projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Fix ENOTEMPTY check in btrfs_rmdir
[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/buffer_head.h>
20 #include <linux/fs.h>
21 #include <linux/pagemap.h>
22 #include <linux/highmem.h>
23 #include <linux/time.h>
24 #include <linux/init.h>
25 #include <linux/string.h>
26 #include <linux/smp_lock.h>
27 #include <linux/backing-dev.h>
28 #include <linux/mpage.h>
29 #include <linux/swap.h>
30 #include <linux/writeback.h>
31 #include <linux/statfs.h>
32 #include <linux/compat.h>
33 #include <linux/bit_spinlock.h>
34 #include <linux/version.h>
35 #include "ctree.h"
36 #include "disk-io.h"
37 #include "transaction.h"
38 #include "btrfs_inode.h"
39 #include "ioctl.h"
40 #include "print-tree.h"
41
42 struct btrfs_iget_args {
43 u64 ino;
44 struct btrfs_root *root;
45 };
46
47 static struct inode_operations btrfs_dir_inode_operations;
48 static struct inode_operations btrfs_symlink_inode_operations;
49 static struct inode_operations btrfs_dir_ro_inode_operations;
50 static struct inode_operations btrfs_special_inode_operations;
51 static struct inode_operations btrfs_file_inode_operations;
52 static struct address_space_operations btrfs_aops;
53 static struct address_space_operations btrfs_symlink_aops;
54 static struct file_operations btrfs_dir_file_operations;
55 static struct extent_map_ops btrfs_extent_map_ops;
56
57 static struct kmem_cache *btrfs_inode_cachep;
58 struct kmem_cache *btrfs_trans_handle_cachep;
59 struct kmem_cache *btrfs_transaction_cachep;
60 struct kmem_cache *btrfs_bit_radix_cachep;
61 struct kmem_cache *btrfs_path_cachep;
62
63 #define S_SHIFT 12
64 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
65 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
66 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
67 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
68 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
69 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
70 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
71 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
72 };
73
74 static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
75 {
76 struct btrfs_root *root = BTRFS_I(inode)->root;
77 struct btrfs_trans_handle *trans;
78 struct btrfs_key ins;
79 u64 alloc_hint = 0;
80 u64 num_bytes;
81 int ret;
82 u64 blocksize = root->sectorsize;
83
84 mutex_lock(&root->fs_info->fs_mutex);
85 trans = btrfs_start_transaction(root, 1);
86 btrfs_set_trans_block_group(trans, inode);
87 BUG_ON(!trans);
88 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
89 ret = btrfs_drop_extents(trans, root, inode,
90 start, start + num_bytes, start, &alloc_hint);
91
92 ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_bytes, 0,
93 alloc_hint, (u64)-1, &ins, 1);
94 if (ret) {
95 WARN_ON(1);
96 goto out;
97 }
98 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
99 start, ins.objectid, ins.offset,
100 ins.offset);
101 out:
102 btrfs_end_transaction(trans, root);
103 mutex_unlock(&root->fs_info->fs_mutex);
104 return ret;
105 }
106
107 int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end)
108 {
109 struct inode *inode = page->mapping->host;
110 struct btrfs_root *root = BTRFS_I(inode)->root;
111 struct btrfs_trans_handle *trans;
112 char *kaddr;
113 int ret;
114 u64 page_start = page->index << PAGE_CACHE_SHIFT;
115 size_t offset = start - page_start;
116
117 mutex_lock(&root->fs_info->fs_mutex);
118 trans = btrfs_start_transaction(root, 1);
119 btrfs_set_trans_block_group(trans, inode);
120 kaddr = kmap(page);
121 btrfs_csum_file_block(trans, root, inode, inode->i_ino,
122 start, kaddr + offset, end - start + 1);
123 kunmap(page);
124 ret = btrfs_end_transaction(trans, root);
125 BUG_ON(ret);
126 mutex_unlock(&root->fs_info->fs_mutex);
127 return ret;
128 }
129
130 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
131 {
132 int ret = 0;
133 struct inode *inode = page->mapping->host;
134 struct btrfs_root *root = BTRFS_I(inode)->root;
135 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
136 struct btrfs_csum_item *item;
137 struct btrfs_path *path = NULL;
138 u32 csum;
139
140 mutex_lock(&root->fs_info->fs_mutex);
141 path = btrfs_alloc_path();
142 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
143 if (IS_ERR(item)) {
144 ret = PTR_ERR(item);
145 /* a csum that isn't present is a preallocated region. */
146 if (ret == -ENOENT || ret == -EFBIG)
147 ret = 0;
148 csum = 0;
149 goto out;
150 }
151 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
152 BTRFS_CRC32_SIZE);
153 set_state_private(em_tree, start, csum);
154 out:
155 if (path)
156 btrfs_free_path(path);
157 mutex_unlock(&root->fs_info->fs_mutex);
158 return ret;
159 }
160
161 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
162 {
163 size_t offset = start - (page->index << PAGE_CACHE_SHIFT);
164 struct inode *inode = page->mapping->host;
165 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
166 char *kaddr;
167 u64 private;
168 int ret;
169 struct btrfs_root *root = BTRFS_I(inode)->root;
170 u32 csum = ~(u32)0;
171 unsigned long flags;
172
173 ret = get_state_private(em_tree, start, &private);
174 local_irq_save(flags);
175 kaddr = kmap_atomic(page, KM_IRQ0);
176 if (ret) {
177 goto zeroit;
178 }
179 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
180 btrfs_csum_final(csum, (char *)&csum);
181 if (csum != private) {
182 goto zeroit;
183 }
184 kunmap_atomic(kaddr, KM_IRQ0);
185 local_irq_restore(flags);
186 return 0;
187
188 zeroit:
189 printk("btrfs csum failed ino %lu off %llu\n",
190 page->mapping->host->i_ino, (unsigned long long)start);
191 memset(kaddr + offset, 1, end - start + 1);
192 flush_dcache_page(page);
193 kunmap_atomic(kaddr, KM_IRQ0);
194 local_irq_restore(flags);
195 return 0;
196 }
197
198 void btrfs_read_locked_inode(struct inode *inode)
199 {
200 struct btrfs_path *path;
201 struct extent_buffer *leaf;
202 struct btrfs_inode_item *inode_item;
203 struct btrfs_inode_timespec *tspec;
204 struct btrfs_root *root = BTRFS_I(inode)->root;
205 struct btrfs_key location;
206 u64 alloc_group_block;
207 u32 rdev;
208 int ret;
209
210 path = btrfs_alloc_path();
211 BUG_ON(!path);
212 mutex_lock(&root->fs_info->fs_mutex);
213
214 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
215 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
216 if (ret)
217 goto make_bad;
218
219 leaf = path->nodes[0];
220 inode_item = btrfs_item_ptr(leaf, path->slots[0],
221 struct btrfs_inode_item);
222
223 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
224 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
225 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
226 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
227 inode->i_size = btrfs_inode_size(leaf, inode_item);
228
229 tspec = btrfs_inode_atime(inode_item);
230 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
231 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
232
233 tspec = btrfs_inode_mtime(inode_item);
234 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
235 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
236
237 tspec = btrfs_inode_ctime(inode_item);
238 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
239 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
240
241 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
242 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
243 inode->i_rdev = 0;
244 rdev = btrfs_inode_rdev(leaf, inode_item);
245
246 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
247 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
248 alloc_group_block);
249
250 btrfs_free_path(path);
251 inode_item = NULL;
252
253 mutex_unlock(&root->fs_info->fs_mutex);
254
255 switch (inode->i_mode & S_IFMT) {
256 case S_IFREG:
257 inode->i_mapping->a_ops = &btrfs_aops;
258 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
259 inode->i_fop = &btrfs_file_operations;
260 inode->i_op = &btrfs_file_inode_operations;
261 break;
262 case S_IFDIR:
263 inode->i_fop = &btrfs_dir_file_operations;
264 if (root == root->fs_info->tree_root)
265 inode->i_op = &btrfs_dir_ro_inode_operations;
266 else
267 inode->i_op = &btrfs_dir_inode_operations;
268 break;
269 case S_IFLNK:
270 inode->i_op = &btrfs_symlink_inode_operations;
271 inode->i_mapping->a_ops = &btrfs_symlink_aops;
272 break;
273 default:
274 init_special_inode(inode, inode->i_mode, rdev);
275 break;
276 }
277 return;
278
279 make_bad:
280 btrfs_release_path(root, path);
281 btrfs_free_path(path);
282 mutex_unlock(&root->fs_info->fs_mutex);
283 make_bad_inode(inode);
284 }
285
286 static void fill_inode_item(struct extent_buffer *leaf,
287 struct btrfs_inode_item *item,
288 struct inode *inode)
289 {
290 btrfs_set_inode_uid(leaf, item, inode->i_uid);
291 btrfs_set_inode_gid(leaf, item, inode->i_gid);
292 btrfs_set_inode_size(leaf, item, inode->i_size);
293 btrfs_set_inode_mode(leaf, item, inode->i_mode);
294 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
295
296 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
297 inode->i_atime.tv_sec);
298 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
299 inode->i_atime.tv_nsec);
300
301 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
302 inode->i_mtime.tv_sec);
303 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
304 inode->i_mtime.tv_nsec);
305
306 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
307 inode->i_ctime.tv_sec);
308 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
309 inode->i_ctime.tv_nsec);
310
311 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
312 btrfs_set_inode_generation(leaf, item, inode->i_generation);
313 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
314 btrfs_set_inode_block_group(leaf, item,
315 BTRFS_I(inode)->block_group->key.objectid);
316 }
317
318 int btrfs_update_inode(struct btrfs_trans_handle *trans,
319 struct btrfs_root *root,
320 struct inode *inode)
321 {
322 struct btrfs_inode_item *inode_item;
323 struct btrfs_path *path;
324 struct extent_buffer *leaf;
325 int ret;
326
327 path = btrfs_alloc_path();
328 BUG_ON(!path);
329 ret = btrfs_lookup_inode(trans, root, path,
330 &BTRFS_I(inode)->location, 1);
331 if (ret) {
332 if (ret > 0)
333 ret = -ENOENT;
334 goto failed;
335 }
336
337 leaf = path->nodes[0];
338 inode_item = btrfs_item_ptr(leaf, path->slots[0],
339 struct btrfs_inode_item);
340
341 fill_inode_item(leaf, inode_item, inode);
342 btrfs_mark_buffer_dirty(leaf);
343 btrfs_set_inode_last_trans(trans, inode);
344 ret = 0;
345 failed:
346 btrfs_release_path(root, path);
347 btrfs_free_path(path);
348 return ret;
349 }
350
351
352 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
353 struct btrfs_root *root,
354 struct inode *dir,
355 struct dentry *dentry)
356 {
357 struct btrfs_path *path;
358 const char *name = dentry->d_name.name;
359 int name_len = dentry->d_name.len;
360 int ret = 0;
361 struct extent_buffer *leaf;
362 struct btrfs_dir_item *di;
363 struct btrfs_key key;
364
365 path = btrfs_alloc_path();
366 if (!path) {
367 ret = -ENOMEM;
368 goto err;
369 }
370
371 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
372 name, name_len, -1);
373 if (IS_ERR(di)) {
374 ret = PTR_ERR(di);
375 goto err;
376 }
377 if (!di) {
378 ret = -ENOENT;
379 goto err;
380 }
381 leaf = path->nodes[0];
382 btrfs_dir_item_key_to_cpu(leaf, di, &key);
383 ret = btrfs_delete_one_dir_name(trans, root, path, di);
384 if (ret)
385 goto err;
386 btrfs_release_path(root, path);
387
388 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
389 key.objectid, name, name_len, -1);
390 if (IS_ERR(di)) {
391 ret = PTR_ERR(di);
392 goto err;
393 }
394 if (!di) {
395 ret = -ENOENT;
396 goto err;
397 }
398 ret = btrfs_delete_one_dir_name(trans, root, path, di);
399
400 dentry->d_inode->i_ctime = dir->i_ctime;
401 err:
402 btrfs_free_path(path);
403 if (!ret) {
404 dir->i_size -= name_len * 2;
405 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
406 btrfs_update_inode(trans, root, dir);
407 drop_nlink(dentry->d_inode);
408 ret = btrfs_update_inode(trans, root, dentry->d_inode);
409 dir->i_sb->s_dirt = 1;
410 }
411 return ret;
412 }
413
414 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
415 {
416 struct btrfs_root *root;
417 struct btrfs_trans_handle *trans;
418 int ret;
419 unsigned long nr;
420
421 root = BTRFS_I(dir)->root;
422 mutex_lock(&root->fs_info->fs_mutex);
423 trans = btrfs_start_transaction(root, 1);
424
425 btrfs_set_trans_block_group(trans, dir);
426 ret = btrfs_unlink_trans(trans, root, dir, dentry);
427 nr = trans->blocks_used;
428
429 btrfs_end_transaction(trans, root);
430 mutex_unlock(&root->fs_info->fs_mutex);
431 btrfs_btree_balance_dirty(root, nr);
432
433 return ret;
434 }
435
436 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
437 {
438 struct inode *inode = dentry->d_inode;
439 int err;
440 int ret;
441 struct btrfs_root *root = BTRFS_I(dir)->root;
442 struct btrfs_path *path;
443 struct btrfs_key key;
444 struct btrfs_trans_handle *trans;
445 struct btrfs_key found_key;
446 int found_type;
447 struct extent_buffer *leaf;
448 char *goodnames = "..";
449 unsigned long nr;
450
451 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
452 return -ENOTEMPTY;
453
454 path = btrfs_alloc_path();
455 BUG_ON(!path);
456 mutex_lock(&root->fs_info->fs_mutex);
457 trans = btrfs_start_transaction(root, 1);
458
459 btrfs_set_trans_block_group(trans, dir);
460 key.objectid = inode->i_ino;
461 key.offset = (u64)-1;
462 key.type = (u8)-1;
463 while(1) {
464 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
465 if (ret < 0) {
466 err = ret;
467 goto out;
468 }
469 BUG_ON(ret == 0);
470 if (path->slots[0] == 0) {
471 err = -ENOENT;
472 goto out;
473 }
474 path->slots[0]--;
475 leaf = path->nodes[0];
476 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
477 found_type = btrfs_key_type(&found_key);
478 if (found_key.objectid != inode->i_ino) {
479 err = -ENOENT;
480 goto out;
481 }
482 if ((found_type != BTRFS_DIR_ITEM_KEY &&
483 found_type != BTRFS_DIR_INDEX_KEY) ||
484 (!btrfs_match_dir_item_name(root, path, goodnames, 2) &&
485 !btrfs_match_dir_item_name(root, path, goodnames, 1))) {
486 err = -ENOTEMPTY;
487 goto out;
488 }
489 ret = btrfs_del_item(trans, root, path);
490 BUG_ON(ret);
491
492 if (found_type == BTRFS_DIR_ITEM_KEY && found_key.offset == 1)
493 break;
494 btrfs_release_path(root, path);
495 }
496 ret = 0;
497 btrfs_release_path(root, path);
498
499 /* now the directory is empty */
500 err = btrfs_unlink_trans(trans, root, dir, dentry);
501 if (!err) {
502 inode->i_size = 0;
503 }
504 out:
505 btrfs_release_path(root, path);
506 btrfs_free_path(path);
507 nr = trans->blocks_used;
508 ret = btrfs_end_transaction(trans, root);
509 mutex_unlock(&root->fs_info->fs_mutex);
510 btrfs_btree_balance_dirty(root, nr);
511 if (ret && !err)
512 err = ret;
513 return err;
514 }
515
516 static int btrfs_free_inode(struct btrfs_trans_handle *trans,
517 struct btrfs_root *root,
518 struct inode *inode)
519 {
520 struct btrfs_path *path;
521 int ret;
522
523 clear_inode(inode);
524
525 path = btrfs_alloc_path();
526 BUG_ON(!path);
527 ret = btrfs_lookup_inode(trans, root, path,
528 &BTRFS_I(inode)->location, -1);
529 if (ret > 0)
530 ret = -ENOENT;
531 if (!ret)
532 ret = btrfs_del_item(trans, root, path);
533 btrfs_free_path(path);
534 return ret;
535 }
536
537 /*
538 * this can truncate away extent items, csum items and directory items.
539 * It starts at a high offset and removes keys until it can't find
540 * any higher than i_size.
541 *
542 * csum items that cross the new i_size are truncated to the new size
543 * as well.
544 */
545 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
546 struct btrfs_root *root,
547 struct inode *inode)
548 {
549 int ret;
550 struct btrfs_path *path;
551 struct btrfs_key key;
552 struct btrfs_key found_key;
553 u32 found_type;
554 struct extent_buffer *leaf;
555 struct btrfs_file_extent_item *fi;
556 u64 extent_start = 0;
557 u64 extent_num_bytes = 0;
558 u64 item_end = 0;
559 int found_extent;
560 int del_item;
561
562 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
563 path = btrfs_alloc_path();
564 path->reada = -1;
565 BUG_ON(!path);
566
567 /* FIXME, add redo link to tree so we don't leak on crash */
568 key.objectid = inode->i_ino;
569 key.offset = (u64)-1;
570 key.type = (u8)-1;
571
572 while(1) {
573 btrfs_init_path(path);
574 fi = NULL;
575 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
576 if (ret < 0) {
577 goto error;
578 }
579 if (ret > 0) {
580 BUG_ON(path->slots[0] == 0);
581 path->slots[0]--;
582 }
583 leaf = path->nodes[0];
584 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
585 found_type = btrfs_key_type(&found_key);
586
587 if (found_key.objectid != inode->i_ino)
588 break;
589
590 if (found_type != BTRFS_CSUM_ITEM_KEY &&
591 found_type != BTRFS_DIR_ITEM_KEY &&
592 found_type != BTRFS_DIR_INDEX_KEY &&
593 found_type != BTRFS_EXTENT_DATA_KEY)
594 break;
595
596 item_end = found_key.offset;
597 if (found_type == BTRFS_EXTENT_DATA_KEY) {
598 fi = btrfs_item_ptr(leaf, path->slots[0],
599 struct btrfs_file_extent_item);
600 if (btrfs_file_extent_type(leaf, fi) !=
601 BTRFS_FILE_EXTENT_INLINE) {
602 item_end +=
603 btrfs_file_extent_num_bytes(leaf, fi);
604 }
605 }
606 if (found_type == BTRFS_CSUM_ITEM_KEY) {
607 ret = btrfs_csum_truncate(trans, root, path,
608 inode->i_size);
609 BUG_ON(ret);
610 }
611 if (item_end < inode->i_size) {
612 if (found_type == BTRFS_DIR_ITEM_KEY) {
613 found_type = BTRFS_INODE_ITEM_KEY;
614 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
615 found_type = BTRFS_CSUM_ITEM_KEY;
616 } else if (found_type) {
617 found_type--;
618 } else {
619 break;
620 }
621 btrfs_set_key_type(&key, found_type);
622 btrfs_release_path(root, path);
623 continue;
624 }
625 if (found_key.offset >= inode->i_size)
626 del_item = 1;
627 else
628 del_item = 0;
629 found_extent = 0;
630
631 /* FIXME, shrink the extent if the ref count is only 1 */
632 if (found_type == BTRFS_EXTENT_DATA_KEY &&
633 btrfs_file_extent_type(leaf, fi) !=
634 BTRFS_FILE_EXTENT_INLINE) {
635 u64 num_dec;
636 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
637 if (!del_item) {
638 u64 orig_num_bytes =
639 btrfs_file_extent_num_bytes(leaf, fi);
640 extent_num_bytes = inode->i_size -
641 found_key.offset + root->sectorsize - 1;
642 btrfs_set_file_extent_num_bytes(leaf, fi,
643 extent_num_bytes);
644 num_dec = (orig_num_bytes -
645 extent_num_bytes) >> 9;
646 if (extent_start != 0) {
647 inode->i_blocks -= num_dec;
648 }
649 btrfs_mark_buffer_dirty(leaf);
650 } else {
651 extent_num_bytes =
652 btrfs_file_extent_disk_num_bytes(leaf,
653 fi);
654 /* FIXME blocksize != 4096 */
655 num_dec = btrfs_file_extent_num_bytes(leaf,
656 fi) >> 9;
657 if (extent_start != 0) {
658 found_extent = 1;
659 inode->i_blocks -= num_dec;
660 }
661 }
662 }
663 if (del_item) {
664 ret = btrfs_del_item(trans, root, path);
665 if (ret)
666 goto error;
667 } else {
668 break;
669 }
670 btrfs_release_path(root, path);
671 if (found_extent) {
672 ret = btrfs_free_extent(trans, root, extent_start,
673 extent_num_bytes, 0);
674 BUG_ON(ret);
675 }
676 }
677 ret = 0;
678 error:
679 btrfs_release_path(root, path);
680 btrfs_free_path(path);
681 inode->i_sb->s_dirt = 1;
682 return ret;
683 }
684
685 static int btrfs_cow_one_page(struct inode *inode, struct page *page,
686 size_t zero_start)
687 {
688 char *kaddr;
689 int ret = 0;
690 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
691 u64 page_start = page->index << PAGE_CACHE_SHIFT;
692 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
693
694 set_page_extent_mapped(page);
695
696 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
697 set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start,
698 page_end, GFP_NOFS);
699 if (zero_start != PAGE_CACHE_SIZE) {
700 kaddr = kmap(page);
701 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
702 flush_dcache_page(page);
703 kunmap(page);
704 }
705 set_page_dirty(page);
706 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
707
708 return ret;
709 }
710
711 /*
712 * taken from block_truncate_page, but does cow as it zeros out
713 * any bytes left in the last page in the file.
714 */
715 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
716 {
717 struct inode *inode = mapping->host;
718 struct btrfs_root *root = BTRFS_I(inode)->root;
719 u32 blocksize = root->sectorsize;
720 pgoff_t index = from >> PAGE_CACHE_SHIFT;
721 unsigned offset = from & (PAGE_CACHE_SIZE-1);
722 struct page *page;
723 int ret = 0;
724 u64 page_start;
725
726 if ((offset & (blocksize - 1)) == 0)
727 goto out;
728
729 down_read(&root->snap_sem);
730 ret = -ENOMEM;
731 page = grab_cache_page(mapping, index);
732 if (!page)
733 goto out;
734 if (!PageUptodate(page)) {
735 ret = btrfs_readpage(NULL, page);
736 lock_page(page);
737 if (!PageUptodate(page)) {
738 ret = -EIO;
739 goto out;
740 }
741 }
742 page_start = page->index << PAGE_CACHE_SHIFT;
743
744 ret = btrfs_cow_one_page(inode, page, offset);
745
746 unlock_page(page);
747 page_cache_release(page);
748 up_read(&BTRFS_I(inode)->root->snap_sem);
749 out:
750 return ret;
751 }
752
753 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
754 {
755 struct inode *inode = dentry->d_inode;
756 int err;
757
758 err = inode_change_ok(inode, attr);
759 if (err)
760 return err;
761
762 if (S_ISREG(inode->i_mode) &&
763 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
764 struct btrfs_trans_handle *trans;
765 struct btrfs_root *root = BTRFS_I(inode)->root;
766 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
767
768 u64 mask = root->sectorsize - 1;
769 u64 pos = (inode->i_size + mask) & ~mask;
770 u64 block_end = attr->ia_size | mask;
771 u64 hole_size;
772 u64 alloc_hint;
773
774 if (attr->ia_size <= pos)
775 goto out;
776
777 btrfs_truncate_page(inode->i_mapping, inode->i_size);
778
779 lock_extent(em_tree, pos, block_end, GFP_NOFS);
780 hole_size = (attr->ia_size - pos + mask) & ~mask;
781
782 mutex_lock(&root->fs_info->fs_mutex);
783 trans = btrfs_start_transaction(root, 1);
784 btrfs_set_trans_block_group(trans, inode);
785 err = btrfs_drop_extents(trans, root, inode,
786 pos, pos + hole_size, pos,
787 &alloc_hint);
788
789 err = btrfs_insert_file_extent(trans, root, inode->i_ino,
790 pos, 0, 0, hole_size);
791 btrfs_end_transaction(trans, root);
792 mutex_unlock(&root->fs_info->fs_mutex);
793 unlock_extent(em_tree, pos, block_end, GFP_NOFS);
794 if (err)
795 return err;
796 }
797 out:
798 err = inode_setattr(inode, attr);
799
800 return err;
801 }
802 void btrfs_delete_inode(struct inode *inode)
803 {
804 struct btrfs_trans_handle *trans;
805 struct btrfs_root *root = BTRFS_I(inode)->root;
806 unsigned long nr;
807 int ret;
808
809 truncate_inode_pages(&inode->i_data, 0);
810 if (is_bad_inode(inode)) {
811 goto no_delete;
812 }
813
814 inode->i_size = 0;
815 mutex_lock(&root->fs_info->fs_mutex);
816 trans = btrfs_start_transaction(root, 1);
817
818 btrfs_set_trans_block_group(trans, inode);
819 ret = btrfs_truncate_in_trans(trans, root, inode);
820 if (ret)
821 goto no_delete_lock;
822 ret = btrfs_free_inode(trans, root, inode);
823 if (ret)
824 goto no_delete_lock;
825 nr = trans->blocks_used;
826
827 btrfs_end_transaction(trans, root);
828 mutex_unlock(&root->fs_info->fs_mutex);
829 btrfs_btree_balance_dirty(root, nr);
830 return;
831
832 no_delete_lock:
833 nr = trans->blocks_used;
834 btrfs_end_transaction(trans, root);
835 mutex_unlock(&root->fs_info->fs_mutex);
836 btrfs_btree_balance_dirty(root, nr);
837 no_delete:
838 clear_inode(inode);
839 }
840
841 /*
842 * this returns the key found in the dir entry in the location pointer.
843 * If no dir entries were found, location->objectid is 0.
844 */
845 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
846 struct btrfs_key *location)
847 {
848 const char *name = dentry->d_name.name;
849 int namelen = dentry->d_name.len;
850 struct btrfs_dir_item *di;
851 struct btrfs_path *path;
852 struct btrfs_root *root = BTRFS_I(dir)->root;
853 int ret = 0;
854
855 path = btrfs_alloc_path();
856 BUG_ON(!path);
857 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
858 namelen, 0);
859 if (IS_ERR(di))
860 ret = PTR_ERR(di);
861 if (!di || IS_ERR(di)) {
862 location->objectid = 0;
863 goto out;
864 }
865 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
866 out:
867 btrfs_release_path(root, path);
868 btrfs_free_path(path);
869 return ret;
870 }
871
872 /*
873 * when we hit a tree root in a directory, the btrfs part of the inode
874 * needs to be changed to reflect the root directory of the tree root. This
875 * is kind of like crossing a mount point.
876 */
877 static int fixup_tree_root_location(struct btrfs_root *root,
878 struct btrfs_key *location,
879 struct btrfs_root **sub_root,
880 struct dentry *dentry)
881 {
882 struct btrfs_path *path;
883 struct btrfs_root_item *ri;
884
885 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
886 return 0;
887 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
888 return 0;
889
890 path = btrfs_alloc_path();
891 BUG_ON(!path);
892 mutex_lock(&root->fs_info->fs_mutex);
893
894 *sub_root = btrfs_read_fs_root(root->fs_info, location,
895 dentry->d_name.name,
896 dentry->d_name.len);
897 if (IS_ERR(*sub_root))
898 return PTR_ERR(*sub_root);
899
900 ri = &(*sub_root)->root_item;
901 location->objectid = btrfs_root_dirid(ri);
902 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
903 location->offset = 0;
904
905 btrfs_free_path(path);
906 mutex_unlock(&root->fs_info->fs_mutex);
907 return 0;
908 }
909
910 static int btrfs_init_locked_inode(struct inode *inode, void *p)
911 {
912 struct btrfs_iget_args *args = p;
913 inode->i_ino = args->ino;
914 BTRFS_I(inode)->root = args->root;
915 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
916 inode->i_mapping, GFP_NOFS);
917 return 0;
918 }
919
920 static int btrfs_find_actor(struct inode *inode, void *opaque)
921 {
922 struct btrfs_iget_args *args = opaque;
923 return (args->ino == inode->i_ino &&
924 args->root == BTRFS_I(inode)->root);
925 }
926
927 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
928 struct btrfs_root *root)
929 {
930 struct inode *inode;
931 struct btrfs_iget_args args;
932 args.ino = objectid;
933 args.root = root;
934
935 inode = iget5_locked(s, objectid, btrfs_find_actor,
936 btrfs_init_locked_inode,
937 (void *)&args);
938 return inode;
939 }
940
941 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
942 struct nameidata *nd)
943 {
944 struct inode * inode;
945 struct btrfs_inode *bi = BTRFS_I(dir);
946 struct btrfs_root *root = bi->root;
947 struct btrfs_root *sub_root = root;
948 struct btrfs_key location;
949 int ret;
950
951 if (dentry->d_name.len > BTRFS_NAME_LEN)
952 return ERR_PTR(-ENAMETOOLONG);
953
954 mutex_lock(&root->fs_info->fs_mutex);
955 ret = btrfs_inode_by_name(dir, dentry, &location);
956 mutex_unlock(&root->fs_info->fs_mutex);
957
958 if (ret < 0)
959 return ERR_PTR(ret);
960
961 inode = NULL;
962 if (location.objectid) {
963 ret = fixup_tree_root_location(root, &location, &sub_root,
964 dentry);
965 if (ret < 0)
966 return ERR_PTR(ret);
967 if (ret > 0)
968 return ERR_PTR(-ENOENT);
969 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
970 sub_root);
971 if (!inode)
972 return ERR_PTR(-EACCES);
973 if (inode->i_state & I_NEW) {
974 /* the inode and parent dir are two different roots */
975 if (sub_root != root) {
976 igrab(inode);
977 sub_root->inode = inode;
978 }
979 BTRFS_I(inode)->root = sub_root;
980 memcpy(&BTRFS_I(inode)->location, &location,
981 sizeof(location));
982 btrfs_read_locked_inode(inode);
983 unlock_new_inode(inode);
984 }
985 }
986 return d_splice_alias(inode, dentry);
987 }
988
989 static unsigned char btrfs_filetype_table[] = {
990 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
991 };
992
993 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
994 {
995 struct inode *inode = filp->f_path.dentry->d_inode;
996 struct btrfs_root *root = BTRFS_I(inode)->root;
997 struct btrfs_item *item;
998 struct btrfs_dir_item *di;
999 struct btrfs_key key;
1000 struct btrfs_key found_key;
1001 struct btrfs_path *path;
1002 int ret;
1003 u32 nritems;
1004 struct extent_buffer *leaf;
1005 int slot;
1006 int advance;
1007 unsigned char d_type;
1008 int over = 0;
1009 u32 di_cur;
1010 u32 di_total;
1011 u32 di_len;
1012 int key_type = BTRFS_DIR_INDEX_KEY;
1013 char tmp_name[32];
1014 char *name_ptr;
1015 int name_len;
1016
1017 /* FIXME, use a real flag for deciding about the key type */
1018 if (root->fs_info->tree_root == root)
1019 key_type = BTRFS_DIR_ITEM_KEY;
1020
1021 mutex_lock(&root->fs_info->fs_mutex);
1022 key.objectid = inode->i_ino;
1023 btrfs_set_key_type(&key, key_type);
1024 key.offset = filp->f_pos;
1025
1026 path = btrfs_alloc_path();
1027 path->reada = 2;
1028 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1029 if (ret < 0)
1030 goto err;
1031 advance = 0;
1032 while(1) {
1033 leaf = path->nodes[0];
1034 nritems = btrfs_header_nritems(leaf);
1035 slot = path->slots[0];
1036 if (advance || slot >= nritems) {
1037 if (slot >= nritems -1) {
1038 ret = btrfs_next_leaf(root, path);
1039 if (ret)
1040 break;
1041 leaf = path->nodes[0];
1042 nritems = btrfs_header_nritems(leaf);
1043 slot = path->slots[0];
1044 } else {
1045 slot++;
1046 path->slots[0]++;
1047 }
1048 }
1049 advance = 1;
1050 item = btrfs_item_nr(leaf, slot);
1051 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1052
1053 if (found_key.objectid != key.objectid)
1054 break;
1055 if (btrfs_key_type(&found_key) != key_type)
1056 break;
1057 if (found_key.offset < filp->f_pos)
1058 continue;
1059
1060 filp->f_pos = found_key.offset;
1061 advance = 1;
1062 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1063 di_cur = 0;
1064 di_total = btrfs_item_size(leaf, item);
1065 while(di_cur < di_total) {
1066 struct btrfs_key location;
1067
1068 name_len = btrfs_dir_name_len(leaf, di);
1069 if (name_len < 32) {
1070 name_ptr = tmp_name;
1071 } else {
1072 name_ptr = kmalloc(name_len, GFP_NOFS);
1073 BUG_ON(!name_ptr);
1074 }
1075 read_extent_buffer(leaf, name_ptr,
1076 (unsigned long)(di + 1), name_len);
1077
1078 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1079 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1080
1081 over = filldir(dirent, name_ptr, name_len,
1082 found_key.offset,
1083 location.objectid,
1084 d_type);
1085
1086 if (name_ptr != tmp_name)
1087 kfree(name_ptr);
1088
1089 if (over)
1090 goto nopos;
1091 di_len = btrfs_dir_name_len(leaf, di) + sizeof(*di);
1092 di_cur += di_len;
1093 di = (struct btrfs_dir_item *)((char *)di + di_len);
1094 }
1095 }
1096 filp->f_pos++;
1097 nopos:
1098 ret = 0;
1099 err:
1100 btrfs_release_path(root, path);
1101 btrfs_free_path(path);
1102 mutex_unlock(&root->fs_info->fs_mutex);
1103 return ret;
1104 }
1105
1106 int btrfs_write_inode(struct inode *inode, int wait)
1107 {
1108 struct btrfs_root *root = BTRFS_I(inode)->root;
1109 struct btrfs_trans_handle *trans;
1110 int ret = 0;
1111
1112 if (wait) {
1113 mutex_lock(&root->fs_info->fs_mutex);
1114 trans = btrfs_start_transaction(root, 1);
1115 btrfs_set_trans_block_group(trans, inode);
1116 ret = btrfs_commit_transaction(trans, root);
1117 mutex_unlock(&root->fs_info->fs_mutex);
1118 }
1119 return ret;
1120 }
1121
1122 /*
1123 * This is somewhat expensive, updating the tree every time the
1124 * inode changes. But, it is most likely to find the inode in cache.
1125 * FIXME, needs more benchmarking...there are no reasons other than performance
1126 * to keep or drop this code.
1127 */
1128 void btrfs_dirty_inode(struct inode *inode)
1129 {
1130 struct btrfs_root *root = BTRFS_I(inode)->root;
1131 struct btrfs_trans_handle *trans;
1132
1133 mutex_lock(&root->fs_info->fs_mutex);
1134 trans = btrfs_start_transaction(root, 1);
1135 btrfs_set_trans_block_group(trans, inode);
1136 btrfs_update_inode(trans, root, inode);
1137 btrfs_end_transaction(trans, root);
1138 mutex_unlock(&root->fs_info->fs_mutex);
1139 }
1140
1141 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1142 struct btrfs_root *root,
1143 u64 objectid,
1144 struct btrfs_block_group_cache *group,
1145 int mode)
1146 {
1147 struct inode *inode;
1148 struct btrfs_inode_item *inode_item;
1149 struct btrfs_key *location;
1150 struct btrfs_path *path;
1151 int ret;
1152 int owner;
1153
1154 path = btrfs_alloc_path();
1155 BUG_ON(!path);
1156
1157 inode = new_inode(root->fs_info->sb);
1158 if (!inode)
1159 return ERR_PTR(-ENOMEM);
1160
1161 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1162 inode->i_mapping, GFP_NOFS);
1163 BTRFS_I(inode)->root = root;
1164
1165 if (mode & S_IFDIR)
1166 owner = 0;
1167 else
1168 owner = 1;
1169 group = btrfs_find_block_group(root, group, 0, 0, owner);
1170 BTRFS_I(inode)->block_group = group;
1171
1172 ret = btrfs_insert_empty_inode(trans, root, path, objectid);
1173 if (ret)
1174 goto fail;
1175
1176 inode->i_uid = current->fsuid;
1177 inode->i_gid = current->fsgid;
1178 inode->i_mode = mode;
1179 inode->i_ino = objectid;
1180 inode->i_blocks = 0;
1181 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1182 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1183 struct btrfs_inode_item);
1184 fill_inode_item(path->nodes[0], inode_item, inode);
1185 btrfs_mark_buffer_dirty(path->nodes[0]);
1186 btrfs_free_path(path);
1187
1188 location = &BTRFS_I(inode)->location;
1189 location->objectid = objectid;
1190 location->offset = 0;
1191 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1192
1193 insert_inode_hash(inode);
1194 return inode;
1195 fail:
1196 btrfs_free_path(path);
1197 return ERR_PTR(ret);
1198 }
1199
1200 static inline u8 btrfs_inode_type(struct inode *inode)
1201 {
1202 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1203 }
1204
1205 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1206 struct dentry *dentry, struct inode *inode)
1207 {
1208 int ret;
1209 struct btrfs_key key;
1210 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1211 struct inode *parent_inode;
1212
1213 key.objectid = inode->i_ino;
1214 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1215 key.offset = 0;
1216
1217 ret = btrfs_insert_dir_item(trans, root,
1218 dentry->d_name.name, dentry->d_name.len,
1219 dentry->d_parent->d_inode->i_ino,
1220 &key, btrfs_inode_type(inode));
1221 if (ret == 0) {
1222 parent_inode = dentry->d_parent->d_inode;
1223 parent_inode->i_size += dentry->d_name.len * 2;
1224 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1225 ret = btrfs_update_inode(trans, root,
1226 dentry->d_parent->d_inode);
1227 }
1228 return ret;
1229 }
1230
1231 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1232 struct dentry *dentry, struct inode *inode)
1233 {
1234 int err = btrfs_add_link(trans, dentry, inode);
1235 if (!err) {
1236 d_instantiate(dentry, inode);
1237 return 0;
1238 }
1239 if (err > 0)
1240 err = -EEXIST;
1241 return err;
1242 }
1243
1244 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1245 int mode, dev_t rdev)
1246 {
1247 struct btrfs_trans_handle *trans;
1248 struct btrfs_root *root = BTRFS_I(dir)->root;
1249 struct inode *inode;
1250 int err;
1251 int drop_inode = 0;
1252 u64 objectid;
1253 unsigned long nr;
1254
1255 if (!new_valid_dev(rdev))
1256 return -EINVAL;
1257
1258 mutex_lock(&root->fs_info->fs_mutex);
1259 trans = btrfs_start_transaction(root, 1);
1260 btrfs_set_trans_block_group(trans, dir);
1261
1262 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1263 if (err) {
1264 err = -ENOSPC;
1265 goto out_unlock;
1266 }
1267
1268 inode = btrfs_new_inode(trans, root, objectid,
1269 BTRFS_I(dir)->block_group, mode);
1270 err = PTR_ERR(inode);
1271 if (IS_ERR(inode))
1272 goto out_unlock;
1273
1274 btrfs_set_trans_block_group(trans, inode);
1275 err = btrfs_add_nondir(trans, dentry, inode);
1276 if (err)
1277 drop_inode = 1;
1278 else {
1279 inode->i_op = &btrfs_special_inode_operations;
1280 init_special_inode(inode, inode->i_mode, rdev);
1281 btrfs_update_inode(trans, root, inode);
1282 }
1283 dir->i_sb->s_dirt = 1;
1284 btrfs_update_inode_block_group(trans, inode);
1285 btrfs_update_inode_block_group(trans, dir);
1286 out_unlock:
1287 nr = trans->blocks_used;
1288 btrfs_end_transaction(trans, root);
1289 mutex_unlock(&root->fs_info->fs_mutex);
1290
1291 if (drop_inode) {
1292 inode_dec_link_count(inode);
1293 iput(inode);
1294 }
1295 btrfs_btree_balance_dirty(root, nr);
1296 return err;
1297 }
1298
1299 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1300 int mode, struct nameidata *nd)
1301 {
1302 struct btrfs_trans_handle *trans;
1303 struct btrfs_root *root = BTRFS_I(dir)->root;
1304 struct inode *inode;
1305 int err;
1306 int drop_inode = 0;
1307 unsigned long nr;
1308 u64 objectid;
1309
1310 mutex_lock(&root->fs_info->fs_mutex);
1311 trans = btrfs_start_transaction(root, 1);
1312 btrfs_set_trans_block_group(trans, dir);
1313
1314 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1315 if (err) {
1316 err = -ENOSPC;
1317 goto out_unlock;
1318 }
1319
1320 inode = btrfs_new_inode(trans, root, objectid,
1321 BTRFS_I(dir)->block_group, mode);
1322 err = PTR_ERR(inode);
1323 if (IS_ERR(inode))
1324 goto out_unlock;
1325
1326 btrfs_set_trans_block_group(trans, inode);
1327 err = btrfs_add_nondir(trans, dentry, inode);
1328 if (err)
1329 drop_inode = 1;
1330 else {
1331 inode->i_mapping->a_ops = &btrfs_aops;
1332 inode->i_fop = &btrfs_file_operations;
1333 inode->i_op = &btrfs_file_inode_operations;
1334 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1335 inode->i_mapping, GFP_NOFS);
1336 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
1337 }
1338 dir->i_sb->s_dirt = 1;
1339 btrfs_update_inode_block_group(trans, inode);
1340 btrfs_update_inode_block_group(trans, dir);
1341 out_unlock:
1342 nr = trans->blocks_used;
1343 btrfs_end_transaction(trans, root);
1344 mutex_unlock(&root->fs_info->fs_mutex);
1345
1346 if (drop_inode) {
1347 inode_dec_link_count(inode);
1348 iput(inode);
1349 }
1350 btrfs_btree_balance_dirty(root, nr);
1351 return err;
1352 }
1353
1354 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1355 struct dentry *dentry)
1356 {
1357 struct btrfs_trans_handle *trans;
1358 struct btrfs_root *root = BTRFS_I(dir)->root;
1359 struct inode *inode = old_dentry->d_inode;
1360 unsigned long nr;
1361 int err;
1362 int drop_inode = 0;
1363
1364 if (inode->i_nlink == 0)
1365 return -ENOENT;
1366
1367 inc_nlink(inode);
1368 mutex_lock(&root->fs_info->fs_mutex);
1369 trans = btrfs_start_transaction(root, 1);
1370
1371 btrfs_set_trans_block_group(trans, dir);
1372 atomic_inc(&inode->i_count);
1373 err = btrfs_add_nondir(trans, dentry, inode);
1374
1375 if (err)
1376 drop_inode = 1;
1377
1378 dir->i_sb->s_dirt = 1;
1379 btrfs_update_inode_block_group(trans, dir);
1380 err = btrfs_update_inode(trans, root, inode);
1381
1382 if (err)
1383 drop_inode = 1;
1384
1385 nr = trans->blocks_used;
1386 btrfs_end_transaction(trans, root);
1387 mutex_unlock(&root->fs_info->fs_mutex);
1388
1389 if (drop_inode) {
1390 inode_dec_link_count(inode);
1391 iput(inode);
1392 }
1393 btrfs_btree_balance_dirty(root, nr);
1394 return err;
1395 }
1396
1397 static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
1398 struct btrfs_root *root,
1399 u64 objectid, u64 dirid)
1400 {
1401 int ret;
1402 char buf[2];
1403 struct btrfs_key key;
1404
1405 buf[0] = '.';
1406 buf[1] = '.';
1407
1408 key.objectid = objectid;
1409 key.offset = 0;
1410 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1411
1412 ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid,
1413 &key, BTRFS_FT_DIR);
1414 if (ret)
1415 goto error;
1416
1417 key.objectid = dirid;
1418 ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid,
1419 &key, BTRFS_FT_DIR);
1420 if (ret)
1421 goto error;
1422 error:
1423 return ret;
1424 }
1425
1426 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1427 {
1428 struct inode *inode;
1429 struct btrfs_trans_handle *trans;
1430 struct btrfs_root *root = BTRFS_I(dir)->root;
1431 int err = 0;
1432 int drop_on_err = 0;
1433 u64 objectid;
1434 unsigned long nr = 1;
1435
1436 mutex_lock(&root->fs_info->fs_mutex);
1437 trans = btrfs_start_transaction(root, 1);
1438 btrfs_set_trans_block_group(trans, dir);
1439
1440 if (IS_ERR(trans)) {
1441 err = PTR_ERR(trans);
1442 goto out_unlock;
1443 }
1444
1445 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1446 if (err) {
1447 err = -ENOSPC;
1448 goto out_unlock;
1449 }
1450
1451 inode = btrfs_new_inode(trans, root, objectid,
1452 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1453 if (IS_ERR(inode)) {
1454 err = PTR_ERR(inode);
1455 goto out_fail;
1456 }
1457
1458 drop_on_err = 1;
1459 inode->i_op = &btrfs_dir_inode_operations;
1460 inode->i_fop = &btrfs_dir_file_operations;
1461 btrfs_set_trans_block_group(trans, inode);
1462
1463 err = btrfs_make_empty_dir(trans, root, inode->i_ino, dir->i_ino);
1464 if (err)
1465 goto out_fail;
1466
1467 inode->i_size = 6;
1468 err = btrfs_update_inode(trans, root, inode);
1469 if (err)
1470 goto out_fail;
1471
1472 err = btrfs_add_link(trans, dentry, inode);
1473 if (err)
1474 goto out_fail;
1475
1476 d_instantiate(dentry, inode);
1477 drop_on_err = 0;
1478 dir->i_sb->s_dirt = 1;
1479 btrfs_update_inode_block_group(trans, inode);
1480 btrfs_update_inode_block_group(trans, dir);
1481
1482 out_fail:
1483 nr = trans->blocks_used;
1484 btrfs_end_transaction(trans, root);
1485
1486 out_unlock:
1487 mutex_unlock(&root->fs_info->fs_mutex);
1488 if (drop_on_err)
1489 iput(inode);
1490 btrfs_btree_balance_dirty(root, nr);
1491 return err;
1492 }
1493
1494 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
1495 size_t page_offset, u64 start, u64 end,
1496 int create)
1497 {
1498 int ret;
1499 int err = 0;
1500 u64 bytenr;
1501 u64 extent_start = 0;
1502 u64 extent_end = 0;
1503 u64 objectid = inode->i_ino;
1504 u32 found_type;
1505 int failed_insert = 0;
1506 struct btrfs_path *path;
1507 struct btrfs_root *root = BTRFS_I(inode)->root;
1508 struct btrfs_file_extent_item *item;
1509 struct extent_buffer *leaf;
1510 struct btrfs_key found_key;
1511 struct extent_map *em = NULL;
1512 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1513 struct btrfs_trans_handle *trans = NULL;
1514
1515 path = btrfs_alloc_path();
1516 BUG_ON(!path);
1517 mutex_lock(&root->fs_info->fs_mutex);
1518
1519 again:
1520 em = lookup_extent_mapping(em_tree, start, end);
1521 if (em) {
1522 goto out;
1523 }
1524 if (!em) {
1525 em = alloc_extent_map(GFP_NOFS);
1526 if (!em) {
1527 err = -ENOMEM;
1528 goto out;
1529 }
1530 em->start = EXTENT_MAP_HOLE;
1531 em->end = EXTENT_MAP_HOLE;
1532 }
1533 em->bdev = inode->i_sb->s_bdev;
1534 ret = btrfs_lookup_file_extent(NULL, root, path,
1535 objectid, start, 0);
1536 if (ret < 0) {
1537 err = ret;
1538 goto out;
1539 }
1540
1541 if (ret != 0) {
1542 if (path->slots[0] == 0)
1543 goto not_found;
1544 path->slots[0]--;
1545 }
1546
1547 leaf = path->nodes[0];
1548 item = btrfs_item_ptr(leaf, path->slots[0],
1549 struct btrfs_file_extent_item);
1550 /* are we inside the extent that was found? */
1551 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1552 found_type = btrfs_key_type(&found_key);
1553 if (found_key.objectid != objectid ||
1554 found_type != BTRFS_EXTENT_DATA_KEY) {
1555 goto not_found;
1556 }
1557
1558 found_type = btrfs_file_extent_type(leaf, item);
1559 extent_start = found_key.offset;
1560 if (found_type == BTRFS_FILE_EXTENT_REG) {
1561 extent_end = extent_start +
1562 btrfs_file_extent_num_bytes(leaf, item);
1563 err = 0;
1564 if (start < extent_start || start >= extent_end) {
1565 em->start = start;
1566 if (start < extent_start) {
1567 if (end < extent_start)
1568 goto not_found;
1569 em->end = extent_end - 1;
1570 } else {
1571 em->end = end;
1572 }
1573 goto not_found_em;
1574 }
1575 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
1576 if (bytenr == 0) {
1577 em->start = extent_start;
1578 em->end = extent_end - 1;
1579 em->block_start = EXTENT_MAP_HOLE;
1580 em->block_end = EXTENT_MAP_HOLE;
1581 goto insert;
1582 }
1583 bytenr += btrfs_file_extent_offset(leaf, item);
1584 em->block_start = bytenr;
1585 em->block_end = em->block_start +
1586 btrfs_file_extent_num_bytes(leaf, item) - 1;
1587 em->start = extent_start;
1588 em->end = extent_end - 1;
1589 goto insert;
1590 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1591 unsigned long ptr;
1592 char *map;
1593 size_t size;
1594 size_t extent_offset;
1595 size_t copy_size;
1596
1597 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
1598 path->slots[0]));
1599
1600 extent_end = (extent_start + size) |
1601 ((u64)root->sectorsize - 1);
1602 if (start < extent_start || start >= extent_end) {
1603 em->start = start;
1604 if (start < extent_start) {
1605 if (end < extent_start)
1606 goto not_found;
1607 em->end = extent_end;
1608 } else {
1609 em->end = end;
1610 }
1611 goto not_found_em;
1612 }
1613
1614 extent_offset = (page->index << PAGE_CACHE_SHIFT) -
1615 extent_start;
1616 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
1617 map = kmap(page);
1618 copy_size = min_t(u64, PAGE_CACHE_SIZE - page_offset,
1619 size - extent_offset);
1620
1621 em->block_start = EXTENT_MAP_INLINE;
1622 em->block_end = EXTENT_MAP_INLINE;
1623 em->start = extent_start + extent_offset;
1624 em->end = (em->start + copy_size -1) |
1625 ((u64)root->sectorsize -1);
1626
1627 if (!page) {
1628 goto insert;
1629 }
1630
1631 read_extent_buffer(leaf, map + page_offset, ptr, copy_size);
1632 /*
1633 memset(map + page_offset + copy_size, 0,
1634 PAGE_CACHE_SIZE - copy_size - page_offset);
1635 */
1636 flush_dcache_page(page);
1637 kunmap(page);
1638 set_extent_uptodate(em_tree, em->start, em->end, GFP_NOFS);
1639 goto insert;
1640 } else {
1641 printk("unkknown found_type %d\n", found_type);
1642 WARN_ON(1);
1643 }
1644 not_found:
1645 em->start = start;
1646 em->end = end;
1647 not_found_em:
1648 em->block_start = EXTENT_MAP_HOLE;
1649 em->block_end = EXTENT_MAP_HOLE;
1650 insert:
1651 btrfs_release_path(root, path);
1652 if (em->start > start || em->end < start) {
1653 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->end, start, end);
1654 err = -EIO;
1655 goto out;
1656 }
1657 ret = add_extent_mapping(em_tree, em);
1658 if (ret == -EEXIST) {
1659 free_extent_map(em);
1660 em = NULL;
1661 failed_insert++;
1662 if (failed_insert > 5) {
1663 printk("failing to insert %Lu %Lu\n", start, end);
1664 err = -EIO;
1665 goto out;
1666 }
1667 goto again;
1668 }
1669 err = 0;
1670 out:
1671 btrfs_free_path(path);
1672 if (trans) {
1673 ret = btrfs_end_transaction(trans, root);
1674 if (!err)
1675 err = ret;
1676 }
1677 mutex_unlock(&root->fs_info->fs_mutex);
1678 if (err) {
1679 free_extent_map(em);
1680 WARN_ON(1);
1681 return ERR_PTR(err);
1682 }
1683 return em;
1684 }
1685
1686 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
1687 {
1688 return extent_bmap(mapping, iblock, btrfs_get_extent);
1689 }
1690
1691 static int btrfs_prepare_write(struct file *file, struct page *page,
1692 unsigned from, unsigned to)
1693 {
1694 return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree,
1695 page->mapping->host, page, from, to,
1696 btrfs_get_extent);
1697 }
1698
1699 int btrfs_readpage(struct file *file, struct page *page)
1700 {
1701 struct extent_map_tree *tree;
1702 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1703 return extent_read_full_page(tree, page, btrfs_get_extent);
1704 }
1705 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
1706 {
1707 struct extent_map_tree *tree;
1708
1709
1710 if (current->flags & PF_MEMALLOC) {
1711 redirty_page_for_writepage(wbc, page);
1712 unlock_page(page);
1713 return 0;
1714 }
1715 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1716 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
1717 }
1718
1719 static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
1720 {
1721 struct extent_map_tree *tree;
1722 int ret;
1723
1724 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1725 ret = try_release_extent_mapping(tree, page);
1726 if (ret == 1) {
1727 ClearPagePrivate(page);
1728 set_page_private(page, 0);
1729 page_cache_release(page);
1730 }
1731 return ret;
1732 }
1733
1734 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
1735 {
1736 struct extent_map_tree *tree;
1737
1738 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1739 extent_invalidatepage(tree, page, offset);
1740 btrfs_releasepage(page, GFP_NOFS);
1741 }
1742
1743 /*
1744 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
1745 * called from a page fault handler when a page is first dirtied. Hence we must
1746 * be careful to check for EOF conditions here. We set the page up correctly
1747 * for a written page which means we get ENOSPC checking when writing into
1748 * holes and correct delalloc and unwritten extent mapping on filesystems that
1749 * support these features.
1750 *
1751 * We are not allowed to take the i_mutex here so we have to play games to
1752 * protect against truncate races as the page could now be beyond EOF. Because
1753 * vmtruncate() writes the inode size before removing pages, once we have the
1754 * page lock we can determine safely if the page is beyond EOF. If it is not
1755 * beyond EOF, then the page is guaranteed safe against truncation until we
1756 * unlock the page.
1757 */
1758 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1759 {
1760 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
1761 unsigned long end;
1762 loff_t size;
1763 int ret = -EINVAL;
1764 u64 page_start;
1765
1766 down_read(&BTRFS_I(inode)->root->snap_sem);
1767 lock_page(page);
1768 wait_on_page_writeback(page);
1769 size = i_size_read(inode);
1770 page_start = page->index << PAGE_CACHE_SHIFT;
1771
1772 if ((page->mapping != inode->i_mapping) ||
1773 (page_start > size)) {
1774 /* page got truncated out from underneath us */
1775 goto out_unlock;
1776 }
1777
1778 /* page is wholly or partially inside EOF */
1779 if (page_start + PAGE_CACHE_SIZE > size)
1780 end = size & ~PAGE_CACHE_MASK;
1781 else
1782 end = PAGE_CACHE_SIZE;
1783
1784 ret = btrfs_cow_one_page(inode, page, end);
1785
1786 out_unlock:
1787 up_read(&BTRFS_I(inode)->root->snap_sem);
1788 unlock_page(page);
1789 return ret;
1790 }
1791
1792 static void btrfs_truncate(struct inode *inode)
1793 {
1794 struct btrfs_root *root = BTRFS_I(inode)->root;
1795 int ret;
1796 struct btrfs_trans_handle *trans;
1797 unsigned long nr;
1798
1799 if (!S_ISREG(inode->i_mode))
1800 return;
1801 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1802 return;
1803
1804 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1805
1806 mutex_lock(&root->fs_info->fs_mutex);
1807 trans = btrfs_start_transaction(root, 1);
1808 btrfs_set_trans_block_group(trans, inode);
1809
1810 /* FIXME, add redo link to tree so we don't leak on crash */
1811 ret = btrfs_truncate_in_trans(trans, root, inode);
1812 btrfs_update_inode(trans, root, inode);
1813 nr = trans->blocks_used;
1814
1815 ret = btrfs_end_transaction(trans, root);
1816 BUG_ON(ret);
1817 mutex_unlock(&root->fs_info->fs_mutex);
1818 btrfs_btree_balance_dirty(root, nr);
1819 }
1820
1821 int btrfs_commit_write(struct file *file, struct page *page,
1822 unsigned from, unsigned to)
1823 {
1824 return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree,
1825 page->mapping->host, page, from, to);
1826 }
1827
1828 static int create_subvol(struct btrfs_root *root, char *name, int namelen)
1829 {
1830 struct btrfs_trans_handle *trans;
1831 struct btrfs_key key;
1832 struct btrfs_root_item root_item;
1833 struct btrfs_inode_item *inode_item;
1834 struct extent_buffer *leaf;
1835 struct btrfs_root *new_root;
1836 struct inode *inode;
1837 struct inode *dir;
1838 int ret;
1839 int err;
1840 u64 objectid;
1841 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
1842 unsigned long nr = 1;
1843
1844 mutex_lock(&root->fs_info->fs_mutex);
1845 trans = btrfs_start_transaction(root, 1);
1846 BUG_ON(!trans);
1847
1848 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, 0);
1849 if (IS_ERR(leaf))
1850 return PTR_ERR(leaf);
1851
1852 btrfs_set_header_nritems(leaf, 0);
1853 btrfs_set_header_level(leaf, 0);
1854 btrfs_set_header_bytenr(leaf, leaf->start);
1855 btrfs_set_header_generation(leaf, trans->transid);
1856 btrfs_set_header_owner(leaf, root->root_key.objectid);
1857 write_extent_buffer(leaf, root->fs_info->fsid,
1858 (unsigned long)btrfs_header_fsid(leaf),
1859 BTRFS_FSID_SIZE);
1860 btrfs_mark_buffer_dirty(leaf);
1861
1862 inode_item = &root_item.inode;
1863 memset(inode_item, 0, sizeof(*inode_item));
1864 inode_item->generation = cpu_to_le64(1);
1865 inode_item->size = cpu_to_le64(3);
1866 inode_item->nlink = cpu_to_le32(1);
1867 inode_item->nblocks = cpu_to_le64(1);
1868 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1869
1870 btrfs_set_root_bytenr(&root_item, leaf->start);
1871 btrfs_set_root_level(&root_item, 0);
1872 btrfs_set_root_refs(&root_item, 1);
1873 btrfs_set_root_used(&root_item, 0);
1874
1875 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
1876 root_item.drop_level = 0;
1877
1878 free_extent_buffer(leaf);
1879 leaf = NULL;
1880
1881 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1882 0, &objectid);
1883 if (ret)
1884 goto fail;
1885
1886 btrfs_set_root_dirid(&root_item, new_dirid);
1887
1888 key.objectid = objectid;
1889 key.offset = 1;
1890 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1891 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1892 &root_item);
1893 if (ret)
1894 goto fail;
1895
1896 /*
1897 * insert the directory item
1898 */
1899 key.offset = (u64)-1;
1900 dir = root->fs_info->sb->s_root->d_inode;
1901 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1902 name, namelen, dir->i_ino, &key,
1903 BTRFS_FT_DIR);
1904 if (ret)
1905 goto fail;
1906
1907 ret = btrfs_commit_transaction(trans, root);
1908 if (ret)
1909 goto fail_commit;
1910
1911 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
1912 BUG_ON(!new_root);
1913
1914 trans = btrfs_start_transaction(new_root, 1);
1915 BUG_ON(!trans);
1916
1917 inode = btrfs_new_inode(trans, new_root, new_dirid,
1918 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
1919 if (IS_ERR(inode))
1920 goto fail;
1921 inode->i_op = &btrfs_dir_inode_operations;
1922 inode->i_fop = &btrfs_dir_file_operations;
1923 new_root->inode = inode;
1924
1925 ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid);
1926 if (ret)
1927 goto fail;
1928
1929 inode->i_nlink = 1;
1930 inode->i_size = 6;
1931 ret = btrfs_update_inode(trans, new_root, inode);
1932 if (ret)
1933 goto fail;
1934 fail:
1935 nr = trans->blocks_used;
1936 err = btrfs_commit_transaction(trans, root);
1937 if (err && !ret)
1938 ret = err;
1939 fail_commit:
1940 mutex_unlock(&root->fs_info->fs_mutex);
1941 btrfs_btree_balance_dirty(root, nr);
1942 return ret;
1943 }
1944
1945 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
1946 {
1947 struct btrfs_trans_handle *trans;
1948 struct btrfs_key key;
1949 struct btrfs_root_item new_root_item;
1950 struct extent_buffer *tmp;
1951 int ret;
1952 int err;
1953 u64 objectid;
1954 unsigned long nr;
1955
1956 if (!root->ref_cows)
1957 return -EINVAL;
1958
1959 down_write(&root->snap_sem);
1960 freeze_bdev(root->fs_info->sb->s_bdev);
1961 thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb);
1962
1963 mutex_lock(&root->fs_info->fs_mutex);
1964 trans = btrfs_start_transaction(root, 1);
1965 BUG_ON(!trans);
1966
1967 ret = btrfs_update_inode(trans, root, root->inode);
1968 if (ret)
1969 goto fail;
1970
1971 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1972 0, &objectid);
1973 if (ret)
1974 goto fail;
1975
1976 memcpy(&new_root_item, &root->root_item,
1977 sizeof(new_root_item));
1978
1979 key.objectid = objectid;
1980 key.offset = 1;
1981 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1982
1983 btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
1984 btrfs_set_root_bytenr(&new_root_item, root->node->start);
1985 btrfs_set_root_level(&new_root_item, btrfs_header_level(root->node));
1986
1987 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1988 &new_root_item);
1989 if (ret)
1990 goto fail;
1991
1992 /*
1993 * insert the directory item
1994 */
1995 key.offset = (u64)-1;
1996 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1997 name, namelen,
1998 root->fs_info->sb->s_root->d_inode->i_ino,
1999 &key, BTRFS_FT_DIR);
2000
2001 if (ret)
2002 goto fail;
2003
2004 ret = btrfs_inc_root_ref(trans, root);
2005 if (ret)
2006 goto fail;
2007 fail:
2008 nr = trans->blocks_used;
2009 err = btrfs_commit_transaction(trans, root);
2010
2011 if (err && !ret)
2012 ret = err;
2013
2014 mutex_unlock(&root->fs_info->fs_mutex);
2015 up_write(&root->snap_sem);
2016 btrfs_btree_balance_dirty(root, nr);
2017 return ret;
2018 }
2019
2020 static unsigned long force_ra(struct address_space *mapping,
2021 struct file_ra_state *ra, struct file *file,
2022 pgoff_t offset, pgoff_t last_index)
2023 {
2024 pgoff_t req_size;
2025
2026 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2027 req_size = last_index - offset + 1;
2028 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2029 return offset;
2030 #else
2031 req_size = min(last_index - offset + 1, (pgoff_t)128);
2032 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2033 return offset + req_size;
2034 #endif
2035 }
2036
2037 int btrfs_defrag_file(struct file *file) {
2038 struct inode *inode = file->f_path.dentry->d_inode;
2039 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2040 struct page *page;
2041 unsigned long last_index;
2042 unsigned long ra_index = 0;
2043 u64 page_start;
2044 u64 page_end;
2045 unsigned long i;
2046
2047 mutex_lock(&inode->i_mutex);
2048 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
2049 for (i = 0; i <= last_index; i++) {
2050 if (i == ra_index) {
2051 ra_index = force_ra(inode->i_mapping, &file->f_ra,
2052 file, ra_index, last_index);
2053 }
2054 page = grab_cache_page(inode->i_mapping, i);
2055 if (!page)
2056 goto out_unlock;
2057 if (!PageUptodate(page)) {
2058 btrfs_readpage(NULL, page);
2059 lock_page(page);
2060 if (!PageUptodate(page)) {
2061 unlock_page(page);
2062 page_cache_release(page);
2063 goto out_unlock;
2064 }
2065 }
2066 page_start = page->index << PAGE_CACHE_SHIFT;
2067 page_end = page_start + PAGE_CACHE_SIZE - 1;
2068
2069 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
2070 set_extent_delalloc(em_tree, page_start,
2071 page_end, GFP_NOFS);
2072 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
2073 set_page_dirty(page);
2074 unlock_page(page);
2075 page_cache_release(page);
2076 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
2077 }
2078
2079 out_unlock:
2080 mutex_unlock(&inode->i_mutex);
2081 return 0;
2082 }
2083
2084 static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg)
2085 {
2086 struct btrfs_ioctl_vol_args vol_args;
2087 struct btrfs_dir_item *di;
2088 struct btrfs_path *path;
2089 int namelen;
2090 u64 root_dirid;
2091
2092 if (copy_from_user(&vol_args, arg, sizeof(vol_args)))
2093 return -EFAULT;
2094
2095 namelen = strlen(vol_args.name);
2096 if (namelen > BTRFS_VOL_NAME_MAX)
2097 return -EINVAL;
2098 if (strchr(vol_args.name, '/'))
2099 return -EINVAL;
2100
2101 path = btrfs_alloc_path();
2102 if (!path)
2103 return -ENOMEM;
2104
2105 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2106 mutex_lock(&root->fs_info->fs_mutex);
2107 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2108 path, root_dirid,
2109 vol_args.name, namelen, 0);
2110 mutex_unlock(&root->fs_info->fs_mutex);
2111 btrfs_free_path(path);
2112 if (di && !IS_ERR(di))
2113 return -EEXIST;
2114 if (IS_ERR(di))
2115 return PTR_ERR(di);
2116
2117 if (root == root->fs_info->tree_root)
2118 return create_subvol(root, vol_args.name, namelen);
2119 return create_snapshot(root, vol_args.name, namelen);
2120 }
2121
2122 static int btrfs_ioctl_defrag(struct file *file)
2123 {
2124 struct inode *inode = file->f_path.dentry->d_inode;
2125 struct btrfs_root *root = BTRFS_I(inode)->root;
2126
2127 switch (inode->i_mode & S_IFMT) {
2128 case S_IFDIR:
2129 mutex_lock(&root->fs_info->fs_mutex);
2130 btrfs_defrag_root(root, 0);
2131 btrfs_defrag_root(root->fs_info->extent_root, 0);
2132 mutex_unlock(&root->fs_info->fs_mutex);
2133 break;
2134 case S_IFREG:
2135 btrfs_defrag_file(file);
2136 break;
2137 }
2138
2139 return 0;
2140 }
2141
2142 long btrfs_ioctl(struct file *file, unsigned int
2143 cmd, unsigned long arg)
2144 {
2145 struct btrfs_root *root = BTRFS_I(file->f_path.dentry->d_inode)->root;
2146
2147 switch (cmd) {
2148 case BTRFS_IOC_SNAP_CREATE:
2149 return btrfs_ioctl_snap_create(root, (void __user *)arg);
2150 case BTRFS_IOC_DEFRAG:
2151 return btrfs_ioctl_defrag(file);
2152 }
2153
2154 return -ENOTTY;
2155 }
2156
2157 /*
2158 * Called inside transaction, so use GFP_NOFS
2159 */
2160 struct inode *btrfs_alloc_inode(struct super_block *sb)
2161 {
2162 struct btrfs_inode *ei;
2163
2164 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2165 if (!ei)
2166 return NULL;
2167 ei->last_trans = 0;
2168 return &ei->vfs_inode;
2169 }
2170
2171 void btrfs_destroy_inode(struct inode *inode)
2172 {
2173 WARN_ON(!list_empty(&inode->i_dentry));
2174 WARN_ON(inode->i_data.nrpages);
2175
2176 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2177 }
2178
2179 static void init_once(void * foo, struct kmem_cache * cachep,
2180 unsigned long flags)
2181 {
2182 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2183
2184 inode_init_once(&ei->vfs_inode);
2185 }
2186
2187 void btrfs_destroy_cachep(void)
2188 {
2189 if (btrfs_inode_cachep)
2190 kmem_cache_destroy(btrfs_inode_cachep);
2191 if (btrfs_trans_handle_cachep)
2192 kmem_cache_destroy(btrfs_trans_handle_cachep);
2193 if (btrfs_transaction_cachep)
2194 kmem_cache_destroy(btrfs_transaction_cachep);
2195 if (btrfs_bit_radix_cachep)
2196 kmem_cache_destroy(btrfs_bit_radix_cachep);
2197 if (btrfs_path_cachep)
2198 kmem_cache_destroy(btrfs_path_cachep);
2199 }
2200
2201 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2202 unsigned long extra_flags,
2203 void (*ctor)(void *, struct kmem_cache *,
2204 unsigned long))
2205 {
2206 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2207 SLAB_MEM_SPREAD | extra_flags), ctor
2208 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2209 ,NULL
2210 #endif
2211 );
2212 }
2213
2214 int btrfs_init_cachep(void)
2215 {
2216 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2217 sizeof(struct btrfs_inode),
2218 0, init_once);
2219 if (!btrfs_inode_cachep)
2220 goto fail;
2221 btrfs_trans_handle_cachep =
2222 btrfs_cache_create("btrfs_trans_handle_cache",
2223 sizeof(struct btrfs_trans_handle),
2224 0, NULL);
2225 if (!btrfs_trans_handle_cachep)
2226 goto fail;
2227 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
2228 sizeof(struct btrfs_transaction),
2229 0, NULL);
2230 if (!btrfs_transaction_cachep)
2231 goto fail;
2232 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
2233 sizeof(struct btrfs_path),
2234 0, NULL);
2235 if (!btrfs_path_cachep)
2236 goto fail;
2237 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
2238 SLAB_DESTROY_BY_RCU, NULL);
2239 if (!btrfs_bit_radix_cachep)
2240 goto fail;
2241 return 0;
2242 fail:
2243 btrfs_destroy_cachep();
2244 return -ENOMEM;
2245 }
2246
2247 static int btrfs_getattr(struct vfsmount *mnt,
2248 struct dentry *dentry, struct kstat *stat)
2249 {
2250 struct inode *inode = dentry->d_inode;
2251 generic_fillattr(inode, stat);
2252 stat->blksize = 256 * 1024;
2253 return 0;
2254 }
2255
2256 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2257 struct inode * new_dir,struct dentry *new_dentry)
2258 {
2259 struct btrfs_trans_handle *trans;
2260 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2261 struct inode *new_inode = new_dentry->d_inode;
2262 struct inode *old_inode = old_dentry->d_inode;
2263 struct timespec ctime = CURRENT_TIME;
2264 struct btrfs_path *path;
2265 struct btrfs_dir_item *di;
2266 int ret;
2267
2268 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2269 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2270 return -ENOTEMPTY;
2271 }
2272
2273 mutex_lock(&root->fs_info->fs_mutex);
2274 trans = btrfs_start_transaction(root, 1);
2275
2276 btrfs_set_trans_block_group(trans, new_dir);
2277 path = btrfs_alloc_path();
2278 if (!path) {
2279 ret = -ENOMEM;
2280 goto out_fail;
2281 }
2282
2283 old_dentry->d_inode->i_nlink++;
2284 old_dir->i_ctime = old_dir->i_mtime = ctime;
2285 new_dir->i_ctime = new_dir->i_mtime = ctime;
2286 old_inode->i_ctime = ctime;
2287
2288 if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) {
2289 struct btrfs_key *location = &BTRFS_I(new_dir)->location;
2290 struct btrfs_key old_parent_key;
2291 di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino,
2292 "..", 2, -1);
2293 if (IS_ERR(di)) {
2294 ret = PTR_ERR(di);
2295 goto out_fail;
2296 }
2297 if (!di) {
2298 ret = -ENOENT;
2299 goto out_fail;
2300 }
2301 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &old_parent_key);
2302 ret = btrfs_del_item(trans, root, path);
2303 if (ret) {
2304 goto out_fail;
2305 }
2306 btrfs_release_path(root, path);
2307
2308 di = btrfs_lookup_dir_index_item(trans, root, path,
2309 old_inode->i_ino,
2310 old_parent_key.objectid,
2311 "..", 2, -1);
2312 if (IS_ERR(di)) {
2313 ret = PTR_ERR(di);
2314 goto out_fail;
2315 }
2316 if (!di) {
2317 ret = -ENOENT;
2318 goto out_fail;
2319 }
2320 ret = btrfs_del_item(trans, root, path);
2321 if (ret) {
2322 goto out_fail;
2323 }
2324 btrfs_release_path(root, path);
2325
2326 ret = btrfs_insert_dir_item(trans, root, "..", 2,
2327 old_inode->i_ino, location,
2328 BTRFS_FT_DIR);
2329 if (ret)
2330 goto out_fail;
2331 }
2332
2333
2334 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2335 if (ret)
2336 goto out_fail;
2337
2338 if (new_inode) {
2339 new_inode->i_ctime = CURRENT_TIME;
2340 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2341 if (ret)
2342 goto out_fail;
2343 }
2344 ret = btrfs_add_link(trans, new_dentry, old_inode);
2345 if (ret)
2346 goto out_fail;
2347
2348 out_fail:
2349 btrfs_free_path(path);
2350 btrfs_end_transaction(trans, root);
2351 mutex_unlock(&root->fs_info->fs_mutex);
2352 return ret;
2353 }
2354
2355 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2356 const char *symname)
2357 {
2358 struct btrfs_trans_handle *trans;
2359 struct btrfs_root *root = BTRFS_I(dir)->root;
2360 struct btrfs_path *path;
2361 struct btrfs_key key;
2362 struct inode *inode;
2363 int err;
2364 int drop_inode = 0;
2365 u64 objectid;
2366 int name_len;
2367 int datasize;
2368 unsigned long ptr;
2369 struct btrfs_file_extent_item *ei;
2370 struct extent_buffer *leaf;
2371 unsigned long nr;
2372
2373 name_len = strlen(symname) + 1;
2374 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2375 return -ENAMETOOLONG;
2376 mutex_lock(&root->fs_info->fs_mutex);
2377 trans = btrfs_start_transaction(root, 1);
2378 btrfs_set_trans_block_group(trans, dir);
2379
2380 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2381 if (err) {
2382 err = -ENOSPC;
2383 goto out_unlock;
2384 }
2385
2386 inode = btrfs_new_inode(trans, root, objectid,
2387 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2388 err = PTR_ERR(inode);
2389 if (IS_ERR(inode))
2390 goto out_unlock;
2391
2392 btrfs_set_trans_block_group(trans, inode);
2393 err = btrfs_add_nondir(trans, dentry, inode);
2394 if (err)
2395 drop_inode = 1;
2396 else {
2397 inode->i_mapping->a_ops = &btrfs_aops;
2398 inode->i_fop = &btrfs_file_operations;
2399 inode->i_op = &btrfs_file_inode_operations;
2400 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
2401 inode->i_mapping, GFP_NOFS);
2402 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
2403 }
2404 dir->i_sb->s_dirt = 1;
2405 btrfs_update_inode_block_group(trans, inode);
2406 btrfs_update_inode_block_group(trans, dir);
2407 if (drop_inode)
2408 goto out_unlock;
2409
2410 path = btrfs_alloc_path();
2411 BUG_ON(!path);
2412 key.objectid = inode->i_ino;
2413 key.offset = 0;
2414 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2415 datasize = btrfs_file_extent_calc_inline_size(name_len);
2416 err = btrfs_insert_empty_item(trans, root, path, &key,
2417 datasize);
2418 if (err) {
2419 drop_inode = 1;
2420 goto out_unlock;
2421 }
2422 leaf = path->nodes[0];
2423 ei = btrfs_item_ptr(leaf, path->slots[0],
2424 struct btrfs_file_extent_item);
2425 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
2426 btrfs_set_file_extent_type(leaf, ei,
2427 BTRFS_FILE_EXTENT_INLINE);
2428 ptr = btrfs_file_extent_inline_start(ei);
2429 write_extent_buffer(leaf, symname, ptr, name_len);
2430 btrfs_mark_buffer_dirty(leaf);
2431 btrfs_free_path(path);
2432
2433 inode->i_op = &btrfs_symlink_inode_operations;
2434 inode->i_mapping->a_ops = &btrfs_symlink_aops;
2435 inode->i_size = name_len - 1;
2436 err = btrfs_update_inode(trans, root, inode);
2437 if (err)
2438 drop_inode = 1;
2439
2440 out_unlock:
2441 nr = trans->blocks_used;
2442 btrfs_end_transaction(trans, root);
2443 mutex_unlock(&root->fs_info->fs_mutex);
2444 if (drop_inode) {
2445 inode_dec_link_count(inode);
2446 iput(inode);
2447 }
2448 btrfs_btree_balance_dirty(root, nr);
2449 return err;
2450 }
2451
2452 static struct inode_operations btrfs_dir_inode_operations = {
2453 .lookup = btrfs_lookup,
2454 .create = btrfs_create,
2455 .unlink = btrfs_unlink,
2456 .link = btrfs_link,
2457 .mkdir = btrfs_mkdir,
2458 .rmdir = btrfs_rmdir,
2459 .rename = btrfs_rename,
2460 .symlink = btrfs_symlink,
2461 .setattr = btrfs_setattr,
2462 .mknod = btrfs_mknod,
2463 };
2464
2465 static struct inode_operations btrfs_dir_ro_inode_operations = {
2466 .lookup = btrfs_lookup,
2467 };
2468
2469 static struct file_operations btrfs_dir_file_operations = {
2470 .llseek = generic_file_llseek,
2471 .read = generic_read_dir,
2472 .readdir = btrfs_readdir,
2473 .unlocked_ioctl = btrfs_ioctl,
2474 #ifdef CONFIG_COMPAT
2475 .compat_ioctl = btrfs_ioctl,
2476 #endif
2477 };
2478
2479 static struct extent_map_ops btrfs_extent_map_ops = {
2480 .fill_delalloc = run_delalloc_range,
2481 .writepage_io_hook = btrfs_writepage_io_hook,
2482 .readpage_io_hook = btrfs_readpage_io_hook,
2483 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
2484 };
2485
2486 static struct address_space_operations btrfs_aops = {
2487 .readpage = btrfs_readpage,
2488 .writepage = btrfs_writepage,
2489 .sync_page = block_sync_page,
2490 .prepare_write = btrfs_prepare_write,
2491 .commit_write = btrfs_commit_write,
2492 .bmap = btrfs_bmap,
2493 .invalidatepage = btrfs_invalidatepage,
2494 .releasepage = btrfs_releasepage,
2495 .set_page_dirty = __set_page_dirty_nobuffers,
2496 };
2497
2498 static struct address_space_operations btrfs_symlink_aops = {
2499 .readpage = btrfs_readpage,
2500 .writepage = btrfs_writepage,
2501 .invalidatepage = btrfs_invalidatepage,
2502 .releasepage = btrfs_releasepage,
2503 };
2504
2505 static struct inode_operations btrfs_file_inode_operations = {
2506 .truncate = btrfs_truncate,
2507 .getattr = btrfs_getattr,
2508 .setattr = btrfs_setattr,
2509 };
2510
2511 static struct inode_operations btrfs_special_inode_operations = {
2512 .getattr = btrfs_getattr,
2513 .setattr = btrfs_setattr,
2514 };
2515
2516 static struct inode_operations btrfs_symlink_inode_operations = {
2517 .readlink = generic_readlink,
2518 .follow_link = page_follow_link_light,
2519 .put_link = page_put_link,
2520 };
Linux kernel - Deliverables
This page took 0.09519 seconds and 6 git commands to generate.