_hook,
3278 .clear_bit_hook = btrfs_clear_bit_hook,
3281 static struct address_space_operations btrfs_aops = {
3282 .readpage = btrfs_readpage,
3283 .writepage = btrfs_writepage,
3284 .writepages = btrfs_writepages,
3285 .readpages = btrfs_readpages,
3286 .sync_page = block_sync_page,
3288 .direct_IO = btrfs_direct_IO,
3289 .invalidatepage = btrfs_invalidatepage,
3290 .releasepage = btrfs_releasepage,
3291 .set_page_dirty = __set_page_dirty_nobuffers,
3294 static struct address_space_operations btrfs_symlink_aops = {
3295 .readpage = btrfs_readpage,
3296 .writepage = btrfs_writepage,
3297 .invalidatepage = btrfs_invalidatepage,
3298 .releasepage = btrfs_releasepage,
3301 static struct inode_operations btrfs_file_inode_operations = {
3302 .truncate = btrfs_truncate,
3303 .getattr = btrfs_getattr,
3304 .setattr = btrfs_setattr,
3305 .setxattr = generic_setxattr,
3306 .getxattr = generic_getxattr,
3307 .listxattr = btrfs_listxattr,
3308 .removexattr = generic_removexattr,
3309 .permission = btrfs_permission,
3311 static struct inode_operations btrfs_special_inode_operations = {
3312 .getattr = btrfs_getattr,
3313 .setattr = btrfs_setattr,
3314 .permission = btrfs_permission,
3316 static struct inode_operations btrfs_symlink_inode_operations = {
3317 .readlink = generic_readlink,
3318 .follow_link = page_follow_link_light,
3319 .put_link = page_put_link,
3320 .permission = btrfs_permission,