projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Introduce btrfs_iget helper
[deliverable/linux.git] / fs / btrfs / inode.c
1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
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.
7 *
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.
12 *
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.
17 */
18
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
23 #include <linux/fs.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>
39 #include <linux/posix_acl.h>
40 #include "ctree.h"
41 #include "disk-io.h"
42 #include "transaction.h"
43 #include "btrfs_inode.h"
44 #include "ioctl.h"
45 #include "print-tree.h"
46 #include "volumes.h"
47 #include "ordered-data.h"
48
49 struct btrfs_iget_args {
50 u64 ino;
51 struct btrfs_root *root;
52 };
53
54 static struct inode_operations btrfs_dir_inode_operations;
55 static struct inode_operations btrfs_symlink_inode_operations;
56 static struct inode_operations btrfs_dir_ro_inode_operations;
57 static struct inode_operations btrfs_special_inode_operations;
58 static struct inode_operations btrfs_file_inode_operations;
59 static struct address_space_operations btrfs_aops;
60 static struct address_space_operations btrfs_symlink_aops;
61 static struct file_operations btrfs_dir_file_operations;
62 static struct extent_io_ops btrfs_extent_io_ops;
63
64 static struct kmem_cache *btrfs_inode_cachep;
65 struct kmem_cache *btrfs_trans_handle_cachep;
66 struct kmem_cache *btrfs_transaction_cachep;
67 struct kmem_cache *btrfs_bit_radix_cachep;
68 struct kmem_cache *btrfs_path_cachep;
69
70 #define S_SHIFT 12
71 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
72 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
73 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
74 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
75 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
76 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
77 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
78 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
79 };
80
81 static void btrfs_truncate(struct inode *inode);
82
83 int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
84 int for_del)
85 {
86 u64 total;
87 u64 used;
88 u64 thresh;
89 unsigned long flags;
90 int ret = 0;
91
92 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
93 total = btrfs_super_total_bytes(&root->fs_info->super_copy);
94 used = btrfs_super_bytes_used(&root->fs_info->super_copy);
95 if (for_del)
96 thresh = total * 90;
97 else
98 thresh = total * 85;
99
100 do_div(thresh, 100);
101
102 if (used + root->fs_info->delalloc_bytes + num_required > thresh)
103 ret = -ENOSPC;
104 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
105 return ret;
106 }
107
108 static int cow_file_range(struct inode *inode, u64 start, u64 end)
109 {
110 struct btrfs_root *root = BTRFS_I(inode)->root;
111 struct btrfs_trans_handle *trans;
112 u64 alloc_hint = 0;
113 u64 num_bytes;
114 u64 cur_alloc_size;
115 u64 blocksize = root->sectorsize;
116 u64 orig_num_bytes;
117 struct btrfs_key ins;
118 struct extent_map *em;
119 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
120 int ret = 0;
121
122 trans = btrfs_join_transaction(root, 1);
123 BUG_ON(!trans);
124 btrfs_set_trans_block_group(trans, inode);
125
126 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
127 num_bytes = max(blocksize, num_bytes);
128 orig_num_bytes = num_bytes;
129
130 if (alloc_hint == EXTENT_MAP_INLINE)
131 goto out;
132
133 BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy));
134 mutex_lock(&BTRFS_I(inode)->extent_mutex);
135 btrfs_drop_extent_cache(inode, start, start + num_bytes - 1);
136 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
137
138 while(num_bytes > 0) {
139 cur_alloc_size = min(num_bytes, root->fs_info->max_extent);
140 ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
141 root->sectorsize, 0, 0,
142 (u64)-1, &ins, 1);
143 if (ret) {
144 WARN_ON(1);
145 goto out;
146 }
147 em = alloc_extent_map(GFP_NOFS);
148 em->start = start;
149 em->len = ins.offset;
150 em->block_start = ins.objectid;
151 em->bdev = root->fs_info->fs_devices->latest_bdev;
152 mutex_lock(&BTRFS_I(inode)->extent_mutex);
153 set_bit(EXTENT_FLAG_PINNED, &em->flags);
154 while(1) {
155 spin_lock(&em_tree->lock);
156 ret = add_extent_mapping(em_tree, em);
157 spin_unlock(&em_tree->lock);
158 if (ret != -EEXIST) {
159 free_extent_map(em);
160 break;
161 }
162 btrfs_drop_extent_cache(inode, start,
163 start + ins.offset - 1);
164 }
165 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
166
167 cur_alloc_size = ins.offset;
168 ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
169 ins.offset, 0);
170 BUG_ON(ret);
171 if (num_bytes < cur_alloc_size) {
172 printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes,
173 cur_alloc_size);
174 break;
175 }
176 num_bytes -= cur_alloc_size;
177 alloc_hint = ins.objectid + ins.offset;
178 start += cur_alloc_size;
179 }
180 out:
181 btrfs_end_transaction(trans, root);
182 return ret;
183 }
184
185 static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end)
186 {
187 u64 extent_start;
188 u64 extent_end;
189 u64 bytenr;
190 u64 loops = 0;
191 u64 total_fs_bytes;
192 struct btrfs_root *root = BTRFS_I(inode)->root;
193 struct btrfs_block_group_cache *block_group;
194 struct btrfs_trans_handle *trans;
195 struct extent_buffer *leaf;
196 int found_type;
197 struct btrfs_path *path;
198 struct btrfs_file_extent_item *item;
199 int ret;
200 int err = 0;
201 struct btrfs_key found_key;
202
203 total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
204 path = btrfs_alloc_path();
205 BUG_ON(!path);
206 trans = btrfs_join_transaction(root, 1);
207 BUG_ON(!trans);
208 again:
209 ret = btrfs_lookup_file_extent(NULL, root, path,
210 inode->i_ino, start, 0);
211 if (ret < 0) {
212 err = ret;
213 goto out;
214 }
215
216 if (ret != 0) {
217 if (path->slots[0] == 0)
218 goto not_found;
219 path->slots[0]--;
220 }
221
222 leaf = path->nodes[0];
223 item = btrfs_item_ptr(leaf, path->slots[0],
224 struct btrfs_file_extent_item);
225
226 /* are we inside the extent that was found? */
227 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
228 found_type = btrfs_key_type(&found_key);
229 if (found_key.objectid != inode->i_ino ||
230 found_type != BTRFS_EXTENT_DATA_KEY)
231 goto not_found;
232
233 found_type = btrfs_file_extent_type(leaf, item);
234 extent_start = found_key.offset;
235 if (found_type == BTRFS_FILE_EXTENT_REG) {
236 u64 extent_num_bytes;
237
238 extent_num_bytes = btrfs_file_extent_num_bytes(leaf, item);
239 extent_end = extent_start + extent_num_bytes;
240 err = 0;
241
242 if (loops && start != extent_start)
243 goto not_found;
244
245 if (start < extent_start || start >= extent_end)
246 goto not_found;
247
248 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
249 if (bytenr == 0)
250 goto not_found;
251
252 if (btrfs_cross_ref_exists(trans, root, &found_key, bytenr))
253 goto not_found;
254 /*
255 * we may be called by the resizer, make sure we're inside
256 * the limits of the FS
257 */
258 block_group = btrfs_lookup_block_group(root->fs_info,
259 bytenr);
260 if (!block_group || block_group->ro)
261 goto not_found;
262
263 bytenr += btrfs_file_extent_offset(leaf, item);
264 extent_num_bytes = min(end + 1, extent_end) - start;
265 ret = btrfs_add_ordered_extent(inode, start, bytenr,
266 extent_num_bytes, 1);
267 if (ret) {
268 err = ret;
269 goto out;
270 }
271
272 btrfs_release_path(root, path);
273 start = extent_end;
274 if (start <= end) {
275 loops++;
276 goto again;
277 }
278 } else {
279 not_found:
280 btrfs_end_transaction(trans, root);
281 btrfs_free_path(path);
282 return cow_file_range(inode, start, end);
283 }
284 out:
285 WARN_ON(err);
286 btrfs_end_transaction(trans, root);
287 btrfs_free_path(path);
288 return err;
289 }
290
291 static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
292 {
293 struct btrfs_root *root = BTRFS_I(inode)->root;
294 int ret;
295
296 if (btrfs_test_opt(root, NODATACOW) ||
297 btrfs_test_flag(inode, NODATACOW))
298 ret = run_delalloc_nocow(inode, start, end);
299 else
300 ret = cow_file_range(inode, start, end);
301
302 return ret;
303 }
304
305 int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
306 unsigned long old, unsigned long bits)
307 {
308 unsigned long flags;
309 if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
310 struct btrfs_root *root = BTRFS_I(inode)->root;
311 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
312 BTRFS_I(inode)->delalloc_bytes += end - start + 1;
313 root->fs_info->delalloc_bytes += end - start + 1;
314 if (list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
315 list_add_tail(&BTRFS_I(inode)->delalloc_inodes,
316 &root->fs_info->delalloc_inodes);
317 }
318 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
319 }
320 return 0;
321 }
322
323 int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end,
324 unsigned long old, unsigned long bits)
325 {
326 if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
327 struct btrfs_root *root = BTRFS_I(inode)->root;
328 unsigned long flags;
329
330 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
331 if (end - start + 1 > root->fs_info->delalloc_bytes) {
332 printk("warning: delalloc account %Lu %Lu\n",
333 end - start + 1, root->fs_info->delalloc_bytes);
334 root->fs_info->delalloc_bytes = 0;
335 BTRFS_I(inode)->delalloc_bytes = 0;
336 } else {
337 root->fs_info->delalloc_bytes -= end - start + 1;
338 BTRFS_I(inode)->delalloc_bytes -= end - start + 1;
339 }
340 if (BTRFS_I(inode)->delalloc_bytes == 0 &&
341 !list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
342 list_del_init(&BTRFS_I(inode)->delalloc_inodes);
343 }
344 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
345 }
346 return 0;
347 }
348
349 int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
350 size_t size, struct bio *bio)
351 {
352 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
353 struct btrfs_mapping_tree *map_tree;
354 u64 logical = bio->bi_sector << 9;
355 u64 length = 0;
356 u64 map_length;
357 int ret;
358
359 length = bio->bi_size;
360 map_tree = &root->fs_info->mapping_tree;
361 map_length = length;
362 ret = btrfs_map_block(map_tree, READ, logical,
363 &map_length, NULL, 0);
364
365 if (map_length < length + size) {
366 return 1;
367 }
368 return 0;
369 }
370
371 int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
372 int mirror_num)
373 {
374 struct btrfs_root *root = BTRFS_I(inode)->root;
375 int ret = 0;
376
377 ret = btrfs_csum_one_bio(root, inode, bio);
378 BUG_ON(ret);
379
380 return btrfs_map_bio(root, rw, bio, mirror_num, 1);
381 }
382
383 int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
384 int mirror_num)
385 {
386 struct btrfs_root *root = BTRFS_I(inode)->root;
387 int ret = 0;
388
389 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
390 BUG_ON(ret);
391
392 if (btrfs_test_opt(root, NODATASUM) ||
393 btrfs_test_flag(inode, NODATASUM)) {
394 goto mapit;
395 }
396
397 if (!(rw & (1 << BIO_RW))) {
398 btrfs_lookup_bio_sums(root, inode, bio);
399 goto mapit;
400 }
401 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
402 inode, rw, bio, mirror_num,
403 __btrfs_submit_bio_hook);
404 mapit:
405 return btrfs_map_bio(root, rw, bio, mirror_num, 0);
406 }
407
408 static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
409 struct inode *inode, u64 file_offset,
410 struct list_head *list)
411 {
412 struct list_head *cur;
413 struct btrfs_ordered_sum *sum;
414
415 btrfs_set_trans_block_group(trans, inode);
416 list_for_each(cur, list) {
417 sum = list_entry(cur, struct btrfs_ordered_sum, list);
418 btrfs_csum_file_blocks(trans, BTRFS_I(inode)->root,
419 inode, sum);
420 }
421 return 0;
422 }
423
424 int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end)
425 {
426 return set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end,
427 GFP_NOFS);
428 }
429
430 struct btrfs_writepage_fixup {
431 struct page *page;
432 struct btrfs_work work;
433 };
434
435 /* see btrfs_writepage_start_hook for details on why this is required */
436 void btrfs_writepage_fixup_worker(struct btrfs_work *work)
437 {
438 struct btrfs_writepage_fixup *fixup;
439 struct btrfs_ordered_extent *ordered;
440 struct page *page;
441 struct inode *inode;
442 u64 page_start;
443 u64 page_end;
444
445 fixup = container_of(work, struct btrfs_writepage_fixup, work);
446 page = fixup->page;
447 again:
448 lock_page(page);
449 if (!page->mapping || !PageDirty(page) || !PageChecked(page)) {
450 ClearPageChecked(page);
451 goto out_page;
452 }
453
454 inode = page->mapping->host;
455 page_start = page_offset(page);
456 page_end = page_offset(page) + PAGE_CACHE_SIZE - 1;
457
458 lock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
459
460 /* already ordered? We're done */
461 if (test_range_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
462 EXTENT_ORDERED, 0)) {
463 goto out;
464 }
465
466 ordered = btrfs_lookup_ordered_extent(inode, page_start);
467 if (ordered) {
468 unlock_extent(&BTRFS_I(inode)->io_tree, page_start,
469 page_end, GFP_NOFS);
470 unlock_page(page);
471 btrfs_start_ordered_extent(inode, ordered, 1);
472 goto again;
473 }
474
475 btrfs_set_extent_delalloc(inode, page_start, page_end);
476 ClearPageChecked(page);
477 out:
478 unlock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
479 out_page:
480 unlock_page(page);
481 page_cache_release(page);
482 }
483
484 /*
485 * There are a few paths in the higher layers of the kernel that directly
486 * set the page dirty bit without asking the filesystem if it is a
487 * good idea. This causes problems because we want to make sure COW
488 * properly happens and the data=ordered rules are followed.
489 *
490 * In our case any range that doesn't have the EXTENT_ORDERED bit set
491 * hasn't been properly setup for IO. We kick off an async process
492 * to fix it up. The async helper will wait for ordered extents, set
493 * the delalloc bit and make it safe to write the page.
494 */
495 int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
496 {
497 struct inode *inode = page->mapping->host;
498 struct btrfs_writepage_fixup *fixup;
499 struct btrfs_root *root = BTRFS_I(inode)->root;
500 int ret;
501
502 ret = test_range_bit(&BTRFS_I(inode)->io_tree, start, end,
503 EXTENT_ORDERED, 0);
504 if (ret)
505 return 0;
506
507 if (PageChecked(page))
508 return -EAGAIN;
509
510 fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
511 if (!fixup)
512 return -EAGAIN;
513
514 SetPageChecked(page);
515 page_cache_get(page);
516 fixup->work.func = btrfs_writepage_fixup_worker;
517 fixup->page = page;
518 btrfs_queue_worker(&root->fs_info->fixup_workers, &fixup->work);
519 return -EAGAIN;
520 }
521
522 static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
523 {
524 struct btrfs_root *root = BTRFS_I(inode)->root;
525 struct btrfs_trans_handle *trans;
526 struct btrfs_ordered_extent *ordered_extent;
527 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
528 u64 alloc_hint = 0;
529 struct list_head list;
530 struct btrfs_key ins;
531 int ret;
532
533 ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1);
534 if (!ret)
535 return 0;
536
537 trans = btrfs_join_transaction(root, 1);
538
539 ordered_extent = btrfs_lookup_ordered_extent(inode, start);
540 BUG_ON(!ordered_extent);
541 if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags))
542 goto nocow;
543
544 lock_extent(io_tree, ordered_extent->file_offset,
545 ordered_extent->file_offset + ordered_extent->len - 1,
546 GFP_NOFS);
547
548 INIT_LIST_HEAD(&list);
549
550 ins.objectid = ordered_extent->start;
551 ins.offset = ordered_extent->len;
552 ins.type = BTRFS_EXTENT_ITEM_KEY;
553
554 ret = btrfs_alloc_reserved_extent(trans, root, root->root_key.objectid,
555 trans->transid, inode->i_ino,
556 ordered_extent->file_offset, &ins);
557 BUG_ON(ret);
558
559 mutex_lock(&BTRFS_I(inode)->extent_mutex);
560
561 ret = btrfs_drop_extents(trans, root, inode,
562 ordered_extent->file_offset,
563 ordered_extent->file_offset +
564 ordered_extent->len,
565 ordered_extent->file_offset, &alloc_hint);
566 BUG_ON(ret);
567 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
568 ordered_extent->file_offset,
569 ordered_extent->start,
570 ordered_extent->len,
571 ordered_extent->len, 0);
572 BUG_ON(ret);
573
574 btrfs_drop_extent_cache(inode, ordered_extent->file_offset,
575 ordered_extent->file_offset +
576 ordered_extent->len - 1);
577 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
578
579 inode->i_blocks += ordered_extent->len >> 9;
580 unlock_extent(io_tree, ordered_extent->file_offset,
581 ordered_extent->file_offset + ordered_extent->len - 1,
582 GFP_NOFS);
583 nocow:
584 add_pending_csums(trans, inode, ordered_extent->file_offset,
585 &ordered_extent->list);
586
587 btrfs_ordered_update_i_size(inode, ordered_extent);
588 btrfs_remove_ordered_extent(inode, ordered_extent);
589
590 /* once for us */
591 btrfs_put_ordered_extent(ordered_extent);
592 /* once for the tree */
593 btrfs_put_ordered_extent(ordered_extent);
594
595 btrfs_update_inode(trans, root, inode);
596 btrfs_end_transaction(trans, root);
597 return 0;
598 }
599
600 int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
601 struct extent_state *state, int uptodate)
602 {
603 return btrfs_finish_ordered_io(page->mapping->host, start, end);
604 }
605
606 struct io_failure_record {
607 struct page *page;
608 u64 start;
609 u64 len;
610 u64 logical;
611 int last_mirror;
612 };
613
614 int btrfs_io_failed_hook(struct bio *failed_bio,
615 struct page *page, u64 start, u64 end,
616 struct extent_state *state)
617 {
618 struct io_failure_record *failrec = NULL;
619 u64 private;
620 struct extent_map *em;
621 struct inode *inode = page->mapping->host;
622 struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
623 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
624 struct bio *bio;
625 int num_copies;
626 int ret;
627 int rw;
628 u64 logical;
629
630 ret = get_state_private(failure_tree, start, &private);
631 if (ret) {
632 failrec = kmalloc(sizeof(*failrec), GFP_NOFS);
633 if (!failrec)
634 return -ENOMEM;
635 failrec->start = start;
636 failrec->len = end - start + 1;
637 failrec->last_mirror = 0;
638
639 spin_lock(&em_tree->lock);
640 em = lookup_extent_mapping(em_tree, start, failrec->len);
641 if (em->start > start || em->start + em->len < start) {
642 free_extent_map(em);
643 em = NULL;
644 }
645 spin_unlock(&em_tree->lock);
646
647 if (!em || IS_ERR(em)) {
648 kfree(failrec);
649 return -EIO;
650 }
651 logical = start - em->start;
652 logical = em->block_start + logical;
653 failrec->logical = logical;
654 free_extent_map(em);
655 set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
656 EXTENT_DIRTY, GFP_NOFS);
657 set_state_private(failure_tree, start,
658 (u64)(unsigned long)failrec);
659 } else {
660 failrec = (struct io_failure_record *)(unsigned long)private;
661 }
662 num_copies = btrfs_num_copies(
663 &BTRFS_I(inode)->root->fs_info->mapping_tree,
664 failrec->logical, failrec->len);
665 failrec->last_mirror++;
666 if (!state) {
667 spin_lock_irq(&BTRFS_I(inode)->io_tree.lock);
668 state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
669 failrec->start,
670 EXTENT_LOCKED);
671 if (state && state->start != failrec->start)
672 state = NULL;
673 spin_unlock_irq(&BTRFS_I(inode)->io_tree.lock);
674 }
675 if (!state || failrec->last_mirror > num_copies) {
676 set_state_private(failure_tree, failrec->start, 0);
677 clear_extent_bits(failure_tree, failrec->start,
678 failrec->start + failrec->len - 1,
679 EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
680 kfree(failrec);
681 return -EIO;
682 }
683 bio = bio_alloc(GFP_NOFS, 1);
684 bio->bi_private = state;
685 bio->bi_end_io = failed_bio->bi_end_io;
686 bio->bi_sector = failrec->logical >> 9;
687 bio->bi_bdev = failed_bio->bi_bdev;
688 bio->bi_size = 0;
689 bio_add_page(bio, page, failrec->len, start - page_offset(page));
690 if (failed_bio->bi_rw & (1 << BIO_RW))
691 rw = WRITE;
692 else
693 rw = READ;
694
695 BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
696 failrec->last_mirror);
697 return 0;
698 }
699
700 int btrfs_clean_io_failures(struct inode *inode, u64 start)
701 {
702 u64 private;
703 u64 private_failure;
704 struct io_failure_record *failure;
705 int ret;
706
707 private = 0;
708 if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
709 (u64)-1, 1, EXTENT_DIRTY)) {
710 ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
711 start, &private_failure);
712 if (ret == 0) {
713 failure = (struct io_failure_record *)(unsigned long)
714 private_failure;
715 set_state_private(&BTRFS_I(inode)->io_failure_tree,
716 failure->start, 0);
717 clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
718 failure->start,
719 failure->start + failure->len - 1,
720 EXTENT_DIRTY | EXTENT_LOCKED,
721 GFP_NOFS);
722 kfree(failure);
723 }
724 }
725 return 0;
726 }
727
728 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
729 struct extent_state *state)
730 {
731 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
732 struct inode *inode = page->mapping->host;
733 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
734 char *kaddr;
735 u64 private = ~(u32)0;
736 int ret;
737 struct btrfs_root *root = BTRFS_I(inode)->root;
738 u32 csum = ~(u32)0;
739 unsigned long flags;
740
741 if (btrfs_test_opt(root, NODATASUM) ||
742 btrfs_test_flag(inode, NODATASUM))
743 return 0;
744 if (state && state->start == start) {
745 private = state->private;
746 ret = 0;
747 } else {
748 ret = get_state_private(io_tree, start, &private);
749 }
750 local_irq_save(flags);
751 kaddr = kmap_atomic(page, KM_IRQ0);
752 if (ret) {
753 goto zeroit;
754 }
755 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
756 btrfs_csum_final(csum, (char *)&csum);
757 if (csum != private) {
758 goto zeroit;
759 }
760 kunmap_atomic(kaddr, KM_IRQ0);
761 local_irq_restore(flags);
762
763 /* if the io failure tree for this inode is non-empty,
764 * check to see if we've recovered from a failed IO
765 */
766 btrfs_clean_io_failures(inode, start);
767 return 0;
768
769 zeroit:
770 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
771 page->mapping->host->i_ino, (unsigned long long)start, csum,
772 private);
773 memset(kaddr + offset, 1, end - start + 1);
774 flush_dcache_page(page);
775 kunmap_atomic(kaddr, KM_IRQ0);
776 local_irq_restore(flags);
777 if (private == 0)
778 return 0;
779 return -EIO;
780 }
781
782 /*
783 * This creates an orphan entry for the given inode in case something goes
784 * wrong in the middle of an unlink/truncate.
785 */
786 int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode)
787 {
788 struct btrfs_root *root = BTRFS_I(inode)->root;
789 int ret = 0;
790
791 spin_lock(&root->list_lock);
792
793 /* already on the orphan list, we're good */
794 if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
795 spin_unlock(&root->list_lock);
796 return 0;
797 }
798
799 list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);
800
801 spin_unlock(&root->list_lock);
802
803 /*
804 * insert an orphan item to track this unlinked/truncated file
805 */
806 ret = btrfs_insert_orphan_item(trans, root, inode->i_ino);
807
808 return ret;
809 }
810
811 /*
812 * We have done the truncate/delete so we can go ahead and remove the orphan
813 * item for this particular inode.
814 */
815 int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode)
816 {
817 struct btrfs_root *root = BTRFS_I(inode)->root;
818 int ret = 0;
819
820 spin_lock(&root->list_lock);
821
822 if (list_empty(&BTRFS_I(inode)->i_orphan)) {
823 spin_unlock(&root->list_lock);
824 return 0;
825 }
826
827 list_del_init(&BTRFS_I(inode)->i_orphan);
828 if (!trans) {
829 spin_unlock(&root->list_lock);
830 return 0;
831 }
832
833 spin_unlock(&root->list_lock);
834
835 ret = btrfs_del_orphan_item(trans, root, inode->i_ino);
836
837 return ret;
838 }
839
840 /*
841 * this cleans up any orphans that may be left on the list from the last use
842 * of this root.
843 */
844 void btrfs_orphan_cleanup(struct btrfs_root *root)
845 {
846 struct btrfs_path *path;
847 struct extent_buffer *leaf;
848 struct btrfs_item *item;
849 struct btrfs_key key, found_key;
850 struct btrfs_trans_handle *trans;
851 struct inode *inode;
852 int ret = 0, nr_unlink = 0, nr_truncate = 0;
853
854 /* don't do orphan cleanup if the fs is readonly. */
855 if (root->inode->i_sb->s_flags & MS_RDONLY)
856 return;
857
858 path = btrfs_alloc_path();
859 if (!path)
860 return;
861 path->reada = -1;
862
863 key.objectid = BTRFS_ORPHAN_OBJECTID;
864 btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY);
865 key.offset = (u64)-1;
866
867 trans = btrfs_start_transaction(root, 1);
868 btrfs_set_trans_block_group(trans, root->inode);
869
870 while (1) {
871 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
872 if (ret < 0) {
873 printk(KERN_ERR "Error searching slot for orphan: %d"
874 "\n", ret);
875 break;
876 }
877
878 /*
879 * if ret == 0 means we found what we were searching for, which
880 * is weird, but possible, so only screw with path if we didnt
881 * find the key and see if we have stuff that matches
882 */
883 if (ret > 0) {
884 if (path->slots[0] == 0)
885 break;
886 path->slots[0]--;
887 }
888
889 /* pull out the item */
890 leaf = path->nodes[0];
891 item = btrfs_item_nr(leaf, path->slots[0]);
892 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
893
894 /* make sure the item matches what we want */
895 if (found_key.objectid != BTRFS_ORPHAN_OBJECTID)
896 break;
897 if (btrfs_key_type(&found_key) != BTRFS_ORPHAN_ITEM_KEY)
898 break;
899
900 /* release the path since we're done with it */
901 btrfs_release_path(root, path);
902
903 /*
904 * this is where we are basically btrfs_lookup, without the
905 * crossing root thing. we store the inode number in the
906 * offset of the orphan item.
907 */
908 inode = btrfs_iget_locked(root->inode->i_sb,
909 found_key.offset, root);
910 if (!inode)
911 break;
912
913 if (inode->i_state & I_NEW) {
914 BTRFS_I(inode)->root = root;
915
916 /* have to set the location manually */
917 BTRFS_I(inode)->location.objectid = inode->i_ino;
918 BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
919 BTRFS_I(inode)->location.offset = 0;
920
921 btrfs_read_locked_inode(inode);
922 unlock_new_inode(inode);
923 }
924
925 /*
926 * add this inode to the orphan list so btrfs_orphan_del does
927 * the proper thing when we hit it
928 */
929 spin_lock(&root->list_lock);
930 list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);
931 spin_unlock(&root->list_lock);
932
933 /*
934 * if this is a bad inode, means we actually succeeded in
935 * removing the inode, but not the orphan record, which means
936 * we need to manually delete the orphan since iput will just
937 * do a destroy_inode
938 */
939 if (is_bad_inode(inode)) {
940 btrfs_orphan_del(trans, inode);
941 iput(inode);
942 continue;
943 }
944
945 /* if we have links, this was a truncate, lets do that */
946 if (inode->i_nlink) {
947 nr_truncate++;
948 btrfs_truncate(inode);
949 } else {
950 nr_unlink++;
951 }
952
953 /* this will do delete_inode and everything for us */
954 iput(inode);
955 }
956
957 if (nr_unlink)
958 printk(KERN_INFO "btrfs: unlinked %d orphans\n", nr_unlink);
959 if (nr_truncate)
960 printk(KERN_INFO "btrfs: truncated %d orphans\n", nr_truncate);
961
962 btrfs_free_path(path);
963 btrfs_end_transaction(trans, root);
964 }
965
966 void btrfs_read_locked_inode(struct inode *inode)
967 {
968 struct btrfs_path *path;
969 struct extent_buffer *leaf;
970 struct btrfs_inode_item *inode_item;
971 struct btrfs_timespec *tspec;
972 struct btrfs_root *root = BTRFS_I(inode)->root;
973 struct btrfs_key location;
974 u64 alloc_group_block;
975 u32 rdev;
976 int ret;
977
978 path = btrfs_alloc_path();
979 BUG_ON(!path);
980 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
981
982 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
983 if (ret)
984 goto make_bad;
985
986 leaf = path->nodes[0];
987 inode_item = btrfs_item_ptr(leaf, path->slots[0],
988 struct btrfs_inode_item);
989
990 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
991 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
992 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
993 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
994 btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item));
995
996 tspec = btrfs_inode_atime(inode_item);
997 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
998 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
999
1000 tspec = btrfs_inode_mtime(inode_item);
1001 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
1002 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
1003
1004 tspec = btrfs_inode_ctime(inode_item);
1005 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
1006 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
1007
1008 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
1009 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
1010 inode->i_rdev = 0;
1011 rdev = btrfs_inode_rdev(leaf, inode_item);
1012
1013 BTRFS_I(inode)->index_cnt = (u64)-1;
1014
1015 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
1016 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
1017 alloc_group_block);
1018 BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
1019 if (!BTRFS_I(inode)->block_group) {
1020 BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
1021 NULL, 0,
1022 BTRFS_BLOCK_GROUP_METADATA, 0);
1023 }
1024 btrfs_free_path(path);
1025 inode_item = NULL;
1026
1027 switch (inode->i_mode & S_IFMT) {
1028 case S_IFREG:
1029 inode->i_mapping->a_ops = &btrfs_aops;
1030 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
1031 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
1032 inode->i_fop = &btrfs_file_operations;
1033 inode->i_op = &btrfs_file_inode_operations;
1034 break;
1035 case S_IFDIR:
1036 inode->i_fop = &btrfs_dir_file_operations;
1037 if (root == root->fs_info->tree_root)
1038 inode->i_op = &btrfs_dir_ro_inode_operations;
1039 else
1040 inode->i_op = &btrfs_dir_inode_operations;
1041 break;
1042 case S_IFLNK:
1043 inode->i_op = &btrfs_symlink_inode_operations;
1044 inode->i_mapping->a_ops = &btrfs_symlink_aops;
1045 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
1046 break;
1047 default:
1048 init_special_inode(inode, inode->i_mode, rdev);
1049 break;
1050 }
1051 return;
1052
1053 make_bad:
1054 btrfs_free_path(path);
1055 make_bad_inode(inode);
1056 }
1057
1058 static void fill_inode_item(struct extent_buffer *leaf,
1059 struct btrfs_inode_item *item,
1060 struct inode *inode)
1061 {
1062 btrfs_set_inode_uid(leaf, item, inode->i_uid);
1063 btrfs_set_inode_gid(leaf, item, inode->i_gid);
1064 btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size);
1065 btrfs_set_inode_mode(leaf, item, inode->i_mode);
1066 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
1067
1068 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
1069 inode->i_atime.tv_sec);
1070 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
1071 inode->i_atime.tv_nsec);
1072
1073 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
1074 inode->i_mtime.tv_sec);
1075 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
1076 inode->i_mtime.tv_nsec);
1077
1078 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
1079 inode->i_ctime.tv_sec);
1080 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
1081 inode->i_ctime.tv_nsec);
1082
1083 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
1084 btrfs_set_inode_generation(leaf, item, inode->i_generation);
1085 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
1086 btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
1087 btrfs_set_inode_block_group(leaf, item,
1088 BTRFS_I(inode)->block_group->key.objectid);
1089 }
1090
1091 int noinline btrfs_update_inode(struct btrfs_trans_handle *trans,
1092 struct btrfs_root *root,
1093 struct inode *inode)
1094 {
1095 struct btrfs_inode_item *inode_item;
1096 struct btrfs_path *path;
1097 struct extent_buffer *leaf;
1098 int ret;
1099
1100 path = btrfs_alloc_path();
1101 BUG_ON(!path);
1102 ret = btrfs_lookup_inode(trans, root, path,
1103 &BTRFS_I(inode)->location, 1);
1104 if (ret) {
1105 if (ret > 0)
1106 ret = -ENOENT;
1107 goto failed;
1108 }
1109
1110 leaf = path->nodes[0];
1111 inode_item = btrfs_item_ptr(leaf, path->slots[0],
1112 struct btrfs_inode_item);
1113
1114 fill_inode_item(leaf, inode_item, inode);
1115 btrfs_mark_buffer_dirty(leaf);
1116 btrfs_set_inode_last_trans(trans, inode);
1117 ret = 0;
1118 failed:
1119 btrfs_free_path(path);
1120 return ret;
1121 }
1122
1123
1124 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
1125 struct btrfs_root *root,
1126 struct inode *dir,
1127 struct dentry *dentry)
1128 {
1129 struct btrfs_path *path;
1130 const char *name = dentry->d_name.name;
1131 int name_len = dentry->d_name.len;
1132 int ret = 0;
1133 struct extent_buffer *leaf;
1134 struct btrfs_dir_item *di;
1135 struct btrfs_key key;
1136 u64 index;
1137
1138 path = btrfs_alloc_path();
1139 if (!path) {
1140 ret = -ENOMEM;
1141 goto err;
1142 }
1143
1144 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
1145 name, name_len, -1);
1146 if (IS_ERR(di)) {
1147 ret = PTR_ERR(di);
1148 goto err;
1149 }
1150 if (!di) {
1151 ret = -ENOENT;
1152 goto err;
1153 }
1154 leaf = path->nodes[0];
1155 btrfs_dir_item_key_to_cpu(leaf, di, &key);
1156 ret = btrfs_delete_one_dir_name(trans, root, path, di);
1157 if (ret)
1158 goto err;
1159 btrfs_release_path(root, path);
1160
1161 ret = btrfs_del_inode_ref(trans, root, name, name_len,
1162 dentry->d_inode->i_ino,
1163 dentry->d_parent->d_inode->i_ino, &index);
1164 if (ret) {
1165 printk("failed to delete reference to %.*s, "
1166 "inode %lu parent %lu\n", name_len, name,
1167 dentry->d_inode->i_ino,
1168 dentry->d_parent->d_inode->i_ino);
1169 goto err;
1170 }
1171
1172 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
1173 index, name, name_len, -1);
1174 if (IS_ERR(di)) {
1175 ret = PTR_ERR(di);
1176 goto err;
1177 }
1178 if (!di) {
1179 ret = -ENOENT;
1180 goto err;
1181 }
1182 ret = btrfs_delete_one_dir_name(trans, root, path, di);
1183 btrfs_release_path(root, path);
1184
1185 dentry->d_inode->i_ctime = dir->i_ctime;
1186 err:
1187 btrfs_free_path(path);
1188 if (!ret) {
1189 btrfs_i_size_write(dir, dir->i_size - name_len * 2);
1190 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
1191 btrfs_update_inode(trans, root, dir);
1192 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1193 dentry->d_inode->i_nlink--;
1194 #else
1195 drop_nlink(dentry->d_inode);
1196 #endif
1197 ret = btrfs_update_inode(trans, root, dentry->d_inode);
1198 dir->i_sb->s_dirt = 1;
1199 }
1200 return ret;
1201 }
1202
1203 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
1204 {
1205 struct btrfs_root *root;
1206 struct btrfs_trans_handle *trans;
1207 struct inode *inode = dentry->d_inode;
1208 int ret;
1209 unsigned long nr = 0;
1210
1211 root = BTRFS_I(dir)->root;
1212
1213 ret = btrfs_check_free_space(root, 1, 1);
1214 if (ret)
1215 goto fail;
1216
1217 trans = btrfs_start_transaction(root, 1);
1218
1219 btrfs_set_trans_block_group(trans, dir);
1220 ret = btrfs_unlink_trans(trans, root, dir, dentry);
1221
1222 if (inode->i_nlink == 0)
1223 ret = btrfs_orphan_add(trans, inode);
1224
1225 nr = trans->blocks_used;
1226
1227 btrfs_end_transaction_throttle(trans, root);
1228 fail:
1229 btrfs_btree_balance_dirty(root, nr);
1230 return ret;
1231 }
1232
1233 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
1234 {
1235 struct inode *inode = dentry->d_inode;
1236 int err = 0;
1237 int ret;
1238 struct btrfs_root *root = BTRFS_I(dir)->root;
1239 struct btrfs_trans_handle *trans;
1240 unsigned long nr = 0;
1241
1242 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
1243 return -ENOTEMPTY;
1244 }
1245
1246 ret = btrfs_check_free_space(root, 1, 1);
1247 if (ret)
1248 goto fail;
1249
1250 trans = btrfs_start_transaction(root, 1);
1251 btrfs_set_trans_block_group(trans, dir);
1252
1253 err = btrfs_orphan_add(trans, inode);
1254 if (err)
1255 goto fail_trans;
1256
1257 /* now the directory is empty */
1258 err = btrfs_unlink_trans(trans, root, dir, dentry);
1259 if (!err) {
1260 btrfs_i_size_write(inode, 0);
1261 }
1262
1263 fail_trans:
1264 nr = trans->blocks_used;
1265 ret = btrfs_end_transaction_throttle(trans, root);
1266 fail:
1267 btrfs_btree_balance_dirty(root, nr);
1268
1269 if (ret && !err)
1270 err = ret;
1271 return err;
1272 }
1273
1274 /*
1275 * this can truncate away extent items, csum items and directory items.
1276 * It starts at a high offset and removes keys until it can't find
1277 * any higher than i_size.
1278 *
1279 * csum items that cross the new i_size are truncated to the new size
1280 * as well.
1281 *
1282 * min_type is the minimum key type to truncate down to. If set to 0, this
1283 * will kill all the items on this inode, including the INODE_ITEM_KEY.
1284 */
1285 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
1286 struct btrfs_root *root,
1287 struct inode *inode,
1288 u32 min_type)
1289 {
1290 int ret;
1291 struct btrfs_path *path;
1292 struct btrfs_key key;
1293 struct btrfs_key found_key;
1294 u32 found_type;
1295 struct extent_buffer *leaf;
1296 struct btrfs_file_extent_item *fi;
1297 u64 extent_start = 0;
1298 u64 extent_num_bytes = 0;
1299 u64 item_end = 0;
1300 u64 root_gen = 0;
1301 u64 root_owner = 0;
1302 int found_extent;
1303 int del_item;
1304 int pending_del_nr = 0;
1305 int pending_del_slot = 0;
1306 int extent_type = -1;
1307 u64 mask = root->sectorsize - 1;
1308
1309 btrfs_drop_extent_cache(inode, inode->i_size & (~mask), (u64)-1);
1310 path = btrfs_alloc_path();
1311 path->reada = -1;
1312 BUG_ON(!path);
1313
1314 /* FIXME, add redo link to tree so we don't leak on crash */
1315 key.objectid = inode->i_ino;
1316 key.offset = (u64)-1;
1317 key.type = (u8)-1;
1318
1319 btrfs_init_path(path);
1320 search_again:
1321 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1322 if (ret < 0) {
1323 goto error;
1324 }
1325 if (ret > 0) {
1326 BUG_ON(path->slots[0] == 0);
1327 path->slots[0]--;
1328 }
1329
1330 while(1) {
1331 fi = NULL;
1332 leaf = path->nodes[0];
1333 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1334 found_type = btrfs_key_type(&found_key);
1335
1336 if (found_key.objectid != inode->i_ino)
1337 break;
1338
1339 if (found_type < min_type)
1340 break;
1341
1342 item_end = found_key.offset;
1343 if (found_type == BTRFS_EXTENT_DATA_KEY) {
1344 fi = btrfs_item_ptr(leaf, path->slots[0],
1345 struct btrfs_file_extent_item);
1346 extent_type = btrfs_file_extent_type(leaf, fi);
1347 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
1348 item_end +=
1349 btrfs_file_extent_num_bytes(leaf, fi);
1350 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1351 struct btrfs_item *item = btrfs_item_nr(leaf,
1352 path->slots[0]);
1353 item_end += btrfs_file_extent_inline_len(leaf,
1354 item);
1355 }
1356 item_end--;
1357 }
1358 if (found_type == BTRFS_CSUM_ITEM_KEY) {
1359 ret = btrfs_csum_truncate(trans, root, path,
1360 inode->i_size);
1361 BUG_ON(ret);
1362 }
1363 if (item_end < inode->i_size) {
1364 if (found_type == BTRFS_DIR_ITEM_KEY) {
1365 found_type = BTRFS_INODE_ITEM_KEY;
1366 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
1367 found_type = BTRFS_CSUM_ITEM_KEY;
1368 } else if (found_type == BTRFS_EXTENT_DATA_KEY) {
1369 found_type = BTRFS_XATTR_ITEM_KEY;
1370 } else if (found_type == BTRFS_XATTR_ITEM_KEY) {
1371 found_type = BTRFS_INODE_REF_KEY;
1372 } else if (found_type) {
1373 found_type--;
1374 } else {
1375 break;
1376 }
1377 btrfs_set_key_type(&key, found_type);
1378 goto next;
1379 }
1380 if (found_key.offset >= inode->i_size)
1381 del_item = 1;
1382 else
1383 del_item = 0;
1384 found_extent = 0;
1385
1386 /* FIXME, shrink the extent if the ref count is only 1 */
1387 if (found_type != BTRFS_EXTENT_DATA_KEY)
1388 goto delete;
1389
1390 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
1391 u64 num_dec;
1392 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
1393 if (!del_item) {
1394 u64 orig_num_bytes =
1395 btrfs_file_extent_num_bytes(leaf, fi);
1396 extent_num_bytes = inode->i_size -
1397 found_key.offset + root->sectorsize - 1;
1398 extent_num_bytes = extent_num_bytes &
1399 ~((u64)root->sectorsize - 1);
1400 btrfs_set_file_extent_num_bytes(leaf, fi,
1401 extent_num_bytes);
1402 num_dec = (orig_num_bytes -
1403 extent_num_bytes);
1404 if (extent_start != 0)
1405 dec_i_blocks(inode, num_dec);
1406 btrfs_mark_buffer_dirty(leaf);
1407 } else {
1408 extent_num_bytes =
1409 btrfs_file_extent_disk_num_bytes(leaf,
1410 fi);
1411 /* FIXME blocksize != 4096 */
1412 num_dec = btrfs_file_extent_num_bytes(leaf, fi);
1413 if (extent_start != 0) {
1414 found_extent = 1;
1415 dec_i_blocks(inode, num_dec);
1416 }
1417 root_gen = btrfs_header_generation(leaf);
1418 root_owner = btrfs_header_owner(leaf);
1419 }
1420 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1421 if (!del_item) {
1422 u32 newsize = inode->i_size - found_key.offset;
1423 dec_i_blocks(inode, item_end + 1 -
1424 found_key.offset - newsize);
1425 newsize =
1426 btrfs_file_extent_calc_inline_size(newsize);
1427 ret = btrfs_truncate_item(trans, root, path,
1428 newsize, 1);
1429 BUG_ON(ret);
1430 } else {
1431 dec_i_blocks(inode, item_end + 1 -
1432 found_key.offset);
1433 }
1434 }
1435 delete:
1436 if (del_item) {
1437 if (!pending_del_nr) {
1438 /* no pending yet, add ourselves */
1439 pending_del_slot = path->slots[0];
1440 pending_del_nr = 1;
1441 } else if (pending_del_nr &&
1442 path->slots[0] + 1 == pending_del_slot) {
1443 /* hop on the pending chunk */
1444 pending_del_nr++;
1445 pending_del_slot = path->slots[0];
1446 } else {
1447 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot);
1448 }
1449 } else {
1450 break;
1451 }
1452 if (found_extent) {
1453 ret = btrfs_free_extent(trans, root, extent_start,
1454 extent_num_bytes,
1455 root_owner,
1456 root_gen, inode->i_ino,
1457 found_key.offset, 0);
1458 BUG_ON(ret);
1459 }
1460 next:
1461 if (path->slots[0] == 0) {
1462 if (pending_del_nr)
1463 goto del_pending;
1464 btrfs_release_path(root, path);
1465 goto search_again;
1466 }
1467
1468 path->slots[0]--;
1469 if (pending_del_nr &&
1470 path->slots[0] + 1 != pending_del_slot) {
1471 struct btrfs_key debug;
1472 del_pending:
1473 btrfs_item_key_to_cpu(path->nodes[0], &debug,
1474 pending_del_slot);
1475 ret = btrfs_del_items(trans, root, path,
1476 pending_del_slot,
1477 pending_del_nr);
1478 BUG_ON(ret);
1479 pending_del_nr = 0;
1480 btrfs_release_path(root, path);
1481 goto search_again;
1482 }
1483 }
1484 ret = 0;
1485 error:
1486 if (pending_del_nr) {
1487 ret = btrfs_del_items(trans, root, path, pending_del_slot,
1488 pending_del_nr);
1489 }
1490 btrfs_free_path(path);
1491 inode->i_sb->s_dirt = 1;
1492 return ret;
1493 }
1494
1495 /*
1496 * taken from block_truncate_page, but does cow as it zeros out
1497 * any bytes left in the last page in the file.
1498 */
1499 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
1500 {
1501 struct inode *inode = mapping->host;
1502 struct btrfs_root *root = BTRFS_I(inode)->root;
1503 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1504 struct btrfs_ordered_extent *ordered;
1505 char *kaddr;
1506 u32 blocksize = root->sectorsize;
1507 pgoff_t index = from >> PAGE_CACHE_SHIFT;
1508 unsigned offset = from & (PAGE_CACHE_SIZE-1);
1509 struct page *page;
1510 int ret = 0;
1511 u64 page_start;
1512 u64 page_end;
1513
1514 if ((offset & (blocksize - 1)) == 0)
1515 goto out;
1516
1517 ret = -ENOMEM;
1518 again:
1519 page = grab_cache_page(mapping, index);
1520 if (!page)
1521 goto out;
1522
1523 page_start = page_offset(page);
1524 page_end = page_start + PAGE_CACHE_SIZE - 1;
1525
1526 if (!PageUptodate(page)) {
1527 ret = btrfs_readpage(NULL, page);
1528 lock_page(page);
1529 if (page->mapping != mapping) {
1530 unlock_page(page);
1531 page_cache_release(page);
1532 goto again;
1533 }
1534 if (!PageUptodate(page)) {
1535 ret = -EIO;
1536 goto out_unlock;
1537 }
1538 }
1539 wait_on_page_writeback(page);
1540
1541 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
1542 set_page_extent_mapped(page);
1543
1544 ordered = btrfs_lookup_ordered_extent(inode, page_start);
1545 if (ordered) {
1546 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
1547 unlock_page(page);
1548 page_cache_release(page);
1549 btrfs_start_ordered_extent(inode, ordered, 1);
1550 btrfs_put_ordered_extent(ordered);
1551 goto again;
1552 }
1553
1554 btrfs_set_extent_delalloc(inode, page_start, page_end);
1555 ret = 0;
1556 if (offset != PAGE_CACHE_SIZE) {
1557 kaddr = kmap(page);
1558 memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
1559 flush_dcache_page(page);
1560 kunmap(page);
1561 }
1562 ClearPageChecked(page);
1563 set_page_dirty(page);
1564 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
1565
1566 out_unlock:
1567 unlock_page(page);
1568 page_cache_release(page);
1569 out:
1570 return ret;
1571 }
1572
1573 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
1574 {
1575 struct inode *inode = dentry->d_inode;
1576 int err;
1577
1578 err = inode_change_ok(inode, attr);
1579 if (err)
1580 return err;
1581
1582 if (S_ISREG(inode->i_mode) &&
1583 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
1584 struct btrfs_trans_handle *trans;
1585 struct btrfs_root *root = BTRFS_I(inode)->root;
1586 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1587
1588 u64 mask = root->sectorsize - 1;
1589 u64 hole_start = (inode->i_size + mask) & ~mask;
1590 u64 block_end = (attr->ia_size + mask) & ~mask;
1591 u64 hole_size;
1592 u64 alloc_hint = 0;
1593
1594 if (attr->ia_size <= hole_start)
1595 goto out;
1596
1597 err = btrfs_check_free_space(root, 1, 0);
1598 if (err)
1599 goto fail;
1600
1601 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1602
1603 hole_size = block_end - hole_start;
1604 while(1) {
1605 struct btrfs_ordered_extent *ordered;
1606 btrfs_wait_ordered_range(inode, hole_start, hole_size);
1607
1608 lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1609 ordered = btrfs_lookup_ordered_extent(inode, hole_start);
1610 if (ordered) {
1611 unlock_extent(io_tree, hole_start,
1612 block_end - 1, GFP_NOFS);
1613 btrfs_put_ordered_extent(ordered);
1614 } else {
1615 break;
1616 }
1617 }
1618
1619 trans = btrfs_start_transaction(root, 1);
1620 btrfs_set_trans_block_group(trans, inode);
1621 mutex_lock(&BTRFS_I(inode)->extent_mutex);
1622 err = btrfs_drop_extents(trans, root, inode,
1623 hole_start, block_end, hole_start,
1624 &alloc_hint);
1625
1626 if (alloc_hint != EXTENT_MAP_INLINE) {
1627 err = btrfs_insert_file_extent(trans, root,
1628 inode->i_ino,
1629 hole_start, 0, 0,
1630 hole_size, 0);
1631 btrfs_drop_extent_cache(inode, hole_start,
1632 (u64)-1);
1633 btrfs_check_file(root, inode);
1634 }
1635 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
1636 btrfs_end_transaction(trans, root);
1637 unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1638 if (err)
1639 return err;
1640 }
1641 out:
1642 err = inode_setattr(inode, attr);
1643
1644 if (!err && ((attr->ia_valid & ATTR_MODE)))
1645 err = btrfs_acl_chmod(inode);
1646 fail:
1647 return err;
1648 }
1649
1650 void btrfs_delete_inode(struct inode *inode)
1651 {
1652 struct btrfs_trans_handle *trans;
1653 struct btrfs_root *root = BTRFS_I(inode)->root;
1654 unsigned long nr;
1655 int ret;
1656
1657 truncate_inode_pages(&inode->i_data, 0);
1658 if (is_bad_inode(inode)) {
1659 btrfs_orphan_del(NULL, inode);
1660 goto no_delete;
1661 }
1662 btrfs_wait_ordered_range(inode, 0, (u64)-1);
1663
1664 btrfs_i_size_write(inode, 0);
1665 trans = btrfs_start_transaction(root, 1);
1666
1667 btrfs_set_trans_block_group(trans, inode);
1668 ret = btrfs_truncate_in_trans(trans, root, inode, 0);
1669 if (ret) {
1670 btrfs_orphan_del(NULL, inode);
1671 goto no_delete_lock;
1672 }
1673
1674 btrfs_orphan_del(trans, inode);
1675
1676 nr = trans->blocks_used;
1677 clear_inode(inode);
1678
1679 btrfs_end_transaction(trans, root);
1680 btrfs_btree_balance_dirty(root, nr);
1681 return;
1682
1683 no_delete_lock:
1684 nr = trans->blocks_used;
1685 btrfs_end_transaction(trans, root);
1686 btrfs_btree_balance_dirty(root, nr);
1687 no_delete:
1688 clear_inode(inode);
1689 }
1690
1691 /*
1692 * this returns the key found in the dir entry in the location pointer.
1693 * If no dir entries were found, location->objectid is 0.
1694 */
1695 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
1696 struct btrfs_key *location)
1697 {
1698 const char *name = dentry->d_name.name;
1699 int namelen = dentry->d_name.len;
1700 struct btrfs_dir_item *di;
1701 struct btrfs_path *path;
1702 struct btrfs_root *root = BTRFS_I(dir)->root;
1703 int ret = 0;
1704
1705 if (namelen == 1 && strcmp(name, ".") == 0) {
1706 location->objectid = dir->i_ino;
1707 location->type = BTRFS_INODE_ITEM_KEY;
1708 location->offset = 0;
1709 return 0;
1710 }
1711 path = btrfs_alloc_path();
1712 BUG_ON(!path);
1713
1714 if (namelen == 2 && strcmp(name, "..") == 0) {
1715 struct btrfs_key key;
1716 struct extent_buffer *leaf;
1717 int slot;
1718
1719 key.objectid = dir->i_ino;
1720 key.offset = (u64)-1;
1721 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1722 if (ret < 0 || path->slots[0] == 0)
1723 goto out_err;
1724 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1725 BUG_ON(ret == 0);
1726 ret = 0;
1727 leaf = path->nodes[0];
1728 slot = path->slots[0] - 1;
1729
1730 btrfs_item_key_to_cpu(leaf, &key, slot);
1731 if (key.objectid != dir->i_ino ||
1732 key.type != BTRFS_INODE_REF_KEY) {
1733 goto out_err;
1734 }
1735 location->objectid = key.offset;
1736 location->type = BTRFS_INODE_ITEM_KEY;
1737 location->offset = 0;
1738 goto out;
1739 }
1740
1741 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
1742 namelen, 0);
1743 if (IS_ERR(di))
1744 ret = PTR_ERR(di);
1745 if (!di || IS_ERR(di)) {
1746 goto out_err;
1747 }
1748 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
1749 out:
1750 btrfs_free_path(path);
1751 return ret;
1752 out_err:
1753 location->objectid = 0;
1754 goto out;
1755 }
1756
1757 /*
1758 * when we hit a tree root in a directory, the btrfs part of the inode
1759 * needs to be changed to reflect the root directory of the tree root. This
1760 * is kind of like crossing a mount point.
1761 */
1762 static int fixup_tree_root_location(struct btrfs_root *root,
1763 struct btrfs_key *location,
1764 struct btrfs_root **sub_root,
1765 struct dentry *dentry)
1766 {
1767 struct btrfs_root_item *ri;
1768
1769 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
1770 return 0;
1771 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1772 return 0;
1773
1774 *sub_root = btrfs_read_fs_root(root->fs_info, location,
1775 dentry->d_name.name,
1776 dentry->d_name.len);
1777 if (IS_ERR(*sub_root))
1778 return PTR_ERR(*sub_root);
1779
1780 ri = &(*sub_root)->root_item;
1781 location->objectid = btrfs_root_dirid(ri);
1782 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1783 location->offset = 0;
1784
1785 return 0;
1786 }
1787
1788 static int btrfs_init_locked_inode(struct inode *inode, void *p)
1789 {
1790 struct btrfs_iget_args *args = p;
1791 inode->i_ino = args->ino;
1792 BTRFS_I(inode)->root = args->root;
1793 BTRFS_I(inode)->delalloc_bytes = 0;
1794 inode->i_mapping->writeback_index = 0;
1795 BTRFS_I(inode)->disk_i_size = 0;
1796 BTRFS_I(inode)->index_cnt = (u64)-1;
1797 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1798 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1799 inode->i_mapping, GFP_NOFS);
1800 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1801 inode->i_mapping, GFP_NOFS);
1802 INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
1803 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
1804 mutex_init(&BTRFS_I(inode)->csum_mutex);
1805 mutex_init(&BTRFS_I(inode)->extent_mutex);
1806 return 0;
1807 }
1808
1809 static int btrfs_find_actor(struct inode *inode, void *opaque)
1810 {
1811 struct btrfs_iget_args *args = opaque;
1812 return (args->ino == inode->i_ino &&
1813 args->root == BTRFS_I(inode)->root);
1814 }
1815
1816 struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
1817 u64 root_objectid)
1818 {
1819 struct btrfs_iget_args args;
1820 args.ino = objectid;
1821 args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
1822
1823 if (!args.root)
1824 return NULL;
1825
1826 return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
1827 }
1828
1829 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
1830 struct btrfs_root *root)
1831 {
1832 struct inode *inode;
1833 struct btrfs_iget_args args;
1834 args.ino = objectid;
1835 args.root = root;
1836
1837 inode = iget5_locked(s, objectid, btrfs_find_actor,
1838 btrfs_init_locked_inode,
1839 (void *)&args);
1840 return inode;
1841 }
1842
1843 /* Get an inode object given its location and corresponding root.
1844 * Returns in *is_new if the inode was read from disk
1845 */
1846 struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location,
1847 struct btrfs_root *root, int *is_new)
1848 {
1849 struct inode *inode;
1850
1851 inode = btrfs_iget_locked(s, location->objectid, root);
1852 if (!inode)
1853 return ERR_PTR(-EACCES);
1854
1855 if (inode->i_state & I_NEW) {
1856 BTRFS_I(inode)->root = root;
1857 memcpy(&BTRFS_I(inode)->location, location, sizeof(*location));
1858 btrfs_read_locked_inode(inode);
1859 unlock_new_inode(inode);
1860 if (is_new)
1861 *is_new = 1;
1862 } else {
1863 if (is_new)
1864 *is_new = 0;
1865 }
1866
1867 return inode;
1868 }
1869
1870 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
1871 struct nameidata *nd)
1872 {
1873 struct inode * inode;
1874 struct btrfs_inode *bi = BTRFS_I(dir);
1875 struct btrfs_root *root = bi->root;
1876 struct btrfs_root *sub_root = root;
1877 struct btrfs_key location;
1878 int ret, new, do_orphan = 0;
1879
1880 if (dentry->d_name.len > BTRFS_NAME_LEN)
1881 return ERR_PTR(-ENAMETOOLONG);
1882
1883 ret = btrfs_inode_by_name(dir, dentry, &location);
1884
1885 if (ret < 0)
1886 return ERR_PTR(ret);
1887
1888 inode = NULL;
1889 if (location.objectid) {
1890 ret = fixup_tree_root_location(root, &location, &sub_root,
1891 dentry);
1892 if (ret < 0)
1893 return ERR_PTR(ret);
1894 if (ret > 0)
1895 return ERR_PTR(-ENOENT);
1896 inode = btrfs_iget(dir->i_sb, &location, sub_root, &new);
1897 if (IS_ERR(inode))
1898 return ERR_CAST(inode);
1899
1900 /* the inode and parent dir are two different roots */
1901 if (new && root != sub_root) {
1902 igrab(inode);
1903 sub_root->inode = inode;
1904 do_orphan = 1;
1905 }
1906 }
1907
1908 if (unlikely(do_orphan))
1909 btrfs_orphan_cleanup(sub_root);
1910
1911 return d_splice_alias(inode, dentry);
1912 }
1913
1914 static unsigned char btrfs_filetype_table[] = {
1915 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1916 };
1917
1918 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1919 {
1920 struct inode *inode = filp->f_dentry->d_inode;
1921 struct btrfs_root *root = BTRFS_I(inode)->root;
1922 struct btrfs_item *item;
1923 struct btrfs_dir_item *di;
1924 struct btrfs_key key;
1925 struct btrfs_key found_key;
1926 struct btrfs_path *path;
1927 int ret;
1928 u32 nritems;
1929 struct extent_buffer *leaf;
1930 int slot;
1931 int advance;
1932 unsigned char d_type;
1933 int over = 0;
1934 u32 di_cur;
1935 u32 di_total;
1936 u32 di_len;
1937 int key_type = BTRFS_DIR_INDEX_KEY;
1938 char tmp_name[32];
1939 char *name_ptr;
1940 int name_len;
1941
1942 /* FIXME, use a real flag for deciding about the key type */
1943 if (root->fs_info->tree_root == root)
1944 key_type = BTRFS_DIR_ITEM_KEY;
1945
1946 /* special case for "." */
1947 if (filp->f_pos == 0) {
1948 over = filldir(dirent, ".", 1,
1949 1, inode->i_ino,
1950 DT_DIR);
1951 if (over)
1952 return 0;
1953 filp->f_pos = 1;
1954 }
1955
1956 key.objectid = inode->i_ino;
1957 path = btrfs_alloc_path();
1958 path->reada = 2;
1959
1960 /* special case for .., just use the back ref */
1961 if (filp->f_pos == 1) {
1962 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1963 key.offset = (u64)-1;
1964 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1965 if (ret < 0 || path->slots[0] == 0) {
1966 btrfs_release_path(root, path);
1967 goto read_dir_items;
1968 }
1969 BUG_ON(ret == 0);
1970 leaf = path->nodes[0];
1971 slot = path->slots[0] - 1;
1972 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1973 btrfs_release_path(root, path);
1974 if (found_key.objectid != key.objectid ||
1975 found_key.type != BTRFS_INODE_REF_KEY)
1976 goto read_dir_items;
1977 over = filldir(dirent, "..", 2,
1978 2, found_key.offset, DT_DIR);
1979 if (over)
1980 goto nopos;
1981 filp->f_pos = 2;
1982 }
1983
1984 read_dir_items:
1985 btrfs_set_key_type(&key, key_type);
1986 key.offset = filp->f_pos;
1987
1988 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1989 if (ret < 0)
1990 goto err;
1991 advance = 0;
1992 while(1) {
1993 leaf = path->nodes[0];
1994 nritems = btrfs_header_nritems(leaf);
1995 slot = path->slots[0];
1996 if (advance || slot >= nritems) {
1997 if (slot >= nritems -1) {
1998 ret = btrfs_next_leaf(root, path);
1999 if (ret)
2000 break;
2001 leaf = path->nodes[0];
2002 nritems = btrfs_header_nritems(leaf);
2003 slot = path->slots[0];
2004 } else {
2005 slot++;
2006 path->slots[0]++;
2007 }
2008 }
2009 advance = 1;
2010 item = btrfs_item_nr(leaf, slot);
2011 btrfs_item_key_to_cpu(leaf, &found_key, slot);
2012
2013 if (found_key.objectid != key.objectid)
2014 break;
2015 if (btrfs_key_type(&found_key) != key_type)
2016 break;
2017 if (found_key.offset < filp->f_pos)
2018 continue;
2019
2020 filp->f_pos = found_key.offset;
2021 advance = 1;
2022 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
2023 di_cur = 0;
2024 di_total = btrfs_item_size(leaf, item);
2025 while(di_cur < di_total) {
2026 struct btrfs_key location;
2027
2028 name_len = btrfs_dir_name_len(leaf, di);
2029 if (name_len < 32) {
2030 name_ptr = tmp_name;
2031 } else {
2032 name_ptr = kmalloc(name_len, GFP_NOFS);
2033 BUG_ON(!name_ptr);
2034 }
2035 read_extent_buffer(leaf, name_ptr,
2036 (unsigned long)(di + 1), name_len);
2037
2038 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
2039 btrfs_dir_item_key_to_cpu(leaf, di, &location);
2040 over = filldir(dirent, name_ptr, name_len,
2041 found_key.offset,
2042 location.objectid,
2043 d_type);
2044
2045 if (name_ptr != tmp_name)
2046 kfree(name_ptr);
2047
2048 if (over)
2049 goto nopos;
2050 di_len = btrfs_dir_name_len(leaf, di) +
2051 btrfs_dir_data_len(leaf, di) +sizeof(*di);
2052 di_cur += di_len;
2053 di = (struct btrfs_dir_item *)((char *)di + di_len);
2054 }
2055 }
2056 if (key_type == BTRFS_DIR_INDEX_KEY)
2057 filp->f_pos = INT_LIMIT(typeof(filp->f_pos));
2058 else
2059 filp->f_pos++;
2060 nopos:
2061 ret = 0;
2062 err:
2063 btrfs_free_path(path);
2064 return ret;
2065 }
2066
2067 int btrfs_write_inode(struct inode *inode, int wait)
2068 {
2069 struct btrfs_root *root = BTRFS_I(inode)->root;
2070 struct btrfs_trans_handle *trans;
2071 int ret = 0;
2072
2073 if (root->fs_info->closing > 1)
2074 return 0;
2075
2076 if (wait) {
2077 trans = btrfs_join_transaction(root, 1);
2078 btrfs_set_trans_block_group(trans, inode);
2079 ret = btrfs_commit_transaction(trans, root);
2080 }
2081 return ret;
2082 }
2083
2084 /*
2085 * This is somewhat expensive, updating the tree every time the
2086 * inode changes. But, it is most likely to find the inode in cache.
2087 * FIXME, needs more benchmarking...there are no reasons other than performance
2088 * to keep or drop this code.
2089 */
2090 void btrfs_dirty_inode(struct inode *inode)
2091 {
2092 struct btrfs_root *root = BTRFS_I(inode)->root;
2093 struct btrfs_trans_handle *trans;
2094
2095 trans = btrfs_join_transaction(root, 1);
2096 btrfs_set_trans_block_group(trans, inode);
2097 btrfs_update_inode(trans, root, inode);
2098 btrfs_end_transaction(trans, root);
2099 }
2100
2101 static int btrfs_set_inode_index_count(struct inode *inode)
2102 {
2103 struct btrfs_root *root = BTRFS_I(inode)->root;
2104 struct btrfs_key key, found_key;
2105 struct btrfs_path *path;
2106 struct extent_buffer *leaf;
2107 int ret;
2108
2109 key.objectid = inode->i_ino;
2110 btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY);
2111 key.offset = (u64)-1;
2112
2113 path = btrfs_alloc_path();
2114 if (!path)
2115 return -ENOMEM;
2116
2117 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2118 if (ret < 0)
2119 goto out;
2120 /* FIXME: we should be able to handle this */
2121 if (ret == 0)
2122 goto out;
2123 ret = 0;
2124
2125 /*
2126 * MAGIC NUMBER EXPLANATION:
2127 * since we search a directory based on f_pos we have to start at 2
2128 * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody
2129 * else has to start at 2
2130 */
2131 if (path->slots[0] == 0) {
2132 BTRFS_I(inode)->index_cnt = 2;
2133 goto out;
2134 }
2135
2136 path->slots[0]--;
2137
2138 leaf = path->nodes[0];
2139 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2140
2141 if (found_key.objectid != inode->i_ino ||
2142 btrfs_key_type(&found_key) != BTRFS_DIR_INDEX_KEY) {
2143 BTRFS_I(inode)->index_cnt = 2;
2144 goto out;
2145 }
2146
2147 BTRFS_I(inode)->index_cnt = found_key.offset + 1;
2148 out:
2149 btrfs_free_path(path);
2150 return ret;
2151 }
2152
2153 static int btrfs_set_inode_index(struct inode *dir, struct inode *inode,
2154 u64 *index)
2155 {
2156 int ret = 0;
2157
2158 if (BTRFS_I(dir)->index_cnt == (u64)-1) {
2159 ret = btrfs_set_inode_index_count(dir);
2160 if (ret)
2161 return ret;
2162 }
2163
2164 *index = BTRFS_I(dir)->index_cnt;
2165 BTRFS_I(dir)->index_cnt++;
2166
2167 return ret;
2168 }
2169
2170 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
2171 struct btrfs_root *root,
2172 struct inode *dir,
2173 const char *name, int name_len,
2174 u64 ref_objectid,
2175 u64 objectid,
2176 struct btrfs_block_group_cache *group,
2177 int mode, u64 *index)
2178 {
2179 struct inode *inode;
2180 struct btrfs_inode_item *inode_item;
2181 struct btrfs_block_group_cache *new_inode_group;
2182 struct btrfs_key *location;
2183 struct btrfs_path *path;
2184 struct btrfs_inode_ref *ref;
2185 struct btrfs_key key[2];
2186 u32 sizes[2];
2187 unsigned long ptr;
2188 int ret;
2189 int owner;
2190
2191 path = btrfs_alloc_path();
2192 BUG_ON(!path);
2193
2194 inode = new_inode(root->fs_info->sb);
2195 if (!inode)
2196 return ERR_PTR(-ENOMEM);
2197
2198 if (dir) {
2199 ret = btrfs_set_inode_index(dir, inode, index);
2200 if (ret)
2201 return ERR_PTR(ret);
2202 }
2203 /*
2204 * index_cnt is ignored for everything but a dir,
2205 * btrfs_get_inode_index_count has an explanation for the magic
2206 * number
2207 */
2208 BTRFS_I(inode)->index_cnt = 2;
2209
2210 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2211 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
2212 inode->i_mapping, GFP_NOFS);
2213 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
2214 inode->i_mapping, GFP_NOFS);
2215 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
2216 INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
2217 mutex_init(&BTRFS_I(inode)->csum_mutex);
2218 mutex_init(&BTRFS_I(inode)->extent_mutex);
2219 BTRFS_I(inode)->delalloc_bytes = 0;
2220 inode->i_mapping->writeback_index = 0;
2221 BTRFS_I(inode)->disk_i_size = 0;
2222 BTRFS_I(inode)->root = root;
2223
2224 if (mode & S_IFDIR)
2225 owner = 0;
2226 else
2227 owner = 1;
2228 new_inode_group = btrfs_find_block_group(root, group, 0,
2229 BTRFS_BLOCK_GROUP_METADATA, owner);
2230 if (!new_inode_group) {
2231 printk("find_block group failed\n");
2232 new_inode_group = group;
2233 }
2234 BTRFS_I(inode)->block_group = new_inode_group;
2235 BTRFS_I(inode)->flags = 0;
2236
2237 key[0].objectid = objectid;
2238 btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
2239 key[0].offset = 0;
2240
2241 key[1].objectid = objectid;
2242 btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
2243 key[1].offset = ref_objectid;
2244
2245 sizes[0] = sizeof(struct btrfs_inode_item);
2246 sizes[1] = name_len + sizeof(*ref);
2247
2248 ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
2249 if (ret != 0)
2250 goto fail;
2251
2252 if (objectid > root->highest_inode)
2253 root->highest_inode = objectid;
2254
2255 inode->i_uid = current->fsuid;
2256 inode->i_gid = current->fsgid;
2257 inode->i_mode = mode;
2258 inode->i_ino = objectid;
2259 inode->i_blocks = 0;
2260 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2261 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
2262 struct btrfs_inode_item);
2263 fill_inode_item(path->nodes[0], inode_item, inode);
2264
2265 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
2266 struct btrfs_inode_ref);
2267 btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
2268 btrfs_set_inode_ref_index(path->nodes[0], ref, *index);
2269 ptr = (unsigned long)(ref + 1);
2270 write_extent_buffer(path->nodes[0], name, ptr, name_len);
2271
2272 btrfs_mark_buffer_dirty(path->nodes[0]);
2273 btrfs_free_path(path);
2274
2275 location = &BTRFS_I(inode)->location;
2276 location->objectid = objectid;
2277 location->offset = 0;
2278 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
2279
2280 insert_inode_hash(inode);
2281 return inode;
2282 fail:
2283 if (dir)
2284 BTRFS_I(dir)->index_cnt--;
2285 btrfs_free_path(path);
2286 return ERR_PTR(ret);
2287 }
2288
2289 static inline u8 btrfs_inode_type(struct inode *inode)
2290 {
2291 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
2292 }
2293
2294 static int btrfs_add_link(struct btrfs_trans_handle *trans,
2295 struct dentry *dentry, struct inode *inode,
2296 int add_backref, u64 index)
2297 {
2298 int ret;
2299 struct btrfs_key key;
2300 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
2301 struct inode *parent_inode = dentry->d_parent->d_inode;
2302
2303 key.objectid = inode->i_ino;
2304 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
2305 key.offset = 0;
2306
2307 ret = btrfs_insert_dir_item(trans, root,
2308 dentry->d_name.name, dentry->d_name.len,
2309 dentry->d_parent->d_inode->i_ino,
2310 &key, btrfs_inode_type(inode),
2311 index);
2312 if (ret == 0) {
2313 if (add_backref) {
2314 ret = btrfs_insert_inode_ref(trans, root,
2315 dentry->d_name.name,
2316 dentry->d_name.len,
2317 inode->i_ino,
2318 parent_inode->i_ino,
2319 index);
2320 }
2321 btrfs_i_size_write(parent_inode, parent_inode->i_size +
2322 dentry->d_name.len * 2);
2323 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
2324 ret = btrfs_update_inode(trans, root,
2325 dentry->d_parent->d_inode);
2326 }
2327 return ret;
2328 }
2329
2330 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
2331 struct dentry *dentry, struct inode *inode,
2332 int backref, u64 index)
2333 {
2334 int err = btrfs_add_link(trans, dentry, inode, backref, index);
2335 if (!err) {
2336 d_instantiate(dentry, inode);
2337 return 0;
2338 }
2339 if (err > 0)
2340 err = -EEXIST;
2341 return err;
2342 }
2343
2344 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
2345 int mode, dev_t rdev)
2346 {
2347 struct btrfs_trans_handle *trans;
2348 struct btrfs_root *root = BTRFS_I(dir)->root;
2349 struct inode *inode = NULL;
2350 int err;
2351 int drop_inode = 0;
2352 u64 objectid;
2353 unsigned long nr = 0;
2354 u64 index = 0;
2355
2356 if (!new_valid_dev(rdev))
2357 return -EINVAL;
2358
2359 err = btrfs_check_free_space(root, 1, 0);
2360 if (err)
2361 goto fail;
2362
2363 trans = btrfs_start_transaction(root, 1);
2364 btrfs_set_trans_block_group(trans, dir);
2365
2366 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2367 if (err) {
2368 err = -ENOSPC;
2369 goto out_unlock;
2370 }
2371
2372 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
2373 dentry->d_name.len,
2374 dentry->d_parent->d_inode->i_ino, objectid,
2375 BTRFS_I(dir)->block_group, mode, &index);
2376 err = PTR_ERR(inode);
2377 if (IS_ERR(inode))
2378 goto out_unlock;
2379
2380 err = btrfs_init_acl(inode, dir);
2381 if (err) {
2382 drop_inode = 1;
2383 goto out_unlock;
2384 }
2385
2386 btrfs_set_trans_block_group(trans, inode);
2387 err = btrfs_add_nondir(trans, dentry, inode, 0, index);
2388 if (err)
2389 drop_inode = 1;
2390 else {
2391 inode->i_op = &btrfs_special_inode_operations;
2392 init_special_inode(inode, inode->i_mode, rdev);
2393 btrfs_update_inode(trans, root, inode);
2394 }
2395 dir->i_sb->s_dirt = 1;
2396 btrfs_update_inode_block_group(trans, inode);
2397 btrfs_update_inode_block_group(trans, dir);
2398 out_unlock:
2399 nr = trans->blocks_used;
2400 btrfs_end_transaction_throttle(trans, root);
2401 fail:
2402 if (drop_inode) {
2403 inode_dec_link_count(inode);
2404 iput(inode);
2405 }
2406 btrfs_btree_balance_dirty(root, nr);
2407 return err;
2408 }
2409
2410 static int btrfs_create(struct inode *dir, struct dentry *dentry,
2411 int mode, struct nameidata *nd)
2412 {
2413 struct btrfs_trans_handle *trans;
2414 struct btrfs_root *root = BTRFS_I(dir)->root;
2415 struct inode *inode = NULL;
2416 int err;
2417 int drop_inode = 0;
2418 unsigned long nr = 0;
2419 u64 objectid;
2420 u64 index = 0;
2421
2422 err = btrfs_check_free_space(root, 1, 0);
2423 if (err)
2424 goto fail;
2425 trans = btrfs_start_transaction(root, 1);
2426 btrfs_set_trans_block_group(trans, dir);
2427
2428 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2429 if (err) {
2430 err = -ENOSPC;
2431 goto out_unlock;
2432 }
2433
2434 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
2435 dentry->d_name.len,
2436 dentry->d_parent->d_inode->i_ino,
2437 objectid, BTRFS_I(dir)->block_group, mode,
2438 &index);
2439 err = PTR_ERR(inode);
2440 if (IS_ERR(inode))
2441 goto out_unlock;
2442
2443 err = btrfs_init_acl(inode, dir);
2444 if (err) {
2445 drop_inode = 1;
2446 goto out_unlock;
2447 }
2448
2449 btrfs_set_trans_block_group(trans, inode);
2450 err = btrfs_add_nondir(trans, dentry, inode, 0, index);
2451 if (err)
2452 drop_inode = 1;
2453 else {
2454 inode->i_mapping->a_ops = &btrfs_aops;
2455 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
2456 inode->i_fop = &btrfs_file_operations;
2457 inode->i_op = &btrfs_file_inode_operations;
2458 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2459 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
2460 inode->i_mapping, GFP_NOFS);
2461 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
2462 inode->i_mapping, GFP_NOFS);
2463 INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
2464 mutex_init(&BTRFS_I(inode)->csum_mutex);
2465 mutex_init(&BTRFS_I(inode)->extent_mutex);
2466 BTRFS_I(inode)->delalloc_bytes = 0;
2467 BTRFS_I(inode)->disk_i_size = 0;
2468 inode->i_mapping->writeback_index = 0;
2469 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
2470 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
2471 }
2472 dir->i_sb->s_dirt = 1;
2473 btrfs_update_inode_block_group(trans, inode);
2474 btrfs_update_inode_block_group(trans, dir);
2475 out_unlock:
2476 nr = trans->blocks_used;
2477 btrfs_end_transaction_throttle(trans, root);
2478 fail:
2479 if (drop_inode) {
2480 inode_dec_link_count(inode);
2481 iput(inode);
2482 }
2483 btrfs_btree_balance_dirty(root, nr);
2484 return err;
2485 }
2486
2487 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
2488 struct dentry *dentry)
2489 {
2490 struct btrfs_trans_handle *trans;
2491 struct btrfs_root *root = BTRFS_I(dir)->root;
2492 struct inode *inode = old_dentry->d_inode;
2493 u64 index;
2494 unsigned long nr = 0;
2495 int err;
2496 int drop_inode = 0;
2497
2498 if (inode->i_nlink == 0)
2499 return -ENOENT;
2500
2501 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2502 inode->i_nlink++;
2503 #else
2504 inc_nlink(inode);
2505 #endif
2506 err = btrfs_check_free_space(root, 1, 0);
2507 if (err)
2508 goto fail;
2509 err = btrfs_set_inode_index(dir, inode, &index);
2510 if (err)
2511 goto fail;
2512
2513 trans = btrfs_start_transaction(root, 1);
2514
2515 btrfs_set_trans_block_group(trans, dir);
2516 atomic_inc(&inode->i_count);
2517
2518 err = btrfs_add_nondir(trans, dentry, inode, 1, index);
2519
2520 if (err)
2521 drop_inode = 1;
2522
2523 dir->i_sb->s_dirt = 1;
2524 btrfs_update_inode_block_group(trans, dir);
2525 err = btrfs_update_inode(trans, root, inode);
2526
2527 if (err)
2528 drop_inode = 1;
2529
2530 nr = trans->blocks_used;
2531 btrfs_end_transaction_throttle(trans, root);
2532 fail:
2533 if (drop_inode) {
2534 inode_dec_link_count(inode);
2535 iput(inode);
2536 }
2537 btrfs_btree_balance_dirty(root, nr);
2538 return err;
2539 }
2540
2541 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
2542 {
2543 struct inode *inode = NULL;
2544 struct btrfs_trans_handle *trans;
2545 struct btrfs_root *root = BTRFS_I(dir)->root;
2546 int err = 0;
2547 int drop_on_err = 0;
2548 u64 objectid = 0;
2549 u64 index = 0;
2550 unsigned long nr = 1;
2551
2552 err = btrfs_check_free_space(root, 1, 0);
2553 if (err)
2554 goto out_unlock;
2555
2556 trans = btrfs_start_transaction(root, 1);
2557 btrfs_set_trans_block_group(trans, dir);
2558
2559 if (IS_ERR(trans)) {
2560 err = PTR_ERR(trans);
2561 goto out_unlock;
2562 }
2563
2564 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2565 if (err) {
2566 err = -ENOSPC;
2567 goto out_unlock;
2568 }
2569
2570 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
2571 dentry->d_name.len,
2572 dentry->d_parent->d_inode->i_ino, objectid,
2573 BTRFS_I(dir)->block_group, S_IFDIR | mode,
2574 &index);
2575 if (IS_ERR(inode)) {
2576 err = PTR_ERR(inode);
2577 goto out_fail;
2578 }
2579
2580 drop_on_err = 1;
2581
2582 err = btrfs_init_acl(inode, dir);
2583 if (err)
2584 goto out_fail;
2585
2586 inode->i_op = &btrfs_dir_inode_operations;
2587 inode->i_fop = &btrfs_dir_file_operations;
2588 btrfs_set_trans_block_group(trans, inode);
2589
2590 btrfs_i_size_write(inode, 0);
2591 err = btrfs_update_inode(trans, root, inode);
2592 if (err)
2593 goto out_fail;
2594
2595 err = btrfs_add_link(trans, dentry, inode, 0, index);
2596 if (err)
2597 goto out_fail;
2598
2599 d_instantiate(dentry, inode);
2600 drop_on_err = 0;
2601 dir->i_sb->s_dirt = 1;
2602 btrfs_update_inode_block_group(trans, inode);
2603 btrfs_update_inode_block_group(trans, dir);
2604
2605 out_fail:
2606 nr = trans->blocks_used;
2607 btrfs_end_transaction_throttle(trans, root);
2608
2609 out_unlock:
2610 if (drop_on_err)
2611 iput(inode);
2612 btrfs_btree_balance_dirty(root, nr);
2613 return err;
2614 }
2615
2616 static int merge_extent_mapping(struct extent_map_tree *em_tree,
2617 struct extent_map *existing,
2618 struct extent_map *em,
2619 u64 map_start, u64 map_len)
2620 {
2621 u64 start_diff;
2622
2623 BUG_ON(map_start < em->start || map_start >= extent_map_end(em));
2624 start_diff = map_start - em->start;
2625 em->start = map_start;
2626 em->len = map_len;
2627 if (em->block_start < EXTENT_MAP_LAST_BYTE)
2628 em->block_start += start_diff;
2629 return add_extent_mapping(em_tree, em);
2630 }
2631
2632 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
2633 size_t pg_offset, u64 start, u64 len,
2634 int create)
2635 {
2636 int ret;
2637 int err = 0;
2638 u64 bytenr;
2639 u64 extent_start = 0;
2640 u64 extent_end = 0;
2641 u64 objectid = inode->i_ino;
2642 u32 found_type;
2643 struct btrfs_path *path = NULL;
2644 struct btrfs_root *root = BTRFS_I(inode)->root;
2645 struct btrfs_file_extent_item *item;
2646 struct extent_buffer *leaf;
2647 struct btrfs_key found_key;
2648 struct extent_map *em = NULL;
2649 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2650 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2651 struct btrfs_trans_handle *trans = NULL;
2652
2653 again:
2654 spin_lock(&em_tree->lock);
2655 em = lookup_extent_mapping(em_tree, start, len);
2656 if (em)
2657 em->bdev = root->fs_info->fs_devices->latest_bdev;
2658 spin_unlock(&em_tree->lock);
2659
2660 if (em) {
2661 if (em->start > start || em->start + em->len <= start)
2662 free_extent_map(em);
2663 else if (em->block_start == EXTENT_MAP_INLINE && page)
2664 free_extent_map(em);
2665 else
2666 goto out;
2667 }
2668 em = alloc_extent_map(GFP_NOFS);
2669 if (!em) {
2670 err = -ENOMEM;
2671 goto out;
2672 }
2673 em->bdev = root->fs_info->fs_devices->latest_bdev;
2674 em->start = EXTENT_MAP_HOLE;
2675 em->len = (u64)-1;
2676
2677 if (!path) {
2678 path = btrfs_alloc_path();
2679 BUG_ON(!path);
2680 }
2681
2682 ret = btrfs_lookup_file_extent(trans, root, path,
2683 objectid, start, trans != NULL);
2684 if (ret < 0) {
2685 err = ret;
2686 goto out;
2687 }
2688
2689 if (ret != 0) {
2690 if (path->slots[0] == 0)
2691 goto not_found;
2692 path->slots[0]--;
2693 }
2694
2695 leaf = path->nodes[0];
2696 item = btrfs_item_ptr(leaf, path->slots[0],
2697 struct btrfs_file_extent_item);
2698 /* are we inside the extent that was found? */
2699 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2700 found_type = btrfs_key_type(&found_key);
2701 if (found_key.objectid != objectid ||
2702 found_type != BTRFS_EXTENT_DATA_KEY) {
2703 goto not_found;
2704 }
2705
2706 found_type = btrfs_file_extent_type(leaf, item);
2707 extent_start = found_key.offset;
2708 if (found_type == BTRFS_FILE_EXTENT_REG) {
2709 extent_end = extent_start +
2710 btrfs_file_extent_num_bytes(leaf, item);
2711 err = 0;
2712 if (start < extent_start || start >= extent_end) {
2713 em->start = start;
2714 if (start < extent_start) {
2715 if (start + len <= extent_start)
2716 goto not_found;
2717 em->len = extent_end - extent_start;
2718 } else {
2719 em->len = len;
2720 }
2721 goto not_found_em;
2722 }
2723 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
2724 if (bytenr == 0) {
2725 em->start = extent_start;
2726 em->len = extent_end - extent_start;
2727 em->block_start = EXTENT_MAP_HOLE;
2728 goto insert;
2729 }
2730 bytenr += btrfs_file_extent_offset(leaf, item);
2731 em->block_start = bytenr;
2732 em->start = extent_start;
2733 em->len = extent_end - extent_start;
2734 goto insert;
2735 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
2736 u64 page_start;
2737 unsigned long ptr;
2738 char *map;
2739 size_t size;
2740 size_t extent_offset;
2741 size_t copy_size;
2742
2743 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
2744 path->slots[0]));
2745 extent_end = (extent_start + size + root->sectorsize - 1) &
2746 ~((u64)root->sectorsize - 1);
2747 if (start < extent_start || start >= extent_end) {
2748 em->start = start;
2749 if (start < extent_start) {
2750 if (start + len <= extent_start)
2751 goto not_found;
2752 em->len = extent_end - extent_start;
2753 } else {
2754 em->len = len;
2755 }
2756 goto not_found_em;
2757 }
2758 em->block_start = EXTENT_MAP_INLINE;
2759
2760 if (!page) {
2761 em->start = extent_start;
2762 em->len = size;
2763 goto out;
2764 }
2765
2766 page_start = page_offset(page) + pg_offset;
2767 extent_offset = page_start - extent_start;
2768 copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
2769 size - extent_offset);
2770 em->start = extent_start + extent_offset;
2771 em->len = (copy_size + root->sectorsize - 1) &
2772 ~((u64)root->sectorsize - 1);
2773 map = kmap(page);
2774 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
2775 if (create == 0 && !PageUptodate(page)) {
2776 read_extent_buffer(leaf, map + pg_offset, ptr,
2777 copy_size);
2778 flush_dcache_page(page);
2779 } else if (create && PageUptodate(page)) {
2780 if (!trans) {
2781 kunmap(page);
2782 free_extent_map(em);
2783 em = NULL;
2784 btrfs_release_path(root, path);
2785 trans = btrfs_join_transaction(root, 1);
2786 goto again;
2787 }
2788 write_extent_buffer(leaf, map + pg_offset, ptr,
2789 copy_size);
2790 btrfs_mark_buffer_dirty(leaf);
2791 }
2792 kunmap(page);
2793 set_extent_uptodate(io_tree, em->start,
2794 extent_map_end(em) - 1, GFP_NOFS);
2795 goto insert;
2796 } else {
2797 printk("unkknown found_type %d\n", found_type);
2798 WARN_ON(1);
2799 }
2800 not_found:
2801 em->start = start;
2802 em->len = len;
2803 not_found_em:
2804 em->block_start = EXTENT_MAP_HOLE;
2805 insert:
2806 btrfs_release_path(root, path);
2807 if (em->start > start || extent_map_end(em) <= start) {
2808 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len);
2809 err = -EIO;
2810 goto out;
2811 }
2812
2813 err = 0;
2814 spin_lock(&em_tree->lock);
2815 ret = add_extent_mapping(em_tree, em);
2816 /* it is possible that someone inserted the extent into the tree
2817 * while we had the lock dropped. It is also possible that
2818 * an overlapping map exists in the tree
2819 */
2820 if (ret == -EEXIST) {
2821 struct extent_map *existing;
2822
2823 ret = 0;
2824
2825 existing = lookup_extent_mapping(em_tree, start, len);
2826 if (existing && (existing->start > start ||
2827 existing->start + existing->len <= start)) {
2828 free_extent_map(existing);
2829 existing = NULL;
2830 }
2831 if (!existing) {
2832 existing = lookup_extent_mapping(em_tree, em->start,
2833 em->len);
2834 if (existing) {
2835 err = merge_extent_mapping(em_tree, existing,
2836 em, start,
2837 root->sectorsize);
2838 free_extent_map(existing);
2839 if (err) {
2840 free_extent_map(em);
2841 em = NULL;
2842 }
2843 } else {
2844 err = -EIO;
2845 printk("failing to insert %Lu %Lu\n",
2846 start, len);
2847 free_extent_map(em);
2848 em = NULL;
2849 }
2850 } else {
2851 free_extent_map(em);
2852 em = existing;
2853 err = 0;
2854 }
2855 }
2856 spin_unlock(&em_tree->lock);
2857 out:
2858 if (path)
2859 btrfs_free_path(path);
2860 if (trans) {
2861 ret = btrfs_end_transaction(trans, root);
2862 if (!err) {
2863 err = ret;
2864 }
2865 }
2866 if (err) {
2867 free_extent_map(em);
2868 WARN_ON(1);
2869 return ERR_PTR(err);
2870 }
2871 return em;
2872 }
2873
2874 #if 0 /* waiting for O_DIRECT reads */
2875 static int btrfs_get_block(struct inode *inode, sector_t iblock,
2876 struct buffer_head *bh_result, int create)
2877 {
2878 struct extent_map *em;
2879 u64 start = (u64)iblock << inode->i_blkbits;
2880 struct btrfs_multi_bio *multi = NULL;
2881 struct btrfs_root *root = BTRFS_I(inode)->root;
2882 u64 len;
2883 u64 logical;
2884 u64 map_length;
2885 int ret = 0;
2886
2887 em = btrfs_get_extent(inode, NULL, 0, start, bh_result->b_size, 0);
2888
2889 if (!em || IS_ERR(em))
2890 goto out;
2891
2892 if (em->start > start || em->start + em->len <= start) {
2893 goto out;
2894 }
2895
2896 if (em->block_start == EXTENT_MAP_INLINE) {
2897 ret = -EINVAL;
2898 goto out;
2899 }
2900
2901 len = em->start + em->len - start;
2902 len = min_t(u64, len, INT_LIMIT(typeof(bh_result->b_size)));
2903
2904 if (em->block_start == EXTENT_MAP_HOLE ||
2905 em->block_start == EXTENT_MAP_DELALLOC) {
2906 bh_result->b_size = len;
2907 goto out;
2908 }
2909
2910 logical = start - em->start;
2911 logical = em->block_start + logical;
2912
2913 map_length = len;
2914 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
2915 logical, &map_length, &multi, 0);
2916 BUG_ON(ret);
2917 bh_result->b_blocknr = multi->stripes[0].physical >> inode->i_blkbits;
2918 bh_result->b_size = min(map_length, len);
2919
2920 bh_result->b_bdev = multi->stripes[0].dev->bdev;
2921 set_buffer_mapped(bh_result);
2922 kfree(multi);
2923 out:
2924 free_extent_map(em);
2925 return ret;
2926 }
2927 #endif
2928
2929 static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
2930 const struct iovec *iov, loff_t offset,
2931 unsigned long nr_segs)
2932 {
2933 return -EINVAL;
2934 #if 0
2935 struct file *file = iocb->ki_filp;
2936 struct inode *inode = file->f_mapping->host;
2937
2938 if (rw == WRITE)
2939 return -EINVAL;
2940
2941 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2942 offset, nr_segs, btrfs_get_block, NULL);
2943 #endif
2944 }
2945
2946 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
2947 {
2948 return extent_bmap(mapping, iblock, btrfs_get_extent);
2949 }
2950
2951 int btrfs_readpage(struct file *file, struct page *page)
2952 {
2953 struct extent_io_tree *tree;
2954 tree = &BTRFS_I(page->mapping->host)->io_tree;
2955 return extent_read_full_page(tree, page, btrfs_get_extent);
2956 }
2957
2958 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
2959 {
2960 struct extent_io_tree *tree;
2961
2962
2963 if (current->flags & PF_MEMALLOC) {
2964 redirty_page_for_writepage(wbc, page);
2965 unlock_page(page);
2966 return 0;
2967 }
2968 tree = &BTRFS_I(page->mapping->host)->io_tree;
2969 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
2970 }
2971
2972 int btrfs_writepages(struct address_space *mapping,
2973 struct writeback_control *wbc)
2974 {
2975 struct extent_io_tree *tree;
2976 tree = &BTRFS_I(mapping->host)->io_tree;
2977 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
2978 }
2979
2980 static int
2981 btrfs_readpages(struct file *file, struct address_space *mapping,
2982 struct list_head *pages, unsigned nr_pages)
2983 {
2984 struct extent_io_tree *tree;
2985 tree = &BTRFS_I(mapping->host)->io_tree;
2986 return extent_readpages(tree, mapping, pages, nr_pages,
2987 btrfs_get_extent);
2988 }
2989 static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
2990 {
2991 struct extent_io_tree *tree;
2992 struct extent_map_tree *map;
2993 int ret;
2994
2995 tree = &BTRFS_I(page->mapping->host)->io_tree;
2996 map = &BTRFS_I(page->mapping->host)->extent_tree;
2997 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
2998 if (ret == 1) {
2999 ClearPagePrivate(page);
3000 set_page_private(page, 0);
3001 page_cache_release(page);
3002 }
3003 return ret;
3004 }
3005
3006 static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
3007 {
3008 return __btrfs_releasepage(page, gfp_flags);
3009 }
3010
3011 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
3012 {
3013 struct extent_io_tree *tree;
3014 struct btrfs_ordered_extent *ordered;
3015 u64 page_start = page_offset(page);
3016 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
3017
3018 wait_on_page_writeback(page);
3019 tree = &BTRFS_I(page->mapping->host)->io_tree;
3020 if (offset) {
3021 btrfs_releasepage(page, GFP_NOFS);
3022 return;
3023 }
3024
3025 lock_extent(tree, page_start, page_end, GFP_NOFS);
3026 ordered = btrfs_lookup_ordered_extent(page->mapping->host,
3027 page_offset(page));
3028 if (ordered) {
3029 /*
3030 * IO on this page will never be started, so we need
3031 * to account for any ordered extents now
3032 */
3033 clear_extent_bit(tree, page_start, page_end,
3034 EXTENT_DIRTY | EXTENT_DELALLOC |
3035 EXTENT_LOCKED, 1, 0, GFP_NOFS);
3036 btrfs_finish_ordered_io(page->mapping->host,
3037 page_start, page_end);
3038 btrfs_put_ordered_extent(ordered);
3039 lock_extent(tree, page_start, page_end, GFP_NOFS);
3040 }
3041 clear_extent_bit(tree, page_start, page_end,
3042 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
3043 EXTENT_ORDERED,
3044 1, 1, GFP_NOFS);
3045 __btrfs_releasepage(page, GFP_NOFS);
3046
3047 ClearPageChecked(page);
3048 if (PagePrivate(page)) {
3049 ClearPagePrivate(page);
3050 set_page_private(page, 0);
3051 page_cache_release(page);
3052 }
3053 }
3054
3055 /*
3056 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
3057 * called from a page fault handler when a page is first dirtied. Hence we must
3058 * be careful to check for EOF conditions here. We set the page up correctly
3059 * for a written page which means we get ENOSPC checking when writing into
3060 * holes and correct delalloc and unwritten extent mapping on filesystems that
3061 * support these features.
3062 *
3063 * We are not allowed to take the i_mutex here so we have to play games to
3064 * protect against truncate races as the page could now be beyond EOF. Because
3065 * vmtruncate() writes the inode size before removing pages, once we have the
3066 * page lock we can determine safely if the page is beyond EOF. If it is not
3067 * beyond EOF, then the page is guaranteed safe against truncation until we
3068 * unlock the page.
3069 */
3070 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
3071 {
3072 struct inode *inode = fdentry(vma->vm_file)->d_inode;
3073 struct btrfs_root *root = BTRFS_I(inode)->root;
3074 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
3075 struct btrfs_ordered_extent *ordered;
3076 char *kaddr;
3077 unsigned long zero_start;
3078 loff_t size;
3079 int ret;
3080 u64 page_start;
3081 u64 page_end;
3082
3083 ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
3084 if (ret)
3085 goto out;
3086
3087 ret = -EINVAL;
3088 again:
3089 lock_page(page);
3090 size = i_size_read(inode);
3091 page_start = page_offset(page);
3092 page_end = page_start + PAGE_CACHE_SIZE - 1;
3093
3094 if ((page->mapping != inode->i_mapping) ||
3095 (page_start >= size)) {
3096 /* page got truncated out from underneath us */
3097 goto out_unlock;
3098 }
3099 wait_on_page_writeback(page);
3100
3101 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
3102 set_page_extent_mapped(page);
3103
3104 /*
3105 * we can't set the delalloc bits if there are pending ordered
3106 * extents. Drop our locks and wait for them to finish
3107 */
3108 ordered = btrfs_lookup_ordered_extent(inode, page_start);
3109 if (ordered) {
3110 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
3111 unlock_page(page);
3112 btrfs_start_ordered_extent(inode, ordered, 1);
3113 btrfs_put_ordered_extent(ordered);
3114 goto again;
3115 }
3116
3117 btrfs_set_extent_delalloc(inode, page_start, page_end);
3118 ret = 0;
3119
3120 /* page is wholly or partially inside EOF */
3121 if (page_start + PAGE_CACHE_SIZE > size)
3122 zero_start = size & ~PAGE_CACHE_MASK;
3123 else
3124 zero_start = PAGE_CACHE_SIZE;
3125
3126 if (zero_start != PAGE_CACHE_SIZE) {
3127 kaddr = kmap(page);
3128 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
3129 flush_dcache_page(page);
3130 kunmap(page);
3131 }
3132 ClearPageChecked(page);
3133 set_page_dirty(page);
3134 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
3135
3136 out_unlock:
3137 unlock_page(page);
3138 out:
3139 return ret;
3140 }
3141
3142 static void btrfs_truncate(struct inode *inode)
3143 {
3144 struct btrfs_root *root = BTRFS_I(inode)->root;
3145 int ret;
3146 struct btrfs_trans_handle *trans;
3147 unsigned long nr;
3148 u64 mask = root->sectorsize - 1;
3149
3150 if (!S_ISREG(inode->i_mode))
3151 return;
3152 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
3153 return;
3154
3155 btrfs_truncate_page(inode->i_mapping, inode->i_size);
3156 btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
3157
3158 trans = btrfs_start_transaction(root, 1);
3159 btrfs_set_trans_block_group(trans, inode);
3160 btrfs_i_size_write(inode, inode->i_size);
3161
3162 ret = btrfs_orphan_add(trans, inode);
3163 if (ret)
3164 goto out;
3165 /* FIXME, add redo link to tree so we don't leak on crash */
3166 ret = btrfs_truncate_in_trans(trans, root, inode,
3167 BTRFS_EXTENT_DATA_KEY);
3168 btrfs_update_inode(trans, root, inode);
3169
3170 ret = btrfs_orphan_del(trans, inode);
3171 BUG_ON(ret);
3172
3173 out:
3174 nr = trans->blocks_used;
3175 ret = btrfs_end_transaction_throttle(trans, root);
3176 BUG_ON(ret);
3177 btrfs_btree_balance_dirty(root, nr);
3178 }
3179
3180 /*
3181 * Invalidate a single dcache entry at the root of the filesystem.
3182 * Needed after creation of snapshot or subvolume.
3183 */
3184 void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name,
3185 int namelen)
3186 {
3187 struct dentry *alias, *entry;
3188 struct qstr qstr;
3189
3190 alias = d_find_alias(root->fs_info->sb->s_root->d_inode);
3191 if (alias) {
3192 qstr.name = name;
3193 qstr.len = namelen;
3194 /* change me if btrfs ever gets a d_hash operation */
3195 qstr.hash = full_name_hash(qstr.name, qstr.len);
3196 entry = d_lookup(alias, &qstr);
3197 dput(alias);
3198 if (entry) {
3199 d_invalidate(entry);
3200 dput(entry);
3201 }
3202 }
3203 }
3204
3205 int btrfs_create_subvol_root(struct btrfs_root *new_root,
3206 struct btrfs_trans_handle *trans, u64 new_dirid,
3207 struct btrfs_block_group_cache *block_group)
3208 {
3209 struct inode *inode;
3210 u64 index = 0;
3211
3212 inode = btrfs_new_inode(trans, new_root, NULL, "..", 2, new_dirid,
3213 new_dirid, block_group, S_IFDIR | 0700, &index);
3214 if (IS_ERR(inode))
3215 return PTR_ERR(inode);
3216 inode->i_op = &btrfs_dir_inode_operations;
3217 inode->i_fop = &btrfs_dir_file_operations;
3218 new_root->inode = inode;
3219
3220 inode->i_nlink = 1;
3221 btrfs_i_size_write(inode, 0);
3222
3223 return btrfs_update_inode(trans, new_root, inode);
3224 }
3225
3226 unsigned long btrfs_force_ra(struct address_space *mapping,
3227 struct file_ra_state *ra, struct file *file,
3228 pgoff_t offset, pgoff_t last_index)
3229 {
3230 pgoff_t req_size = last_index - offset + 1;
3231
3232 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3233 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
3234 return offset;
3235 #else
3236 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
3237 return offset + req_size;
3238 #endif
3239 }
3240
3241 struct inode *btrfs_alloc_inode(struct super_block *sb)
3242 {
3243 struct btrfs_inode *ei;
3244
3245 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
3246 if (!ei)
3247 return NULL;
3248 ei->last_trans = 0;
3249 btrfs_ordered_inode_tree_init(&ei->ordered_tree);
3250 ei->i_acl = BTRFS_ACL_NOT_CACHED;
3251 ei->i_default_acl = BTRFS_ACL_NOT_CACHED;
3252 INIT_LIST_HEAD(&ei->i_orphan);
3253 return &ei->vfs_inode;
3254 }
3255
3256 void btrfs_destroy_inode(struct inode *inode)
3257 {
3258 struct btrfs_ordered_extent *ordered;
3259 WARN_ON(!list_empty(&inode->i_dentry));
3260 WARN_ON(inode->i_data.nrpages);
3261
3262 if (BTRFS_I(inode)->i_acl &&
3263 BTRFS_I(inode)->i_acl != BTRFS_ACL_NOT_CACHED)
3264 posix_acl_release(BTRFS_I(inode)->i_acl);
3265 if (BTRFS_I(inode)->i_default_acl &&
3266 BTRFS_I(inode)->i_default_acl != BTRFS_ACL_NOT_CACHED)
3267 posix_acl_release(BTRFS_I(inode)->i_default_acl);
3268
3269 spin_lock(&BTRFS_I(inode)->root->list_lock);
3270 if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
3271 printk(KERN_ERR "BTRFS: inode %lu: inode still on the orphan"
3272 " list\n", inode->i_ino);
3273 dump_stack();
3274 }
3275 spin_unlock(&BTRFS_I(inode)->root->list_lock);
3276
3277 while(1) {
3278 ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
3279 if (!ordered)
3280 break;
3281 else {
3282 printk("found ordered extent %Lu %Lu\n",
3283 ordered->file_offset, ordered->len);
3284 btrfs_remove_ordered_extent(inode, ordered);
3285 btrfs_put_ordered_extent(ordered);
3286 btrfs_put_ordered_extent(ordered);
3287 }
3288 }
3289 btrfs_drop_extent_cache(inode, 0, (u64)-1);
3290 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
3291 }
3292
3293 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3294 static void init_once(void *foo)
3295 #elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3296 static void init_once(struct kmem_cache * cachep, void *foo)
3297 #else
3298 static void init_once(void * foo, struct kmem_cache * cachep,
3299 unsigned long flags)
3300 #endif
3301 {
3302 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
3303
3304 inode_init_once(&ei->vfs_inode);
3305 }
3306
3307 void btrfs_destroy_cachep(void)
3308 {
3309 if (btrfs_inode_cachep)
3310 kmem_cache_destroy(btrfs_inode_cachep);
3311 if (btrfs_trans_handle_cachep)
3312 kmem_cache_destroy(btrfs_trans_handle_cachep);
3313 if (btrfs_transaction_cachep)
3314 kmem_cache_destroy(btrfs_transaction_cachep);
3315 if (btrfs_bit_radix_cachep)
3316 kmem_cache_destroy(btrfs_bit_radix_cachep);
3317 if (btrfs_path_cachep)
3318 kmem_cache_destroy(btrfs_path_cachep);
3319 }
3320
3321 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
3322 unsigned long extra_flags,
3323 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3324 void (*ctor)(void *)
3325 #elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
3326 void (*ctor)(struct kmem_cache *, void *)
3327 #else
3328 void (*ctor)(void *, struct kmem_cache *,
3329 unsigned long)
3330 #endif
3331 )
3332 {
3333 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
3334 SLAB_MEM_SPREAD | extra_flags), ctor
3335 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3336 ,NULL
3337 #endif
3338 );
3339 }
3340
3341 int btrfs_init_cachep(void)
3342 {
3343 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
3344 sizeof(struct btrfs_inode),
3345 0, init_once);
3346 if (!btrfs_inode_cachep)
3347 goto fail;
3348 btrfs_trans_handle_cachep =
3349 btrfs_cache_create("btrfs_trans_handle_cache",
3350 sizeof(struct btrfs_trans_handle),
3351 0, NULL);
3352 if (!btrfs_trans_handle_cachep)
3353 goto fail;
3354 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
3355 sizeof(struct btrfs_transaction),
3356 0, NULL);
3357 if (!btrfs_transaction_cachep)
3358 goto fail;
3359 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
3360 sizeof(struct btrfs_path),
3361 0, NULL);
3362 if (!btrfs_path_cachep)
3363 goto fail;
3364 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
3365 SLAB_DESTROY_BY_RCU, NULL);
3366 if (!btrfs_bit_radix_cachep)
3367 goto fail;
3368 return 0;
3369 fail:
3370 btrfs_destroy_cachep();
3371 return -ENOMEM;
3372 }
3373
3374 static int btrfs_getattr(struct vfsmount *mnt,
3375 struct dentry *dentry, struct kstat *stat)
3376 {
3377 struct inode *inode = dentry->d_inode;
3378 generic_fillattr(inode, stat);
3379 stat->blksize = PAGE_CACHE_SIZE;
3380 stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9);
3381 return 0;
3382 }
3383
3384 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
3385 struct inode * new_dir,struct dentry *new_dentry)
3386 {
3387 struct btrfs_trans_handle *trans;
3388 struct btrfs_root *root = BTRFS_I(old_dir)->root;
3389 struct inode *new_inode = new_dentry->d_inode;
3390 struct inode *old_inode = old_dentry->d_inode;
3391 struct timespec ctime = CURRENT_TIME;
3392 u64 index = 0;
3393 int ret;
3394
3395 if (S_ISDIR(old_inode->i_mode) && new_inode &&
3396 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
3397 return -ENOTEMPTY;
3398 }
3399
3400 ret = btrfs_check_free_space(root, 1, 0);
3401 if (ret)
3402 goto out_unlock;
3403
3404 trans = btrfs_start_transaction(root, 1);
3405
3406 btrfs_set_trans_block_group(trans, new_dir);
3407
3408 old_dentry->d_inode->i_nlink++;
3409 old_dir->i_ctime = old_dir->i_mtime = ctime;
3410 new_dir->i_ctime = new_dir->i_mtime = ctime;
3411 old_inode->i_ctime = ctime;
3412
3413 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
3414 if (ret)
3415 goto out_fail;
3416
3417 if (new_inode) {
3418 new_inode->i_ctime = CURRENT_TIME;
3419 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
3420 if (ret)
3421 goto out_fail;
3422 if (new_inode->i_nlink == 0) {
3423 ret = btrfs_orphan_add(trans, new_inode);
3424 if (ret)
3425 goto out_fail;
3426 }
3427 }
3428 ret = btrfs_set_inode_index(new_dir, old_inode, &index);
3429 if (ret)
3430 goto out_fail;
3431
3432 ret = btrfs_add_link(trans, new_dentry, old_inode, 1, index);
3433 if (ret)
3434 goto out_fail;
3435
3436 out_fail:
3437 btrfs_end_transaction_throttle(trans, root);
3438 out_unlock:
3439 return ret;
3440 }
3441
3442 int btrfs_start_delalloc_inodes(struct btrfs_root *root)
3443 {
3444 struct list_head *head = &root->fs_info->delalloc_inodes;
3445 struct btrfs_inode *binode;
3446 unsigned long flags;
3447
3448 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
3449 while(!list_empty(head)) {
3450 binode = list_entry(head->next, struct btrfs_inode,
3451 delalloc_inodes);
3452 atomic_inc(&binode->vfs_inode.i_count);
3453 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
3454 filemap_write_and_wait(binode->vfs_inode.i_mapping);
3455 iput(&binode->vfs_inode);
3456 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
3457 }
3458 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
3459 return 0;
3460 }
3461
3462 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
3463 const char *symname)
3464 {
3465 struct btrfs_trans_handle *trans;
3466 struct btrfs_root *root = BTRFS_I(dir)->root;
3467 struct btrfs_path *path;
3468 struct btrfs_key key;
3469 struct inode *inode = NULL;
3470 int err;
3471 int drop_inode = 0;
3472 u64 objectid;
3473 u64 index = 0 ;
3474 int name_len;
3475 int datasize;
3476 unsigned long ptr;
3477 struct btrfs_file_extent_item *ei;
3478 struct extent_buffer *leaf;
3479 unsigned long nr = 0;
3480
3481 name_len = strlen(symname) + 1;
3482 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
3483 return -ENAMETOOLONG;
3484
3485 err = btrfs_check_free_space(root, 1, 0);
3486 if (err)
3487 goto out_fail;
3488
3489 trans = btrfs_start_transaction(root, 1);
3490 btrfs_set_trans_block_group(trans, dir);
3491
3492 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
3493 if (err) {
3494 err = -ENOSPC;
3495 goto out_unlock;
3496 }
3497
3498 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
3499 dentry->d_name.len,
3500 dentry->d_parent->d_inode->i_ino, objectid,
3501 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO,
3502 &index);
3503 err = PTR_ERR(inode);
3504 if (IS_ERR(inode))
3505 goto out_unlock;
3506
3507 err = btrfs_init_acl(inode, dir);
3508 if (err) {
3509 drop_inode = 1;
3510 goto out_unlock;
3511 }
3512
3513 btrfs_set_trans_block_group(trans, inode);
3514 err = btrfs_add_nondir(trans, dentry, inode, 0, index);
3515 if (err)
3516 drop_inode = 1;
3517 else {
3518 inode->i_mapping->a_ops = &btrfs_aops;
3519 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3520 inode->i_fop = &btrfs_file_operations;
3521 inode->i_op = &btrfs_file_inode_operations;
3522 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
3523 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
3524 inode->i_mapping, GFP_NOFS);
3525 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
3526 inode->i_mapping, GFP_NOFS);
3527 INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
3528 mutex_init(&BTRFS_I(inode)->csum_mutex);
3529 mutex_init(&BTRFS_I(inode)->extent_mutex);
3530 BTRFS_I(inode)->delalloc_bytes = 0;
3531 BTRFS_I(inode)->disk_i_size = 0;
3532 inode->i_mapping->writeback_index = 0;
3533 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
3534 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
3535 }
3536 dir->i_sb->s_dirt = 1;
3537 btrfs_update_inode_block_group(trans, inode);
3538 btrfs_update_inode_block_group(trans, dir);
3539 if (drop_inode)
3540 goto out_unlock;
3541
3542 path = btrfs_alloc_path();
3543 BUG_ON(!path);
3544 key.objectid = inode->i_ino;
3545 key.offset = 0;
3546 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
3547 datasize = btrfs_file_extent_calc_inline_size(name_len);
3548 err = btrfs_insert_empty_item(trans, root, path, &key,
3549 datasize);
3550 if (err) {
3551 drop_inode = 1;
3552 goto out_unlock;
3553 }
3554 leaf = path->nodes[0];
3555 ei = btrfs_item_ptr(leaf, path->slots[0],
3556 struct btrfs_file_extent_item);
3557 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
3558 btrfs_set_file_extent_type(leaf, ei,
3559 BTRFS_FILE_EXTENT_INLINE);
3560 ptr = btrfs_file_extent_inline_start(ei);
3561 write_extent_buffer(leaf, symname, ptr, name_len);
3562 btrfs_mark_buffer_dirty(leaf);
3563 btrfs_free_path(path);
3564
3565 inode->i_op = &btrfs_symlink_inode_operations;
3566 inode->i_mapping->a_ops = &btrfs_symlink_aops;
3567 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3568 btrfs_i_size_write(inode, name_len - 1);
3569 err = btrfs_update_inode(trans, root, inode);
3570 if (err)
3571 drop_inode = 1;
3572
3573 out_unlock:
3574 nr = trans->blocks_used;
3575 btrfs_end_transaction_throttle(trans, root);
3576 out_fail:
3577 if (drop_inode) {
3578 inode_dec_link_count(inode);
3579 iput(inode);
3580 }
3581 btrfs_btree_balance_dirty(root, nr);
3582 return err;
3583 }
3584
3585 static int btrfs_set_page_dirty(struct page *page)
3586 {
3587 return __set_page_dirty_nobuffers(page);
3588 }
3589
3590 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3591 static int btrfs_permission(struct inode *inode, int mask)
3592 #else
3593 static int btrfs_permission(struct inode *inode, int mask,
3594 struct nameidata *nd)
3595 #endif
3596 {
3597 if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
3598 return -EACCES;
3599 return generic_permission(inode, mask, btrfs_check_acl);
3600 }
3601
3602 static struct inode_operations btrfs_dir_inode_operations = {
3603 .lookup = btrfs_lookup,
3604 .create = btrfs_create,
3605 .unlink = btrfs_unlink,
3606 .link = btrfs_link,
3607 .mkdir = btrfs_mkdir,
3608 .rmdir = btrfs_rmdir,
3609 .rename = btrfs_rename,
3610 .symlink = btrfs_symlink,
3611 .setattr = btrfs_setattr,
3612 .mknod = btrfs_mknod,
3613 .setxattr = generic_setxattr,
3614 .getxattr = generic_getxattr,
3615 .listxattr = btrfs_listxattr,
3616 .removexattr = generic_removexattr,
3617 .permission = btrfs_permission,
3618 };
3619 static struct inode_operations btrfs_dir_ro_inode_operations = {
3620 .lookup = btrfs_lookup,
3621 .permission = btrfs_permission,
3622 };
3623 static struct file_operations btrfs_dir_file_operations = {
3624 .llseek = generic_file_llseek,
3625 .read = generic_read_dir,
3626 .readdir = btrfs_readdir,
3627 .unlocked_ioctl = btrfs_ioctl,
3628 #ifdef CONFIG_COMPAT
3629 .compat_ioctl = btrfs_ioctl,
3630 #endif
3631 .release = btrfs_release_file,
3632 };
3633
3634 static struct extent_io_ops btrfs_extent_io_ops = {
3635 .fill_delalloc = run_delalloc_range,
3636 .submit_bio_hook = btrfs_submit_bio_hook,
3637 .merge_bio_hook = btrfs_merge_bio_hook,
3638 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
3639 .writepage_end_io_hook = btrfs_writepage_end_io_hook,
3640 .writepage_start_hook = btrfs_writepage_start_hook,
3641 .readpage_io_failed_hook = btrfs_io_failed_hook,
3642 .set_bit_hook = btrfs_set_bit_hook,
3643 .clear_bit_hook = btrfs_clear_bit_hook,
3644 };
3645
3646 static struct address_space_operations btrfs_aops = {
3647 .readpage = btrfs_readpage,
3648 .writepage = btrfs_writepage,
3649 .writepages = btrfs_writepages,
3650 .readpages = btrfs_readpages,
3651 .sync_page = block_sync_page,
3652 .bmap = btrfs_bmap,
3653 .direct_IO = btrfs_direct_IO,
3654 .invalidatepage = btrfs_invalidatepage,
3655 .releasepage = btrfs_releasepage,
3656 .set_page_dirty = btrfs_set_page_dirty,
3657 };
3658
3659 static struct address_space_operations btrfs_symlink_aops = {
3660 .readpage = btrfs_readpage,
3661 .writepage = btrfs_writepage,
3662 .invalidatepage = btrfs_invalidatepage,
3663 .releasepage = btrfs_releasepage,
3664 };
3665
3666 static struct inode_operations btrfs_file_inode_operations = {
3667 .truncate = btrfs_truncate,
3668 .getattr = btrfs_getattr,
3669 .setattr = btrfs_setattr,
3670 .setxattr = generic_setxattr,
3671 .getxattr = generic_getxattr,
3672 .listxattr = btrfs_listxattr,
3673 .removexattr = generic_removexattr,
3674 .permission = btrfs_permission,
3675 };
3676 static struct inode_operations btrfs_special_inode_operations = {
3677 .getattr = btrfs_getattr,
3678 .setattr = btrfs_setattr,
3679 .permission = btrfs_permission,
3680 .setxattr = generic_setxattr,
3681 .getxattr = generic_getxattr,
3682 .listxattr = btrfs_listxattr,
3683 .removexattr = generic_removexattr,
3684 };
3685 static struct inode_operations btrfs_symlink_inode_operations = {
3686 .readlink = generic_readlink,
3687 .follow_link = page_follow_link_light,
3688 .put_link = page_put_link,
3689 .permission = btrfs_permission,
3690 };
Linux kernel - Deliverables
This page took 0.103005 seconds and 6 git commands to generate.