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;
334 if (!(rw & (1 << BIO_RW))) {
335 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
340 if (btrfs_test_opt(root, NODATASUM) ||
341 btrfs_test_flag(inode, NODATASUM)) {
345 mutex_lock(&root->fs_info->fs_mutex);
346 trans = btrfs_start_transaction(root, 1);
347 btrfs_set_trans_block_group(trans, inode);
348 btrfs_csum_file_blocks(trans, root, inode, bio);
349 ret = btrfs_end_transaction(trans, root);
351 mutex_unlock(&root->fs_info->fs_mutex);
353 return btrfs_map_bio(root, rw, bio, mirror_num);
356 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
359 struct inode *inode = page->mapping->host;
360 struct btrfs_root *root = BTRFS_I(inode)->root;
361 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
362 struct btrfs_csum_item *item;
363 struct btrfs_path *path = NULL;
365 if (btrfs_test_opt(root, NODATASUM) ||
366 btrfs_test_flag(inode, NODATASUM))
368 mutex_lock(&root->fs_info->fs_mutex);
369 path = btrfs_alloc_path();
370 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
373 /* a csum that isn't present is a preallocated region. */
374 if (ret == -ENOENT || ret == -EFBIG)
377 printk("no csum found for inode %lu start %Lu\n", inode->i_ino, start);
380 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
382 set_state_private(io_tree, start, csum);
385 btrfs_free_path(path);
386 mutex_unlock(&root->fs_info->fs_mutex);
390 struct io_failure_record {
398 int btrfs_readpage_io_failed_hook(struct bio *failed_bio,
399 struct page *page, u64 start, u64 end,
400 struct extent_state *state)
402 struct io_failure_record *failrec = NULL;
404 struct extent_map *em;
405 struct inode *inode = page->mapping->host;
406 struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
412 ret = get_state_private(failure_tree, start, &private);
414 size_t pg_offset = start - page_offset(page);
415 failrec = kmalloc(sizeof(*failrec), GFP_NOFS);
418 failrec->start = start;
419 failrec->len = end - start + 1;
420 failrec->last_mirror = 0;
422 em = btrfs_get_extent(inode, NULL, pg_offset, start,
425 if (!em || IS_ERR(em)) {
429 logical = start - em->start;
430 logical = em->block_start + logical;
431 failrec->logical = logical;
433 set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
434 EXTENT_DIRTY, GFP_NOFS);
435 set_state_private(failure_tree, start,
436 (u64)(unsigned long)failrec);
438 failrec = (struct io_failure_record *)(unsigned long)private;
440 num_copies = btrfs_num_copies(
441 &BTRFS_I(inode)->root->fs_info->mapping_tree,
442 failrec->logical, failrec->len);
443 failrec->last_mirror++;
445 spin_lock_irq(&BTRFS_I(inode)->io_tree.lock);
446 state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
449 if (state && state->start != failrec->start)
451 spin_unlock_irq(&BTRFS_I(inode)->io_tree.lock);
453 if (!state || failrec->last_mirror > num_copies) {
454 set_state_private(failure_tree, failrec->start, 0);
455 clear_extent_bits(failure_tree, failrec->start,
456 failrec->start + failrec->len - 1,
457 EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
461 bio = bio_alloc(GFP_NOFS, 1);
462 bio->bi_private = state;
463 bio->bi_end_io = failed_bio->bi_end_io;
464 bio->bi_sector = failrec->logical >> 9;
465 bio->bi_bdev = failed_bio->bi_bdev;
466 bio_add_page(bio, page, failrec->len, start - page_offset(page));
467 btrfs_submit_bio_hook(inode, READ, bio, failrec->last_mirror);
471 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
472 struct extent_state *state)
474 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
475 struct inode *inode = page->mapping->host;
476 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
478 u64 private = ~(u32)0;
480 struct btrfs_root *root = BTRFS_I(inode)->root;
484 if (btrfs_test_opt(root, NODATASUM) ||
485 btrfs_test_flag(inode, NODATASUM))
487 if (state && state->start == start) {
488 private = state->private;
491 ret = get_state_private(io_tree, start, &private);
493 local_irq_save(flags);
494 kaddr = kmap_atomic(page, KM_IRQ0);
498 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
499 btrfs_csum_final(csum, (char *)&csum);
500 if (csum != private) {
503 kunmap_atomic(kaddr, KM_IRQ0);
504 local_irq_restore(flags);
506 /* if the io failure tree for this inode is non-empty,
507 * check to see if we've recovered from a failed IO
510 if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
511 (u64)-1, 1, EXTENT_DIRTY)) {
513 struct io_failure_record *failure;
514 ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
515 start, &private_failure);
517 failure = (struct io_failure_record *)(unsigned long)
519 set_state_private(&BTRFS_I(inode)->io_failure_tree,
521 clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
523 failure->start + failure->len - 1,
524 EXTENT_DIRTY | EXTENT_LOCKED,
532 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
533 page->mapping->host->i_ino, (unsigned long long)start, csum,
535 memset(kaddr + offset, 1, end - start + 1);
536 flush_dcache_page(page);
537 kunmap_atomic(kaddr, KM_IRQ0);
538 local_irq_restore(flags);
542 void btrfs_read_locked_inode(struct inode *inode)
544 struct btrfs_path *path;
545 struct extent_buffer *leaf;
546 struct btrfs_inode_item *inode_item;
547 struct btrfs_timespec *tspec;
548 struct btrfs_root *root = BTRFS_I(inode)->root;
549 struct btrfs_key location;
550 u64 alloc_group_block;
554 path = btrfs_alloc_path();
556 mutex_lock(&root->fs_info->fs_mutex);
557 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
559 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
563 leaf = path->nodes[0];
564 inode_item = btrfs_item_ptr(leaf, path->slots[0],
565 struct btrfs_inode_item);
567 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
568 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
569 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
570 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
571 inode->i_size = btrfs_inode_size(leaf, inode_item);
573 tspec = btrfs_inode_atime(inode_item);
574 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
575 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
577 tspec = btrfs_inode_mtime(inode_item);
578 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
579 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
581 tspec = btrfs_inode_ctime(inode_item);
582 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
583 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
585 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
586 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
588 rdev = btrfs_inode_rdev(leaf, inode_item);
590 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
591 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
593 BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
594 if (!BTRFS_I(inode)->block_group) {
595 BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
597 BTRFS_BLOCK_GROUP_METADATA, 0);
599 btrfs_free_path(path);
602 mutex_unlock(&root->fs_info->fs_mutex);
604 switch (inode->i_mode & S_IFMT) {
606 inode->i_mapping->a_ops = &btrfs_aops;
607 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
608 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
609 inode->i_fop = &btrfs_file_operations;
610 inode->i_op = &btrfs_file_inode_operations;
613 inode->i_fop = &btrfs_dir_file_operations;
614 if (root == root->fs_info->tree_root)
615 inode->i_op = &btrfs_dir_ro_inode_operations;
617 inode->i_op = &btrfs_dir_inode_operations;
620 inode->i_op = &btrfs_symlink_inode_operations;
621 inode->i_mapping->a_ops = &btrfs_symlink_aops;
622 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
625 init_special_inode(inode, inode->i_mode, rdev);
631 btrfs_release_path(root, path);
632 btrfs_free_path(path);
633 mutex_unlock(&root->fs_info->fs_mutex);
634 make_bad_inode(inode);
637 static void fill_inode_item(struct extent_buffer *leaf,
638 struct btrfs_inode_item *item,
641 btrfs_set_inode_uid(leaf, item, inode->i_uid);
642 btrfs_set_inode_gid(leaf, item, inode->i_gid);
643 btrfs_set_inode_size(leaf, item, inode->i_size);
644 btrfs_set_inode_mode(leaf, item, inode->i_mode);
645 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
647 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
648 inode->i_atime.tv_sec);
649 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
650 inode->i_atime.tv_nsec);
652 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
653 inode->i_mtime.tv_sec);
654 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
655 inode->i_mtime.tv_nsec);
657 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
658 inode->i_ctime.tv_sec);
659 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
660 inode->i_ctime.tv_nsec);
662 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
663 btrfs_set_inode_generation(leaf, item, inode->i_generation);
664 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
665 btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
666 btrfs_set_inode_block_group(leaf, item,
667 BTRFS_I(inode)->block_group->key.objectid);
670 int btrfs_update_inode(struct btrfs_trans_handle *trans,
671 struct btrfs_root *root,
674 struct btrfs_inode_item *inode_item;
675 struct btrfs_path *path;
676 struct extent_buffer *leaf;
679 path = btrfs_alloc_path();
681 ret = btrfs_lookup_inode(trans, root, path,
682 &BTRFS_I(inode)->location, 1);
689 leaf = path->nodes[0];
690 inode_item = btrfs_item_ptr(leaf, path->slots[0],
691 struct btrfs_inode_item);
693 fill_inode_item(leaf, inode_item, inode);
694 btrfs_mark_buffer_dirty(leaf);
695 btrfs_set_inode_last_trans(trans, inode);
698 btrfs_release_path(root, path);
699 btrfs_free_path(path);
704 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
705 struct btrfs_root *root,
707 struct dentry *dentry)
709 struct btrfs_path *path;
710 const char *name = dentry->d_name.name;
711 int name_len = dentry->d_name.len;
713 struct extent_buffer *leaf;
714 struct btrfs_dir_item *di;
715 struct btrfs_key key;
717 path = btrfs_alloc_path();
723 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
733 leaf = path->nodes[0];
734 btrfs_dir_item_key_to_cpu(leaf, di, &key);
735 ret = btrfs_delete_one_dir_name(trans, root, path, di);
738 btrfs_release_path(root, path);
740 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
741 key.objectid, name, name_len, -1);
750 ret = btrfs_delete_one_dir_name(trans, root, path, di);
752 dentry->d_inode->i_ctime = dir->i_ctime;
753 ret = btrfs_del_inode_ref(trans, root, name, name_len,
754 dentry->d_inode->i_ino,
755 dentry->d_parent->d_inode->i_ino);
757 printk("failed to delete reference to %.*s, "
758 "inode %lu parent %lu\n", name_len, name,
759 dentry->d_inode->i_ino,
760 dentry->d_parent->d_inode->i_ino);
763 btrfs_free_path(path);
765 dir->i_size -= name_len * 2;
766 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
767 btrfs_update_inode(trans, root, dir);
768 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
769 dentry->d_inode->i_nlink--;
771 drop_nlink(dentry->d_inode);
773 ret = btrfs_update_inode(trans, root, dentry->d_inode);
774 dir->i_sb->s_dirt = 1;
779 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
781 struct btrfs_root *root;
782 struct btrfs_trans_handle *trans;
783 struct inode *inode = dentry->d_inode;
785 unsigned long nr = 0;
787 root = BTRFS_I(dir)->root;
788 mutex_lock(&root->fs_info->fs_mutex);
790 ret = btrfs_check_free_space(root, 1, 1);
794 trans = btrfs_start_transaction(root, 1);
796 btrfs_set_trans_block_group(trans, dir);
797 ret = btrfs_unlink_trans(trans, root, dir, dentry);
798 nr = trans->blocks_used;
800 if (inode->i_nlink == 0) {
802 /* if the inode isn't linked anywhere,
803 * we don't need to worry about
806 found = btrfs_del_ordered_inode(inode);
808 atomic_dec(&inode->i_count);
812 btrfs_end_transaction(trans, root);
814 mutex_unlock(&root->fs_info->fs_mutex);
815 btrfs_btree_balance_dirty(root, nr);
816 btrfs_throttle(root);
820 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
822 struct inode *inode = dentry->d_inode;
825 struct btrfs_root *root = BTRFS_I(dir)->root;
826 struct btrfs_trans_handle *trans;
827 unsigned long nr = 0;
829 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
832 mutex_lock(&root->fs_info->fs_mutex);
833 ret = btrfs_check_free_space(root, 1, 1);
837 trans = btrfs_start_transaction(root, 1);
838 btrfs_set_trans_block_group(trans, dir);
840 /* now the directory is empty */
841 err = btrfs_unlink_trans(trans, root, dir, dentry);
846 nr = trans->blocks_used;
847 ret = btrfs_end_transaction(trans, root);
849 mutex_unlock(&root->fs_info->fs_mutex);
850 btrfs_btree_balance_dirty(root, nr);
851 btrfs_throttle(root);
859 * this can truncate away extent items, csum items and directory items.
860 * It starts at a high offset and removes keys until it can't find
861 * any higher than i_size.
863 * csum items that cross the new i_size are truncated to the new size
866 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
867 struct btrfs_root *root,
872 struct btrfs_path *path;
873 struct btrfs_key key;
874 struct btrfs_key found_key;
876 struct extent_buffer *leaf;
877 struct btrfs_file_extent_item *fi;
878 u64 extent_start = 0;
879 u64 extent_num_bytes = 0;
885 int pending_del_nr = 0;
886 int pending_del_slot = 0;
887 int extent_type = -1;
889 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
890 path = btrfs_alloc_path();
894 /* FIXME, add redo link to tree so we don't leak on crash */
895 key.objectid = inode->i_ino;
896 key.offset = (u64)-1;
899 btrfs_init_path(path);
901 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
906 BUG_ON(path->slots[0] == 0);
912 leaf = path->nodes[0];
913 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
914 found_type = btrfs_key_type(&found_key);
916 if (found_key.objectid != inode->i_ino)
919 if (found_type < min_type)
922 item_end = found_key.offset;
923 if (found_type == BTRFS_EXTENT_DATA_KEY) {
924 fi = btrfs_item_ptr(leaf, path->slots[0],
925 struct btrfs_file_extent_item);
926 extent_type = btrfs_file_extent_type(leaf, fi);
927 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
929 btrfs_file_extent_num_bytes(leaf, fi);
930 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
931 struct btrfs_item *item = btrfs_item_nr(leaf,
933 item_end += btrfs_file_extent_inline_len(leaf,
938 if (found_type == BTRFS_CSUM_ITEM_KEY) {
939 ret = btrfs_csum_truncate(trans, root, path,
943 if (item_end < inode->i_size) {
944 if (found_type == BTRFS_DIR_ITEM_KEY) {
945 found_type = BTRFS_INODE_ITEM_KEY;
946 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
947 found_type = BTRFS_CSUM_ITEM_KEY;
948 } else if (found_type == BTRFS_EXTENT_DATA_KEY) {
949 found_type = BTRFS_XATTR_ITEM_KEY;
950 } else if (found_type == BTRFS_XATTR_ITEM_KEY) {
951 found_type = BTRFS_INODE_REF_KEY;
952 } else if (found_type) {
957 btrfs_set_key_type(&key, found_type);
960 if (found_key.offset >= inode->i_size)
966 /* FIXME, shrink the extent if the ref count is only 1 */
967 if (found_type != BTRFS_EXTENT_DATA_KEY)
970 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
972 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
975 btrfs_file_extent_num_bytes(leaf, fi);
976 extent_num_bytes = inode->i_size -
977 found_key.offset + root->sectorsize - 1;
978 extent_num_bytes = extent_num_bytes &
979 ~((u64)root->sectorsize - 1);
980 btrfs_set_file_extent_num_bytes(leaf, fi,
982 num_dec = (orig_num_bytes -
984 if (extent_start != 0)
985 dec_i_blocks(inode, num_dec);
986 btrfs_mark_buffer_dirty(leaf);
989 btrfs_file_extent_disk_num_bytes(leaf,
991 /* FIXME blocksize != 4096 */
992 num_dec = btrfs_file_extent_num_bytes(leaf, fi);
993 if (extent_start != 0) {
995 dec_i_blocks(inode, num_dec);
997 root_gen = btrfs_header_generation(leaf);
998 root_owner = btrfs_header_owner(leaf);
1000 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1002 u32 newsize = inode->i_size - found_key.offset;
1003 dec_i_blocks(inode, item_end + 1 -
1004 found_key.offset - newsize);
1006 btrfs_file_extent_calc_inline_size(newsize);
1007 ret = btrfs_truncate_item(trans, root, path,
1011 dec_i_blocks(inode, item_end + 1 -
1017 if (!pending_del_nr) {
1018 /* no pending yet, add ourselves */
1019 pending_del_slot = path->slots[0];
1021 } else if (pending_del_nr &&
1022 path->slots[0] + 1 == pending_del_slot) {
1023 /* hop on the pending chunk */
1025 pending_del_slot = path->slots[0];
1027 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot);
1033 ret = btrfs_free_extent(trans, root, extent_start,
1036 root_gen, inode->i_ino,
1037 found_key.offset, 0);
1041 if (path->slots[0] == 0) {
1044 btrfs_release_path(root, path);
1049 if (pending_del_nr &&
1050 path->slots[0] + 1 != pending_del_slot) {
1051 struct btrfs_key debug;
1053 btrfs_item_key_to_cpu(path->nodes[0], &debug,
1055 ret = btrfs_del_items(trans, root, path,
1060 btrfs_release_path(root, path);
1066 if (pending_del_nr) {
1067 ret = btrfs_del_items(trans, root, path, pending_del_slot,
1070 btrfs_release_path(root, path);
1071 btrfs_free_path(path);
1072 inode->i_sb->s_dirt = 1;
1076 static int btrfs_cow_one_page(struct inode *inode, struct page *page,
1080 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1081 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
1082 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
1085 WARN_ON(!PageLocked(page));
1086 set_page_extent_mapped(page);
1088 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
1089 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
1090 page_end, GFP_NOFS);
1092 if (zero_start != PAGE_CACHE_SIZE) {
1094 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
1095 flush_dcache_page(page);
1098 set_page_dirty(page);
1099 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
1105 * taken from block_truncate_page, but does cow as it zeros out
1106 * any bytes left in the last page in the file.
1108 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
1110 struct inode *inode = mapping->host;
1111 struct btrfs_root *root = BTRFS_I(inode)->root;
1112 u32 blocksize = root->sectorsize;
1113 pgoff_t index = from >> PAGE_CACHE_SHIFT;
1114 unsigned offset = from & (PAGE_CACHE_SIZE-1);
1119 if ((offset & (blocksize - 1)) == 0)
1123 page = grab_cache_page(mapping, index);
1126 if (!PageUptodate(page)) {
1127 ret = btrfs_readpage(NULL, page);
1129 if (!PageUptodate(page)) {
1134 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
1136 ret = btrfs_cow_one_page(inode, page, offset);
1139 page_cache_release(page);
1144 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
1146 struct inode *inode = dentry->d_inode;
1149 err = inode_change_ok(inode, attr);
1153 if (S_ISREG(inode->i_mode) &&
1154 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
1155 struct btrfs_trans_handle *trans;
1156 struct btrfs_root *root = BTRFS_I(inode)->root;
1157 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1159 u64 mask = root->sectorsize - 1;
1160 u64 hole_start = (inode->i_size + mask) & ~mask;
1161 u64 block_end = (attr->ia_size + mask) & ~mask;
1165 if (attr->ia_size <= hole_start)
1168 mutex_lock(&root->fs_info->fs_mutex);
1169 err = btrfs_check_free_space(root, 1, 0);
1170 mutex_unlock(&root->fs_info->fs_mutex);
1174 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1176 lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1177 hole_size = block_end - hole_start;
1179 mutex_lock(&root->fs_info->fs_mutex);
1180 trans = btrfs_start_transaction(root, 1);
1181 btrfs_set_trans_block_group(trans, inode);
1182 err = btrfs_drop_extents(trans, root, inode,
1183 hole_start, block_end, hole_start,
1186 if (alloc_hint != EXTENT_MAP_INLINE) {
1187 err = btrfs_insert_file_extent(trans, root,
1191 btrfs_drop_extent_cache(inode, hole_start,
1193 btrfs_check_file(root, inode);
1195 btrfs_end_transaction(trans, root);
1196 mutex_unlock(&root->fs_info->fs_mutex);
1197 unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1202 err = inode_setattr(inode, attr);
1207 void btrfs_put_inode(struct inode *inode)
1211 if (!BTRFS_I(inode)->ordered_trans) {
1215 if (mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY) ||
1216 mapping_tagged(inode->i_mapping, PAGECACHE_TAG_WRITEBACK))
1219 ret = btrfs_del_ordered_inode(inode);
1221 atomic_dec(&inode->i_count);
1225 void btrfs_delete_inode(struct inode *inode)
1227 struct btrfs_trans_handle *trans;
1228 struct btrfs_root *root = BTRFS_I(inode)->root;
1232 truncate_inode_pages(&inode->i_data, 0);
1233 if (is_bad_inode(inode)) {
1238 mutex_lock(&root->fs_info->fs_mutex);
1239 trans = btrfs_start_transaction(root, 1);
1241 btrfs_set_trans_block_group(trans, inode);
1242 ret = btrfs_truncate_in_trans(trans, root, inode, 0);
1244 goto no_delete_lock;
1246 nr = trans->blocks_used;
1249 btrfs_end_transaction(trans, root);
1250 mutex_unlock(&root->fs_info->fs_mutex);
1251 btrfs_btree_balance_dirty(root, nr);
1252 btrfs_throttle(root);
1256 nr = trans->blocks_used;
1257 btrfs_end_transaction(trans, root);
1258 mutex_unlock(&root->fs_info->fs_mutex);
1259 btrfs_btree_balance_dirty(root, nr);
1260 btrfs_throttle(root);
1266 * this returns the key found in the dir entry in the location pointer.
1267 * If no dir entries were found, location->objectid is 0.
1269 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
1270 struct btrfs_key *location)
1272 const char *name = dentry->d_name.name;
1273 int namelen = dentry->d_name.len;
1274 struct btrfs_dir_item *di;
1275 struct btrfs_path *path;
1276 struct btrfs_root *root = BTRFS_I(dir)->root;
1279 if (namelen == 1 && strcmp(name, ".") == 0) {
1280 location->objectid = dir->i_ino;
1281 location->type = BTRFS_INODE_ITEM_KEY;
1282 location->offset = 0;
1285 path = btrfs_alloc_path();
1288 if (namelen == 2 && strcmp(name, "..") == 0) {
1289 struct btrfs_key key;
1290 struct extent_buffer *leaf;
1294 key.objectid = dir->i_ino;
1295 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1297 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1301 leaf = path->nodes[0];
1302 slot = path->slots[0];
1303 nritems = btrfs_header_nritems(leaf);
1304 if (slot >= nritems)
1307 btrfs_item_key_to_cpu(leaf, &key, slot);
1308 if (key.objectid != dir->i_ino ||
1309 key.type != BTRFS_INODE_REF_KEY) {
1312 location->objectid = key.offset;
1313 location->type = BTRFS_INODE_ITEM_KEY;
1314 location->offset = 0;
1318 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
1322 if (!di || IS_ERR(di)) {
1325 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
1327 btrfs_free_path(path);
1330 location->objectid = 0;
1335 * when we hit a tree root in a directory, the btrfs part of the inode
1336 * needs to be changed to reflect the root directory of the tree root. This
1337 * is kind of like crossing a mount point.
1339 static int fixup_tree_root_location(struct btrfs_root *root,
1340 struct btrfs_key *location,
1341 struct btrfs_root **sub_root,
1342 struct dentry *dentry)
1344 struct btrfs_path *path;
1345 struct btrfs_root_item *ri;
1347 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
1349 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1352 path = btrfs_alloc_path();
1354 mutex_lock(&root->fs_info->fs_mutex);
1356 *sub_root = btrfs_read_fs_root(root->fs_info, location,
1357 dentry->d_name.name,
1358 dentry->d_name.len);
1359 if (IS_ERR(*sub_root))
1360 return PTR_ERR(*sub_root);
1362 ri = &(*sub_root)->root_item;
1363 location->objectid = btrfs_root_dirid(ri);
1364 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1365 location->offset = 0;
1367 btrfs_free_path(path);
1368 mutex_unlock(&root->fs_info->fs_mutex);
1372 static int btrfs_init_locked_inode(struct inode *inode, void *p)
1374 struct btrfs_iget_args *args = p;
1375 inode->i_ino = args->ino;
1376 BTRFS_I(inode)->root = args->root;
1377 BTRFS_I(inode)->delalloc_bytes = 0;
1378 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1379 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1380 inode->i_mapping, GFP_NOFS);
1381 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1382 inode->i_mapping, GFP_NOFS);
1386 static int btrfs_find_actor(struct inode *inode, void *opaque)
1388 struct btrfs_iget_args *args = opaque;
1389 return (args->ino == inode->i_ino &&
1390 args->root == BTRFS_I(inode)->root);
1393 struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
1396 struct btrfs_iget_args args;
1397 args.ino = objectid;
1398 args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
1403 return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
1406 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
1407 struct btrfs_root *root)
1409 struct inode *inode;
1410 struct btrfs_iget_args args;
1411 args.ino = objectid;
1414 inode = iget5_locked(s, objectid, btrfs_find_actor,
1415 btrfs_init_locked_inode,
1420 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
1421 struct nameidata *nd)
1423 struct inode * inode;
1424 struct btrfs_inode *bi = BTRFS_I(dir);
1425 struct btrfs_root *root = bi->root;
1426 struct btrfs_root *sub_root = root;
1427 struct btrfs_key location;
1430 if (dentry->d_name.len > BTRFS_NAME_LEN)
1431 return ERR_PTR(-ENAMETOOLONG);
1433 mutex_lock(&root->fs_info->fs_mutex);
1434 ret = btrfs_inode_by_name(dir, dentry, &location);
1435 mutex_unlock(&root->fs_info->fs_mutex);
1438 return ERR_PTR(ret);
1441 if (location.objectid) {
1442 ret = fixup_tree_root_location(root, &location, &sub_root,
1445 return ERR_PTR(ret);
1447 return ERR_PTR(-ENOENT);
1448 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
1451 return ERR_PTR(-EACCES);
1452 if (inode->i_state & I_NEW) {
1453 /* the inode and parent dir are two different roots */
1454 if (sub_root != root) {
1456 sub_root->inode = inode;
1458 BTRFS_I(inode)->root = sub_root;
1459 memcpy(&BTRFS_I(inode)->location, &location,
1461 btrfs_read_locked_inode(inode);
1462 unlock_new_inode(inode);
1465 return d_splice_alias(inode, dentry);
1468 static unsigned char btrfs_filetype_table[] = {
1469 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1472 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1474 struct inode *inode = filp->f_dentry->d_inode;
1475 struct btrfs_root *root = BTRFS_I(inode)->root;
1476 struct btrfs_item *item;
1477 struct btrfs_dir_item *di;
1478 struct btrfs_key key;
1479 struct btrfs_key found_key;
1480 struct btrfs_path *path;
1483 struct extent_buffer *leaf;
1486 unsigned char d_type;
1491 int key_type = BTRFS_DIR_INDEX_KEY;
1496 /* FIXME, use a real flag for deciding about the key type */
1497 if (root->fs_info->tree_root == root)
1498 key_type = BTRFS_DIR_ITEM_KEY;
1500 /* special case for "." */
1501 if (filp->f_pos == 0) {
1502 over = filldir(dirent, ".", 1,
1510 mutex_lock(&root->fs_info->fs_mutex);
1511 key.objectid = inode->i_ino;
1512 path = btrfs_alloc_path();
1515 /* special case for .., just use the back ref */
1516 if (filp->f_pos == 1) {
1517 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1519 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1521 leaf = path->nodes[0];
1522 slot = path->slots[0];
1523 nritems = btrfs_header_nritems(leaf);
1524 if (slot >= nritems) {
1525 btrfs_release_path(root, path);
1526 goto read_dir_items;
1528 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1529 btrfs_release_path(root, path);
1530 if (found_key.objectid != key.objectid ||
1531 found_key.type != BTRFS_INODE_REF_KEY)
1532 goto read_dir_items;
1533 over = filldir(dirent, "..", 2,
1534 2, found_key.offset, DT_DIR);
1541 btrfs_set_key_type(&key, key_type);
1542 key.offset = filp->f_pos;
1544 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1549 leaf = path->nodes[0];
1550 nritems = btrfs_header_nritems(leaf);
1551 slot = path->slots[0];
1552 if (advance || slot >= nritems) {
1553 if (slot >= nritems -1) {
1554 ret = btrfs_next_leaf(root, path);
1557 leaf = path->nodes[0];
1558 nritems = btrfs_header_nritems(leaf);
1559 slot = path->slots[0];
1566 item = btrfs_item_nr(leaf, slot);
1567 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1569 if (found_key.objectid != key.objectid)
1571 if (btrfs_key_type(&found_key) != key_type)
1573 if (found_key.offset < filp->f_pos)
1576 filp->f_pos = found_key.offset;
1578 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1580 di_total = btrfs_item_size(leaf, item);
1581 while(di_cur < di_total) {
1582 struct btrfs_key location;
1584 name_len = btrfs_dir_name_len(leaf, di);
1585 if (name_len < 32) {
1586 name_ptr = tmp_name;
1588 name_ptr = kmalloc(name_len, GFP_NOFS);
1591 read_extent_buffer(leaf, name_ptr,
1592 (unsigned long)(di + 1), name_len);
1594 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1595 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1596 over = filldir(dirent, name_ptr, name_len,
1601 if (name_ptr != tmp_name)
1606 di_len = btrfs_dir_name_len(leaf, di) +
1607 btrfs_dir_data_len(leaf, di) +sizeof(*di);
1609 di = (struct btrfs_dir_item *)((char *)di + di_len);
1612 if (key_type == BTRFS_DIR_INDEX_KEY)
1613 filp->f_pos = INT_LIMIT(typeof(filp->f_pos));
1619 btrfs_release_path(root, path);
1620 btrfs_free_path(path);
1621 mutex_unlock(&root->fs_info->fs_mutex);
1625 int btrfs_write_inode(struct inode *inode, int wait)
1627 struct btrfs_root *root = BTRFS_I(inode)->root;
1628 struct btrfs_trans_handle *trans;
1632 mutex_lock(&root->fs_info->fs_mutex);
1633 trans = btrfs_start_transaction(root, 1);
1634 btrfs_set_trans_block_group(trans, inode);
1635 ret = btrfs_commit_transaction(trans, root);
1636 mutex_unlock(&root->fs_info->fs_mutex);
1642 * This is somewhat expensive, updating the tree every time the
1643 * inode changes. But, it is most likely to find the inode in cache.
1644 * FIXME, needs more benchmarking...there are no reasons other than performance
1645 * to keep or drop this code.
1647 void btrfs_dirty_inode(struct inode *inode)
1649 struct btrfs_root *root = BTRFS_I(inode)->root;
1650 struct btrfs_trans_handle *trans;
1652 mutex_lock(&root->fs_info->fs_mutex);
1653 trans = btrfs_start_transaction(root, 1);
1654 btrfs_set_trans_block_group(trans, inode);
1655 btrfs_update_inode(trans, root, inode);
1656 btrfs_end_transaction(trans, root);
1657 mutex_unlock(&root->fs_info->fs_mutex);
1660 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1661 struct btrfs_root *root,
1662 const char *name, int name_len,
1665 struct btrfs_block_group_cache *group,
1668 struct inode *inode;
1669 struct btrfs_inode_item *inode_item;
1670 struct btrfs_block_group_cache *new_inode_group;
1671 struct btrfs_key *location;
1672 struct btrfs_path *path;
1673 struct btrfs_inode_ref *ref;
1674 struct btrfs_key key[2];
1680 path = btrfs_alloc_path();
1683 inode = new_inode(root->fs_info->sb);
1685 return ERR_PTR(-ENOMEM);
1687 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1688 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1689 inode->i_mapping, GFP_NOFS);
1690 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1691 inode->i_mapping, GFP_NOFS);
1692 BTRFS_I(inode)->delalloc_bytes = 0;
1693 BTRFS_I(inode)->root = root;
1699 new_inode_group = btrfs_find_block_group(root, group, 0,
1700 BTRFS_BLOCK_GROUP_METADATA, owner);
1701 if (!new_inode_group) {
1702 printk("find_block group failed\n");
1703 new_inode_group = group;
1705 BTRFS_I(inode)->block_group = new_inode_group;
1706 BTRFS_I(inode)->flags = 0;
1708 key[0].objectid = objectid;
1709 btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
1712 key[1].objectid = objectid;
1713 btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
1714 key[1].offset = ref_objectid;
1716 sizes[0] = sizeof(struct btrfs_inode_item);
1717 sizes[1] = name_len + sizeof(*ref);
1719 ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
1723 if (objectid > root->highest_inode)
1724 root->highest_inode = objectid;
1726 inode->i_uid = current->fsuid;
1727 inode->i_gid = current->fsgid;
1728 inode->i_mode = mode;
1729 inode->i_ino = objectid;
1730 inode->i_blocks = 0;
1731 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1732 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1733 struct btrfs_inode_item);
1734 fill_inode_item(path->nodes[0], inode_item, inode);
1736 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
1737 struct btrfs_inode_ref);
1738 btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
1739 ptr = (unsigned long)(ref + 1);
1740 write_extent_buffer(path->nodes[0], name, ptr, name_len);
1742 btrfs_mark_buffer_dirty(path->nodes[0]);
1743 btrfs_free_path(path);
1745 location = &BTRFS_I(inode)->location;
1746 location->objectid = objectid;
1747 location->offset = 0;
1748 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1750 insert_inode_hash(inode);
1753 btrfs_free_path(path);
1754 return ERR_PTR(ret);
1757 static inline u8 btrfs_inode_type(struct inode *inode)
1759 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1762 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1763 struct dentry *dentry, struct inode *inode,
1767 struct btrfs_key key;
1768 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1769 struct inode *parent_inode;
1771 key.objectid = inode->i_ino;
1772 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1775 ret = btrfs_insert_dir_item(trans, root,
1776 dentry->d_name.name, dentry->d_name.len,
1777 dentry->d_parent->d_inode->i_ino,
1778 &key, btrfs_inode_type(inode));
1781 ret = btrfs_insert_inode_ref(trans, root,
1782 dentry->d_name.name,
1785 dentry->d_parent->d_inode->i_ino);
1787 parent_inode = dentry->d_parent->d_inode;
1788 parent_inode->i_size += dentry->d_name.len * 2;
1789 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1790 ret = btrfs_update_inode(trans, root,
1791 dentry->d_parent->d_inode);
1796 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1797 struct dentry *dentry, struct inode *inode,
1800 int err = btrfs_add_link(trans, dentry, inode, backref);
1802 d_instantiate(dentry, inode);
1810 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1811 int mode, dev_t rdev)
1813 struct btrfs_trans_handle *trans;
1814 struct btrfs_root *root = BTRFS_I(dir)->root;
1815 struct inode *inode = NULL;
1819 unsigned long nr = 0;
1821 if (!new_valid_dev(rdev))
1824 mutex_lock(&root->fs_info->fs_mutex);
1825 err = btrfs_check_free_space(root, 1, 0);
1829 trans = btrfs_start_transaction(root, 1);
1830 btrfs_set_trans_block_group(trans, dir);
1832 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1838 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1840 dentry->d_parent->d_inode->i_ino, objectid,
1841 BTRFS_I(dir)->block_group, mode);
1842 err = PTR_ERR(inode);
1846 btrfs_set_trans_block_group(trans, inode);
1847 err = btrfs_add_nondir(trans, dentry, inode, 0);
1851 inode->i_op = &btrfs_special_inode_operations;
1852 init_special_inode(inode, inode->i_mode, rdev);
1853 btrfs_update_inode(trans, root, inode);
1855 dir->i_sb->s_dirt = 1;
1856 btrfs_update_inode_block_group(trans, inode);
1857 btrfs_update_inode_block_group(trans, dir);
1859 nr = trans->blocks_used;
1860 btrfs_end_transaction(trans, root);
1862 mutex_unlock(&root->fs_info->fs_mutex);
1865 inode_dec_link_count(inode);
1868 btrfs_btree_balance_dirty(root, nr);
1869 btrfs_throttle(root);
1873 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1874 int mode, struct nameidata *nd)
1876 struct btrfs_trans_handle *trans;
1877 struct btrfs_root *root = BTRFS_I(dir)->root;
1878 struct inode *inode = NULL;
1881 unsigned long nr = 0;
1884 mutex_lock(&root->fs_info->fs_mutex);
1885 err = btrfs_check_free_space(root, 1, 0);
1888 trans = btrfs_start_transaction(root, 1);
1889 btrfs_set_trans_block_group(trans, dir);
1891 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1897 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1899 dentry->d_parent->d_inode->i_ino,
1900 objectid, BTRFS_I(dir)->block_group, mode);
1901 err = PTR_ERR(inode);
1905 btrfs_set_trans_block_group(trans, inode);
1906 err = btrfs_add_nondir(trans, dentry, inode, 0);
1910 inode->i_mapping->a_ops = &btrfs_aops;
1911 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
1912 inode->i_fop = &btrfs_file_operations;
1913 inode->i_op = &btrfs_file_inode_operations;
1914 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1915 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1916 inode->i_mapping, GFP_NOFS);
1917 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1918 inode->i_mapping, GFP_NOFS);
1919 BTRFS_I(inode)->delalloc_bytes = 0;
1920 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
1922 dir->i_sb->s_dirt = 1;
1923 btrfs_update_inode_block_group(trans, inode);
1924 btrfs_update_inode_block_group(trans, dir);
1926 nr = trans->blocks_used;
1927 btrfs_end_transaction(trans, root);
1929 mutex_unlock(&root->fs_info->fs_mutex);
1932 inode_dec_link_count(inode);
1935 btrfs_btree_balance_dirty(root, nr);
1936 btrfs_throttle(root);
1940 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1941 struct dentry *dentry)
1943 struct btrfs_trans_handle *trans;
1944 struct btrfs_root *root = BTRFS_I(dir)->root;
1945 struct inode *inode = old_dentry->d_inode;
1946 unsigned long nr = 0;
1950 if (inode->i_nlink == 0)
1953 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1958 mutex_lock(&root->fs_info->fs_mutex);
1959 err = btrfs_check_free_space(root, 1, 0);
1962 trans = btrfs_start_transaction(root, 1);
1964 btrfs_set_trans_block_group(trans, dir);
1965 atomic_inc(&inode->i_count);
1966 err = btrfs_add_nondir(trans, dentry, inode, 1);
1971 dir->i_sb->s_dirt = 1;
1972 btrfs_update_inode_block_group(trans, dir);
1973 err = btrfs_update_inode(trans, root, inode);
1978 nr = trans->blocks_used;
1979 btrfs_end_transaction(trans, root);
1981 mutex_unlock(&root->fs_info->fs_mutex);
1984 inode_dec_link_count(inode);
1987 btrfs_btree_balance_dirty(root, nr);
1988 btrfs_throttle(root);
1992 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1994 struct inode *inode;
1995 struct btrfs_trans_handle *trans;
1996 struct btrfs_root *root = BTRFS_I(dir)->root;
1998 int drop_on_err = 0;
2000 unsigned long nr = 1;
2002 mutex_lock(&root->fs_info->fs_mutex);
2003 err = btrfs_check_free_space(root, 1, 0);
2007 trans = btrfs_start_transaction(root, 1);
2008 btrfs_set_trans_block_group(trans, dir);
2010 if (IS_ERR(trans)) {
2011 err = PTR_ERR(trans);
2015 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2021 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2023 dentry->d_parent->d_inode->i_ino, objectid,
2024 BTRFS_I(dir)->block_group, S_IFDIR | mode);
2025 if (IS_ERR(inode)) {
2026 err = PTR_ERR(inode);
2031 inode->i_op = &btrfs_dir_inode_operations;
2032 inode->i_fop = &btrfs_dir_file_operations;
2033 btrfs_set_trans_block_group(trans, inode);
2036 err = btrfs_update_inode(trans, root, inode);
2040 err = btrfs_add_link(trans, dentry, inode, 0);
2044 d_instantiate(dentry, inode);
2046 dir->i_sb->s_dirt = 1;
2047 btrfs_update_inode_block_group(trans, inode);
2048 btrfs_update_inode_block_group(trans, dir);
2051 nr = trans->blocks_used;
2052 btrfs_end_transaction(trans, root);
2055 mutex_unlock(&root->fs_info->fs_mutex);
2058 btrfs_btree_balance_dirty(root, nr);
2059 btrfs_throttle(root);
2063 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
2064 size_t pg_offset, u64 start, u64 len,
2070 u64 extent_start = 0;
2072 u64 objectid = inode->i_ino;
2074 struct btrfs_path *path;
2075 struct btrfs_root *root = BTRFS_I(inode)->root;
2076 struct btrfs_file_extent_item *item;
2077 struct extent_buffer *leaf;
2078 struct btrfs_key found_key;
2079 struct extent_map *em = NULL;
2080 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2081 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2082 struct btrfs_trans_handle *trans = NULL;
2084 path = btrfs_alloc_path();
2086 mutex_lock(&root->fs_info->fs_mutex);
2089 spin_lock(&em_tree->lock);
2090 em = lookup_extent_mapping(em_tree, start, len);
2091 spin_unlock(&em_tree->lock);
2094 if (em->start > start) {
2095 printk("get_extent lookup [%Lu %Lu] em [%Lu %Lu]\n",
2096 start, len, em->start, em->len);
2099 if (em->block_start == EXTENT_MAP_INLINE && page)
2100 free_extent_map(em);
2104 em = alloc_extent_map(GFP_NOFS);
2110 em->start = EXTENT_MAP_HOLE;
2112 em->bdev = inode->i_sb->s_bdev;
2113 ret = btrfs_lookup_file_extent(trans, root, path,
2114 objectid, start, trans != NULL);
2121 if (path->slots[0] == 0)
2126 leaf = path->nodes[0];
2127 item = btrfs_item_ptr(leaf, path->slots[0],
2128 struct btrfs_file_extent_item);
2129 /* are we inside the extent that was found? */
2130 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2131 found_type = btrfs_key_type(&found_key);
2132 if (found_key.objectid != objectid ||
2133 found_type != BTRFS_EXTENT_DATA_KEY) {
2137 found_type = btrfs_file_extent_type(leaf, item);
2138 extent_start = found_key.offset;
2139 if (found_type == BTRFS_FILE_EXTENT_REG) {
2140 extent_end = extent_start +
2141 btrfs_file_extent_num_bytes(leaf, item);
2143 if (start < extent_start || start >= extent_end) {
2145 if (start < extent_start) {
2146 if (start + len <= extent_start)
2148 em->len = extent_end - extent_start;
2154 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
2156 em->start = extent_start;
2157 em->len = extent_end - extent_start;
2158 em->block_start = EXTENT_MAP_HOLE;
2161 bytenr += btrfs_file_extent_offset(leaf, item);
2162 em->block_start = bytenr;
2163 em->start = extent_start;
2164 em->len = extent_end - extent_start;
2166 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
2171 size_t extent_offset;
2174 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
2176 extent_end = (extent_start + size + root->sectorsize - 1) &
2177 ~((u64)root->sectorsize - 1);
2178 if (start < extent_start || start >= extent_end) {
2180 if (start < extent_start) {
2181 if (start + len <= extent_start)
2183 em->len = extent_end - extent_start;
2189 em->block_start = EXTENT_MAP_INLINE;
2192 em->start = extent_start;
2197 page_start = page_offset(page) + pg_offset;
2198 extent_offset = page_start - extent_start;
2199 copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
2200 size - extent_offset);
2201 em->start = extent_start + extent_offset;
2202 em->len = (copy_size + root->sectorsize - 1) &
2203 ~((u64)root->sectorsize - 1);
2205 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
2206 if (create == 0 && !PageUptodate(page)) {
2207 read_extent_buffer(leaf, map + pg_offset, ptr,
2209 flush_dcache_page(page);
2210 } else if (create && PageUptodate(page)) {
2213 free_extent_map(em);
2215 btrfs_release_path(root, path);
2216 trans = btrfs_start_transaction(root, 1);
2219 write_extent_buffer(leaf, map + pg_offset, ptr,
2221 btrfs_mark_buffer_dirty(leaf);
2224 set_extent_uptodate(io_tree, em->start,
2225 extent_map_end(em) - 1, GFP_NOFS);
2228 printk("unkknown found_type %d\n", found_type);
2235 em->block_start = EXTENT_MAP_HOLE;
2237 btrfs_release_path(root, path);
2238 if (em->start > start || extent_map_end(em) <= start) {
2239 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len);
2245 spin_lock(&em_tree->lock);
2246 ret = add_extent_mapping(em_tree, em);
2247 if (ret == -EEXIST) {
2248 free_extent_map(em);
2249 em = lookup_extent_mapping(em_tree, start, len);
2252 printk("failing to insert %Lu %Lu\n", start, len);
2255 spin_unlock(&em_tree->lock);
2257 btrfs_free_path(path);
2259 ret = btrfs_end_transaction(trans, root);
2263 mutex_unlock(&root->fs_info->fs_mutex);
2265 free_extent_map(em);
2267 return ERR_PTR(err);
2272 static int btrfs_get_block(struct inode *inode, sector_t iblock,
2273 struct buffer_head *bh_result, int create)
2275 struct extent_map *em;
2276 u64 start = (u64)iblock << inode->i_blkbits;
2277 struct btrfs_multi_bio *multi = NULL;
2278 struct btrfs_root *root = BTRFS_I(inode)->root;
2284 em = btrfs_get_extent(inode, NULL, 0, start, bh_result->b_size, 0);
2286 if (!em || IS_ERR(em))
2289 if (em->start > start || em->start + em->len <= start)
2292 if (em->block_start == EXTENT_MAP_INLINE) {
2297 if (em->block_start == EXTENT_MAP_HOLE ||
2298 em->block_start == EXTENT_MAP_DELALLOC) {
2302 len = em->start + em->len - start;
2303 len = min_t(u64, len, INT_LIMIT(typeof(bh_result->b_size)));
2305 logical = start - em->start;
2306 logical = em->block_start + logical;
2309 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
2310 logical, &map_length, &multi, 0);
2312 bh_result->b_blocknr = multi->stripes[0].physical >> inode->i_blkbits;
2313 bh_result->b_size = min(map_length, len);
2314 bh_result->b_bdev = multi->stripes[0].dev->bdev;
2315 set_buffer_mapped(bh_result);
2318 free_extent_map(em);
2322 static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
2323 const struct iovec *iov, loff_t offset,
2324 unsigned long nr_segs)
2326 struct file *file = iocb->ki_filp;
2327 struct inode *inode = file->f_mapping->host;
2332 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2333 offset, nr_segs, btrfs_get_block, NULL);
2336 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
2338 return extent_bmap(mapping, iblock, btrfs_get_extent);
2341 int btrfs_readpage(struct file *file, struct page *page)
2343 struct extent_io_tree *tree;
2344 tree = &BTRFS_I(page->mapping->host)->io_tree;
2345 return extent_read_full_page(tree, page, btrfs_get_extent);
2348 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
2350 struct extent_io_tree *tree;
2353 if (current->flags & PF_MEMALLOC) {
2354 redirty_page_for_writepage(wbc, page);
2358 tree = &BTRFS_I(page->mapping->host)->io_tree;
2359 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
2362 static int btrfs_writepages(struct address_space *mapping,
2363 struct writeback_control *wbc)
2365 struct extent_io_tree *tree;
2366 tree = &BTRFS_I(mapping->host)->io_tree;
2367 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
2371 btrfs_readpages(struct file *file, struct address_space *mapping,
2372 struct list_head *pages, unsigned nr_pages)
2374 struct extent_io_tree *tree;
2375 tree = &BTRFS_I(mapping->host)->io_tree;
2376 return extent_readpages(tree, mapping, pages, nr_pages,
2380 static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
2382 struct extent_io_tree *tree;
2383 struct extent_map_tree *map;
2386 tree = &BTRFS_I(page->mapping->host)->io_tree;
2387 map = &BTRFS_I(page->mapping->host)->extent_tree;
2388 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
2390 ClearPagePrivate(page);
2391 set_page_private(page, 0);
2392 page_cache_release(page);
2397 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
2399 struct extent_io_tree *tree;
2401 tree = &BTRFS_I(page->mapping->host)->io_tree;
2402 extent_invalidatepage(tree, page, offset);
2403 btrfs_releasepage(page, GFP_NOFS);
2407 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2408 * called from a page fault handler when a page is first dirtied. Hence we must
2409 * be careful to check for EOF conditions here. We set the page up correctly
2410 * for a written page which means we get ENOSPC checking when writing into
2411 * holes and correct delalloc and unwritten extent mapping on filesystems that
2412 * support these features.
2414 * We are not allowed to take the i_mutex here so we have to play games to
2415 * protect against truncate races as the page could now be beyond EOF. Because
2416 * vmtruncate() writes the inode size before removing pages, once we have the
2417 * page lock we can determine safely if the page is beyond EOF. If it is not
2418 * beyond EOF, then the page is guaranteed safe against truncation until we
2421 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
2423 struct inode *inode = fdentry(vma->vm_file)->d_inode;
2424 struct btrfs_root *root = BTRFS_I(inode)->root;
2430 mutex_lock(&root->fs_info->fs_mutex);
2431 ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
2432 mutex_unlock(&root->fs_info->fs_mutex);
2439 wait_on_page_writeback(page);
2440 size = i_size_read(inode);
2441 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2443 if ((page->mapping != inode->i_mapping) ||
2444 (page_start > size)) {
2445 /* page got truncated out from underneath us */
2449 /* page is wholly or partially inside EOF */
2450 if (page_start + PAGE_CACHE_SIZE > size)
2451 end = size & ~PAGE_CACHE_MASK;
2453 end = PAGE_CACHE_SIZE;
2455 ret = btrfs_cow_one_page(inode, page, end);
2463 static void btrfs_truncate(struct inode *inode)
2465 struct btrfs_root *root = BTRFS_I(inode)->root;
2467 struct btrfs_trans_handle *trans;
2470 if (!S_ISREG(inode->i_mode))
2472 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
2475 btrfs_truncate_page(inode->i_mapping, inode->i_size);
2477 mutex_lock(&root->fs_info->fs_mutex);
2478 trans = btrfs_start_transaction(root, 1);
2479 btrfs_set_trans_block_group(trans, inode);
2481 /* FIXME, add redo link to tree so we don't leak on crash */
2482 ret = btrfs_truncate_in_trans(trans, root, inode,
2483 BTRFS_EXTENT_DATA_KEY);
2484 btrfs_update_inode(trans, root, inode);
2485 nr = trans->blocks_used;
2487 ret = btrfs_end_transaction(trans, root);
2489 mutex_unlock(&root->fs_info->fs_mutex);
2490 btrfs_btree_balance_dirty(root, nr);
2491 btrfs_throttle(root);
2494 static int noinline create_subvol(struct btrfs_root *root, char *name,
2497 struct btrfs_trans_handle *trans;
2498 struct btrfs_key key;
2499 struct btrfs_root_item root_item;
2500 struct btrfs_inode_item *inode_item;
2501 struct extent_buffer *leaf;
2502 struct btrfs_root *new_root = root;
2503 struct inode *inode;
2508 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
2509 unsigned long nr = 1;
2511 mutex_lock(&root->fs_info->fs_mutex);
2512 ret = btrfs_check_free_space(root, 1, 0);
2516 trans = btrfs_start_transaction(root, 1);
2519 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
2524 leaf = __btrfs_alloc_free_block(trans, root, root->leafsize,
2525 objectid, trans->transid, 0, 0,
2528 return PTR_ERR(leaf);
2530 btrfs_set_header_nritems(leaf, 0);
2531 btrfs_set_header_level(leaf, 0);
2532 btrfs_set_header_bytenr(leaf, leaf->start);
2533 btrfs_set_header_generation(leaf, trans->transid);
2534 btrfs_set_header_owner(leaf, objectid);
2536 write_extent_buffer(leaf, root->fs_info->fsid,
2537 (unsigned long)btrfs_header_fsid(leaf),
2539 btrfs_mark_buffer_dirty(leaf);
2541 inode_item = &root_item.inode;
2542 memset(inode_item, 0, sizeof(*inode_item));
2543 inode_item->generation = cpu_to_le64(1);
2544 inode_item->size = cpu_to_le64(3);
2545 inode_item->nlink = cpu_to_le32(1);
2546 inode_item->nblocks = cpu_to_le64(1);
2547 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
2549 btrfs_set_root_bytenr(&root_item, leaf->start);
2550 btrfs_set_root_level(&root_item, 0);
2551 btrfs_set_root_refs(&root_item, 1);
2552 btrfs_set_root_used(&root_item, 0);
2554 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
2555 root_item.drop_level = 0;
2557 free_extent_buffer(leaf);
2560 btrfs_set_root_dirid(&root_item, new_dirid);
2562 key.objectid = objectid;
2564 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
2565 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
2571 * insert the directory item
2573 key.offset = (u64)-1;
2574 dir = root->fs_info->sb->s_root->d_inode;
2575 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
2576 name, namelen, dir->i_ino, &key,
2581 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
2582 name, namelen, objectid,
2583 root->fs_info->sb->s_root->d_inode->i_ino);
2587 ret = btrfs_commit_transaction(trans, root);
2591 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
2594 trans = btrfs_start_transaction(new_root, 1);
2597 inode = btrfs_new_inode(trans, new_root, "..", 2, new_dirid,
2599 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
2602 inode->i_op = &btrfs_dir_inode_operations;
2603 inode->i_fop = &btrfs_dir_file_operations;
2604 new_root->inode = inode;
2606 ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid,
2610 ret = btrfs_update_inode(trans, new_root, inode);
2614 nr = trans->blocks_used;
2615 err = btrfs_commit_transaction(trans, new_root);
2619 mutex_unlock(&root->fs_info->fs_mutex);
2620 btrfs_btree_balance_dirty(root, nr);
2621 btrfs_throttle(root);
2625 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
2627 struct btrfs_pending_snapshot *pending_snapshot;
2628 struct btrfs_trans_handle *trans;
2631 unsigned long nr = 0;
2633 if (!root->ref_cows)
2636 mutex_lock(&root->fs_info->fs_mutex);
2637 ret = btrfs_check_free_space(root, 1, 0);
2641 pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS);
2642 if (!pending_snapshot) {
2646 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
2647 if (!pending_snapshot->name) {
2649 kfree(pending_snapshot);
2652 memcpy(pending_snapshot->name, name, namelen);
2653 pending_snapshot->name[namelen] = '\0';
2654 trans = btrfs_start_transaction(root, 1);
2656 pending_snapshot->root = root;
2657 list_add(&pending_snapshot->list,
2658 &trans->transaction->pending_snapshots);
2659 ret = btrfs_update_inode(trans, root, root->inode);
2660 err = btrfs_commit_transaction(trans, root);
2663 mutex_unlock(&root->fs_info->fs_mutex);
2664 btrfs_btree_balance_dirty(root, nr);
2665 btrfs_throttle(root);
2669 unsigned long btrfs_force_ra(struct address_space *mapping,
2670 struct file_ra_state *ra, struct file *file,
2671 pgoff_t offset, pgoff_t last_index)
2675 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2676 req_size = last_index - offset + 1;
2677 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2680 req_size = min(last_index - offset + 1, (pgoff_t)128);
2681 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2682 return offset + req_size;
2686 int btrfs_defrag_file(struct file *file) {
2687 struct inode *inode = fdentry(file)->d_inode;
2688 struct btrfs_root *root = BTRFS_I(inode)->root;
2689 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2691 unsigned long last_index;
2692 unsigned long ra_index = 0;
2698 mutex_lock(&root->fs_info->fs_mutex);
2699 ret = btrfs_check_free_space(root, inode->i_size, 0);
2700 mutex_unlock(&root->fs_info->fs_mutex);
2704 mutex_lock(&inode->i_mutex);
2705 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
2706 for (i = 0; i <= last_index; i++) {
2707 if (i == ra_index) {
2708 ra_index = btrfs_force_ra(inode->i_mapping,
2710 file, ra_index, last_index);
2712 page = grab_cache_page(inode->i_mapping, i);
2715 if (!PageUptodate(page)) {
2716 btrfs_readpage(NULL, page);
2718 if (!PageUptodate(page)) {
2720 page_cache_release(page);
2724 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2725 page_end = page_start + PAGE_CACHE_SIZE - 1;
2727 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
2728 set_extent_delalloc(io_tree, page_start,
2729 page_end, GFP_NOFS);
2731 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
2732 set_page_dirty(page);
2734 page_cache_release(page);
2735 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
2739 mutex_unlock(&inode->i_mutex);
2743 static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
2747 struct btrfs_ioctl_vol_args *vol_args;
2748 struct btrfs_trans_handle *trans;
2754 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
2759 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
2763 namelen = strlen(vol_args->name);
2764 if (namelen > BTRFS_VOL_NAME_MAX) {
2769 sizestr = vol_args->name;
2770 if (!strcmp(sizestr, "max"))
2771 new_size = root->fs_info->sb->s_bdev->bd_inode->i_size;
2773 if (sizestr[0] == '-') {
2776 } else if (sizestr[0] == '+') {
2780 new_size = btrfs_parse_size(sizestr);
2781 if (new_size == 0) {
2787 mutex_lock(&root->fs_info->fs_mutex);
2788 old_size = btrfs_super_total_bytes(&root->fs_info->super_copy);
2791 if (new_size > old_size) {
2795 new_size = old_size - new_size;
2796 } else if (mod > 0) {
2797 new_size = old_size + new_size;
2800 if (new_size < 256 * 1024 * 1024) {
2804 if (new_size > root->fs_info->sb->s_bdev->bd_inode->i_size) {
2809 do_div(new_size, root->sectorsize);
2810 new_size *= root->sectorsize;
2812 printk("new size is %Lu\n", new_size);
2813 if (new_size > old_size) {
2814 trans = btrfs_start_transaction(root, 1);
2815 ret = btrfs_grow_extent_tree(trans, root, new_size);
2816 btrfs_commit_transaction(trans, root);
2818 ret = btrfs_shrink_extent_tree(root, new_size);
2822 mutex_unlock(&root->fs_info->fs_mutex);
2828 static int noinline btrfs_ioctl_snap_create(struct btrfs_root *root,
2831 struct btrfs_ioctl_vol_args *vol_args;
2832 struct btrfs_dir_item *di;
2833 struct btrfs_path *path;
2838 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
2843 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
2848 namelen = strlen(vol_args->name);
2849 if (namelen > BTRFS_VOL_NAME_MAX) {
2853 if (strchr(vol_args->name, '/')) {
2858 path = btrfs_alloc_path();
2864 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2865 mutex_lock(&root->fs_info->fs_mutex);
2866 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2868 vol_args->name, namelen, 0);
2869 mutex_unlock(&root->fs_info->fs_mutex);
2870 btrfs_free_path(path);
2872 if (di && !IS_ERR(di)) {
2882 if (root == root->fs_info->tree_root)
2883 ret = create_subvol(root, vol_args->name, namelen);
2885 ret = create_snapshot(root, vol_args->name, namelen);
2891 static int btrfs_ioctl_defrag(struct file *file)
2893 struct inode *inode = fdentry(file)->d_inode;
2894 struct btrfs_root *root = BTRFS_I(inode)->root;
2896 switch (inode->i_mode & S_IFMT) {
2898 mutex_lock(&root->fs_info->fs_mutex);
2899 btrfs_defrag_root(root, 0);
2900 btrfs_defrag_root(root->fs_info->extent_root, 0);
2901 mutex_unlock(&root->fs_info->fs_mutex);
2904 btrfs_defrag_file(file);
2911 long btrfs_ioctl(struct file *file, unsigned int
2912 cmd, unsigned long arg)
2914 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2917 case BTRFS_IOC_SNAP_CREATE:
2918 return btrfs_ioctl_snap_create(root, (void __user *)arg);
2919 case BTRFS_IOC_DEFRAG:
2920 return btrfs_ioctl_defrag(file);
2921 case BTRFS_IOC_RESIZE:
2922 return btrfs_ioctl_resize(root, (void __user *)arg);
2929 * Called inside transaction, so use GFP_NOFS
2931 struct inode *btrfs_alloc_inode(struct super_block *sb)
2933 struct btrfs_inode *ei;
2935 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2939 ei->ordered_trans = 0;
2940 return &ei->vfs_inode;
2943 void btrfs_destroy_inode(struct inode *inode)
2945 WARN_ON(!list_empty(&inode->i_dentry));
2946 WARN_ON(inode->i_data.nrpages);
2948 btrfs_drop_extent_cache(inode, 0, (u64)-1);
2949 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2952 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2953 static void init_once(struct kmem_cache * cachep, void *foo)
2955 static void init_once(void * foo, struct kmem_cache * cachep,
2956 unsigned long flags)
2959 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2961 inode_init_once(&ei->vfs_inode);
2964 void btrfs_destroy_cachep(void)
2966 if (btrfs_inode_cachep)
2967 kmem_cache_destroy(btrfs_inode_cachep);
2968 if (btrfs_trans_handle_cachep)
2969 kmem_cache_destroy(btrfs_trans_handle_cachep);
2970 if (btrfs_transaction_cachep)
2971 kmem_cache_destroy(btrfs_transaction_cachep);
2972 if (btrfs_bit_radix_cachep)
2973 kmem_cache_destroy(btrfs_bit_radix_cachep);
2974 if (btrfs_path_cachep)
2975 kmem_cache_destroy(btrfs_path_cachep);
2978 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2979 unsigned long extra_flags,
2980 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2981 void (*ctor)(struct kmem_cache *, void *)
2983 void (*ctor)(void *, struct kmem_cache *,
2988 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2989 SLAB_MEM_SPREAD | extra_flags), ctor
2990 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2996 int btrfs_init_cachep(void)
2998 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2999 sizeof(struct btrfs_inode),
3001 if (!btrfs_inode_cachep)
3003 btrfs_trans_handle_cachep =
3004 btrfs_cache_create("btrfs_trans_handle_cache",
3005 sizeof(struct btrfs_trans_handle),
3007 if (!btrfs_trans_handle_cachep)
3009 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
3010 sizeof(struct btrfs_transaction),
3012 if (!btrfs_transaction_cachep)
3014 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
3015 sizeof(struct btrfs_path),
3017 if (!btrfs_path_cachep)
3019 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
3020 SLAB_DESTROY_BY_RCU, NULL);
3021 if (!btrfs_bit_radix_cachep)
3025 btrfs_destroy_cachep();
3029 static int btrfs_getattr(struct vfsmount *mnt,
3030 struct dentry *dentry, struct kstat *stat)
3032 struct inode *inode = dentry->d_inode;
3033 generic_fillattr(inode, stat);
3034 stat->blksize = PAGE_CACHE_SIZE;
3035 stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9);
3039 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
3040 struct inode * new_dir,struct dentry *new_dentry)
3042 struct btrfs_trans_handle *trans;
3043 struct btrfs_root *root = BTRFS_I(old_dir)->root;
3044 struct inode *new_inode = new_dentry->d_inode;
3045 struct inode *old_inode = old_dentry->d_inode;
3046 struct timespec ctime = CURRENT_TIME;
3047 struct btrfs_path *path;
3050 if (S_ISDIR(old_inode->i_mode) && new_inode &&
3051 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
3055 mutex_lock(&root->fs_info->fs_mutex);
3056 ret = btrfs_check_free_space(root, 1, 0);
3060 trans = btrfs_start_transaction(root, 1);
3062 btrfs_set_trans_block_group(trans, new_dir);
3063 path = btrfs_alloc_path();
3069 old_dentry->d_inode->i_nlink++;
3070 old_dir->i_ctime = old_dir->i_mtime = ctime;
3071 new_dir->i_ctime = new_dir->i_mtime = ctime;
3072 old_inode->i_ctime = ctime;
3074 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
3079 new_inode->i_ctime = CURRENT_TIME;
3080 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
3084 ret = btrfs_add_link(trans, new_dentry, old_inode, 1);
3089 btrfs_free_path(path);
3090 btrfs_end_transaction(trans, root);
3092 mutex_unlock(&root->fs_info->fs_mutex);
3096 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
3097 const char *symname)
3099 struct btrfs_trans_handle *trans;
3100 struct btrfs_root *root = BTRFS_I(dir)->root;
3101 struct btrfs_path *path;
3102 struct btrfs_key key;
3103 struct inode *inode = NULL;
3110 struct btrfs_file_extent_item *ei;
3111 struct extent_buffer *leaf;
3112 unsigned long nr = 0;
3114 name_len = strlen(symname) + 1;
3115 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
3116 return -ENAMETOOLONG;
3118 mutex_lock(&root->fs_info->fs_mutex);
3119 err = btrfs_check_free_space(root, 1, 0);
3123 trans = btrfs_start_transaction(root, 1);
3124 btrfs_set_trans_block_group(trans, dir);
3126 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
3132 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
3134 dentry->d_parent->d_inode->i_ino, objectid,
3135 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
3136 err = PTR_ERR(inode);
3140 btrfs_set_trans_block_group(trans, inode);
3141 err = btrfs_add_nondir(trans, dentry, inode, 0);
3145 inode->i_mapping->a_ops = &btrfs_aops;
3146 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3147 inode->i_fop = &btrfs_file_operations;
3148 inode->i_op = &btrfs_file_inode_operations;
3149 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
3150 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
3151 inode->i_mapping, GFP_NOFS);
3152 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
3153 inode->i_mapping, GFP_NOFS);
3154 BTRFS_I(inode)->delalloc_bytes = 0;
3155 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
3157 dir->i_sb->s_dirt = 1;
3158 btrfs_update_inode_block_group(trans, inode);
3159 btrfs_update_inode_block_group(trans, dir);
3163 path = btrfs_alloc_path();
3165 key.objectid = inode->i_ino;
3167 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
3168 datasize = btrfs_file_extent_calc_inline_size(name_len);
3169 err = btrfs_insert_empty_item(trans, root, path, &key,
3175 leaf = path->nodes[0];
3176 ei = btrfs_item_ptr(leaf, path->slots[0],
3177 struct btrfs_file_extent_item);
3178 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
3179 btrfs_set_file_extent_type(leaf, ei,
3180 BTRFS_FILE_EXTENT_INLINE);
3181 ptr = btrfs_file_extent_inline_start(ei);
3182 write_extent_buffer(leaf, symname, ptr, name_len);
3183 btrfs_mark_buffer_dirty(leaf);
3184 btrfs_free_path(path);
3186 inode->i_op = &btrfs_symlink_inode_operations;
3187 inode->i_mapping->a_ops = &btrfs_symlink_aops;
3188 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3189 inode->i_size = name_len - 1;
3190 err = btrfs_update_inode(trans, root, inode);
3195 nr = trans->blocks_used;
3196 btrfs_end_transaction(trans, root);
3198 mutex_unlock(&root->fs_info->fs_mutex);
3200 inode_dec_link_count(inode);
3203 btrfs_btree_balance_dirty(root, nr);
3204 btrfs_throttle(root);
3208 static int btrfs_permission(struct inode *inode, int mask,
3209 struct nameidata *nd)
3211 if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
3213 return generic_permission(inode, mask, NULL);
3216 static struct inode_operations btrfs_dir_inode_operations = {
3217 .lookup = btrfs_lookup,
3218 .create = btrfs_create,
3219 .unlink = btrfs_unlink,
3221 .mkdir = btrfs_mkdir,
3222 .rmdir = btrfs_rmdir,
3223 .rename = btrfs_rename,
3224 .symlink = btrfs_symlink,
3225 .setattr = btrfs_setattr,
3226 .mknod = btrfs_mknod,
3227 .setxattr = generic_setxattr,
3228 .getxattr = generic_getxattr,
3229 .listxattr = btrfs_listxattr,
3230 .removexattr = generic_removexattr,
3231 .permission = btrfs_permission,
3233 static struct inode_operations btrfs_dir_ro_inode_operations = {
3234 .lookup = btrfs_lookup,
3235 .permission = btrfs_permission,
3237 static struct file_operations btrfs_dir_file_operations = {
3238 .llseek = generic_file_llseek,
3239 .read = generic_read_dir,
3240 .readdir = btrfs_readdir,
3241 .unlocked_ioctl = btrfs_ioctl,
3242 #ifdef CONFIG_COMPAT
3243 .compat_ioctl = btrfs_ioctl,
3247 static struct extent_io_ops btrfs_extent_io_ops = {
3248 .fill_delalloc = run_delalloc_range,
3249 .submit_bio_hook = btrfs_submit_bio_hook,
3250 .merge_bio_hook = btrfs_merge_bio_hook,
3251 .readpage_io_hook = btrfs_readpage_io_hook,
3252 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
3253 .readpage_io_failed_hook = btrfs_readpage_io_failed_hook,
3254 .set_bit_hook = btrfs_set_bit_hook,
3255 .clear_bit_hook = btrfs_clear_bit_hook,
3258 static struct address_space_operations btrfs_aops = {
3259 .readpage = btrfs_readpage,
3260 .writepage = btrfs_writepage,
3261 .writepages = btrfs_writepages,
3262 .readpages = btrfs_readpages,
3263 .sync_page = block_sync_page,
3265 .direct_IO = btrfs_direct_IO,
3266 .invalidatepage = btrfs_invalidatepage,
3267 .releasepage = btrfs_releasepage,
3268 .set_page_dirty = __set_page_dirty_nobuffers,
3271 static struct address_space_operations btrfs_symlink_aops = {
3272 .readpage = btrfs_readpage,
3273 .writepage = btrfs_writepage,
3274 .invalidatepage = btrfs_invalidatepage,
3275 .releasepage = btrfs_releasepage,
3278 static struct inode_operations btrfs_file_inode_operations = {
3279 .truncate = btrfs_truncate,
3280 .getattr = btrfs_getattr,
3281 .setattr = btrfs_setattr,
3282 .setxattr = generic_setxattr,
3283 .getxattr = generic_getxattr,
3284 .listxattr = btrfs_listxattr,
3285 .removexattr = generic_removexattr,
3286 .permission = btrfs_permission,
3288 static struct inode_operations btrfs_special_inode_operations = {
3289 .getattr = btrfs_getattr,
3290 .setattr = btrfs_setattr,
3291 .permission = btrfs_permission,
3293 static struct inode_operations btrfs_symlink_inode_operations = {
3294 .readlink = generic_readlink,
3295 .follow_link = page_follow_link_light,
3296 .put_link = page_put_link,
3297 .permission = btrfs_permission,