2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/pagemap.h>
25 #include <linux/highmem.h>
26 #include <linux/time.h>
27 #include <linux/init.h>
28 #include <linux/string.h>
29 #include <linux/smp_lock.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/bit_spinlock.h>
37 #include <linux/version.h>
38 #include <linux/xattr.h>
41 #include "transaction.h"
42 #include "btrfs_inode.h"
44 #include "print-tree.h"
46 #include "ordered-data.h"
48 struct btrfs_iget_args {
50 struct btrfs_root *root;
53 static struct inode_operations btrfs_dir_inode_operations;
54 static struct inode_operations btrfs_symlink_inode_operations;
55 static struct inode_operations btrfs_dir_ro_inode_operations;
56 static struct inode_operations btrfs_special_inode_operations;
57 static struct inode_operations btrfs_file_inode_operations;
58 static struct address_space_operations btrfs_aops;
59 static struct address_space_operations btrfs_symlink_aops;
60 static struct file_operations btrfs_dir_file_operations;
61 static struct extent_io_ops btrfs_extent_io_ops;
63 static struct kmem_cache *btrfs_inode_cachep;
64 struct kmem_cache *btrfs_trans_handle_cachep;
65 struct kmem_cache *btrfs_transaction_cachep;
66 struct kmem_cache *btrfs_bit_radix_cachep;
67 struct kmem_cache *btrfs_path_cachep;
70 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
71 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
72 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
73 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
74 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
75 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
76 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
77 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
80 int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
89 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
90 total = btrfs_super_total_bytes(&root->fs_info->super_copy);
91 used = btrfs_super_bytes_used(&root->fs_info->super_copy);
99 if (used + root->fs_info->delalloc_bytes + num_required > thresh)
101 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
105 static int cow_file_range(struct inode *inode, u64 start, u64 end)
107 struct btrfs_root *root = BTRFS_I(inode)->root;
108 struct btrfs_trans_handle *trans;
112 u64 blocksize = root->sectorsize;
114 struct btrfs_key ins;
115 struct extent_map *em;
116 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
119 trans = btrfs_join_transaction(root, 1);
121 btrfs_set_trans_block_group(trans, inode);
123 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
124 num_bytes = max(blocksize, num_bytes);
125 orig_num_bytes = num_bytes;
127 if (alloc_hint == EXTENT_MAP_INLINE)
130 BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy));
131 mutex_lock(&BTRFS_I(inode)->extent_mutex);
132 btrfs_drop_extent_cache(inode, start, start + num_bytes - 1);
133 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
135 while(num_bytes > 0) {
136 cur_alloc_size = min(num_bytes, root->fs_info->max_extent);
137 ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
138 root->sectorsize, 0, 0,
144 em = alloc_extent_map(GFP_NOFS);
146 em->len = ins.offset;
147 em->block_start = ins.objectid;
148 em->bdev = root->fs_info->fs_devices->latest_bdev;
149 mutex_lock(&BTRFS_I(inode)->extent_mutex);
150 set_bit(EXTENT_FLAG_PINNED, &em->flags);
152 spin_lock(&em_tree->lock);
153 ret = add_extent_mapping(em_tree, em);
154 spin_unlock(&em_tree->lock);
155 if (ret != -EEXIST) {
159 btrfs_drop_extent_cache(inode, start,
160 start + ins.offset - 1);
162 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
164 cur_alloc_size = ins.offset;
165 ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
168 if (num_bytes < cur_alloc_size) {
169 printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes,
173 num_bytes -= cur_alloc_size;
174 alloc_hint = ins.objectid + ins.offset;
175 start += cur_alloc_size;
178 btrfs_end_transaction(trans, root);
182 static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end)
190 struct btrfs_root *root = BTRFS_I(inode)->root;
191 struct btrfs_block_group_cache *block_group;
192 struct extent_buffer *leaf;
194 struct btrfs_path *path;
195 struct btrfs_file_extent_item *item;
198 struct btrfs_key found_key;
200 total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
201 path = btrfs_alloc_path();
204 ret = btrfs_lookup_file_extent(NULL, root, path,
205 inode->i_ino, start, 0);
207 btrfs_free_path(path);
213 if (path->slots[0] == 0)
218 leaf = path->nodes[0];
219 item = btrfs_item_ptr(leaf, path->slots[0],
220 struct btrfs_file_extent_item);
222 /* are we inside the extent that was found? */
223 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
224 found_type = btrfs_key_type(&found_key);
225 if (found_key.objectid != inode->i_ino ||
226 found_type != BTRFS_EXTENT_DATA_KEY)
229 found_type = btrfs_file_extent_type(leaf, item);
230 extent_start = found_key.offset;
231 if (found_type == BTRFS_FILE_EXTENT_REG) {
232 u64 extent_num_bytes;
234 extent_num_bytes = btrfs_file_extent_num_bytes(leaf, item);
235 extent_end = extent_start + extent_num_bytes;
238 if (loops && start != extent_start)
241 if (start < extent_start || start >= extent_end)
244 cow_end = min(end, extent_end - 1);
245 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
249 if (btrfs_count_snapshots_in_path(root, path, inode->i_ino,
255 * we may be called by the resizer, make sure we're inside
256 * the limits of the FS
258 block_group = btrfs_lookup_block_group(root->fs_info,
260 if (!block_group || block_group->ro)
269 btrfs_free_path(path);
272 btrfs_release_path(root, path);
277 cow_file_range(inode, start, end);
282 static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
284 struct btrfs_root *root = BTRFS_I(inode)->root;
287 if (btrfs_test_opt(root, NODATACOW) ||
288 btrfs_test_flag(inode, NODATACOW))
289 ret = run_delalloc_nocow(inode, start, end);
291 ret = cow_file_range(inode, start, end);
296 int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
297 unsigned long old, unsigned long bits)
300 if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
301 struct btrfs_root *root = BTRFS_I(inode)->root;
302 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
303 BTRFS_I(inode)->delalloc_bytes += end - start + 1;
304 root->fs_info->delalloc_bytes += end - start + 1;
305 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
310 int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end,
311 unsigned long old, unsigned long bits)
313 if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
314 struct btrfs_root *root = BTRFS_I(inode)->root;
317 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
318 if (end - start + 1 > root->fs_info->delalloc_bytes) {
319 printk("warning: delalloc account %Lu %Lu\n",
320 end - start + 1, root->fs_info->delalloc_bytes);
321 root->fs_info->delalloc_bytes = 0;
322 BTRFS_I(inode)->delalloc_bytes = 0;
324 root->fs_info->delalloc_bytes -= end - start + 1;
325 BTRFS_I(inode)->delalloc_bytes -= end - start + 1;
327 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
332 int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
333 size_t size, struct bio *bio)
335 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
336 struct btrfs_mapping_tree *map_tree;
337 u64 logical = bio->bi_sector << 9;
342 length = bio->bi_size;
343 map_tree = &root->fs_info->mapping_tree;
345 ret = btrfs_map_block(map_tree, READ, logical,
346 &map_length, NULL, 0);
348 if (map_length < length + size) {
354 int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
357 struct btrfs_root *root = BTRFS_I(inode)->root;
360 ret = btrfs_csum_one_bio(root, inode, bio);
363 return btrfs_map_bio(root, rw, bio, mirror_num, 1);
366 int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
369 struct btrfs_root *root = BTRFS_I(inode)->root;
372 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
375 if (!(rw & (1 << BIO_RW))) {
379 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
380 inode, rw, bio, mirror_num,
381 __btrfs_submit_bio_hook);
383 return btrfs_map_bio(root, rw, bio, mirror_num, 0);
386 static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
387 struct inode *inode, u64 file_offset,
388 struct list_head *list)
390 struct list_head *cur;
391 struct btrfs_ordered_sum *sum;
393 btrfs_set_trans_block_group(trans, inode);
394 list_for_each(cur, list) {
395 sum = list_entry(cur, struct btrfs_ordered_sum, list);
396 mutex_lock(&BTRFS_I(inode)->csum_mutex);
397 btrfs_csum_file_blocks(trans, BTRFS_I(inode)->root,
399 mutex_unlock(&BTRFS_I(inode)->csum_mutex);
404 struct btrfs_writepage_fixup {
406 struct btrfs_work work;
409 /* see btrfs_writepage_start_hook for details on why this is required */
410 void btrfs_writepage_fixup_worker(struct btrfs_work *work)
412 struct btrfs_writepage_fixup *fixup;
413 struct btrfs_ordered_extent *ordered;
419 fixup = container_of(work, struct btrfs_writepage_fixup, work);
423 if (!page->mapping || !PageDirty(page) || !PageChecked(page)) {
424 ClearPageChecked(page);
428 inode = page->mapping->host;
429 page_start = page_offset(page);
430 page_end = page_offset(page) + PAGE_CACHE_SIZE - 1;
432 lock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
434 /* already ordered? We're done */
435 if (test_range_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
436 EXTENT_ORDERED, 0)) {
440 ordered = btrfs_lookup_ordered_extent(inode, page_start);
442 unlock_extent(&BTRFS_I(inode)->io_tree, page_start,
445 btrfs_start_ordered_extent(inode, ordered, 1);
449 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, page_end,
451 ClearPageChecked(page);
453 unlock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
456 page_cache_release(page);
460 * There are a few paths in the higher layers of the kernel that directly
461 * set the page dirty bit without asking the filesystem if it is a
462 * good idea. This causes problems because we want to make sure COW
463 * properly happens and the data=ordered rules are followed.
465 * In our case any range that doesn't have the EXTENT_ORDERED bit set
466 * hasn't been properly setup for IO. We kick off an async process
467 * to fix it up. The async helper will wait for ordered extents, set
468 * the delalloc bit and make it safe to write the page.
470 int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
472 struct inode *inode = page->mapping->host;
473 struct btrfs_writepage_fixup *fixup;
474 struct btrfs_root *root = BTRFS_I(inode)->root;
477 ret = test_range_bit(&BTRFS_I(inode)->io_tree, start, end,
482 if (PageChecked(page))
485 fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
488 printk("queueing worker to fixup page %lu %Lu\n", inode->i_ino, page_offset(page));
489 SetPageChecked(page);
490 page_cache_get(page);
491 fixup->work.func = btrfs_writepage_fixup_worker;
493 btrfs_queue_worker(&root->fs_info->fixup_workers, &fixup->work);
497 static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
499 struct btrfs_root *root = BTRFS_I(inode)->root;
500 struct btrfs_trans_handle *trans;
501 struct btrfs_ordered_extent *ordered_extent;
502 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
503 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
504 struct extent_map *em;
508 struct list_head list;
509 struct btrfs_key ins;
512 ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1);
516 trans = btrfs_join_transaction(root, 1);
518 ordered_extent = btrfs_lookup_ordered_extent(inode, start);
519 BUG_ON(!ordered_extent);
521 lock_extent(io_tree, ordered_extent->file_offset,
522 ordered_extent->file_offset + ordered_extent->len - 1,
525 INIT_LIST_HEAD(&list);
527 ins.objectid = ordered_extent->start;
528 ins.offset = ordered_extent->len;
529 ins.type = BTRFS_EXTENT_ITEM_KEY;
531 ret = btrfs_alloc_reserved_extent(trans, root, root->root_key.objectid,
532 trans->transid, inode->i_ino,
533 ordered_extent->file_offset, &ins);
536 mutex_lock(&BTRFS_I(inode)->extent_mutex);
538 ret = btrfs_drop_extents(trans, root, inode,
539 ordered_extent->file_offset,
540 ordered_extent->file_offset +
542 ordered_extent->file_offset, &alloc_hint);
544 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
545 ordered_extent->file_offset,
546 ordered_extent->start,
548 ordered_extent->len, 0);
551 spin_lock(&em_tree->lock);
552 clear_start = ordered_extent->file_offset;
553 clear_end = ordered_extent->file_offset + ordered_extent->len;
554 while(clear_start < clear_end) {
555 em = lookup_extent_mapping(em_tree, clear_start,
556 clear_end - clear_start);
558 clear_bit(EXTENT_FLAG_PINNED, &em->flags);
559 clear_start = em->start + em->len;
565 spin_unlock(&em_tree->lock);
567 btrfs_drop_extent_cache(inode, ordered_extent->file_offset,
568 ordered_extent->file_offset +
569 ordered_extent->len - 1);
570 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
572 inode->i_blocks += ordered_extent->len >> 9;
573 unlock_extent(io_tree, ordered_extent->file_offset,
574 ordered_extent->file_offset + ordered_extent->len - 1,
576 add_pending_csums(trans, inode, ordered_extent->file_offset,
577 &ordered_extent->list);
579 btrfs_ordered_update_i_size(inode, ordered_extent);
580 btrfs_remove_ordered_extent(inode, ordered_extent);
583 btrfs_put_ordered_extent(ordered_extent);
584 /* once for the tree */
585 btrfs_put_ordered_extent(ordered_extent);
587 btrfs_update_inode(trans, root, inode);
588 btrfs_end_transaction(trans, root);
592 int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
593 struct extent_state *state, int uptodate)
595 return btrfs_finish_ordered_io(page->mapping->host, start, end);
598 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
601 struct inode *inode = page->mapping->host;
602 struct btrfs_root *root = BTRFS_I(inode)->root;
603 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
604 struct btrfs_csum_item *item;
605 struct btrfs_path *path = NULL;
608 if (btrfs_test_opt(root, NODATASUM) ||
609 btrfs_test_flag(inode, NODATASUM))
612 path = btrfs_alloc_path();
613 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
616 * It is possible there is an ordered extent that has
617 * not yet finished for this range in the file. If so,
618 * that extent will have a csum cached, and it will insert
619 * the sum after all the blocks in the extent are fully
620 * on disk. So, look for an ordered extent and use the
623 ret = btrfs_find_ordered_sum(inode, start, &csum);
628 /* a csum that isn't present is a preallocated region. */
629 if (ret == -ENOENT || ret == -EFBIG)
632 printk("no csum found for inode %lu start %Lu\n", inode->i_ino,
636 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
639 set_state_private(io_tree, start, csum);
642 btrfs_free_path(path);
646 struct io_failure_record {
654 int btrfs_io_failed_hook(struct bio *failed_bio,
655 struct page *page, u64 start, u64 end,
656 struct extent_state *state)
658 struct io_failure_record *failrec = NULL;
660 struct extent_map *em;
661 struct inode *inode = page->mapping->host;
662 struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
663 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
670 ret = get_state_private(failure_tree, start, &private);
672 failrec = kmalloc(sizeof(*failrec), GFP_NOFS);
675 failrec->start = start;
676 failrec->len = end - start + 1;
677 failrec->last_mirror = 0;
679 spin_lock(&em_tree->lock);
680 em = lookup_extent_mapping(em_tree, start, failrec->len);
681 if (em->start > start || em->start + em->len < start) {
685 spin_unlock(&em_tree->lock);
687 if (!em || IS_ERR(em)) {
691 logical = start - em->start;
692 logical = em->block_start + logical;
693 failrec->logical = logical;
695 set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
696 EXTENT_DIRTY, GFP_NOFS);
697 set_state_private(failure_tree, start,
698 (u64)(unsigned long)failrec);
700 failrec = (struct io_failure_record *)(unsigned long)private;
702 num_copies = btrfs_num_copies(
703 &BTRFS_I(inode)->root->fs_info->mapping_tree,
704 failrec->logical, failrec->len);
705 failrec->last_mirror++;
707 spin_lock_irq(&BTRFS_I(inode)->io_tree.lock);
708 state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
711 if (state && state->start != failrec->start)
713 spin_unlock_irq(&BTRFS_I(inode)->io_tree.lock);
715 if (!state || failrec->last_mirror > num_copies) {
716 set_state_private(failure_tree, failrec->start, 0);
717 clear_extent_bits(failure_tree, failrec->start,
718 failrec->start + failrec->len - 1,
719 EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
723 bio = bio_alloc(GFP_NOFS, 1);
724 bio->bi_private = state;
725 bio->bi_end_io = failed_bio->bi_end_io;
726 bio->bi_sector = failrec->logical >> 9;
727 bio->bi_bdev = failed_bio->bi_bdev;
729 bio_add_page(bio, page, failrec->len, start - page_offset(page));
730 if (failed_bio->bi_rw & (1 << BIO_RW))
735 BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
736 failrec->last_mirror);
740 int btrfs_clean_io_failures(struct inode *inode, u64 start)
744 struct io_failure_record *failure;
748 if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
749 (u64)-1, 1, EXTENT_DIRTY)) {
750 ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
751 start, &private_failure);
753 failure = (struct io_failure_record *)(unsigned long)
755 set_state_private(&BTRFS_I(inode)->io_failure_tree,
757 clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
759 failure->start + failure->len - 1,
760 EXTENT_DIRTY | EXTENT_LOCKED,
768 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
769 struct extent_state *state)
771 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
772 struct inode *inode = page->mapping->host;
773 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
775 u64 private = ~(u32)0;
777 struct btrfs_root *root = BTRFS_I(inode)->root;
781 if (btrfs_test_opt(root, NODATASUM) ||
782 btrfs_test_flag(inode, NODATASUM))
784 if (state && state->start == start) {
785 private = state->private;
788 ret = get_state_private(io_tree, start, &private);
790 local_irq_save(flags);
791 kaddr = kmap_atomic(page, KM_IRQ0);
795 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
796 btrfs_csum_final(csum, (char *)&csum);
797 if (csum != private) {
800 kunmap_atomic(kaddr, KM_IRQ0);
801 local_irq_restore(flags);
803 /* if the io failure tree for this inode is non-empty,
804 * check to see if we've recovered from a failed IO
806 btrfs_clean_io_failures(inode, start);
810 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
811 page->mapping->host->i_ino, (unsigned long long)start, csum,
813 memset(kaddr + offset, 1, end - start + 1);
814 flush_dcache_page(page);
815 kunmap_atomic(kaddr, KM_IRQ0);
816 local_irq_restore(flags);
822 void btrfs_read_locked_inode(struct inode *inode)
824 struct btrfs_path *path;
825 struct extent_buffer *leaf;
826 struct btrfs_inode_item *inode_item;
827 struct btrfs_timespec *tspec;
828 struct btrfs_root *root = BTRFS_I(inode)->root;
829 struct btrfs_key location;
830 u64 alloc_group_block;
834 path = btrfs_alloc_path();
836 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
838 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
842 leaf = path->nodes[0];
843 inode_item = btrfs_item_ptr(leaf, path->slots[0],
844 struct btrfs_inode_item);
846 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
847 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
848 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
849 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
850 btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item));
852 tspec = btrfs_inode_atime(inode_item);
853 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
854 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
856 tspec = btrfs_inode_mtime(inode_item);
857 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
858 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
860 tspec = btrfs_inode_ctime(inode_item);
861 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
862 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
864 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
865 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
867 rdev = btrfs_inode_rdev(leaf, inode_item);
869 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
870 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
872 BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
873 if (!BTRFS_I(inode)->block_group) {
874 BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
876 BTRFS_BLOCK_GROUP_METADATA, 0);
878 btrfs_free_path(path);
881 switch (inode->i_mode & S_IFMT) {
883 inode->i_mapping->a_ops = &btrfs_aops;
884 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
885 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
886 inode->i_fop = &btrfs_file_operations;
887 inode->i_op = &btrfs_file_inode_operations;
890 inode->i_fop = &btrfs_dir_file_operations;
891 if (root == root->fs_info->tree_root)
892 inode->i_op = &btrfs_dir_ro_inode_operations;
894 inode->i_op = &btrfs_dir_inode_operations;
897 inode->i_op = &btrfs_symlink_inode_operations;
898 inode->i_mapping->a_ops = &btrfs_symlink_aops;
899 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
902 init_special_inode(inode, inode->i_mode, rdev);
908 btrfs_free_path(path);
909 make_bad_inode(inode);
912 static void fill_inode_item(struct extent_buffer *leaf,
913 struct btrfs_inode_item *item,
916 btrfs_set_inode_uid(leaf, item, inode->i_uid);
917 btrfs_set_inode_gid(leaf, item, inode->i_gid);
918 btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size);
919 btrfs_set_inode_mode(leaf, item, inode->i_mode);
920 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
922 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
923 inode->i_atime.tv_sec);
924 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
925 inode->i_atime.tv_nsec);
927 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
928 inode->i_mtime.tv_sec);
929 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
930 inode->i_mtime.tv_nsec);
932 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
933 inode->i_ctime.tv_sec);
934 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
935 inode->i_ctime.tv_nsec);
937 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
938 btrfs_set_inode_generation(leaf, item, inode->i_generation);
939 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
940 btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
941 btrfs_set_inode_block_group(leaf, item,
942 BTRFS_I(inode)->block_group->key.objectid);
945 int noinline btrfs_update_inode(struct btrfs_trans_handle *trans,
946 struct btrfs_root *root,
949 struct btrfs_inode_item *inode_item;
950 struct btrfs_path *path;
951 struct extent_buffer *leaf;
954 path = btrfs_alloc_path();
956 ret = btrfs_lookup_inode(trans, root, path,
957 &BTRFS_I(inode)->location, 1);
964 leaf = path->nodes[0];
965 inode_item = btrfs_item_ptr(leaf, path->slots[0],
966 struct btrfs_inode_item);
968 fill_inode_item(leaf, inode_item, inode);
969 btrfs_mark_buffer_dirty(leaf);
970 btrfs_set_inode_last_trans(trans, inode);
973 btrfs_free_path(path);
978 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
979 struct btrfs_root *root,
981 struct dentry *dentry)
983 struct btrfs_path *path;
984 const char *name = dentry->d_name.name;
985 int name_len = dentry->d_name.len;
987 struct extent_buffer *leaf;
988 struct btrfs_dir_item *di;
989 struct btrfs_key key;
991 path = btrfs_alloc_path();
997 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
1007 leaf = path->nodes[0];
1008 btrfs_dir_item_key_to_cpu(leaf, di, &key);
1009 ret = btrfs_delete_one_dir_name(trans, root, path, di);
1012 btrfs_release_path(root, path);
1014 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
1015 key.objectid, name, name_len, -1);
1024 ret = btrfs_delete_one_dir_name(trans, root, path, di);
1025 btrfs_release_path(root, path);
1027 dentry->d_inode->i_ctime = dir->i_ctime;
1028 ret = btrfs_del_inode_ref(trans, root, name, name_len,
1029 dentry->d_inode->i_ino,
1030 dentry->d_parent->d_inode->i_ino);
1032 printk("failed to delete reference to %.*s, "
1033 "inode %lu parent %lu\n", name_len, name,
1034 dentry->d_inode->i_ino,
1035 dentry->d_parent->d_inode->i_ino);
1038 btrfs_free_path(path);
1040 btrfs_i_size_write(dir, dir->i_size - name_len * 2);
1041 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
1042 btrfs_update_inode(trans, root, dir);
1043 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1044 dentry->d_inode->i_nlink--;
1046 drop_nlink(dentry->d_inode);
1048 ret = btrfs_update_inode(trans, root, dentry->d_inode);
1049 dir->i_sb->s_dirt = 1;
1054 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
1056 struct btrfs_root *root;
1057 struct btrfs_trans_handle *trans;
1059 unsigned long nr = 0;
1061 root = BTRFS_I(dir)->root;
1063 ret = btrfs_check_free_space(root, 1, 1);
1067 trans = btrfs_start_transaction(root, 1);
1069 btrfs_set_trans_block_group(trans, dir);
1070 ret = btrfs_unlink_trans(trans, root, dir, dentry);
1071 nr = trans->blocks_used;
1073 btrfs_end_transaction_throttle(trans, root);
1075 btrfs_btree_balance_dirty(root, nr);
1079 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
1081 struct inode *inode = dentry->d_inode;
1084 struct btrfs_root *root = BTRFS_I(dir)->root;
1085 struct btrfs_trans_handle *trans;
1086 unsigned long nr = 0;
1088 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
1092 ret = btrfs_check_free_space(root, 1, 1);
1096 trans = btrfs_start_transaction(root, 1);
1097 btrfs_set_trans_block_group(trans, dir);
1099 /* now the directory is empty */
1100 err = btrfs_unlink_trans(trans, root, dir, dentry);
1102 btrfs_i_size_write(inode, 0);
1105 nr = trans->blocks_used;
1106 ret = btrfs_end_transaction_throttle(trans, root);
1108 btrfs_btree_balance_dirty(root, nr);
1116 * this can truncate away extent items, csum items and directory items.
1117 * It starts at a high offset and removes keys until it can't find
1118 * any higher than i_size.
1120 * csum items that cross the new i_size are truncated to the new size
1123 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
1124 struct btrfs_root *root,
1125 struct inode *inode,
1129 struct btrfs_path *path;
1130 struct btrfs_key key;
1131 struct btrfs_key found_key;
1133 struct extent_buffer *leaf;
1134 struct btrfs_file_extent_item *fi;
1135 u64 extent_start = 0;
1136 u64 extent_num_bytes = 0;
1142 int pending_del_nr = 0;
1143 int pending_del_slot = 0;
1144 int extent_type = -1;
1145 u64 mask = root->sectorsize - 1;
1147 btrfs_drop_extent_cache(inode, inode->i_size & (~mask), (u64)-1);
1148 path = btrfs_alloc_path();
1152 /* FIXME, add redo link to tree so we don't leak on crash */
1153 key.objectid = inode->i_ino;
1154 key.offset = (u64)-1;
1157 btrfs_init_path(path);
1159 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1164 BUG_ON(path->slots[0] == 0);
1170 leaf = path->nodes[0];
1171 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1172 found_type = btrfs_key_type(&found_key);
1174 if (found_key.objectid != inode->i_ino)
1177 if (found_type < min_type)
1180 item_end = found_key.offset;
1181 if (found_type == BTRFS_EXTENT_DATA_KEY) {
1182 fi = btrfs_item_ptr(leaf, path->slots[0],
1183 struct btrfs_file_extent_item);
1184 extent_type = btrfs_file_extent_type(leaf, fi);
1185 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
1187 btrfs_file_extent_num_bytes(leaf, fi);
1188 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1189 struct btrfs_item *item = btrfs_item_nr(leaf,
1191 item_end += btrfs_file_extent_inline_len(leaf,
1196 if (found_type == BTRFS_CSUM_ITEM_KEY) {
1197 ret = btrfs_csum_truncate(trans, root, path,
1201 if (item_end < inode->i_size) {
1202 if (found_type == BTRFS_DIR_ITEM_KEY) {
1203 found_type = BTRFS_INODE_ITEM_KEY;
1204 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
1205 found_type = BTRFS_CSUM_ITEM_KEY;
1206 } else if (found_type == BTRFS_EXTENT_DATA_KEY) {
1207 found_type = BTRFS_XATTR_ITEM_KEY;
1208 } else if (found_type == BTRFS_XATTR_ITEM_KEY) {
1209 found_type = BTRFS_INODE_REF_KEY;
1210 } else if (found_type) {
1215 btrfs_set_key_type(&key, found_type);
1218 if (found_key.offset >= inode->i_size)
1224 /* FIXME, shrink the extent if the ref count is only 1 */
1225 if (found_type != BTRFS_EXTENT_DATA_KEY)
1228 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
1230 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
1232 u64 orig_num_bytes =
1233 btrfs_file_extent_num_bytes(leaf, fi);
1234 extent_num_bytes = inode->i_size -
1235 found_key.offset + root->sectorsize - 1;
1236 extent_num_bytes = extent_num_bytes &
1237 ~((u64)root->sectorsize - 1);
1238 btrfs_set_file_extent_num_bytes(leaf, fi,
1240 num_dec = (orig_num_bytes -
1242 if (extent_start != 0)
1243 dec_i_blocks(inode, num_dec);
1244 btrfs_mark_buffer_dirty(leaf);
1247 btrfs_file_extent_disk_num_bytes(leaf,
1249 /* FIXME blocksize != 4096 */
1250 num_dec = btrfs_file_extent_num_bytes(leaf, fi);
1251 if (extent_start != 0) {
1253 dec_i_blocks(inode, num_dec);
1255 root_gen = btrfs_header_generation(leaf);
1256 root_owner = btrfs_header_owner(leaf);
1258 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1260 u32 newsize = inode->i_size - found_key.offset;
1261 dec_i_blocks(inode, item_end + 1 -
1262 found_key.offset - newsize);
1264 btrfs_file_extent_calc_inline_size(newsize);
1265 ret = btrfs_truncate_item(trans, root, path,
1269 dec_i_blocks(inode, item_end + 1 -
1275 if (!pending_del_nr) {
1276 /* no pending yet, add ourselves */
1277 pending_del_slot = path->slots[0];
1279 } else if (pending_del_nr &&
1280 path->slots[0] + 1 == pending_del_slot) {
1281 /* hop on the pending chunk */
1283 pending_del_slot = path->slots[0];
1285 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot);
1291 ret = btrfs_free_extent(trans, root, extent_start,
1294 root_gen, inode->i_ino,
1295 found_key.offset, 0);
1299 if (path->slots[0] == 0) {
1302 btrfs_release_path(root, path);
1307 if (pending_del_nr &&
1308 path->slots[0] + 1 != pending_del_slot) {
1309 struct btrfs_key debug;
1311 btrfs_item_key_to_cpu(path->nodes[0], &debug,
1313 ret = btrfs_del_items(trans, root, path,
1318 btrfs_release_path(root, path);
1324 if (pending_del_nr) {
1325 ret = btrfs_del_items(trans, root, path, pending_del_slot,
1328 btrfs_free_path(path);
1329 inode->i_sb->s_dirt = 1;
1334 * taken from block_truncate_page, but does cow as it zeros out
1335 * any bytes left in the last page in the file.
1337 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
1339 struct inode *inode = mapping->host;
1340 struct btrfs_root *root = BTRFS_I(inode)->root;
1341 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1342 struct btrfs_ordered_extent *ordered;
1344 u32 blocksize = root->sectorsize;
1345 pgoff_t index = from >> PAGE_CACHE_SHIFT;
1346 unsigned offset = from & (PAGE_CACHE_SIZE-1);
1352 if ((offset & (blocksize - 1)) == 0)
1357 page = grab_cache_page(mapping, index);
1361 page_start = page_offset(page);
1362 page_end = page_start + PAGE_CACHE_SIZE - 1;
1364 if (!PageUptodate(page)) {
1365 ret = btrfs_readpage(NULL, page);
1367 if (page->mapping != mapping) {
1369 page_cache_release(page);
1372 if (!PageUptodate(page)) {
1377 wait_on_page_writeback(page);
1379 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
1380 set_page_extent_mapped(page);
1382 ordered = btrfs_lookup_ordered_extent(inode, page_start);
1384 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
1386 page_cache_release(page);
1387 btrfs_start_ordered_extent(inode, ordered, 1);
1388 btrfs_put_ordered_extent(ordered);
1392 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
1393 page_end, GFP_NOFS);
1395 if (offset != PAGE_CACHE_SIZE) {
1397 memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
1398 flush_dcache_page(page);
1401 ClearPageChecked(page);
1402 set_page_dirty(page);
1403 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
1406 page_cache_release(page);
1411 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
1413 struct inode *inode = dentry->d_inode;
1416 err = inode_change_ok(inode, attr);
1420 if (S_ISREG(inode->i_mode) &&
1421 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
1422 struct btrfs_trans_handle *trans;
1423 struct btrfs_root *root = BTRFS_I(inode)->root;
1424 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1426 u64 mask = root->sectorsize - 1;
1427 u64 hole_start = (inode->i_size + mask) & ~mask;
1428 u64 block_end = (attr->ia_size + mask) & ~mask;
1432 if (attr->ia_size <= hole_start)
1435 err = btrfs_check_free_space(root, 1, 0);
1439 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1441 hole_size = block_end - hole_start;
1442 btrfs_wait_ordered_range(inode, hole_start, hole_size);
1443 lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1445 trans = btrfs_start_transaction(root, 1);
1446 btrfs_set_trans_block_group(trans, inode);
1447 mutex_lock(&BTRFS_I(inode)->extent_mutex);
1448 err = btrfs_drop_extents(trans, root, inode,
1449 hole_start, block_end, hole_start,
1452 if (alloc_hint != EXTENT_MAP_INLINE) {
1453 err = btrfs_insert_file_extent(trans, root,
1457 btrfs_drop_extent_cache(inode, hole_start,
1459 btrfs_check_file(root, inode);
1461 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
1462 btrfs_end_transaction(trans, root);
1463 unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1468 err = inode_setattr(inode, attr);
1473 void btrfs_delete_inode(struct inode *inode)
1475 struct btrfs_trans_handle *trans;
1476 struct btrfs_root *root = BTRFS_I(inode)->root;
1480 truncate_inode_pages(&inode->i_data, 0);
1481 if (is_bad_inode(inode)) {
1484 btrfs_wait_ordered_range(inode, 0, (u64)-1);
1486 btrfs_i_size_write(inode, 0);
1487 trans = btrfs_start_transaction(root, 1);
1489 btrfs_set_trans_block_group(trans, inode);
1490 ret = btrfs_truncate_in_trans(trans, root, inode, 0);
1492 goto no_delete_lock;
1494 nr = trans->blocks_used;
1497 btrfs_end_transaction(trans, root);
1498 btrfs_btree_balance_dirty(root, nr);
1502 nr = trans->blocks_used;
1503 btrfs_end_transaction(trans, root);
1504 btrfs_btree_balance_dirty(root, nr);
1510 * this returns the key found in the dir entry in the location pointer.
1511 * If no dir entries were found, location->objectid is 0.
1513 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
1514 struct btrfs_key *location)
1516 const char *name = dentry->d_name.name;
1517 int namelen = dentry->d_name.len;
1518 struct btrfs_dir_item *di;
1519 struct btrfs_path *path;
1520 struct btrfs_root *root = BTRFS_I(dir)->root;
1523 if (namelen == 1 && strcmp(name, ".") == 0) {
1524 location->objectid = dir->i_ino;
1525 location->type = BTRFS_INODE_ITEM_KEY;
1526 location->offset = 0;
1529 path = btrfs_alloc_path();
1532 if (namelen == 2 && strcmp(name, "..") == 0) {
1533 struct btrfs_key key;
1534 struct extent_buffer *leaf;
1538 key.objectid = dir->i_ino;
1539 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1541 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1545 leaf = path->nodes[0];
1546 slot = path->slots[0];
1547 nritems = btrfs_header_nritems(leaf);
1548 if (slot >= nritems)
1551 btrfs_item_key_to_cpu(leaf, &key, slot);
1552 if (key.objectid != dir->i_ino ||
1553 key.type != BTRFS_INODE_REF_KEY) {
1556 location->objectid = key.offset;
1557 location->type = BTRFS_INODE_ITEM_KEY;
1558 location->offset = 0;
1562 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
1566 if (!di || IS_ERR(di)) {
1569 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
1571 btrfs_free_path(path);
1574 location->objectid = 0;
1579 * when we hit a tree root in a directory, the btrfs part of the inode
1580 * needs to be changed to reflect the root directory of the tree root. This
1581 * is kind of like crossing a mount point.
1583 static int fixup_tree_root_location(struct btrfs_root *root,
1584 struct btrfs_key *location,
1585 struct btrfs_root **sub_root,
1586 struct dentry *dentry)
1588 struct btrfs_path *path;
1589 struct btrfs_root_item *ri;
1591 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
1593 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1596 path = btrfs_alloc_path();
1599 *sub_root = btrfs_read_fs_root(root->fs_info, location,
1600 dentry->d_name.name,
1601 dentry->d_name.len);
1602 if (IS_ERR(*sub_root))
1603 return PTR_ERR(*sub_root);
1605 ri = &(*sub_root)->root_item;
1606 location->objectid = btrfs_root_dirid(ri);
1607 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1608 location->offset = 0;
1610 btrfs_free_path(path);
1614 static int btrfs_init_locked_inode(struct inode *inode, void *p)
1616 struct btrfs_iget_args *args = p;
1617 inode->i_ino = args->ino;
1618 BTRFS_I(inode)->root = args->root;
1619 BTRFS_I(inode)->delalloc_bytes = 0;
1620 BTRFS_I(inode)->disk_i_size = 0;
1621 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1622 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1623 inode->i_mapping, GFP_NOFS);
1624 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1625 inode->i_mapping, GFP_NOFS);
1626 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
1627 mutex_init(&BTRFS_I(inode)->csum_mutex);
1628 mutex_init(&BTRFS_I(inode)->extent_mutex);
1632 static int btrfs_find_actor(struct inode *inode, void *opaque)
1634 struct btrfs_iget_args *args = opaque;
1635 return (args->ino == inode->i_ino &&
1636 args->root == BTRFS_I(inode)->root);
1639 struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
1642 struct btrfs_iget_args args;
1643 args.ino = objectid;
1644 args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
1649 return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
1652 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
1653 struct btrfs_root *root)
1655 struct inode *inode;
1656 struct btrfs_iget_args args;
1657 args.ino = objectid;
1660 inode = iget5_locked(s, objectid, btrfs_find_actor,
1661 btrfs_init_locked_inode,
1666 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
1667 struct nameidata *nd)
1669 struct inode * inode;
1670 struct btrfs_inode *bi = BTRFS_I(dir);
1671 struct btrfs_root *root = bi->root;
1672 struct btrfs_root *sub_root = root;
1673 struct btrfs_key location;
1676 if (dentry->d_name.len > BTRFS_NAME_LEN)
1677 return ERR_PTR(-ENAMETOOLONG);
1679 ret = btrfs_inode_by_name(dir, dentry, &location);
1682 return ERR_PTR(ret);
1685 if (location.objectid) {
1686 ret = fixup_tree_root_location(root, &location, &sub_root,
1689 return ERR_PTR(ret);
1691 return ERR_PTR(-ENOENT);
1692 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
1695 return ERR_PTR(-EACCES);
1696 if (inode->i_state & I_NEW) {
1697 /* the inode and parent dir are two different roots */
1698 if (sub_root != root) {
1700 sub_root->inode = inode;
1702 BTRFS_I(inode)->root = sub_root;
1703 memcpy(&BTRFS_I(inode)->location, &location,
1705 btrfs_read_locked_inode(inode);
1706 unlock_new_inode(inode);
1709 return d_splice_alias(inode, dentry);
1712 static unsigned char btrfs_filetype_table[] = {
1713 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1716 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1718 struct inode *inode = filp->f_dentry->d_inode;
1719 struct btrfs_root *root = BTRFS_I(inode)->root;
1720 struct btrfs_item *item;
1721 struct btrfs_dir_item *di;
1722 struct btrfs_key key;
1723 struct btrfs_key found_key;
1724 struct btrfs_path *path;
1727 struct extent_buffer *leaf;
1730 unsigned char d_type;
1735 int key_type = BTRFS_DIR_INDEX_KEY;
1740 /* FIXME, use a real flag for deciding about the key type */
1741 if (root->fs_info->tree_root == root)
1742 key_type = BTRFS_DIR_ITEM_KEY;
1744 /* special case for "." */
1745 if (filp->f_pos == 0) {
1746 over = filldir(dirent, ".", 1,
1754 key.objectid = inode->i_ino;
1755 path = btrfs_alloc_path();
1758 /* special case for .., just use the back ref */
1759 if (filp->f_pos == 1) {
1760 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1762 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1764 leaf = path->nodes[0];
1765 slot = path->slots[0];
1766 nritems = btrfs_header_nritems(leaf);
1767 if (slot >= nritems) {
1768 btrfs_release_path(root, path);
1769 goto read_dir_items;
1771 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1772 btrfs_release_path(root, path);
1773 if (found_key.objectid != key.objectid ||
1774 found_key.type != BTRFS_INODE_REF_KEY)
1775 goto read_dir_items;
1776 over = filldir(dirent, "..", 2,
1777 2, found_key.offset, DT_DIR);
1784 btrfs_set_key_type(&key, key_type);
1785 key.offset = filp->f_pos;
1787 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1792 leaf = path->nodes[0];
1793 nritems = btrfs_header_nritems(leaf);
1794 slot = path->slots[0];
1795 if (advance || slot >= nritems) {
1796 if (slot >= nritems -1) {
1797 ret = btrfs_next_leaf(root, path);
1800 leaf = path->nodes[0];
1801 nritems = btrfs_header_nritems(leaf);
1802 slot = path->slots[0];
1809 item = btrfs_item_nr(leaf, slot);
1810 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1812 if (found_key.objectid != key.objectid)
1814 if (btrfs_key_type(&found_key) != key_type)
1816 if (found_key.offset < filp->f_pos)
1819 filp->f_pos = found_key.offset;
1821 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1823 di_total = btrfs_item_size(leaf, item);
1824 while(di_cur < di_total) {
1825 struct btrfs_key location;
1827 name_len = btrfs_dir_name_len(leaf, di);
1828 if (name_len < 32) {
1829 name_ptr = tmp_name;
1831 name_ptr = kmalloc(name_len, GFP_NOFS);
1834 read_extent_buffer(leaf, name_ptr,
1835 (unsigned long)(di + 1), name_len);
1837 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1838 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1839 over = filldir(dirent, name_ptr, name_len,
1844 if (name_ptr != tmp_name)
1849 di_len = btrfs_dir_name_len(leaf, di) +
1850 btrfs_dir_data_len(leaf, di) +sizeof(*di);
1852 di = (struct btrfs_dir_item *)((char *)di + di_len);
1855 if (key_type == BTRFS_DIR_INDEX_KEY)
1856 filp->f_pos = INT_LIMIT(typeof(filp->f_pos));
1862 btrfs_free_path(path);
1866 int btrfs_write_inode(struct inode *inode, int wait)
1868 struct btrfs_root *root = BTRFS_I(inode)->root;
1869 struct btrfs_trans_handle *trans;
1873 trans = btrfs_join_transaction(root, 1);
1874 btrfs_set_trans_block_group(trans, inode);
1875 ret = btrfs_commit_transaction(trans, root);
1881 * This is somewhat expensive, updating the tree every time the
1882 * inode changes. But, it is most likely to find the inode in cache.
1883 * FIXME, needs more benchmarking...there are no reasons other than performance
1884 * to keep or drop this code.
1886 void btrfs_dirty_inode(struct inode *inode)
1888 struct btrfs_root *root = BTRFS_I(inode)->root;
1889 struct btrfs_trans_handle *trans;
1891 trans = btrfs_join_transaction(root, 1);
1892 btrfs_set_trans_block_group(trans, inode);
1893 btrfs_update_inode(trans, root, inode);
1894 btrfs_end_transaction(trans, root);
1897 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1898 struct btrfs_root *root,
1899 const char *name, int name_len,
1902 struct btrfs_block_group_cache *group,
1905 struct inode *inode;
1906 struct btrfs_inode_item *inode_item;
1907 struct btrfs_block_group_cache *new_inode_group;
1908 struct btrfs_key *location;
1909 struct btrfs_path *path;
1910 struct btrfs_inode_ref *ref;
1911 struct btrfs_key key[2];
1917 path = btrfs_alloc_path();
1920 inode = new_inode(root->fs_info->sb);
1922 return ERR_PTR(-ENOMEM);
1924 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1925 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1926 inode->i_mapping, GFP_NOFS);
1927 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1928 inode->i_mapping, GFP_NOFS);
1929 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
1930 mutex_init(&BTRFS_I(inode)->csum_mutex);
1931 mutex_init(&BTRFS_I(inode)->extent_mutex);
1932 BTRFS_I(inode)->delalloc_bytes = 0;
1933 BTRFS_I(inode)->disk_i_size = 0;
1934 BTRFS_I(inode)->root = root;
1940 new_inode_group = btrfs_find_block_group(root, group, 0,
1941 BTRFS_BLOCK_GROUP_METADATA, owner);
1942 if (!new_inode_group) {
1943 printk("find_block group failed\n");
1944 new_inode_group = group;
1946 BTRFS_I(inode)->block_group = new_inode_group;
1947 BTRFS_I(inode)->flags = 0;
1949 key[0].objectid = objectid;
1950 btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
1953 key[1].objectid = objectid;
1954 btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
1955 key[1].offset = ref_objectid;
1957 sizes[0] = sizeof(struct btrfs_inode_item);
1958 sizes[1] = name_len + sizeof(*ref);
1960 ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
1964 if (objectid > root->highest_inode)
1965 root->highest_inode = objectid;
1967 inode->i_uid = current->fsuid;
1968 inode->i_gid = current->fsgid;
1969 inode->i_mode = mode;
1970 inode->i_ino = objectid;
1971 inode->i_blocks = 0;
1972 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1973 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1974 struct btrfs_inode_item);
1975 fill_inode_item(path->nodes[0], inode_item, inode);
1977 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
1978 struct btrfs_inode_ref);
1979 btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
1980 ptr = (unsigned long)(ref + 1);
1981 write_extent_buffer(path->nodes[0], name, ptr, name_len);
1983 btrfs_mark_buffer_dirty(path->nodes[0]);
1984 btrfs_free_path(path);
1986 location = &BTRFS_I(inode)->location;
1987 location->objectid = objectid;
1988 location->offset = 0;
1989 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1991 insert_inode_hash(inode);
1994 btrfs_free_path(path);
1995 return ERR_PTR(ret);
1998 static inline u8 btrfs_inode_type(struct inode *inode)
2000 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
2003 static int btrfs_add_link(struct btrfs_trans_handle *trans,
2004 struct dentry *dentry, struct inode *inode,
2008 struct btrfs_key key;
2009 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
2010 struct inode *parent_inode;
2012 key.objectid = inode->i_ino;
2013 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
2016 ret = btrfs_insert_dir_item(trans, root,
2017 dentry->d_name.name, dentry->d_name.len,
2018 dentry->d_parent->d_inode->i_ino,
2019 &key, btrfs_inode_type(inode));
2022 ret = btrfs_insert_inode_ref(trans, root,
2023 dentry->d_name.name,
2026 dentry->d_parent->d_inode->i_ino);
2028 parent_inode = dentry->d_parent->d_inode;
2029 btrfs_i_size_write(parent_inode, parent_inode->i_size +
2030 dentry->d_name.len * 2);
2031 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
2032 ret = btrfs_update_inode(trans, root,
2033 dentry->d_parent->d_inode);
2038 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
2039 struct dentry *dentry, struct inode *inode,
2042 int err = btrfs_add_link(trans, dentry, inode, backref);
2044 d_instantiate(dentry, inode);
2052 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
2053 int mode, dev_t rdev)
2055 struct btrfs_trans_handle *trans;
2056 struct btrfs_root *root = BTRFS_I(dir)->root;
2057 struct inode *inode = NULL;
2061 unsigned long nr = 0;
2063 if (!new_valid_dev(rdev))
2066 err = btrfs_check_free_space(root, 1, 0);
2070 trans = btrfs_start_transaction(root, 1);
2071 btrfs_set_trans_block_group(trans, dir);
2073 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2079 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2081 dentry->d_parent->d_inode->i_ino, objectid,
2082 BTRFS_I(dir)->block_group, mode);
2083 err = PTR_ERR(inode);
2087 btrfs_set_trans_block_group(trans, inode);
2088 err = btrfs_add_nondir(trans, dentry, inode, 0);
2092 inode->i_op = &btrfs_special_inode_operations;
2093 init_special_inode(inode, inode->i_mode, rdev);
2094 btrfs_update_inode(trans, root, inode);
2096 dir->i_sb->s_dirt = 1;
2097 btrfs_update_inode_block_group(trans, inode);
2098 btrfs_update_inode_block_group(trans, dir);
2100 nr = trans->blocks_used;
2101 btrfs_end_transaction_throttle(trans, root);
2104 inode_dec_link_count(inode);
2107 btrfs_btree_balance_dirty(root, nr);
2111 static int btrfs_create(struct inode *dir, struct dentry *dentry,
2112 int mode, struct nameidata *nd)
2114 struct btrfs_trans_handle *trans;
2115 struct btrfs_root *root = BTRFS_I(dir)->root;
2116 struct inode *inode = NULL;
2119 unsigned long nr = 0;
2122 err = btrfs_check_free_space(root, 1, 0);
2125 trans = btrfs_start_transaction(root, 1);
2126 btrfs_set_trans_block_group(trans, dir);
2128 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2134 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2136 dentry->d_parent->d_inode->i_ino,
2137 objectid, BTRFS_I(dir)->block_group, mode);
2138 err = PTR_ERR(inode);
2142 btrfs_set_trans_block_group(trans, inode);
2143 err = btrfs_add_nondir(trans, dentry, inode, 0);
2147 inode->i_mapping->a_ops = &btrfs_aops;
2148 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
2149 inode->i_fop = &btrfs_file_operations;
2150 inode->i_op = &btrfs_file_inode_operations;
2151 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2152 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
2153 inode->i_mapping, GFP_NOFS);
2154 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
2155 inode->i_mapping, GFP_NOFS);
2156 mutex_init(&BTRFS_I(inode)->csum_mutex);
2157 mutex_init(&BTRFS_I(inode)->extent_mutex);
2158 BTRFS_I(inode)->delalloc_bytes = 0;
2159 BTRFS_I(inode)->disk_i_size = 0;
2160 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
2161 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
2163 dir->i_sb->s_dirt = 1;
2164 btrfs_update_inode_block_group(trans, inode);
2165 btrfs_update_inode_block_group(trans, dir);
2167 nr = trans->blocks_used;
2168 btrfs_end_transaction_throttle(trans, root);
2171 inode_dec_link_count(inode);
2174 btrfs_btree_balance_dirty(root, nr);
2178 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
2179 struct dentry *dentry)
2181 struct btrfs_trans_handle *trans;
2182 struct btrfs_root *root = BTRFS_I(dir)->root;
2183 struct inode *inode = old_dentry->d_inode;
2184 unsigned long nr = 0;
2188 if (inode->i_nlink == 0)
2191 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2196 err = btrfs_check_free_space(root, 1, 0);
2199 trans = btrfs_start_transaction(root, 1);
2201 btrfs_set_trans_block_group(trans, dir);
2202 atomic_inc(&inode->i_count);
2203 err = btrfs_add_nondir(trans, dentry, inode, 1);
2208 dir->i_sb->s_dirt = 1;
2209 btrfs_update_inode_block_group(trans, dir);
2210 err = btrfs_update_inode(trans, root, inode);
2215 nr = trans->blocks_used;
2216 btrfs_end_transaction_throttle(trans, root);
2219 inode_dec_link_count(inode);
2222 btrfs_btree_balance_dirty(root, nr);
2226 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
2228 struct inode *inode = NULL;
2229 struct btrfs_trans_handle *trans;
2230 struct btrfs_root *root = BTRFS_I(dir)->root;
2232 int drop_on_err = 0;
2234 unsigned long nr = 1;
2236 err = btrfs_check_free_space(root, 1, 0);
2240 trans = btrfs_start_transaction(root, 1);
2241 btrfs_set_trans_block_group(trans, dir);
2243 if (IS_ERR(trans)) {
2244 err = PTR_ERR(trans);
2248 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2254 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2256 dentry->d_parent->d_inode->i_ino, objectid,
2257 BTRFS_I(dir)->block_group, S_IFDIR | mode);
2258 if (IS_ERR(inode)) {
2259 err = PTR_ERR(inode);
2264 inode->i_op = &btrfs_dir_inode_operations;
2265 inode->i_fop = &btrfs_dir_file_operations;
2266 btrfs_set_trans_block_group(trans, inode);
2268 btrfs_i_size_write(inode, 0);
2269 err = btrfs_update_inode(trans, root, inode);
2273 err = btrfs_add_link(trans, dentry, inode, 0);
2277 d_instantiate(dentry, inode);
2279 dir->i_sb->s_dirt = 1;
2280 btrfs_update_inode_block_group(trans, inode);
2281 btrfs_update_inode_block_group(trans, dir);
2284 nr = trans->blocks_used;
2285 btrfs_end_transaction_throttle(trans, root);
2290 btrfs_btree_balance_dirty(root, nr);
2294 static int merge_extent_mapping(struct extent_map_tree *em_tree,
2295 struct extent_map *existing,
2296 struct extent_map *em,
2297 u64 map_start, u64 map_len)
2301 BUG_ON(map_start < em->start || map_start >= extent_map_end(em));
2302 start_diff = map_start - em->start;
2303 em->start = map_start;
2305 if (em->block_start < EXTENT_MAP_LAST_BYTE)
2306 em->block_start += start_diff;
2307 return add_extent_mapping(em_tree, em);
2310 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
2311 size_t pg_offset, u64 start, u64 len,
2317 u64 extent_start = 0;
2319 u64 objectid = inode->i_ino;
2321 struct btrfs_path *path;
2322 struct btrfs_root *root = BTRFS_I(inode)->root;
2323 struct btrfs_file_extent_item *item;
2324 struct extent_buffer *leaf;
2325 struct btrfs_key found_key;
2326 struct extent_map *em = NULL;
2327 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2328 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2329 struct btrfs_trans_handle *trans = NULL;
2331 path = btrfs_alloc_path();
2335 spin_lock(&em_tree->lock);
2336 em = lookup_extent_mapping(em_tree, start, len);
2338 em->bdev = root->fs_info->fs_devices->latest_bdev;
2339 spin_unlock(&em_tree->lock);
2342 if (em->start > start || em->start + em->len <= start)
2343 free_extent_map(em);
2344 else if (em->block_start == EXTENT_MAP_INLINE && page)
2345 free_extent_map(em);
2349 em = alloc_extent_map(GFP_NOFS);
2354 em->bdev = root->fs_info->fs_devices->latest_bdev;
2355 em->start = EXTENT_MAP_HOLE;
2357 ret = btrfs_lookup_file_extent(trans, root, path,
2358 objectid, start, trans != NULL);
2365 if (path->slots[0] == 0)
2370 leaf = path->nodes[0];
2371 item = btrfs_item_ptr(leaf, path->slots[0],
2372 struct btrfs_file_extent_item);
2373 /* are we inside the extent that was found? */
2374 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2375 found_type = btrfs_key_type(&found_key);
2376 if (found_key.objectid != objectid ||
2377 found_type != BTRFS_EXTENT_DATA_KEY) {
2381 found_type = btrfs_file_extent_type(leaf, item);
2382 extent_start = found_key.offset;
2383 if (found_type == BTRFS_FILE_EXTENT_REG) {
2384 extent_end = extent_start +
2385 btrfs_file_extent_num_bytes(leaf, item);
2387 if (start < extent_start || start >= extent_end) {
2389 if (start < extent_start) {
2390 if (start + len <= extent_start)
2392 em->len = extent_end - extent_start;
2398 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
2400 em->start = extent_start;
2401 em->len = extent_end - extent_start;
2402 em->block_start = EXTENT_MAP_HOLE;
2405 bytenr += btrfs_file_extent_offset(leaf, item);
2406 em->block_start = bytenr;
2407 em->start = extent_start;
2408 em->len = extent_end - extent_start;
2410 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
2415 size_t extent_offset;
2418 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
2420 extent_end = (extent_start + size + root->sectorsize - 1) &
2421 ~((u64)root->sectorsize - 1);
2422 if (start < extent_start || start >= extent_end) {
2424 if (start < extent_start) {
2425 if (start + len <= extent_start)
2427 em->len = extent_end - extent_start;
2433 em->block_start = EXTENT_MAP_INLINE;
2436 em->start = extent_start;
2441 page_start = page_offset(page) + pg_offset;
2442 extent_offset = page_start - extent_start;
2443 copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
2444 size - extent_offset);
2445 em->start = extent_start + extent_offset;
2446 em->len = (copy_size + root->sectorsize - 1) &
2447 ~((u64)root->sectorsize - 1);
2449 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
2450 if (create == 0 && !PageUptodate(page)) {
2451 read_extent_buffer(leaf, map + pg_offset, ptr,
2453 flush_dcache_page(page);
2454 } else if (create && PageUptodate(page)) {
2457 free_extent_map(em);
2459 btrfs_release_path(root, path);
2460 trans = btrfs_join_transaction(root, 1);
2463 write_extent_buffer(leaf, map + pg_offset, ptr,
2465 btrfs_mark_buffer_dirty(leaf);
2468 set_extent_uptodate(io_tree, em->start,
2469 extent_map_end(em) - 1, GFP_NOFS);
2472 printk("unkknown found_type %d\n", found_type);
2479 em->block_start = EXTENT_MAP_HOLE;
2481 btrfs_release_path(root, path);
2482 if (em->start > start || extent_map_end(em) <= start) {
2483 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len);
2489 spin_lock(&em_tree->lock);
2490 ret = add_extent_mapping(em_tree, em);
2491 /* it is possible that someone inserted the extent into the tree
2492 * while we had the lock dropped. It is also possible that
2493 * an overlapping map exists in the tree
2495 if (ret == -EEXIST) {
2496 struct extent_map *existing;
2500 existing = lookup_extent_mapping(em_tree, start, len);
2501 if (existing && (existing->start > start ||
2502 existing->start + existing->len <= start)) {
2503 free_extent_map(existing);
2507 existing = lookup_extent_mapping(em_tree, em->start,
2510 err = merge_extent_mapping(em_tree, existing,
2513 free_extent_map(existing);
2515 free_extent_map(em);
2520 printk("failing to insert %Lu %Lu\n",
2522 free_extent_map(em);
2526 free_extent_map(em);
2531 spin_unlock(&em_tree->lock);
2533 btrfs_free_path(path);
2535 ret = btrfs_end_transaction(trans, root);
2541 free_extent_map(em);
2543 return ERR_PTR(err);
2548 #if 0 /* waiting for O_DIRECT reads */
2549 static int btrfs_get_block(struct inode *inode, sector_t iblock,
2550 struct buffer_head *bh_result, int create)
2552 struct extent_map *em;
2553 u64 start = (u64)iblock << inode->i_blkbits;
2554 struct btrfs_multi_bio *multi = NULL;
2555 struct btrfs_root *root = BTRFS_I(inode)->root;
2561 em = btrfs_get_extent(inode, NULL, 0, start, bh_result->b_size, 0);
2563 if (!em || IS_ERR(em))
2566 if (em->start > start || em->start + em->len <= start) {
2570 if (em->block_start == EXTENT_MAP_INLINE) {
2575 len = em->start + em->len - start;
2576 len = min_t(u64, len, INT_LIMIT(typeof(bh_result->b_size)));
2578 if (em->block_start == EXTENT_MAP_HOLE ||
2579 em->block_start == EXTENT_MAP_DELALLOC) {
2580 bh_result->b_size = len;
2584 logical = start - em->start;
2585 logical = em->block_start + logical;
2588 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
2589 logical, &map_length, &multi, 0);
2591 bh_result->b_blocknr = multi->stripes[0].physical >> inode->i_blkbits;
2592 bh_result->b_size = min(map_length, len);
2594 bh_result->b_bdev = multi->stripes[0].dev->bdev;
2595 set_buffer_mapped(bh_result);
2598 free_extent_map(em);
2603 static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
2604 const struct iovec *iov, loff_t offset,
2605 unsigned long nr_segs)
2609 struct file *file = iocb->ki_filp;
2610 struct inode *inode = file->f_mapping->host;
2615 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2616 offset, nr_segs, btrfs_get_block, NULL);
2620 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
2622 return extent_bmap(mapping, iblock, btrfs_get_extent);
2625 int btrfs_readpage(struct file *file, struct page *page)
2627 struct extent_io_tree *tree;
2628 tree = &BTRFS_I(page->mapping->host)->io_tree;
2629 return extent_read_full_page(tree, page, btrfs_get_extent);
2632 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
2634 struct extent_io_tree *tree;
2637 if (current->flags & PF_MEMALLOC) {
2638 redirty_page_for_writepage(wbc, page);
2642 tree = &BTRFS_I(page->mapping->host)->io_tree;
2643 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
2646 static int btrfs_writepages(struct address_space *mapping,
2647 struct writeback_control *wbc)
2649 struct extent_io_tree *tree;
2650 tree = &BTRFS_I(mapping->host)->io_tree;
2651 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
2655 btrfs_readpages(struct file *file, struct address_space *mapping,
2656 struct list_head *pages, unsigned nr_pages)
2658 struct extent_io_tree *tree;
2659 tree = &BTRFS_I(mapping->host)->io_tree;
2660 return extent_readpages(tree, mapping, pages, nr_pages,
2663 static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
2665 struct extent_io_tree *tree;
2666 struct extent_map_tree *map;
2669 tree = &BTRFS_I(page->mapping->host)->io_tree;
2670 map = &BTRFS_I(page->mapping->host)->extent_tree;
2671 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
2673 ClearPagePrivate(page);
2674 set_page_private(page, 0);
2675 page_cache_release(page);
2680 static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
2682 return __btrfs_releasepage(page, gfp_flags);
2685 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
2687 struct extent_io_tree *tree;
2688 struct btrfs_ordered_extent *ordered;
2689 u64 page_start = page_offset(page);
2690 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
2692 wait_on_page_writeback(page);
2693 tree = &BTRFS_I(page->mapping->host)->io_tree;
2695 btrfs_releasepage(page, GFP_NOFS);
2699 lock_extent(tree, page_start, page_end, GFP_NOFS);
2700 ordered = btrfs_lookup_ordered_extent(page->mapping->host,
2704 * IO on this page will never be started, so we need
2705 * to account for any ordered extents now
2707 clear_extent_bit(tree, page_start, page_end,
2708 EXTENT_DIRTY | EXTENT_DELALLOC |
2709 EXTENT_LOCKED, 1, 0, GFP_NOFS);
2710 btrfs_finish_ordered_io(page->mapping->host,
2711 page_start, page_end);
2712 btrfs_put_ordered_extent(ordered);
2713 lock_extent(tree, page_start, page_end, GFP_NOFS);
2715 clear_extent_bit(tree, page_start, page_end,
2716 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
2719 __btrfs_releasepage(page, GFP_NOFS);
2721 ClearPageChecked(page);
2722 if (PagePrivate(page)) {
2723 ClearPagePrivate(page);
2724 set_page_private(page, 0);
2725 page_cache_release(page);
2730 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2731 * called from a page fault handler when a page is first dirtied. Hence we must
2732 * be careful to check for EOF conditions here. We set the page up correctly
2733 * for a written page which means we get ENOSPC checking when writing into
2734 * holes and correct delalloc and unwritten extent mapping on filesystems that
2735 * support these features.
2737 * We are not allowed to take the i_mutex here so we have to play games to
2738 * protect against truncate races as the page could now be beyond EOF. Because
2739 * vmtruncate() writes the inode size before removing pages, once we have the
2740 * page lock we can determine safely if the page is beyond EOF. If it is not
2741 * beyond EOF, then the page is guaranteed safe against truncation until we
2744 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
2746 struct inode *inode = fdentry(vma->vm_file)->d_inode;
2747 struct btrfs_root *root = BTRFS_I(inode)->root;
2748 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2749 struct btrfs_ordered_extent *ordered;
2751 unsigned long zero_start;
2757 ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
2764 size = i_size_read(inode);
2765 page_start = page_offset(page);
2766 page_end = page_start + PAGE_CACHE_SIZE - 1;
2768 if ((page->mapping != inode->i_mapping) ||
2769 (page_start >= size)) {
2770 /* page got truncated out from underneath us */
2773 wait_on_page_writeback(page);
2775 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
2776 set_page_extent_mapped(page);
2779 * we can't set the delalloc bits if there are pending ordered
2780 * extents. Drop our locks and wait for them to finish
2782 ordered = btrfs_lookup_ordered_extent(inode, page_start);
2784 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
2786 btrfs_start_ordered_extent(inode, ordered, 1);
2787 btrfs_put_ordered_extent(ordered);
2791 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
2792 page_end, GFP_NOFS);
2795 /* page is wholly or partially inside EOF */
2796 if (page_start + PAGE_CACHE_SIZE > size)
2797 zero_start = size & ~PAGE_CACHE_MASK;
2799 zero_start = PAGE_CACHE_SIZE;
2801 if (zero_start != PAGE_CACHE_SIZE) {
2803 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
2804 flush_dcache_page(page);
2807 ClearPageChecked(page);
2808 set_page_dirty(page);
2809 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
2817 static void btrfs_truncate(struct inode *inode)
2819 struct btrfs_root *root = BTRFS_I(inode)->root;
2821 struct btrfs_trans_handle *trans;
2823 u64 mask = root->sectorsize - 1;
2825 if (!S_ISREG(inode->i_mode))
2827 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
2830 btrfs_truncate_page(inode->i_mapping, inode->i_size);
2831 btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
2833 trans = btrfs_start_transaction(root, 1);
2834 btrfs_set_trans_block_group(trans, inode);
2835 btrfs_i_size_write(inode, inode->i_size);
2837 /* FIXME, add redo link to tree so we don't leak on crash */
2838 ret = btrfs_truncate_in_trans(trans, root, inode,
2839 BTRFS_EXTENT_DATA_KEY);
2840 btrfs_update_inode(trans, root, inode);
2841 nr = trans->blocks_used;
2843 ret = btrfs_end_transaction_throttle(trans, root);
2845 btrfs_btree_balance_dirty(root, nr);
2849 * Invalidate a single dcache entry at the root of the filesystem.
2850 * Needed after creation of snapshot or subvolume.
2852 void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name,
2855 struct dentry *alias, *entry;
2858 alias = d_find_alias(root->fs_info->sb->s_root->d_inode);
2862 /* change me if btrfs ever gets a d_hash operation */
2863 qstr.hash = full_name_hash(qstr.name, qstr.len);
2864 entry = d_lookup(alias, &qstr);
2867 d_invalidate(entry);
2873 int btrfs_create_subvol_root(struct btrfs_root *new_root,
2874 struct btrfs_trans_handle *trans, u64 new_dirid,
2875 struct btrfs_block_group_cache *block_group)
2877 struct inode *inode;
2880 inode = btrfs_new_inode(trans, new_root, "..", 2, new_dirid,
2881 new_dirid, block_group, S_IFDIR | 0700);
2883 return PTR_ERR(inode);
2884 inode->i_op = &btrfs_dir_inode_operations;
2885 inode->i_fop = &btrfs_dir_file_operations;
2886 new_root->inode = inode;
2888 ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid,
2891 btrfs_i_size_write(inode, 0);
2893 return btrfs_update_inode(trans, new_root, inode);
2896 unsigned long btrfs_force_ra(struct address_space *mapping,
2897 struct file_ra_state *ra, struct file *file,
2898 pgoff_t offset, pgoff_t last_index)
2900 pgoff_t req_size = last_index - offset + 1;
2902 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2903 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2906 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2907 return offset + req_size;
2911 struct inode *btrfs_alloc_inode(struct super_block *sb)
2913 struct btrfs_inode *ei;
2915 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2919 btrfs_ordered_inode_tree_init(&ei->ordered_tree);
2920 return &ei->vfs_inode;
2923 void btrfs_destroy_inode(struct inode *inode)
2925 struct btrfs_ordered_extent *ordered;
2926 WARN_ON(!list_empty(&inode->i_dentry));
2927 WARN_ON(inode->i_data.nrpages);
2930 ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
2934 printk("found ordered extent %Lu %Lu\n",
2935 ordered->file_offset, ordered->len);
2936 btrfs_remove_ordered_extent(inode, ordered);
2937 btrfs_put_ordered_extent(ordered);
2938 btrfs_put_ordered_extent(ordered);
2941 btrfs_drop_extent_cache(inode, 0, (u64)-1);
2942 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2945 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2946 static void init_once(struct kmem_cache * cachep, void *foo)
2948 static void init_once(void * foo, struct kmem_cache * cachep,
2949 unsigned long flags)
2952 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2954 inode_init_once(&ei->vfs_inode);
2957 void btrfs_destroy_cachep(void)
2959 if (btrfs_inode_cachep)
2960 kmem_cache_destroy(btrfs_inode_cachep);
2961 if (btrfs_trans_handle_cachep)
2962 kmem_cache_destroy(btrfs_trans_handle_cachep);
2963 if (btrfs_transaction_cachep)
2964 kmem_cache_destroy(btrfs_transaction_cachep);
2965 if (btrfs_bit_radix_cachep)
2966 kmem_cache_destroy(btrfs_bit_radix_cachep);
2967 if (btrfs_path_cachep)
2968 kmem_cache_destroy(btrfs_path_cachep);
2971 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2972 unsigned long extra_flags,
2973 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2974 void (*ctor)(struct kmem_cache *, void *)
2976 void (*ctor)(void *, struct kmem_cache *,
2981 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2982 SLAB_MEM_SPREAD | extra_flags), ctor
2983 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2989 int btrfs_init_cachep(void)
2991 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2992 sizeof(struct btrfs_inode),
2994 if (!btrfs_inode_cachep)
2996 btrfs_trans_handle_cachep =
2997 btrfs_cache_create("btrfs_trans_handle_cache",
2998 sizeof(struct btrfs_trans_handle),
3000 if (!btrfs_trans_handle_cachep)
3002 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
3003 sizeof(struct btrfs_transaction),
3005 if (!btrfs_transaction_cachep)
3007 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
3008 sizeof(struct btrfs_path),
3010 if (!btrfs_path_cachep)
3012 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
3013 SLAB_DESTROY_BY_RCU, NULL);
3014 if (!btrfs_bit_radix_cachep)
3018 btrfs_destroy_cachep();
3022 static int btrfs_getattr(struct vfsmount *mnt,
3023 struct dentry *dentry, struct kstat *stat)
3025 struct inode *inode = dentry->d_inode;
3026 generic_fillattr(inode, stat);
3027 stat->blksize = PAGE_CACHE_SIZE;
3028 stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9);
3032 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
3033 struct inode * new_dir,struct dentry *new_dentry)
3035 struct btrfs_trans_handle *trans;
3036 struct btrfs_root *root = BTRFS_I(old_dir)->root;
3037 struct inode *new_inode = new_dentry->d_inode;
3038 struct inode *old_inode = old_dentry->d_inode;
3039 struct timespec ctime = CURRENT_TIME;
3042 if (S_ISDIR(old_inode->i_mode) && new_inode &&
3043 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
3047 ret = btrfs_check_free_space(root, 1, 0);
3051 trans = btrfs_start_transaction(root, 1);
3053 btrfs_set_trans_block_group(trans, new_dir);
3055 old_dentry->d_inode->i_nlink++;
3056 old_dir->i_ctime = old_dir->i_mtime = ctime;
3057 new_dir->i_ctime = new_dir->i_mtime = ctime;
3058 old_inode->i_ctime = ctime;
3060 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
3065 new_inode->i_ctime = CURRENT_TIME;
3066 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
3070 ret = btrfs_add_link(trans, new_dentry, old_inode, 1);
3075 btrfs_end_transaction(trans, root);
3080 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
3081 const char *symname)
3083 struct btrfs_trans_handle *trans;
3084 struct btrfs_root *root = BTRFS_I(dir)->root;
3085 struct btrfs_path *path;
3086 struct btrfs_key key;
3087 struct inode *inode = NULL;
3094 struct btrfs_file_extent_item *ei;
3095 struct extent_buffer *leaf;
3096 unsigned long nr = 0;
3098 name_len = strlen(symname) + 1;
3099 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
3100 return -ENAMETOOLONG;
3102 err = btrfs_check_free_space(root, 1, 0);
3106 trans = btrfs_start_transaction(root, 1);
3107 btrfs_set_trans_block_group(trans, dir);
3109 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
3115 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
3117 dentry->d_parent->d_inode->i_ino, objectid,
3118 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
3119 err = PTR_ERR(inode);
3123 btrfs_set_trans_block_group(trans, inode);
3124 err = btrfs_add_nondir(trans, dentry, inode, 0);
3128 inode->i_mapping->a_ops = &btrfs_aops;
3129 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3130 inode->i_fop = &btrfs_file_operations;
3131 inode->i_op = &btrfs_file_inode_operations;
3132 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
3133 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
3134 inode->i_mapping, GFP_NOFS);
3135 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
3136 inode->i_mapping, GFP_NOFS);
3137 mutex_init(&BTRFS_I(inode)->csum_mutex);
3138 mutex_init(&BTRFS_I(inode)->extent_mutex);
3139 BTRFS_I(inode)->delalloc_bytes = 0;
3140 BTRFS_I(inode)->disk_i_size = 0;
3141 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
3142 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
3144 dir->i_sb->s_dirt = 1;
3145 btrfs_update_inode_block_group(trans, inode);
3146 btrfs_update_inode_block_group(trans, dir);
3150 path = btrfs_alloc_path();
3152 key.objectid = inode->i_ino;
3154 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
3155 datasize = btrfs_file_extent_calc_inline_size(name_len);
3156 err = btrfs_insert_empty_item(trans, root, path, &key,
3162 leaf = path->nodes[0];
3163 ei = btrfs_item_ptr(leaf, path->slots[0],
3164 struct btrfs_file_extent_item);
3165 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
3166 btrfs_set_file_extent_type(leaf, ei,
3167 BTRFS_FILE_EXTENT_INLINE);
3168 ptr = btrfs_file_extent_inline_start(ei);
3169 write_extent_buffer(leaf, symname, ptr, name_len);
3170 btrfs_mark_buffer_dirty(leaf);
3171 btrfs_free_path(path);
3173 inode->i_op = &btrfs_symlink_inode_operations;
3174 inode->i_mapping->a_ops = &btrfs_symlink_aops;
3175 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3176 btrfs_i_size_write(inode, name_len - 1);
3177 err = btrfs_update_inode(trans, root, inode);
3182 nr = trans->blocks_used;
3183 btrfs_end_transaction_throttle(trans, root);
3186 inode_dec_link_count(inode);
3189 btrfs_btree_balance_dirty(root, nr);
3193 static int btrfs_set_page_dirty(struct page *page)
3195 return __set_page_dirty_nobuffers(page);
3198 static int btrfs_permission(struct inode *inode, int mask,
3199 struct nameidata *nd)
3201 if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
3203 return generic_permission(inode, mask, NULL);
3206 static struct inode_operations btrfs_dir_inode_operations = {
3207 .lookup = btrfs_lookup,
3208 .create = btrfs_create,
3209 .unlink = btrfs_unlink,
3211 .mkdir = btrfs_mkdir,
3212 .rmdir = btrfs_rmdir,
3213 .rename = btrfs_rename,
3214 .symlink = btrfs_symlink,
3215 .setattr = btrfs_setattr,
3216 .mknod = btrfs_mknod,
3217 .setxattr = generic_setxattr,
3218 .getxattr = generic_getxattr,
3219 .listxattr = btrfs_listxattr,
3220 .removexattr = generic_removexattr,
3221 .permission = btrfs_permission,
3223 static struct inode_operations btrfs_dir_ro_inode_operations = {
3224 .lookup = btrfs_lookup,
3225 .permission = btrfs_permission,
3227 static struct file_operations btrfs_dir_file_operations = {
3228 .llseek = generic_file_llseek,
3229 .read = generic_read_dir,
3230 .readdir = btrfs_readdir,
3231 .unlocked_ioctl = btrfs_ioctl,
3232 #ifdef CONFIG_COMPAT
3233 .compat_ioctl = btrfs_ioctl,
3235 .release = btrfs_release_file,
3238 static struct extent_io_ops btrfs_extent_io_ops = {
3239 .fill_delalloc = run_delalloc_range,
3240 .submit_bio_hook = btrfs_submit_bio_hook,
3241 .merge_bio_hook = btrfs_merge_bio_hook,
3242 .readpage_io_hook = btrfs_readpage_io_hook,
3243 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
3244 .writepage_end_io_hook = btrfs_writepage_end_io_hook,
3245 .writepage_start_hook = btrfs_writepage_start_hook,
3246 .readpage_io_failed_hook = btrfs_io_failed_hook,
3247 .set_bit_hook = btrfs_set_bit_hook,
3248 .clear_bit_hook = btrfs_clear_bit_hook,
3251 static struct address_space_operations btrfs_aops = {
3252 .readpage = btrfs_readpage,
3253 .writepage = btrfs_writepage,
3254 .writepages = btrfs_writepages,
3255 .readpages = btrfs_readpages,
3256 .sync_page = block_sync_page,
3258 .direct_IO = btrfs_direct_IO,
3259 .invalidatepage = btrfs_invalidatepage,
3260 .releasepage = btrfs_releasepage,
3261 .set_page_dirty = btrfs_set_page_dirty,
3264 static struct address_space_operations btrfs_symlink_aops = {
3265 .readpage = btrfs_readpage,
3266 .writepage = btrfs_writepage,
3267 .invalidatepage = btrfs_invalidatepage,
3268 .releasepage = btrfs_releasepage,
3271 static struct inode_operations btrfs_file_inode_operations = {
3272 .truncate = btrfs_truncate,
3273 .getattr = btrfs_getattr,
3274 .setattr = btrfs_setattr,
3275 .setxattr = generic_setxattr,
3276 .getxattr = generic_getxattr,
3277 .listxattr = btrfs_listxattr,
3278 .removexattr = generic_removexattr,
3279 .permission = btrfs_permission,
3281 static struct inode_operations btrfs_special_inode_operations = {
3282 .getattr = btrfs_getattr,
3283 .setattr = btrfs_setattr,
3284 .permission = btrfs_permission,
3286 static struct inode_operations btrfs_symlink_inode_operations = {
3287 .readlink = generic_readlink,
3288 .follow_link = page_follow_link_light,
3289 .put_link = page_put_link,
3290 .permission = btrfs_permission,