]> 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: Avoid extent_buffer lru corruption
[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;
07157aac 55static struct extent_map_ops btrfs_extent_map_ops;
39279cc3
CM		 56
57static struct kmem_cache *btrfs_inode_cachep;
58struct kmem_cache *btrfs_trans_handle_cachep;
59struct kmem_cache *btrfs_transaction_cachep;
60struct kmem_cache *btrfs_bit_radix_cachep;
61struct kmem_cache *btrfs_path_cachep;
62
63#define S_SHIFT 12
64static 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
b888db2b
CM		 74static 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;
db94535d 80 u64 num_bytes;
b888db2b 81 int ret;
db94535d 82 u64 blocksize = root->sectorsize;
b888db2b
CM		 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);
db94535d 88 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
b888db2b 89 ret = btrfs_drop_extents(trans, root, inode,
3326d1b0 90 start, start + num_bytes, start, &alloc_hint);
db94535d 91
179e29e4
CM		 92 if (alloc_hint == EXTENT_MAP_INLINE)
93 goto out;
94
db94535d 95 ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_bytes, 0,
b888db2b
CM		 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);
104out:
105 btrfs_end_transaction(trans, root);
106 mutex_unlock(&root->fs_info->fs_mutex);
107 return ret;
108}
109
07157aac
CM		 110int 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;
35ebb934 117 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
07157aac
CM		 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);
f578d4bd 124 btrfs_csum_file_block(trans, root, inode, inode->i_ino,
07157aac
CM		 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
133int 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;
ff79f819 141 u32 csum;
07157aac
CM		 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;
ff79f819 151 csum = 0;
07157aac
CM		 152 goto out;
153 }
ff79f819
CM		 154 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
155 BTRFS_CRC32_SIZE);
156 set_state_private(em_tree, start, csum);
07157aac
CM		 157out:
158 if (path)
159 btrfs_free_path(path);
160 mutex_unlock(&root->fs_info->fs_mutex);
161 return ret;
162}
163
164int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
165{
35ebb934 166 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
07157aac 167 struct inode *inode = page->mapping->host;
07157aac
CM		 168 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
169 char *kaddr;
170 u64 private;
171 int ret;
ff79f819
CM		 172 struct btrfs_root *root = BTRFS_I(inode)->root;
173 u32 csum = ~(u32)0;
bbf0d006 174 unsigned long flags;
07157aac
CM		 175
176 ret = get_state_private(em_tree, start, &private);
bbf0d006 177 local_irq_save(flags);
07157aac
CM		 178 kaddr = kmap_atomic(page, KM_IRQ0);
179 if (ret) {
180 goto zeroit;
181 }
ff79f819
CM		 182 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
183 btrfs_csum_final(csum, (char *)&csum);
184 if (csum != private) {
07157aac
CM		 185 goto zeroit;
186 }
187 kunmap_atomic(kaddr, KM_IRQ0);
bbf0d006 188 local_irq_restore(flags);
07157aac
CM		 189 return 0;
190
191zeroit:
192 printk("btrfs csum failed ino %lu off %llu\n",
193 page->mapping->host->i_ino, (unsigned long long)start);
db94535d
CM		 194 memset(kaddr + offset, 1, end - start + 1);
195 flush_dcache_page(page);
07157aac 196 kunmap_atomic(kaddr, KM_IRQ0);
bbf0d006 197 local_irq_restore(flags);
07157aac
CM		 198 return 0;
199}
b888db2b 200
39279cc3
CM		 201void btrfs_read_locked_inode(struct inode *inode)
202{
203 struct btrfs_path *path;
5f39d397 204 struct extent_buffer *leaf;
39279cc3 205 struct btrfs_inode_item *inode_item;
5f39d397 206 struct btrfs_inode_timespec *tspec;
39279cc3
CM		 207 struct btrfs_root *root = BTRFS_I(inode)->root;
208 struct btrfs_key location;
209 u64 alloc_group_block;
618e21d5 210 u32 rdev;
39279cc3
CM		 211 int ret;
212
213 path = btrfs_alloc_path();
214 BUG_ON(!path);
39279cc3
CM		 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);
5f39d397 219 if (ret)
39279cc3 220 goto make_bad;
39279cc3 221
5f39d397
CM		 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);
618e21d5 246 inode->i_rdev = 0;
5f39d397
CM		 247 rdev = btrfs_inode_rdev(leaf, inode_item);
248
249 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
39279cc3
CM		 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) {
39279cc3
CM		 259 case S_IFREG:
260 inode->i_mapping->a_ops = &btrfs_aops;
07157aac 261 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
39279cc3
CM		 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;
618e21d5
JB		 276 default:
277 init_special_inode(inode, inode->i_mode, rdev);
278 break;
39279cc3
CM		 279 }
280 return;
281
282make_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
5f39d397
CM		 289static void fill_inode_item(struct extent_buffer *leaf,
290 struct btrfs_inode_item *item,
39279cc3
CM		 291 struct inode *inode)
292{
5f39d397
CM		 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,
39279cc3
CM		 318 BTRFS_I(inode)->block_group->key.objectid);
319}
320
a52d9a80 321int btrfs_update_inode(struct btrfs_trans_handle *trans,
39279cc3
CM		 322 struct btrfs_root *root,
323 struct inode *inode)
324{
325 struct btrfs_inode_item *inode_item;
326 struct btrfs_path *path;
5f39d397 327 struct extent_buffer *leaf;
39279cc3
CM		 328 int ret;
329
330 path = btrfs_alloc_path();
331 BUG_ON(!path);
39279cc3
CM		 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
5f39d397
CM		 340 leaf = path->nodes[0];
341 inode_item = btrfs_item_ptr(leaf, path->slots[0],
39279cc3
CM		 342 struct btrfs_inode_item);
343
5f39d397
CM		 344 fill_inode_item(leaf, inode_item, inode);
345 btrfs_mark_buffer_dirty(leaf);
15ee9bc7 346 btrfs_set_inode_last_trans(trans, inode);
39279cc3
CM		 347 ret = 0;
348failed:
349 btrfs_release_path(root, path);
350 btrfs_free_path(path);
351 return ret;
352}
353
354
355static 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;
5f39d397 364 struct extent_buffer *leaf;
39279cc3 365 struct btrfs_dir_item *di;
5f39d397 366 struct btrfs_key key;
39279cc3
CM		 367
368 path = btrfs_alloc_path();
54aa1f4d
CM		 369 if (!path) {
370 ret = -ENOMEM;
371 goto err;
372 }
373
39279cc3
CM		 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 }
5f39d397
CM		 384 leaf = path->nodes[0];
385 btrfs_dir_item_key_to_cpu(leaf, di, &key);
39279cc3 386 ret = btrfs_delete_one_dir_name(trans, root, path, di);
54aa1f4d
CM		 387 if (ret)
388 goto err;
39279cc3
CM		 389 btrfs_release_path(root, path);
390
391 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
5f39d397 392 key.objectid, name, name_len, -1);
39279cc3
CM		 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);
39279cc3
CM		 402
403 dentry->d_inode->i_ctime = dir->i_ctime;
404err:
405 btrfs_free_path(path);
406 if (!ret) {
407 dir->i_size -= name_len * 2;
79c44584 408 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
39279cc3
CM		 409 btrfs_update_inode(trans, root, dir);
410 drop_nlink(dentry->d_inode);
54aa1f4d 411 ret = btrfs_update_inode(trans, root, dentry->d_inode);
39279cc3
CM		 412 dir->i_sb->s_dirt = 1;
413 }
414 return ret;
415}
416
417static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
418{
419 struct btrfs_root *root;
420 struct btrfs_trans_handle *trans;
421 int ret;
d3c2fdcf 422 unsigned long nr;
39279cc3
CM		 423
424 root = BTRFS_I(dir)->root;
425 mutex_lock(&root->fs_info->fs_mutex);
426 trans = btrfs_start_transaction(root, 1);
5f39d397 427
39279cc3
CM		 428 btrfs_set_trans_block_group(trans, dir);
429 ret = btrfs_unlink_trans(trans, root, dir, dentry);
d3c2fdcf 430 nr = trans->blocks_used;
5f39d397 431
39279cc3
CM		 432 btrfs_end_transaction(trans, root);
433 mutex_unlock(&root->fs_info->fs_mutex);
d3c2fdcf 434 btrfs_btree_balance_dirty(root, nr);
5f39d397 435
39279cc3
CM		 436 return ret;
437}
438
439static 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;
5f39d397 450 struct extent_buffer *leaf;
39279cc3 451 char *goodnames = "..";
d3c2fdcf 452 unsigned long nr;
39279cc3 453
134d4512
Y		 454 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
455 return -ENOTEMPTY;
456
39279cc3
CM		 457 path = btrfs_alloc_path();
458 BUG_ON(!path);
39279cc3
CM		 459 mutex_lock(&root->fs_info->fs_mutex);
460 trans = btrfs_start_transaction(root, 1);
5f39d397 461
39279cc3
CM		 462 btrfs_set_trans_block_group(trans, dir);
463 key.objectid = inode->i_ino;
464 key.offset = (u64)-1;
5f39d397 465 key.type = (u8)-1;
39279cc3
CM		 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]--;
5f39d397
CM		 478 leaf = path->nodes[0];
479 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
39279cc3
CM		 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 }
507out:
508 btrfs_release_path(root, path);
509 btrfs_free_path(path);
d3c2fdcf 510 nr = trans->blocks_used;
39279cc3 511 ret = btrfs_end_transaction(trans, root);
134d4512 512 mutex_unlock(&root->fs_info->fs_mutex);
d3c2fdcf 513 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM		 514 if (ret && !err)
515 err = ret;
516 return err;
517}
518
519static 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);
39279cc3
CM		 530 ret = btrfs_lookup_inode(trans, root, path,
531 &BTRFS_I(inode)->location, -1);
54aa1f4d
CM		 532 if (ret > 0)
533 ret = -ENOENT;
534 if (!ret)
535 ret = btrfs_del_item(trans, root, path);
39279cc3
CM		 536 btrfs_free_path(path);
537 return ret;
538}
539
39279cc3
CM		 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 */
548static 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;
5f39d397 555 struct btrfs_key found_key;
39279cc3 556 u32 found_type;
5f39d397 557 struct extent_buffer *leaf;
39279cc3
CM		 558 struct btrfs_file_extent_item *fi;
559 u64 extent_start = 0;
db94535d 560 u64 extent_num_bytes = 0;
39279cc3
CM		 561 u64 item_end = 0;
562 int found_extent;
563 int del_item;
179e29e4 564 int extent_type = -1;
39279cc3 565
a52d9a80 566 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
39279cc3 567 path = btrfs_alloc_path();
3c69faec 568 path->reada = -1;
39279cc3 569 BUG_ON(!path);
5f39d397 570
39279cc3
CM		 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;
5f39d397
CM		 574 key.type = (u8)-1;
575
39279cc3
CM		 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 }
5f39d397
CM		 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);
39279cc3 590
5f39d397 591 if (found_key.objectid != inode->i_ino)
39279cc3 592 break;
5f39d397 593
39279cc3
CM		 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
5f39d397 600 item_end = found_key.offset;
39279cc3 601 if (found_type == BTRFS_EXTENT_DATA_KEY) {
5f39d397 602 fi = btrfs_item_ptr(leaf, path->slots[0],
39279cc3 603 struct btrfs_file_extent_item);
179e29e4
CM		 604 extent_type = btrfs_file_extent_type(leaf, fi);
605 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
5f39d397 606 item_end +=
db94535d 607 btrfs_file_extent_num_bytes(leaf, fi);
179e29e4
CM		 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);
39279cc3 613 }
008630c1 614 item_end--;
39279cc3
CM		 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 }
008630c1 621 if (item_end < inode->i_size) {
b888db2b
CM		 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;
39279cc3 630 }
a61721d5 631 btrfs_set_key_type(&key, found_type);
65555a06 632 btrfs_release_path(root, path);
b888db2b 633 continue;
39279cc3 634 }
5f39d397 635 if (found_key.offset >= inode->i_size)
39279cc3
CM		 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 */
179e29e4
CM		 642 if (found_type != BTRFS_EXTENT_DATA_KEY)
643 goto delete;
644
645 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
39279cc3 646 u64 num_dec;
db94535d 647 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
39279cc3 648 if (!del_item) {
db94535d
CM		 649 u64 orig_num_bytes =
650 btrfs_file_extent_num_bytes(leaf, fi);
651 extent_num_bytes = inode->i_size -
5f39d397 652 found_key.offset + root->sectorsize - 1;
db94535d
CM		 653 btrfs_set_file_extent_num_bytes(leaf, fi,
654 extent_num_bytes);
655 num_dec = (orig_num_bytes -
656 extent_num_bytes) >> 9;
bab9fb03
Y		 657 if (extent_start != 0) {
658 inode->i_blocks -= num_dec;
659 }
5f39d397 660 btrfs_mark_buffer_dirty(leaf);
39279cc3 661 } else {
db94535d
CM		 662 extent_num_bytes =
663 btrfs_file_extent_disk_num_bytes(leaf,
664 fi);
39279cc3 665 /* FIXME blocksize != 4096 */
db94535d
CM		 666 num_dec = btrfs_file_extent_num_bytes(leaf,
667 fi) >> 9;
39279cc3
CM		 668 if (extent_start != 0) {
669 found_extent = 1;
670 inode->i_blocks -= num_dec;
671 }
672 }
179e29e4
CM		 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);
39279cc3 680 }
179e29e4 681delete:
39279cc3
CM		 682 if (del_item) {
683 ret = btrfs_del_item(trans, root, path);
54aa1f4d
CM		 684 if (ret)
685 goto error;
39279cc3
CM		 686 } else {
687 break;
688 }
689 btrfs_release_path(root, path);
690 if (found_extent) {
691 ret = btrfs_free_extent(trans, root, extent_start,
db94535d 692 extent_num_bytes, 0);
39279cc3
CM		 693 BUG_ON(ret);
694 }
695 }
696 ret = 0;
697error:
698 btrfs_release_path(root, path);
699 btrfs_free_path(path);
700 inode->i_sb->s_dirt = 1;
701 return ret;
702}
703
b888db2b 704static int btrfs_cow_one_page(struct inode *inode, struct page *page,
a52d9a80
CM		 705 size_t zero_start)
706{
707 char *kaddr;
708 int ret = 0;
b888db2b 709 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
35ebb934 710 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
b888db2b 711 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
a52d9a80 712
b3cfa35a 713 set_page_extent_mapped(page);
a52d9a80 714
b888db2b
CM		 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);
a52d9a80 718 if (zero_start != PAGE_CACHE_SIZE) {
b888db2b 719 kaddr = kmap(page);
a52d9a80
CM		 720 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
721 flush_dcache_page(page);
b888db2b 722 kunmap(page);
a52d9a80 723 }
b888db2b
CM		 724 set_page_dirty(page);
725 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
a52d9a80 726
a52d9a80
CM		 727 return ret;
728}
729
39279cc3
CM		 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 */
734static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
735{
736 struct inode *inode = mapping->host;
db94535d
CM		 737 struct btrfs_root *root = BTRFS_I(inode)->root;
738 u32 blocksize = root->sectorsize;
39279cc3
CM		 739 pgoff_t index = from >> PAGE_CACHE_SHIFT;
740 unsigned offset = from & (PAGE_CACHE_SIZE-1);
741 struct page *page;
39279cc3 742 int ret = 0;
a52d9a80 743 u64 page_start;
39279cc3
CM		 744
745 if ((offset & (blocksize - 1)) == 0)
746 goto out;
747
db94535d 748 down_read(&root->snap_sem);
39279cc3
CM		 749 ret = -ENOMEM;
750 page = grab_cache_page(mapping, index);
751 if (!page)
752 goto out;
39279cc3 753 if (!PageUptodate(page)) {
9ebefb18 754 ret = btrfs_readpage(NULL, page);
39279cc3
CM		 755 lock_page(page);
756 if (!PageUptodate(page)) {
757 ret = -EIO;
758 goto out;
759 }
760 }
35ebb934 761 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
a52d9a80 762
b888db2b 763 ret = btrfs_cow_one_page(inode, page, offset);
39279cc3 764
39279cc3
CM		 765 unlock_page(page);
766 page_cache_release(page);
011410bd 767 up_read(&BTRFS_I(inode)->root->snap_sem);
39279cc3
CM		 768out:
769 return ret;
770}
771
772static 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;
2bf5a725
CM		 785 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
786
5f39d397 787 u64 mask = root->sectorsize - 1;
39279cc3 788 u64 pos = (inode->i_size + mask) & ~mask;
2bf5a725 789 u64 block_end = attr->ia_size | mask;
39279cc3 790 u64 hole_size;
179e29e4 791 u64 alloc_hint = 0;
39279cc3
CM		 792
793 if (attr->ia_size <= pos)
794 goto out;
795
796 btrfs_truncate_page(inode->i_mapping, inode->i_size);
797
2bf5a725 798 lock_extent(em_tree, pos, block_end, GFP_NOFS);
39279cc3 799 hole_size = (attr->ia_size - pos + mask) & ~mask;
39279cc3
CM		 800
801 mutex_lock(&root->fs_info->fs_mutex);
802 trans = btrfs_start_transaction(root, 1);
803 btrfs_set_trans_block_group(trans, inode);
2bf5a725 804 err = btrfs_drop_extents(trans, root, inode,
3326d1b0
CM		 805 pos, pos + hole_size, pos,
806 &alloc_hint);
2bf5a725 807
179e29e4
CM		 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 }
39279cc3
CM		 813 btrfs_end_transaction(trans, root);
814 mutex_unlock(&root->fs_info->fs_mutex);
2bf5a725 815 unlock_extent(em_tree, pos, block_end, GFP_NOFS);
54aa1f4d
CM		 816 if (err)
817 return err;
39279cc3
CM		 818 }
819out:
820 err = inode_setattr(inode, attr);
821
822 return err;
823}
824void btrfs_delete_inode(struct inode *inode)
825{
826 struct btrfs_trans_handle *trans;
827 struct btrfs_root *root = BTRFS_I(inode)->root;
d3c2fdcf 828 unsigned long nr;
39279cc3
CM		 829 int ret;
830
831 truncate_inode_pages(&inode->i_data, 0);
832 if (is_bad_inode(inode)) {
833 goto no_delete;
834 }
5f39d397 835
39279cc3
CM		 836 inode->i_size = 0;
837 mutex_lock(&root->fs_info->fs_mutex);
838 trans = btrfs_start_transaction(root, 1);
5f39d397 839
39279cc3
CM		 840 btrfs_set_trans_block_group(trans, inode);
841 ret = btrfs_truncate_in_trans(trans, root, inode);
54aa1f4d
CM		 842 if (ret)
843 goto no_delete_lock;
844 ret = btrfs_free_inode(trans, root, inode);
845 if (ret)
846 goto no_delete_lock;
d3c2fdcf 847 nr = trans->blocks_used;
5f39d397 848
39279cc3
CM		 849 btrfs_end_transaction(trans, root);
850 mutex_unlock(&root->fs_info->fs_mutex);
d3c2fdcf 851 btrfs_btree_balance_dirty(root, nr);
39279cc3 852 return;
54aa1f4d
CM		 853
854no_delete_lock:
d3c2fdcf 855 nr = trans->blocks_used;
54aa1f4d
CM		 856 btrfs_end_transaction(trans, root);
857 mutex_unlock(&root->fs_info->fs_mutex);
d3c2fdcf 858 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM		 859no_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 */
867static 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;
0d9f7f3e 875 int ret = 0;
39279cc3
CM		 876
877 path = btrfs_alloc_path();
878 BUG_ON(!path);
39279cc3
CM		 879 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
880 namelen, 0);
0d9f7f3e
Y		 881 if (IS_ERR(di))
882 ret = PTR_ERR(di);
39279cc3
CM		 883 if (!di || IS_ERR(di)) {
884 location->objectid = 0;
39279cc3
CM		 885 goto out;
886 }
5f39d397 887 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
39279cc3
CM		 888out:
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 */
899static int fixup_tree_root_location(struct btrfs_root *root,
900 struct btrfs_key *location,
58176a96
JB		 901 struct btrfs_root **sub_root,
902 struct dentry *dentry)
39279cc3
CM		 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
58176a96
JB		 916 *sub_root = btrfs_read_fs_root(root->fs_info, location,
917 dentry->d_name.name,
918 dentry->d_name.len);
39279cc3
CM		 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);
39279cc3
CM		 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
932static 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;
b888db2b
CM		 937 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
938 inode->i_mapping, GFP_NOFS);
39279cc3
CM		 939 return 0;
940}
941
942static 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
949struct 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
963static 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);
5f39d397 975
39279cc3
CM		 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);
5f39d397 979
39279cc3
CM		 980 if (ret < 0)
981 return ERR_PTR(ret);
5f39d397 982
39279cc3
CM		 983 inode = NULL;
984 if (location.objectid) {
58176a96
JB		 985 ret = fixup_tree_root_location(root, &location, &sub_root,
986 dentry);
39279cc3
CM		 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
39279cc3
CM		 1011static unsigned char btrfs_filetype_table[] = {
1012 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1013};
1014
1015static 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;
5f39d397 1022 struct btrfs_key found_key;
39279cc3
CM		 1023 struct btrfs_path *path;
1024 int ret;
1025 u32 nritems;
5f39d397 1026 struct extent_buffer *leaf;
39279cc3
CM		 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;
5f39d397
CM		 1035 char tmp_name[32];
1036 char *name_ptr;
1037 int name_len;
39279cc3
CM		 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;
5f39d397 1042
39279cc3
CM		 1043 mutex_lock(&root->fs_info->fs_mutex);
1044 key.objectid = inode->i_ino;
39279cc3
CM		 1045 btrfs_set_key_type(&key, key_type);
1046 key.offset = filp->f_pos;
5f39d397 1047
39279cc3 1048 path = btrfs_alloc_path();
2cc58cf2 1049 path->reada = 2;
39279cc3
CM		 1050 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1051 if (ret < 0)
1052 goto err;
1053 advance = 0;
39279cc3 1054 while(1) {
5f39d397
CM		 1055 leaf = path->nodes[0];
1056 nritems = btrfs_header_nritems(leaf);
39279cc3
CM		 1057 slot = path->slots[0];
1058 if (advance || slot >= nritems) {
1059 if (slot >= nritems -1) {
39279cc3
CM		 1060 ret = btrfs_next_leaf(root, path);
1061 if (ret)
1062 break;
5f39d397
CM		 1063 leaf = path->nodes[0];
1064 nritems = btrfs_header_nritems(leaf);
39279cc3
CM		 1065 slot = path->slots[0];
1066 } else {
1067 slot++;
1068 path->slots[0]++;
1069 }
1070 }
1071 advance = 1;
5f39d397
CM		 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)
39279cc3 1076 break;
5f39d397 1077 if (btrfs_key_type(&found_key) != key_type)
39279cc3 1078 break;
5f39d397 1079 if (found_key.offset < filp->f_pos)
39279cc3 1080 continue;
5f39d397
CM		 1081
1082 filp->f_pos = found_key.offset;
39279cc3
CM		 1083 advance = 1;
1084 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1085 di_cur = 0;
5f39d397 1086 di_total = btrfs_item_size(leaf, item);
39279cc3 1087 while(di_cur < di_total) {
5f39d397
CM		 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,
39279cc3 1106 d_type);
5f39d397
CM		 1107
1108 if (name_ptr != tmp_name)
1109 kfree(name_ptr);
1110
39279cc3
CM		 1111 if (over)
1112 goto nopos;
5f39d397 1113 di_len = btrfs_dir_name_len(leaf, di) + sizeof(*di);
39279cc3
CM		 1114 di_cur += di_len;
1115 di = (struct btrfs_dir_item *)((char *)di + di_len);
1116 }
1117 }
1118 filp->f_pos++;
1119nopos:
1120 ret = 0;
1121err:
1122 btrfs_release_path(root, path);
1123 btrfs_free_path(path);
1124 mutex_unlock(&root->fs_info->fs_mutex);
1125 return ret;
1126}
1127
1128int 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/*
54aa1f4d 1145 * This is somewhat expensive, updating the tree every time the
39279cc3
CM		 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 */
1150void 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);
39279cc3
CM		 1161}
1162
1163static 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;
5f39d397 1170 struct btrfs_inode_item *inode_item;
39279cc3 1171 struct btrfs_key *location;
5f39d397 1172 struct btrfs_path *path;
39279cc3
CM		 1173 int ret;
1174 int owner;
1175
5f39d397
CM		 1176 path = btrfs_alloc_path();
1177 BUG_ON(!path);
1178
39279cc3
CM		 1179 inode = new_inode(root->fs_info->sb);
1180 if (!inode)
1181 return ERR_PTR(-ENOMEM);
1182
b888db2b
CM		 1183 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1184 inode->i_mapping, GFP_NOFS);
39279cc3 1185 BTRFS_I(inode)->root = root;
b888db2b 1186
39279cc3
CM		 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
5f39d397
CM		 1194 ret = btrfs_insert_empty_inode(trans, root, path, objectid);
1195 if (ret)
1196 goto fail;
1197
39279cc3
CM		 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;
5f39d397
CM		 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
39279cc3
CM		 1210 location = &BTRFS_I(inode)->location;
1211 location->objectid = objectid;
39279cc3
CM		 1212 location->offset = 0;
1213 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1214
39279cc3
CM		 1215 insert_inode_hash(inode);
1216 return inode;
5f39d397
CM		 1217fail:
1218 btrfs_free_path(path);
1219 return ERR_PTR(ret);
39279cc3
CM		 1220}
1221
1222static inline u8 btrfs_inode_type(struct inode *inode)
1223{
1224 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1225}
1226
1227static 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;
79c44584 1233 struct inode *parent_inode;
5f39d397 1234
39279cc3 1235 key.objectid = inode->i_ino;
39279cc3
CM		 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) {
79c44584
CM		 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;
39279cc3
CM		 1247 ret = btrfs_update_inode(trans, root,
1248 dentry->d_parent->d_inode);
1249 }
1250 return ret;
1251}
1252
1253static 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
618e21d5
JB		 1266static 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;
d3c2fdcf 1275 unsigned long nr;
618e21d5
JB		 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);
1b4ab1bb 1303 btrfs_update_inode(trans, root, inode);
618e21d5
JB		 1304 }
1305 dir->i_sb->s_dirt = 1;
1306 btrfs_update_inode_block_group(trans, inode);
1307 btrfs_update_inode_block_group(trans, dir);
1308out_unlock:
d3c2fdcf 1309 nr = trans->blocks_used;
618e21d5
JB		 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 }
d3c2fdcf 1317 btrfs_btree_balance_dirty(root, nr);
618e21d5
JB		 1318 return err;
1319}
1320
39279cc3
CM		 1321static 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;
d3c2fdcf 1329 unsigned long nr;
39279cc3
CM		 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;
a52d9a80
CM		 1356 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1357 inode->i_mapping, GFP_NOFS);
07157aac 1358 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
39279cc3
CM		 1359 }
1360 dir->i_sb->s_dirt = 1;
1361 btrfs_update_inode_block_group(trans, inode);
1362 btrfs_update_inode_block_group(trans, dir);
1363out_unlock:
d3c2fdcf 1364 nr = trans->blocks_used;
39279cc3
CM		 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 }
d3c2fdcf 1372 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM		 1373 return err;
1374}
1375
1376static 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;
d3c2fdcf 1382 unsigned long nr;
39279cc3
CM		 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);
5f39d397 1392
39279cc3
CM		 1393 btrfs_set_trans_block_group(trans, dir);
1394 atomic_inc(&inode->i_count);
1395 err = btrfs_add_nondir(trans, dentry, inode);
5f39d397 1396
39279cc3
CM		 1397 if (err)
1398 drop_inode = 1;
5f39d397 1399
39279cc3
CM		 1400 dir->i_sb->s_dirt = 1;
1401 btrfs_update_inode_block_group(trans, dir);
54aa1f4d 1402 err = btrfs_update_inode(trans, root, inode);
5f39d397 1403
54aa1f4d
CM		 1404 if (err)
1405 drop_inode = 1;
39279cc3 1406
d3c2fdcf 1407 nr = trans->blocks_used;
39279cc3
CM		 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 }
d3c2fdcf 1415 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM		 1416 return err;
1417}
1418
1419static 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;
39279cc3
CM		 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;
5f39d397 1438
39279cc3
CM		 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;
1444error:
1445 return ret;
1446}
1447
1448static 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;
d3c2fdcf 1456 unsigned long nr = 1;
39279cc3
CM		 1457
1458 mutex_lock(&root->fs_info->fs_mutex);
1459 trans = btrfs_start_transaction(root, 1);
1460 btrfs_set_trans_block_group(trans, dir);
5f39d397 1461
39279cc3
CM		 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 }
5f39d397 1479
39279cc3
CM		 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;
5f39d397 1493
39279cc3
CM		 1494 err = btrfs_add_link(trans, dentry, inode);
1495 if (err)
1496 goto out_fail;
5f39d397 1497
39279cc3
CM		 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
1504out_fail:
d3c2fdcf 1505 nr = trans->blocks_used;
39279cc3 1506 btrfs_end_transaction(trans, root);
5f39d397 1507
39279cc3
CM		 1508out_unlock:
1509 mutex_unlock(&root->fs_info->fs_mutex);
1510 if (drop_on_err)
1511 iput(inode);
d3c2fdcf 1512 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM		 1513 return err;
1514}
1515
a52d9a80
CM		 1516struct 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;
db94535d 1522 u64 bytenr;
a52d9a80
CM		 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;
5f39d397
CM		 1531 struct extent_buffer *leaf;
1532 struct btrfs_key found_key;
a52d9a80
CM		 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
1541again:
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 }
5f39d397
CM		 1552 em->start = EXTENT_MAP_HOLE;
1553 em->end = EXTENT_MAP_HOLE;
a52d9a80
CM		 1554 }
1555 em->bdev = inode->i_sb->s_bdev;
179e29e4
CM		 1556 ret = btrfs_lookup_file_extent(trans, root, path,
1557 objectid, start, trans != NULL);
a52d9a80
CM		 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
5f39d397
CM		 1569 leaf = path->nodes[0];
1570 item = btrfs_item_ptr(leaf, path->slots[0],
a52d9a80 1571 struct btrfs_file_extent_item);
a52d9a80 1572 /* are we inside the extent that was found? */
5f39d397
CM		 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 ||
a52d9a80
CM		 1576 found_type != BTRFS_EXTENT_DATA_KEY) {
1577 goto not_found;
1578 }
1579
5f39d397
CM		 1580 found_type = btrfs_file_extent_type(leaf, item);
1581 extent_start = found_key.offset;
a52d9a80
CM		 1582 if (found_type == BTRFS_FILE_EXTENT_REG) {
1583 extent_end = extent_start +
db94535d 1584 btrfs_file_extent_num_bytes(leaf, item);
a52d9a80 1585 err = 0;
b888db2b 1586 if (start < extent_start || start >= extent_end) {
a52d9a80
CM		 1587 em->start = start;
1588 if (start < extent_start) {
b888db2b
CM		 1589 if (end < extent_start)
1590 goto not_found;
a52d9a80
CM		 1591 em->end = extent_end - 1;
1592 } else {
1593 em->end = end;
1594 }
1595 goto not_found_em;
1596 }
db94535d
CM		 1597 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
1598 if (bytenr == 0) {
a52d9a80
CM		 1599 em->start = extent_start;
1600 em->end = extent_end - 1;
5f39d397
CM		 1601 em->block_start = EXTENT_MAP_HOLE;
1602 em->block_end = EXTENT_MAP_HOLE;
a52d9a80
CM		 1603 goto insert;
1604 }
db94535d
CM		 1605 bytenr += btrfs_file_extent_offset(leaf, item);
1606 em->block_start = bytenr;
a52d9a80 1607 em->block_end = em->block_start +
db94535d 1608 btrfs_file_extent_num_bytes(leaf, item) - 1;
a52d9a80
CM		 1609 em->start = extent_start;
1610 em->end = extent_end - 1;
1611 goto insert;
1612 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
5f39d397 1613 unsigned long ptr;
a52d9a80 1614 char *map;
3326d1b0
CM		 1615 size_t size;
1616 size_t extent_offset;
1617 size_t copy_size;
a52d9a80 1618
5f39d397
CM		 1619 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
1620 path->slots[0]));
689f9346 1621 extent_end = (extent_start + size - 1) |
db94535d 1622 ((u64)root->sectorsize - 1);
b888db2b 1623 if (start < extent_start || start >= extent_end) {
a52d9a80
CM		 1624 em->start = start;
1625 if (start < extent_start) {
b888db2b
CM		 1626 if (end < extent_start)
1627 goto not_found;
50b78c24 1628 em->end = extent_end;
a52d9a80
CM		 1629 } else {
1630 em->end = end;
1631 }
1632 goto not_found_em;
1633 }
689f9346
Y		 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 }
5f39d397 1642
35ebb934 1643 extent_offset = ((u64)page->index << PAGE_CACHE_SHIFT) -
689f9346 1644 extent_start + page_offset;
ae2f5411 1645 copy_size = min_t(u64, PAGE_CACHE_SIZE - page_offset,
3326d1b0 1646 size - extent_offset);
3326d1b0
CM		 1647 em->start = extent_start + extent_offset;
1648 em->end = (em->start + copy_size -1) |
1649 ((u64)root->sectorsize -1);
689f9346
Y		 1650 map = kmap(page);
1651 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
179e29e4
CM		 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);
a52d9a80 1668 }
a52d9a80 1669 kunmap(page);
3326d1b0 1670 set_extent_uptodate(em_tree, em->start, em->end, GFP_NOFS);
a52d9a80
CM		 1671 goto insert;
1672 } else {
1673 printk("unkknown found_type %d\n", found_type);
1674 WARN_ON(1);
1675 }
1676not_found:
1677 em->start = start;
1678 em->end = end;
1679not_found_em:
5f39d397
CM		 1680 em->block_start = EXTENT_MAP_HOLE;
1681 em->block_end = EXTENT_MAP_HOLE;
a52d9a80
CM		 1682insert:
1683 btrfs_release_path(root, path);
1684 if (em->start > start || em->end < start) {
b888db2b 1685 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->end, start, end);
a52d9a80
CM		 1686 err = -EIO;
1687 goto out;
1688 }
1689 ret = add_extent_mapping(em_tree, em);
1690 if (ret == -EEXIST) {
1691 free_extent_map(em);
2bf5a725 1692 em = NULL;
a52d9a80
CM		 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 }
a52d9a80
CM		 1699 goto again;
1700 }
1701 err = 0;
1702out:
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
d396c6f5 1718static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
39279cc3 1719{
d396c6f5 1720 return extent_bmap(mapping, iblock, btrfs_get_extent);
39279cc3
CM		 1721}
1722
1723static int btrfs_prepare_write(struct file *file, struct page *page,
1724 unsigned from, unsigned to)
1725{
a52d9a80
CM		 1726 return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree,
1727 page->mapping->host, page, from, to,
1728 btrfs_get_extent);
39279cc3
CM		 1729}
1730
a52d9a80 1731int btrfs_readpage(struct file *file, struct page *page)
9ebefb18 1732{
a52d9a80
CM		 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);
9ebefb18 1736}
a52d9a80 1737static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
39279cc3 1738{
a52d9a80 1739 struct extent_map_tree *tree;
b888db2b
CM		 1740
1741
1742 if (current->flags & PF_MEMALLOC) {
1743 redirty_page_for_writepage(wbc, page);
1744 unlock_page(page);
1745 return 0;
1746 }
a52d9a80
CM		 1747 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1748 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
9ebefb18
CM		 1749}
1750
b293f02e
CM		 1751
1752static int btrfs_writepages(struct address_space *mapping,
1753 struct writeback_control *wbc)
1754{
1755 struct extent_map_tree *tree;
1756 tree = &BTRFS_I(mapping->host)->extent_tree;
1757 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
1758}
1759
a52d9a80 1760static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
9ebefb18 1761{
a52d9a80
CM		 1762 struct extent_map_tree *tree;
1763 int ret;
8c2383c3 1764
a52d9a80
CM		 1765 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1766 ret = try_release_extent_mapping(tree, page);
1767 if (ret == 1) {
1768 ClearPagePrivate(page);
1769 set_page_private(page, 0);
1770 page_cache_release(page);
39279cc3 1771 }
a52d9a80 1772 return ret;
39279cc3
CM		 1773}
1774
a52d9a80 1775static void btrfs_invalidatepage(struct page *page, unsigned long offset)
39279cc3 1776{
a52d9a80 1777 struct extent_map_tree *tree;
39279cc3 1778
a52d9a80
CM		 1779 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1780 extent_invalidatepage(tree, page, offset);
1781 btrfs_releasepage(page, GFP_NOFS);
39279cc3
CM		 1782}
1783
9ebefb18
CM		 1784/*
1785 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
1786 * called from a page fault handler when a page is first dirtied. Hence we must
1787 * be careful to check for EOF conditions here. We set the page up correctly
1788 * for a written page which means we get ENOSPC checking when writing into
1789 * holes and correct delalloc and unwritten extent mapping on filesystems that
1790 * support these features.
1791 *
1792 * We are not allowed to take the i_mutex here so we have to play games to
1793 * protect against truncate races as the page could now be beyond EOF. Because
1794 * vmtruncate() writes the inode size before removing pages, once we have the
1795 * page lock we can determine safely if the page is beyond EOF. If it is not
1796 * beyond EOF, then the page is guaranteed safe against truncation until we
1797 * unlock the page.
1798 */
1799int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1800{
1801 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
1802 unsigned long end;
1803 loff_t size;
1804 int ret = -EINVAL;
a52d9a80 1805 u64 page_start;
9ebefb18 1806
011410bd 1807 down_read(&BTRFS_I(inode)->root->snap_sem);
9ebefb18
CM		 1808 lock_page(page);
1809 wait_on_page_writeback(page);
9ebefb18 1810 size = i_size_read(inode);
35ebb934 1811 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
a52d9a80 1812
9ebefb18 1813 if ((page->mapping != inode->i_mapping) ||
a52d9a80 1814 (page_start > size)) {
9ebefb18
CM		 1815 /* page got truncated out from underneath us */
1816 goto out_unlock;
1817 }
1818
1819 /* page is wholly or partially inside EOF */
a52d9a80 1820 if (page_start + PAGE_CACHE_SIZE > size)
9ebefb18
CM		 1821 end = size & ~PAGE_CACHE_MASK;
1822 else
1823 end = PAGE_CACHE_SIZE;
1824
b888db2b 1825 ret = btrfs_cow_one_page(inode, page, end);
9ebefb18
CM		 1826
1827out_unlock:
011410bd 1828 up_read(&BTRFS_I(inode)->root->snap_sem);
9ebefb18
CM		 1829 unlock_page(page);
1830 return ret;
1831}
1832
39279cc3
CM		 1833static void btrfs_truncate(struct inode *inode)
1834{
1835 struct btrfs_root *root = BTRFS_I(inode)->root;
1836 int ret;
1837 struct btrfs_trans_handle *trans;
d3c2fdcf 1838 unsigned long nr;
39279cc3
CM		 1839
1840 if (!S_ISREG(inode->i_mode))
1841 return;
1842 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1843 return;
1844
1845 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1846
1847 mutex_lock(&root->fs_info->fs_mutex);
1848 trans = btrfs_start_transaction(root, 1);
1849 btrfs_set_trans_block_group(trans, inode);
1850
1851 /* FIXME, add redo link to tree so we don't leak on crash */
1852 ret = btrfs_truncate_in_trans(trans, root, inode);
39279cc3 1853 btrfs_update_inode(trans, root, inode);
d3c2fdcf 1854 nr = trans->blocks_used;
5f39d397 1855
39279cc3
CM		 1856 ret = btrfs_end_transaction(trans, root);
1857 BUG_ON(ret);
1858 mutex_unlock(&root->fs_info->fs_mutex);
d3c2fdcf 1859 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM		 1860}
1861
1862int btrfs_commit_write(struct file *file, struct page *page,
1863 unsigned from, unsigned to)
1864{
a52d9a80
CM		 1865 return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree,
1866 page->mapping->host, page, from, to);
39279cc3
CM		 1867}
1868
1869static int create_subvol(struct btrfs_root *root, char *name, int namelen)
1870{
1871 struct btrfs_trans_handle *trans;
1872 struct btrfs_key key;
1873 struct btrfs_root_item root_item;
1874 struct btrfs_inode_item *inode_item;
5f39d397 1875 struct extent_buffer *leaf;
39279cc3
CM		 1876 struct btrfs_root *new_root;
1877 struct inode *inode;
1878 struct inode *dir;
1879 int ret;
54aa1f4d 1880 int err;
39279cc3
CM		 1881 u64 objectid;
1882 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
d3c2fdcf 1883 unsigned long nr = 1;
39279cc3
CM		 1884
1885 mutex_lock(&root->fs_info->fs_mutex);
1886 trans = btrfs_start_transaction(root, 1);
1887 BUG_ON(!trans);
1888
db94535d 1889 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, 0);
5f39d397
CM		 1890 if (IS_ERR(leaf))
1891 return PTR_ERR(leaf);
1892
1893 btrfs_set_header_nritems(leaf, 0);
1894 btrfs_set_header_level(leaf, 0);
db94535d 1895 btrfs_set_header_bytenr(leaf, leaf->start);
5f39d397
CM		 1896 btrfs_set_header_generation(leaf, trans->transid);
1897 btrfs_set_header_owner(leaf, root->root_key.objectid);
1898 write_extent_buffer(leaf, root->fs_info->fsid,
1899 (unsigned long)btrfs_header_fsid(leaf),
1900 BTRFS_FSID_SIZE);
1901 btrfs_mark_buffer_dirty(leaf);
39279cc3
CM		 1902
1903 inode_item = &root_item.inode;
1904 memset(inode_item, 0, sizeof(*inode_item));
5f39d397
CM		 1905 inode_item->generation = cpu_to_le64(1);
1906 inode_item->size = cpu_to_le64(3);
1907 inode_item->nlink = cpu_to_le32(1);
1908 inode_item->nblocks = cpu_to_le64(1);
1909 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
39279cc3 1910
db94535d
CM		 1911 btrfs_set_root_bytenr(&root_item, leaf->start);
1912 btrfs_set_root_level(&root_item, 0);
39279cc3 1913 btrfs_set_root_refs(&root_item, 1);
5f39d397
CM		 1914 btrfs_set_root_used(&root_item, 0);
1915
5eda7b5e
CM		 1916 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
1917 root_item.drop_level = 0;
5f39d397
CM		 1918
1919 free_extent_buffer(leaf);
1920 leaf = NULL;
39279cc3
CM		 1921
1922 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1923 0, &objectid);
54aa1f4d
CM		 1924 if (ret)
1925 goto fail;
39279cc3
CM		 1926
1927 btrfs_set_root_dirid(&root_item, new_dirid);
1928
1929 key.objectid = objectid;
1930 key.offset = 1;
39279cc3
CM		 1931 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1932 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1933 &root_item);
54aa1f4d
CM		 1934 if (ret)
1935 goto fail;
39279cc3
CM		 1936
1937 /*
1938 * insert the directory item
1939 */
1940 key.offset = (u64)-1;
1941 dir = root->fs_info->sb->s_root->d_inode;
1942 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1943 name, namelen, dir->i_ino, &key,
1944 BTRFS_FT_DIR);
54aa1f4d
CM		 1945 if (ret)
1946 goto fail;
39279cc3
CM		 1947
1948 ret = btrfs_commit_transaction(trans, root);
54aa1f4d
CM		 1949 if (ret)
1950 goto fail_commit;
39279cc3 1951
58176a96 1952 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
39279cc3
CM		 1953 BUG_ON(!new_root);
1954
1955 trans = btrfs_start_transaction(new_root, 1);
1956 BUG_ON(!trans);
1957
1958 inode = btrfs_new_inode(trans, new_root, new_dirid,
1959 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
54aa1f4d
CM		 1960 if (IS_ERR(inode))
1961 goto fail;
39279cc3
CM		 1962 inode->i_op = &btrfs_dir_inode_operations;
1963 inode->i_fop = &btrfs_dir_file_operations;
34088780 1964 new_root->inode = inode;
39279cc3
CM		 1965
1966 ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid);
54aa1f4d
CM		 1967 if (ret)
1968 goto fail;
39279cc3
CM		 1969
1970 inode->i_nlink = 1;
1971 inode->i_size = 6;
1972 ret = btrfs_update_inode(trans, new_root, inode);
54aa1f4d
CM		 1973 if (ret)
1974 goto fail;
1975fail:
d3c2fdcf 1976 nr = trans->blocks_used;
54aa1f4d
CM		 1977 err = btrfs_commit_transaction(trans, root);
1978 if (err && !ret)
1979 ret = err;
1980fail_commit:
39279cc3 1981 mutex_unlock(&root->fs_info->fs_mutex);
d3c2fdcf 1982 btrfs_btree_balance_dirty(root, nr);
54aa1f4d 1983 return ret;
39279cc3
CM		 1984}
1985
1986static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
1987{
1988 struct btrfs_trans_handle *trans;
1989 struct btrfs_key key;
1990 struct btrfs_root_item new_root_item;
5f39d397 1991 struct extent_buffer *tmp;
39279cc3 1992 int ret;
54aa1f4d 1993 int err;
39279cc3 1994 u64 objectid;
d3c2fdcf 1995 unsigned long nr;
39279cc3
CM		 1996
1997 if (!root->ref_cows)
1998 return -EINVAL;
1999
011410bd
CM		 2000 down_write(&root->snap_sem);
2001 freeze_bdev(root->fs_info->sb->s_bdev);
2002 thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb);
2003
39279cc3
CM		 2004 mutex_lock(&root->fs_info->fs_mutex);
2005 trans = btrfs_start_transaction(root, 1);
2006 BUG_ON(!trans);
2007
2008 ret = btrfs_update_inode(trans, root, root->inode);
54aa1f4d
CM		 2009 if (ret)
2010 goto fail;
39279cc3
CM		 2011
2012 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
2013 0, &objectid);
54aa1f4d
CM		 2014 if (ret)
2015 goto fail;
39279cc3
CM		 2016
2017 memcpy(&new_root_item, &root->root_item,
2018 sizeof(new_root_item));
2019
2020 key.objectid = objectid;
2021 key.offset = 1;
39279cc3 2022 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
5f39d397 2023
83df7c1d 2024 btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
db94535d
CM		 2025 btrfs_set_root_bytenr(&new_root_item, root->node->start);
2026 btrfs_set_root_level(&new_root_item, btrfs_header_level(root->node));
39279cc3
CM		 2027
2028 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
2029 &new_root_item);
54aa1f4d
CM		 2030 if (ret)
2031 goto fail;
39279cc3
CM		 2032
2033 /*
2034 * insert the directory item
2035 */
2036 key.offset = (u64)-1;
2037 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
2038 name, namelen,
2039 root->fs_info->sb->s_root->d_inode->i_ino,
2040 &key, BTRFS_FT_DIR);
2041
54aa1f4d
CM		 2042 if (ret)
2043 goto fail;
39279cc3
CM		 2044
2045 ret = btrfs_inc_root_ref(trans, root);
54aa1f4d
CM		 2046 if (ret)
2047 goto fail;
54aa1f4d 2048fail:
d3c2fdcf 2049 nr = trans->blocks_used;
54aa1f4d 2050 err = btrfs_commit_transaction(trans, root);
5f39d397 2051
54aa1f4d
CM		 2052 if (err && !ret)
2053 ret = err;
5f39d397 2054
39279cc3 2055 mutex_unlock(&root->fs_info->fs_mutex);
011410bd 2056 up_write(&root->snap_sem);
d3c2fdcf 2057 btrfs_btree_balance_dirty(root, nr);
54aa1f4d 2058 return ret;
39279cc3
CM		 2059}
2060
86479a04
CM		 2061static unsigned long force_ra(struct address_space *mapping,
2062 struct file_ra_state *ra, struct file *file,
2063 pgoff_t offset, pgoff_t last_index)
2064{
2065 pgoff_t req_size;
2066
2067#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2068 req_size = last_index - offset + 1;
2069 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2070 return offset;
2071#else
2072 req_size = min(last_index - offset + 1, (pgoff_t)128);
2073 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2074 return offset + req_size;
2075#endif
2076}
2077
2078int btrfs_defrag_file(struct file *file) {
2079 struct inode *inode = file->f_path.dentry->d_inode;
2080 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2081 struct page *page;
2082 unsigned long last_index;
2083 unsigned long ra_index = 0;
2084 u64 page_start;
2085 u64 page_end;
2086 unsigned long i;
2087
2088 mutex_lock(&inode->i_mutex);
2089 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
2090 for (i = 0; i <= last_index; i++) {
2091 if (i == ra_index) {
2092 ra_index = force_ra(inode->i_mapping, &file->f_ra,
2093 file, ra_index, last_index);
2094 }
2095 page = grab_cache_page(inode->i_mapping, i);
2096 if (!page)
2097 goto out_unlock;
2098 if (!PageUptodate(page)) {
2099 btrfs_readpage(NULL, page);
2100 lock_page(page);
2101 if (!PageUptodate(page)) {
2102 unlock_page(page);
2103 page_cache_release(page);
2104 goto out_unlock;
2105 }
2106 }
35ebb934 2107 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
86479a04
CM		 2108 page_end = page_start + PAGE_CACHE_SIZE - 1;
2109
2110 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
2111 set_extent_delalloc(em_tree, page_start,
2112 page_end, GFP_NOFS);
2113 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
2114 set_page_dirty(page);
2115 unlock_page(page);
2116 page_cache_release(page);
2117 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
2118 }
2119
2120out_unlock:
2121 mutex_unlock(&inode->i_mutex);
2122 return 0;
2123}
2124
d03581f4 2125static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg)
39279cc3 2126{
39279cc3 2127 struct btrfs_ioctl_vol_args vol_args;
39279cc3 2128 struct btrfs_dir_item *di;
39279cc3 2129 struct btrfs_path *path;
d03581f4 2130 int namelen;
39279cc3
CM		 2131 u64 root_dirid;
2132
d03581f4
CH		 2133 if (copy_from_user(&vol_args, arg, sizeof(vol_args)))
2134 return -EFAULT;
5f39d397 2135
d03581f4
CH		 2136 namelen = strlen(vol_args.name);
2137 if (namelen > BTRFS_VOL_NAME_MAX)
2138 return -EINVAL;
2139 if (strchr(vol_args.name, '/'))
2140 return -EINVAL;
2141
2142 path = btrfs_alloc_path();
2143 if (!path)
2144 return -ENOMEM;
2145
2146 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2147 mutex_lock(&root->fs_info->fs_mutex);
2148 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2149 path, root_dirid,
2150 vol_args.name, namelen, 0);
2151 mutex_unlock(&root->fs_info->fs_mutex);
2152 btrfs_free_path(path);
2153 if (di && !IS_ERR(di))
2154 return -EEXIST;
2155 if (IS_ERR(di))
2156 return PTR_ERR(di);
2157
2158 if (root == root->fs_info->tree_root)
2159 return create_subvol(root, vol_args.name, namelen);
2160 return create_snapshot(root, vol_args.name, namelen);
2161}
2162
2163static int btrfs_ioctl_defrag(struct file *file)
2164{
2165 struct inode *inode = file->f_path.dentry->d_inode;
2166 struct btrfs_root *root = BTRFS_I(inode)->root;
2167
2168 switch (inode->i_mode & S_IFMT) {
2169 case S_IFDIR:
39279cc3 2170 mutex_lock(&root->fs_info->fs_mutex);
d03581f4
CH		 2171 btrfs_defrag_root(root, 0);
2172 btrfs_defrag_root(root->fs_info->extent_root, 0);
39279cc3 2173 mutex_unlock(&root->fs_info->fs_mutex);
39279cc3 2174 break;
d03581f4
CH		 2175 case S_IFREG:
2176 btrfs_defrag_file(file);
2177 break;
2178 }
2179
2180 return 0;
2181}
6702ed49 2182
d03581f4
CH		 2183long btrfs_ioctl(struct file *file, unsigned int
2184 cmd, unsigned long arg)
2185{
2186 struct btrfs_root *root = BTRFS_I(file->f_path.dentry->d_inode)->root;
2187
2188 switch (cmd) {
2189 case BTRFS_IOC_SNAP_CREATE:
2190 return btrfs_ioctl_snap_create(root, (void __user *)arg);
6702ed49 2191 case BTRFS_IOC_DEFRAG:
d03581f4 2192 return btrfs_ioctl_defrag(file);
39279cc3 2193 }
d03581f4
CH		 2194
2195 return -ENOTTY;
39279cc3
CM		 2196}
2197
39279cc3
CM		 2198/*
2199 * Called inside transaction, so use GFP_NOFS
2200 */
2201struct inode *btrfs_alloc_inode(struct super_block *sb)
2202{
2203 struct btrfs_inode *ei;
2204
2205 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2206 if (!ei)
2207 return NULL;
15ee9bc7 2208 ei->last_trans = 0;
39279cc3
CM		 2209 return &ei->vfs_inode;
2210}
2211
2212void btrfs_destroy_inode(struct inode *inode)
2213{
2214 WARN_ON(!list_empty(&inode->i_dentry));
2215 WARN_ON(inode->i_data.nrpages);
2216
2217 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2218}
2219
44ec0b71
CM		 2220#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2221static void init_once(struct kmem_cache * cachep, void *foo)
2222#else
39279cc3
CM		 2223static void init_once(void * foo, struct kmem_cache * cachep,
2224 unsigned long flags)
44ec0b71 2225#endif
39279cc3
CM		 2226{
2227 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2228
2229 inode_init_once(&ei->vfs_inode);
2230}
2231
2232void btrfs_destroy_cachep(void)
2233{
2234 if (btrfs_inode_cachep)
2235 kmem_cache_destroy(btrfs_inode_cachep);
2236 if (btrfs_trans_handle_cachep)
2237 kmem_cache_destroy(btrfs_trans_handle_cachep);
2238 if (btrfs_transaction_cachep)
2239 kmem_cache_destroy(btrfs_transaction_cachep);
2240 if (btrfs_bit_radix_cachep)
2241 kmem_cache_destroy(btrfs_bit_radix_cachep);
2242 if (btrfs_path_cachep)
2243 kmem_cache_destroy(btrfs_path_cachep);
2244}
2245
86479a04 2246struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
92fee66d 2247 unsigned long extra_flags,
44ec0b71
CM		 2248#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2249 void (*ctor)(struct kmem_cache *, void *)
2250#else
92fee66d 2251 void (*ctor)(void *, struct kmem_cache *,
44ec0b71
CM		 2252 unsigned long)
2253#endif
2254 )
92fee66d
CM		 2255{
2256 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2257 SLAB_MEM_SPREAD | extra_flags), ctor
2258#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2259 ,NULL
2260#endif
2261 );
2262}
2263
39279cc3
CM		 2264int btrfs_init_cachep(void)
2265{
86479a04 2266 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
92fee66d
CM		 2267 sizeof(struct btrfs_inode),
2268 0, init_once);
39279cc3
CM		 2269 if (!btrfs_inode_cachep)
2270 goto fail;
86479a04
CM		 2271 btrfs_trans_handle_cachep =
2272 btrfs_cache_create("btrfs_trans_handle_cache",
2273 sizeof(struct btrfs_trans_handle),
2274 0, NULL);
39279cc3
CM		 2275 if (!btrfs_trans_handle_cachep)
2276 goto fail;
86479a04 2277 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
39279cc3 2278 sizeof(struct btrfs_transaction),
92fee66d 2279 0, NULL);
39279cc3
CM		 2280 if (!btrfs_transaction_cachep)
2281 goto fail;
86479a04 2282 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
23223584 2283 sizeof(struct btrfs_path),
92fee66d 2284 0, NULL);
39279cc3
CM		 2285 if (!btrfs_path_cachep)
2286 goto fail;
86479a04 2287 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
92fee66d 2288 SLAB_DESTROY_BY_RCU, NULL);
39279cc3
CM		 2289 if (!btrfs_bit_radix_cachep)
2290 goto fail;
2291 return 0;
2292fail:
2293 btrfs_destroy_cachep();
2294 return -ENOMEM;
2295}
2296
2297static int btrfs_getattr(struct vfsmount *mnt,
2298 struct dentry *dentry, struct kstat *stat)
2299{
2300 struct inode *inode = dentry->d_inode;
2301 generic_fillattr(inode, stat);
2302 stat->blksize = 256 * 1024;
2303 return 0;
2304}
2305
2306static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2307 struct inode * new_dir,struct dentry *new_dentry)
2308{
2309 struct btrfs_trans_handle *trans;
2310 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2311 struct inode *new_inode = new_dentry->d_inode;
2312 struct inode *old_inode = old_dentry->d_inode;
2313 struct timespec ctime = CURRENT_TIME;
2314 struct btrfs_path *path;
2315 struct btrfs_dir_item *di;
2316 int ret;
2317
2318 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2319 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2320 return -ENOTEMPTY;
2321 }
5f39d397 2322
39279cc3
CM		 2323 mutex_lock(&root->fs_info->fs_mutex);
2324 trans = btrfs_start_transaction(root, 1);
5f39d397 2325
39279cc3
CM		 2326 btrfs_set_trans_block_group(trans, new_dir);
2327 path = btrfs_alloc_path();
2328 if (!path) {
2329 ret = -ENOMEM;
2330 goto out_fail;
2331 }
2332
2333 old_dentry->d_inode->i_nlink++;
2334 old_dir->i_ctime = old_dir->i_mtime = ctime;
2335 new_dir->i_ctime = new_dir->i_mtime = ctime;
2336 old_inode->i_ctime = ctime;
5f39d397 2337
39279cc3
CM		 2338 if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) {
2339 struct btrfs_key *location = &BTRFS_I(new_dir)->location;
5f39d397 2340 struct btrfs_key old_parent_key;
39279cc3
CM		 2341 di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino,
2342 "..", 2, -1);
2343 if (IS_ERR(di)) {
2344 ret = PTR_ERR(di);
2345 goto out_fail;
2346 }
2347 if (!di) {
2348 ret = -ENOENT;
2349 goto out_fail;
2350 }
5f39d397 2351 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &old_parent_key);
39279cc3
CM		 2352 ret = btrfs_del_item(trans, root, path);
2353 if (ret) {
39279cc3
CM		 2354 goto out_fail;
2355 }
2356 btrfs_release_path(root, path);
2357
2358 di = btrfs_lookup_dir_index_item(trans, root, path,
2359 old_inode->i_ino,
5f39d397 2360 old_parent_key.objectid,
39279cc3
CM		 2361 "..", 2, -1);
2362 if (IS_ERR(di)) {
2363 ret = PTR_ERR(di);
2364 goto out_fail;
2365 }
2366 if (!di) {
2367 ret = -ENOENT;
2368 goto out_fail;
2369 }
2370 ret = btrfs_del_item(trans, root, path);
2371 if (ret) {
39279cc3
CM		 2372 goto out_fail;
2373 }
2374 btrfs_release_path(root, path);
2375
2376 ret = btrfs_insert_dir_item(trans, root, "..", 2,
2377 old_inode->i_ino, location,
2378 BTRFS_FT_DIR);
2379 if (ret)
2380 goto out_fail;
2381 }
2382
2383
2384 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2385 if (ret)
2386 goto out_fail;
2387
2388 if (new_inode) {
2389 new_inode->i_ctime = CURRENT_TIME;
2390 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2391 if (ret)
2392 goto out_fail;
39279cc3
CM		 2393 }
2394 ret = btrfs_add_link(trans, new_dentry, old_inode);
2395 if (ret)
2396 goto out_fail;
2397
2398out_fail:
2399 btrfs_free_path(path);
2400 btrfs_end_transaction(trans, root);
2401 mutex_unlock(&root->fs_info->fs_mutex);
2402 return ret;
2403}
2404
2405static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2406 const char *symname)
2407{
2408 struct btrfs_trans_handle *trans;
2409 struct btrfs_root *root = BTRFS_I(dir)->root;
2410 struct btrfs_path *path;
2411 struct btrfs_key key;
2412 struct inode *inode;
2413 int err;
2414 int drop_inode = 0;
2415 u64 objectid;
2416 int name_len;
2417 int datasize;
5f39d397 2418 unsigned long ptr;
39279cc3 2419 struct btrfs_file_extent_item *ei;
5f39d397 2420 struct extent_buffer *leaf;
d3c2fdcf 2421 unsigned long nr;
39279cc3
CM		 2422
2423 name_len = strlen(symname) + 1;
2424 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2425 return -ENAMETOOLONG;
2426 mutex_lock(&root->fs_info->fs_mutex);
2427 trans = btrfs_start_transaction(root, 1);
2428 btrfs_set_trans_block_group(trans, dir);
2429
2430 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2431 if (err) {
2432 err = -ENOSPC;
2433 goto out_unlock;
2434 }
2435
2436 inode = btrfs_new_inode(trans, root, objectid,
2437 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2438 err = PTR_ERR(inode);
2439 if (IS_ERR(inode))
2440 goto out_unlock;
2441
2442 btrfs_set_trans_block_group(trans, inode);
2443 err = btrfs_add_nondir(trans, dentry, inode);
2444 if (err)
2445 drop_inode = 1;
2446 else {
2447 inode->i_mapping->a_ops = &btrfs_aops;
2448 inode->i_fop = &btrfs_file_operations;
2449 inode->i_op = &btrfs_file_inode_operations;
a52d9a80
CM		 2450 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
2451 inode->i_mapping, GFP_NOFS);
07157aac 2452 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
39279cc3
CM		 2453 }
2454 dir->i_sb->s_dirt = 1;
2455 btrfs_update_inode_block_group(trans, inode);
2456 btrfs_update_inode_block_group(trans, dir);
2457 if (drop_inode)
2458 goto out_unlock;
2459
2460 path = btrfs_alloc_path();
2461 BUG_ON(!path);
2462 key.objectid = inode->i_ino;
2463 key.offset = 0;
39279cc3
CM		 2464 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2465 datasize = btrfs_file_extent_calc_inline_size(name_len);
2466 err = btrfs_insert_empty_item(trans, root, path, &key,
2467 datasize);
54aa1f4d
CM		 2468 if (err) {
2469 drop_inode = 1;
2470 goto out_unlock;
2471 }
5f39d397
CM		 2472 leaf = path->nodes[0];
2473 ei = btrfs_item_ptr(leaf, path->slots[0],
2474 struct btrfs_file_extent_item);
2475 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
2476 btrfs_set_file_extent_type(leaf, ei,
39279cc3
CM		 2477 BTRFS_FILE_EXTENT_INLINE);
2478 ptr = btrfs_file_extent_inline_start(ei);
5f39d397
CM		 2479 write_extent_buffer(leaf, symname, ptr, name_len);
2480 btrfs_mark_buffer_dirty(leaf);
39279cc3 2481 btrfs_free_path(path);
5f39d397 2482
39279cc3
CM		 2483 inode->i_op = &btrfs_symlink_inode_operations;
2484 inode->i_mapping->a_ops = &btrfs_symlink_aops;
2485 inode->i_size = name_len - 1;
54aa1f4d
CM		 2486 err = btrfs_update_inode(trans, root, inode);
2487 if (err)
2488 drop_inode = 1;
39279cc3
CM		 2489
2490out_unlock:
d3c2fdcf 2491 nr = trans->blocks_used;
39279cc3
CM		 2492 btrfs_end_transaction(trans, root);
2493 mutex_unlock(&root->fs_info->fs_mutex);
39279cc3
CM		 2494 if (drop_inode) {
2495 inode_dec_link_count(inode);
2496 iput(inode);
2497 }
d3c2fdcf 2498 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM		 2499 return err;
2500}
2501
2502static struct inode_operations btrfs_dir_inode_operations = {
2503 .lookup = btrfs_lookup,
2504 .create = btrfs_create,
2505 .unlink = btrfs_unlink,
2506 .link = btrfs_link,
2507 .mkdir = btrfs_mkdir,
2508 .rmdir = btrfs_rmdir,
2509 .rename = btrfs_rename,
2510 .symlink = btrfs_symlink,
2511 .setattr = btrfs_setattr,
618e21d5 2512 .mknod = btrfs_mknod,
39279cc3
CM		 2513};
2514
2515static struct inode_operations btrfs_dir_ro_inode_operations = {
2516 .lookup = btrfs_lookup,
2517};
2518
2519static struct file_operations btrfs_dir_file_operations = {
2520 .llseek = generic_file_llseek,
2521 .read = generic_read_dir,
2522 .readdir = btrfs_readdir,
34287aa3 2523 .unlocked_ioctl = btrfs_ioctl,
39279cc3 2524#ifdef CONFIG_COMPAT
34287aa3 2525 .compat_ioctl = btrfs_ioctl,
39279cc3
CM		 2526#endif
2527};
2528
07157aac
CM		 2529static struct extent_map_ops btrfs_extent_map_ops = {
2530 .fill_delalloc = run_delalloc_range,
2531 .writepage_io_hook = btrfs_writepage_io_hook,
2532 .readpage_io_hook = btrfs_readpage_io_hook,
2533 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
2534};
2535
39279cc3
CM		 2536static struct address_space_operations btrfs_aops = {
2537 .readpage = btrfs_readpage,
2538 .writepage = btrfs_writepage,
b293f02e 2539 .writepages = btrfs_writepages,
39279cc3
CM		 2540 .sync_page = block_sync_page,
2541 .prepare_write = btrfs_prepare_write,
2542 .commit_write = btrfs_commit_write,
2543 .bmap = btrfs_bmap,
a52d9a80
CM		 2544 .invalidatepage = btrfs_invalidatepage,
2545 .releasepage = btrfs_releasepage,
2546 .set_page_dirty = __set_page_dirty_nobuffers,
39279cc3
CM		 2547};
2548
2549static struct address_space_operations btrfs_symlink_aops = {
2550 .readpage = btrfs_readpage,
2551 .writepage = btrfs_writepage,
2bf5a725
CM		 2552 .invalidatepage = btrfs_invalidatepage,
2553 .releasepage = btrfs_releasepage,
39279cc3
CM		 2554};
2555
2556static struct inode_operations btrfs_file_inode_operations = {
2557 .truncate = btrfs_truncate,
2558 .getattr = btrfs_getattr,
2559 .setattr = btrfs_setattr,
2560};
2561
618e21d5
JB		 2562static struct inode_operations btrfs_special_inode_operations = {
2563 .getattr = btrfs_getattr,
2564 .setattr = btrfs_setattr,
2565};
2566
39279cc3
CM		 2567static struct inode_operations btrfs_symlink_inode_operations = {
2568 .readlink = generic_readlink,
2569 .follow_link = page_follow_link_light,
2570 .put_link = page_put_link,
2571};
Clone of davem's net-next-2.6 tree