]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/btrfs/inode.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: Use mount -o subvol to select the subvol directory instead of dev:
[net-next-2.6.git] / fs / btrfs / inode.c
CommitLineData
6cbd5570
CM		 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
39279cc3
CM		 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>
9ebefb18 33#include <linux/bit_spinlock.h>
92fee66d 34#include <linux/version.h>
39279cc3
CM		 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
42struct btrfs_iget_args {
43 u64 ino;
44 struct btrfs_root *root;
45};
46
47static struct inode_operations btrfs_dir_inode_operations;
48static struct inode_operations btrfs_symlink_inode_operations;
49static struct inode_operations btrfs_dir_ro_inode_operations;
618e21d5 50static struct inode_operations btrfs_special_inode_operations;
39279cc3
CM		 51static struct inode_operations btrfs_file_inode_operations;
52static struct address_space_operations btrfs_aops;
53static struct address_space_operations btrfs_symlink_aops;
54static struct file_operations btrfs_dir_file_operations;
55
56static struct kmem_cache *btrfs_inode_cachep;
57struct kmem_cache *btrfs_trans_handle_cachep;
58struct kmem_cache *btrfs_transaction_cachep;
59struct kmem_cache *btrfs_bit_radix_cachep;
60struct kmem_cache *btrfs_path_cachep;
61
62#define S_SHIFT 12
63static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
64 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
65 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
66 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
67 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
68 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
69 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
70 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
71};
72
b888db2b
CM		 73static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
74{
75 struct btrfs_root *root = BTRFS_I(inode)->root;
76 struct btrfs_trans_handle *trans;
77 struct btrfs_key ins;
78 u64 alloc_hint = 0;
79 u64 num_blocks;
80 int ret;
81 u64 blocksize = 1 << inode->i_blkbits;
82
83 mutex_lock(&root->fs_info->fs_mutex);
84 trans = btrfs_start_transaction(root, 1);
85 btrfs_set_trans_block_group(trans, inode);
86 BUG_ON(!trans);
87 num_blocks = (end - start + blocksize) & ~(blocksize - 1);
88 ret = btrfs_drop_extents(trans, root, inode,
89 start, start + num_blocks, &alloc_hint);
90 num_blocks = num_blocks >> inode->i_blkbits;
91 ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_blocks, 0,
92 alloc_hint, (u64)-1, &ins, 1);
93 if (ret) {
94 WARN_ON(1);
95 goto out;
96 }
97 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
98 start, ins.objectid, ins.offset,
99 ins.offset);
100out:
101 btrfs_end_transaction(trans, root);
102 mutex_unlock(&root->fs_info->fs_mutex);
103 return ret;
104}
105
106
39279cc3
CM		 107void btrfs_read_locked_inode(struct inode *inode)
108{
109 struct btrfs_path *path;
110 struct btrfs_inode_item *inode_item;
111 struct btrfs_root *root = BTRFS_I(inode)->root;
112 struct btrfs_key location;
113 u64 alloc_group_block;
618e21d5 114 u32 rdev;
39279cc3
CM		 115 int ret;
116
117 path = btrfs_alloc_path();
118 BUG_ON(!path);
39279cc3
CM		 119 mutex_lock(&root->fs_info->fs_mutex);
120
121 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
122 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
123 if (ret) {
124 btrfs_free_path(path);
125 goto make_bad;
126 }
127 inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
128 path->slots[0],
129 struct btrfs_inode_item);
130
131 inode->i_mode = btrfs_inode_mode(inode_item);
132 inode->i_nlink = btrfs_inode_nlink(inode_item);
133 inode->i_uid = btrfs_inode_uid(inode_item);
134 inode->i_gid = btrfs_inode_gid(inode_item);
135 inode->i_size = btrfs_inode_size(inode_item);
136 inode->i_atime.tv_sec = btrfs_timespec_sec(&inode_item->atime);
137 inode->i_atime.tv_nsec = btrfs_timespec_nsec(&inode_item->atime);
138 inode->i_mtime.tv_sec = btrfs_timespec_sec(&inode_item->mtime);
139 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(&inode_item->mtime);
140 inode->i_ctime.tv_sec = btrfs_timespec_sec(&inode_item->ctime);
141 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime);
142 inode->i_blocks = btrfs_inode_nblocks(inode_item);
143 inode->i_generation = btrfs_inode_generation(inode_item);
618e21d5
JB		 144 inode->i_rdev = 0;
145 rdev = btrfs_inode_rdev(inode_item);
39279cc3
CM		 146 alloc_group_block = btrfs_inode_block_group(inode_item);
147 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
148 alloc_group_block);
149
150 btrfs_free_path(path);
151 inode_item = NULL;
152
153 mutex_unlock(&root->fs_info->fs_mutex);
154
155 switch (inode->i_mode & S_IFMT) {
39279cc3
CM		 156 case S_IFREG:
157 inode->i_mapping->a_ops = &btrfs_aops;
b888db2b 158 BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range;
39279cc3
CM		 159 inode->i_fop = &btrfs_file_operations;
160 inode->i_op = &btrfs_file_inode_operations;
161 break;
162 case S_IFDIR:
163 inode->i_fop = &btrfs_dir_file_operations;
164 if (root == root->fs_info->tree_root)
165 inode->i_op = &btrfs_dir_ro_inode_operations;
166 else
167 inode->i_op = &btrfs_dir_inode_operations;
168 break;
169 case S_IFLNK:
170 inode->i_op = &btrfs_symlink_inode_operations;
171 inode->i_mapping->a_ops = &btrfs_symlink_aops;
172 break;
618e21d5
JB		 173 default:
174 init_special_inode(inode, inode->i_mode, rdev);
175 break;
39279cc3
CM		 176 }
177 return;
178
179make_bad:
180 btrfs_release_path(root, path);
181 btrfs_free_path(path);
182 mutex_unlock(&root->fs_info->fs_mutex);
183 make_bad_inode(inode);
184}
185
186static void fill_inode_item(struct btrfs_inode_item *item,
187 struct inode *inode)
188{
189 btrfs_set_inode_uid(item, inode->i_uid);
190 btrfs_set_inode_gid(item, inode->i_gid);
191 btrfs_set_inode_size(item, inode->i_size);
192 btrfs_set_inode_mode(item, inode->i_mode);
193 btrfs_set_inode_nlink(item, inode->i_nlink);
194 btrfs_set_timespec_sec(&item->atime, inode->i_atime.tv_sec);
195 btrfs_set_timespec_nsec(&item->atime, inode->i_atime.tv_nsec);
196 btrfs_set_timespec_sec(&item->mtime, inode->i_mtime.tv_sec);
197 btrfs_set_timespec_nsec(&item->mtime, inode->i_mtime.tv_nsec);
198 btrfs_set_timespec_sec(&item->ctime, inode->i_ctime.tv_sec);
199 btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec);
200 btrfs_set_inode_nblocks(item, inode->i_blocks);
201 btrfs_set_inode_generation(item, inode->i_generation);
618e21d5 202 btrfs_set_inode_rdev(item, inode->i_rdev);
39279cc3
CM		 203 btrfs_set_inode_block_group(item,
204 BTRFS_I(inode)->block_group->key.objectid);
205}
206
a52d9a80 207int btrfs_update_inode(struct btrfs_trans_handle *trans,
39279cc3
CM		 208 struct btrfs_root *root,
209 struct inode *inode)
210{
211 struct btrfs_inode_item *inode_item;
212 struct btrfs_path *path;
213 int ret;
214
215 path = btrfs_alloc_path();
216 BUG_ON(!path);
39279cc3
CM		 217 ret = btrfs_lookup_inode(trans, root, path,
218 &BTRFS_I(inode)->location, 1);
219 if (ret) {
220 if (ret > 0)
221 ret = -ENOENT;
222 goto failed;
223 }
224
225 inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
226 path->slots[0],
227 struct btrfs_inode_item);
228
229 fill_inode_item(inode_item, inode);
230 btrfs_mark_buffer_dirty(path->nodes[0]);
15ee9bc7 231 btrfs_set_inode_last_trans(trans, inode);
39279cc3
CM		 232 ret = 0;
233failed:
234 btrfs_release_path(root, path);
235 btrfs_free_path(path);
236 return ret;
237}
238
239
240static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
241 struct btrfs_root *root,
242 struct inode *dir,
243 struct dentry *dentry)
244{
245 struct btrfs_path *path;
246 const char *name = dentry->d_name.name;
247 int name_len = dentry->d_name.len;
248 int ret = 0;
249 u64 objectid;
250 struct btrfs_dir_item *di;
251
252 path = btrfs_alloc_path();
54aa1f4d
CM		 253 if (!path) {
254 ret = -ENOMEM;
255 goto err;
256 }
257
39279cc3
CM		 258 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
259 name, name_len, -1);
260 if (IS_ERR(di)) {
261 ret = PTR_ERR(di);
262 goto err;
263 }
264 if (!di) {
265 ret = -ENOENT;
266 goto err;
267 }
268 objectid = btrfs_disk_key_objectid(&di->location);
269 ret = btrfs_delete_one_dir_name(trans, root, path, di);
54aa1f4d
CM		 270 if (ret)
271 goto err;
39279cc3
CM		 272 btrfs_release_path(root, path);
273
274 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
275 objectid, name, name_len, -1);
276 if (IS_ERR(di)) {
277 ret = PTR_ERR(di);
278 goto err;
279 }
280 if (!di) {
281 ret = -ENOENT;
282 goto err;
283 }
284 ret = btrfs_delete_one_dir_name(trans, root, path, di);
39279cc3
CM		 285
286 dentry->d_inode->i_ctime = dir->i_ctime;
287err:
288 btrfs_free_path(path);
289 if (!ret) {
290 dir->i_size -= name_len * 2;
79c44584 291 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
39279cc3
CM		 292 btrfs_update_inode(trans, root, dir);
293 drop_nlink(dentry->d_inode);
54aa1f4d 294 ret = btrfs_update_inode(trans, root, dentry->d_inode);
39279cc3
CM		 295 dir->i_sb->s_dirt = 1;
296 }
297 return ret;
298}
299
300static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
301{
302 struct btrfs_root *root;
303 struct btrfs_trans_handle *trans;
304 int ret;
305
306 root = BTRFS_I(dir)->root;
307 mutex_lock(&root->fs_info->fs_mutex);
308 trans = btrfs_start_transaction(root, 1);
309 btrfs_set_trans_block_group(trans, dir);
310 ret = btrfs_unlink_trans(trans, root, dir, dentry);
311 btrfs_end_transaction(trans, root);
312 mutex_unlock(&root->fs_info->fs_mutex);
313 btrfs_btree_balance_dirty(root);
314 return ret;
315}
316
317static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
318{
319 struct inode *inode = dentry->d_inode;
320 int err;
321 int ret;
322 struct btrfs_root *root = BTRFS_I(dir)->root;
323 struct btrfs_path *path;
324 struct btrfs_key key;
325 struct btrfs_trans_handle *trans;
326 struct btrfs_key found_key;
327 int found_type;
328 struct btrfs_leaf *leaf;
329 char *goodnames = "..";
330
331 path = btrfs_alloc_path();
332 BUG_ON(!path);
39279cc3
CM		 333 mutex_lock(&root->fs_info->fs_mutex);
334 trans = btrfs_start_transaction(root, 1);
335 btrfs_set_trans_block_group(trans, dir);
336 key.objectid = inode->i_ino;
337 key.offset = (u64)-1;
338 key.flags = (u32)-1;
339 while(1) {
340 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
341 if (ret < 0) {
342 err = ret;
343 goto out;
344 }
345 BUG_ON(ret == 0);
346 if (path->slots[0] == 0) {
347 err = -ENOENT;
348 goto out;
349 }
350 path->slots[0]--;
351 leaf = btrfs_buffer_leaf(path->nodes[0]);
352 btrfs_disk_key_to_cpu(&found_key,
353 &leaf->items[path->slots[0]].key);
354 found_type = btrfs_key_type(&found_key);
355 if (found_key.objectid != inode->i_ino) {
356 err = -ENOENT;
357 goto out;
358 }
359 if ((found_type != BTRFS_DIR_ITEM_KEY &&
360 found_type != BTRFS_DIR_INDEX_KEY) ||
361 (!btrfs_match_dir_item_name(root, path, goodnames, 2) &&
362 !btrfs_match_dir_item_name(root, path, goodnames, 1))) {
363 err = -ENOTEMPTY;
364 goto out;
365 }
366 ret = btrfs_del_item(trans, root, path);
367 BUG_ON(ret);
368
369 if (found_type == BTRFS_DIR_ITEM_KEY && found_key.offset == 1)
370 break;
371 btrfs_release_path(root, path);
372 }
373 ret = 0;
374 btrfs_release_path(root, path);
375
376 /* now the directory is empty */
377 err = btrfs_unlink_trans(trans, root, dir, dentry);
378 if (!err) {
379 inode->i_size = 0;
380 }
381out:
382 btrfs_release_path(root, path);
383 btrfs_free_path(path);
384 mutex_unlock(&root->fs_info->fs_mutex);
385 ret = btrfs_end_transaction(trans, root);
386 btrfs_btree_balance_dirty(root);
387 if (ret && !err)
388 err = ret;
389 return err;
390}
391
392static int btrfs_free_inode(struct btrfs_trans_handle *trans,
393 struct btrfs_root *root,
394 struct inode *inode)
395{
396 struct btrfs_path *path;
397 int ret;
398
399 clear_inode(inode);
400
401 path = btrfs_alloc_path();
402 BUG_ON(!path);
39279cc3
CM		 403 ret = btrfs_lookup_inode(trans, root, path,
404 &BTRFS_I(inode)->location, -1);
54aa1f4d
CM		 405 if (ret > 0)
406 ret = -ENOENT;
407 if (!ret)
408 ret = btrfs_del_item(trans, root, path);
39279cc3
CM		 409 btrfs_free_path(path);
410 return ret;
411}
412
39279cc3
CM		 413/*
414 * this can truncate away extent items, csum items and directory items.
415 * It starts at a high offset and removes keys until it can't find
416 * any higher than i_size.
417 *
418 * csum items that cross the new i_size are truncated to the new size
419 * as well.
420 */
421static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
422 struct btrfs_root *root,
423 struct inode *inode)
424{
425 int ret;
426 struct btrfs_path *path;
427 struct btrfs_key key;
428 struct btrfs_disk_key *found_key;
429 u32 found_type;
430 struct btrfs_leaf *leaf;
431 struct btrfs_file_extent_item *fi;
432 u64 extent_start = 0;
433 u64 extent_num_blocks = 0;
434 u64 item_end = 0;
435 int found_extent;
436 int del_item;
437
a52d9a80 438 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
39279cc3 439 path = btrfs_alloc_path();
3c69faec 440 path->reada = -1;
39279cc3
CM		 441 BUG_ON(!path);
442 /* FIXME, add redo link to tree so we don't leak on crash */
443 key.objectid = inode->i_ino;
444 key.offset = (u64)-1;
445 key.flags = (u32)-1;
446 while(1) {
447 btrfs_init_path(path);
448 fi = NULL;
449 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
450 if (ret < 0) {
451 goto error;
452 }
453 if (ret > 0) {
454 BUG_ON(path->slots[0] == 0);
455 path->slots[0]--;
456 }
39279cc3
CM		 457 leaf = btrfs_buffer_leaf(path->nodes[0]);
458 found_key = &leaf->items[path->slots[0]].key;
459 found_type = btrfs_disk_key_type(found_key);
460
461 if (btrfs_disk_key_objectid(found_key) != inode->i_ino)
462 break;
463 if (found_type != BTRFS_CSUM_ITEM_KEY &&
464 found_type != BTRFS_DIR_ITEM_KEY &&
465 found_type != BTRFS_DIR_INDEX_KEY &&
466 found_type != BTRFS_EXTENT_DATA_KEY)
467 break;
468
469 item_end = btrfs_disk_key_offset(found_key);
470 if (found_type == BTRFS_EXTENT_DATA_KEY) {
471 fi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
472 path->slots[0],
473 struct btrfs_file_extent_item);
474 if (btrfs_file_extent_type(fi) !=
475 BTRFS_FILE_EXTENT_INLINE) {
476 item_end += btrfs_file_extent_num_blocks(fi) <<
477 inode->i_blkbits;
478 }
479 }
480 if (found_type == BTRFS_CSUM_ITEM_KEY) {
481 ret = btrfs_csum_truncate(trans, root, path,
482 inode->i_size);
483 BUG_ON(ret);
484 }
485 if (item_end < inode->i_size) {
b888db2b
CM		 486 if (found_type == BTRFS_DIR_ITEM_KEY) {
487 found_type = BTRFS_INODE_ITEM_KEY;
488 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
489 found_type = BTRFS_CSUM_ITEM_KEY;
490 } else if (found_type) {
491 found_type--;
492 } else {
493 break;
39279cc3 494 }
b888db2b
CM		 495 btrfs_set_key_type(&key, found_type - 1);
496 continue;
39279cc3
CM		 497 }
498 if (btrfs_disk_key_offset(found_key) >= inode->i_size)
499 del_item = 1;
500 else
501 del_item = 0;
502 found_extent = 0;
503
504 /* FIXME, shrink the extent if the ref count is only 1 */
505 if (found_type == BTRFS_EXTENT_DATA_KEY &&
506 btrfs_file_extent_type(fi) !=
507 BTRFS_FILE_EXTENT_INLINE) {
508 u64 num_dec;
509 if (!del_item) {
510 u64 orig_num_blocks =
511 btrfs_file_extent_num_blocks(fi);
512 extent_num_blocks = inode->i_size -
513 btrfs_disk_key_offset(found_key) +
514 root->blocksize - 1;
515 extent_num_blocks >>= inode->i_blkbits;
516 btrfs_set_file_extent_num_blocks(fi,
517 extent_num_blocks);
518 inode->i_blocks -= (orig_num_blocks -
519 extent_num_blocks) << 3;
ccd467d6 520 btrfs_mark_buffer_dirty(path->nodes[0]);
39279cc3
CM		 521 } else {
522 extent_start =
523 btrfs_file_extent_disk_blocknr(fi);
524 extent_num_blocks =
525 btrfs_file_extent_disk_num_blocks(fi);
526 /* FIXME blocksize != 4096 */
527 num_dec = btrfs_file_extent_num_blocks(fi) << 3;
528 if (extent_start != 0) {
529 found_extent = 1;
530 inode->i_blocks -= num_dec;
531 }
532 }
533 }
534 if (del_item) {
535 ret = btrfs_del_item(trans, root, path);
54aa1f4d
CM		 536 if (ret)
537 goto error;
39279cc3
CM		 538 } else {
539 break;
540 }
541 btrfs_release_path(root, path);
542 if (found_extent) {
543 ret = btrfs_free_extent(trans, root, extent_start,
544 extent_num_blocks, 0);
545 BUG_ON(ret);
546 }
547 }
548 ret = 0;
549error:
550 btrfs_release_path(root, path);
551 btrfs_free_path(path);
552 inode->i_sb->s_dirt = 1;
553 return ret;
554}
555
b888db2b 556static int btrfs_cow_one_page(struct inode *inode, struct page *page,
a52d9a80
CM		 557 size_t zero_start)
558{
559 char *kaddr;
560 int ret = 0;
b888db2b 561 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
a52d9a80 562 u64 page_start = page->index << PAGE_CACHE_SHIFT;
b888db2b 563 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
a52d9a80
CM		 564
565 if (!PagePrivate(page)) {
566 SetPagePrivate(page);
567 set_page_private(page, 1);
b888db2b 568 WARN_ON(!page->mapping->a_ops->invalidatepage);
a52d9a80
CM		 569 page_cache_get(page);
570 }
571
b888db2b
CM		 572 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
573 set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start,
574 page_end, GFP_NOFS);
a52d9a80 575 if (zero_start != PAGE_CACHE_SIZE) {
b888db2b 576 kaddr = kmap(page);
a52d9a80
CM		 577 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
578 flush_dcache_page(page);
b888db2b 579 kunmap(page);
a52d9a80 580 }
b888db2b
CM		 581 set_page_dirty(page);
582 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
a52d9a80 583
a52d9a80
CM		 584 return ret;
585}
586
39279cc3
CM		 587/*
588 * taken from block_truncate_page, but does cow as it zeros out
589 * any bytes left in the last page in the file.
590 */
591static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
592{
593 struct inode *inode = mapping->host;
594 unsigned blocksize = 1 << inode->i_blkbits;
595 pgoff_t index = from >> PAGE_CACHE_SHIFT;
596 unsigned offset = from & (PAGE_CACHE_SIZE-1);
597 struct page *page;
39279cc3 598 int ret = 0;
a52d9a80 599 u64 page_start;
39279cc3
CM		 600
601 if ((offset & (blocksize - 1)) == 0)
602 goto out;
603
604 ret = -ENOMEM;
605 page = grab_cache_page(mapping, index);
606 if (!page)
607 goto out;
39279cc3 608 if (!PageUptodate(page)) {
9ebefb18 609 ret = btrfs_readpage(NULL, page);
39279cc3
CM		 610 lock_page(page);
611 if (!PageUptodate(page)) {
612 ret = -EIO;
613 goto out;
614 }
615 }
a52d9a80
CM		 616 page_start = page->index << PAGE_CACHE_SHIFT;
617
b888db2b 618 ret = btrfs_cow_one_page(inode, page, offset);
39279cc3 619
39279cc3
CM		 620 unlock_page(page);
621 page_cache_release(page);
622out:
623 return ret;
624}
625
626static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
627{
628 struct inode *inode = dentry->d_inode;
629 int err;
630
631 err = inode_change_ok(inode, attr);
632 if (err)
633 return err;
634
635 if (S_ISREG(inode->i_mode) &&
636 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
637 struct btrfs_trans_handle *trans;
638 struct btrfs_root *root = BTRFS_I(inode)->root;
639 u64 mask = root->blocksize - 1;
640 u64 pos = (inode->i_size + mask) & ~mask;
641 u64 hole_size;
642
643 if (attr->ia_size <= pos)
644 goto out;
645
646 btrfs_truncate_page(inode->i_mapping, inode->i_size);
647
648 hole_size = (attr->ia_size - pos + mask) & ~mask;
649 hole_size >>= inode->i_blkbits;
650
651 mutex_lock(&root->fs_info->fs_mutex);
652 trans = btrfs_start_transaction(root, 1);
653 btrfs_set_trans_block_group(trans, inode);
654 err = btrfs_insert_file_extent(trans, root, inode->i_ino,
655 pos, 0, 0, hole_size);
39279cc3
CM		 656 btrfs_end_transaction(trans, root);
657 mutex_unlock(&root->fs_info->fs_mutex);
54aa1f4d
CM		 658 if (err)
659 return err;
39279cc3
CM		 660 }
661out:
662 err = inode_setattr(inode, attr);
663
664 return err;
665}
666void btrfs_delete_inode(struct inode *inode)
667{
668 struct btrfs_trans_handle *trans;
669 struct btrfs_root *root = BTRFS_I(inode)->root;
670 int ret;
671
672 truncate_inode_pages(&inode->i_data, 0);
673 if (is_bad_inode(inode)) {
674 goto no_delete;
675 }
676 inode->i_size = 0;
677 mutex_lock(&root->fs_info->fs_mutex);
678 trans = btrfs_start_transaction(root, 1);
679 btrfs_set_trans_block_group(trans, inode);
680 ret = btrfs_truncate_in_trans(trans, root, inode);
54aa1f4d
CM		 681 if (ret)
682 goto no_delete_lock;
683 ret = btrfs_free_inode(trans, root, inode);
684 if (ret)
685 goto no_delete_lock;
39279cc3
CM		 686 btrfs_end_transaction(trans, root);
687 mutex_unlock(&root->fs_info->fs_mutex);
688 btrfs_btree_balance_dirty(root);
689 return;
54aa1f4d
CM		 690
691no_delete_lock:
692 btrfs_end_transaction(trans, root);
693 mutex_unlock(&root->fs_info->fs_mutex);
694 btrfs_btree_balance_dirty(root);
39279cc3
CM		 695no_delete:
696 clear_inode(inode);
697}
698
699/*
700 * this returns the key found in the dir entry in the location pointer.
701 * If no dir entries were found, location->objectid is 0.
702 */
703static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
704 struct btrfs_key *location)
705{
706 const char *name = dentry->d_name.name;
707 int namelen = dentry->d_name.len;
708 struct btrfs_dir_item *di;
709 struct btrfs_path *path;
710 struct btrfs_root *root = BTRFS_I(dir)->root;
711 int ret;
712
713 path = btrfs_alloc_path();
714 BUG_ON(!path);
39279cc3
CM		 715 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
716 namelen, 0);
717 if (!di || IS_ERR(di)) {
718 location->objectid = 0;
719 ret = 0;
720 goto out;
721 }
722 btrfs_disk_key_to_cpu(location, &di->location);
723out:
724 btrfs_release_path(root, path);
725 btrfs_free_path(path);
726 return ret;
727}
728
729/*
730 * when we hit a tree root in a directory, the btrfs part of the inode
731 * needs to be changed to reflect the root directory of the tree root. This
732 * is kind of like crossing a mount point.
733 */
734static int fixup_tree_root_location(struct btrfs_root *root,
735 struct btrfs_key *location,
58176a96
JB		 736 struct btrfs_root **sub_root,
737 struct dentry *dentry)
39279cc3
CM		 738{
739 struct btrfs_path *path;
740 struct btrfs_root_item *ri;
741
742 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
743 return 0;
744 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
745 return 0;
746
747 path = btrfs_alloc_path();
748 BUG_ON(!path);
749 mutex_lock(&root->fs_info->fs_mutex);
750
58176a96
JB		 751 *sub_root = btrfs_read_fs_root(root->fs_info, location,
752 dentry->d_name.name,
753 dentry->d_name.len);
39279cc3
CM		 754 if (IS_ERR(*sub_root))
755 return PTR_ERR(*sub_root);
756
757 ri = &(*sub_root)->root_item;
758 location->objectid = btrfs_root_dirid(ri);
759 location->flags = 0;
760 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
761 location->offset = 0;
762
763 btrfs_free_path(path);
764 mutex_unlock(&root->fs_info->fs_mutex);
765 return 0;
766}
767
768static int btrfs_init_locked_inode(struct inode *inode, void *p)
769{
770 struct btrfs_iget_args *args = p;
771 inode->i_ino = args->ino;
772 BTRFS_I(inode)->root = args->root;
b888db2b
CM		 773 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
774 inode->i_mapping, GFP_NOFS);
39279cc3
CM		 775 return 0;
776}
777
778static int btrfs_find_actor(struct inode *inode, void *opaque)
779{
780 struct btrfs_iget_args *args = opaque;
781 return (args->ino == inode->i_ino &&
782 args->root == BTRFS_I(inode)->root);
783}
784
785struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
786 struct btrfs_root *root)
787{
788 struct inode *inode;
789 struct btrfs_iget_args args;
790 args.ino = objectid;
791 args.root = root;
792
793 inode = iget5_locked(s, objectid, btrfs_find_actor,
794 btrfs_init_locked_inode,
795 (void *)&args);
796 return inode;
797}
798
799static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
800 struct nameidata *nd)
801{
802 struct inode * inode;
803 struct btrfs_inode *bi = BTRFS_I(dir);
804 struct btrfs_root *root = bi->root;
805 struct btrfs_root *sub_root = root;
806 struct btrfs_key location;
807 int ret;
808
809 if (dentry->d_name.len > BTRFS_NAME_LEN)
810 return ERR_PTR(-ENAMETOOLONG);
811 mutex_lock(&root->fs_info->fs_mutex);
812 ret = btrfs_inode_by_name(dir, dentry, &location);
813 mutex_unlock(&root->fs_info->fs_mutex);
814 if (ret < 0)
815 return ERR_PTR(ret);
816 inode = NULL;
817 if (location.objectid) {
58176a96
JB		 818 ret = fixup_tree_root_location(root, &location, &sub_root,
819 dentry);
39279cc3
CM		 820 if (ret < 0)
821 return ERR_PTR(ret);
822 if (ret > 0)
823 return ERR_PTR(-ENOENT);
824 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
825 sub_root);
826 if (!inode)
827 return ERR_PTR(-EACCES);
828 if (inode->i_state & I_NEW) {
829 /* the inode and parent dir are two different roots */
830 if (sub_root != root) {
831 igrab(inode);
832 sub_root->inode = inode;
833 }
834 BTRFS_I(inode)->root = sub_root;
835 memcpy(&BTRFS_I(inode)->location, &location,
836 sizeof(location));
837 btrfs_read_locked_inode(inode);
838 unlock_new_inode(inode);
839 }
840 }
841 return d_splice_alias(inode, dentry);
842}
843
39279cc3
CM		 844static unsigned char btrfs_filetype_table[] = {
845 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
846};
847
848static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
849{
850 struct inode *inode = filp->f_path.dentry->d_inode;
851 struct btrfs_root *root = BTRFS_I(inode)->root;
852 struct btrfs_item *item;
853 struct btrfs_dir_item *di;
854 struct btrfs_key key;
855 struct btrfs_path *path;
856 int ret;
857 u32 nritems;
858 struct btrfs_leaf *leaf;
859 int slot;
860 int advance;
861 unsigned char d_type;
862 int over = 0;
863 u32 di_cur;
864 u32 di_total;
865 u32 di_len;
866 int key_type = BTRFS_DIR_INDEX_KEY;
867
868 /* FIXME, use a real flag for deciding about the key type */
869 if (root->fs_info->tree_root == root)
870 key_type = BTRFS_DIR_ITEM_KEY;
871 mutex_lock(&root->fs_info->fs_mutex);
872 key.objectid = inode->i_ino;
873 key.flags = 0;
874 btrfs_set_key_type(&key, key_type);
875 key.offset = filp->f_pos;
876 path = btrfs_alloc_path();
2cc58cf2 877 path->reada = 2;
39279cc3
CM		 878 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
879 if (ret < 0)
880 goto err;
881 advance = 0;
39279cc3
CM		 882 while(1) {
883 leaf = btrfs_buffer_leaf(path->nodes[0]);
884 nritems = btrfs_header_nritems(&leaf->header);
885 slot = path->slots[0];
886 if (advance || slot >= nritems) {
887 if (slot >= nritems -1) {
39279cc3
CM		 888 ret = btrfs_next_leaf(root, path);
889 if (ret)
890 break;
891 leaf = btrfs_buffer_leaf(path->nodes[0]);
892 nritems = btrfs_header_nritems(&leaf->header);
893 slot = path->slots[0];
894 } else {
895 slot++;
896 path->slots[0]++;
897 }
898 }
899 advance = 1;
900 item = leaf->items + slot;
901 if (btrfs_disk_key_objectid(&item->key) != key.objectid)
902 break;
903 if (btrfs_disk_key_type(&item->key) != key_type)
904 break;
905 if (btrfs_disk_key_offset(&item->key) < filp->f_pos)
906 continue;
907 filp->f_pos = btrfs_disk_key_offset(&item->key);
908 advance = 1;
909 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
910 di_cur = 0;
911 di_total = btrfs_item_size(leaf->items + slot);
912 while(di_cur < di_total) {
913 d_type = btrfs_filetype_table[btrfs_dir_type(di)];
914 over = filldir(dirent, (const char *)(di + 1),
915 btrfs_dir_name_len(di),
916 btrfs_disk_key_offset(&item->key),
917 btrfs_disk_key_objectid(&di->location),
918 d_type);
919 if (over)
920 goto nopos;
921 di_len = btrfs_dir_name_len(di) + sizeof(*di);
922 di_cur += di_len;
923 di = (struct btrfs_dir_item *)((char *)di + di_len);
924 }
925 }
926 filp->f_pos++;
927nopos:
928 ret = 0;
929err:
930 btrfs_release_path(root, path);
931 btrfs_free_path(path);
932 mutex_unlock(&root->fs_info->fs_mutex);
933 return ret;
934}
935
936int btrfs_write_inode(struct inode *inode, int wait)
937{
938 struct btrfs_root *root = BTRFS_I(inode)->root;
939 struct btrfs_trans_handle *trans;
940 int ret = 0;
941
942 if (wait) {
943 mutex_lock(&root->fs_info->fs_mutex);
944 trans = btrfs_start_transaction(root, 1);
945 btrfs_set_trans_block_group(trans, inode);
946 ret = btrfs_commit_transaction(trans, root);
947 mutex_unlock(&root->fs_info->fs_mutex);
948 }
949 return ret;
950}
951
952/*
54aa1f4d 953 * This is somewhat expensive, updating the tree every time the
39279cc3
CM		 954 * inode changes. But, it is most likely to find the inode in cache.
955 * FIXME, needs more benchmarking...there are no reasons other than performance
956 * to keep or drop this code.
957 */
958void btrfs_dirty_inode(struct inode *inode)
959{
960 struct btrfs_root *root = BTRFS_I(inode)->root;
961 struct btrfs_trans_handle *trans;
962
963 mutex_lock(&root->fs_info->fs_mutex);
964 trans = btrfs_start_transaction(root, 1);
965 btrfs_set_trans_block_group(trans, inode);
966 btrfs_update_inode(trans, root, inode);
967 btrfs_end_transaction(trans, root);
968 mutex_unlock(&root->fs_info->fs_mutex);
39279cc3
CM		 969}
970
971static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
972 struct btrfs_root *root,
973 u64 objectid,
974 struct btrfs_block_group_cache *group,
975 int mode)
976{
977 struct inode *inode;
978 struct btrfs_inode_item inode_item;
979 struct btrfs_key *location;
980 int ret;
981 int owner;
982
983 inode = new_inode(root->fs_info->sb);
984 if (!inode)
985 return ERR_PTR(-ENOMEM);
986
b888db2b
CM		 987 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
988 inode->i_mapping, GFP_NOFS);
39279cc3 989 BTRFS_I(inode)->root = root;
b888db2b 990
39279cc3
CM		 991 if (mode & S_IFDIR)
992 owner = 0;
993 else
994 owner = 1;
995 group = btrfs_find_block_group(root, group, 0, 0, owner);
996 BTRFS_I(inode)->block_group = group;
997
998 inode->i_uid = current->fsuid;
999 inode->i_gid = current->fsgid;
1000 inode->i_mode = mode;
1001 inode->i_ino = objectid;
1002 inode->i_blocks = 0;
1003 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1004 fill_inode_item(&inode_item, inode);
1005 location = &BTRFS_I(inode)->location;
1006 location->objectid = objectid;
1007 location->flags = 0;
1008 location->offset = 0;
1009 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1010
1011 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
54aa1f4d
CM		 1012 if (ret)
1013 return ERR_PTR(ret);
39279cc3
CM		 1014 insert_inode_hash(inode);
1015 return inode;
1016}
1017
1018static inline u8 btrfs_inode_type(struct inode *inode)
1019{
1020 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1021}
1022
1023static int btrfs_add_link(struct btrfs_trans_handle *trans,
1024 struct dentry *dentry, struct inode *inode)
1025{
1026 int ret;
1027 struct btrfs_key key;
1028 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
79c44584 1029 struct inode *parent_inode;
39279cc3
CM		 1030 key.objectid = inode->i_ino;
1031 key.flags = 0;
1032 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1033 key.offset = 0;
1034
1035 ret = btrfs_insert_dir_item(trans, root,
1036 dentry->d_name.name, dentry->d_name.len,
1037 dentry->d_parent->d_inode->i_ino,
1038 &key, btrfs_inode_type(inode));
1039 if (ret == 0) {
79c44584
CM		 1040 parent_inode = dentry->d_parent->d_inode;
1041 parent_inode->i_size += dentry->d_name.len * 2;
1042 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
39279cc3
CM		 1043 ret = btrfs_update_inode(trans, root,
1044 dentry->d_parent->d_inode);
1045 }
1046 return ret;
1047}
1048
1049static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1050 struct dentry *dentry, struct inode *inode)
1051{
1052 int err = btrfs_add_link(trans, dentry, inode);
1053 if (!err) {
1054 d_instantiate(dentry, inode);
1055 return 0;
1056 }
1057 if (err > 0)
1058 err = -EEXIST;
1059 return err;
1060}
1061
618e21d5
JB		 1062static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1063 int mode, dev_t rdev)
1064{
1065 struct btrfs_trans_handle *trans;
1066 struct btrfs_root *root = BTRFS_I(dir)->root;
1067 struct inode *inode;
1068 int err;
1069 int drop_inode = 0;
1070 u64 objectid;
1071
1072 if (!new_valid_dev(rdev))
1073 return -EINVAL;
1074
1075 mutex_lock(&root->fs_info->fs_mutex);
1076 trans = btrfs_start_transaction(root, 1);
1077 btrfs_set_trans_block_group(trans, dir);
1078
1079 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1080 if (err) {
1081 err = -ENOSPC;
1082 goto out_unlock;
1083 }
1084
1085 inode = btrfs_new_inode(trans, root, objectid,
1086 BTRFS_I(dir)->block_group, mode);
1087 err = PTR_ERR(inode);
1088 if (IS_ERR(inode))
1089 goto out_unlock;
1090
1091 btrfs_set_trans_block_group(trans, inode);
1092 err = btrfs_add_nondir(trans, dentry, inode);
1093 if (err)
1094 drop_inode = 1;
1095 else {
1096 inode->i_op = &btrfs_special_inode_operations;
1097 init_special_inode(inode, inode->i_mode, rdev);
1b4ab1bb 1098 btrfs_update_inode(trans, root, inode);
618e21d5
JB		 1099 }
1100 dir->i_sb->s_dirt = 1;
1101 btrfs_update_inode_block_group(trans, inode);
1102 btrfs_update_inode_block_group(trans, dir);
1103out_unlock:
1104 btrfs_end_transaction(trans, root);
1105 mutex_unlock(&root->fs_info->fs_mutex);
1106
1107 if (drop_inode) {
1108 inode_dec_link_count(inode);
1109 iput(inode);
1110 }
1111 btrfs_btree_balance_dirty(root);
1112 return err;
1113}
1114
39279cc3
CM		 1115static int btrfs_create(struct inode *dir, struct dentry *dentry,
1116 int mode, struct nameidata *nd)
1117{
1118 struct btrfs_trans_handle *trans;
1119 struct btrfs_root *root = BTRFS_I(dir)->root;
1120 struct inode *inode;
1121 int err;
1122 int drop_inode = 0;
1123 u64 objectid;
1124
1125 mutex_lock(&root->fs_info->fs_mutex);
1126 trans = btrfs_start_transaction(root, 1);
1127 btrfs_set_trans_block_group(trans, dir);
1128
1129 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1130 if (err) {
1131 err = -ENOSPC;
1132 goto out_unlock;
1133 }
1134
1135 inode = btrfs_new_inode(trans, root, objectid,
1136 BTRFS_I(dir)->block_group, mode);
1137 err = PTR_ERR(inode);
1138 if (IS_ERR(inode))
1139 goto out_unlock;
1140
1141 btrfs_set_trans_block_group(trans, inode);
1142 err = btrfs_add_nondir(trans, dentry, inode);
1143 if (err)
1144 drop_inode = 1;
1145 else {
1146 inode->i_mapping->a_ops = &btrfs_aops;
1147 inode->i_fop = &btrfs_file_operations;
1148 inode->i_op = &btrfs_file_inode_operations;
a52d9a80
CM		 1149 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1150 inode->i_mapping, GFP_NOFS);
b888db2b 1151 BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range;
39279cc3
CM		 1152 }
1153 dir->i_sb->s_dirt = 1;
1154 btrfs_update_inode_block_group(trans, inode);
1155 btrfs_update_inode_block_group(trans, dir);
1156out_unlock:
1157 btrfs_end_transaction(trans, root);
1158 mutex_unlock(&root->fs_info->fs_mutex);
1159
1160 if (drop_inode) {
1161 inode_dec_link_count(inode);
1162 iput(inode);
1163 }
1164 btrfs_btree_balance_dirty(root);
1165 return err;
1166}
1167
1168static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1169 struct dentry *dentry)
1170{
1171 struct btrfs_trans_handle *trans;
1172 struct btrfs_root *root = BTRFS_I(dir)->root;
1173 struct inode *inode = old_dentry->d_inode;
1174 int err;
1175 int drop_inode = 0;
1176
1177 if (inode->i_nlink == 0)
1178 return -ENOENT;
1179
1180 inc_nlink(inode);
1181 mutex_lock(&root->fs_info->fs_mutex);
1182 trans = btrfs_start_transaction(root, 1);
1183 btrfs_set_trans_block_group(trans, dir);
1184 atomic_inc(&inode->i_count);
1185 err = btrfs_add_nondir(trans, dentry, inode);
1186 if (err)
1187 drop_inode = 1;
1188 dir->i_sb->s_dirt = 1;
1189 btrfs_update_inode_block_group(trans, dir);
54aa1f4d
CM		 1190 err = btrfs_update_inode(trans, root, inode);
1191 if (err)
1192 drop_inode = 1;
39279cc3
CM		 1193
1194 btrfs_end_transaction(trans, root);
1195 mutex_unlock(&root->fs_info->fs_mutex);
1196
1197 if (drop_inode) {
1198 inode_dec_link_count(inode);
1199 iput(inode);
1200 }
1201 btrfs_btree_balance_dirty(root);
1202 return err;
1203}
1204
1205static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
1206 struct btrfs_root *root,
1207 u64 objectid, u64 dirid)
1208{
1209 int ret;
1210 char buf[2];
1211 struct btrfs_key key;
1212
1213 buf[0] = '.';
1214 buf[1] = '.';
1215
1216 key.objectid = objectid;
1217 key.offset = 0;
1218 key.flags = 0;
1219 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1220
1221 ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid,
1222 &key, BTRFS_FT_DIR);
1223 if (ret)
1224 goto error;
1225 key.objectid = dirid;
1226 ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid,
1227 &key, BTRFS_FT_DIR);
1228 if (ret)
1229 goto error;
1230error:
1231 return ret;
1232}
1233
1234static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1235{
1236 struct inode *inode;
1237 struct btrfs_trans_handle *trans;
1238 struct btrfs_root *root = BTRFS_I(dir)->root;
1239 int err = 0;
1240 int drop_on_err = 0;
1241 u64 objectid;
1242
1243 mutex_lock(&root->fs_info->fs_mutex);
1244 trans = btrfs_start_transaction(root, 1);
1245 btrfs_set_trans_block_group(trans, dir);
1246 if (IS_ERR(trans)) {
1247 err = PTR_ERR(trans);
1248 goto out_unlock;
1249 }
1250
1251 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1252 if (err) {
1253 err = -ENOSPC;
1254 goto out_unlock;
1255 }
1256
1257 inode = btrfs_new_inode(trans, root, objectid,
1258 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1259 if (IS_ERR(inode)) {
1260 err = PTR_ERR(inode);
1261 goto out_fail;
1262 }
1263 drop_on_err = 1;
1264 inode->i_op = &btrfs_dir_inode_operations;
1265 inode->i_fop = &btrfs_dir_file_operations;
1266 btrfs_set_trans_block_group(trans, inode);
1267
1268 err = btrfs_make_empty_dir(trans, root, inode->i_ino, dir->i_ino);
1269 if (err)
1270 goto out_fail;
1271
1272 inode->i_size = 6;
1273 err = btrfs_update_inode(trans, root, inode);
1274 if (err)
1275 goto out_fail;
1276 err = btrfs_add_link(trans, dentry, inode);
1277 if (err)
1278 goto out_fail;
1279 d_instantiate(dentry, inode);
1280 drop_on_err = 0;
1281 dir->i_sb->s_dirt = 1;
1282 btrfs_update_inode_block_group(trans, inode);
1283 btrfs_update_inode_block_group(trans, dir);
1284
1285out_fail:
1286 btrfs_end_transaction(trans, root);
1287out_unlock:
1288 mutex_unlock(&root->fs_info->fs_mutex);
1289 if (drop_on_err)
1290 iput(inode);
1291 btrfs_btree_balance_dirty(root);
1292 return err;
1293}
1294
a52d9a80
CM		 1295struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
1296 size_t page_offset, u64 start, u64 end,
1297 int create)
1298{
1299 int ret;
1300 int err = 0;
1301 u64 blocknr;
1302 u64 extent_start = 0;
1303 u64 extent_end = 0;
1304 u64 objectid = inode->i_ino;
1305 u32 found_type;
1306 int failed_insert = 0;
1307 struct btrfs_path *path;
1308 struct btrfs_root *root = BTRFS_I(inode)->root;
1309 struct btrfs_file_extent_item *item;
1310 struct btrfs_leaf *leaf;
1311 struct btrfs_disk_key *found_key;
1312 struct extent_map *em = NULL;
1313 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1314 struct btrfs_trans_handle *trans = NULL;
1315
1316 path = btrfs_alloc_path();
1317 BUG_ON(!path);
1318 mutex_lock(&root->fs_info->fs_mutex);
1319
1320again:
1321 em = lookup_extent_mapping(em_tree, start, end);
1322 if (em) {
1323 goto out;
1324 }
1325 if (!em) {
1326 em = alloc_extent_map(GFP_NOFS);
1327 if (!em) {
1328 err = -ENOMEM;
1329 goto out;
1330 }
1331 em->start = 0;
1332 em->end = 0;
1333 }
1334 em->bdev = inode->i_sb->s_bdev;
1335 ret = btrfs_lookup_file_extent(NULL, root, path,
1336 objectid, start, 0);
1337 if (ret < 0) {
1338 err = ret;
1339 goto out;
1340 }
1341
1342 if (ret != 0) {
1343 if (path->slots[0] == 0)
1344 goto not_found;
1345 path->slots[0]--;
1346 }
1347
1348 item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
1349 struct btrfs_file_extent_item);
1350 leaf = btrfs_buffer_leaf(path->nodes[0]);
1351 blocknr = btrfs_file_extent_disk_blocknr(item);
1352 blocknr += btrfs_file_extent_offset(item);
1353
1354 /* are we inside the extent that was found? */
1355 found_key = &leaf->items[path->slots[0]].key;
1356 found_type = btrfs_disk_key_type(found_key);
1357 if (btrfs_disk_key_objectid(found_key) != objectid ||
1358 found_type != BTRFS_EXTENT_DATA_KEY) {
1359 goto not_found;
1360 }
1361
1362 found_type = btrfs_file_extent_type(item);
1363 extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
1364 if (found_type == BTRFS_FILE_EXTENT_REG) {
1365 extent_end = extent_start +
1366 (btrfs_file_extent_num_blocks(item) << inode->i_blkbits);
1367 err = 0;
b888db2b 1368 if (start < extent_start || start >= extent_end) {
a52d9a80
CM		 1369 em->start = start;
1370 if (start < extent_start) {
b888db2b
CM		 1371 if (end < extent_start)
1372 goto not_found;
a52d9a80
CM		 1373 em->end = extent_end - 1;
1374 } else {
1375 em->end = end;
1376 }
1377 goto not_found_em;
1378 }
1379 if (btrfs_file_extent_disk_blocknr(item) == 0) {
1380 em->start = extent_start;
1381 em->end = extent_end - 1;
1382 em->block_start = 0;
1383 em->block_end = 0;
1384 goto insert;
1385 }
1386 em->block_start = blocknr << inode->i_blkbits;
1387 em->block_end = em->block_start +
1388 (btrfs_file_extent_num_blocks(item) <<
1389 inode->i_blkbits) - 1;
1390 em->start = extent_start;
1391 em->end = extent_end - 1;
1392 goto insert;
1393 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1394 char *ptr;
1395 char *map;
1396 u32 size;
1397
1398 size = btrfs_file_extent_inline_len(leaf->items +
1399 path->slots[0]);
1400 extent_end = extent_start + size;
b888db2b 1401 if (start < extent_start || start >= extent_end) {
a52d9a80
CM		 1402 em->start = start;
1403 if (start < extent_start) {
b888db2b
CM		 1404 if (end < extent_start)
1405 goto not_found;
a52d9a80
CM		 1406 em->end = extent_end - 1;
1407 } else {
1408 em->end = end;
1409 }
1410 goto not_found_em;
1411 }
1412 em->block_start = EXTENT_MAP_INLINE;
1413 em->block_end = EXTENT_MAP_INLINE;
1414 em->start = extent_start;
1415 em->end = extent_end - 1;
1416 if (!page) {
1417 goto insert;
1418 }
1419 ptr = btrfs_file_extent_inline_start(item);
1420 map = kmap(page);
1421 memcpy(map + page_offset, ptr, size);
1422 flush_dcache_page(result->b_page);
1423 kunmap(page);
1424 set_extent_uptodate(em_tree, extent_start,
1425 extent_end, GFP_NOFS);
1426 goto insert;
1427 } else {
1428 printk("unkknown found_type %d\n", found_type);
1429 WARN_ON(1);
1430 }
1431not_found:
1432 em->start = start;
1433 em->end = end;
1434not_found_em:
1435 em->block_start = 0;
1436 em->block_end = 0;
1437insert:
1438 btrfs_release_path(root, path);
1439 if (em->start > start || em->end < start) {
b888db2b 1440 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->end, start, end);
a52d9a80
CM		 1441 err = -EIO;
1442 goto out;
1443 }
1444 ret = add_extent_mapping(em_tree, em);
1445 if (ret == -EEXIST) {
1446 free_extent_map(em);
1447 failed_insert++;
1448 if (failed_insert > 5) {
1449 printk("failing to insert %Lu %Lu\n", start, end);
1450 err = -EIO;
1451 goto out;
1452 }
1453 em = NULL;
1454 goto again;
1455 }
1456 err = 0;
1457out:
1458 btrfs_free_path(path);
1459 if (trans) {
1460 ret = btrfs_end_transaction(trans, root);
1461 if (!err)
1462 err = ret;
1463 }
1464 mutex_unlock(&root->fs_info->fs_mutex);
1465 if (err) {
1466 free_extent_map(em);
1467 WARN_ON(1);
1468 return ERR_PTR(err);
1469 }
1470 return em;
1471}
1472
1473
39279cc3
CM		 1474/*
1475 * FIBMAP and others want to pass in a fake buffer head. They need to
1476 * use BTRFS_GET_BLOCK_NO_DIRECT to make sure we don't try to memcpy
1477 * any packed file data into the fake bh
1478 */
1479#define BTRFS_GET_BLOCK_NO_CREATE 0
1480#define BTRFS_GET_BLOCK_CREATE 1
1481#define BTRFS_GET_BLOCK_NO_DIRECT 2
1482
1483/*
1484 * FIXME create==1 doe not work.
1485 */
1486static int btrfs_get_block_lock(struct inode *inode, sector_t iblock,
1487 struct buffer_head *result, int create)
1488{
1489 int ret;
1490 int err = 0;
1491 u64 blocknr;
1492 u64 extent_start = 0;
1493 u64 extent_end = 0;
1494 u64 objectid = inode->i_ino;
1495 u32 found_type;
1496 u64 alloc_hint = 0;
1497 struct btrfs_path *path;
1498 struct btrfs_root *root = BTRFS_I(inode)->root;
1499 struct btrfs_file_extent_item *item;
1500 struct btrfs_leaf *leaf;
1501 struct btrfs_disk_key *found_key;
1502 struct btrfs_trans_handle *trans = NULL;
1503
1504 path = btrfs_alloc_path();
1505 BUG_ON(!path);
39279cc3 1506 if (create & BTRFS_GET_BLOCK_CREATE) {
9ebefb18
CM		 1507 /*
1508 * danger!, this only works if the page is properly up
1509 * to date somehow
1510 */
39279cc3
CM		 1511 trans = btrfs_start_transaction(root, 1);
1512 if (!trans) {
1513 err = -ENOMEM;
1514 goto out;
1515 }
1516 ret = btrfs_drop_extents(trans, root, inode,
1517 iblock << inode->i_blkbits,
1518 (iblock + 1) << inode->i_blkbits,
1519 &alloc_hint);
1520 BUG_ON(ret);
1521 }
1522
1523 ret = btrfs_lookup_file_extent(NULL, root, path,
f1ace244 1524 objectid,
39279cc3
CM		 1525 iblock << inode->i_blkbits, 0);
1526 if (ret < 0) {
1527 err = ret;
1528 goto out;
1529 }
1530
1531 if (ret != 0) {
1532 if (path->slots[0] == 0) {
1533 btrfs_release_path(root, path);
1534 goto not_found;
1535 }
1536 path->slots[0]--;
1537 }
1538
1539 item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
1540 struct btrfs_file_extent_item);
1541 leaf = btrfs_buffer_leaf(path->nodes[0]);
1542 blocknr = btrfs_file_extent_disk_blocknr(item);
1543 blocknr += btrfs_file_extent_offset(item);
1544
1545 /* are we inside the extent that was found? */
1546 found_key = &leaf->items[path->slots[0]].key;
1547 found_type = btrfs_disk_key_type(found_key);
1548 if (btrfs_disk_key_objectid(found_key) != objectid ||
1549 found_type != BTRFS_EXTENT_DATA_KEY) {
1550 extent_end = 0;
1551 extent_start = 0;
1552 goto not_found;
1553 }
1554 found_type = btrfs_file_extent_type(item);
1555 extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
1556 if (found_type == BTRFS_FILE_EXTENT_REG) {
1557 extent_start = extent_start >> inode->i_blkbits;
1558 extent_end = extent_start + btrfs_file_extent_num_blocks(item);
1559 err = 0;
1560 if (btrfs_file_extent_disk_blocknr(item) == 0)
1561 goto out;
1562 if (iblock >= extent_start && iblock < extent_end) {
1563 btrfs_map_bh_to_logical(root, result, blocknr +
1564 iblock - extent_start);
1565 goto out;
1566 }
1567 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1568 char *ptr;
1569 char *map;
1570 u32 size;
1571
1572 if (create & BTRFS_GET_BLOCK_NO_DIRECT) {
1573 err = -EINVAL;
1574 goto out;
1575 }
1576 size = btrfs_file_extent_inline_len(leaf->items +
1577 path->slots[0]);
1578 extent_end = (extent_start + size) >> inode->i_blkbits;
1579 extent_start >>= inode->i_blkbits;
1580 if (iblock < extent_start || iblock > extent_end) {
1581 goto not_found;
1582 }
1583 ptr = btrfs_file_extent_inline_start(item);
1584 map = kmap(result->b_page);
1585 memcpy(map, ptr, size);
1586 memset(map + size, 0, PAGE_CACHE_SIZE - size);
1587 flush_dcache_page(result->b_page);
1588 kunmap(result->b_page);
1589 set_buffer_uptodate(result);
1590 SetPageChecked(result->b_page);
1591 btrfs_map_bh_to_logical(root, result, 0);
1592 }
1593not_found:
1594 if (create & BTRFS_GET_BLOCK_CREATE) {
1595 struct btrfs_key ins;
1596 ret = btrfs_alloc_extent(trans, root, inode->i_ino,
6702ed49 1597 1, 0, alloc_hint, (u64)-1,
39279cc3 1598 &ins, 1);
54aa1f4d
CM		 1599 if (ret) {
1600 err = ret;
1601 goto out;
1602 }
39279cc3
CM		 1603 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
1604 iblock << inode->i_blkbits,
1605 ins.objectid, ins.offset,
1606 ins.offset);
54aa1f4d
CM		 1607 if (ret) {
1608 err = ret;
1609 goto out;
1610 }
39279cc3
CM		 1611 btrfs_map_bh_to_logical(root, result, ins.objectid);
1612 }
1613out:
54aa1f4d
CM		 1614 if (trans) {
1615 ret = btrfs_end_transaction(trans, root);
1616 if (!err)
1617 err = ret;
1618 }
39279cc3
CM		 1619 btrfs_free_path(path);
1620 return err;
1621}
1622
1623int btrfs_get_block(struct inode *inode, sector_t iblock,
1624 struct buffer_head *result, int create)
1625{
1626 int err;
1627 struct btrfs_root *root = BTRFS_I(inode)->root;
1628 mutex_lock(&root->fs_info->fs_mutex);
1629 err = btrfs_get_block_lock(inode, iblock, result, create);
1630 mutex_unlock(&root->fs_info->fs_mutex);
1631 return err;
1632}
1633
1634static int btrfs_get_block_bmap(struct inode *inode, sector_t iblock,
1635 struct buffer_head *result, int create)
1636{
1637 struct btrfs_root *root = BTRFS_I(inode)->root;
a52d9a80
CM		 1638 u64 start = iblock << inode->i_blkbits;
1639 u64 end = start + root->blocksize -1;
1640 struct extent_map *em;
1641
1642 em = btrfs_get_extent(inode, NULL, 0, start, end, 0);
1643 if (em && !IS_ERR(em) && em->block_start != EXTENT_MAP_INLINE &&
1644 em->block_start != 0) {
1645 u64 offset;
1646 offset = start - em->start;
1647 start = (em->block_start + offset) >> inode->i_blkbits;
1648 btrfs_map_bh_to_logical(root, result, start);
1649 }
39279cc3
CM		 1650 return 0;
1651}
1652
1653static sector_t btrfs_bmap(struct address_space *as, sector_t block)
1654{
1655 return generic_block_bmap(as, block, btrfs_get_block_bmap);
1656}
1657
1658static int btrfs_prepare_write(struct file *file, struct page *page,
1659 unsigned from, unsigned to)
1660{
a52d9a80
CM		 1661 return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree,
1662 page->mapping->host, page, from, to,
1663 btrfs_get_extent);
39279cc3
CM		 1664}
1665
a52d9a80 1666int btrfs_readpage(struct file *file, struct page *page)
9ebefb18 1667{
a52d9a80
CM		 1668 struct extent_map_tree *tree;
1669 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1670 return extent_read_full_page(tree, page, btrfs_get_extent);
9ebefb18 1671}
a52d9a80 1672static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
39279cc3 1673{
a52d9a80 1674 struct extent_map_tree *tree;
b888db2b
CM		 1675
1676
1677 if (current->flags & PF_MEMALLOC) {
1678 redirty_page_for_writepage(wbc, page);
1679 unlock_page(page);
1680 return 0;
1681 }
a52d9a80
CM		 1682 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1683 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
9ebefb18
CM		 1684}
1685
a52d9a80 1686static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
9ebefb18 1687{
a52d9a80
CM		 1688 struct extent_map_tree *tree;
1689 int ret;
8c2383c3 1690
a52d9a80
CM		 1691 if (page->private != 1) {
1692 WARN_ON(1);
1693 return try_to_free_buffers(page);
39279cc3 1694 }
a52d9a80
CM		 1695 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1696 ret = try_release_extent_mapping(tree, page);
1697 if (ret == 1) {
1698 ClearPagePrivate(page);
1699 set_page_private(page, 0);
1700 page_cache_release(page);
39279cc3 1701 }
a52d9a80 1702 return ret;
39279cc3
CM		 1703}
1704
a52d9a80 1705static void btrfs_invalidatepage(struct page *page, unsigned long offset)
39279cc3 1706{
a52d9a80 1707 struct extent_map_tree *tree;
39279cc3 1708
a52d9a80
CM		 1709 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1710 extent_invalidatepage(tree, page, offset);
1711 btrfs_releasepage(page, GFP_NOFS);
39279cc3
CM		 1712}
1713
9ebefb18
CM		 1714/*
1715 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
1716 * called from a page fault handler when a page is first dirtied. Hence we must
1717 * be careful to check for EOF conditions here. We set the page up correctly
1718 * for a written page which means we get ENOSPC checking when writing into
1719 * holes and correct delalloc and unwritten extent mapping on filesystems that
1720 * support these features.
1721 *
1722 * We are not allowed to take the i_mutex here so we have to play games to
1723 * protect against truncate races as the page could now be beyond EOF. Because
1724 * vmtruncate() writes the inode size before removing pages, once we have the
1725 * page lock we can determine safely if the page is beyond EOF. If it is not
1726 * beyond EOF, then the page is guaranteed safe against truncation until we
1727 * unlock the page.
1728 */
1729int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1730{
1731 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
1732 unsigned long end;
1733 loff_t size;
1734 int ret = -EINVAL;
a52d9a80 1735 u64 page_start;
9ebefb18
CM		 1736
1737 lock_page(page);
1738 wait_on_page_writeback(page);
9ebefb18 1739 size = i_size_read(inode);
a52d9a80
CM		 1740 page_start = page->index << PAGE_CACHE_SHIFT;
1741
9ebefb18 1742 if ((page->mapping != inode->i_mapping) ||
a52d9a80 1743 (page_start > size)) {
9ebefb18
CM		 1744 /* page got truncated out from underneath us */
1745 goto out_unlock;
1746 }
1747
1748 /* page is wholly or partially inside EOF */
a52d9a80 1749 if (page_start + PAGE_CACHE_SIZE > size)
9ebefb18
CM		 1750 end = size & ~PAGE_CACHE_MASK;
1751 else
1752 end = PAGE_CACHE_SIZE;
1753
b888db2b 1754 ret = btrfs_cow_one_page(inode, page, end);
9ebefb18
CM		 1755
1756out_unlock:
1757 unlock_page(page);
1758 return ret;
1759}
1760
39279cc3
CM		 1761static void btrfs_truncate(struct inode *inode)
1762{
1763 struct btrfs_root *root = BTRFS_I(inode)->root;
1764 int ret;
1765 struct btrfs_trans_handle *trans;
1766
1767 if (!S_ISREG(inode->i_mode))
1768 return;
1769 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1770 return;
1771
1772 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1773
1774 mutex_lock(&root->fs_info->fs_mutex);
1775 trans = btrfs_start_transaction(root, 1);
1776 btrfs_set_trans_block_group(trans, inode);
1777
1778 /* FIXME, add redo link to tree so we don't leak on crash */
1779 ret = btrfs_truncate_in_trans(trans, root, inode);
39279cc3
CM		 1780 btrfs_update_inode(trans, root, inode);
1781 ret = btrfs_end_transaction(trans, root);
1782 BUG_ON(ret);
1783 mutex_unlock(&root->fs_info->fs_mutex);
1784 btrfs_btree_balance_dirty(root);
1785}
1786
1787int btrfs_commit_write(struct file *file, struct page *page,
1788 unsigned from, unsigned to)
1789{
a52d9a80
CM		 1790 return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree,
1791 page->mapping->host, page, from, to);
39279cc3
CM		 1792}
1793
1794static int create_subvol(struct btrfs_root *root, char *name, int namelen)
1795{
1796 struct btrfs_trans_handle *trans;
1797 struct btrfs_key key;
1798 struct btrfs_root_item root_item;
1799 struct btrfs_inode_item *inode_item;
1800 struct buffer_head *subvol;
1801 struct btrfs_leaf *leaf;
1802 struct btrfs_root *new_root;
1803 struct inode *inode;
1804 struct inode *dir;
1805 int ret;
54aa1f4d 1806 int err;
39279cc3
CM		 1807 u64 objectid;
1808 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
1809
1810 mutex_lock(&root->fs_info->fs_mutex);
1811 trans = btrfs_start_transaction(root, 1);
1812 BUG_ON(!trans);
1813
6702ed49 1814 subvol = btrfs_alloc_free_block(trans, root, 0, 0);
54aa1f4d
CM		 1815 if (IS_ERR(subvol))
1816 return PTR_ERR(subvol);
39279cc3
CM		 1817 leaf = btrfs_buffer_leaf(subvol);
1818 btrfs_set_header_nritems(&leaf->header, 0);
1819 btrfs_set_header_level(&leaf->header, 0);
1820 btrfs_set_header_blocknr(&leaf->header, bh_blocknr(subvol));
1821 btrfs_set_header_generation(&leaf->header, trans->transid);
1822 btrfs_set_header_owner(&leaf->header, root->root_key.objectid);
1823 memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid,
1824 sizeof(leaf->header.fsid));
ccd467d6 1825 btrfs_mark_buffer_dirty(subvol);
39279cc3
CM		 1826
1827 inode_item = &root_item.inode;
1828 memset(inode_item, 0, sizeof(*inode_item));
1829 btrfs_set_inode_generation(inode_item, 1);
1830 btrfs_set_inode_size(inode_item, 3);
1831 btrfs_set_inode_nlink(inode_item, 1);
1832 btrfs_set_inode_nblocks(inode_item, 1);
1833 btrfs_set_inode_mode(inode_item, S_IFDIR | 0755);
1834
1835 btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol));
1836 btrfs_set_root_refs(&root_item, 1);
58176a96 1837 btrfs_set_root_blocks_used(&root_item, 0);
5eda7b5e
CM		 1838 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
1839 root_item.drop_level = 0;
39279cc3
CM		 1840 brelse(subvol);
1841 subvol = NULL;
1842
1843 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1844 0, &objectid);
54aa1f4d
CM		 1845 if (ret)
1846 goto fail;
39279cc3
CM		 1847
1848 btrfs_set_root_dirid(&root_item, new_dirid);
1849
1850 key.objectid = objectid;
1851 key.offset = 1;
1852 key.flags = 0;
1853 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1854 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1855 &root_item);
54aa1f4d
CM		 1856 if (ret)
1857 goto fail;
39279cc3
CM		 1858
1859 /*
1860 * insert the directory item
1861 */
1862 key.offset = (u64)-1;
1863 dir = root->fs_info->sb->s_root->d_inode;
1864 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1865 name, namelen, dir->i_ino, &key,
1866 BTRFS_FT_DIR);
54aa1f4d
CM		 1867 if (ret)
1868 goto fail;
39279cc3
CM		 1869
1870 ret = btrfs_commit_transaction(trans, root);
54aa1f4d
CM		 1871 if (ret)
1872 goto fail_commit;
39279cc3 1873
58176a96 1874 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
39279cc3
CM		 1875 BUG_ON(!new_root);
1876
1877 trans = btrfs_start_transaction(new_root, 1);
1878 BUG_ON(!trans);
1879
1880 inode = btrfs_new_inode(trans, new_root, new_dirid,
1881 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
54aa1f4d
CM		 1882 if (IS_ERR(inode))
1883 goto fail;
39279cc3
CM		 1884 inode->i_op = &btrfs_dir_inode_operations;
1885 inode->i_fop = &btrfs_dir_file_operations;
34088780 1886 new_root->inode = inode;
39279cc3
CM		 1887
1888 ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid);
54aa1f4d
CM		 1889 if (ret)
1890 goto fail;
39279cc3
CM		 1891
1892 inode->i_nlink = 1;
1893 inode->i_size = 6;
1894 ret = btrfs_update_inode(trans, new_root, inode);
54aa1f4d
CM		 1895 if (ret)
1896 goto fail;
1897fail:
1898 err = btrfs_commit_transaction(trans, root);
1899 if (err && !ret)
1900 ret = err;
1901fail_commit:
39279cc3
CM		 1902 mutex_unlock(&root->fs_info->fs_mutex);
1903 btrfs_btree_balance_dirty(root);
54aa1f4d 1904 return ret;
39279cc3
CM		 1905}
1906
1907static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
1908{
1909 struct btrfs_trans_handle *trans;
1910 struct btrfs_key key;
1911 struct btrfs_root_item new_root_item;
83df7c1d 1912 struct buffer_head *tmp;
39279cc3 1913 int ret;
54aa1f4d 1914 int err;
39279cc3
CM		 1915 u64 objectid;
1916
1917 if (!root->ref_cows)
1918 return -EINVAL;
1919
1920 mutex_lock(&root->fs_info->fs_mutex);
1921 trans = btrfs_start_transaction(root, 1);
1922 BUG_ON(!trans);
1923
1924 ret = btrfs_update_inode(trans, root, root->inode);
54aa1f4d
CM		 1925 if (ret)
1926 goto fail;
39279cc3
CM		 1927
1928 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1929 0, &objectid);
54aa1f4d
CM		 1930 if (ret)
1931 goto fail;
39279cc3
CM		 1932
1933 memcpy(&new_root_item, &root->root_item,
1934 sizeof(new_root_item));
1935
1936 key.objectid = objectid;
1937 key.offset = 1;
1938 key.flags = 0;
1939 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
83df7c1d 1940 btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
39279cc3
CM		 1941 btrfs_set_root_blocknr(&new_root_item, bh_blocknr(root->node));
1942
1943 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1944 &new_root_item);
54aa1f4d
CM		 1945 if (ret)
1946 goto fail;
39279cc3
CM		 1947
1948 /*
1949 * insert the directory item
1950 */
1951 key.offset = (u64)-1;
1952 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1953 name, namelen,
1954 root->fs_info->sb->s_root->d_inode->i_ino,
1955 &key, BTRFS_FT_DIR);
1956
54aa1f4d
CM		 1957 if (ret)
1958 goto fail;
39279cc3
CM		 1959
1960 ret = btrfs_inc_root_ref(trans, root);
54aa1f4d
CM		 1961 if (ret)
1962 goto fail;
39279cc3 1963
54aa1f4d
CM		 1964fail:
1965 err = btrfs_commit_transaction(trans, root);
1966 if (err && !ret)
1967 ret = err;
39279cc3
CM		 1968 mutex_unlock(&root->fs_info->fs_mutex);
1969 btrfs_btree_balance_dirty(root);
54aa1f4d 1970 return ret;
39279cc3
CM		 1971}
1972
1973int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int
1974 cmd, unsigned long arg)
1975{
1976 struct btrfs_root *root = BTRFS_I(inode)->root;
1977 struct btrfs_ioctl_vol_args vol_args;
1978 int ret = 0;
1979 struct btrfs_dir_item *di;
1980 int namelen;
1981 struct btrfs_path *path;
1982 u64 root_dirid;
1983
1984 switch (cmd) {
1985 case BTRFS_IOC_SNAP_CREATE:
1986 if (copy_from_user(&vol_args,
1987 (struct btrfs_ioctl_vol_args __user *)arg,
1988 sizeof(vol_args)))
1989 return -EFAULT;
1990 namelen = strlen(vol_args.name);
1991 if (namelen > BTRFS_VOL_NAME_MAX)
1992 return -EINVAL;
8a712645
CM		 1993 if (strchr(vol_args.name, '/'))
1994 return -EINVAL;
39279cc3
CM		 1995 path = btrfs_alloc_path();
1996 if (!path)
1997 return -ENOMEM;
1998 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
1999 mutex_lock(&root->fs_info->fs_mutex);
2000 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2001 path, root_dirid,
2002 vol_args.name, namelen, 0);
2003 mutex_unlock(&root->fs_info->fs_mutex);
2004 btrfs_free_path(path);
2005 if (di && !IS_ERR(di))
2006 return -EEXIST;
54aa1f4d
CM		 2007 if (IS_ERR(di))
2008 return PTR_ERR(di);
39279cc3
CM		 2009
2010 if (root == root->fs_info->tree_root)
2011 ret = create_subvol(root, vol_args.name, namelen);
2012 else
2013 ret = create_snapshot(root, vol_args.name, namelen);
39279cc3 2014 break;
6702ed49
CM		 2015
2016 case BTRFS_IOC_DEFRAG:
2017 mutex_lock(&root->fs_info->fs_mutex);
e9d0b13b
CM		 2018 btrfs_defrag_root(root, 0);
2019 btrfs_defrag_root(root->fs_info->extent_root, 0);
6702ed49
CM		 2020 mutex_unlock(&root->fs_info->fs_mutex);
2021 ret = 0;
2022 break;
39279cc3
CM		 2023 default:
2024 return -ENOTTY;
2025 }
2026 return ret;
2027}
2028
2029#ifdef CONFIG_COMPAT
2030long btrfs_compat_ioctl(struct file *file, unsigned int cmd,
2031 unsigned long arg)
2032{
2033 struct inode *inode = file->f_path.dentry->d_inode;
2034 int ret;
2035 lock_kernel();
2036 ret = btrfs_ioctl(inode, file, cmd, (unsigned long) compat_ptr(arg));
2037 unlock_kernel();
2038 return ret;
2039
2040}
2041#endif
2042
2043/*
2044 * Called inside transaction, so use GFP_NOFS
2045 */
2046struct inode *btrfs_alloc_inode(struct super_block *sb)
2047{
2048 struct btrfs_inode *ei;
2049
2050 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2051 if (!ei)
2052 return NULL;
15ee9bc7 2053 ei->last_trans = 0;
39279cc3
CM		 2054 return &ei->vfs_inode;
2055}
2056
2057void btrfs_destroy_inode(struct inode *inode)
2058{
2059 WARN_ON(!list_empty(&inode->i_dentry));
2060 WARN_ON(inode->i_data.nrpages);
2061
2062 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2063}
2064
2065static void init_once(void * foo, struct kmem_cache * cachep,
2066 unsigned long flags)
2067{
2068 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2069
2070 inode_init_once(&ei->vfs_inode);
2071}
2072
2073void btrfs_destroy_cachep(void)
2074{
2075 if (btrfs_inode_cachep)
2076 kmem_cache_destroy(btrfs_inode_cachep);
2077 if (btrfs_trans_handle_cachep)
2078 kmem_cache_destroy(btrfs_trans_handle_cachep);
2079 if (btrfs_transaction_cachep)
2080 kmem_cache_destroy(btrfs_transaction_cachep);
2081 if (btrfs_bit_radix_cachep)
2082 kmem_cache_destroy(btrfs_bit_radix_cachep);
2083 if (btrfs_path_cachep)
2084 kmem_cache_destroy(btrfs_path_cachep);
2085}
2086
92fee66d
CM		 2087static struct kmem_cache *cache_create(const char *name, size_t size,
2088 unsigned long extra_flags,
2089 void (*ctor)(void *, struct kmem_cache *,
2090 unsigned long))
2091{
2092 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2093 SLAB_MEM_SPREAD | extra_flags), ctor
2094#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2095 ,NULL
2096#endif
2097 );
2098}
2099
39279cc3
CM		 2100int btrfs_init_cachep(void)
2101{
92fee66d
CM		 2102 btrfs_inode_cachep = cache_create("btrfs_inode_cache",
2103 sizeof(struct btrfs_inode),
2104 0, init_once);
39279cc3
CM		 2105 if (!btrfs_inode_cachep)
2106 goto fail;
92fee66d 2107 btrfs_trans_handle_cachep = cache_create("btrfs_trans_handle_cache",
39279cc3 2108 sizeof(struct btrfs_trans_handle),
92fee66d 2109 0, NULL);
39279cc3
CM		 2110 if (!btrfs_trans_handle_cachep)
2111 goto fail;
92fee66d 2112 btrfs_transaction_cachep = cache_create("btrfs_transaction_cache",
39279cc3 2113 sizeof(struct btrfs_transaction),
92fee66d 2114 0, NULL);
39279cc3
CM		 2115 if (!btrfs_transaction_cachep)
2116 goto fail;
92fee66d
CM		 2117 btrfs_path_cachep = cache_create("btrfs_path_cache",
2118 sizeof(struct btrfs_transaction),
2119 0, NULL);
39279cc3
CM		 2120 if (!btrfs_path_cachep)
2121 goto fail;
92fee66d
CM		 2122 btrfs_bit_radix_cachep = cache_create("btrfs_radix", 256,
2123 SLAB_DESTROY_BY_RCU, NULL);
39279cc3
CM		 2124 if (!btrfs_bit_radix_cachep)
2125 goto fail;
2126 return 0;
2127fail:
2128 btrfs_destroy_cachep();
2129 return -ENOMEM;
2130}
2131
2132static int btrfs_getattr(struct vfsmount *mnt,
2133 struct dentry *dentry, struct kstat *stat)
2134{
2135 struct inode *inode = dentry->d_inode;
2136 generic_fillattr(inode, stat);
2137 stat->blksize = 256 * 1024;
2138 return 0;
2139}
2140
2141static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2142 struct inode * new_dir,struct dentry *new_dentry)
2143{
2144 struct btrfs_trans_handle *trans;
2145 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2146 struct inode *new_inode = new_dentry->d_inode;
2147 struct inode *old_inode = old_dentry->d_inode;
2148 struct timespec ctime = CURRENT_TIME;
2149 struct btrfs_path *path;
2150 struct btrfs_dir_item *di;
2151 int ret;
2152
2153 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2154 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2155 return -ENOTEMPTY;
2156 }
2157 mutex_lock(&root->fs_info->fs_mutex);
2158 trans = btrfs_start_transaction(root, 1);
2159 btrfs_set_trans_block_group(trans, new_dir);
2160 path = btrfs_alloc_path();
2161 if (!path) {
2162 ret = -ENOMEM;
2163 goto out_fail;
2164 }
2165
2166 old_dentry->d_inode->i_nlink++;
2167 old_dir->i_ctime = old_dir->i_mtime = ctime;
2168 new_dir->i_ctime = new_dir->i_mtime = ctime;
2169 old_inode->i_ctime = ctime;
2170 if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) {
2171 struct btrfs_key *location = &BTRFS_I(new_dir)->location;
2172 u64 old_parent_oid;
2173 di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino,
2174 "..", 2, -1);
2175 if (IS_ERR(di)) {
2176 ret = PTR_ERR(di);
2177 goto out_fail;
2178 }
2179 if (!di) {
2180 ret = -ENOENT;
2181 goto out_fail;
2182 }
2183 old_parent_oid = btrfs_disk_key_objectid(&di->location);
2184 ret = btrfs_del_item(trans, root, path);
2185 if (ret) {
39279cc3
CM		 2186 goto out_fail;
2187 }
2188 btrfs_release_path(root, path);
2189
2190 di = btrfs_lookup_dir_index_item(trans, root, path,
2191 old_inode->i_ino,
2192 old_parent_oid,
2193 "..", 2, -1);
2194 if (IS_ERR(di)) {
2195 ret = PTR_ERR(di);
2196 goto out_fail;
2197 }
2198 if (!di) {
2199 ret = -ENOENT;
2200 goto out_fail;
2201 }
2202 ret = btrfs_del_item(trans, root, path);
2203 if (ret) {
39279cc3
CM		 2204 goto out_fail;
2205 }
2206 btrfs_release_path(root, path);
2207
2208 ret = btrfs_insert_dir_item(trans, root, "..", 2,
2209 old_inode->i_ino, location,
2210 BTRFS_FT_DIR);
2211 if (ret)
2212 goto out_fail;
2213 }
2214
2215
2216 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2217 if (ret)
2218 goto out_fail;
2219
2220 if (new_inode) {
2221 new_inode->i_ctime = CURRENT_TIME;
2222 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2223 if (ret)
2224 goto out_fail;
2225 if (S_ISDIR(new_inode->i_mode))
2226 clear_nlink(new_inode);
2227 else
2228 drop_nlink(new_inode);
54aa1f4d
CM		 2229 ret = btrfs_update_inode(trans, root, new_inode);
2230 if (ret)
2231 goto out_fail;
39279cc3
CM		 2232 }
2233 ret = btrfs_add_link(trans, new_dentry, old_inode);
2234 if (ret)
2235 goto out_fail;
2236
2237out_fail:
2238 btrfs_free_path(path);
2239 btrfs_end_transaction(trans, root);
2240 mutex_unlock(&root->fs_info->fs_mutex);
2241 return ret;
2242}
2243
2244static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2245 const char *symname)
2246{
2247 struct btrfs_trans_handle *trans;
2248 struct btrfs_root *root = BTRFS_I(dir)->root;
2249 struct btrfs_path *path;
2250 struct btrfs_key key;
2251 struct inode *inode;
2252 int err;
2253 int drop_inode = 0;
2254 u64 objectid;
2255 int name_len;
2256 int datasize;
2257 char *ptr;
2258 struct btrfs_file_extent_item *ei;
2259
2260 name_len = strlen(symname) + 1;
2261 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2262 return -ENAMETOOLONG;
2263 mutex_lock(&root->fs_info->fs_mutex);
2264 trans = btrfs_start_transaction(root, 1);
2265 btrfs_set_trans_block_group(trans, dir);
2266
2267 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2268 if (err) {
2269 err = -ENOSPC;
2270 goto out_unlock;
2271 }
2272
2273 inode = btrfs_new_inode(trans, root, objectid,
2274 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2275 err = PTR_ERR(inode);
2276 if (IS_ERR(inode))
2277 goto out_unlock;
2278
2279 btrfs_set_trans_block_group(trans, inode);
2280 err = btrfs_add_nondir(trans, dentry, inode);
2281 if (err)
2282 drop_inode = 1;
2283 else {
2284 inode->i_mapping->a_ops = &btrfs_aops;
2285 inode->i_fop = &btrfs_file_operations;
2286 inode->i_op = &btrfs_file_inode_operations;
a52d9a80
CM		 2287 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
2288 inode->i_mapping, GFP_NOFS);
b888db2b 2289 BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range;
39279cc3
CM		 2290 }
2291 dir->i_sb->s_dirt = 1;
2292 btrfs_update_inode_block_group(trans, inode);
2293 btrfs_update_inode_block_group(trans, dir);
2294 if (drop_inode)
2295 goto out_unlock;
2296
2297 path = btrfs_alloc_path();
2298 BUG_ON(!path);
2299 key.objectid = inode->i_ino;
2300 key.offset = 0;
2301 key.flags = 0;
2302 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2303 datasize = btrfs_file_extent_calc_inline_size(name_len);
2304 err = btrfs_insert_empty_item(trans, root, path, &key,
2305 datasize);
54aa1f4d
CM		 2306 if (err) {
2307 drop_inode = 1;
2308 goto out_unlock;
2309 }
39279cc3
CM		 2310 ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
2311 path->slots[0], struct btrfs_file_extent_item);
2312 btrfs_set_file_extent_generation(ei, trans->transid);
2313 btrfs_set_file_extent_type(ei,
2314 BTRFS_FILE_EXTENT_INLINE);
2315 ptr = btrfs_file_extent_inline_start(ei);
2316 btrfs_memcpy(root, path->nodes[0]->b_data,
2317 ptr, symname, name_len);
ccd467d6 2318 btrfs_mark_buffer_dirty(path->nodes[0]);
39279cc3
CM		 2319 btrfs_free_path(path);
2320 inode->i_op = &btrfs_symlink_inode_operations;
2321 inode->i_mapping->a_ops = &btrfs_symlink_aops;
2322 inode->i_size = name_len - 1;
54aa1f4d
CM		 2323 err = btrfs_update_inode(trans, root, inode);
2324 if (err)
2325 drop_inode = 1;
39279cc3
CM		 2326
2327out_unlock:
2328 btrfs_end_transaction(trans, root);
2329 mutex_unlock(&root->fs_info->fs_mutex);
39279cc3
CM		 2330 if (drop_inode) {
2331 inode_dec_link_count(inode);
2332 iput(inode);
2333 }
2334 btrfs_btree_balance_dirty(root);
2335 return err;
2336}
2337
2338static struct inode_operations btrfs_dir_inode_operations = {
2339 .lookup = btrfs_lookup,
2340 .create = btrfs_create,
2341 .unlink = btrfs_unlink,
2342 .link = btrfs_link,
2343 .mkdir = btrfs_mkdir,
2344 .rmdir = btrfs_rmdir,
2345 .rename = btrfs_rename,
2346 .symlink = btrfs_symlink,
2347 .setattr = btrfs_setattr,
618e21d5 2348 .mknod = btrfs_mknod,
39279cc3
CM		 2349};
2350
2351static struct inode_operations btrfs_dir_ro_inode_operations = {
2352 .lookup = btrfs_lookup,
2353};
2354
2355static struct file_operations btrfs_dir_file_operations = {
2356 .llseek = generic_file_llseek,
2357 .read = generic_read_dir,
2358 .readdir = btrfs_readdir,
2359 .ioctl = btrfs_ioctl,
2360#ifdef CONFIG_COMPAT
2361 .compat_ioctl = btrfs_compat_ioctl,
2362#endif
2363};
2364
2365static struct address_space_operations btrfs_aops = {
2366 .readpage = btrfs_readpage,
2367 .writepage = btrfs_writepage,
2368 .sync_page = block_sync_page,
2369 .prepare_write = btrfs_prepare_write,
2370 .commit_write = btrfs_commit_write,
2371 .bmap = btrfs_bmap,
a52d9a80
CM		 2372 .invalidatepage = btrfs_invalidatepage,
2373 .releasepage = btrfs_releasepage,
2374 .set_page_dirty = __set_page_dirty_nobuffers,
39279cc3
CM		 2375};
2376
2377static struct address_space_operations btrfs_symlink_aops = {
2378 .readpage = btrfs_readpage,
2379 .writepage = btrfs_writepage,
2380};
2381
2382static struct inode_operations btrfs_file_inode_operations = {
2383 .truncate = btrfs_truncate,
2384 .getattr = btrfs_getattr,
2385 .setattr = btrfs_setattr,
2386};
2387
618e21d5
JB		 2388static struct inode_operations btrfs_special_inode_operations = {
2389 .getattr = btrfs_getattr,
2390 .setattr = btrfs_setattr,
2391};
2392
39279cc3
CM		 2393static struct inode_operations btrfs_symlink_inode_operations = {
2394 .readlink = generic_readlink,
2395 .follow_link = page_follow_link_light,
2396 .put_link = page_put_link,
2397};
Clone of davem's net-next-2.6 tree