projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Btrfs: Add readpages 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 item_end--;
615 }
616 if (found_type == BTRFS_CSUM_ITEM_KEY) {
617 ret = btrfs_csum_truncate(trans, root, path,
618 inode->i_size);
619 BUG_ON(ret);
620 }
621 if (item_end < inode->i_size) {
622 if (found_type == BTRFS_DIR_ITEM_KEY) {
623 found_type = BTRFS_INODE_ITEM_KEY;
624 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
625 found_type = BTRFS_CSUM_ITEM_KEY;
626 } else if (found_type) {
627 found_type--;
628 } else {
629 break;
630 }
631 btrfs_set_key_type(&key, found_type);
632 btrfs_release_path(root, path);
633 continue;
634 }
635 if (found_key.offset >= inode->i_size)
636 del_item = 1;
637 else
638 del_item = 0;
639 found_extent = 0;
640
641 /* FIXME, shrink the extent if the ref count is only 1 */
642 if (found_type != BTRFS_EXTENT_DATA_KEY)
643 goto delete;
644
645 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
646 u64 num_dec;
647 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
648 if (!del_item) {
649 u64 orig_num_bytes =
650 btrfs_file_extent_num_bytes(leaf, fi);
651 extent_num_bytes = inode->i_size -
652 found_key.offset + root->sectorsize - 1;
653 btrfs_set_file_extent_num_bytes(leaf, fi,
654 extent_num_bytes);
655 num_dec = (orig_num_bytes -
656 extent_num_bytes) >> 9;
657 if (extent_start != 0) {
658 inode->i_blocks -= num_dec;
659 }
660 btrfs_mark_buffer_dirty(leaf);
661 } else {
662 extent_num_bytes =
663 btrfs_file_extent_disk_num_bytes(leaf,
664 fi);
665 /* FIXME blocksize != 4096 */
666 num_dec = btrfs_file_extent_num_bytes(leaf,
667 fi) >> 9;
668 if (extent_start != 0) {
669 found_extent = 1;
670 inode->i_blocks -= num_dec;
671 }
672 }
673 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE &&
674 !del_item) {
675 u32 newsize = inode->i_size - found_key.offset;
676 newsize = btrfs_file_extent_calc_inline_size(newsize);
677 ret = btrfs_truncate_item(trans, root, path,
678 newsize, 1);
679 BUG_ON(ret);
680 }
681 delete:
682 if (del_item) {
683 ret = btrfs_del_item(trans, root, path);
684 if (ret)
685 goto error;
686 } else {
687 break;
688 }
689 btrfs_release_path(root, path);
690 if (found_extent) {
691 ret = btrfs_free_extent(trans, root, extent_start,
692 extent_num_bytes, 0);
693 BUG_ON(ret);
694 }
695 }
696 ret = 0;
697 error:
698 btrfs_release_path(root, path);
699 btrfs_free_path(path);
700 inode->i_sb->s_dirt = 1;
701 return ret;
702 }
703
704 static int btrfs_cow_one_page(struct inode *inode, struct page *page,
705 size_t zero_start)
706 {
707 char *kaddr;
708 int ret = 0;
709 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
710 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
711 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
712
713 set_page_extent_mapped(page);
714
715 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
716 set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start,
717 page_end, GFP_NOFS);
718 if (zero_start != PAGE_CACHE_SIZE) {
719 kaddr = kmap(page);
720 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
721 flush_dcache_page(page);
722 kunmap(page);
723 }
724 set_page_dirty(page);
725 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
726
727 return ret;
728 }
729
730 /*
731 * taken from block_truncate_page, but does cow as it zeros out
732 * any bytes left in the last page in the file.
733 */
734 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
735 {
736 struct inode *inode = mapping->host;
737 struct btrfs_root *root = BTRFS_I(inode)->root;
738 u32 blocksize = root->sectorsize;
739 pgoff_t index = from >> PAGE_CACHE_SHIFT;
740 unsigned offset = from & (PAGE_CACHE_SIZE-1);
741 struct page *page;
742 int ret = 0;
743 u64 page_start;
744
745 if ((offset & (blocksize - 1)) == 0)
746 goto out;
747
748 down_read(&root->snap_sem);
749 ret = -ENOMEM;
750 page = grab_cache_page(mapping, index);
751 if (!page)
752 goto out;
753 if (!PageUptodate(page)) {
754 ret = btrfs_readpage(NULL, page);
755 lock_page(page);
756 if (!PageUptodate(page)) {
757 ret = -EIO;
758 goto out;
759 }
760 }
761 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
762
763 ret = btrfs_cow_one_page(inode, page, offset);
764
765 unlock_page(page);
766 page_cache_release(page);
767 up_read(&BTRFS_I(inode)->root->snap_sem);
768 out:
769 return ret;
770 }
771
772 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
773 {
774 struct inode *inode = dentry->d_inode;
775 int err;
776
777 err = inode_change_ok(inode, attr);
778 if (err)
779 return err;
780
781 if (S_ISREG(inode->i_mode) &&
782 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
783 struct btrfs_trans_handle *trans;
784 struct btrfs_root *root = BTRFS_I(inode)->root;
785 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
786
787 u64 mask = root->sectorsize - 1;
788 u64 pos = (inode->i_size + mask) & ~mask;
789 u64 block_end = attr->ia_size | mask;
790 u64 hole_size;
791 u64 alloc_hint = 0;
792
793 if (attr->ia_size <= pos)
794 goto out;
795
796 btrfs_truncate_page(inode->i_mapping, inode->i_size);
797
798 lock_extent(em_tree, pos, block_end, GFP_NOFS);
799 hole_size = (attr->ia_size - pos + mask) & ~mask;
800
801 mutex_lock(&root->fs_info->fs_mutex);
802 trans = btrfs_start_transaction(root, 1);
803 btrfs_set_trans_block_group(trans, inode);
804 err = btrfs_drop_extents(trans, root, inode,
805 pos, pos + hole_size, pos,
806 &alloc_hint);
807
808 if (alloc_hint != EXTENT_MAP_INLINE) {
809 err = btrfs_insert_file_extent(trans, root,
810 inode->i_ino,
811 pos, 0, 0, hole_size);
812 }
813 btrfs_end_transaction(trans, root);
814 mutex_unlock(&root->fs_info->fs_mutex);
815 unlock_extent(em_tree, pos, block_end, GFP_NOFS);
816 if (err)
817 return err;
818 }
819 out:
820 err = inode_setattr(inode, attr);
821
822 return err;
823 }
824 void btrfs_delete_inode(struct inode *inode)
825 {
826 struct btrfs_trans_handle *trans;
827 struct btrfs_root *root = BTRFS_I(inode)->root;
828 unsigned long nr;
829 int ret;
830
831 truncate_inode_pages(&inode->i_data, 0);
832 if (is_bad_inode(inode)) {
833 goto no_delete;
834 }
835
836 inode->i_size = 0;
837 mutex_lock(&root->fs_info->fs_mutex);
838 trans = btrfs_start_transaction(root, 1);
839
840 btrfs_set_trans_block_group(trans, inode);
841 ret = btrfs_truncate_in_trans(trans, root, inode);
842 if (ret)
843 goto no_delete_lock;
844 ret = btrfs_free_inode(trans, root, inode);
845 if (ret)
846 goto no_delete_lock;
847 nr = trans->blocks_used;
848
849 btrfs_end_transaction(trans, root);
850 mutex_unlock(&root->fs_info->fs_mutex);
851 btrfs_btree_balance_dirty(root, nr);
852 return;
853
854 no_delete_lock:
855 nr = trans->blocks_used;
856 btrfs_end_transaction(trans, root);
857 mutex_unlock(&root->fs_info->fs_mutex);
858 btrfs_btree_balance_dirty(root, nr);
859 no_delete:
860 clear_inode(inode);
861 }
862
863 /*
864 * this returns the key found in the dir entry in the location pointer.
865 * If no dir entries were found, location->objectid is 0.
866 */
867 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
868 struct btrfs_key *location)
869 {
870 const char *name = dentry->d_name.name;
871 int namelen = dentry->d_name.len;
872 struct btrfs_dir_item *di;
873 struct btrfs_path *path;
874 struct btrfs_root *root = BTRFS_I(dir)->root;
875 int ret = 0;
876
877 path = btrfs_alloc_path();
878 BUG_ON(!path);
879 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
880 namelen, 0);
881 if (IS_ERR(di))
882 ret = PTR_ERR(di);
883 if (!di || IS_ERR(di)) {
884 location->objectid = 0;
885 goto out;
886 }
887 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
888 out:
889 btrfs_release_path(root, path);
890 btrfs_free_path(path);
891 return ret;
892 }
893
894 /*
895 * when we hit a tree root in a directory, the btrfs part of the inode
896 * needs to be changed to reflect the root directory of the tree root. This
897 * is kind of like crossing a mount point.
898 */
899 static int fixup_tree_root_location(struct btrfs_root *root,
900 struct btrfs_key *location,
901 struct btrfs_root **sub_root,
902 struct dentry *dentry)
903 {
904 struct btrfs_path *path;
905 struct btrfs_root_item *ri;
906
907 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
908 return 0;
909 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
910 return 0;
911
912 path = btrfs_alloc_path();
913 BUG_ON(!path);
914 mutex_lock(&root->fs_info->fs_mutex);
915
916 *sub_root = btrfs_read_fs_root(root->fs_info, location,
917 dentry->d_name.name,
918 dentry->d_name.len);
919 if (IS_ERR(*sub_root))
920 return PTR_ERR(*sub_root);
921
922 ri = &(*sub_root)->root_item;
923 location->objectid = btrfs_root_dirid(ri);
924 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
925 location->offset = 0;
926
927 btrfs_free_path(path);
928 mutex_unlock(&root->fs_info->fs_mutex);
929 return 0;
930 }
931
932 static int btrfs_init_locked_inode(struct inode *inode, void *p)
933 {
934 struct btrfs_iget_args *args = p;
935 inode->i_ino = args->ino;
936 BTRFS_I(inode)->root = args->root;
937 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
938 inode->i_mapping, GFP_NOFS);
939 return 0;
940 }
941
942 static int btrfs_find_actor(struct inode *inode, void *opaque)
943 {
944 struct btrfs_iget_args *args = opaque;
945 return (args->ino == inode->i_ino &&
946 args->root == BTRFS_I(inode)->root);
947 }
948
949 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
950 struct btrfs_root *root)
951 {
952 struct inode *inode;
953 struct btrfs_iget_args args;
954 args.ino = objectid;
955 args.root = root;
956
957 inode = iget5_locked(s, objectid, btrfs_find_actor,
958 btrfs_init_locked_inode,
959 (void *)&args);
960 return inode;
961 }
962
963 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
964 struct nameidata *nd)
965 {
966 struct inode * inode;
967 struct btrfs_inode *bi = BTRFS_I(dir);
968 struct btrfs_root *root = bi->root;
969 struct btrfs_root *sub_root = root;
970 struct btrfs_key location;
971 int ret;
972
973 if (dentry->d_name.len > BTRFS_NAME_LEN)
974 return ERR_PTR(-ENAMETOOLONG);
975
976 mutex_lock(&root->fs_info->fs_mutex);
977 ret = btrfs_inode_by_name(dir, dentry, &location);
978 mutex_unlock(&root->fs_info->fs_mutex);
979
980 if (ret < 0)
981 return ERR_PTR(ret);
982
983 inode = NULL;
984 if (location.objectid) {
985 ret = fixup_tree_root_location(root, &location, &sub_root,
986 dentry);
987 if (ret < 0)
988 return ERR_PTR(ret);
989 if (ret > 0)
990 return ERR_PTR(-ENOENT);
991 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
992 sub_root);
993 if (!inode)
994 return ERR_PTR(-EACCES);
995 if (inode->i_state & I_NEW) {
996 /* the inode and parent dir are two different roots */
997 if (sub_root != root) {
998 igrab(inode);
999 sub_root->inode = inode;
1000 }
1001 BTRFS_I(inode)->root = sub_root;
1002 memcpy(&BTRFS_I(inode)->location, &location,
1003 sizeof(location));
1004 btrfs_read_locked_inode(inode);
1005 unlock_new_inode(inode);
1006 }
1007 }
1008 return d_splice_alias(inode, dentry);
1009 }
1010
1011 static unsigned char btrfs_filetype_table[] = {
1012 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1013 };
1014
1015 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1016 {
1017 struct inode *inode = filp->f_path.dentry->d_inode;
1018 struct btrfs_root *root = BTRFS_I(inode)->root;
1019 struct btrfs_item *item;
1020 struct btrfs_dir_item *di;
1021 struct btrfs_key key;
1022 struct btrfs_key found_key;
1023 struct btrfs_path *path;
1024 int ret;
1025 u32 nritems;
1026 struct extent_buffer *leaf;
1027 int slot;
1028 int advance;
1029 unsigned char d_type;
1030 int over = 0;
1031 u32 di_cur;
1032 u32 di_total;
1033 u32 di_len;
1034 int key_type = BTRFS_DIR_INDEX_KEY;
1035 char tmp_name[32];
1036 char *name_ptr;
1037 int name_len;
1038
1039 /* FIXME, use a real flag for deciding about the key type */
1040 if (root->fs_info->tree_root == root)
1041 key_type = BTRFS_DIR_ITEM_KEY;
1042
1043 mutex_lock(&root->fs_info->fs_mutex);
1044 key.objectid = inode->i_ino;
1045 btrfs_set_key_type(&key, key_type);
1046 key.offset = filp->f_pos;
1047
1048 path = btrfs_alloc_path();
1049 path->reada = 2;
1050 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1051 if (ret < 0)
1052 goto err;
1053 advance = 0;
1054 while(1) {
1055 leaf = path->nodes[0];
1056 nritems = btrfs_header_nritems(leaf);
1057 slot = path->slots[0];
1058 if (advance || slot >= nritems) {
1059 if (slot >= nritems -1) {
1060 ret = btrfs_next_leaf(root, path);
1061 if (ret)
1062 break;
1063 leaf = path->nodes[0];
1064 nritems = btrfs_header_nritems(leaf);
1065 slot = path->slots[0];
1066 } else {
1067 slot++;
1068 path->slots[0]++;
1069 }
1070 }
1071 advance = 1;
1072 item = btrfs_item_nr(leaf, slot);
1073 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1074
1075 if (found_key.objectid != key.objectid)
1076 break;
1077 if (btrfs_key_type(&found_key) != key_type)
1078 break;
1079 if (found_key.offset < filp->f_pos)
1080 continue;
1081
1082 filp->f_pos = found_key.offset;
1083 advance = 1;
1084 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1085 di_cur = 0;
1086 di_total = btrfs_item_size(leaf, item);
1087 while(di_cur < di_total) {
1088 struct btrfs_key location;
1089
1090 name_len = btrfs_dir_name_len(leaf, di);
1091 if (name_len < 32) {
1092 name_ptr = tmp_name;
1093 } else {
1094 name_ptr = kmalloc(name_len, GFP_NOFS);
1095 BUG_ON(!name_ptr);
1096 }
1097 read_extent_buffer(leaf, name_ptr,
1098 (unsigned long)(di + 1), name_len);
1099
1100 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1101 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1102
1103 over = filldir(dirent, name_ptr, name_len,
1104 found_key.offset,
1105 location.objectid,
1106 d_type);
1107
1108 if (name_ptr != tmp_name)
1109 kfree(name_ptr);
1110
1111 if (over)
1112 goto nopos;
1113 di_len = btrfs_dir_name_len(leaf, di) + sizeof(*di);
1114 di_cur += di_len;
1115 di = (struct btrfs_dir_item *)((char *)di + di_len);
1116 }
1117 }
1118 filp->f_pos++;
1119 nopos:
1120 ret = 0;
1121 err:
1122 btrfs_release_path(root, path);
1123 btrfs_free_path(path);
1124 mutex_unlock(&root->fs_info->fs_mutex);
1125 return ret;
1126 }
1127
1128 int btrfs_write_inode(struct inode *inode, int wait)
1129 {
1130 struct btrfs_root *root = BTRFS_I(inode)->root;
1131 struct btrfs_trans_handle *trans;
1132 int ret = 0;
1133
1134 if (wait) {
1135 mutex_lock(&root->fs_info->fs_mutex);
1136 trans = btrfs_start_transaction(root, 1);
1137 btrfs_set_trans_block_group(trans, inode);
1138 ret = btrfs_commit_transaction(trans, root);
1139 mutex_unlock(&root->fs_info->fs_mutex);
1140 }
1141 return ret;
1142 }
1143
1144 /*
1145 * This is somewhat expensive, updating the tree every time the
1146 * inode changes. But, it is most likely to find the inode in cache.
1147 * FIXME, needs more benchmarking...there are no reasons other than performance
1148 * to keep or drop this code.
1149 */
1150 void btrfs_dirty_inode(struct inode *inode)
1151 {
1152 struct btrfs_root *root = BTRFS_I(inode)->root;
1153 struct btrfs_trans_handle *trans;
1154
1155 mutex_lock(&root->fs_info->fs_mutex);
1156 trans = btrfs_start_transaction(root, 1);
1157 btrfs_set_trans_block_group(trans, inode);
1158 btrfs_update_inode(trans, root, inode);
1159 btrfs_end_transaction(trans, root);
1160 mutex_unlock(&root->fs_info->fs_mutex);
1161 }
1162
1163 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1164 struct btrfs_root *root,
1165 u64 objectid,
1166 struct btrfs_block_group_cache *group,
1167 int mode)
1168 {
1169 struct inode *inode;
1170 struct btrfs_inode_item *inode_item;
1171 struct btrfs_key *location;
1172 struct btrfs_path *path;
1173 int ret;
1174 int owner;
1175
1176 path = btrfs_alloc_path();
1177 BUG_ON(!path);
1178
1179 inode = new_inode(root->fs_info->sb);
1180 if (!inode)
1181 return ERR_PTR(-ENOMEM);
1182
1183 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1184 inode->i_mapping, GFP_NOFS);
1185 BTRFS_I(inode)->root = root;
1186
1187 if (mode & S_IFDIR)
1188 owner = 0;
1189 else
1190 owner = 1;
1191 group = btrfs_find_block_group(root, group, 0, 0, owner);
1192 BTRFS_I(inode)->block_group = group;
1193
1194 ret = btrfs_insert_empty_inode(trans, root, path, objectid);
1195 if (ret)
1196 goto fail;
1197
1198 inode->i_uid = current->fsuid;
1199 inode->i_gid = current->fsgid;
1200 inode->i_mode = mode;
1201 inode->i_ino = objectid;
1202 inode->i_blocks = 0;
1203 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1204 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1205 struct btrfs_inode_item);
1206 fill_inode_item(path->nodes[0], inode_item, inode);
1207 btrfs_mark_buffer_dirty(path->nodes[0]);
1208 btrfs_free_path(path);
1209
1210 location = &BTRFS_I(inode)->location;
1211 location->objectid = objectid;
1212 location->offset = 0;
1213 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1214
1215 insert_inode_hash(inode);
1216 return inode;
1217 fail:
1218 btrfs_free_path(path);
1219 return ERR_PTR(ret);
1220 }
1221
1222 static inline u8 btrfs_inode_type(struct inode *inode)
1223 {
1224 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1225 }
1226
1227 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1228 struct dentry *dentry, struct inode *inode)
1229 {
1230 int ret;
1231 struct btrfs_key key;
1232 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1233 struct inode *parent_inode;
1234
1235 key.objectid = inode->i_ino;
1236 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1237 key.offset = 0;
1238
1239 ret = btrfs_insert_dir_item(trans, root,
1240 dentry->d_name.name, dentry->d_name.len,
1241 dentry->d_parent->d_inode->i_ino,
1242 &key, btrfs_inode_type(inode));
1243 if (ret == 0) {
1244 parent_inode = dentry->d_parent->d_inode;
1245 parent_inode->i_size += dentry->d_name.len * 2;
1246 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1247 ret = btrfs_update_inode(trans, root,
1248 dentry->d_parent->d_inode);
1249 }
1250 return ret;
1251 }
1252
1253 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1254 struct dentry *dentry, struct inode *inode)
1255 {
1256 int err = btrfs_add_link(trans, dentry, inode);
1257 if (!err) {
1258 d_instantiate(dentry, inode);
1259 return 0;
1260 }
1261 if (err > 0)
1262 err = -EEXIST;
1263 return err;
1264 }
1265
1266 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1267 int mode, dev_t rdev)
1268 {
1269 struct btrfs_trans_handle *trans;
1270 struct btrfs_root *root = BTRFS_I(dir)->root;
1271 struct inode *inode;
1272 int err;
1273 int drop_inode = 0;
1274 u64 objectid;
1275 unsigned long nr;
1276
1277 if (!new_valid_dev(rdev))
1278 return -EINVAL;
1279
1280 mutex_lock(&root->fs_info->fs_mutex);
1281 trans = btrfs_start_transaction(root, 1);
1282 btrfs_set_trans_block_group(trans, dir);
1283
1284 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1285 if (err) {
1286 err = -ENOSPC;
1287 goto out_unlock;
1288 }
1289
1290 inode = btrfs_new_inode(trans, root, objectid,
1291 BTRFS_I(dir)->block_group, mode);
1292 err = PTR_ERR(inode);
1293 if (IS_ERR(inode))
1294 goto out_unlock;
1295
1296 btrfs_set_trans_block_group(trans, inode);
1297 err = btrfs_add_nondir(trans, dentry, inode);
1298 if (err)
1299 drop_inode = 1;
1300 else {
1301 inode->i_op = &btrfs_special_inode_operations;
1302 init_special_inode(inode, inode->i_mode, rdev);
1303 btrfs_update_inode(trans, root, inode);
1304 }
1305 dir->i_sb->s_dirt = 1;
1306 btrfs_update_inode_block_group(trans, inode);
1307 btrfs_update_inode_block_group(trans, dir);
1308 out_unlock:
1309 nr = trans->blocks_used;
1310 btrfs_end_transaction(trans, root);
1311 mutex_unlock(&root->fs_info->fs_mutex);
1312
1313 if (drop_inode) {
1314 inode_dec_link_count(inode);
1315 iput(inode);
1316 }
1317 btrfs_btree_balance_dirty(root, nr);
1318 return err;
1319 }
1320
1321 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1322 int mode, struct nameidata *nd)
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 unsigned long nr;
1330 u64 objectid;
1331
1332 mutex_lock(&root->fs_info->fs_mutex);
1333 trans = btrfs_start_transaction(root, 1);
1334 btrfs_set_trans_block_group(trans, dir);
1335
1336 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1337 if (err) {
1338 err = -ENOSPC;
1339 goto out_unlock;
1340 }
1341
1342 inode = btrfs_new_inode(trans, root, objectid,
1343 BTRFS_I(dir)->block_group, mode);
1344 err = PTR_ERR(inode);
1345 if (IS_ERR(inode))
1346 goto out_unlock;
1347
1348 btrfs_set_trans_block_group(trans, inode);
1349 err = btrfs_add_nondir(trans, dentry, inode);
1350 if (err)
1351 drop_inode = 1;
1352 else {
1353 inode->i_mapping->a_ops = &btrfs_aops;
1354 inode->i_fop = &btrfs_file_operations;
1355 inode->i_op = &btrfs_file_inode_operations;
1356 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1357 inode->i_mapping, GFP_NOFS);
1358 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
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_link(struct dentry *old_dentry, struct inode *dir,
1377 struct dentry *dentry)
1378 {
1379 struct btrfs_trans_handle *trans;
1380 struct btrfs_root *root = BTRFS_I(dir)->root;
1381 struct inode *inode = old_dentry->d_inode;
1382 unsigned long nr;
1383 int err;
1384 int drop_inode = 0;
1385
1386 if (inode->i_nlink == 0)
1387 return -ENOENT;
1388
1389 inc_nlink(inode);
1390 mutex_lock(&root->fs_info->fs_mutex);
1391 trans = btrfs_start_transaction(root, 1);
1392
1393 btrfs_set_trans_block_group(trans, dir);
1394 atomic_inc(&inode->i_count);
1395 err = btrfs_add_nondir(trans, dentry, inode);
1396
1397 if (err)
1398 drop_inode = 1;
1399
1400 dir->i_sb->s_dirt = 1;
1401 btrfs_update_inode_block_group(trans, dir);
1402 err = btrfs_update_inode(trans, root, inode);
1403
1404 if (err)
1405 drop_inode = 1;
1406
1407 nr = trans->blocks_used;
1408 btrfs_end_transaction(trans, root);
1409 mutex_unlock(&root->fs_info->fs_mutex);
1410
1411 if (drop_inode) {
1412 inode_dec_link_count(inode);
1413 iput(inode);
1414 }
1415 btrfs_btree_balance_dirty(root, nr);
1416 return err;
1417 }
1418
1419 static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
1420 struct btrfs_root *root,
1421 u64 objectid, u64 dirid)
1422 {
1423 int ret;
1424 char buf[2];
1425 struct btrfs_key key;
1426
1427 buf[0] = '.';
1428 buf[1] = '.';
1429
1430 key.objectid = objectid;
1431 key.offset = 0;
1432 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1433
1434 ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid,
1435 &key, BTRFS_FT_DIR);
1436 if (ret)
1437 goto error;
1438
1439 key.objectid = dirid;
1440 ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid,
1441 &key, BTRFS_FT_DIR);
1442 if (ret)
1443 goto error;
1444 error:
1445 return ret;
1446 }
1447
1448 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1449 {
1450 struct inode *inode;
1451 struct btrfs_trans_handle *trans;
1452 struct btrfs_root *root = BTRFS_I(dir)->root;
1453 int err = 0;
1454 int drop_on_err = 0;
1455 u64 objectid;
1456 unsigned long nr = 1;
1457
1458 mutex_lock(&root->fs_info->fs_mutex);
1459 trans = btrfs_start_transaction(root, 1);
1460 btrfs_set_trans_block_group(trans, dir);
1461
1462 if (IS_ERR(trans)) {
1463 err = PTR_ERR(trans);
1464 goto out_unlock;
1465 }
1466
1467 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1468 if (err) {
1469 err = -ENOSPC;
1470 goto out_unlock;
1471 }
1472
1473 inode = btrfs_new_inode(trans, root, objectid,
1474 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1475 if (IS_ERR(inode)) {
1476 err = PTR_ERR(inode);
1477 goto out_fail;
1478 }
1479
1480 drop_on_err = 1;
1481 inode->i_op = &btrfs_dir_inode_operations;
1482 inode->i_fop = &btrfs_dir_file_operations;
1483 btrfs_set_trans_block_group(trans, inode);
1484
1485 err = btrfs_make_empty_dir(trans, root, inode->i_ino, dir->i_ino);
1486 if (err)
1487 goto out_fail;
1488
1489 inode->i_size = 6;
1490 err = btrfs_update_inode(trans, root, inode);
1491 if (err)
1492 goto out_fail;
1493
1494 err = btrfs_add_link(trans, dentry, inode);
1495 if (err)
1496 goto out_fail;
1497
1498 d_instantiate(dentry, inode);
1499 drop_on_err = 0;
1500 dir->i_sb->s_dirt = 1;
1501 btrfs_update_inode_block_group(trans, inode);
1502 btrfs_update_inode_block_group(trans, dir);
1503
1504 out_fail:
1505 nr = trans->blocks_used;
1506 btrfs_end_transaction(trans, root);
1507
1508 out_unlock:
1509 mutex_unlock(&root->fs_info->fs_mutex);
1510 if (drop_on_err)
1511 iput(inode);
1512 btrfs_btree_balance_dirty(root, nr);
1513 return err;
1514 }
1515
1516 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
1517 size_t page_offset, u64 start, u64 end,
1518 int create)
1519 {
1520 int ret;
1521 int err = 0;
1522 u64 bytenr;
1523 u64 extent_start = 0;
1524 u64 extent_end = 0;
1525 u64 objectid = inode->i_ino;
1526 u32 found_type;
1527 int failed_insert = 0;
1528 struct btrfs_path *path;
1529 struct btrfs_root *root = BTRFS_I(inode)->root;
1530 struct btrfs_file_extent_item *item;
1531 struct extent_buffer *leaf;
1532 struct btrfs_key found_key;
1533 struct extent_map *em = NULL;
1534 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1535 struct btrfs_trans_handle *trans = NULL;
1536
1537 path = btrfs_alloc_path();
1538 BUG_ON(!path);
1539 mutex_lock(&root->fs_info->fs_mutex);
1540
1541 again:
1542 em = lookup_extent_mapping(em_tree, start, end);
1543 if (em) {
1544 goto out;
1545 }
1546 if (!em) {
1547 em = alloc_extent_map(GFP_NOFS);
1548 if (!em) {
1549 err = -ENOMEM;
1550 goto out;
1551 }
1552 em->start = EXTENT_MAP_HOLE;
1553 em->end = EXTENT_MAP_HOLE;
1554 }
1555 em->bdev = inode->i_sb->s_bdev;
1556 ret = btrfs_lookup_file_extent(trans, root, path,
1557 objectid, start, trans != NULL);
1558 if (ret < 0) {
1559 err = ret;
1560 goto out;
1561 }
1562
1563 if (ret != 0) {
1564 if (path->slots[0] == 0)
1565 goto not_found;
1566 path->slots[0]--;
1567 }
1568
1569 leaf = path->nodes[0];
1570 item = btrfs_item_ptr(leaf, path->slots[0],
1571 struct btrfs_file_extent_item);
1572 /* are we inside the extent that was found? */
1573 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1574 found_type = btrfs_key_type(&found_key);
1575 if (found_key.objectid != objectid ||
1576 found_type != BTRFS_EXTENT_DATA_KEY) {
1577 goto not_found;
1578 }
1579
1580 found_type = btrfs_file_extent_type(leaf, item);
1581 extent_start = found_key.offset;
1582 if (found_type == BTRFS_FILE_EXTENT_REG) {
1583 extent_end = extent_start +
1584 btrfs_file_extent_num_bytes(leaf, item);
1585 err = 0;
1586 if (start < extent_start || start >= extent_end) {
1587 em->start = start;
1588 if (start < extent_start) {
1589 if (end < extent_start)
1590 goto not_found;
1591 em->end = extent_end - 1;
1592 } else {
1593 em->end = end;
1594 }
1595 goto not_found_em;
1596 }
1597 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
1598 if (bytenr == 0) {
1599 em->start = extent_start;
1600 em->end = extent_end - 1;
1601 em->block_start = EXTENT_MAP_HOLE;
1602 em->block_end = EXTENT_MAP_HOLE;
1603 goto insert;
1604 }
1605 bytenr += btrfs_file_extent_offset(leaf, item);
1606 em->block_start = bytenr;
1607 em->block_end = em->block_start +
1608 btrfs_file_extent_num_bytes(leaf, item) - 1;
1609 em->start = extent_start;
1610 em->end = extent_end - 1;
1611 goto insert;
1612 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1613 unsigned long ptr;
1614 char *map;
1615 size_t size;
1616 size_t extent_offset;
1617 size_t copy_size;
1618
1619 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
1620 path->slots[0]));
1621 extent_end = (extent_start + size - 1) |
1622 ((u64)root->sectorsize - 1);
1623 if (start < extent_start || start >= extent_end) {
1624 em->start = start;
1625 if (start < extent_start) {
1626 if (end < extent_start)
1627 goto not_found;
1628 em->end = extent_end;
1629 } else {
1630 em->end = end;
1631 }
1632 goto not_found_em;
1633 }
1634 em->block_start = EXTENT_MAP_INLINE;
1635 em->block_end = EXTENT_MAP_INLINE;
1636
1637 if (!page) {
1638 em->start = extent_start;
1639 em->end = extent_start + size - 1;
1640 goto out;
1641 }
1642
1643 extent_offset = ((u64)page->index << PAGE_CACHE_SHIFT) -
1644 extent_start + page_offset;
1645 copy_size = min_t(u64, PAGE_CACHE_SIZE - page_offset,
1646 size - extent_offset);
1647 em->start = extent_start + extent_offset;
1648 em->end = (em->start + copy_size -1) |
1649 ((u64)root->sectorsize -1);
1650 map = kmap(page);
1651 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
1652 if (create == 0 && !PageUptodate(page)) {
1653 read_extent_buffer(leaf, map + page_offset, ptr,
1654 copy_size);
1655 flush_dcache_page(page);
1656 } else if (create && PageUptodate(page)) {
1657 if (!trans) {
1658 kunmap(page);
1659 free_extent_map(em);
1660 em = NULL;
1661 btrfs_release_path(root, path);
1662 trans = btrfs_start_transaction(root, 1);
1663 goto again;
1664 }
1665 write_extent_buffer(leaf, map + page_offset, ptr,
1666 copy_size);
1667 btrfs_mark_buffer_dirty(leaf);
1668 }
1669 kunmap(page);
1670 set_extent_uptodate(em_tree, em->start, em->end, GFP_NOFS);
1671 goto insert;
1672 } else {
1673 printk("unkknown found_type %d\n", found_type);
1674 WARN_ON(1);
1675 }
1676 not_found:
1677 em->start = start;
1678 em->end = end;
1679 not_found_em:
1680 em->block_start = EXTENT_MAP_HOLE;
1681 em->block_end = EXTENT_MAP_HOLE;
1682 insert:
1683 btrfs_release_path(root, path);
1684 if (em->start > start || em->end < start) {
1685 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->end, start, end);
1686 err = -EIO;
1687 goto out;
1688 }
1689 ret = add_extent_mapping(em_tree, em);
1690 if (ret == -EEXIST) {
1691 free_extent_map(em);
1692 em = NULL;
1693 failed_insert++;
1694 if (failed_insert > 5) {
1695 printk("failing to insert %Lu %Lu\n", start, end);
1696 err = -EIO;
1697 goto out;
1698 }
1699 goto again;
1700 }
1701 err = 0;
1702 out:
1703 btrfs_free_path(path);
1704 if (trans) {
1705 ret = btrfs_end_transaction(trans, root);
1706 if (!err)
1707 err = ret;
1708 }
1709 mutex_unlock(&root->fs_info->fs_mutex);
1710 if (err) {
1711 free_extent_map(em);
1712 WARN_ON(1);
1713 return ERR_PTR(err);
1714 }
1715 return em;
1716 }
1717
1718 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
1719 {
1720 return extent_bmap(mapping, iblock, btrfs_get_extent);
1721 }
1722
1723 static int btrfs_prepare_write(struct file *file, struct page *page,
1724 unsigned from, unsigned to)
1725 {
1726 return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree,
1727 page->mapping->host, page, from, to,
1728 btrfs_get_extent);
1729 }
1730
1731 int btrfs_readpage(struct file *file, struct page *page)
1732 {
1733 struct extent_map_tree *tree;
1734 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1735 return extent_read_full_page(tree, page, btrfs_get_extent);
1736 }
1737 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
1738 {
1739 struct extent_map_tree *tree;
1740
1741
1742 if (current->flags & PF_MEMALLOC) {
1743 redirty_page_for_writepage(wbc, page);
1744 unlock_page(page);
1745 return 0;
1746 }
1747 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1748 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
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
1760 btrfs_readpages(struct file *file, struct address_space *mapping,
1761 struct list_head *pages, unsigned nr_pages)
1762 {
1763 struct extent_map_tree *tree;
1764 tree = &BTRFS_I(mapping->host)->extent_tree;
1765 return extent_readpages(tree, mapping, pages, nr_pages,
1766 btrfs_get_extent);
1767 }
1768
1769 static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
1770 {
1771 struct extent_map_tree *tree;
1772 int ret;
1773
1774 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1775 ret = try_release_extent_mapping(tree, page);
1776 if (ret == 1) {
1777 ClearPagePrivate(page);
1778 set_page_private(page, 0);
1779 page_cache_release(page);
1780 }
1781 return ret;
1782 }
1783
1784 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
1785 {
1786 struct extent_map_tree *tree;
1787
1788 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1789 extent_invalidatepage(tree, page, offset);
1790 btrfs_releasepage(page, GFP_NOFS);
1791 }
1792
1793 /*
1794 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
1795 * called from a page fault handler when a page is first dirtied. Hence we must
1796 * be careful to check for EOF conditions here. We set the page up correctly
1797 * for a written page which means we get ENOSPC checking when writing into
1798 * holes and correct delalloc and unwritten extent mapping on filesystems that
1799 * support these features.
1800 *
1801 * We are not allowed to take the i_mutex here so we have to play games to
1802 * protect against truncate races as the page could now be beyond EOF. Because
1803 * vmtruncate() writes the inode size before removing pages, once we have the
1804 * page lock we can determine safely if the page is beyond EOF. If it is not
1805 * beyond EOF, then the page is guaranteed safe against truncation until we
1806 * unlock the page.
1807 */
1808 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1809 {
1810 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
1811 unsigned long end;
1812 loff_t size;
1813 int ret = -EINVAL;
1814 u64 page_start;
1815
1816 down_read(&BTRFS_I(inode)->root->snap_sem);
1817 lock_page(page);
1818 wait_on_page_writeback(page);
1819 size = i_size_read(inode);
1820 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
1821
1822 if ((page->mapping != inode->i_mapping) ||
1823 (page_start > size)) {
1824 /* page got truncated out from underneath us */
1825 goto out_unlock;
1826 }
1827
1828 /* page is wholly or partially inside EOF */
1829 if (page_start + PAGE_CACHE_SIZE > size)
1830 end = size & ~PAGE_CACHE_MASK;
1831 else
1832 end = PAGE_CACHE_SIZE;
1833
1834 ret = btrfs_cow_one_page(inode, page, end);
1835
1836 out_unlock:
1837 up_read(&BTRFS_I(inode)->root->snap_sem);
1838 unlock_page(page);
1839 return ret;
1840 }
1841
1842 static void btrfs_truncate(struct inode *inode)
1843 {
1844 struct btrfs_root *root = BTRFS_I(inode)->root;
1845 int ret;
1846 struct btrfs_trans_handle *trans;
1847 unsigned long nr;
1848
1849 if (!S_ISREG(inode->i_mode))
1850 return;
1851 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1852 return;
1853
1854 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1855
1856 mutex_lock(&root->fs_info->fs_mutex);
1857 trans = btrfs_start_transaction(root, 1);
1858 btrfs_set_trans_block_group(trans, inode);
1859
1860 /* FIXME, add redo link to tree so we don't leak on crash */
1861 ret = btrfs_truncate_in_trans(trans, root, inode);
1862 btrfs_update_inode(trans, root, inode);
1863 nr = trans->blocks_used;
1864
1865 ret = btrfs_end_transaction(trans, root);
1866 BUG_ON(ret);
1867 mutex_unlock(&root->fs_info->fs_mutex);
1868 btrfs_btree_balance_dirty(root, nr);
1869 }
1870
1871 int btrfs_commit_write(struct file *file, struct page *page,
1872 unsigned from, unsigned to)
1873 {
1874 return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree,
1875 page->mapping->host, page, from, to);
1876 }
1877
1878 static int create_subvol(struct btrfs_root *root, char *name, int namelen)
1879 {
1880 struct btrfs_trans_handle *trans;
1881 struct btrfs_key key;
1882 struct btrfs_root_item root_item;
1883 struct btrfs_inode_item *inode_item;
1884 struct extent_buffer *leaf;
1885 struct btrfs_root *new_root;
1886 struct inode *inode;
1887 struct inode *dir;
1888 int ret;
1889 int err;
1890 u64 objectid;
1891 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
1892 unsigned long nr = 1;
1893
1894 mutex_lock(&root->fs_info->fs_mutex);
1895 trans = btrfs_start_transaction(root, 1);
1896 BUG_ON(!trans);
1897
1898 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, 0);
1899 if (IS_ERR(leaf))
1900 return PTR_ERR(leaf);
1901
1902 btrfs_set_header_nritems(leaf, 0);
1903 btrfs_set_header_level(leaf, 0);
1904 btrfs_set_header_bytenr(leaf, leaf->start);
1905 btrfs_set_header_generation(leaf, trans->transid);
1906 btrfs_set_header_owner(leaf, root->root_key.objectid);
1907 write_extent_buffer(leaf, root->fs_info->fsid,
1908 (unsigned long)btrfs_header_fsid(leaf),
1909 BTRFS_FSID_SIZE);
1910 btrfs_mark_buffer_dirty(leaf);
1911
1912 inode_item = &root_item.inode;
1913 memset(inode_item, 0, sizeof(*inode_item));
1914 inode_item->generation = cpu_to_le64(1);
1915 inode_item->size = cpu_to_le64(3);
1916 inode_item->nlink = cpu_to_le32(1);
1917 inode_item->nblocks = cpu_to_le64(1);
1918 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1919
1920 btrfs_set_root_bytenr(&root_item, leaf->start);
1921 btrfs_set_root_level(&root_item, 0);
1922 btrfs_set_root_refs(&root_item, 1);
1923 btrfs_set_root_used(&root_item, 0);
1924
1925 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
1926 root_item.drop_level = 0;
1927
1928 free_extent_buffer(leaf);
1929 leaf = NULL;
1930
1931 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1932 0, &objectid);
1933 if (ret)
1934 goto fail;
1935
1936 btrfs_set_root_dirid(&root_item, new_dirid);
1937
1938 key.objectid = objectid;
1939 key.offset = 1;
1940 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1941 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1942 &root_item);
1943 if (ret)
1944 goto fail;
1945
1946 /*
1947 * insert the directory item
1948 */
1949 key.offset = (u64)-1;
1950 dir = root->fs_info->sb->s_root->d_inode;
1951 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1952 name, namelen, dir->i_ino, &key,
1953 BTRFS_FT_DIR);
1954 if (ret)
1955 goto fail;
1956
1957 ret = btrfs_commit_transaction(trans, root);
1958 if (ret)
1959 goto fail_commit;
1960
1961 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
1962 BUG_ON(!new_root);
1963
1964 trans = btrfs_start_transaction(new_root, 1);
1965 BUG_ON(!trans);
1966
1967 inode = btrfs_new_inode(trans, new_root, new_dirid,
1968 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
1969 if (IS_ERR(inode))
1970 goto fail;
1971 inode->i_op = &btrfs_dir_inode_operations;
1972 inode->i_fop = &btrfs_dir_file_operations;
1973 new_root->inode = inode;
1974
1975 ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid);
1976 if (ret)
1977 goto fail;
1978
1979 inode->i_nlink = 1;
1980 inode->i_size = 6;
1981 ret = btrfs_update_inode(trans, new_root, inode);
1982 if (ret)
1983 goto fail;
1984 fail:
1985 nr = trans->blocks_used;
1986 err = btrfs_commit_transaction(trans, root);
1987 if (err && !ret)
1988 ret = err;
1989 fail_commit:
1990 mutex_unlock(&root->fs_info->fs_mutex);
1991 btrfs_btree_balance_dirty(root, nr);
1992 return ret;
1993 }
1994
1995 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
1996 {
1997 struct btrfs_trans_handle *trans;
1998 struct btrfs_key key;
1999 struct btrfs_root_item new_root_item;
2000 struct extent_buffer *tmp;
2001 int ret;
2002 int err;
2003 u64 objectid;
2004 unsigned long nr;
2005
2006 if (!root->ref_cows)
2007 return -EINVAL;
2008
2009 down_write(&root->snap_sem);
2010 freeze_bdev(root->fs_info->sb->s_bdev);
2011 thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb);
2012
2013 mutex_lock(&root->fs_info->fs_mutex);
2014 trans = btrfs_start_transaction(root, 1);
2015 BUG_ON(!trans);
2016
2017 ret = btrfs_update_inode(trans, root, root->inode);
2018 if (ret)
2019 goto fail;
2020
2021 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
2022 0, &objectid);
2023 if (ret)
2024 goto fail;
2025
2026 memcpy(&new_root_item, &root->root_item,
2027 sizeof(new_root_item));
2028
2029 key.objectid = objectid;
2030 key.offset = 1;
2031 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
2032
2033 btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
2034 btrfs_set_root_bytenr(&new_root_item, root->node->start);
2035 btrfs_set_root_level(&new_root_item, btrfs_header_level(root->node));
2036
2037 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
2038 &new_root_item);
2039 if (ret)
2040 goto fail;
2041
2042 /*
2043 * insert the directory item
2044 */
2045 key.offset = (u64)-1;
2046 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
2047 name, namelen,
2048 root->fs_info->sb->s_root->d_inode->i_ino,
2049 &key, BTRFS_FT_DIR);
2050
2051 if (ret)
2052 goto fail;
2053
2054 ret = btrfs_inc_root_ref(trans, root);
2055 if (ret)
2056 goto fail;
2057 fail:
2058 nr = trans->blocks_used;
2059 err = btrfs_commit_transaction(trans, root);
2060
2061 if (err && !ret)
2062 ret = err;
2063
2064 mutex_unlock(&root->fs_info->fs_mutex);
2065 up_write(&root->snap_sem);
2066 btrfs_btree_balance_dirty(root, nr);
2067 return ret;
2068 }
2069
2070 static unsigned long force_ra(struct address_space *mapping,
2071 struct file_ra_state *ra, struct file *file,
2072 pgoff_t offset, pgoff_t last_index)
2073 {
2074 pgoff_t req_size;
2075
2076 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2077 req_size = last_index - offset + 1;
2078 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2079 return offset;
2080 #else
2081 req_size = min(last_index - offset + 1, (pgoff_t)128);
2082 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2083 return offset + req_size;
2084 #endif
2085 }
2086
2087 int btrfs_defrag_file(struct file *file) {
2088 struct inode *inode = file->f_path.dentry->d_inode;
2089 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2090 struct page *page;
2091 unsigned long last_index;
2092 unsigned long ra_index = 0;
2093 u64 page_start;
2094 u64 page_end;
2095 unsigned long i;
2096
2097 mutex_lock(&inode->i_mutex);
2098 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
2099 for (i = 0; i <= last_index; i++) {
2100 if (i == ra_index) {
2101 ra_index = force_ra(inode->i_mapping, &file->f_ra,
2102 file, ra_index, last_index);
2103 }
2104 page = grab_cache_page(inode->i_mapping, i);
2105 if (!page)
2106 goto out_unlock;
2107 if (!PageUptodate(page)) {
2108 btrfs_readpage(NULL, page);
2109 lock_page(page);
2110 if (!PageUptodate(page)) {
2111 unlock_page(page);
2112 page_cache_release(page);
2113 goto out_unlock;
2114 }
2115 }
2116 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2117 page_end = page_start + PAGE_CACHE_SIZE - 1;
2118
2119 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
2120 set_extent_delalloc(em_tree, page_start,
2121 page_end, GFP_NOFS);
2122 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
2123 set_page_dirty(page);
2124 unlock_page(page);
2125 page_cache_release(page);
2126 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
2127 }
2128
2129 out_unlock:
2130 mutex_unlock(&inode->i_mutex);
2131 return 0;
2132 }
2133
2134 static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg)
2135 {
2136 struct btrfs_ioctl_vol_args vol_args;
2137 struct btrfs_dir_item *di;
2138 struct btrfs_path *path;
2139 int namelen;
2140 u64 root_dirid;
2141
2142 if (copy_from_user(&vol_args, arg, sizeof(vol_args)))
2143 return -EFAULT;
2144
2145 namelen = strlen(vol_args.name);
2146 if (namelen > BTRFS_VOL_NAME_MAX)
2147 return -EINVAL;
2148 if (strchr(vol_args.name, '/'))
2149 return -EINVAL;
2150
2151 path = btrfs_alloc_path();
2152 if (!path)
2153 return -ENOMEM;
2154
2155 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2156 mutex_lock(&root->fs_info->fs_mutex);
2157 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2158 path, root_dirid,
2159 vol_args.name, namelen, 0);
2160 mutex_unlock(&root->fs_info->fs_mutex);
2161 btrfs_free_path(path);
2162 if (di && !IS_ERR(di))
2163 return -EEXIST;
2164 if (IS_ERR(di))
2165 return PTR_ERR(di);
2166
2167 if (root == root->fs_info->tree_root)
2168 return create_subvol(root, vol_args.name, namelen);
2169 return create_snapshot(root, vol_args.name, namelen);
2170 }
2171
2172 static int btrfs_ioctl_defrag(struct file *file)
2173 {
2174 struct inode *inode = file->f_path.dentry->d_inode;
2175 struct btrfs_root *root = BTRFS_I(inode)->root;
2176
2177 switch (inode->i_mode & S_IFMT) {
2178 case S_IFDIR:
2179 mutex_lock(&root->fs_info->fs_mutex);
2180 btrfs_defrag_root(root, 0);
2181 btrfs_defrag_root(root->fs_info->extent_root, 0);
2182 mutex_unlock(&root->fs_info->fs_mutex);
2183 break;
2184 case S_IFREG:
2185 btrfs_defrag_file(file);
2186 break;
2187 }
2188
2189 return 0;
2190 }
2191
2192 long btrfs_ioctl(struct file *file, unsigned int
2193 cmd, unsigned long arg)
2194 {
2195 struct btrfs_root *root = BTRFS_I(file->f_path.dentry->d_inode)->root;
2196
2197 switch (cmd) {
2198 case BTRFS_IOC_SNAP_CREATE:
2199 return btrfs_ioctl_snap_create(root, (void __user *)arg);
2200 case BTRFS_IOC_DEFRAG:
2201 return btrfs_ioctl_defrag(file);
2202 }
2203
2204 return -ENOTTY;
2205 }
2206
2207 /*
2208 * Called inside transaction, so use GFP_NOFS
2209 */
2210 struct inode *btrfs_alloc_inode(struct super_block *sb)
2211 {
2212 struct btrfs_inode *ei;
2213
2214 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2215 if (!ei)
2216 return NULL;
2217 ei->last_trans = 0;
2218 return &ei->vfs_inode;
2219 }
2220
2221 void btrfs_destroy_inode(struct inode *inode)
2222 {
2223 WARN_ON(!list_empty(&inode->i_dentry));
2224 WARN_ON(inode->i_data.nrpages);
2225
2226 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2227 }
2228
2229 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2230 static void init_once(struct kmem_cache * cachep, void *foo)
2231 #else
2232 static void init_once(void * foo, struct kmem_cache * cachep,
2233 unsigned long flags)
2234 #endif
2235 {
2236 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2237
2238 inode_init_once(&ei->vfs_inode);
2239 }
2240
2241 void btrfs_destroy_cachep(void)
2242 {
2243 if (btrfs_inode_cachep)
2244 kmem_cache_destroy(btrfs_inode_cachep);
2245 if (btrfs_trans_handle_cachep)
2246 kmem_cache_destroy(btrfs_trans_handle_cachep);
2247 if (btrfs_transaction_cachep)
2248 kmem_cache_destroy(btrfs_transaction_cachep);
2249 if (btrfs_bit_radix_cachep)
2250 kmem_cache_destroy(btrfs_bit_radix_cachep);
2251 if (btrfs_path_cachep)
2252 kmem_cache_destroy(btrfs_path_cachep);
2253 }
2254
2255 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2256 unsigned long extra_flags,
2257 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2258 void (*ctor)(struct kmem_cache *, void *)
2259 #else
2260 void (*ctor)(void *, struct kmem_cache *,
2261 unsigned long)
2262 #endif
2263 )
2264 {
2265 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2266 SLAB_MEM_SPREAD | extra_flags), ctor
2267 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2268 ,NULL
2269 #endif
2270 );
2271 }
2272
2273 int btrfs_init_cachep(void)
2274 {
2275 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2276 sizeof(struct btrfs_inode),
2277 0, init_once);
2278 if (!btrfs_inode_cachep)
2279 goto fail;
2280 btrfs_trans_handle_cachep =
2281 btrfs_cache_create("btrfs_trans_handle_cache",
2282 sizeof(struct btrfs_trans_handle),
2283 0, NULL);
2284 if (!btrfs_trans_handle_cachep)
2285 goto fail;
2286 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
2287 sizeof(struct btrfs_transaction),
2288 0, NULL);
2289 if (!btrfs_transaction_cachep)
2290 goto fail;
2291 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
2292 sizeof(struct btrfs_path),
2293 0, NULL);
2294 if (!btrfs_path_cachep)
2295 goto fail;
2296 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
2297 SLAB_DESTROY_BY_RCU, NULL);
2298 if (!btrfs_bit_radix_cachep)
2299 goto fail;
2300 return 0;
2301 fail:
2302 btrfs_destroy_cachep();
2303 return -ENOMEM;
2304 }
2305
2306 static int btrfs_getattr(struct vfsmount *mnt,
2307 struct dentry *dentry, struct kstat *stat)
2308 {
2309 struct inode *inode = dentry->d_inode;
2310 generic_fillattr(inode, stat);
2311 stat->blksize = 256 * 1024;
2312 return 0;
2313 }
2314
2315 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2316 struct inode * new_dir,struct dentry *new_dentry)
2317 {
2318 struct btrfs_trans_handle *trans;
2319 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2320 struct inode *new_inode = new_dentry->d_inode;
2321 struct inode *old_inode = old_dentry->d_inode;
2322 struct timespec ctime = CURRENT_TIME;
2323 struct btrfs_path *path;
2324 struct btrfs_dir_item *di;
2325 int ret;
2326
2327 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2328 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2329 return -ENOTEMPTY;
2330 }
2331
2332 mutex_lock(&root->fs_info->fs_mutex);
2333 trans = btrfs_start_transaction(root, 1);
2334
2335 btrfs_set_trans_block_group(trans, new_dir);
2336 path = btrfs_alloc_path();
2337 if (!path) {
2338 ret = -ENOMEM;
2339 goto out_fail;
2340 }
2341
2342 old_dentry->d_inode->i_nlink++;
2343 old_dir->i_ctime = old_dir->i_mtime = ctime;
2344 new_dir->i_ctime = new_dir->i_mtime = ctime;
2345 old_inode->i_ctime = ctime;
2346
2347 if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) {
2348 struct btrfs_key *location = &BTRFS_I(new_dir)->location;
2349 struct btrfs_key old_parent_key;
2350 di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino,
2351 "..", 2, -1);
2352 if (IS_ERR(di)) {
2353 ret = PTR_ERR(di);
2354 goto out_fail;
2355 }
2356 if (!di) {
2357 ret = -ENOENT;
2358 goto out_fail;
2359 }
2360 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &old_parent_key);
2361 ret = btrfs_del_item(trans, root, path);
2362 if (ret) {
2363 goto out_fail;
2364 }
2365 btrfs_release_path(root, path);
2366
2367 di = btrfs_lookup_dir_index_item(trans, root, path,
2368 old_inode->i_ino,
2369 old_parent_key.objectid,
2370 "..", 2, -1);
2371 if (IS_ERR(di)) {
2372 ret = PTR_ERR(di);
2373 goto out_fail;
2374 }
2375 if (!di) {
2376 ret = -ENOENT;
2377 goto out_fail;
2378 }
2379 ret = btrfs_del_item(trans, root, path);
2380 if (ret) {
2381 goto out_fail;
2382 }
2383 btrfs_release_path(root, path);
2384
2385 ret = btrfs_insert_dir_item(trans, root, "..", 2,
2386 old_inode->i_ino, location,
2387 BTRFS_FT_DIR);
2388 if (ret)
2389 goto out_fail;
2390 }
2391
2392
2393 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2394 if (ret)
2395 goto out_fail;
2396
2397 if (new_inode) {
2398 new_inode->i_ctime = CURRENT_TIME;
2399 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2400 if (ret)
2401 goto out_fail;
2402 }
2403 ret = btrfs_add_link(trans, new_dentry, old_inode);
2404 if (ret)
2405 goto out_fail;
2406
2407 out_fail:
2408 btrfs_free_path(path);
2409 btrfs_end_transaction(trans, root);
2410 mutex_unlock(&root->fs_info->fs_mutex);
2411 return ret;
2412 }
2413
2414 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2415 const char *symname)
2416 {
2417 struct btrfs_trans_handle *trans;
2418 struct btrfs_root *root = BTRFS_I(dir)->root;
2419 struct btrfs_path *path;
2420 struct btrfs_key key;
2421 struct inode *inode;
2422 int err;
2423 int drop_inode = 0;
2424 u64 objectid;
2425 int name_len;
2426 int datasize;
2427 unsigned long ptr;
2428 struct btrfs_file_extent_item *ei;
2429 struct extent_buffer *leaf;
2430 unsigned long nr;
2431
2432 name_len = strlen(symname) + 1;
2433 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2434 return -ENAMETOOLONG;
2435 mutex_lock(&root->fs_info->fs_mutex);
2436 trans = btrfs_start_transaction(root, 1);
2437 btrfs_set_trans_block_group(trans, dir);
2438
2439 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2440 if (err) {
2441 err = -ENOSPC;
2442 goto out_unlock;
2443 }
2444
2445 inode = btrfs_new_inode(trans, root, objectid,
2446 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2447 err = PTR_ERR(inode);
2448 if (IS_ERR(inode))
2449 goto out_unlock;
2450
2451 btrfs_set_trans_block_group(trans, inode);
2452 err = btrfs_add_nondir(trans, dentry, inode);
2453 if (err)
2454 drop_inode = 1;
2455 else {
2456 inode->i_mapping->a_ops = &btrfs_aops;
2457 inode->i_fop = &btrfs_file_operations;
2458 inode->i_op = &btrfs_file_inode_operations;
2459 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
2460 inode->i_mapping, GFP_NOFS);
2461 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
2462 }
2463 dir->i_sb->s_dirt = 1;
2464 btrfs_update_inode_block_group(trans, inode);
2465 btrfs_update_inode_block_group(trans, dir);
2466 if (drop_inode)
2467 goto out_unlock;
2468
2469 path = btrfs_alloc_path();
2470 BUG_ON(!path);
2471 key.objectid = inode->i_ino;
2472 key.offset = 0;
2473 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2474 datasize = btrfs_file_extent_calc_inline_size(name_len);
2475 err = btrfs_insert_empty_item(trans, root, path, &key,
2476 datasize);
2477 if (err) {
2478 drop_inode = 1;
2479 goto out_unlock;
2480 }
2481 leaf = path->nodes[0];
2482 ei = btrfs_item_ptr(leaf, path->slots[0],
2483 struct btrfs_file_extent_item);
2484 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
2485 btrfs_set_file_extent_type(leaf, ei,
2486 BTRFS_FILE_EXTENT_INLINE);
2487 ptr = btrfs_file_extent_inline_start(ei);
2488 write_extent_buffer(leaf, symname, ptr, name_len);
2489 btrfs_mark_buffer_dirty(leaf);
2490 btrfs_free_path(path);
2491
2492 inode->i_op = &btrfs_symlink_inode_operations;
2493 inode->i_mapping->a_ops = &btrfs_symlink_aops;
2494 inode->i_size = name_len - 1;
2495 err = btrfs_update_inode(trans, root, inode);
2496 if (err)
2497 drop_inode = 1;
2498
2499 out_unlock:
2500 nr = trans->blocks_used;
2501 btrfs_end_transaction(trans, root);
2502 mutex_unlock(&root->fs_info->fs_mutex);
2503 if (drop_inode) {
2504 inode_dec_link_count(inode);
2505 iput(inode);
2506 }
2507 btrfs_btree_balance_dirty(root, nr);
2508 return err;
2509 }
2510
2511 static struct inode_operations btrfs_dir_inode_operations = {
2512 .lookup = btrfs_lookup,
2513 .create = btrfs_create,
2514 .unlink = btrfs_unlink,
2515 .link = btrfs_link,
2516 .mkdir = btrfs_mkdir,
2517 .rmdir = btrfs_rmdir,
2518 .rename = btrfs_rename,
2519 .symlink = btrfs_symlink,
2520 .setattr = btrfs_setattr,
2521 .mknod = btrfs_mknod,
2522 };
2523
2524 static struct inode_operations btrfs_dir_ro_inode_operations = {
2525 .lookup = btrfs_lookup,
2526 };
2527
2528 static struct file_operations btrfs_dir_file_operations = {
2529 .llseek = generic_file_llseek,
2530 .read = generic_read_dir,
2531 .readdir = btrfs_readdir,
2532 .unlocked_ioctl = btrfs_ioctl,
2533 #ifdef CONFIG_COMPAT
2534 .compat_ioctl = btrfs_ioctl,
2535 #endif
2536 };
2537
2538 static struct extent_map_ops btrfs_extent_map_ops = {
2539 .fill_delalloc = run_delalloc_range,
2540 .writepage_io_hook = btrfs_writepage_io_hook,
2541 .readpage_io_hook = btrfs_readpage_io_hook,
2542 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
2543 };
2544
2545 static struct address_space_operations btrfs_aops = {
2546 .readpage = btrfs_readpage,
2547 .writepage = btrfs_writepage,
2548 .writepages = btrfs_writepages,
2549 .readpages = btrfs_readpages,
2550 .sync_page = block_sync_page,
2551 .prepare_write = btrfs_prepare_write,
2552 .commit_write = btrfs_commit_write,
2553 .bmap = btrfs_bmap,
2554 .invalidatepage = btrfs_invalidatepage,
2555 .releasepage = btrfs_releasepage,
2556 .set_page_dirty = __set_page_dirty_nobuffers,
2557 };
2558
2559 static struct address_space_operations btrfs_symlink_aops = {
2560 .readpage = btrfs_readpage,
2561 .writepage = btrfs_writepage,
2562 .invalidatepage = btrfs_invalidatepage,
2563 .releasepage = btrfs_releasepage,
2564 };
2565
2566 static struct inode_operations btrfs_file_inode_operations = {
2567 .truncate = btrfs_truncate,
2568 .getattr = btrfs_getattr,
2569 .setattr = btrfs_setattr,
2570 };
2571
2572 static struct inode_operations btrfs_special_inode_operations = {
2573 .getattr = btrfs_getattr,
2574 .setattr = btrfs_setattr,
2575 };
2576
2577 static struct inode_operations btrfs_symlink_inode_operations = {
2578 .readlink = generic_readlink,
2579 .follow_link = page_follow_link_light,
2580 .put_link = page_put_link,
2581 };
Linux kernel - Deliverables
This page took 0.099743 seconds and 6 git commands to generate.