projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Btrfs: Fix typo in .. check (thanks Yan)
[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 <linux/xattr.h>
36 #include "ctree.h"
37 #include "disk-io.h"
38 #include "transaction.h"
39 #include "btrfs_inode.h"
40 #include "ioctl.h"
41 #include "print-tree.h"
42
43 struct btrfs_iget_args {
44 u64 ino;
45 struct btrfs_root *root;
46 };
47
48 static struct inode_operations btrfs_dir_inode_operations;
49 static struct inode_operations btrfs_symlink_inode_operations;
50 static struct inode_operations btrfs_dir_ro_inode_operations;
51 static struct inode_operations btrfs_special_inode_operations;
52 static struct inode_operations btrfs_file_inode_operations;
53 static struct address_space_operations btrfs_aops;
54 static struct address_space_operations btrfs_symlink_aops;
55 static struct file_operations btrfs_dir_file_operations;
56 static struct extent_map_ops btrfs_extent_map_ops;
57
58 static struct kmem_cache *btrfs_inode_cachep;
59 struct kmem_cache *btrfs_trans_handle_cachep;
60 struct kmem_cache *btrfs_transaction_cachep;
61 struct kmem_cache *btrfs_bit_radix_cachep;
62 struct kmem_cache *btrfs_path_cachep;
63
64 #define S_SHIFT 12
65 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
66 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
67 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
68 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
69 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
70 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
71 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
72 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
73 };
74
75 static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
76 {
77 struct btrfs_root *root = BTRFS_I(inode)->root;
78 struct btrfs_trans_handle *trans;
79 struct btrfs_key ins;
80 u64 alloc_hint = 0;
81 u64 num_bytes;
82 int ret;
83 u64 blocksize = root->sectorsize;
84
85 mutex_lock(&root->fs_info->fs_mutex);
86 trans = btrfs_start_transaction(root, 1);
87 btrfs_set_trans_block_group(trans, inode);
88 BUG_ON(!trans);
89 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
90 ret = btrfs_drop_extents(trans, root, inode,
91 start, start + num_bytes, start, &alloc_hint);
92
93 if (alloc_hint == EXTENT_MAP_INLINE)
94 goto out;
95
96 ret = btrfs_alloc_extent(trans, root, num_bytes,
97 root->root_key.objectid, trans->transid,
98 inode->i_ino, start, 0,
99 alloc_hint, (u64)-1, &ins, 1);
100 if (ret) {
101 WARN_ON(1);
102 goto out;
103 }
104 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
105 start, ins.objectid, ins.offset,
106 ins.offset);
107 out:
108 btrfs_end_transaction(trans, root);
109 mutex_unlock(&root->fs_info->fs_mutex);
110 return ret;
111 }
112
113 int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end)
114 {
115 struct inode *inode = page->mapping->host;
116 struct btrfs_root *root = BTRFS_I(inode)->root;
117 struct btrfs_trans_handle *trans;
118 char *kaddr;
119 int ret;
120 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
121 size_t offset = start - page_start;
122
123 mutex_lock(&root->fs_info->fs_mutex);
124 trans = btrfs_start_transaction(root, 1);
125 btrfs_set_trans_block_group(trans, inode);
126 kaddr = kmap(page);
127 btrfs_csum_file_block(trans, root, inode, inode->i_ino,
128 start, kaddr + offset, end - start + 1);
129 kunmap(page);
130 ret = btrfs_end_transaction(trans, root);
131 BUG_ON(ret);
132 mutex_unlock(&root->fs_info->fs_mutex);
133 return ret;
134 }
135
136 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
137 {
138 int ret = 0;
139 struct inode *inode = page->mapping->host;
140 struct btrfs_root *root = BTRFS_I(inode)->root;
141 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
142 struct btrfs_csum_item *item;
143 struct btrfs_path *path = NULL;
144 u32 csum;
145
146 mutex_lock(&root->fs_info->fs_mutex);
147 path = btrfs_alloc_path();
148 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
149 if (IS_ERR(item)) {
150 ret = PTR_ERR(item);
151 /* a csum that isn't present is a preallocated region. */
152 if (ret == -ENOENT || ret == -EFBIG)
153 ret = 0;
154 csum = 0;
155 goto out;
156 }
157 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
158 BTRFS_CRC32_SIZE);
159 set_state_private(em_tree, start, csum);
160 out:
161 if (path)
162 btrfs_free_path(path);
163 mutex_unlock(&root->fs_info->fs_mutex);
164 return ret;
165 }
166
167 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
168 {
169 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
170 struct inode *inode = page->mapping->host;
171 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
172 char *kaddr;
173 u64 private;
174 int ret;
175 struct btrfs_root *root = BTRFS_I(inode)->root;
176 u32 csum = ~(u32)0;
177 unsigned long flags;
178
179 ret = get_state_private(em_tree, start, &private);
180 local_irq_save(flags);
181 kaddr = kmap_atomic(page, KM_IRQ0);
182 if (ret) {
183 goto zeroit;
184 }
185 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
186 btrfs_csum_final(csum, (char *)&csum);
187 if (csum != private) {
188 goto zeroit;
189 }
190 kunmap_atomic(kaddr, KM_IRQ0);
191 local_irq_restore(flags);
192 return 0;
193
194 zeroit:
195 printk("btrfs csum failed ino %lu off %llu\n",
196 page->mapping->host->i_ino, (unsigned long long)start);
197 memset(kaddr + offset, 1, end - start + 1);
198 flush_dcache_page(page);
199 kunmap_atomic(kaddr, KM_IRQ0);
200 local_irq_restore(flags);
201 return 0;
202 }
203
204 void btrfs_read_locked_inode(struct inode *inode)
205 {
206 struct btrfs_path *path;
207 struct extent_buffer *leaf;
208 struct btrfs_inode_item *inode_item;
209 struct btrfs_inode_timespec *tspec;
210 struct btrfs_root *root = BTRFS_I(inode)->root;
211 struct btrfs_key location;
212 u64 alloc_group_block;
213 u32 rdev;
214 int ret;
215
216 path = btrfs_alloc_path();
217 BUG_ON(!path);
218 mutex_lock(&root->fs_info->fs_mutex);
219
220 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
221 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
222 if (ret)
223 goto make_bad;
224
225 leaf = path->nodes[0];
226 inode_item = btrfs_item_ptr(leaf, path->slots[0],
227 struct btrfs_inode_item);
228
229 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
230 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
231 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
232 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
233 inode->i_size = btrfs_inode_size(leaf, inode_item);
234
235 tspec = btrfs_inode_atime(inode_item);
236 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
237 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
238
239 tspec = btrfs_inode_mtime(inode_item);
240 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
241 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
242
243 tspec = btrfs_inode_ctime(inode_item);
244 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
245 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
246
247 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
248 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
249 inode->i_rdev = 0;
250 rdev = btrfs_inode_rdev(leaf, inode_item);
251
252 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
253 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
254 alloc_group_block);
255
256 btrfs_free_path(path);
257 inode_item = NULL;
258
259 mutex_unlock(&root->fs_info->fs_mutex);
260
261 switch (inode->i_mode & S_IFMT) {
262 case S_IFREG:
263 inode->i_mapping->a_ops = &btrfs_aops;
264 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
265 inode->i_fop = &btrfs_file_operations;
266 inode->i_op = &btrfs_file_inode_operations;
267 break;
268 case S_IFDIR:
269 inode->i_fop = &btrfs_dir_file_operations;
270 if (root == root->fs_info->tree_root)
271 inode->i_op = &btrfs_dir_ro_inode_operations;
272 else
273 inode->i_op = &btrfs_dir_inode_operations;
274 break;
275 case S_IFLNK:
276 inode->i_op = &btrfs_symlink_inode_operations;
277 inode->i_mapping->a_ops = &btrfs_symlink_aops;
278 break;
279 default:
280 init_special_inode(inode, inode->i_mode, rdev);
281 break;
282 }
283 return;
284
285 make_bad:
286 btrfs_release_path(root, path);
287 btrfs_free_path(path);
288 mutex_unlock(&root->fs_info->fs_mutex);
289 make_bad_inode(inode);
290 }
291
292 static void fill_inode_item(struct extent_buffer *leaf,
293 struct btrfs_inode_item *item,
294 struct inode *inode)
295 {
296 btrfs_set_inode_uid(leaf, item, inode->i_uid);
297 btrfs_set_inode_gid(leaf, item, inode->i_gid);
298 btrfs_set_inode_size(leaf, item, inode->i_size);
299 btrfs_set_inode_mode(leaf, item, inode->i_mode);
300 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
301
302 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
303 inode->i_atime.tv_sec);
304 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
305 inode->i_atime.tv_nsec);
306
307 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
308 inode->i_mtime.tv_sec);
309 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
310 inode->i_mtime.tv_nsec);
311
312 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
313 inode->i_ctime.tv_sec);
314 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
315 inode->i_ctime.tv_nsec);
316
317 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
318 btrfs_set_inode_generation(leaf, item, inode->i_generation);
319 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
320 btrfs_set_inode_block_group(leaf, item,
321 BTRFS_I(inode)->block_group->key.objectid);
322 }
323
324 int btrfs_update_inode(struct btrfs_trans_handle *trans,
325 struct btrfs_root *root,
326 struct inode *inode)
327 {
328 struct btrfs_inode_item *inode_item;
329 struct btrfs_path *path;
330 struct extent_buffer *leaf;
331 int ret;
332
333 path = btrfs_alloc_path();
334 BUG_ON(!path);
335 ret = btrfs_lookup_inode(trans, root, path,
336 &BTRFS_I(inode)->location, 1);
337 if (ret) {
338 if (ret > 0)
339 ret = -ENOENT;
340 goto failed;
341 }
342
343 leaf = path->nodes[0];
344 inode_item = btrfs_item_ptr(leaf, path->slots[0],
345 struct btrfs_inode_item);
346
347 fill_inode_item(leaf, inode_item, inode);
348 btrfs_mark_buffer_dirty(leaf);
349 btrfs_set_inode_last_trans(trans, inode);
350 ret = 0;
351 failed:
352 btrfs_release_path(root, path);
353 btrfs_free_path(path);
354 return ret;
355 }
356
357
358 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
359 struct btrfs_root *root,
360 struct inode *dir,
361 struct dentry *dentry)
362 {
363 struct btrfs_path *path;
364 const char *name = dentry->d_name.name;
365 int name_len = dentry->d_name.len;
366 int ret = 0;
367 struct extent_buffer *leaf;
368 struct btrfs_dir_item *di;
369 struct btrfs_key key;
370
371 path = btrfs_alloc_path();
372 if (!path) {
373 ret = -ENOMEM;
374 goto err;
375 }
376
377 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
378 name, name_len, -1);
379 if (IS_ERR(di)) {
380 ret = PTR_ERR(di);
381 goto err;
382 }
383 if (!di) {
384 ret = -ENOENT;
385 goto err;
386 }
387 leaf = path->nodes[0];
388 btrfs_dir_item_key_to_cpu(leaf, di, &key);
389 ret = btrfs_delete_one_dir_name(trans, root, path, di);
390 if (ret)
391 goto err;
392 btrfs_release_path(root, path);
393
394 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
395 key.objectid, name, name_len, -1);
396 if (IS_ERR(di)) {
397 ret = PTR_ERR(di);
398 goto err;
399 }
400 if (!di) {
401 ret = -ENOENT;
402 goto err;
403 }
404 ret = btrfs_delete_one_dir_name(trans, root, path, di);
405
406 dentry->d_inode->i_ctime = dir->i_ctime;
407 ret = btrfs_del_inode_ref(trans, root, name, name_len,
408 dentry->d_inode->i_ino,
409 dentry->d_parent->d_inode->i_ino);
410 if (ret) {
411 printk("failed to delete reference to %.*s, "
412 "inode %lu parent %lu\n", name_len, name,
413 dentry->d_inode->i_ino,
414 dentry->d_parent->d_inode->i_ino);
415 }
416 err:
417 btrfs_free_path(path);
418 if (!ret) {
419 dir->i_size -= name_len * 2;
420 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
421 btrfs_update_inode(trans, root, dir);
422 drop_nlink(dentry->d_inode);
423 ret = btrfs_update_inode(trans, root, dentry->d_inode);
424 dir->i_sb->s_dirt = 1;
425 }
426 return ret;
427 }
428
429 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
430 {
431 struct btrfs_root *root;
432 struct btrfs_trans_handle *trans;
433 int ret;
434 unsigned long nr;
435
436 root = BTRFS_I(dir)->root;
437 mutex_lock(&root->fs_info->fs_mutex);
438 trans = btrfs_start_transaction(root, 1);
439
440 btrfs_set_trans_block_group(trans, dir);
441 ret = btrfs_unlink_trans(trans, root, dir, dentry);
442 nr = trans->blocks_used;
443
444 btrfs_end_transaction(trans, root);
445 mutex_unlock(&root->fs_info->fs_mutex);
446 btrfs_btree_balance_dirty(root, nr);
447
448 return ret;
449 }
450
451 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
452 {
453 struct inode *inode = dentry->d_inode;
454 int err;
455 int ret;
456 struct btrfs_root *root = BTRFS_I(dir)->root;
457 struct btrfs_trans_handle *trans;
458 unsigned long nr;
459
460 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
461 return -ENOTEMPTY;
462
463 mutex_lock(&root->fs_info->fs_mutex);
464 trans = btrfs_start_transaction(root, 1);
465 btrfs_set_trans_block_group(trans, dir);
466
467 /* now the directory is empty */
468 err = btrfs_unlink_trans(trans, root, dir, dentry);
469 if (!err) {
470 inode->i_size = 0;
471 }
472
473 nr = trans->blocks_used;
474 ret = btrfs_end_transaction(trans, root);
475 mutex_unlock(&root->fs_info->fs_mutex);
476 btrfs_btree_balance_dirty(root, nr);
477
478 if (ret && !err)
479 err = ret;
480 return err;
481 }
482
483 static int btrfs_free_inode(struct btrfs_trans_handle *trans,
484 struct btrfs_root *root,
485 struct inode *inode)
486 {
487 struct btrfs_path *path;
488 int ret;
489
490 clear_inode(inode);
491
492 path = btrfs_alloc_path();
493 BUG_ON(!path);
494 ret = btrfs_lookup_inode(trans, root, path,
495 &BTRFS_I(inode)->location, -1);
496 if (ret > 0)
497 ret = -ENOENT;
498 if (!ret)
499 ret = btrfs_del_item(trans, root, path);
500 btrfs_free_path(path);
501 return ret;
502 }
503
504 /*
505 * this can truncate away extent items, csum items and directory items.
506 * It starts at a high offset and removes keys until it can't find
507 * any higher than i_size.
508 *
509 * csum items that cross the new i_size are truncated to the new size
510 * as well.
511 */
512 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
513 struct btrfs_root *root,
514 struct inode *inode)
515 {
516 int ret;
517 struct btrfs_path *path;
518 struct btrfs_key key;
519 struct btrfs_key found_key;
520 u32 found_type;
521 struct extent_buffer *leaf;
522 struct btrfs_file_extent_item *fi;
523 u64 extent_start = 0;
524 u64 extent_num_bytes = 0;
525 u64 item_end = 0;
526 u64 root_gen = 0;
527 u64 root_owner = 0;
528 int found_extent;
529 int del_item;
530 int extent_type = -1;
531
532 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
533 path = btrfs_alloc_path();
534 path->reada = -1;
535 BUG_ON(!path);
536
537 /* FIXME, add redo link to tree so we don't leak on crash */
538 key.objectid = inode->i_ino;
539 key.offset = (u64)-1;
540 key.type = (u8)-1;
541
542 while(1) {
543 btrfs_init_path(path);
544 fi = NULL;
545 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
546 if (ret < 0) {
547 goto error;
548 }
549 if (ret > 0) {
550 BUG_ON(path->slots[0] == 0);
551 path->slots[0]--;
552 }
553 leaf = path->nodes[0];
554 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
555 found_type = btrfs_key_type(&found_key);
556
557 if (found_key.objectid != inode->i_ino)
558 break;
559
560 if (found_type != BTRFS_CSUM_ITEM_KEY &&
561 found_type != BTRFS_DIR_ITEM_KEY &&
562 found_type != BTRFS_DIR_INDEX_KEY &&
563 found_type != BTRFS_EXTENT_DATA_KEY)
564 break;
565
566 item_end = found_key.offset;
567 if (found_type == BTRFS_EXTENT_DATA_KEY) {
568 fi = btrfs_item_ptr(leaf, path->slots[0],
569 struct btrfs_file_extent_item);
570 extent_type = btrfs_file_extent_type(leaf, fi);
571 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
572 item_end +=
573 btrfs_file_extent_num_bytes(leaf, fi);
574 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
575 struct btrfs_item *item = btrfs_item_nr(leaf,
576 path->slots[0]);
577 item_end += btrfs_file_extent_inline_len(leaf,
578 item);
579 }
580 item_end--;
581 }
582 if (found_type == BTRFS_CSUM_ITEM_KEY) {
583 ret = btrfs_csum_truncate(trans, root, path,
584 inode->i_size);
585 BUG_ON(ret);
586 }
587 if (item_end < inode->i_size) {
588 if (found_type == BTRFS_DIR_ITEM_KEY) {
589 found_type = BTRFS_INODE_ITEM_KEY;
590 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
591 found_type = BTRFS_CSUM_ITEM_KEY;
592 } else if (found_type) {
593 found_type--;
594 } else {
595 break;
596 }
597 btrfs_set_key_type(&key, found_type);
598 btrfs_release_path(root, path);
599 continue;
600 }
601 if (found_key.offset >= inode->i_size)
602 del_item = 1;
603 else
604 del_item = 0;
605 found_extent = 0;
606
607 /* FIXME, shrink the extent if the ref count is only 1 */
608 if (found_type != BTRFS_EXTENT_DATA_KEY)
609 goto delete;
610
611 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
612 u64 num_dec;
613 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
614 if (!del_item) {
615 u64 orig_num_bytes =
616 btrfs_file_extent_num_bytes(leaf, fi);
617 extent_num_bytes = inode->i_size -
618 found_key.offset + root->sectorsize - 1;
619 btrfs_set_file_extent_num_bytes(leaf, fi,
620 extent_num_bytes);
621 num_dec = (orig_num_bytes -
622 extent_num_bytes) >> 9;
623 if (extent_start != 0) {
624 inode->i_blocks -= num_dec;
625 }
626 btrfs_mark_buffer_dirty(leaf);
627 } else {
628 extent_num_bytes =
629 btrfs_file_extent_disk_num_bytes(leaf,
630 fi);
631 /* FIXME blocksize != 4096 */
632 num_dec = btrfs_file_extent_num_bytes(leaf,
633 fi) >> 9;
634 if (extent_start != 0) {
635 found_extent = 1;
636 inode->i_blocks -= num_dec;
637 }
638 root_gen = btrfs_header_generation(leaf);
639 root_owner = btrfs_header_owner(leaf);
640 }
641 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE &&
642 !del_item) {
643 u32 newsize = inode->i_size - found_key.offset;
644 newsize = btrfs_file_extent_calc_inline_size(newsize);
645 ret = btrfs_truncate_item(trans, root, path,
646 newsize, 1);
647 BUG_ON(ret);
648 }
649 delete:
650 if (del_item) {
651 ret = btrfs_del_item(trans, root, path);
652 if (ret)
653 goto error;
654 } else {
655 break;
656 }
657 btrfs_release_path(root, path);
658 if (found_extent) {
659 ret = btrfs_free_extent(trans, root, extent_start,
660 extent_num_bytes,
661 root_owner,
662 root_gen, inode->i_ino,
663 found_key.offset, 0);
664 BUG_ON(ret);
665 }
666 }
667 ret = 0;
668 error:
669 btrfs_release_path(root, path);
670 btrfs_free_path(path);
671 inode->i_sb->s_dirt = 1;
672 return ret;
673 }
674
675 static int btrfs_cow_one_page(struct inode *inode, struct page *page,
676 size_t zero_start)
677 {
678 char *kaddr;
679 int ret = 0;
680 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
681 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
682 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
683
684 set_page_extent_mapped(page);
685
686 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
687 set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start,
688 page_end, GFP_NOFS);
689 if (zero_start != PAGE_CACHE_SIZE) {
690 kaddr = kmap(page);
691 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
692 flush_dcache_page(page);
693 kunmap(page);
694 }
695 set_page_dirty(page);
696 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
697
698 return ret;
699 }
700
701 /*
702 * taken from block_truncate_page, but does cow as it zeros out
703 * any bytes left in the last page in the file.
704 */
705 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
706 {
707 struct inode *inode = mapping->host;
708 struct btrfs_root *root = BTRFS_I(inode)->root;
709 u32 blocksize = root->sectorsize;
710 pgoff_t index = from >> PAGE_CACHE_SHIFT;
711 unsigned offset = from & (PAGE_CACHE_SIZE-1);
712 struct page *page;
713 int ret = 0;
714 u64 page_start;
715
716 if ((offset & (blocksize - 1)) == 0)
717 goto out;
718
719 down_read(&root->snap_sem);
720 ret = -ENOMEM;
721 page = grab_cache_page(mapping, index);
722 if (!page)
723 goto out;
724 if (!PageUptodate(page)) {
725 ret = btrfs_readpage(NULL, page);
726 lock_page(page);
727 if (!PageUptodate(page)) {
728 ret = -EIO;
729 goto out;
730 }
731 }
732 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
733
734 ret = btrfs_cow_one_page(inode, page, offset);
735
736 unlock_page(page);
737 page_cache_release(page);
738 up_read(&BTRFS_I(inode)->root->snap_sem);
739 out:
740 return ret;
741 }
742
743 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
744 {
745 struct inode *inode = dentry->d_inode;
746 int err;
747
748 err = inode_change_ok(inode, attr);
749 if (err)
750 return err;
751
752 if (S_ISREG(inode->i_mode) &&
753 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
754 struct btrfs_trans_handle *trans;
755 struct btrfs_root *root = BTRFS_I(inode)->root;
756 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
757
758 u64 mask = root->sectorsize - 1;
759 u64 pos = (inode->i_size + mask) & ~mask;
760 u64 block_end = attr->ia_size | mask;
761 u64 hole_size;
762 u64 alloc_hint = 0;
763
764 if (attr->ia_size <= pos)
765 goto out;
766
767 btrfs_truncate_page(inode->i_mapping, inode->i_size);
768
769 lock_extent(em_tree, pos, block_end, GFP_NOFS);
770 hole_size = (attr->ia_size - pos + mask) & ~mask;
771
772 mutex_lock(&root->fs_info->fs_mutex);
773 trans = btrfs_start_transaction(root, 1);
774 btrfs_set_trans_block_group(trans, inode);
775 err = btrfs_drop_extents(trans, root, inode,
776 pos, pos + hole_size, pos,
777 &alloc_hint);
778
779 if (alloc_hint != EXTENT_MAP_INLINE) {
780 err = btrfs_insert_file_extent(trans, root,
781 inode->i_ino,
782 pos, 0, 0, hole_size);
783 }
784 btrfs_end_transaction(trans, root);
785 mutex_unlock(&root->fs_info->fs_mutex);
786 unlock_extent(em_tree, pos, block_end, GFP_NOFS);
787 if (err)
788 return err;
789 }
790 out:
791 err = inode_setattr(inode, attr);
792
793 return err;
794 }
795 void btrfs_delete_inode(struct inode *inode)
796 {
797 struct btrfs_trans_handle *trans;
798 struct btrfs_root *root = BTRFS_I(inode)->root;
799 unsigned long nr;
800 int ret;
801
802 truncate_inode_pages(&inode->i_data, 0);
803 if (is_bad_inode(inode)) {
804 goto no_delete;
805 }
806
807 inode->i_size = 0;
808 mutex_lock(&root->fs_info->fs_mutex);
809 trans = btrfs_start_transaction(root, 1);
810
811 btrfs_set_trans_block_group(trans, inode);
812 ret = btrfs_truncate_in_trans(trans, root, inode);
813 if (ret)
814 goto no_delete_lock;
815 ret = btrfs_delete_xattrs(trans, root, inode);
816 if (ret)
817 goto no_delete_lock;
818 ret = btrfs_free_inode(trans, root, inode);
819 if (ret)
820 goto no_delete_lock;
821 nr = trans->blocks_used;
822
823 btrfs_end_transaction(trans, root);
824 mutex_unlock(&root->fs_info->fs_mutex);
825 btrfs_btree_balance_dirty(root, nr);
826 return;
827
828 no_delete_lock:
829 nr = trans->blocks_used;
830 btrfs_end_transaction(trans, root);
831 mutex_unlock(&root->fs_info->fs_mutex);
832 btrfs_btree_balance_dirty(root, nr);
833 no_delete:
834 clear_inode(inode);
835 }
836
837 /*
838 * this returns the key found in the dir entry in the location pointer.
839 * If no dir entries were found, location->objectid is 0.
840 */
841 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
842 struct btrfs_key *location)
843 {
844 const char *name = dentry->d_name.name;
845 int namelen = dentry->d_name.len;
846 struct btrfs_dir_item *di;
847 struct btrfs_path *path;
848 struct btrfs_root *root = BTRFS_I(dir)->root;
849 int ret = 0;
850
851 if (namelen == 1 && strcmp(name, ".") == 0) {
852 location->objectid = dir->i_ino;
853 location->type = BTRFS_INODE_ITEM_KEY;
854 location->offset = 0;
855 return 0;
856 }
857 path = btrfs_alloc_path();
858 BUG_ON(!path);
859
860 if (namelen == 2 && strcmp(name, "..") == 0) {
861 struct btrfs_key key;
862 struct extent_buffer *leaf;
863 u32 nritems;
864 int slot;
865
866 key.objectid = dir->i_ino;
867 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
868 key.offset = 0;
869 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
870 BUG_ON(ret == 0);
871 ret = 0;
872
873 leaf = path->nodes[0];
874 slot = path->slots[0];
875 nritems = btrfs_header_nritems(leaf);
876 if (slot >= nritems)
877 goto out_err;
878
879 btrfs_item_key_to_cpu(leaf, &key, slot);
880 if (key.objectid != dir->i_ino ||
881 key.type != BTRFS_INODE_REF_KEY) {
882 goto out_err;
883 }
884 location->objectid = key.offset;
885 location->type = BTRFS_INODE_ITEM_KEY;
886 location->offset = 0;
887 goto out;
888 }
889
890 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
891 namelen, 0);
892 if (IS_ERR(di))
893 ret = PTR_ERR(di);
894 if (!di || IS_ERR(di)) {
895 goto out_err;
896 }
897 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
898 out:
899 btrfs_free_path(path);
900 return ret;
901 out_err:
902 location->objectid = 0;
903 goto out;
904 }
905
906 /*
907 * when we hit a tree root in a directory, the btrfs part of the inode
908 * needs to be changed to reflect the root directory of the tree root. This
909 * is kind of like crossing a mount point.
910 */
911 static int fixup_tree_root_location(struct btrfs_root *root,
912 struct btrfs_key *location,
913 struct btrfs_root **sub_root,
914 struct dentry *dentry)
915 {
916 struct btrfs_path *path;
917 struct btrfs_root_item *ri;
918
919 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
920 return 0;
921 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
922 return 0;
923
924 path = btrfs_alloc_path();
925 BUG_ON(!path);
926 mutex_lock(&root->fs_info->fs_mutex);
927
928 *sub_root = btrfs_read_fs_root(root->fs_info, location,
929 dentry->d_name.name,
930 dentry->d_name.len);
931 if (IS_ERR(*sub_root))
932 return PTR_ERR(*sub_root);
933
934 ri = &(*sub_root)->root_item;
935 location->objectid = btrfs_root_dirid(ri);
936 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
937 location->offset = 0;
938
939 btrfs_free_path(path);
940 mutex_unlock(&root->fs_info->fs_mutex);
941 return 0;
942 }
943
944 static int btrfs_init_locked_inode(struct inode *inode, void *p)
945 {
946 struct btrfs_iget_args *args = p;
947 inode->i_ino = args->ino;
948 BTRFS_I(inode)->root = args->root;
949 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
950 inode->i_mapping, GFP_NOFS);
951 return 0;
952 }
953
954 static int btrfs_find_actor(struct inode *inode, void *opaque)
955 {
956 struct btrfs_iget_args *args = opaque;
957 return (args->ino == inode->i_ino &&
958 args->root == BTRFS_I(inode)->root);
959 }
960
961 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
962 struct btrfs_root *root)
963 {
964 struct inode *inode;
965 struct btrfs_iget_args args;
966 args.ino = objectid;
967 args.root = root;
968
969 inode = iget5_locked(s, objectid, btrfs_find_actor,
970 btrfs_init_locked_inode,
971 (void *)&args);
972 return inode;
973 }
974
975 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
976 struct nameidata *nd)
977 {
978 struct inode * inode;
979 struct btrfs_inode *bi = BTRFS_I(dir);
980 struct btrfs_root *root = bi->root;
981 struct btrfs_root *sub_root = root;
982 struct btrfs_key location;
983 int ret;
984
985 if (dentry->d_name.len > BTRFS_NAME_LEN)
986 return ERR_PTR(-ENAMETOOLONG);
987
988 mutex_lock(&root->fs_info->fs_mutex);
989 ret = btrfs_inode_by_name(dir, dentry, &location);
990 mutex_unlock(&root->fs_info->fs_mutex);
991
992 if (ret < 0)
993 return ERR_PTR(ret);
994
995 inode = NULL;
996 if (location.objectid) {
997 ret = fixup_tree_root_location(root, &location, &sub_root,
998 dentry);
999 if (ret < 0)
1000 return ERR_PTR(ret);
1001 if (ret > 0)
1002 return ERR_PTR(-ENOENT);
1003 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
1004 sub_root);
1005 if (!inode)
1006 return ERR_PTR(-EACCES);
1007 if (inode->i_state & I_NEW) {
1008 /* the inode and parent dir are two different roots */
1009 if (sub_root != root) {
1010 igrab(inode);
1011 sub_root->inode = inode;
1012 }
1013 BTRFS_I(inode)->root = sub_root;
1014 memcpy(&BTRFS_I(inode)->location, &location,
1015 sizeof(location));
1016 btrfs_read_locked_inode(inode);
1017 unlock_new_inode(inode);
1018 }
1019 }
1020 return d_splice_alias(inode, dentry);
1021 }
1022
1023 static unsigned char btrfs_filetype_table[] = {
1024 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1025 };
1026
1027 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1028 {
1029 struct inode *inode = filp->f_path.dentry->d_inode;
1030 struct btrfs_root *root = BTRFS_I(inode)->root;
1031 struct btrfs_item *item;
1032 struct btrfs_dir_item *di;
1033 struct btrfs_key key;
1034 struct btrfs_key found_key;
1035 struct btrfs_path *path;
1036 int ret;
1037 u32 nritems;
1038 struct extent_buffer *leaf;
1039 int slot;
1040 int advance;
1041 unsigned char d_type;
1042 int over = 0;
1043 u32 di_cur;
1044 u32 di_total;
1045 u32 di_len;
1046 int key_type = BTRFS_DIR_INDEX_KEY;
1047 char tmp_name[32];
1048 char *name_ptr;
1049 int name_len;
1050
1051 /* FIXME, use a real flag for deciding about the key type */
1052 if (root->fs_info->tree_root == root)
1053 key_type = BTRFS_DIR_ITEM_KEY;
1054
1055 /* special case for "." */
1056 if (filp->f_pos == 0) {
1057 over = filldir(dirent, ".", 1,
1058 1, inode->i_ino,
1059 DT_DIR);
1060 if (over)
1061 return 0;
1062 filp->f_pos = 1;
1063 }
1064
1065 mutex_lock(&root->fs_info->fs_mutex);
1066 key.objectid = inode->i_ino;
1067 path = btrfs_alloc_path();
1068 path->reada = 2;
1069
1070 /* special case for .., just use the back ref */
1071 if (filp->f_pos == 1) {
1072 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1073 key.offset = 0;
1074 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1075 BUG_ON(ret == 0);
1076 leaf = path->nodes[0];
1077 slot = path->slots[0];
1078 nritems = btrfs_header_nritems(leaf);
1079 if (slot >= nritems) {
1080 btrfs_release_path(root, path);
1081 goto read_dir_items;
1082 }
1083 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1084 btrfs_release_path(root, path);
1085 if (found_key.objectid != key.objectid ||
1086 found_key.type != BTRFS_INODE_REF_KEY)
1087 goto read_dir_items;
1088 over = filldir(dirent, "..", 2,
1089 2, found_key.offset, DT_DIR);
1090 if (over)
1091 goto nopos;
1092 filp->f_pos = 2;
1093 }
1094
1095 read_dir_items:
1096 btrfs_set_key_type(&key, key_type);
1097 key.offset = filp->f_pos;
1098
1099 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1100 if (ret < 0)
1101 goto err;
1102 advance = 0;
1103 while(1) {
1104 leaf = path->nodes[0];
1105 nritems = btrfs_header_nritems(leaf);
1106 slot = path->slots[0];
1107 if (advance || slot >= nritems) {
1108 if (slot >= nritems -1) {
1109 ret = btrfs_next_leaf(root, path);
1110 if (ret)
1111 break;
1112 leaf = path->nodes[0];
1113 nritems = btrfs_header_nritems(leaf);
1114 slot = path->slots[0];
1115 } else {
1116 slot++;
1117 path->slots[0]++;
1118 }
1119 }
1120 advance = 1;
1121 item = btrfs_item_nr(leaf, slot);
1122 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1123
1124 if (found_key.objectid != key.objectid)
1125 break;
1126 if (btrfs_key_type(&found_key) != key_type)
1127 break;
1128 if (found_key.offset < filp->f_pos)
1129 continue;
1130
1131 filp->f_pos = found_key.offset;
1132 advance = 1;
1133 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1134 di_cur = 0;
1135 di_total = btrfs_item_size(leaf, item);
1136 while(di_cur < di_total) {
1137 struct btrfs_key location;
1138
1139 name_len = btrfs_dir_name_len(leaf, di);
1140 if (name_len < 32) {
1141 name_ptr = tmp_name;
1142 } else {
1143 name_ptr = kmalloc(name_len, GFP_NOFS);
1144 BUG_ON(!name_ptr);
1145 }
1146 read_extent_buffer(leaf, name_ptr,
1147 (unsigned long)(di + 1), name_len);
1148
1149 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1150 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1151
1152 over = filldir(dirent, name_ptr, name_len,
1153 found_key.offset,
1154 location.objectid,
1155 d_type);
1156
1157 if (name_ptr != tmp_name)
1158 kfree(name_ptr);
1159
1160 if (over)
1161 goto nopos;
1162 di_len = btrfs_dir_name_len(leaf, di) +
1163 btrfs_dir_data_len(leaf, di) +sizeof(*di);
1164 di_cur += di_len;
1165 di = (struct btrfs_dir_item *)((char *)di + di_len);
1166 }
1167 }
1168 filp->f_pos++;
1169 nopos:
1170 ret = 0;
1171 err:
1172 btrfs_release_path(root, path);
1173 btrfs_free_path(path);
1174 mutex_unlock(&root->fs_info->fs_mutex);
1175 return ret;
1176 }
1177
1178 int btrfs_write_inode(struct inode *inode, int wait)
1179 {
1180 struct btrfs_root *root = BTRFS_I(inode)->root;
1181 struct btrfs_trans_handle *trans;
1182 int ret = 0;
1183
1184 if (wait) {
1185 mutex_lock(&root->fs_info->fs_mutex);
1186 trans = btrfs_start_transaction(root, 1);
1187 btrfs_set_trans_block_group(trans, inode);
1188 ret = btrfs_commit_transaction(trans, root);
1189 mutex_unlock(&root->fs_info->fs_mutex);
1190 }
1191 return ret;
1192 }
1193
1194 /*
1195 * This is somewhat expensive, updating the tree every time the
1196 * inode changes. But, it is most likely to find the inode in cache.
1197 * FIXME, needs more benchmarking...there are no reasons other than performance
1198 * to keep or drop this code.
1199 */
1200 void btrfs_dirty_inode(struct inode *inode)
1201 {
1202 struct btrfs_root *root = BTRFS_I(inode)->root;
1203 struct btrfs_trans_handle *trans;
1204
1205 mutex_lock(&root->fs_info->fs_mutex);
1206 trans = btrfs_start_transaction(root, 1);
1207 btrfs_set_trans_block_group(trans, inode);
1208 btrfs_update_inode(trans, root, inode);
1209 btrfs_end_transaction(trans, root);
1210 mutex_unlock(&root->fs_info->fs_mutex);
1211 }
1212
1213 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1214 struct btrfs_root *root,
1215 u64 objectid,
1216 struct btrfs_block_group_cache *group,
1217 int mode)
1218 {
1219 struct inode *inode;
1220 struct btrfs_inode_item *inode_item;
1221 struct btrfs_key *location;
1222 struct btrfs_path *path;
1223 int ret;
1224 int owner;
1225
1226 path = btrfs_alloc_path();
1227 BUG_ON(!path);
1228
1229 inode = new_inode(root->fs_info->sb);
1230 if (!inode)
1231 return ERR_PTR(-ENOMEM);
1232
1233 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1234 inode->i_mapping, GFP_NOFS);
1235 BTRFS_I(inode)->root = root;
1236
1237 if (mode & S_IFDIR)
1238 owner = 0;
1239 else
1240 owner = 1;
1241 group = btrfs_find_block_group(root, group, 0, 0, owner);
1242 BTRFS_I(inode)->block_group = group;
1243
1244 ret = btrfs_insert_empty_inode(trans, root, path, objectid);
1245 if (ret)
1246 goto fail;
1247
1248 inode->i_uid = current->fsuid;
1249 inode->i_gid = current->fsgid;
1250 inode->i_mode = mode;
1251 inode->i_ino = objectid;
1252 inode->i_blocks = 0;
1253 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1254 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1255 struct btrfs_inode_item);
1256 fill_inode_item(path->nodes[0], inode_item, inode);
1257 btrfs_mark_buffer_dirty(path->nodes[0]);
1258 btrfs_free_path(path);
1259
1260 location = &BTRFS_I(inode)->location;
1261 location->objectid = objectid;
1262 location->offset = 0;
1263 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1264
1265 insert_inode_hash(inode);
1266 return inode;
1267 fail:
1268 btrfs_free_path(path);
1269 return ERR_PTR(ret);
1270 }
1271
1272 static inline u8 btrfs_inode_type(struct inode *inode)
1273 {
1274 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1275 }
1276
1277 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1278 struct dentry *dentry, struct inode *inode)
1279 {
1280 int ret;
1281 struct btrfs_key key;
1282 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1283 struct inode *parent_inode;
1284
1285 key.objectid = inode->i_ino;
1286 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1287 key.offset = 0;
1288
1289 ret = btrfs_insert_dir_item(trans, root,
1290 dentry->d_name.name, dentry->d_name.len,
1291 dentry->d_parent->d_inode->i_ino,
1292 &key, btrfs_inode_type(inode));
1293 if (ret == 0) {
1294 ret = btrfs_insert_inode_ref(trans, root,
1295 dentry->d_name.name,
1296 dentry->d_name.len,
1297 inode->i_ino,
1298 dentry->d_parent->d_inode->i_ino);
1299 parent_inode = dentry->d_parent->d_inode;
1300 parent_inode->i_size += dentry->d_name.len * 2;
1301 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1302 ret = btrfs_update_inode(trans, root,
1303 dentry->d_parent->d_inode);
1304 }
1305 return ret;
1306 }
1307
1308 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1309 struct dentry *dentry, struct inode *inode)
1310 {
1311 int err = btrfs_add_link(trans, dentry, inode);
1312 if (!err) {
1313 d_instantiate(dentry, inode);
1314 return 0;
1315 }
1316 if (err > 0)
1317 err = -EEXIST;
1318 return err;
1319 }
1320
1321 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1322 int mode, dev_t rdev)
1323 {
1324 struct btrfs_trans_handle *trans;
1325 struct btrfs_root *root = BTRFS_I(dir)->root;
1326 struct inode *inode;
1327 int err;
1328 int drop_inode = 0;
1329 u64 objectid;
1330 unsigned long nr;
1331
1332 if (!new_valid_dev(rdev))
1333 return -EINVAL;
1334
1335 mutex_lock(&root->fs_info->fs_mutex);
1336 trans = btrfs_start_transaction(root, 1);
1337 btrfs_set_trans_block_group(trans, dir);
1338
1339 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1340 if (err) {
1341 err = -ENOSPC;
1342 goto out_unlock;
1343 }
1344
1345 inode = btrfs_new_inode(trans, root, objectid,
1346 BTRFS_I(dir)->block_group, mode);
1347 err = PTR_ERR(inode);
1348 if (IS_ERR(inode))
1349 goto out_unlock;
1350
1351 btrfs_set_trans_block_group(trans, inode);
1352 err = btrfs_add_nondir(trans, dentry, inode);
1353 if (err)
1354 drop_inode = 1;
1355 else {
1356 inode->i_op = &btrfs_special_inode_operations;
1357 init_special_inode(inode, inode->i_mode, rdev);
1358 btrfs_update_inode(trans, root, inode);
1359 }
1360 dir->i_sb->s_dirt = 1;
1361 btrfs_update_inode_block_group(trans, inode);
1362 btrfs_update_inode_block_group(trans, dir);
1363 out_unlock:
1364 nr = trans->blocks_used;
1365 btrfs_end_transaction(trans, root);
1366 mutex_unlock(&root->fs_info->fs_mutex);
1367
1368 if (drop_inode) {
1369 inode_dec_link_count(inode);
1370 iput(inode);
1371 }
1372 btrfs_btree_balance_dirty(root, nr);
1373 return err;
1374 }
1375
1376 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1377 int mode, struct nameidata *nd)
1378 {
1379 struct btrfs_trans_handle *trans;
1380 struct btrfs_root *root = BTRFS_I(dir)->root;
1381 struct inode *inode;
1382 int err;
1383 int drop_inode = 0;
1384 unsigned long nr;
1385 u64 objectid;
1386
1387 mutex_lock(&root->fs_info->fs_mutex);
1388 trans = btrfs_start_transaction(root, 1);
1389 btrfs_set_trans_block_group(trans, dir);
1390
1391 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1392 if (err) {
1393 err = -ENOSPC;
1394 goto out_unlock;
1395 }
1396
1397 inode = btrfs_new_inode(trans, root, objectid,
1398 BTRFS_I(dir)->block_group, mode);
1399 err = PTR_ERR(inode);
1400 if (IS_ERR(inode))
1401 goto out_unlock;
1402
1403 btrfs_set_trans_block_group(trans, inode);
1404 err = btrfs_add_nondir(trans, dentry, inode);
1405 if (err)
1406 drop_inode = 1;
1407 else {
1408 inode->i_mapping->a_ops = &btrfs_aops;
1409 inode->i_fop = &btrfs_file_operations;
1410 inode->i_op = &btrfs_file_inode_operations;
1411 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1412 inode->i_mapping, GFP_NOFS);
1413 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
1414 }
1415 dir->i_sb->s_dirt = 1;
1416 btrfs_update_inode_block_group(trans, inode);
1417 btrfs_update_inode_block_group(trans, dir);
1418 out_unlock:
1419 nr = trans->blocks_used;
1420 btrfs_end_transaction(trans, root);
1421 mutex_unlock(&root->fs_info->fs_mutex);
1422
1423 if (drop_inode) {
1424 inode_dec_link_count(inode);
1425 iput(inode);
1426 }
1427 btrfs_btree_balance_dirty(root, nr);
1428 return err;
1429 }
1430
1431 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1432 struct dentry *dentry)
1433 {
1434 struct btrfs_trans_handle *trans;
1435 struct btrfs_root *root = BTRFS_I(dir)->root;
1436 struct inode *inode = old_dentry->d_inode;
1437 unsigned long nr;
1438 int err;
1439 int drop_inode = 0;
1440
1441 if (inode->i_nlink == 0)
1442 return -ENOENT;
1443
1444 inc_nlink(inode);
1445 mutex_lock(&root->fs_info->fs_mutex);
1446 trans = btrfs_start_transaction(root, 1);
1447
1448 btrfs_set_trans_block_group(trans, dir);
1449 atomic_inc(&inode->i_count);
1450 err = btrfs_add_nondir(trans, dentry, inode);
1451
1452 if (err)
1453 drop_inode = 1;
1454
1455 dir->i_sb->s_dirt = 1;
1456 btrfs_update_inode_block_group(trans, dir);
1457 err = btrfs_update_inode(trans, root, inode);
1458
1459 if (err)
1460 drop_inode = 1;
1461
1462 nr = trans->blocks_used;
1463 btrfs_end_transaction(trans, root);
1464 mutex_unlock(&root->fs_info->fs_mutex);
1465
1466 if (drop_inode) {
1467 inode_dec_link_count(inode);
1468 iput(inode);
1469 }
1470 btrfs_btree_balance_dirty(root, nr);
1471 return err;
1472 }
1473
1474 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1475 {
1476 struct inode *inode;
1477 struct btrfs_trans_handle *trans;
1478 struct btrfs_root *root = BTRFS_I(dir)->root;
1479 int err = 0;
1480 int drop_on_err = 0;
1481 u64 objectid;
1482 unsigned long nr = 1;
1483
1484 mutex_lock(&root->fs_info->fs_mutex);
1485 trans = btrfs_start_transaction(root, 1);
1486 btrfs_set_trans_block_group(trans, dir);
1487
1488 if (IS_ERR(trans)) {
1489 err = PTR_ERR(trans);
1490 goto out_unlock;
1491 }
1492
1493 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1494 if (err) {
1495 err = -ENOSPC;
1496 goto out_unlock;
1497 }
1498
1499 inode = btrfs_new_inode(trans, root, objectid,
1500 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1501 if (IS_ERR(inode)) {
1502 err = PTR_ERR(inode);
1503 goto out_fail;
1504 }
1505
1506 drop_on_err = 1;
1507 inode->i_op = &btrfs_dir_inode_operations;
1508 inode->i_fop = &btrfs_dir_file_operations;
1509 btrfs_set_trans_block_group(trans, inode);
1510
1511 inode->i_size = 0;
1512 err = btrfs_update_inode(trans, root, inode);
1513 if (err)
1514 goto out_fail;
1515
1516 err = btrfs_add_link(trans, dentry, inode);
1517 if (err)
1518 goto out_fail;
1519
1520 d_instantiate(dentry, inode);
1521 drop_on_err = 0;
1522 dir->i_sb->s_dirt = 1;
1523 btrfs_update_inode_block_group(trans, inode);
1524 btrfs_update_inode_block_group(trans, dir);
1525
1526 out_fail:
1527 nr = trans->blocks_used;
1528 btrfs_end_transaction(trans, root);
1529
1530 out_unlock:
1531 mutex_unlock(&root->fs_info->fs_mutex);
1532 if (drop_on_err)
1533 iput(inode);
1534 btrfs_btree_balance_dirty(root, nr);
1535 return err;
1536 }
1537
1538 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
1539 size_t page_offset, u64 start, u64 end,
1540 int create)
1541 {
1542 int ret;
1543 int err = 0;
1544 u64 bytenr;
1545 u64 extent_start = 0;
1546 u64 extent_end = 0;
1547 u64 objectid = inode->i_ino;
1548 u32 found_type;
1549 int failed_insert = 0;
1550 struct btrfs_path *path;
1551 struct btrfs_root *root = BTRFS_I(inode)->root;
1552 struct btrfs_file_extent_item *item;
1553 struct extent_buffer *leaf;
1554 struct btrfs_key found_key;
1555 struct extent_map *em = NULL;
1556 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1557 struct btrfs_trans_handle *trans = NULL;
1558
1559 path = btrfs_alloc_path();
1560 BUG_ON(!path);
1561 mutex_lock(&root->fs_info->fs_mutex);
1562
1563 again:
1564 em = lookup_extent_mapping(em_tree, start, end);
1565 if (em) {
1566 goto out;
1567 }
1568 if (!em) {
1569 em = alloc_extent_map(GFP_NOFS);
1570 if (!em) {
1571 err = -ENOMEM;
1572 goto out;
1573 }
1574 em->start = EXTENT_MAP_HOLE;
1575 em->end = EXTENT_MAP_HOLE;
1576 }
1577 em->bdev = inode->i_sb->s_bdev;
1578 ret = btrfs_lookup_file_extent(trans, root, path,
1579 objectid, start, trans != NULL);
1580 if (ret < 0) {
1581 err = ret;
1582 goto out;
1583 }
1584
1585 if (ret != 0) {
1586 if (path->slots[0] == 0)
1587 goto not_found;
1588 path->slots[0]--;
1589 }
1590
1591 leaf = path->nodes[0];
1592 item = btrfs_item_ptr(leaf, path->slots[0],
1593 struct btrfs_file_extent_item);
1594 /* are we inside the extent that was found? */
1595 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1596 found_type = btrfs_key_type(&found_key);
1597 if (found_key.objectid != objectid ||
1598 found_type != BTRFS_EXTENT_DATA_KEY) {
1599 goto not_found;
1600 }
1601
1602 found_type = btrfs_file_extent_type(leaf, item);
1603 extent_start = found_key.offset;
1604 if (found_type == BTRFS_FILE_EXTENT_REG) {
1605 extent_end = extent_start +
1606 btrfs_file_extent_num_bytes(leaf, item);
1607 err = 0;
1608 if (start < extent_start || start >= extent_end) {
1609 em->start = start;
1610 if (start < extent_start) {
1611 if (end < extent_start)
1612 goto not_found;
1613 em->end = extent_end - 1;
1614 } else {
1615 em->end = end;
1616 }
1617 goto not_found_em;
1618 }
1619 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
1620 if (bytenr == 0) {
1621 em->start = extent_start;
1622 em->end = extent_end - 1;
1623 em->block_start = EXTENT_MAP_HOLE;
1624 em->block_end = EXTENT_MAP_HOLE;
1625 goto insert;
1626 }
1627 bytenr += btrfs_file_extent_offset(leaf, item);
1628 em->block_start = bytenr;
1629 em->block_end = em->block_start +
1630 btrfs_file_extent_num_bytes(leaf, item) - 1;
1631 em->start = extent_start;
1632 em->end = extent_end - 1;
1633 goto insert;
1634 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1635 unsigned long ptr;
1636 char *map;
1637 size_t size;
1638 size_t extent_offset;
1639 size_t copy_size;
1640
1641 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
1642 path->slots[0]));
1643 extent_end = (extent_start + size - 1) |
1644 ((u64)root->sectorsize - 1);
1645 if (start < extent_start || start >= extent_end) {
1646 em->start = start;
1647 if (start < extent_start) {
1648 if (end < extent_start)
1649 goto not_found;
1650 em->end = extent_end;
1651 } else {
1652 em->end = end;
1653 }
1654 goto not_found_em;
1655 }
1656 em->block_start = EXTENT_MAP_INLINE;
1657 em->block_end = EXTENT_MAP_INLINE;
1658
1659 if (!page) {
1660 em->start = extent_start;
1661 em->end = extent_start + size - 1;
1662 goto out;
1663 }
1664
1665 extent_offset = ((u64)page->index << PAGE_CACHE_SHIFT) -
1666 extent_start + page_offset;
1667 copy_size = min_t(u64, PAGE_CACHE_SIZE - page_offset,
1668 size - extent_offset);
1669 em->start = extent_start + extent_offset;
1670 em->end = (em->start + copy_size -1) |
1671 ((u64)root->sectorsize -1);
1672 map = kmap(page);
1673 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
1674 if (create == 0 && !PageUptodate(page)) {
1675 read_extent_buffer(leaf, map + page_offset, ptr,
1676 copy_size);
1677 flush_dcache_page(page);
1678 } else if (create && PageUptodate(page)) {
1679 if (!trans) {
1680 kunmap(page);
1681 free_extent_map(em);
1682 em = NULL;
1683 btrfs_release_path(root, path);
1684 trans = btrfs_start_transaction(root, 1);
1685 goto again;
1686 }
1687 write_extent_buffer(leaf, map + page_offset, ptr,
1688 copy_size);
1689 btrfs_mark_buffer_dirty(leaf);
1690 }
1691 kunmap(page);
1692 set_extent_uptodate(em_tree, em->start, em->end, GFP_NOFS);
1693 goto insert;
1694 } else {
1695 printk("unkknown found_type %d\n", found_type);
1696 WARN_ON(1);
1697 }
1698 not_found:
1699 em->start = start;
1700 em->end = end;
1701 not_found_em:
1702 em->block_start = EXTENT_MAP_HOLE;
1703 em->block_end = EXTENT_MAP_HOLE;
1704 insert:
1705 btrfs_release_path(root, path);
1706 if (em->start > start || em->end < start) {
1707 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->end, start, end);
1708 err = -EIO;
1709 goto out;
1710 }
1711 ret = add_extent_mapping(em_tree, em);
1712 if (ret == -EEXIST) {
1713 free_extent_map(em);
1714 em = NULL;
1715 failed_insert++;
1716 if (failed_insert > 5) {
1717 printk("failing to insert %Lu %Lu\n", start, end);
1718 err = -EIO;
1719 goto out;
1720 }
1721 goto again;
1722 }
1723 err = 0;
1724 out:
1725 btrfs_free_path(path);
1726 if (trans) {
1727 ret = btrfs_end_transaction(trans, root);
1728 if (!err)
1729 err = ret;
1730 }
1731 mutex_unlock(&root->fs_info->fs_mutex);
1732 if (err) {
1733 free_extent_map(em);
1734 WARN_ON(1);
1735 return ERR_PTR(err);
1736 }
1737 return em;
1738 }
1739
1740 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
1741 {
1742 return extent_bmap(mapping, iblock, btrfs_get_extent);
1743 }
1744
1745 static int btrfs_prepare_write(struct file *file, struct page *page,
1746 unsigned from, unsigned to)
1747 {
1748 return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree,
1749 page->mapping->host, page, from, to,
1750 btrfs_get_extent);
1751 }
1752
1753 int btrfs_readpage(struct file *file, struct page *page)
1754 {
1755 struct extent_map_tree *tree;
1756 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1757 return extent_read_full_page(tree, page, btrfs_get_extent);
1758 }
1759 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
1760 {
1761 struct extent_map_tree *tree;
1762
1763
1764 if (current->flags & PF_MEMALLOC) {
1765 redirty_page_for_writepage(wbc, page);
1766 unlock_page(page);
1767 return 0;
1768 }
1769 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1770 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
1771 }
1772
1773 static int btrfs_writepages(struct address_space *mapping,
1774 struct writeback_control *wbc)
1775 {
1776 struct extent_map_tree *tree;
1777 tree = &BTRFS_I(mapping->host)->extent_tree;
1778 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
1779 }
1780
1781 static int
1782 btrfs_readpages(struct file *file, struct address_space *mapping,
1783 struct list_head *pages, unsigned nr_pages)
1784 {
1785 struct extent_map_tree *tree;
1786 tree = &BTRFS_I(mapping->host)->extent_tree;
1787 return extent_readpages(tree, mapping, pages, nr_pages,
1788 btrfs_get_extent);
1789 }
1790
1791 static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
1792 {
1793 struct extent_map_tree *tree;
1794 int ret;
1795
1796 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1797 ret = try_release_extent_mapping(tree, page);
1798 if (ret == 1) {
1799 ClearPagePrivate(page);
1800 set_page_private(page, 0);
1801 page_cache_release(page);
1802 }
1803 return ret;
1804 }
1805
1806 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
1807 {
1808 struct extent_map_tree *tree;
1809
1810 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1811 extent_invalidatepage(tree, page, offset);
1812 btrfs_releasepage(page, GFP_NOFS);
1813 }
1814
1815 /*
1816 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
1817 * called from a page fault handler when a page is first dirtied. Hence we must
1818 * be careful to check for EOF conditions here. We set the page up correctly
1819 * for a written page which means we get ENOSPC checking when writing into
1820 * holes and correct delalloc and unwritten extent mapping on filesystems that
1821 * support these features.
1822 *
1823 * We are not allowed to take the i_mutex here so we have to play games to
1824 * protect against truncate races as the page could now be beyond EOF. Because
1825 * vmtruncate() writes the inode size before removing pages, once we have the
1826 * page lock we can determine safely if the page is beyond EOF. If it is not
1827 * beyond EOF, then the page is guaranteed safe against truncation until we
1828 * unlock the page.
1829 */
1830 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1831 {
1832 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
1833 unsigned long end;
1834 loff_t size;
1835 int ret = -EINVAL;
1836 u64 page_start;
1837
1838 down_read(&BTRFS_I(inode)->root->snap_sem);
1839 lock_page(page);
1840 wait_on_page_writeback(page);
1841 size = i_size_read(inode);
1842 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
1843
1844 if ((page->mapping != inode->i_mapping) ||
1845 (page_start > size)) {
1846 /* page got truncated out from underneath us */
1847 goto out_unlock;
1848 }
1849
1850 /* page is wholly or partially inside EOF */
1851 if (page_start + PAGE_CACHE_SIZE > size)
1852 end = size & ~PAGE_CACHE_MASK;
1853 else
1854 end = PAGE_CACHE_SIZE;
1855
1856 ret = btrfs_cow_one_page(inode, page, end);
1857
1858 out_unlock:
1859 up_read(&BTRFS_I(inode)->root->snap_sem);
1860 unlock_page(page);
1861 return ret;
1862 }
1863
1864 static void btrfs_truncate(struct inode *inode)
1865 {
1866 struct btrfs_root *root = BTRFS_I(inode)->root;
1867 int ret;
1868 struct btrfs_trans_handle *trans;
1869 unsigned long nr;
1870
1871 if (!S_ISREG(inode->i_mode))
1872 return;
1873 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1874 return;
1875
1876 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1877
1878 mutex_lock(&root->fs_info->fs_mutex);
1879 trans = btrfs_start_transaction(root, 1);
1880 btrfs_set_trans_block_group(trans, inode);
1881
1882 /* FIXME, add redo link to tree so we don't leak on crash */
1883 ret = btrfs_truncate_in_trans(trans, root, inode);
1884 btrfs_update_inode(trans, root, inode);
1885 nr = trans->blocks_used;
1886
1887 ret = btrfs_end_transaction(trans, root);
1888 BUG_ON(ret);
1889 mutex_unlock(&root->fs_info->fs_mutex);
1890 btrfs_btree_balance_dirty(root, nr);
1891 }
1892
1893 int btrfs_commit_write(struct file *file, struct page *page,
1894 unsigned from, unsigned to)
1895 {
1896 return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree,
1897 page->mapping->host, page, from, to);
1898 }
1899
1900 static int create_subvol(struct btrfs_root *root, char *name, int namelen)
1901 {
1902 struct btrfs_trans_handle *trans;
1903 struct btrfs_key key;
1904 struct btrfs_root_item root_item;
1905 struct btrfs_inode_item *inode_item;
1906 struct extent_buffer *leaf;
1907 struct btrfs_root *new_root;
1908 struct inode *inode;
1909 struct inode *dir;
1910 int ret;
1911 int err;
1912 u64 objectid;
1913 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
1914 unsigned long nr = 1;
1915
1916 mutex_lock(&root->fs_info->fs_mutex);
1917 trans = btrfs_start_transaction(root, 1);
1918 BUG_ON(!trans);
1919
1920 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1921 0, &objectid);
1922 if (ret)
1923 goto fail;
1924
1925 leaf = __btrfs_alloc_free_block(trans, root, root->leafsize,
1926 objectid, trans->transid, 0, 0,
1927 0, 0);
1928 if (IS_ERR(leaf))
1929 return PTR_ERR(leaf);
1930
1931 btrfs_set_header_nritems(leaf, 0);
1932 btrfs_set_header_level(leaf, 0);
1933 btrfs_set_header_bytenr(leaf, leaf->start);
1934 btrfs_set_header_generation(leaf, trans->transid);
1935 btrfs_set_header_owner(leaf, objectid);
1936
1937 write_extent_buffer(leaf, root->fs_info->fsid,
1938 (unsigned long)btrfs_header_fsid(leaf),
1939 BTRFS_FSID_SIZE);
1940 btrfs_mark_buffer_dirty(leaf);
1941
1942 inode_item = &root_item.inode;
1943 memset(inode_item, 0, sizeof(*inode_item));
1944 inode_item->generation = cpu_to_le64(1);
1945 inode_item->size = cpu_to_le64(3);
1946 inode_item->nlink = cpu_to_le32(1);
1947 inode_item->nblocks = cpu_to_le64(1);
1948 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1949
1950 btrfs_set_root_bytenr(&root_item, leaf->start);
1951 btrfs_set_root_level(&root_item, 0);
1952 btrfs_set_root_refs(&root_item, 1);
1953 btrfs_set_root_used(&root_item, 0);
1954
1955 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
1956 root_item.drop_level = 0;
1957
1958 free_extent_buffer(leaf);
1959 leaf = NULL;
1960
1961 btrfs_set_root_dirid(&root_item, new_dirid);
1962
1963 key.objectid = objectid;
1964 key.offset = 1;
1965 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1966 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1967 &root_item);
1968 if (ret)
1969 goto fail;
1970
1971 /*
1972 * insert the directory item
1973 */
1974 key.offset = (u64)-1;
1975 dir = root->fs_info->sb->s_root->d_inode;
1976 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1977 name, namelen, dir->i_ino, &key,
1978 BTRFS_FT_DIR);
1979 if (ret)
1980 goto fail;
1981
1982 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
1983 name, namelen, objectid,
1984 root->fs_info->sb->s_root->d_inode->i_ino);
1985 if (ret)
1986 goto fail;
1987
1988 ret = btrfs_commit_transaction(trans, root);
1989 if (ret)
1990 goto fail_commit;
1991
1992 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
1993 BUG_ON(!new_root);
1994
1995 trans = btrfs_start_transaction(new_root, 1);
1996 BUG_ON(!trans);
1997
1998 inode = btrfs_new_inode(trans, new_root, new_dirid,
1999 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
2000 if (IS_ERR(inode))
2001 goto fail;
2002 inode->i_op = &btrfs_dir_inode_operations;
2003 inode->i_fop = &btrfs_dir_file_operations;
2004 new_root->inode = inode;
2005
2006 ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid,
2007 new_dirid);
2008 inode->i_nlink = 1;
2009 inode->i_size = 0;
2010 ret = btrfs_update_inode(trans, new_root, inode);
2011 if (ret)
2012 goto fail;
2013 fail:
2014 nr = trans->blocks_used;
2015 err = btrfs_commit_transaction(trans, root);
2016 if (err && !ret)
2017 ret = err;
2018 fail_commit:
2019 mutex_unlock(&root->fs_info->fs_mutex);
2020 btrfs_btree_balance_dirty(root, nr);
2021 return ret;
2022 }
2023
2024 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
2025 {
2026 struct btrfs_trans_handle *trans;
2027 struct btrfs_key key;
2028 struct btrfs_root_item new_root_item;
2029 struct extent_buffer *tmp;
2030 int ret;
2031 int err;
2032 u64 objectid;
2033 unsigned long nr;
2034
2035 if (!root->ref_cows)
2036 return -EINVAL;
2037
2038 down_write(&root->snap_sem);
2039 freeze_bdev(root->fs_info->sb->s_bdev);
2040 thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb);
2041
2042 mutex_lock(&root->fs_info->fs_mutex);
2043 trans = btrfs_start_transaction(root, 1);
2044 BUG_ON(!trans);
2045
2046 ret = btrfs_update_inode(trans, root, root->inode);
2047 if (ret)
2048 goto fail;
2049
2050 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
2051 0, &objectid);
2052 if (ret)
2053 goto fail;
2054
2055 memcpy(&new_root_item, &root->root_item,
2056 sizeof(new_root_item));
2057
2058 key.objectid = objectid;
2059 key.offset = 1;
2060 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
2061 extent_buffer_get(root->node);
2062 btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
2063 free_extent_buffer(tmp);
2064 btrfs_set_root_bytenr(&new_root_item, root->node->start);
2065 btrfs_set_root_level(&new_root_item, btrfs_header_level(root->node));
2066 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
2067 &new_root_item);
2068 if (ret)
2069 goto fail;
2070
2071 /*
2072 * insert the directory item
2073 */
2074 key.offset = (u64)-1;
2075 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
2076 name, namelen,
2077 root->fs_info->sb->s_root->d_inode->i_ino,
2078 &key, BTRFS_FT_DIR);
2079
2080 if (ret)
2081 goto fail;
2082
2083 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
2084 name, namelen, objectid,
2085 root->fs_info->sb->s_root->d_inode->i_ino);
2086
2087 if (ret)
2088 goto fail;
2089
2090 ret = btrfs_inc_root_ref(trans, root, objectid);
2091 if (ret)
2092 goto fail;
2093 fail:
2094 nr = trans->blocks_used;
2095 err = btrfs_commit_transaction(trans, root);
2096
2097 if (err && !ret)
2098 ret = err;
2099
2100 mutex_unlock(&root->fs_info->fs_mutex);
2101 up_write(&root->snap_sem);
2102 btrfs_btree_balance_dirty(root, nr);
2103 return ret;
2104 }
2105
2106 static unsigned long force_ra(struct address_space *mapping,
2107 struct file_ra_state *ra, struct file *file,
2108 pgoff_t offset, pgoff_t last_index)
2109 {
2110 pgoff_t req_size;
2111
2112 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2113 req_size = last_index - offset + 1;
2114 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2115 return offset;
2116 #else
2117 req_size = min(last_index - offset + 1, (pgoff_t)128);
2118 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2119 return offset + req_size;
2120 #endif
2121 }
2122
2123 int btrfs_defrag_file(struct file *file) {
2124 struct inode *inode = file->f_path.dentry->d_inode;
2125 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2126 struct page *page;
2127 unsigned long last_index;
2128 unsigned long ra_index = 0;
2129 u64 page_start;
2130 u64 page_end;
2131 unsigned long i;
2132
2133 mutex_lock(&inode->i_mutex);
2134 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
2135 for (i = 0; i <= last_index; i++) {
2136 if (i == ra_index) {
2137 ra_index = force_ra(inode->i_mapping, &file->f_ra,
2138 file, ra_index, last_index);
2139 }
2140 page = grab_cache_page(inode->i_mapping, i);
2141 if (!page)
2142 goto out_unlock;
2143 if (!PageUptodate(page)) {
2144 btrfs_readpage(NULL, page);
2145 lock_page(page);
2146 if (!PageUptodate(page)) {
2147 unlock_page(page);
2148 page_cache_release(page);
2149 goto out_unlock;
2150 }
2151 }
2152 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2153 page_end = page_start + PAGE_CACHE_SIZE - 1;
2154
2155 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
2156 set_extent_delalloc(em_tree, page_start,
2157 page_end, GFP_NOFS);
2158 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
2159 set_page_dirty(page);
2160 unlock_page(page);
2161 page_cache_release(page);
2162 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
2163 }
2164
2165 out_unlock:
2166 mutex_unlock(&inode->i_mutex);
2167 return 0;
2168 }
2169
2170 static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg)
2171 {
2172 struct btrfs_ioctl_vol_args vol_args;
2173 struct btrfs_dir_item *di;
2174 struct btrfs_path *path;
2175 int namelen;
2176 u64 root_dirid;
2177
2178 if (copy_from_user(&vol_args, arg, sizeof(vol_args)))
2179 return -EFAULT;
2180
2181 namelen = strlen(vol_args.name);
2182 if (namelen > BTRFS_VOL_NAME_MAX)
2183 return -EINVAL;
2184 if (strchr(vol_args.name, '/'))
2185 return -EINVAL;
2186
2187 path = btrfs_alloc_path();
2188 if (!path)
2189 return -ENOMEM;
2190
2191 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2192 mutex_lock(&root->fs_info->fs_mutex);
2193 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2194 path, root_dirid,
2195 vol_args.name, namelen, 0);
2196 mutex_unlock(&root->fs_info->fs_mutex);
2197 btrfs_free_path(path);
2198 if (di && !IS_ERR(di))
2199 return -EEXIST;
2200 if (IS_ERR(di))
2201 return PTR_ERR(di);
2202
2203 if (root == root->fs_info->tree_root)
2204 return create_subvol(root, vol_args.name, namelen);
2205 return create_snapshot(root, vol_args.name, namelen);
2206 }
2207
2208 static int btrfs_ioctl_defrag(struct file *file)
2209 {
2210 struct inode *inode = file->f_path.dentry->d_inode;
2211 struct btrfs_root *root = BTRFS_I(inode)->root;
2212
2213 switch (inode->i_mode & S_IFMT) {
2214 case S_IFDIR:
2215 mutex_lock(&root->fs_info->fs_mutex);
2216 btrfs_defrag_root(root, 0);
2217 btrfs_defrag_root(root->fs_info->extent_root, 0);
2218 mutex_unlock(&root->fs_info->fs_mutex);
2219 break;
2220 case S_IFREG:
2221 btrfs_defrag_file(file);
2222 break;
2223 }
2224
2225 return 0;
2226 }
2227
2228 long btrfs_ioctl(struct file *file, unsigned int
2229 cmd, unsigned long arg)
2230 {
2231 struct btrfs_root *root = BTRFS_I(file->f_path.dentry->d_inode)->root;
2232
2233 switch (cmd) {
2234 case BTRFS_IOC_SNAP_CREATE:
2235 return btrfs_ioctl_snap_create(root, (void __user *)arg);
2236 case BTRFS_IOC_DEFRAG:
2237 return btrfs_ioctl_defrag(file);
2238 }
2239
2240 return -ENOTTY;
2241 }
2242
2243 /*
2244 * Called inside transaction, so use GFP_NOFS
2245 */
2246 struct inode *btrfs_alloc_inode(struct super_block *sb)
2247 {
2248 struct btrfs_inode *ei;
2249
2250 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2251 if (!ei)
2252 return NULL;
2253 ei->last_trans = 0;
2254 return &ei->vfs_inode;
2255 }
2256
2257 void btrfs_destroy_inode(struct inode *inode)
2258 {
2259 WARN_ON(!list_empty(&inode->i_dentry));
2260 WARN_ON(inode->i_data.nrpages);
2261
2262 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2263 }
2264
2265 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2266 static void init_once(struct kmem_cache * cachep, void *foo)
2267 #else
2268 static void init_once(void * foo, struct kmem_cache * cachep,
2269 unsigned long flags)
2270 #endif
2271 {
2272 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2273
2274 inode_init_once(&ei->vfs_inode);
2275 }
2276
2277 void btrfs_destroy_cachep(void)
2278 {
2279 if (btrfs_inode_cachep)
2280 kmem_cache_destroy(btrfs_inode_cachep);
2281 if (btrfs_trans_handle_cachep)
2282 kmem_cache_destroy(btrfs_trans_handle_cachep);
2283 if (btrfs_transaction_cachep)
2284 kmem_cache_destroy(btrfs_transaction_cachep);
2285 if (btrfs_bit_radix_cachep)
2286 kmem_cache_destroy(btrfs_bit_radix_cachep);
2287 if (btrfs_path_cachep)
2288 kmem_cache_destroy(btrfs_path_cachep);
2289 }
2290
2291 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2292 unsigned long extra_flags,
2293 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2294 void (*ctor)(struct kmem_cache *, void *)
2295 #else
2296 void (*ctor)(void *, struct kmem_cache *,
2297 unsigned long)
2298 #endif
2299 )
2300 {
2301 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2302 SLAB_MEM_SPREAD | extra_flags), ctor
2303 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2304 ,NULL
2305 #endif
2306 );
2307 }
2308
2309 int btrfs_init_cachep(void)
2310 {
2311 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2312 sizeof(struct btrfs_inode),
2313 0, init_once);
2314 if (!btrfs_inode_cachep)
2315 goto fail;
2316 btrfs_trans_handle_cachep =
2317 btrfs_cache_create("btrfs_trans_handle_cache",
2318 sizeof(struct btrfs_trans_handle),
2319 0, NULL);
2320 if (!btrfs_trans_handle_cachep)
2321 goto fail;
2322 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
2323 sizeof(struct btrfs_transaction),
2324 0, NULL);
2325 if (!btrfs_transaction_cachep)
2326 goto fail;
2327 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
2328 sizeof(struct btrfs_path),
2329 0, NULL);
2330 if (!btrfs_path_cachep)
2331 goto fail;
2332 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
2333 SLAB_DESTROY_BY_RCU, NULL);
2334 if (!btrfs_bit_radix_cachep)
2335 goto fail;
2336 return 0;
2337 fail:
2338 btrfs_destroy_cachep();
2339 return -ENOMEM;
2340 }
2341
2342 static int btrfs_getattr(struct vfsmount *mnt,
2343 struct dentry *dentry, struct kstat *stat)
2344 {
2345 struct inode *inode = dentry->d_inode;
2346 generic_fillattr(inode, stat);
2347 stat->blksize = 256 * 1024;
2348 return 0;
2349 }
2350
2351 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2352 struct inode * new_dir,struct dentry *new_dentry)
2353 {
2354 struct btrfs_trans_handle *trans;
2355 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2356 struct inode *new_inode = new_dentry->d_inode;
2357 struct inode *old_inode = old_dentry->d_inode;
2358 struct timespec ctime = CURRENT_TIME;
2359 struct btrfs_path *path;
2360 int ret;
2361
2362 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2363 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2364 return -ENOTEMPTY;
2365 }
2366
2367 mutex_lock(&root->fs_info->fs_mutex);
2368 trans = btrfs_start_transaction(root, 1);
2369
2370 btrfs_set_trans_block_group(trans, new_dir);
2371 path = btrfs_alloc_path();
2372 if (!path) {
2373 ret = -ENOMEM;
2374 goto out_fail;
2375 }
2376
2377 old_dentry->d_inode->i_nlink++;
2378 old_dir->i_ctime = old_dir->i_mtime = ctime;
2379 new_dir->i_ctime = new_dir->i_mtime = ctime;
2380 old_inode->i_ctime = ctime;
2381
2382 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2383 if (ret)
2384 goto out_fail;
2385
2386 if (new_inode) {
2387 new_inode->i_ctime = CURRENT_TIME;
2388 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2389 if (ret)
2390 goto out_fail;
2391 }
2392 ret = btrfs_add_link(trans, new_dentry, old_inode);
2393 if (ret)
2394 goto out_fail;
2395
2396 out_fail:
2397 btrfs_free_path(path);
2398 btrfs_end_transaction(trans, root);
2399 mutex_unlock(&root->fs_info->fs_mutex);
2400 return ret;
2401 }
2402
2403 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2404 const char *symname)
2405 {
2406 struct btrfs_trans_handle *trans;
2407 struct btrfs_root *root = BTRFS_I(dir)->root;
2408 struct btrfs_path *path;
2409 struct btrfs_key key;
2410 struct inode *inode;
2411 int err;
2412 int drop_inode = 0;
2413 u64 objectid;
2414 int name_len;
2415 int datasize;
2416 unsigned long ptr;
2417 struct btrfs_file_extent_item *ei;
2418 struct extent_buffer *leaf;
2419 unsigned long nr;
2420
2421 name_len = strlen(symname) + 1;
2422 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2423 return -ENAMETOOLONG;
2424 mutex_lock(&root->fs_info->fs_mutex);
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, objectid,
2435 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2436 err = PTR_ERR(inode);
2437 if (IS_ERR(inode))
2438 goto out_unlock;
2439
2440 btrfs_set_trans_block_group(trans, inode);
2441 err = btrfs_add_nondir(trans, dentry, inode);
2442 if (err)
2443 drop_inode = 1;
2444 else {
2445 inode->i_mapping->a_ops = &btrfs_aops;
2446 inode->i_fop = &btrfs_file_operations;
2447 inode->i_op = &btrfs_file_inode_operations;
2448 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
2449 inode->i_mapping, GFP_NOFS);
2450 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
2451 }
2452 dir->i_sb->s_dirt = 1;
2453 btrfs_update_inode_block_group(trans, inode);
2454 btrfs_update_inode_block_group(trans, dir);
2455 if (drop_inode)
2456 goto out_unlock;
2457
2458 path = btrfs_alloc_path();
2459 BUG_ON(!path);
2460 key.objectid = inode->i_ino;
2461 key.offset = 0;
2462 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2463 datasize = btrfs_file_extent_calc_inline_size(name_len);
2464 err = btrfs_insert_empty_item(trans, root, path, &key,
2465 datasize);
2466 if (err) {
2467 drop_inode = 1;
2468 goto out_unlock;
2469 }
2470 leaf = path->nodes[0];
2471 ei = btrfs_item_ptr(leaf, path->slots[0],
2472 struct btrfs_file_extent_item);
2473 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
2474 btrfs_set_file_extent_type(leaf, ei,
2475 BTRFS_FILE_EXTENT_INLINE);
2476 ptr = btrfs_file_extent_inline_start(ei);
2477 write_extent_buffer(leaf, symname, ptr, name_len);
2478 btrfs_mark_buffer_dirty(leaf);
2479 btrfs_free_path(path);
2480
2481 inode->i_op = &btrfs_symlink_inode_operations;
2482 inode->i_mapping->a_ops = &btrfs_symlink_aops;
2483 inode->i_size = name_len - 1;
2484 err = btrfs_update_inode(trans, root, inode);
2485 if (err)
2486 drop_inode = 1;
2487
2488 out_unlock:
2489 nr = trans->blocks_used;
2490 btrfs_end_transaction(trans, root);
2491 mutex_unlock(&root->fs_info->fs_mutex);
2492 if (drop_inode) {
2493 inode_dec_link_count(inode);
2494 iput(inode);
2495 }
2496 btrfs_btree_balance_dirty(root, nr);
2497 return err;
2498 }
2499
2500 static struct inode_operations btrfs_dir_inode_operations = {
2501 .lookup = btrfs_lookup,
2502 .create = btrfs_create,
2503 .unlink = btrfs_unlink,
2504 .link = btrfs_link,
2505 .mkdir = btrfs_mkdir,
2506 .rmdir = btrfs_rmdir,
2507 .rename = btrfs_rename,
2508 .symlink = btrfs_symlink,
2509 .setattr = btrfs_setattr,
2510 .mknod = btrfs_mknod,
2511 .setxattr = generic_setxattr,
2512 .getxattr = generic_getxattr,
2513 .listxattr = btrfs_listxattr,
2514 .removexattr = generic_removexattr,
2515 };
2516
2517 static struct inode_operations btrfs_dir_ro_inode_operations = {
2518 .lookup = btrfs_lookup,
2519 };
2520
2521 static struct file_operations btrfs_dir_file_operations = {
2522 .llseek = generic_file_llseek,
2523 .read = generic_read_dir,
2524 .readdir = btrfs_readdir,
2525 .unlocked_ioctl = btrfs_ioctl,
2526 #ifdef CONFIG_COMPAT
2527 .compat_ioctl = btrfs_ioctl,
2528 #endif
2529 };
2530
2531 static struct extent_map_ops btrfs_extent_map_ops = {
2532 .fill_delalloc = run_delalloc_range,
2533 .writepage_io_hook = btrfs_writepage_io_hook,
2534 .readpage_io_hook = btrfs_readpage_io_hook,
2535 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
2536 };
2537
2538 static struct address_space_operations btrfs_aops = {
2539 .readpage = btrfs_readpage,
2540 .writepage = btrfs_writepage,
2541 .writepages = btrfs_writepages,
2542 .readpages = btrfs_readpages,
2543 .sync_page = block_sync_page,
2544 .prepare_write = btrfs_prepare_write,
2545 .commit_write = btrfs_commit_write,
2546 .bmap = btrfs_bmap,
2547 .invalidatepage = btrfs_invalidatepage,
2548 .releasepage = btrfs_releasepage,
2549 .set_page_dirty = __set_page_dirty_nobuffers,
2550 };
2551
2552 static struct address_space_operations btrfs_symlink_aops = {
2553 .readpage = btrfs_readpage,
2554 .writepage = btrfs_writepage,
2555 .invalidatepage = btrfs_invalidatepage,
2556 .releasepage = btrfs_releasepage,
2557 };
2558
2559 static struct inode_operations btrfs_file_inode_operations = {
2560 .truncate = btrfs_truncate,
2561 .getattr = btrfs_getattr,
2562 .setattr = btrfs_setattr,
2563 .setxattr = generic_setxattr,
2564 .getxattr = generic_getxattr,
2565 .listxattr = btrfs_listxattr,
2566 .removexattr = generic_removexattr,
2567 };
2568
2569 static struct inode_operations btrfs_special_inode_operations = {
2570 .getattr = btrfs_getattr,
2571 .setattr = btrfs_setattr,
2572 };
2573
2574 static struct inode_operations btrfs_symlink_inode_operations = {
2575 .readlink = generic_readlink,
2576 .follow_link = page_follow_link_light,
2577 .put_link = page_put_link,
2578 };
Linux kernel - Deliverables
This page took 0.125422 seconds and 6 git commands to generate.