projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Btrfs: Add writepages support
[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 if (alloc_hint == EXTENT_MAP_INLINE)
93 goto out;
94
95 ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_bytes, 0,
96 alloc_hint, (u64)-1, &ins, 1);
97 if (ret) {
98 WARN_ON(1);
99 goto out;
100 }
101 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
102 start, ins.objectid, ins.offset,
103 ins.offset);
104 out:
105 btrfs_end_transaction(trans, root);
106 mutex_unlock(&root->fs_info->fs_mutex);
107 return ret;
108 }
109
110 int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end)
111 {
112 struct inode *inode = page->mapping->host;
113 struct btrfs_root *root = BTRFS_I(inode)->root;
114 struct btrfs_trans_handle *trans;
115 char *kaddr;
116 int ret;
117 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
118 size_t offset = start - page_start;
119
120 mutex_lock(&root->fs_info->fs_mutex);
121 trans = btrfs_start_transaction(root, 1);
122 btrfs_set_trans_block_group(trans, inode);
123 kaddr = kmap(page);
124 btrfs_csum_file_block(trans, root, inode, inode->i_ino,
125 start, kaddr + offset, end - start + 1);
126 kunmap(page);
127 ret = btrfs_end_transaction(trans, root);
128 BUG_ON(ret);
129 mutex_unlock(&root->fs_info->fs_mutex);
130 return ret;
131 }
132
133 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
134 {
135 int ret = 0;
136 struct inode *inode = page->mapping->host;
137 struct btrfs_root *root = BTRFS_I(inode)->root;
138 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
139 struct btrfs_csum_item *item;
140 struct btrfs_path *path = NULL;
141 u32 csum;
142
143 mutex_lock(&root->fs_info->fs_mutex);
144 path = btrfs_alloc_path();
145 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
146 if (IS_ERR(item)) {
147 ret = PTR_ERR(item);
148 /* a csum that isn't present is a preallocated region. */
149 if (ret == -ENOENT || ret == -EFBIG)
150 ret = 0;
151 csum = 0;
152 goto out;
153 }
154 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
155 BTRFS_CRC32_SIZE);
156 set_state_private(em_tree, start, csum);
157 out:
158 if (path)
159 btrfs_free_path(path);
160 mutex_unlock(&root->fs_info->fs_mutex);
161 return ret;
162 }
163
164 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
165 {
166 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
167 struct inode *inode = page->mapping->host;
168 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
169 char *kaddr;
170 u64 private;
171 int ret;
172 struct btrfs_root *root = BTRFS_I(inode)->root;
173 u32 csum = ~(u32)0;
174 unsigned long flags;
175
176 ret = get_state_private(em_tree, start, &private);
177 local_irq_save(flags);
178 kaddr = kmap_atomic(page, KM_IRQ0);
179 if (ret) {
180 goto zeroit;
181 }
182 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
183 btrfs_csum_final(csum, (char *)&csum);
184 if (csum != private) {
185 goto zeroit;
186 }
187 kunmap_atomic(kaddr, KM_IRQ0);
188 local_irq_restore(flags);
189 return 0;
190
191 zeroit:
192 printk("btrfs csum failed ino %lu off %llu\n",
193 page->mapping->host->i_ino, (unsigned long long)start);
194 memset(kaddr + offset, 1, end - start + 1);
195 flush_dcache_page(page);
196 kunmap_atomic(kaddr, KM_IRQ0);
197 local_irq_restore(flags);
198 return 0;
199 }
200
201 void btrfs_read_locked_inode(struct inode *inode)
202 {
203 struct btrfs_path *path;
204 struct extent_buffer *leaf;
205 struct btrfs_inode_item *inode_item;
206 struct btrfs_inode_timespec *tspec;
207 struct btrfs_root *root = BTRFS_I(inode)->root;
208 struct btrfs_key location;
209 u64 alloc_group_block;
210 u32 rdev;
211 int ret;
212
213 path = btrfs_alloc_path();
214 BUG_ON(!path);
215 mutex_lock(&root->fs_info->fs_mutex);
216
217 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
218 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
219 if (ret)
220 goto make_bad;
221
222 leaf = path->nodes[0];
223 inode_item = btrfs_item_ptr(leaf, path->slots[0],
224 struct btrfs_inode_item);
225
226 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
227 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
228 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
229 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
230 inode->i_size = btrfs_inode_size(leaf, inode_item);
231
232 tspec = btrfs_inode_atime(inode_item);
233 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
234 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
235
236 tspec = btrfs_inode_mtime(inode_item);
237 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
238 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
239
240 tspec = btrfs_inode_ctime(inode_item);
241 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
242 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
243
244 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
245 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
246 inode->i_rdev = 0;
247 rdev = btrfs_inode_rdev(leaf, inode_item);
248
249 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
250 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
251 alloc_group_block);
252
253 btrfs_free_path(path);
254 inode_item = NULL;
255
256 mutex_unlock(&root->fs_info->fs_mutex);
257
258 switch (inode->i_mode & S_IFMT) {
259 case S_IFREG:
260 inode->i_mapping->a_ops = &btrfs_aops;
261 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
262 inode->i_fop = &btrfs_file_operations;
263 inode->i_op = &btrfs_file_inode_operations;
264 break;
265 case S_IFDIR:
266 inode->i_fop = &btrfs_dir_file_operations;
267 if (root == root->fs_info->tree_root)
268 inode->i_op = &btrfs_dir_ro_inode_operations;
269 else
270 inode->i_op = &btrfs_dir_inode_operations;
271 break;
272 case S_IFLNK:
273 inode->i_op = &btrfs_symlink_inode_operations;
274 inode->i_mapping->a_ops = &btrfs_symlink_aops;
275 break;
276 default:
277 init_special_inode(inode, inode->i_mode, rdev);
278 break;
279 }
280 return;
281
282 make_bad:
283 btrfs_release_path(root, path);
284 btrfs_free_path(path);
285 mutex_unlock(&root->fs_info->fs_mutex);
286 make_bad_inode(inode);
287 }
288
289 static void fill_inode_item(struct extent_buffer *leaf,
290 struct btrfs_inode_item *item,
291 struct inode *inode)
292 {
293 btrfs_set_inode_uid(leaf, item, inode->i_uid);
294 btrfs_set_inode_gid(leaf, item, inode->i_gid);
295 btrfs_set_inode_size(leaf, item, inode->i_size);
296 btrfs_set_inode_mode(leaf, item, inode->i_mode);
297 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
298
299 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
300 inode->i_atime.tv_sec);
301 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
302 inode->i_atime.tv_nsec);
303
304 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
305 inode->i_mtime.tv_sec);
306 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
307 inode->i_mtime.tv_nsec);
308
309 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
310 inode->i_ctime.tv_sec);
311 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
312 inode->i_ctime.tv_nsec);
313
314 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
315 btrfs_set_inode_generation(leaf, item, inode->i_generation);
316 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
317 btrfs_set_inode_block_group(leaf, item,
318 BTRFS_I(inode)->block_group->key.objectid);
319 }
320
321 int btrfs_update_inode(struct btrfs_trans_handle *trans,
322 struct btrfs_root *root,
323 struct inode *inode)
324 {
325 struct btrfs_inode_item *inode_item;
326 struct btrfs_path *path;
327 struct extent_buffer *leaf;
328 int ret;
329
330 path = btrfs_alloc_path();
331 BUG_ON(!path);
332 ret = btrfs_lookup_inode(trans, root, path,
333 &BTRFS_I(inode)->location, 1);
334 if (ret) {
335 if (ret > 0)
336 ret = -ENOENT;
337 goto failed;
338 }
339
340 leaf = path->nodes[0];
341 inode_item = btrfs_item_ptr(leaf, path->slots[0],
342 struct btrfs_inode_item);
343
344 fill_inode_item(leaf, inode_item, inode);
345 btrfs_mark_buffer_dirty(leaf);
346 btrfs_set_inode_last_trans(trans, inode);
347 ret = 0;
348 failed:
349 btrfs_release_path(root, path);
350 btrfs_free_path(path);
351 return ret;
352 }
353
354
355 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
356 struct btrfs_root *root,
357 struct inode *dir,
358 struct dentry *dentry)
359 {
360 struct btrfs_path *path;
361 const char *name = dentry->d_name.name;
362 int name_len = dentry->d_name.len;
363 int ret = 0;
364 struct extent_buffer *leaf;
365 struct btrfs_dir_item *di;
366 struct btrfs_key key;
367
368 path = btrfs_alloc_path();
369 if (!path) {
370 ret = -ENOMEM;
371 goto err;
372 }
373
374 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
375 name, name_len, -1);
376 if (IS_ERR(di)) {
377 ret = PTR_ERR(di);
378 goto err;
379 }
380 if (!di) {
381 ret = -ENOENT;
382 goto err;
383 }
384 leaf = path->nodes[0];
385 btrfs_dir_item_key_to_cpu(leaf, di, &key);
386 ret = btrfs_delete_one_dir_name(trans, root, path, di);
387 if (ret)
388 goto err;
389 btrfs_release_path(root, path);
390
391 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
392 key.objectid, name, name_len, -1);
393 if (IS_ERR(di)) {
394 ret = PTR_ERR(di);
395 goto err;
396 }
397 if (!di) {
398 ret = -ENOENT;
399 goto err;
400 }
401 ret = btrfs_delete_one_dir_name(trans, root, path, di);
402
403 dentry->d_inode->i_ctime = dir->i_ctime;
404 err:
405 btrfs_free_path(path);
406 if (!ret) {
407 dir->i_size -= name_len * 2;
408 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
409 btrfs_update_inode(trans, root, dir);
410 drop_nlink(dentry->d_inode);
411 ret = btrfs_update_inode(trans, root, dentry->d_inode);
412 dir->i_sb->s_dirt = 1;
413 }
414 return ret;
415 }
416
417 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
418 {
419 struct btrfs_root *root;
420 struct btrfs_trans_handle *trans;
421 int ret;
422 unsigned long nr;
423
424 root = BTRFS_I(dir)->root;
425 mutex_lock(&root->fs_info->fs_mutex);
426 trans = btrfs_start_transaction(root, 1);
427
428 btrfs_set_trans_block_group(trans, dir);
429 ret = btrfs_unlink_trans(trans, root, dir, dentry);
430 nr = trans->blocks_used;
431
432 btrfs_end_transaction(trans, root);
433 mutex_unlock(&root->fs_info->fs_mutex);
434 btrfs_btree_balance_dirty(root, nr);
435
436 return ret;
437 }
438
439 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
440 {
441 struct inode *inode = dentry->d_inode;
442 int err;
443 int ret;
444 struct btrfs_root *root = BTRFS_I(dir)->root;
445 struct btrfs_path *path;
446 struct btrfs_key key;
447 struct btrfs_trans_handle *trans;
448 struct btrfs_key found_key;
449 int found_type;
450 struct extent_buffer *leaf;
451 char *goodnames = "..";
452 unsigned long nr;
453
454 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
455 return -ENOTEMPTY;
456
457 path = btrfs_alloc_path();
458 BUG_ON(!path);
459 mutex_lock(&root->fs_info->fs_mutex);
460 trans = btrfs_start_transaction(root, 1);
461
462 btrfs_set_trans_block_group(trans, dir);
463 key.objectid = inode->i_ino;
464 key.offset = (u64)-1;
465 key.type = (u8)-1;
466 while(1) {
467 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
468 if (ret < 0) {
469 err = ret;
470 goto out;
471 }
472 BUG_ON(ret == 0);
473 if (path->slots[0] == 0) {
474 err = -ENOENT;
475 goto out;
476 }
477 path->slots[0]--;
478 leaf = path->nodes[0];
479 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
480 found_type = btrfs_key_type(&found_key);
481 if (found_key.objectid != inode->i_ino) {
482 err = -ENOENT;
483 goto out;
484 }
485 if ((found_type != BTRFS_DIR_ITEM_KEY &&
486 found_type != BTRFS_DIR_INDEX_KEY) ||
487 (!btrfs_match_dir_item_name(root, path, goodnames, 2) &&
488 !btrfs_match_dir_item_name(root, path, goodnames, 1))) {
489 err = -ENOTEMPTY;
490 goto out;
491 }
492 ret = btrfs_del_item(trans, root, path);
493 BUG_ON(ret);
494
495 if (found_type == BTRFS_DIR_ITEM_KEY && found_key.offset == 1)
496 break;
497 btrfs_release_path(root, path);
498 }
499 ret = 0;
500 btrfs_release_path(root, path);
501
502 /* now the directory is empty */
503 err = btrfs_unlink_trans(trans, root, dir, dentry);
504 if (!err) {
505 inode->i_size = 0;
506 }
507 out:
508 btrfs_release_path(root, path);
509 btrfs_free_path(path);
510 nr = trans->blocks_used;
511 ret = btrfs_end_transaction(trans, root);
512 mutex_unlock(&root->fs_info->fs_mutex);
513 btrfs_btree_balance_dirty(root, nr);
514 if (ret && !err)
515 err = ret;
516 return err;
517 }
518
519 static int btrfs_free_inode(struct btrfs_trans_handle *trans,
520 struct btrfs_root *root,
521 struct inode *inode)
522 {
523 struct btrfs_path *path;
524 int ret;
525
526 clear_inode(inode);
527
528 path = btrfs_alloc_path();
529 BUG_ON(!path);
530 ret = btrfs_lookup_inode(trans, root, path,
531 &BTRFS_I(inode)->location, -1);
532 if (ret > 0)
533 ret = -ENOENT;
534 if (!ret)
535 ret = btrfs_del_item(trans, root, path);
536 btrfs_free_path(path);
537 return ret;
538 }
539
540 /*
541 * this can truncate away extent items, csum items and directory items.
542 * It starts at a high offset and removes keys until it can't find
543 * any higher than i_size.
544 *
545 * csum items that cross the new i_size are truncated to the new size
546 * as well.
547 */
548 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
549 struct btrfs_root *root,
550 struct inode *inode)
551 {
552 int ret;
553 struct btrfs_path *path;
554 struct btrfs_key key;
555 struct btrfs_key found_key;
556 u32 found_type;
557 struct extent_buffer *leaf;
558 struct btrfs_file_extent_item *fi;
559 u64 extent_start = 0;
560 u64 extent_num_bytes = 0;
561 u64 item_end = 0;
562 int found_extent;
563 int del_item;
564 int extent_type = -1;
565
566 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
567 path = btrfs_alloc_path();
568 path->reada = -1;
569 BUG_ON(!path);
570
571 /* FIXME, add redo link to tree so we don't leak on crash */
572 key.objectid = inode->i_ino;
573 key.offset = (u64)-1;
574 key.type = (u8)-1;
575
576 while(1) {
577 btrfs_init_path(path);
578 fi = NULL;
579 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
580 if (ret < 0) {
581 goto error;
582 }
583 if (ret > 0) {
584 BUG_ON(path->slots[0] == 0);
585 path->slots[0]--;
586 }
587 leaf = path->nodes[0];
588 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
589 found_type = btrfs_key_type(&found_key);
590
591 if (found_key.objectid != inode->i_ino)
592 break;
593
594 if (found_type != BTRFS_CSUM_ITEM_KEY &&
595 found_type != BTRFS_DIR_ITEM_KEY &&
596 found_type != BTRFS_DIR_INDEX_KEY &&
597 found_type != BTRFS_EXTENT_DATA_KEY)
598 break;
599
600 item_end = found_key.offset;
601 if (found_type == BTRFS_EXTENT_DATA_KEY) {
602 fi = btrfs_item_ptr(leaf, path->slots[0],
603 struct btrfs_file_extent_item);
604 extent_type = btrfs_file_extent_type(leaf, fi);
605 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
606 item_end +=
607 btrfs_file_extent_num_bytes(leaf, fi);
608 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
609 struct btrfs_item *item = btrfs_item_nr(leaf,
610 path->slots[0]);
611 item_end += btrfs_file_extent_inline_len(leaf,
612 item);
613 }
614 }
615 if (found_type == BTRFS_CSUM_ITEM_KEY) {
616 ret = btrfs_csum_truncate(trans, root, path,
617 inode->i_size);
618 BUG_ON(ret);
619 }
620 if (item_end <= inode->i_size) {
621 if (found_type == BTRFS_DIR_ITEM_KEY) {
622 found_type = BTRFS_INODE_ITEM_KEY;
623 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
624 found_type = BTRFS_CSUM_ITEM_KEY;
625 } else if (found_type) {
626 found_type--;
627 } else {
628 break;
629 }
630 btrfs_set_key_type(&key, found_type);
631 btrfs_release_path(root, path);
632 continue;
633 }
634 if (found_key.offset >= inode->i_size)
635 del_item = 1;
636 else
637 del_item = 0;
638 found_extent = 0;
639
640 /* FIXME, shrink the extent if the ref count is only 1 */
641 if (found_type != BTRFS_EXTENT_DATA_KEY)
642 goto delete;
643
644 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
645 u64 num_dec;
646 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
647 if (!del_item) {
648 u64 orig_num_bytes =
649 btrfs_file_extent_num_bytes(leaf, fi);
650 extent_num_bytes = inode->i_size -
651 found_key.offset + root->sectorsize - 1;
652 btrfs_set_file_extent_num_bytes(leaf, fi,
653 extent_num_bytes);
654 num_dec = (orig_num_bytes -
655 extent_num_bytes) >> 9;
656 if (extent_start != 0) {
657 inode->i_blocks -= num_dec;
658 }
659 btrfs_mark_buffer_dirty(leaf);
660 } else {
661 extent_num_bytes =
662 btrfs_file_extent_disk_num_bytes(leaf,
663 fi);
664 /* FIXME blocksize != 4096 */
665 num_dec = btrfs_file_extent_num_bytes(leaf,
666 fi) >> 9;
667 if (extent_start != 0) {
668 found_extent = 1;
669 inode->i_blocks -= num_dec;
670 }
671 }
672 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE &&
673 !del_item) {
674 u32 newsize = inode->i_size - found_key.offset;
675 newsize = btrfs_file_extent_calc_inline_size(newsize);
676 ret = btrfs_truncate_item(trans, root, path,
677 newsize, 1);
678 BUG_ON(ret);
679 }
680 delete:
681 if (del_item) {
682 ret = btrfs_del_item(trans, root, path);
683 if (ret)
684 goto error;
685 } else {
686 break;
687 }
688 btrfs_release_path(root, path);
689 if (found_extent) {
690 ret = btrfs_free_extent(trans, root, extent_start,
691 extent_num_bytes, 0);
692 BUG_ON(ret);
693 }
694 }
695 ret = 0;
696 error:
697 btrfs_release_path(root, path);
698 btrfs_free_path(path);
699 inode->i_sb->s_dirt = 1;
700 return ret;
701 }
702
703 static int btrfs_cow_one_page(struct inode *inode, struct page *page,
704 size_t zero_start)
705 {
706 char *kaddr;
707 int ret = 0;
708 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
709 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
710 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
711
712 set_page_extent_mapped(page);
713
714 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
715 set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start,
716 page_end, GFP_NOFS);
717 if (zero_start != PAGE_CACHE_SIZE) {
718 kaddr = kmap(page);
719 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
720 flush_dcache_page(page);
721 kunmap(page);
722 }
723 set_page_dirty(page);
724 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
725
726 return ret;
727 }
728
729 /*
730 * taken from block_truncate_page, but does cow as it zeros out
731 * any bytes left in the last page in the file.
732 */
733 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
734 {
735 struct inode *inode = mapping->host;
736 struct btrfs_root *root = BTRFS_I(inode)->root;
737 u32 blocksize = root->sectorsize;
738 pgoff_t index = from >> PAGE_CACHE_SHIFT;
739 unsigned offset = from & (PAGE_CACHE_SIZE-1);
740 struct page *page;
741 int ret = 0;
742 u64 page_start;
743
744 if ((offset & (blocksize - 1)) == 0)
745 goto out;
746
747 down_read(&root->snap_sem);
748 ret = -ENOMEM;
749 page = grab_cache_page(mapping, index);
750 if (!page)
751 goto out;
752 if (!PageUptodate(page)) {
753 ret = btrfs_readpage(NULL, page);
754 lock_page(page);
755 if (!PageUptodate(page)) {
756 ret = -EIO;
757 goto out;
758 }
759 }
760 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
761
762 ret = btrfs_cow_one_page(inode, page, offset);
763
764 unlock_page(page);
765 page_cache_release(page);
766 up_read(&BTRFS_I(inode)->root->snap_sem);
767 out:
768 return ret;
769 }
770
771 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
772 {
773 struct inode *inode = dentry->d_inode;
774 int err;
775
776 err = inode_change_ok(inode, attr);
777 if (err)
778 return err;
779
780 if (S_ISREG(inode->i_mode) &&
781 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
782 struct btrfs_trans_handle *trans;
783 struct btrfs_root *root = BTRFS_I(inode)->root;
784 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
785
786 u64 mask = root->sectorsize - 1;
787 u64 pos = (inode->i_size + mask) & ~mask;
788 u64 block_end = attr->ia_size | mask;
789 u64 hole_size;
790 u64 alloc_hint = 0;
791
792 if (attr->ia_size <= pos)
793 goto out;
794
795 btrfs_truncate_page(inode->i_mapping, inode->i_size);
796
797 lock_extent(em_tree, pos, block_end, GFP_NOFS);
798 hole_size = (attr->ia_size - pos + mask) & ~mask;
799
800 mutex_lock(&root->fs_info->fs_mutex);
801 trans = btrfs_start_transaction(root, 1);
802 btrfs_set_trans_block_group(trans, inode);
803 err = btrfs_drop_extents(trans, root, inode,
804 pos, pos + hole_size, pos,
805 &alloc_hint);
806
807 if (alloc_hint != EXTENT_MAP_INLINE) {
808 err = btrfs_insert_file_extent(trans, root,
809 inode->i_ino,
810 pos, 0, 0, hole_size);
811 }
812 btrfs_end_transaction(trans, root);
813 mutex_unlock(&root->fs_info->fs_mutex);
814 unlock_extent(em_tree, pos, block_end, GFP_NOFS);
815 if (err)
816 return err;
817 }
818 out:
819 err = inode_setattr(inode, attr);
820
821 return err;
822 }
823 void btrfs_delete_inode(struct inode *inode)
824 {
825 struct btrfs_trans_handle *trans;
826 struct btrfs_root *root = BTRFS_I(inode)->root;
827 unsigned long nr;
828 int ret;
829
830 truncate_inode_pages(&inode->i_data, 0);
831 if (is_bad_inode(inode)) {
832 goto no_delete;
833 }
834
835 inode->i_size = 0;
836 mutex_lock(&root->fs_info->fs_mutex);
837 trans = btrfs_start_transaction(root, 1);
838
839 btrfs_set_trans_block_group(trans, inode);
840 ret = btrfs_truncate_in_trans(trans, root, inode);
841 if (ret)
842 goto no_delete_lock;
843 ret = btrfs_free_inode(trans, root, inode);
844 if (ret)
845 goto no_delete_lock;
846 nr = trans->blocks_used;
847
848 btrfs_end_transaction(trans, root);
849 mutex_unlock(&root->fs_info->fs_mutex);
850 btrfs_btree_balance_dirty(root, nr);
851 return;
852
853 no_delete_lock:
854 nr = trans->blocks_used;
855 btrfs_end_transaction(trans, root);
856 mutex_unlock(&root->fs_info->fs_mutex);
857 btrfs_btree_balance_dirty(root, nr);
858 no_delete:
859 clear_inode(inode);
860 }
861
862 /*
863 * this returns the key found in the dir entry in the location pointer.
864 * If no dir entries were found, location->objectid is 0.
865 */
866 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
867 struct btrfs_key *location)
868 {
869 const char *name = dentry->d_name.name;
870 int namelen = dentry->d_name.len;
871 struct btrfs_dir_item *di;
872 struct btrfs_path *path;
873 struct btrfs_root *root = BTRFS_I(dir)->root;
874 int ret = 0;
875
876 path = btrfs_alloc_path();
877 BUG_ON(!path);
878 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
879 namelen, 0);
880 if (IS_ERR(di))
881 ret = PTR_ERR(di);
882 if (!di || IS_ERR(di)) {
883 location->objectid = 0;
884 goto out;
885 }
886 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
887 out:
888 btrfs_release_path(root, path);
889 btrfs_free_path(path);
890 return ret;
891 }
892
893 /*
894 * when we hit a tree root in a directory, the btrfs part of the inode
895 * needs to be changed to reflect the root directory of the tree root. This
896 * is kind of like crossing a mount point.
897 */
898 static int fixup_tree_root_location(struct btrfs_root *root,
899 struct btrfs_key *location,
900 struct btrfs_root **sub_root,
901 struct dentry *dentry)
902 {
903 struct btrfs_path *path;
904 struct btrfs_root_item *ri;
905
906 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
907 return 0;
908 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
909 return 0;
910
911 path = btrfs_alloc_path();
912 BUG_ON(!path);
913 mutex_lock(&root->fs_info->fs_mutex);
914
915 *sub_root = btrfs_read_fs_root(root->fs_info, location,
916 dentry->d_name.name,
917 dentry->d_name.len);
918 if (IS_ERR(*sub_root))
919 return PTR_ERR(*sub_root);
920
921 ri = &(*sub_root)->root_item;
922 location->objectid = btrfs_root_dirid(ri);
923 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
924 location->offset = 0;
925
926 btrfs_free_path(path);
927 mutex_unlock(&root->fs_info->fs_mutex);
928 return 0;
929 }
930
931 static int btrfs_init_locked_inode(struct inode *inode, void *p)
932 {
933 struct btrfs_iget_args *args = p;
934 inode->i_ino = args->ino;
935 BTRFS_I(inode)->root = args->root;
936 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
937 inode->i_mapping, GFP_NOFS);
938 return 0;
939 }
940
941 static int btrfs_find_actor(struct inode *inode, void *opaque)
942 {
943 struct btrfs_iget_args *args = opaque;
944 return (args->ino == inode->i_ino &&
945 args->root == BTRFS_I(inode)->root);
946 }
947
948 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
949 struct btrfs_root *root)
950 {
951 struct inode *inode;
952 struct btrfs_iget_args args;
953 args.ino = objectid;
954 args.root = root;
955
956 inode = iget5_locked(s, objectid, btrfs_find_actor,
957 btrfs_init_locked_inode,
958 (void *)&args);
959 return inode;
960 }
961
962 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
963 struct nameidata *nd)
964 {
965 struct inode * inode;
966 struct btrfs_inode *bi = BTRFS_I(dir);
967 struct btrfs_root *root = bi->root;
968 struct btrfs_root *sub_root = root;
969 struct btrfs_key location;
970 int ret;
971
972 if (dentry->d_name.len > BTRFS_NAME_LEN)
973 return ERR_PTR(-ENAMETOOLONG);
974
975 mutex_lock(&root->fs_info->fs_mutex);
976 ret = btrfs_inode_by_name(dir, dentry, &location);
977 mutex_unlock(&root->fs_info->fs_mutex);
978
979 if (ret < 0)
980 return ERR_PTR(ret);
981
982 inode = NULL;
983 if (location.objectid) {
984 ret = fixup_tree_root_location(root, &location, &sub_root,
985 dentry);
986 if (ret < 0)
987 return ERR_PTR(ret);
988 if (ret > 0)
989 return ERR_PTR(-ENOENT);
990 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
991 sub_root);
992 if (!inode)
993 return ERR_PTR(-EACCES);
994 if (inode->i_state & I_NEW) {
995 /* the inode and parent dir are two different roots */
996 if (sub_root != root) {
997 igrab(inode);
998 sub_root->inode = inode;
999 }
1000 BTRFS_I(inode)->root = sub_root;
1001 memcpy(&BTRFS_I(inode)->location, &location,
1002 sizeof(location));
1003 btrfs_read_locked_inode(inode);
1004 unlock_new_inode(inode);
1005 }
1006 }
1007 return d_splice_alias(inode, dentry);
1008 }
1009
1010 static unsigned char btrfs_filetype_table[] = {
1011 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1012 };
1013
1014 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1015 {
1016 struct inode *inode = filp->f_path.dentry->d_inode;
1017 struct btrfs_root *root = BTRFS_I(inode)->root;
1018 struct btrfs_item *item;
1019 struct btrfs_dir_item *di;
1020 struct btrfs_key key;
1021 struct btrfs_key found_key;
1022 struct btrfs_path *path;
1023 int ret;
1024 u32 nritems;
1025 struct extent_buffer *leaf;
1026 int slot;
1027 int advance;
1028 unsigned char d_type;
1029 int over = 0;
1030 u32 di_cur;
1031 u32 di_total;
1032 u32 di_len;
1033 int key_type = BTRFS_DIR_INDEX_KEY;
1034 char tmp_name[32];
1035 char *name_ptr;
1036 int name_len;
1037
1038 /* FIXME, use a real flag for deciding about the key type */
1039 if (root->fs_info->tree_root == root)
1040 key_type = BTRFS_DIR_ITEM_KEY;
1041
1042 mutex_lock(&root->fs_info->fs_mutex);
1043 key.objectid = inode->i_ino;
1044 btrfs_set_key_type(&key, key_type);
1045 key.offset = filp->f_pos;
1046
1047 path = btrfs_alloc_path();
1048 path->reada = 2;
1049 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1050 if (ret < 0)
1051 goto err;
1052 advance = 0;
1053 while(1) {
1054 leaf = path->nodes[0];
1055 nritems = btrfs_header_nritems(leaf);
1056 slot = path->slots[0];
1057 if (advance || slot >= nritems) {
1058 if (slot >= nritems -1) {
1059 ret = btrfs_next_leaf(root, path);
1060 if (ret)
1061 break;
1062 leaf = path->nodes[0];
1063 nritems = btrfs_header_nritems(leaf);
1064 slot = path->slots[0];
1065 } else {
1066 slot++;
1067 path->slots[0]++;
1068 }
1069 }
1070 advance = 1;
1071 item = btrfs_item_nr(leaf, slot);
1072 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1073
1074 if (found_key.objectid != key.objectid)
1075 break;
1076 if (btrfs_key_type(&found_key) != key_type)
1077 break;
1078 if (found_key.offset < filp->f_pos)
1079 continue;
1080
1081 filp->f_pos = found_key.offset;
1082 advance = 1;
1083 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1084 di_cur = 0;
1085 di_total = btrfs_item_size(leaf, item);
1086 while(di_cur < di_total) {
1087 struct btrfs_key location;
1088
1089 name_len = btrfs_dir_name_len(leaf, di);
1090 if (name_len < 32) {
1091 name_ptr = tmp_name;
1092 } else {
1093 name_ptr = kmalloc(name_len, GFP_NOFS);
1094 BUG_ON(!name_ptr);
1095 }
1096 read_extent_buffer(leaf, name_ptr,
1097 (unsigned long)(di + 1), name_len);
1098
1099 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1100 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1101
1102 over = filldir(dirent, name_ptr, name_len,
1103 found_key.offset,
1104 location.objectid,
1105 d_type);
1106
1107 if (name_ptr != tmp_name)
1108 kfree(name_ptr);
1109
1110 if (over)
1111 goto nopos;
1112 di_len = btrfs_dir_name_len(leaf, di) + sizeof(*di);
1113 di_cur += di_len;
1114 di = (struct btrfs_dir_item *)((char *)di + di_len);
1115 }
1116 }
1117 filp->f_pos++;
1118 nopos:
1119 ret = 0;
1120 err:
1121 btrfs_release_path(root, path);
1122 btrfs_free_path(path);
1123 mutex_unlock(&root->fs_info->fs_mutex);
1124 return ret;
1125 }
1126
1127 int btrfs_write_inode(struct inode *inode, int wait)
1128 {
1129 struct btrfs_root *root = BTRFS_I(inode)->root;
1130 struct btrfs_trans_handle *trans;
1131 int ret = 0;
1132
1133 if (wait) {
1134 mutex_lock(&root->fs_info->fs_mutex);
1135 trans = btrfs_start_transaction(root, 1);
1136 btrfs_set_trans_block_group(trans, inode);
1137 ret = btrfs_commit_transaction(trans, root);
1138 mutex_unlock(&root->fs_info->fs_mutex);
1139 }
1140 return ret;
1141 }
1142
1143 /*
1144 * This is somewhat expensive, updating the tree every time the
1145 * inode changes. But, it is most likely to find the inode in cache.
1146 * FIXME, needs more benchmarking...there are no reasons other than performance
1147 * to keep or drop this code.
1148 */
1149 void btrfs_dirty_inode(struct inode *inode)
1150 {
1151 struct btrfs_root *root = BTRFS_I(inode)->root;
1152 struct btrfs_trans_handle *trans;
1153
1154 mutex_lock(&root->fs_info->fs_mutex);
1155 trans = btrfs_start_transaction(root, 1);
1156 btrfs_set_trans_block_group(trans, inode);
1157 btrfs_update_inode(trans, root, inode);
1158 btrfs_end_transaction(trans, root);
1159 mutex_unlock(&root->fs_info->fs_mutex);
1160 }
1161
1162 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1163 struct btrfs_root *root,
1164 u64 objectid,
1165 struct btrfs_block_group_cache *group,
1166 int mode)
1167 {
1168 struct inode *inode;
1169 struct btrfs_inode_item *inode_item;
1170 struct btrfs_key *location;
1171 struct btrfs_path *path;
1172 int ret;
1173 int owner;
1174
1175 path = btrfs_alloc_path();
1176 BUG_ON(!path);
1177
1178 inode = new_inode(root->fs_info->sb);
1179 if (!inode)
1180 return ERR_PTR(-ENOMEM);
1181
1182 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1183 inode->i_mapping, GFP_NOFS);
1184 BTRFS_I(inode)->root = root;
1185
1186 if (mode & S_IFDIR)
1187 owner = 0;
1188 else
1189 owner = 1;
1190 group = btrfs_find_block_group(root, group, 0, 0, owner);
1191 BTRFS_I(inode)->block_group = group;
1192
1193 ret = btrfs_insert_empty_inode(trans, root, path, objectid);
1194 if (ret)
1195 goto fail;
1196
1197 inode->i_uid = current->fsuid;
1198 inode->i_gid = current->fsgid;
1199 inode->i_mode = mode;
1200 inode->i_ino = objectid;
1201 inode->i_blocks = 0;
1202 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1203 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1204 struct btrfs_inode_item);
1205 fill_inode_item(path->nodes[0], inode_item, inode);
1206 btrfs_mark_buffer_dirty(path->nodes[0]);
1207 btrfs_free_path(path);
1208
1209 location = &BTRFS_I(inode)->location;
1210 location->objectid = objectid;
1211 location->offset = 0;
1212 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1213
1214 insert_inode_hash(inode);
1215 return inode;
1216 fail:
1217 btrfs_free_path(path);
1218 return ERR_PTR(ret);
1219 }
1220
1221 static inline u8 btrfs_inode_type(struct inode *inode)
1222 {
1223 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1224 }
1225
1226 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1227 struct dentry *dentry, struct inode *inode)
1228 {
1229 int ret;
1230 struct btrfs_key key;
1231 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1232 struct inode *parent_inode;
1233
1234 key.objectid = inode->i_ino;
1235 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1236 key.offset = 0;
1237
1238 ret = btrfs_insert_dir_item(trans, root,
1239 dentry->d_name.name, dentry->d_name.len,
1240 dentry->d_parent->d_inode->i_ino,
1241 &key, btrfs_inode_type(inode));
1242 if (ret == 0) {
1243 parent_inode = dentry->d_parent->d_inode;
1244 parent_inode->i_size += dentry->d_name.len * 2;
1245 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1246 ret = btrfs_update_inode(trans, root,
1247 dentry->d_parent->d_inode);
1248 }
1249 return ret;
1250 }
1251
1252 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1253 struct dentry *dentry, struct inode *inode)
1254 {
1255 int err = btrfs_add_link(trans, dentry, inode);
1256 if (!err) {
1257 d_instantiate(dentry, inode);
1258 return 0;
1259 }
1260 if (err > 0)
1261 err = -EEXIST;
1262 return err;
1263 }
1264
1265 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1266 int mode, dev_t rdev)
1267 {
1268 struct btrfs_trans_handle *trans;
1269 struct btrfs_root *root = BTRFS_I(dir)->root;
1270 struct inode *inode;
1271 int err;
1272 int drop_inode = 0;
1273 u64 objectid;
1274 unsigned long nr;
1275
1276 if (!new_valid_dev(rdev))
1277 return -EINVAL;
1278
1279 mutex_lock(&root->fs_info->fs_mutex);
1280 trans = btrfs_start_transaction(root, 1);
1281 btrfs_set_trans_block_group(trans, dir);
1282
1283 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1284 if (err) {
1285 err = -ENOSPC;
1286 goto out_unlock;
1287 }
1288
1289 inode = btrfs_new_inode(trans, root, objectid,
1290 BTRFS_I(dir)->block_group, mode);
1291 err = PTR_ERR(inode);
1292 if (IS_ERR(inode))
1293 goto out_unlock;
1294
1295 btrfs_set_trans_block_group(trans, inode);
1296 err = btrfs_add_nondir(trans, dentry, inode);
1297 if (err)
1298 drop_inode = 1;
1299 else {
1300 inode->i_op = &btrfs_special_inode_operations;
1301 init_special_inode(inode, inode->i_mode, rdev);
1302 btrfs_update_inode(trans, root, inode);
1303 }
1304 dir->i_sb->s_dirt = 1;
1305 btrfs_update_inode_block_group(trans, inode);
1306 btrfs_update_inode_block_group(trans, dir);
1307 out_unlock:
1308 nr = trans->blocks_used;
1309 btrfs_end_transaction(trans, root);
1310 mutex_unlock(&root->fs_info->fs_mutex);
1311
1312 if (drop_inode) {
1313 inode_dec_link_count(inode);
1314 iput(inode);
1315 }
1316 btrfs_btree_balance_dirty(root, nr);
1317 return err;
1318 }
1319
1320 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1321 int mode, struct nameidata *nd)
1322 {
1323 struct btrfs_trans_handle *trans;
1324 struct btrfs_root *root = BTRFS_I(dir)->root;
1325 struct inode *inode;
1326 int err;
1327 int drop_inode = 0;
1328 unsigned long nr;
1329 u64 objectid;
1330
1331 mutex_lock(&root->fs_info->fs_mutex);
1332 trans = btrfs_start_transaction(root, 1);
1333 btrfs_set_trans_block_group(trans, dir);
1334
1335 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1336 if (err) {
1337 err = -ENOSPC;
1338 goto out_unlock;
1339 }
1340
1341 inode = btrfs_new_inode(trans, root, objectid,
1342 BTRFS_I(dir)->block_group, mode);
1343 err = PTR_ERR(inode);
1344 if (IS_ERR(inode))
1345 goto out_unlock;
1346
1347 btrfs_set_trans_block_group(trans, inode);
1348 err = btrfs_add_nondir(trans, dentry, inode);
1349 if (err)
1350 drop_inode = 1;
1351 else {
1352 inode->i_mapping->a_ops = &btrfs_aops;
1353 inode->i_fop = &btrfs_file_operations;
1354 inode->i_op = &btrfs_file_inode_operations;
1355 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1356 inode->i_mapping, GFP_NOFS);
1357 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
1358 }
1359 dir->i_sb->s_dirt = 1;
1360 btrfs_update_inode_block_group(trans, inode);
1361 btrfs_update_inode_block_group(trans, dir);
1362 out_unlock:
1363 nr = trans->blocks_used;
1364 btrfs_end_transaction(trans, root);
1365 mutex_unlock(&root->fs_info->fs_mutex);
1366
1367 if (drop_inode) {
1368 inode_dec_link_count(inode);
1369 iput(inode);
1370 }
1371 btrfs_btree_balance_dirty(root, nr);
1372 return err;
1373 }
1374
1375 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1376 struct dentry *dentry)
1377 {
1378 struct btrfs_trans_handle *trans;
1379 struct btrfs_root *root = BTRFS_I(dir)->root;
1380 struct inode *inode = old_dentry->d_inode;
1381 unsigned long nr;
1382 int err;
1383 int drop_inode = 0;
1384
1385 if (inode->i_nlink == 0)
1386 return -ENOENT;
1387
1388 inc_nlink(inode);
1389 mutex_lock(&root->fs_info->fs_mutex);
1390 trans = btrfs_start_transaction(root, 1);
1391
1392 btrfs_set_trans_block_group(trans, dir);
1393 atomic_inc(&inode->i_count);
1394 err = btrfs_add_nondir(trans, dentry, inode);
1395
1396 if (err)
1397 drop_inode = 1;
1398
1399 dir->i_sb->s_dirt = 1;
1400 btrfs_update_inode_block_group(trans, dir);
1401 err = btrfs_update_inode(trans, root, inode);
1402
1403 if (err)
1404 drop_inode = 1;
1405
1406 nr = trans->blocks_used;
1407 btrfs_end_transaction(trans, root);
1408 mutex_unlock(&root->fs_info->fs_mutex);
1409
1410 if (drop_inode) {
1411 inode_dec_link_count(inode);
1412 iput(inode);
1413 }
1414 btrfs_btree_balance_dirty(root, nr);
1415 return err;
1416 }
1417
1418 static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
1419 struct btrfs_root *root,
1420 u64 objectid, u64 dirid)
1421 {
1422 int ret;
1423 char buf[2];
1424 struct btrfs_key key;
1425
1426 buf[0] = '.';
1427 buf[1] = '.';
1428
1429 key.objectid = objectid;
1430 key.offset = 0;
1431 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1432
1433 ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid,
1434 &key, BTRFS_FT_DIR);
1435 if (ret)
1436 goto error;
1437
1438 key.objectid = dirid;
1439 ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid,
1440 &key, BTRFS_FT_DIR);
1441 if (ret)
1442 goto error;
1443 error:
1444 return ret;
1445 }
1446
1447 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1448 {
1449 struct inode *inode;
1450 struct btrfs_trans_handle *trans;
1451 struct btrfs_root *root = BTRFS_I(dir)->root;
1452 int err = 0;
1453 int drop_on_err = 0;
1454 u64 objectid;
1455 unsigned long nr = 1;
1456
1457 mutex_lock(&root->fs_info->fs_mutex);
1458 trans = btrfs_start_transaction(root, 1);
1459 btrfs_set_trans_block_group(trans, dir);
1460
1461 if (IS_ERR(trans)) {
1462 err = PTR_ERR(trans);
1463 goto out_unlock;
1464 }
1465
1466 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1467 if (err) {
1468 err = -ENOSPC;
1469 goto out_unlock;
1470 }
1471
1472 inode = btrfs_new_inode(trans, root, objectid,
1473 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1474 if (IS_ERR(inode)) {
1475 err = PTR_ERR(inode);
1476 goto out_fail;
1477 }
1478
1479 drop_on_err = 1;
1480 inode->i_op = &btrfs_dir_inode_operations;
1481 inode->i_fop = &btrfs_dir_file_operations;
1482 btrfs_set_trans_block_group(trans, inode);
1483
1484 err = btrfs_make_empty_dir(trans, root, inode->i_ino, dir->i_ino);
1485 if (err)
1486 goto out_fail;
1487
1488 inode->i_size = 6;
1489 err = btrfs_update_inode(trans, root, inode);
1490 if (err)
1491 goto out_fail;
1492
1493 err = btrfs_add_link(trans, dentry, inode);
1494 if (err)
1495 goto out_fail;
1496
1497 d_instantiate(dentry, inode);
1498 drop_on_err = 0;
1499 dir->i_sb->s_dirt = 1;
1500 btrfs_update_inode_block_group(trans, inode);
1501 btrfs_update_inode_block_group(trans, dir);
1502
1503 out_fail:
1504 nr = trans->blocks_used;
1505 btrfs_end_transaction(trans, root);
1506
1507 out_unlock:
1508 mutex_unlock(&root->fs_info->fs_mutex);
1509 if (drop_on_err)
1510 iput(inode);
1511 btrfs_btree_balance_dirty(root, nr);
1512 return err;
1513 }
1514
1515 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
1516 size_t page_offset, u64 start, u64 end,
1517 int create)
1518 {
1519 int ret;
1520 int err = 0;
1521 u64 bytenr;
1522 u64 extent_start = 0;
1523 u64 extent_end = 0;
1524 u64 objectid = inode->i_ino;
1525 u32 found_type;
1526 int failed_insert = 0;
1527 struct btrfs_path *path;
1528 struct btrfs_root *root = BTRFS_I(inode)->root;
1529 struct btrfs_file_extent_item *item;
1530 struct extent_buffer *leaf;
1531 struct btrfs_key found_key;
1532 struct extent_map *em = NULL;
1533 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1534 struct btrfs_trans_handle *trans = NULL;
1535
1536 path = btrfs_alloc_path();
1537 BUG_ON(!path);
1538 mutex_lock(&root->fs_info->fs_mutex);
1539
1540 again:
1541 em = lookup_extent_mapping(em_tree, start, end);
1542 if (em) {
1543 goto out;
1544 }
1545 if (!em) {
1546 em = alloc_extent_map(GFP_NOFS);
1547 if (!em) {
1548 err = -ENOMEM;
1549 goto out;
1550 }
1551 em->start = EXTENT_MAP_HOLE;
1552 em->end = EXTENT_MAP_HOLE;
1553 }
1554 em->bdev = inode->i_sb->s_bdev;
1555 ret = btrfs_lookup_file_extent(trans, root, path,
1556 objectid, start, trans != NULL);
1557 if (ret < 0) {
1558 err = ret;
1559 goto out;
1560 }
1561
1562 if (ret != 0) {
1563 if (path->slots[0] == 0)
1564 goto not_found;
1565 path->slots[0]--;
1566 }
1567
1568 leaf = path->nodes[0];
1569 item = btrfs_item_ptr(leaf, path->slots[0],
1570 struct btrfs_file_extent_item);
1571 /* are we inside the extent that was found? */
1572 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1573 found_type = btrfs_key_type(&found_key);
1574 if (found_key.objectid != objectid ||
1575 found_type != BTRFS_EXTENT_DATA_KEY) {
1576 goto not_found;
1577 }
1578
1579 found_type = btrfs_file_extent_type(leaf, item);
1580 extent_start = found_key.offset;
1581 if (found_type == BTRFS_FILE_EXTENT_REG) {
1582 extent_end = extent_start +
1583 btrfs_file_extent_num_bytes(leaf, item);
1584 err = 0;
1585 if (start < extent_start || start >= extent_end) {
1586 em->start = start;
1587 if (start < extent_start) {
1588 if (end < extent_start)
1589 goto not_found;
1590 em->end = extent_end - 1;
1591 } else {
1592 em->end = end;
1593 }
1594 goto not_found_em;
1595 }
1596 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
1597 if (bytenr == 0) {
1598 em->start = extent_start;
1599 em->end = extent_end - 1;
1600 em->block_start = EXTENT_MAP_HOLE;
1601 em->block_end = EXTENT_MAP_HOLE;
1602 goto insert;
1603 }
1604 bytenr += btrfs_file_extent_offset(leaf, item);
1605 em->block_start = bytenr;
1606 em->block_end = em->block_start +
1607 btrfs_file_extent_num_bytes(leaf, item) - 1;
1608 em->start = extent_start;
1609 em->end = extent_end - 1;
1610 goto insert;
1611 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1612 unsigned long ptr;
1613 char *map;
1614 size_t size;
1615 size_t extent_offset;
1616 size_t copy_size;
1617
1618 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
1619 path->slots[0]));
1620 extent_end = (extent_start + size - 1) |
1621 ((u64)root->sectorsize - 1);
1622 if (start < extent_start || start >= extent_end) {
1623 em->start = start;
1624 if (start < extent_start) {
1625 if (end < extent_start)
1626 goto not_found;
1627 em->end = extent_end;
1628 } else {
1629 em->end = end;
1630 }
1631 goto not_found_em;
1632 }
1633 em->block_start = EXTENT_MAP_INLINE;
1634 em->block_end = EXTENT_MAP_INLINE;
1635
1636 if (!page) {
1637 em->start = extent_start;
1638 em->end = extent_start + size - 1;
1639 goto out;
1640 }
1641
1642 extent_offset = ((u64)page->index << PAGE_CACHE_SHIFT) -
1643 extent_start + page_offset;
1644 copy_size = min_t(u64, PAGE_CACHE_SIZE - page_offset,
1645 size - extent_offset);
1646 em->start = extent_start + extent_offset;
1647 em->end = (em->start + copy_size -1) |
1648 ((u64)root->sectorsize -1);
1649 map = kmap(page);
1650 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
1651 if (create == 0 && !PageUptodate(page)) {
1652 read_extent_buffer(leaf, map + page_offset, ptr,
1653 copy_size);
1654 flush_dcache_page(page);
1655 } else if (create && PageUptodate(page)) {
1656 if (!trans) {
1657 kunmap(page);
1658 free_extent_map(em);
1659 em = NULL;
1660 btrfs_release_path(root, path);
1661 trans = btrfs_start_transaction(root, 1);
1662 goto again;
1663 }
1664 write_extent_buffer(leaf, map + page_offset, ptr,
1665 copy_size);
1666 btrfs_mark_buffer_dirty(leaf);
1667 }
1668 kunmap(page);
1669 set_extent_uptodate(em_tree, em->start, em->end, GFP_NOFS);
1670 goto insert;
1671 } else {
1672 printk("unkknown found_type %d\n", found_type);
1673 WARN_ON(1);
1674 }
1675 not_found:
1676 em->start = start;
1677 em->end = end;
1678 not_found_em:
1679 em->block_start = EXTENT_MAP_HOLE;
1680 em->block_end = EXTENT_MAP_HOLE;
1681 insert:
1682 btrfs_release_path(root, path);
1683 if (em->start > start || em->end < start) {
1684 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->end, start, end);
1685 err = -EIO;
1686 goto out;
1687 }
1688 ret = add_extent_mapping(em_tree, em);
1689 if (ret == -EEXIST) {
1690 free_extent_map(em);
1691 em = NULL;
1692 failed_insert++;
1693 if (failed_insert > 5) {
1694 printk("failing to insert %Lu %Lu\n", start, end);
1695 err = -EIO;
1696 goto out;
1697 }
1698 goto again;
1699 }
1700 err = 0;
1701 out:
1702 btrfs_free_path(path);
1703 if (trans) {
1704 ret = btrfs_end_transaction(trans, root);
1705 if (!err)
1706 err = ret;
1707 }
1708 mutex_unlock(&root->fs_info->fs_mutex);
1709 if (err) {
1710 free_extent_map(em);
1711 WARN_ON(1);
1712 return ERR_PTR(err);
1713 }
1714 return em;
1715 }
1716
1717 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
1718 {
1719 return extent_bmap(mapping, iblock, btrfs_get_extent);
1720 }
1721
1722 static int btrfs_prepare_write(struct file *file, struct page *page,
1723 unsigned from, unsigned to)
1724 {
1725 return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree,
1726 page->mapping->host, page, from, to,
1727 btrfs_get_extent);
1728 }
1729
1730 int btrfs_readpage(struct file *file, struct page *page)
1731 {
1732 struct extent_map_tree *tree;
1733 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1734 return extent_read_full_page(tree, page, btrfs_get_extent);
1735 }
1736 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
1737 {
1738 struct extent_map_tree *tree;
1739
1740
1741 if (current->flags & PF_MEMALLOC) {
1742 redirty_page_for_writepage(wbc, page);
1743 unlock_page(page);
1744 return 0;
1745 }
1746 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1747 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
1748 }
1749
1750
1751 static int btrfs_writepages(struct address_space *mapping,
1752 struct writeback_control *wbc)
1753 {
1754 struct extent_map_tree *tree;
1755 tree = &BTRFS_I(mapping->host)->extent_tree;
1756 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
1757 }
1758
1759 static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
1760 {
1761 struct extent_map_tree *tree;
1762 int ret;
1763
1764 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1765 ret = try_release_extent_mapping(tree, page);
1766 if (ret == 1) {
1767 ClearPagePrivate(page);
1768 set_page_private(page, 0);
1769 page_cache_release(page);
1770 }
1771 return ret;
1772 }
1773
1774 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
1775 {
1776 struct extent_map_tree *tree;
1777
1778 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1779 extent_invalidatepage(tree, page, offset);
1780 btrfs_releasepage(page, GFP_NOFS);
1781 }
1782
1783 /*
1784 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
1785 * called from a page fault handler when a page is first dirtied. Hence we must
1786 * be careful to check for EOF conditions here. We set the page up correctly
1787 * for a written page which means we get ENOSPC checking when writing into
1788 * holes and correct delalloc and unwritten extent mapping on filesystems that
1789 * support these features.
1790 *
1791 * We are not allowed to take the i_mutex here so we have to play games to
1792 * protect against truncate races as the page could now be beyond EOF. Because
1793 * vmtruncate() writes the inode size before removing pages, once we have the
1794 * page lock we can determine safely if the page is beyond EOF. If it is not
1795 * beyond EOF, then the page is guaranteed safe against truncation until we
1796 * unlock the page.
1797 */
1798 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1799 {
1800 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
1801 unsigned long end;
1802 loff_t size;
1803 int ret = -EINVAL;
1804 u64 page_start;
1805
1806 down_read(&BTRFS_I(inode)->root->snap_sem);
1807 lock_page(page);
1808 wait_on_page_writeback(page);
1809 size = i_size_read(inode);
1810 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
1811
1812 if ((page->mapping != inode->i_mapping) ||
1813 (page_start > size)) {
1814 /* page got truncated out from underneath us */
1815 goto out_unlock;
1816 }
1817
1818 /* page is wholly or partially inside EOF */
1819 if (page_start + PAGE_CACHE_SIZE > size)
1820 end = size & ~PAGE_CACHE_MASK;
1821 else
1822 end = PAGE_CACHE_SIZE;
1823
1824 ret = btrfs_cow_one_page(inode, page, end);
1825
1826 out_unlock:
1827 up_read(&BTRFS_I(inode)->root->snap_sem);
1828 unlock_page(page);
1829 return ret;
1830 }
1831
1832 static void btrfs_truncate(struct inode *inode)
1833 {
1834 struct btrfs_root *root = BTRFS_I(inode)->root;
1835 int ret;
1836 struct btrfs_trans_handle *trans;
1837 unsigned long nr;
1838
1839 if (!S_ISREG(inode->i_mode))
1840 return;
1841 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1842 return;
1843
1844 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1845
1846 mutex_lock(&root->fs_info->fs_mutex);
1847 trans = btrfs_start_transaction(root, 1);
1848 btrfs_set_trans_block_group(trans, inode);
1849
1850 /* FIXME, add redo link to tree so we don't leak on crash */
1851 ret = btrfs_truncate_in_trans(trans, root, inode);
1852 btrfs_update_inode(trans, root, inode);
1853 nr = trans->blocks_used;
1854
1855 ret = btrfs_end_transaction(trans, root);
1856 BUG_ON(ret);
1857 mutex_unlock(&root->fs_info->fs_mutex);
1858 btrfs_btree_balance_dirty(root, nr);
1859 }
1860
1861 int btrfs_commit_write(struct file *file, struct page *page,
1862 unsigned from, unsigned to)
1863 {
1864 return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree,
1865 page->mapping->host, page, from, to);
1866 }
1867
1868 static int create_subvol(struct btrfs_root *root, char *name, int namelen)
1869 {
1870 struct btrfs_trans_handle *trans;
1871 struct btrfs_key key;
1872 struct btrfs_root_item root_item;
1873 struct btrfs_inode_item *inode_item;
1874 struct extent_buffer *leaf;
1875 struct btrfs_root *new_root;
1876 struct inode *inode;
1877 struct inode *dir;
1878 int ret;
1879 int err;
1880 u64 objectid;
1881 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
1882 unsigned long nr = 1;
1883
1884 mutex_lock(&root->fs_info->fs_mutex);
1885 trans = btrfs_start_transaction(root, 1);
1886 BUG_ON(!trans);
1887
1888 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, 0);
1889 if (IS_ERR(leaf))
1890 return PTR_ERR(leaf);
1891
1892 btrfs_set_header_nritems(leaf, 0);
1893 btrfs_set_header_level(leaf, 0);
1894 btrfs_set_header_bytenr(leaf, leaf->start);
1895 btrfs_set_header_generation(leaf, trans->transid);
1896 btrfs_set_header_owner(leaf, root->root_key.objectid);
1897 write_extent_buffer(leaf, root->fs_info->fsid,
1898 (unsigned long)btrfs_header_fsid(leaf),
1899 BTRFS_FSID_SIZE);
1900 btrfs_mark_buffer_dirty(leaf);
1901
1902 inode_item = &root_item.inode;
1903 memset(inode_item, 0, sizeof(*inode_item));
1904 inode_item->generation = cpu_to_le64(1);
1905 inode_item->size = cpu_to_le64(3);
1906 inode_item->nlink = cpu_to_le32(1);
1907 inode_item->nblocks = cpu_to_le64(1);
1908 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1909
1910 btrfs_set_root_bytenr(&root_item, leaf->start);
1911 btrfs_set_root_level(&root_item, 0);
1912 btrfs_set_root_refs(&root_item, 1);
1913 btrfs_set_root_used(&root_item, 0);
1914
1915 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
1916 root_item.drop_level = 0;
1917
1918 free_extent_buffer(leaf);
1919 leaf = NULL;
1920
1921 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1922 0, &objectid);
1923 if (ret)
1924 goto fail;
1925
1926 btrfs_set_root_dirid(&root_item, new_dirid);
1927
1928 key.objectid = objectid;
1929 key.offset = 1;
1930 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1931 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1932 &root_item);
1933 if (ret)
1934 goto fail;
1935
1936 /*
1937 * insert the directory item
1938 */
1939 key.offset = (u64)-1;
1940 dir = root->fs_info->sb->s_root->d_inode;
1941 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1942 name, namelen, dir->i_ino, &key,
1943 BTRFS_FT_DIR);
1944 if (ret)
1945 goto fail;
1946
1947 ret = btrfs_commit_transaction(trans, root);
1948 if (ret)
1949 goto fail_commit;
1950
1951 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
1952 BUG_ON(!new_root);
1953
1954 trans = btrfs_start_transaction(new_root, 1);
1955 BUG_ON(!trans);
1956
1957 inode = btrfs_new_inode(trans, new_root, new_dirid,
1958 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
1959 if (IS_ERR(inode))
1960 goto fail;
1961 inode->i_op = &btrfs_dir_inode_operations;
1962 inode->i_fop = &btrfs_dir_file_operations;
1963 new_root->inode = inode;
1964
1965 ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid);
1966 if (ret)
1967 goto fail;
1968
1969 inode->i_nlink = 1;
1970 inode->i_size = 6;
1971 ret = btrfs_update_inode(trans, new_root, inode);
1972 if (ret)
1973 goto fail;
1974 fail:
1975 nr = trans->blocks_used;
1976 err = btrfs_commit_transaction(trans, root);
1977 if (err && !ret)
1978 ret = err;
1979 fail_commit:
1980 mutex_unlock(&root->fs_info->fs_mutex);
1981 btrfs_btree_balance_dirty(root, nr);
1982 return ret;
1983 }
1984
1985 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
1986 {
1987 struct btrfs_trans_handle *trans;
1988 struct btrfs_key key;
1989 struct btrfs_root_item new_root_item;
1990 struct extent_buffer *tmp;
1991 int ret;
1992 int err;
1993 u64 objectid;
1994 unsigned long nr;
1995
1996 if (!root->ref_cows)
1997 return -EINVAL;
1998
1999 down_write(&root->snap_sem);
2000 freeze_bdev(root->fs_info->sb->s_bdev);
2001 thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb);
2002
2003 mutex_lock(&root->fs_info->fs_mutex);
2004 trans = btrfs_start_transaction(root, 1);
2005 BUG_ON(!trans);
2006
2007 ret = btrfs_update_inode(trans, root, root->inode);
2008 if (ret)
2009 goto fail;
2010
2011 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
2012 0, &objectid);
2013 if (ret)
2014 goto fail;
2015
2016 memcpy(&new_root_item, &root->root_item,
2017 sizeof(new_root_item));
2018
2019 key.objectid = objectid;
2020 key.offset = 1;
2021 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
2022
2023 btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
2024 btrfs_set_root_bytenr(&new_root_item, root->node->start);
2025 btrfs_set_root_level(&new_root_item, btrfs_header_level(root->node));
2026
2027 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
2028 &new_root_item);
2029 if (ret)
2030 goto fail;
2031
2032 /*
2033 * insert the directory item
2034 */
2035 key.offset = (u64)-1;
2036 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
2037 name, namelen,
2038 root->fs_info->sb->s_root->d_inode->i_ino,
2039 &key, BTRFS_FT_DIR);
2040
2041 if (ret)
2042 goto fail;
2043
2044 ret = btrfs_inc_root_ref(trans, root);
2045 if (ret)
2046 goto fail;
2047 fail:
2048 nr = trans->blocks_used;
2049 err = btrfs_commit_transaction(trans, root);
2050
2051 if (err && !ret)
2052 ret = err;
2053
2054 mutex_unlock(&root->fs_info->fs_mutex);
2055 up_write(&root->snap_sem);
2056 btrfs_btree_balance_dirty(root, nr);
2057 return ret;
2058 }
2059
2060 static unsigned long force_ra(struct address_space *mapping,
2061 struct file_ra_state *ra, struct file *file,
2062 pgoff_t offset, pgoff_t last_index)
2063 {
2064 pgoff_t req_size;
2065
2066 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2067 req_size = last_index - offset + 1;
2068 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2069 return offset;
2070 #else
2071 req_size = min(last_index - offset + 1, (pgoff_t)128);
2072 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2073 return offset + req_size;
2074 #endif
2075 }
2076
2077 int btrfs_defrag_file(struct file *file) {
2078 struct inode *inode = file->f_path.dentry->d_inode;
2079 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2080 struct page *page;
2081 unsigned long last_index;
2082 unsigned long ra_index = 0;
2083 u64 page_start;
2084 u64 page_end;
2085 unsigned long i;
2086
2087 mutex_lock(&inode->i_mutex);
2088 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
2089 for (i = 0; i <= last_index; i++) {
2090 if (i == ra_index) {
2091 ra_index = force_ra(inode->i_mapping, &file->f_ra,
2092 file, ra_index, last_index);
2093 }
2094 page = grab_cache_page(inode->i_mapping, i);
2095 if (!page)
2096 goto out_unlock;
2097 if (!PageUptodate(page)) {
2098 btrfs_readpage(NULL, page);
2099 lock_page(page);
2100 if (!PageUptodate(page)) {
2101 unlock_page(page);
2102 page_cache_release(page);
2103 goto out_unlock;
2104 }
2105 }
2106 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2107 page_end = page_start + PAGE_CACHE_SIZE - 1;
2108
2109 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
2110 set_extent_delalloc(em_tree, page_start,
2111 page_end, GFP_NOFS);
2112 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
2113 set_page_dirty(page);
2114 unlock_page(page);
2115 page_cache_release(page);
2116 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
2117 }
2118
2119 out_unlock:
2120 mutex_unlock(&inode->i_mutex);
2121 return 0;
2122 }
2123
2124 static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg)
2125 {
2126 struct btrfs_ioctl_vol_args vol_args;
2127 struct btrfs_dir_item *di;
2128 struct btrfs_path *path;
2129 int namelen;
2130 u64 root_dirid;
2131
2132 if (copy_from_user(&vol_args, arg, sizeof(vol_args)))
2133 return -EFAULT;
2134
2135 namelen = strlen(vol_args.name);
2136 if (namelen > BTRFS_VOL_NAME_MAX)
2137 return -EINVAL;
2138 if (strchr(vol_args.name, '/'))
2139 return -EINVAL;
2140
2141 path = btrfs_alloc_path();
2142 if (!path)
2143 return -ENOMEM;
2144
2145 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2146 mutex_lock(&root->fs_info->fs_mutex);
2147 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2148 path, root_dirid,
2149 vol_args.name, namelen, 0);
2150 mutex_unlock(&root->fs_info->fs_mutex);
2151 btrfs_free_path(path);
2152 if (di && !IS_ERR(di))
2153 return -EEXIST;
2154 if (IS_ERR(di))
2155 return PTR_ERR(di);
2156
2157 if (root == root->fs_info->tree_root)
2158 return create_subvol(root, vol_args.name, namelen);
2159 return create_snapshot(root, vol_args.name, namelen);
2160 }
2161
2162 static int btrfs_ioctl_defrag(struct file *file)
2163 {
2164 struct inode *inode = file->f_path.dentry->d_inode;
2165 struct btrfs_root *root = BTRFS_I(inode)->root;
2166
2167 switch (inode->i_mode & S_IFMT) {
2168 case S_IFDIR:
2169 mutex_lock(&root->fs_info->fs_mutex);
2170 btrfs_defrag_root(root, 0);
2171 btrfs_defrag_root(root->fs_info->extent_root, 0);
2172 mutex_unlock(&root->fs_info->fs_mutex);
2173 break;
2174 case S_IFREG:
2175 btrfs_defrag_file(file);
2176 break;
2177 }
2178
2179 return 0;
2180 }
2181
2182 long btrfs_ioctl(struct file *file, unsigned int
2183 cmd, unsigned long arg)
2184 {
2185 struct btrfs_root *root = BTRFS_I(file->f_path.dentry->d_inode)->root;
2186
2187 switch (cmd) {
2188 case BTRFS_IOC_SNAP_CREATE:
2189 return btrfs_ioctl_snap_create(root, (void __user *)arg);
2190 case BTRFS_IOC_DEFRAG:
2191 return btrfs_ioctl_defrag(file);
2192 }
2193
2194 return -ENOTTY;
2195 }
2196
2197 /*
2198 * Called inside transaction, so use GFP_NOFS
2199 */
2200 struct inode *btrfs_alloc_inode(struct super_block *sb)
2201 {
2202 struct btrfs_inode *ei;
2203
2204 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2205 if (!ei)
2206 return NULL;
2207 ei->last_trans = 0;
2208 return &ei->vfs_inode;
2209 }
2210
2211 void btrfs_destroy_inode(struct inode *inode)
2212 {
2213 WARN_ON(!list_empty(&inode->i_dentry));
2214 WARN_ON(inode->i_data.nrpages);
2215
2216 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2217 }
2218
2219 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2220 static void init_once(struct kmem_cache * cachep, void *foo)
2221 #else
2222 static void init_once(void * foo, struct kmem_cache * cachep,
2223 unsigned long flags)
2224 #endif
2225 {
2226 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2227
2228 inode_init_once(&ei->vfs_inode);
2229 }
2230
2231 void btrfs_destroy_cachep(void)
2232 {
2233 if (btrfs_inode_cachep)
2234 kmem_cache_destroy(btrfs_inode_cachep);
2235 if (btrfs_trans_handle_cachep)
2236 kmem_cache_destroy(btrfs_trans_handle_cachep);
2237 if (btrfs_transaction_cachep)
2238 kmem_cache_destroy(btrfs_transaction_cachep);
2239 if (btrfs_bit_radix_cachep)
2240 kmem_cache_destroy(btrfs_bit_radix_cachep);
2241 if (btrfs_path_cachep)
2242 kmem_cache_destroy(btrfs_path_cachep);
2243 }
2244
2245 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2246 unsigned long extra_flags,
2247 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2248 void (*ctor)(struct kmem_cache *, void *)
2249 #else
2250 void (*ctor)(void *, struct kmem_cache *,
2251 unsigned long)
2252 #endif
2253 )
2254 {
2255 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2256 SLAB_MEM_SPREAD | extra_flags), ctor
2257 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2258 ,NULL
2259 #endif
2260 );
2261 }
2262
2263 int btrfs_init_cachep(void)
2264 {
2265 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2266 sizeof(struct btrfs_inode),
2267 0, init_once);
2268 if (!btrfs_inode_cachep)
2269 goto fail;
2270 btrfs_trans_handle_cachep =
2271 btrfs_cache_create("btrfs_trans_handle_cache",
2272 sizeof(struct btrfs_trans_handle),
2273 0, NULL);
2274 if (!btrfs_trans_handle_cachep)
2275 goto fail;
2276 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
2277 sizeof(struct btrfs_transaction),
2278 0, NULL);
2279 if (!btrfs_transaction_cachep)
2280 goto fail;
2281 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
2282 sizeof(struct btrfs_path),
2283 0, NULL);
2284 if (!btrfs_path_cachep)
2285 goto fail;
2286 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
2287 SLAB_DESTROY_BY_RCU, NULL);
2288 if (!btrfs_bit_radix_cachep)
2289 goto fail;
2290 return 0;
2291 fail:
2292 btrfs_destroy_cachep();
2293 return -ENOMEM;
2294 }
2295
2296 static int btrfs_getattr(struct vfsmount *mnt,
2297 struct dentry *dentry, struct kstat *stat)
2298 {
2299 struct inode *inode = dentry->d_inode;
2300 generic_fillattr(inode, stat);
2301 stat->blksize = 256 * 1024;
2302 return 0;
2303 }
2304
2305 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2306 struct inode * new_dir,struct dentry *new_dentry)
2307 {
2308 struct btrfs_trans_handle *trans;
2309 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2310 struct inode *new_inode = new_dentry->d_inode;
2311 struct inode *old_inode = old_dentry->d_inode;
2312 struct timespec ctime = CURRENT_TIME;
2313 struct btrfs_path *path;
2314 struct btrfs_dir_item *di;
2315 int ret;
2316
2317 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2318 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2319 return -ENOTEMPTY;
2320 }
2321
2322 mutex_lock(&root->fs_info->fs_mutex);
2323 trans = btrfs_start_transaction(root, 1);
2324
2325 btrfs_set_trans_block_group(trans, new_dir);
2326 path = btrfs_alloc_path();
2327 if (!path) {
2328 ret = -ENOMEM;
2329 goto out_fail;
2330 }
2331
2332 old_dentry->d_inode->i_nlink++;
2333 old_dir->i_ctime = old_dir->i_mtime = ctime;
2334 new_dir->i_ctime = new_dir->i_mtime = ctime;
2335 old_inode->i_ctime = ctime;
2336
2337 if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) {
2338 struct btrfs_key *location = &BTRFS_I(new_dir)->location;
2339 struct btrfs_key old_parent_key;
2340 di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino,
2341 "..", 2, -1);
2342 if (IS_ERR(di)) {
2343 ret = PTR_ERR(di);
2344 goto out_fail;
2345 }
2346 if (!di) {
2347 ret = -ENOENT;
2348 goto out_fail;
2349 }
2350 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &old_parent_key);
2351 ret = btrfs_del_item(trans, root, path);
2352 if (ret) {
2353 goto out_fail;
2354 }
2355 btrfs_release_path(root, path);
2356
2357 di = btrfs_lookup_dir_index_item(trans, root, path,
2358 old_inode->i_ino,
2359 old_parent_key.objectid,
2360 "..", 2, -1);
2361 if (IS_ERR(di)) {
2362 ret = PTR_ERR(di);
2363 goto out_fail;
2364 }
2365 if (!di) {
2366 ret = -ENOENT;
2367 goto out_fail;
2368 }
2369 ret = btrfs_del_item(trans, root, path);
2370 if (ret) {
2371 goto out_fail;
2372 }
2373 btrfs_release_path(root, path);
2374
2375 ret = btrfs_insert_dir_item(trans, root, "..", 2,
2376 old_inode->i_ino, location,
2377 BTRFS_FT_DIR);
2378 if (ret)
2379 goto out_fail;
2380 }
2381
2382
2383 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2384 if (ret)
2385 goto out_fail;
2386
2387 if (new_inode) {
2388 new_inode->i_ctime = CURRENT_TIME;
2389 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2390 if (ret)
2391 goto out_fail;
2392 }
2393 ret = btrfs_add_link(trans, new_dentry, old_inode);
2394 if (ret)
2395 goto out_fail;
2396
2397 out_fail:
2398 btrfs_free_path(path);
2399 btrfs_end_transaction(trans, root);
2400 mutex_unlock(&root->fs_info->fs_mutex);
2401 return ret;
2402 }
2403
2404 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2405 const char *symname)
2406 {
2407 struct btrfs_trans_handle *trans;
2408 struct btrfs_root *root = BTRFS_I(dir)->root;
2409 struct btrfs_path *path;
2410 struct btrfs_key key;
2411 struct inode *inode;
2412 int err;
2413 int drop_inode = 0;
2414 u64 objectid;
2415 int name_len;
2416 int datasize;
2417 unsigned long ptr;
2418 struct btrfs_file_extent_item *ei;
2419 struct extent_buffer *leaf;
2420 unsigned long nr;
2421
2422 name_len = strlen(symname) + 1;
2423 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2424 return -ENAMETOOLONG;
2425 mutex_lock(&root->fs_info->fs_mutex);
2426 trans = btrfs_start_transaction(root, 1);
2427 btrfs_set_trans_block_group(trans, dir);
2428
2429 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2430 if (err) {
2431 err = -ENOSPC;
2432 goto out_unlock;
2433 }
2434
2435 inode = btrfs_new_inode(trans, root, objectid,
2436 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2437 err = PTR_ERR(inode);
2438 if (IS_ERR(inode))
2439 goto out_unlock;
2440
2441 btrfs_set_trans_block_group(trans, inode);
2442 err = btrfs_add_nondir(trans, dentry, inode);
2443 if (err)
2444 drop_inode = 1;
2445 else {
2446 inode->i_mapping->a_ops = &btrfs_aops;
2447 inode->i_fop = &btrfs_file_operations;
2448 inode->i_op = &btrfs_file_inode_operations;
2449 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
2450 inode->i_mapping, GFP_NOFS);
2451 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
2452 }
2453 dir->i_sb->s_dirt = 1;
2454 btrfs_update_inode_block_group(trans, inode);
2455 btrfs_update_inode_block_group(trans, dir);
2456 if (drop_inode)
2457 goto out_unlock;
2458
2459 path = btrfs_alloc_path();
2460 BUG_ON(!path);
2461 key.objectid = inode->i_ino;
2462 key.offset = 0;
2463 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2464 datasize = btrfs_file_extent_calc_inline_size(name_len);
2465 err = btrfs_insert_empty_item(trans, root, path, &key,
2466 datasize);
2467 if (err) {
2468 drop_inode = 1;
2469 goto out_unlock;
2470 }
2471 leaf = path->nodes[0];
2472 ei = btrfs_item_ptr(leaf, path->slots[0],
2473 struct btrfs_file_extent_item);
2474 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
2475 btrfs_set_file_extent_type(leaf, ei,
2476 BTRFS_FILE_EXTENT_INLINE);
2477 ptr = btrfs_file_extent_inline_start(ei);
2478 write_extent_buffer(leaf, symname, ptr, name_len);
2479 btrfs_mark_buffer_dirty(leaf);
2480 btrfs_free_path(path);
2481
2482 inode->i_op = &btrfs_symlink_inode_operations;
2483 inode->i_mapping->a_ops = &btrfs_symlink_aops;
2484 inode->i_size = name_len - 1;
2485 err = btrfs_update_inode(trans, root, inode);
2486 if (err)
2487 drop_inode = 1;
2488
2489 out_unlock:
2490 nr = trans->blocks_used;
2491 btrfs_end_transaction(trans, root);
2492 mutex_unlock(&root->fs_info->fs_mutex);
2493 if (drop_inode) {
2494 inode_dec_link_count(inode);
2495 iput(inode);
2496 }
2497 btrfs_btree_balance_dirty(root, nr);
2498 return err;
2499 }
2500
2501 static struct inode_operations btrfs_dir_inode_operations = {
2502 .lookup = btrfs_lookup,
2503 .create = btrfs_create,
2504 .unlink = btrfs_unlink,
2505 .link = btrfs_link,
2506 .mkdir = btrfs_mkdir,
2507 .rmdir = btrfs_rmdir,
2508 .rename = btrfs_rename,
2509 .symlink = btrfs_symlink,
2510 .setattr = btrfs_setattr,
2511 .mknod = btrfs_mknod,
2512 };
2513
2514 static struct inode_operations btrfs_dir_ro_inode_operations = {
2515 .lookup = btrfs_lookup,
2516 };
2517
2518 static struct file_operations btrfs_dir_file_operations = {
2519 .llseek = generic_file_llseek,
2520 .read = generic_read_dir,
2521 .readdir = btrfs_readdir,
2522 .unlocked_ioctl = btrfs_ioctl,
2523 #ifdef CONFIG_COMPAT
2524 .compat_ioctl = btrfs_ioctl,
2525 #endif
2526 };
2527
2528 static struct extent_map_ops btrfs_extent_map_ops = {
2529 .fill_delalloc = run_delalloc_range,
2530 .writepage_io_hook = btrfs_writepage_io_hook,
2531 .readpage_io_hook = btrfs_readpage_io_hook,
2532 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
2533 };
2534
2535 static struct address_space_operations btrfs_aops = {
2536 .readpage = btrfs_readpage,
2537 .writepage = btrfs_writepage,
2538 .writepages = btrfs_writepages,
2539 .sync_page = block_sync_page,
2540 .prepare_write = btrfs_prepare_write,
2541 .commit_write = btrfs_commit_write,
2542 .bmap = btrfs_bmap,
2543 .invalidatepage = btrfs_invalidatepage,
2544 .releasepage = btrfs_releasepage,
2545 .set_page_dirty = __set_page_dirty_nobuffers,
2546 };
2547
2548 static struct address_space_operations btrfs_symlink_aops = {
2549 .readpage = btrfs_readpage,
2550 .writepage = btrfs_writepage,
2551 .invalidatepage = btrfs_invalidatepage,
2552 .releasepage = btrfs_releasepage,
2553 };
2554
2555 static struct inode_operations btrfs_file_inode_operations = {
2556 .truncate = btrfs_truncate,
2557 .getattr = btrfs_getattr,
2558 .setattr = btrfs_setattr,
2559 };
2560
2561 static struct inode_operations btrfs_special_inode_operations = {
2562 .getattr = btrfs_getattr,
2563 .setattr = btrfs_setattr,
2564 };
2565
2566 static struct inode_operations btrfs_symlink_inode_operations = {
2567 .readlink = generic_readlink,
2568 .follow_link = page_follow_link_light,
2569 .put_link = page_put_link,
2570 };
Linux kernel - Deliverables
This page took 0.080528 seconds and 6 git commands to generate.