]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/btrfs/inode.c
projects / mirror_ubuntu-bionic-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: Simplify makefile
[mirror_ubuntu-bionic-kernel.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;
80 u64 num_blocks;
81 int ret;
82 u64 blocksize = 1 << inode->i_blkbits;
83
84 mutex_lock(&root->fs_info->fs_mutex);
85 trans = btrfs_start_transaction(root, 1);
86 btrfs_set_trans_block_group(trans, inode);
87 BUG_ON(!trans);
88 num_blocks = (end - start + blocksize) & ~(blocksize - 1);
89 ret = btrfs_drop_extents(trans, root, inode,
90 start, start + num_blocks, &alloc_hint);
91 num_blocks = num_blocks >> inode->i_blkbits;
92 ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_blocks, 0,
93 alloc_hint, (u64)-1, &ins, 1);
94 if (ret) {
95 WARN_ON(1);
96 goto out;
97 }
98 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
99 start, ins.objectid, ins.offset,
100 ins.offset);
101out:
102 btrfs_end_transaction(trans, root);
103 mutex_unlock(&root->fs_info->fs_mutex);
104 return ret;
105}
106
07157aac
CM		 107int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end)
108{
109 struct inode *inode = page->mapping->host;
110 struct btrfs_root *root = BTRFS_I(inode)->root;
111 struct btrfs_trans_handle *trans;
112 char *kaddr;
113 int ret;
114 u64 page_start = page->index << PAGE_CACHE_SHIFT;
115 size_t offset = start - page_start;
116
117 mutex_lock(&root->fs_info->fs_mutex);
118 trans = btrfs_start_transaction(root, 1);
119 btrfs_set_trans_block_group(trans, inode);
120 kaddr = kmap(page);
121 btrfs_csum_file_block(trans, root, inode->i_ino,
122 start, kaddr + offset, end - start + 1);
123 kunmap(page);
124 ret = btrfs_end_transaction(trans, root);
125 BUG_ON(ret);
126 mutex_unlock(&root->fs_info->fs_mutex);
127 return ret;
128}
129
130int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
131{
132 int ret = 0;
133 struct inode *inode = page->mapping->host;
134 struct btrfs_root *root = BTRFS_I(inode)->root;
135 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
136 struct btrfs_csum_item *item;
137 struct btrfs_path *path = NULL;
138 u64 private;
139
140 mutex_lock(&root->fs_info->fs_mutex);
141 path = btrfs_alloc_path();
142 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
143 if (IS_ERR(item)) {
144 ret = PTR_ERR(item);
145 /* a csum that isn't present is a preallocated region. */
146 if (ret == -ENOENT || ret == -EFBIG)
147 ret = 0;
148 private = 0;
149 goto out;
150 }
151 memcpy((char *)&private, &item->csum, BTRFS_CRC32_SIZE);
152 set_state_private(em_tree, start, private);
153out:
154 if (path)
155 btrfs_free_path(path);
156 mutex_unlock(&root->fs_info->fs_mutex);
157 return ret;
158}
159
160int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
161{
162 char csum[BTRFS_CRC32_SIZE];
163 size_t offset = start - (page->index << PAGE_CACHE_SHIFT);
164 struct inode *inode = page->mapping->host;
165 struct btrfs_root *root = BTRFS_I(inode)->root;
166 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
167 char *kaddr;
168 u64 private;
169 int ret;
170
171 ret = get_state_private(em_tree, start, &private);
172 kaddr = kmap_atomic(page, KM_IRQ0);
173 if (ret) {
174 goto zeroit;
175 }
176 ret = btrfs_csum_data(root, kaddr + offset, end - start + 1, csum);
177 BUG_ON(ret);
178 if (memcmp(csum, &private, BTRFS_CRC32_SIZE)) {
179 goto zeroit;
180 }
181 kunmap_atomic(kaddr, KM_IRQ0);
182 return 0;
183
184zeroit:
185 printk("btrfs csum failed ino %lu off %llu\n",
186 page->mapping->host->i_ino, (unsigned long long)start);
187 memset(kaddr + offset, 1, end - start + 1); flush_dcache_page(page);
188 kunmap_atomic(kaddr, KM_IRQ0);
189 return 0;
190}
b888db2b 191
39279cc3
CM		 192void btrfs_read_locked_inode(struct inode *inode)
193{
194 struct btrfs_path *path;
195 struct btrfs_inode_item *inode_item;
196 struct btrfs_root *root = BTRFS_I(inode)->root;
197 struct btrfs_key location;
198 u64 alloc_group_block;
618e21d5 199 u32 rdev;
39279cc3
CM		 200 int ret;
201
202 path = btrfs_alloc_path();
203 BUG_ON(!path);
39279cc3
CM		 204 mutex_lock(&root->fs_info->fs_mutex);
205
206 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
207 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
208 if (ret) {
209 btrfs_free_path(path);
210 goto make_bad;
211 }
212 inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
213 path->slots[0],
214 struct btrfs_inode_item);
215
216 inode->i_mode = btrfs_inode_mode(inode_item);
217 inode->i_nlink = btrfs_inode_nlink(inode_item);
218 inode->i_uid = btrfs_inode_uid(inode_item);
219 inode->i_gid = btrfs_inode_gid(inode_item);
220 inode->i_size = btrfs_inode_size(inode_item);
221 inode->i_atime.tv_sec = btrfs_timespec_sec(&inode_item->atime);
222 inode->i_atime.tv_nsec = btrfs_timespec_nsec(&inode_item->atime);
223 inode->i_mtime.tv_sec = btrfs_timespec_sec(&inode_item->mtime);
224 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(&inode_item->mtime);
225 inode->i_ctime.tv_sec = btrfs_timespec_sec(&inode_item->ctime);
226 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime);
227 inode->i_blocks = btrfs_inode_nblocks(inode_item);
228 inode->i_generation = btrfs_inode_generation(inode_item);
618e21d5
JB		 229 inode->i_rdev = 0;
230 rdev = btrfs_inode_rdev(inode_item);
39279cc3
CM		 231 alloc_group_block = btrfs_inode_block_group(inode_item);
232 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
233 alloc_group_block);
234
235 btrfs_free_path(path);
236 inode_item = NULL;
237
238 mutex_unlock(&root->fs_info->fs_mutex);
239
240 switch (inode->i_mode & S_IFMT) {
39279cc3
CM		 241 case S_IFREG:
242 inode->i_mapping->a_ops = &btrfs_aops;
07157aac 243 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
39279cc3
CM		 244 inode->i_fop = &btrfs_file_operations;
245 inode->i_op = &btrfs_file_inode_operations;
246 break;
247 case S_IFDIR:
248 inode->i_fop = &btrfs_dir_file_operations;
249 if (root == root->fs_info->tree_root)
250 inode->i_op = &btrfs_dir_ro_inode_operations;
251 else
252 inode->i_op = &btrfs_dir_inode_operations;
253 break;
254 case S_IFLNK:
255 inode->i_op = &btrfs_symlink_inode_operations;
256 inode->i_mapping->a_ops = &btrfs_symlink_aops;
257 break;
618e21d5
JB		 258 default:
259 init_special_inode(inode, inode->i_mode, rdev);
260 break;
39279cc3
CM		 261 }
262 return;
263
264make_bad:
265 btrfs_release_path(root, path);
266 btrfs_free_path(path);
267 mutex_unlock(&root->fs_info->fs_mutex);
268 make_bad_inode(inode);
269}
270
271static void fill_inode_item(struct btrfs_inode_item *item,
272 struct inode *inode)
273{
274 btrfs_set_inode_uid(item, inode->i_uid);
275 btrfs_set_inode_gid(item, inode->i_gid);
276 btrfs_set_inode_size(item, inode->i_size);
277 btrfs_set_inode_mode(item, inode->i_mode);
278 btrfs_set_inode_nlink(item, inode->i_nlink);
279 btrfs_set_timespec_sec(&item->atime, inode->i_atime.tv_sec);
280 btrfs_set_timespec_nsec(&item->atime, inode->i_atime.tv_nsec);
281 btrfs_set_timespec_sec(&item->mtime, inode->i_mtime.tv_sec);
282 btrfs_set_timespec_nsec(&item->mtime, inode->i_mtime.tv_nsec);
283 btrfs_set_timespec_sec(&item->ctime, inode->i_ctime.tv_sec);
284 btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec);
285 btrfs_set_inode_nblocks(item, inode->i_blocks);
286 btrfs_set_inode_generation(item, inode->i_generation);
618e21d5 287 btrfs_set_inode_rdev(item, inode->i_rdev);
39279cc3
CM		 288 btrfs_set_inode_block_group(item,
289 BTRFS_I(inode)->block_group->key.objectid);
290}
291
a52d9a80 292int btrfs_update_inode(struct btrfs_trans_handle *trans,
39279cc3
CM		 293 struct btrfs_root *root,
294 struct inode *inode)
295{
296 struct btrfs_inode_item *inode_item;
297 struct btrfs_path *path;
298 int ret;
299
300 path = btrfs_alloc_path();
301 BUG_ON(!path);
39279cc3
CM		 302 ret = btrfs_lookup_inode(trans, root, path,
303 &BTRFS_I(inode)->location, 1);
304 if (ret) {
305 if (ret > 0)
306 ret = -ENOENT;
307 goto failed;
308 }
309
310 inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
311 path->slots[0],
312 struct btrfs_inode_item);
313
314 fill_inode_item(inode_item, inode);
315 btrfs_mark_buffer_dirty(path->nodes[0]);
15ee9bc7 316 btrfs_set_inode_last_trans(trans, inode);
39279cc3
CM		 317 ret = 0;
318failed:
319 btrfs_release_path(root, path);
320 btrfs_free_path(path);
321 return ret;
322}
323
324
325static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
326 struct btrfs_root *root,
327 struct inode *dir,
328 struct dentry *dentry)
329{
330 struct btrfs_path *path;
331 const char *name = dentry->d_name.name;
332 int name_len = dentry->d_name.len;
333 int ret = 0;
334 u64 objectid;
335 struct btrfs_dir_item *di;
336
337 path = btrfs_alloc_path();
54aa1f4d
CM		 338 if (!path) {
339 ret = -ENOMEM;
340 goto err;
341 }
342
39279cc3
CM		 343 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
344 name, name_len, -1);
345 if (IS_ERR(di)) {
346 ret = PTR_ERR(di);
347 goto err;
348 }
349 if (!di) {
350 ret = -ENOENT;
351 goto err;
352 }
353 objectid = btrfs_disk_key_objectid(&di->location);
354 ret = btrfs_delete_one_dir_name(trans, root, path, di);
54aa1f4d
CM		 355 if (ret)
356 goto err;
39279cc3
CM		 357 btrfs_release_path(root, path);
358
359 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
360 objectid, name, name_len, -1);
361 if (IS_ERR(di)) {
362 ret = PTR_ERR(di);
363 goto err;
364 }
365 if (!di) {
366 ret = -ENOENT;
367 goto err;
368 }
369 ret = btrfs_delete_one_dir_name(trans, root, path, di);
39279cc3
CM		 370
371 dentry->d_inode->i_ctime = dir->i_ctime;
372err:
373 btrfs_free_path(path);
374 if (!ret) {
375 dir->i_size -= name_len * 2;
79c44584 376 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
39279cc3
CM		 377 btrfs_update_inode(trans, root, dir);
378 drop_nlink(dentry->d_inode);
54aa1f4d 379 ret = btrfs_update_inode(trans, root, dentry->d_inode);
39279cc3
CM		 380 dir->i_sb->s_dirt = 1;
381 }
382 return ret;
383}
384
385static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
386{
387 struct btrfs_root *root;
388 struct btrfs_trans_handle *trans;
389 int ret;
390
391 root = BTRFS_I(dir)->root;
392 mutex_lock(&root->fs_info->fs_mutex);
393 trans = btrfs_start_transaction(root, 1);
394 btrfs_set_trans_block_group(trans, dir);
395 ret = btrfs_unlink_trans(trans, root, dir, dentry);
396 btrfs_end_transaction(trans, root);
397 mutex_unlock(&root->fs_info->fs_mutex);
398 btrfs_btree_balance_dirty(root);
399 return ret;
400}
401
402static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
403{
404 struct inode *inode = dentry->d_inode;
405 int err;
406 int ret;
407 struct btrfs_root *root = BTRFS_I(dir)->root;
408 struct btrfs_path *path;
409 struct btrfs_key key;
410 struct btrfs_trans_handle *trans;
411 struct btrfs_key found_key;
412 int found_type;
413 struct btrfs_leaf *leaf;
414 char *goodnames = "..";
415
416 path = btrfs_alloc_path();
417 BUG_ON(!path);
39279cc3
CM		 418 mutex_lock(&root->fs_info->fs_mutex);
419 trans = btrfs_start_transaction(root, 1);
420 btrfs_set_trans_block_group(trans, dir);
421 key.objectid = inode->i_ino;
422 key.offset = (u64)-1;
423 key.flags = (u32)-1;
424 while(1) {
425 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
426 if (ret < 0) {
427 err = ret;
428 goto out;
429 }
430 BUG_ON(ret == 0);
431 if (path->slots[0] == 0) {
432 err = -ENOENT;
433 goto out;
434 }
435 path->slots[0]--;
436 leaf = btrfs_buffer_leaf(path->nodes[0]);
437 btrfs_disk_key_to_cpu(&found_key,
438 &leaf->items[path->slots[0]].key);
439 found_type = btrfs_key_type(&found_key);
440 if (found_key.objectid != inode->i_ino) {
441 err = -ENOENT;
442 goto out;
443 }
444 if ((found_type != BTRFS_DIR_ITEM_KEY &&
445 found_type != BTRFS_DIR_INDEX_KEY) ||
446 (!btrfs_match_dir_item_name(root, path, goodnames, 2) &&
447 !btrfs_match_dir_item_name(root, path, goodnames, 1))) {
448 err = -ENOTEMPTY;
449 goto out;
450 }
451 ret = btrfs_del_item(trans, root, path);
452 BUG_ON(ret);
453
454 if (found_type == BTRFS_DIR_ITEM_KEY && found_key.offset == 1)
455 break;
456 btrfs_release_path(root, path);
457 }
458 ret = 0;
459 btrfs_release_path(root, path);
460
461 /* now the directory is empty */
462 err = btrfs_unlink_trans(trans, root, dir, dentry);
463 if (!err) {
464 inode->i_size = 0;
465 }
466out:
467 btrfs_release_path(root, path);
468 btrfs_free_path(path);
469 mutex_unlock(&root->fs_info->fs_mutex);
470 ret = btrfs_end_transaction(trans, root);
471 btrfs_btree_balance_dirty(root);
472 if (ret && !err)
473 err = ret;
474 return err;
475}
476
477static int btrfs_free_inode(struct btrfs_trans_handle *trans,
478 struct btrfs_root *root,
479 struct inode *inode)
480{
481 struct btrfs_path *path;
482 int ret;
483
484 clear_inode(inode);
485
486 path = btrfs_alloc_path();
487 BUG_ON(!path);
39279cc3
CM		 488 ret = btrfs_lookup_inode(trans, root, path,
489 &BTRFS_I(inode)->location, -1);
54aa1f4d
CM		 490 if (ret > 0)
491 ret = -ENOENT;
492 if (!ret)
493 ret = btrfs_del_item(trans, root, path);
39279cc3
CM		 494 btrfs_free_path(path);
495 return ret;
496}
497
39279cc3
CM		 498/*
499 * this can truncate away extent items, csum items and directory items.
500 * It starts at a high offset and removes keys until it can't find
501 * any higher than i_size.
502 *
503 * csum items that cross the new i_size are truncated to the new size
504 * as well.
505 */
506static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
507 struct btrfs_root *root,
508 struct inode *inode)
509{
510 int ret;
511 struct btrfs_path *path;
512 struct btrfs_key key;
513 struct btrfs_disk_key *found_key;
514 u32 found_type;
515 struct btrfs_leaf *leaf;
516 struct btrfs_file_extent_item *fi;
517 u64 extent_start = 0;
518 u64 extent_num_blocks = 0;
519 u64 item_end = 0;
520 int found_extent;
521 int del_item;
522
a52d9a80 523 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
39279cc3 524 path = btrfs_alloc_path();
3c69faec 525 path->reada = -1;
39279cc3
CM		 526 BUG_ON(!path);
527 /* FIXME, add redo link to tree so we don't leak on crash */
528 key.objectid = inode->i_ino;
529 key.offset = (u64)-1;
530 key.flags = (u32)-1;
531 while(1) {
532 btrfs_init_path(path);
533 fi = NULL;
534 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
535 if (ret < 0) {
536 goto error;
537 }
538 if (ret > 0) {
539 BUG_ON(path->slots[0] == 0);
540 path->slots[0]--;
541 }
39279cc3
CM		 542 leaf = btrfs_buffer_leaf(path->nodes[0]);
543 found_key = &leaf->items[path->slots[0]].key;
544 found_type = btrfs_disk_key_type(found_key);
545
546 if (btrfs_disk_key_objectid(found_key) != inode->i_ino)
547 break;
548 if (found_type != BTRFS_CSUM_ITEM_KEY &&
549 found_type != BTRFS_DIR_ITEM_KEY &&
550 found_type != BTRFS_DIR_INDEX_KEY &&
551 found_type != BTRFS_EXTENT_DATA_KEY)
552 break;
553
554 item_end = btrfs_disk_key_offset(found_key);
555 if (found_type == BTRFS_EXTENT_DATA_KEY) {
556 fi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
557 path->slots[0],
558 struct btrfs_file_extent_item);
559 if (btrfs_file_extent_type(fi) !=
560 BTRFS_FILE_EXTENT_INLINE) {
561 item_end += btrfs_file_extent_num_blocks(fi) <<
562 inode->i_blkbits;
563 }
564 }
565 if (found_type == BTRFS_CSUM_ITEM_KEY) {
566 ret = btrfs_csum_truncate(trans, root, path,
567 inode->i_size);
568 BUG_ON(ret);
569 }
570 if (item_end < inode->i_size) {
b888db2b
CM		 571 if (found_type == BTRFS_DIR_ITEM_KEY) {
572 found_type = BTRFS_INODE_ITEM_KEY;
573 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
574 found_type = BTRFS_CSUM_ITEM_KEY;
575 } else if (found_type) {
576 found_type--;
577 } else {
578 break;
39279cc3 579 }
b888db2b
CM		 580 btrfs_set_key_type(&key, found_type - 1);
581 continue;
39279cc3
CM		 582 }
583 if (btrfs_disk_key_offset(found_key) >= inode->i_size)
584 del_item = 1;
585 else
586 del_item = 0;
587 found_extent = 0;
588
589 /* FIXME, shrink the extent if the ref count is only 1 */
590 if (found_type == BTRFS_EXTENT_DATA_KEY &&
591 btrfs_file_extent_type(fi) !=
592 BTRFS_FILE_EXTENT_INLINE) {
593 u64 num_dec;
594 if (!del_item) {
595 u64 orig_num_blocks =
596 btrfs_file_extent_num_blocks(fi);
597 extent_num_blocks = inode->i_size -
598 btrfs_disk_key_offset(found_key) +
599 root->blocksize - 1;
600 extent_num_blocks >>= inode->i_blkbits;
601 btrfs_set_file_extent_num_blocks(fi,
602 extent_num_blocks);
603 inode->i_blocks -= (orig_num_blocks -
604 extent_num_blocks) << 3;
ccd467d6 605 btrfs_mark_buffer_dirty(path->nodes[0]);
39279cc3
CM		 606 } else {
607 extent_start =
608 btrfs_file_extent_disk_blocknr(fi);
609 extent_num_blocks =
610 btrfs_file_extent_disk_num_blocks(fi);
611 /* FIXME blocksize != 4096 */
612 num_dec = btrfs_file_extent_num_blocks(fi) << 3;
613 if (extent_start != 0) {
614 found_extent = 1;
615 inode->i_blocks -= num_dec;
616 }
617 }
618 }
619 if (del_item) {
620 ret = btrfs_del_item(trans, root, path);
54aa1f4d
CM		 621 if (ret)
622 goto error;
39279cc3
CM		 623 } else {
624 break;
625 }
626 btrfs_release_path(root, path);
627 if (found_extent) {
628 ret = btrfs_free_extent(trans, root, extent_start,
629 extent_num_blocks, 0);
630 BUG_ON(ret);
631 }
632 }
633 ret = 0;
634error:
635 btrfs_release_path(root, path);
636 btrfs_free_path(path);
637 inode->i_sb->s_dirt = 1;
638 return ret;
639}
640
b888db2b 641static int btrfs_cow_one_page(struct inode *inode, struct page *page,
a52d9a80
CM		 642 size_t zero_start)
643{
644 char *kaddr;
645 int ret = 0;
b888db2b 646 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
a52d9a80 647 u64 page_start = page->index << PAGE_CACHE_SHIFT;
b888db2b 648 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
a52d9a80
CM		 649
650 if (!PagePrivate(page)) {
651 SetPagePrivate(page);
652 set_page_private(page, 1);
b888db2b 653 WARN_ON(!page->mapping->a_ops->invalidatepage);
a52d9a80
CM		 654 page_cache_get(page);
655 }
656
b888db2b
CM		 657 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
658 set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start,
659 page_end, GFP_NOFS);
a52d9a80 660 if (zero_start != PAGE_CACHE_SIZE) {
b888db2b 661 kaddr = kmap(page);
a52d9a80
CM		 662 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
663 flush_dcache_page(page);
b888db2b 664 kunmap(page);
a52d9a80 665 }
b888db2b
CM		 666 set_page_dirty(page);
667 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
a52d9a80 668
a52d9a80
CM		 669 return ret;
670}
671
39279cc3
CM		 672/*
673 * taken from block_truncate_page, but does cow as it zeros out
674 * any bytes left in the last page in the file.
675 */
676static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
677{
678 struct inode *inode = mapping->host;
679 unsigned blocksize = 1 << inode->i_blkbits;
680 pgoff_t index = from >> PAGE_CACHE_SHIFT;
681 unsigned offset = from & (PAGE_CACHE_SIZE-1);
682 struct page *page;
39279cc3 683 int ret = 0;
a52d9a80 684 u64 page_start;
39279cc3
CM		 685
686 if ((offset & (blocksize - 1)) == 0)
687 goto out;
688
011410bd 689 down_read(&BTRFS_I(inode)->root->snap_sem);
39279cc3
CM		 690 ret = -ENOMEM;
691 page = grab_cache_page(mapping, index);
692 if (!page)
693 goto out;
39279cc3 694 if (!PageUptodate(page)) {
9ebefb18 695 ret = btrfs_readpage(NULL, page);
39279cc3
CM		 696 lock_page(page);
697 if (!PageUptodate(page)) {
698 ret = -EIO;
699 goto out;
700 }
701 }
a52d9a80
CM		 702 page_start = page->index << PAGE_CACHE_SHIFT;
703
b888db2b 704 ret = btrfs_cow_one_page(inode, page, offset);
39279cc3 705
39279cc3
CM		 706 unlock_page(page);
707 page_cache_release(page);
011410bd 708 up_read(&BTRFS_I(inode)->root->snap_sem);
39279cc3
CM		 709out:
710 return ret;
711}
712
713static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
714{
715 struct inode *inode = dentry->d_inode;
716 int err;
717
718 err = inode_change_ok(inode, attr);
719 if (err)
720 return err;
721
722 if (S_ISREG(inode->i_mode) &&
723 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
724 struct btrfs_trans_handle *trans;
725 struct btrfs_root *root = BTRFS_I(inode)->root;
2bf5a725
CM		 726 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
727
39279cc3
CM		 728 u64 mask = root->blocksize - 1;
729 u64 pos = (inode->i_size + mask) & ~mask;
2bf5a725 730 u64 block_end = attr->ia_size | mask;
39279cc3 731 u64 hole_size;
2bf5a725 732 u64 alloc_hint;
39279cc3
CM		 733
734 if (attr->ia_size <= pos)
735 goto out;
736
737 btrfs_truncate_page(inode->i_mapping, inode->i_size);
738
2bf5a725 739 lock_extent(em_tree, pos, block_end, GFP_NOFS);
39279cc3 740 hole_size = (attr->ia_size - pos + mask) & ~mask;
39279cc3
CM		 741
742 mutex_lock(&root->fs_info->fs_mutex);
743 trans = btrfs_start_transaction(root, 1);
744 btrfs_set_trans_block_group(trans, inode);
2bf5a725
CM		 745 err = btrfs_drop_extents(trans, root, inode,
746 pos, pos + hole_size, &alloc_hint);
747
748 hole_size >>= inode->i_blkbits;
749
39279cc3
CM		 750 err = btrfs_insert_file_extent(trans, root, inode->i_ino,
751 pos, 0, 0, hole_size);
39279cc3
CM		 752 btrfs_end_transaction(trans, root);
753 mutex_unlock(&root->fs_info->fs_mutex);
2bf5a725 754 unlock_extent(em_tree, pos, block_end, GFP_NOFS);
54aa1f4d
CM		 755 if (err)
756 return err;
39279cc3
CM		 757 }
758out:
759 err = inode_setattr(inode, attr);
760
761 return err;
762}
763void btrfs_delete_inode(struct inode *inode)
764{
765 struct btrfs_trans_handle *trans;
766 struct btrfs_root *root = BTRFS_I(inode)->root;
767 int ret;
768
769 truncate_inode_pages(&inode->i_data, 0);
770 if (is_bad_inode(inode)) {
771 goto no_delete;
772 }
773 inode->i_size = 0;
774 mutex_lock(&root->fs_info->fs_mutex);
775 trans = btrfs_start_transaction(root, 1);
776 btrfs_set_trans_block_group(trans, inode);
777 ret = btrfs_truncate_in_trans(trans, root, inode);
54aa1f4d
CM		 778 if (ret)
779 goto no_delete_lock;
780 ret = btrfs_free_inode(trans, root, inode);
781 if (ret)
782 goto no_delete_lock;
39279cc3
CM		 783 btrfs_end_transaction(trans, root);
784 mutex_unlock(&root->fs_info->fs_mutex);
785 btrfs_btree_balance_dirty(root);
786 return;
54aa1f4d
CM		 787
788no_delete_lock:
789 btrfs_end_transaction(trans, root);
790 mutex_unlock(&root->fs_info->fs_mutex);
791 btrfs_btree_balance_dirty(root);
39279cc3
CM		 792no_delete:
793 clear_inode(inode);
794}
795
796/*
797 * this returns the key found in the dir entry in the location pointer.
798 * If no dir entries were found, location->objectid is 0.
799 */
800static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
801 struct btrfs_key *location)
802{
803 const char *name = dentry->d_name.name;
804 int namelen = dentry->d_name.len;
805 struct btrfs_dir_item *di;
806 struct btrfs_path *path;
807 struct btrfs_root *root = BTRFS_I(dir)->root;
808 int ret;
809
810 path = btrfs_alloc_path();
811 BUG_ON(!path);
39279cc3
CM		 812 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
813 namelen, 0);
814 if (!di || IS_ERR(di)) {
815 location->objectid = 0;
816 ret = 0;
817 goto out;
818 }
819 btrfs_disk_key_to_cpu(location, &di->location);
820out:
821 btrfs_release_path(root, path);
822 btrfs_free_path(path);
823 return ret;
824}
825
826/*
827 * when we hit a tree root in a directory, the btrfs part of the inode
828 * needs to be changed to reflect the root directory of the tree root. This
829 * is kind of like crossing a mount point.
830 */
831static int fixup_tree_root_location(struct btrfs_root *root,
832 struct btrfs_key *location,
58176a96
JB		 833 struct btrfs_root **sub_root,
834 struct dentry *dentry)
39279cc3
CM		 835{
836 struct btrfs_path *path;
837 struct btrfs_root_item *ri;
838
839 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
840 return 0;
841 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
842 return 0;
843
844 path = btrfs_alloc_path();
845 BUG_ON(!path);
846 mutex_lock(&root->fs_info->fs_mutex);
847
58176a96
JB		 848 *sub_root = btrfs_read_fs_root(root->fs_info, location,
849 dentry->d_name.name,
850 dentry->d_name.len);
39279cc3
CM		 851 if (IS_ERR(*sub_root))
852 return PTR_ERR(*sub_root);
853
854 ri = &(*sub_root)->root_item;
855 location->objectid = btrfs_root_dirid(ri);
856 location->flags = 0;
857 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
858 location->offset = 0;
859
860 btrfs_free_path(path);
861 mutex_unlock(&root->fs_info->fs_mutex);
862 return 0;
863}
864
865static int btrfs_init_locked_inode(struct inode *inode, void *p)
866{
867 struct btrfs_iget_args *args = p;
868 inode->i_ino = args->ino;
869 BTRFS_I(inode)->root = args->root;
b888db2b
CM		 870 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
871 inode->i_mapping, GFP_NOFS);
39279cc3
CM		 872 return 0;
873}
874
875static int btrfs_find_actor(struct inode *inode, void *opaque)
876{
877 struct btrfs_iget_args *args = opaque;
878 return (args->ino == inode->i_ino &&
879 args->root == BTRFS_I(inode)->root);
880}
881
882struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
883 struct btrfs_root *root)
884{
885 struct inode *inode;
886 struct btrfs_iget_args args;
887 args.ino = objectid;
888 args.root = root;
889
890 inode = iget5_locked(s, objectid, btrfs_find_actor,
891 btrfs_init_locked_inode,
892 (void *)&args);
893 return inode;
894}
895
896static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
897 struct nameidata *nd)
898{
899 struct inode * inode;
900 struct btrfs_inode *bi = BTRFS_I(dir);
901 struct btrfs_root *root = bi->root;
902 struct btrfs_root *sub_root = root;
903 struct btrfs_key location;
904 int ret;
905
906 if (dentry->d_name.len > BTRFS_NAME_LEN)
907 return ERR_PTR(-ENAMETOOLONG);
908 mutex_lock(&root->fs_info->fs_mutex);
909 ret = btrfs_inode_by_name(dir, dentry, &location);
910 mutex_unlock(&root->fs_info->fs_mutex);
911 if (ret < 0)
912 return ERR_PTR(ret);
913 inode = NULL;
914 if (location.objectid) {
58176a96
JB		 915 ret = fixup_tree_root_location(root, &location, &sub_root,
916 dentry);
39279cc3
CM		 917 if (ret < 0)
918 return ERR_PTR(ret);
919 if (ret > 0)
920 return ERR_PTR(-ENOENT);
921 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
922 sub_root);
923 if (!inode)
924 return ERR_PTR(-EACCES);
925 if (inode->i_state & I_NEW) {
926 /* the inode and parent dir are two different roots */
927 if (sub_root != root) {
928 igrab(inode);
929 sub_root->inode = inode;
930 }
931 BTRFS_I(inode)->root = sub_root;
932 memcpy(&BTRFS_I(inode)->location, &location,
933 sizeof(location));
934 btrfs_read_locked_inode(inode);
935 unlock_new_inode(inode);
936 }
937 }
938 return d_splice_alias(inode, dentry);
939}
940
39279cc3
CM		 941static unsigned char btrfs_filetype_table[] = {
942 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
943};
944
945static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
946{
947 struct inode *inode = filp->f_path.dentry->d_inode;
948 struct btrfs_root *root = BTRFS_I(inode)->root;
949 struct btrfs_item *item;
950 struct btrfs_dir_item *di;
951 struct btrfs_key key;
952 struct btrfs_path *path;
953 int ret;
954 u32 nritems;
955 struct btrfs_leaf *leaf;
956 int slot;
957 int advance;
958 unsigned char d_type;
959 int over = 0;
960 u32 di_cur;
961 u32 di_total;
962 u32 di_len;
963 int key_type = BTRFS_DIR_INDEX_KEY;
964
965 /* FIXME, use a real flag for deciding about the key type */
966 if (root->fs_info->tree_root == root)
967 key_type = BTRFS_DIR_ITEM_KEY;
968 mutex_lock(&root->fs_info->fs_mutex);
969 key.objectid = inode->i_ino;
970 key.flags = 0;
971 btrfs_set_key_type(&key, key_type);
972 key.offset = filp->f_pos;
973 path = btrfs_alloc_path();
2cc58cf2 974 path->reada = 2;
39279cc3
CM		 975 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
976 if (ret < 0)
977 goto err;
978 advance = 0;
39279cc3
CM		 979 while(1) {
980 leaf = btrfs_buffer_leaf(path->nodes[0]);
981 nritems = btrfs_header_nritems(&leaf->header);
982 slot = path->slots[0];
983 if (advance || slot >= nritems) {
984 if (slot >= nritems -1) {
39279cc3
CM		 985 ret = btrfs_next_leaf(root, path);
986 if (ret)
987 break;
988 leaf = btrfs_buffer_leaf(path->nodes[0]);
989 nritems = btrfs_header_nritems(&leaf->header);
990 slot = path->slots[0];
991 } else {
992 slot++;
993 path->slots[0]++;
994 }
995 }
996 advance = 1;
997 item = leaf->items + slot;
998 if (btrfs_disk_key_objectid(&item->key) != key.objectid)
999 break;
1000 if (btrfs_disk_key_type(&item->key) != key_type)
1001 break;
1002 if (btrfs_disk_key_offset(&item->key) < filp->f_pos)
1003 continue;
1004 filp->f_pos = btrfs_disk_key_offset(&item->key);
1005 advance = 1;
1006 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1007 di_cur = 0;
1008 di_total = btrfs_item_size(leaf->items + slot);
1009 while(di_cur < di_total) {
1010 d_type = btrfs_filetype_table[btrfs_dir_type(di)];
1011 over = filldir(dirent, (const char *)(di + 1),
1012 btrfs_dir_name_len(di),
1013 btrfs_disk_key_offset(&item->key),
1014 btrfs_disk_key_objectid(&di->location),
1015 d_type);
1016 if (over)
1017 goto nopos;
1018 di_len = btrfs_dir_name_len(di) + sizeof(*di);
1019 di_cur += di_len;
1020 di = (struct btrfs_dir_item *)((char *)di + di_len);
1021 }
1022 }
1023 filp->f_pos++;
1024nopos:
1025 ret = 0;
1026err:
1027 btrfs_release_path(root, path);
1028 btrfs_free_path(path);
1029 mutex_unlock(&root->fs_info->fs_mutex);
1030 return ret;
1031}
1032
1033int btrfs_write_inode(struct inode *inode, int wait)
1034{
1035 struct btrfs_root *root = BTRFS_I(inode)->root;
1036 struct btrfs_trans_handle *trans;
1037 int ret = 0;
1038
1039 if (wait) {
1040 mutex_lock(&root->fs_info->fs_mutex);
1041 trans = btrfs_start_transaction(root, 1);
1042 btrfs_set_trans_block_group(trans, inode);
1043 ret = btrfs_commit_transaction(trans, root);
1044 mutex_unlock(&root->fs_info->fs_mutex);
1045 }
1046 return ret;
1047}
1048
1049/*
54aa1f4d 1050 * This is somewhat expensive, updating the tree every time the
39279cc3
CM		 1051 * inode changes. But, it is most likely to find the inode in cache.
1052 * FIXME, needs more benchmarking...there are no reasons other than performance
1053 * to keep or drop this code.
1054 */
1055void btrfs_dirty_inode(struct inode *inode)
1056{
1057 struct btrfs_root *root = BTRFS_I(inode)->root;
1058 struct btrfs_trans_handle *trans;
1059
1060 mutex_lock(&root->fs_info->fs_mutex);
1061 trans = btrfs_start_transaction(root, 1);
1062 btrfs_set_trans_block_group(trans, inode);
1063 btrfs_update_inode(trans, root, inode);
1064 btrfs_end_transaction(trans, root);
1065 mutex_unlock(&root->fs_info->fs_mutex);
39279cc3
CM		 1066}
1067
1068static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1069 struct btrfs_root *root,
1070 u64 objectid,
1071 struct btrfs_block_group_cache *group,
1072 int mode)
1073{
1074 struct inode *inode;
1075 struct btrfs_inode_item inode_item;
1076 struct btrfs_key *location;
1077 int ret;
1078 int owner;
1079
1080 inode = new_inode(root->fs_info->sb);
1081 if (!inode)
1082 return ERR_PTR(-ENOMEM);
1083
b888db2b
CM		 1084 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1085 inode->i_mapping, GFP_NOFS);
39279cc3 1086 BTRFS_I(inode)->root = root;
b888db2b 1087
39279cc3
CM		 1088 if (mode & S_IFDIR)
1089 owner = 0;
1090 else
1091 owner = 1;
1092 group = btrfs_find_block_group(root, group, 0, 0, owner);
1093 BTRFS_I(inode)->block_group = group;
1094
1095 inode->i_uid = current->fsuid;
1096 inode->i_gid = current->fsgid;
1097 inode->i_mode = mode;
1098 inode->i_ino = objectid;
1099 inode->i_blocks = 0;
1100 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1101 fill_inode_item(&inode_item, inode);
1102 location = &BTRFS_I(inode)->location;
1103 location->objectid = objectid;
1104 location->flags = 0;
1105 location->offset = 0;
1106 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1107
1108 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
54aa1f4d
CM		 1109 if (ret)
1110 return ERR_PTR(ret);
39279cc3
CM		 1111 insert_inode_hash(inode);
1112 return inode;
1113}
1114
1115static inline u8 btrfs_inode_type(struct inode *inode)
1116{
1117 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1118}
1119
1120static int btrfs_add_link(struct btrfs_trans_handle *trans,
1121 struct dentry *dentry, struct inode *inode)
1122{
1123 int ret;
1124 struct btrfs_key key;
1125 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
79c44584 1126 struct inode *parent_inode;
39279cc3
CM		 1127 key.objectid = inode->i_ino;
1128 key.flags = 0;
1129 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1130 key.offset = 0;
1131
1132 ret = btrfs_insert_dir_item(trans, root,
1133 dentry->d_name.name, dentry->d_name.len,
1134 dentry->d_parent->d_inode->i_ino,
1135 &key, btrfs_inode_type(inode));
1136 if (ret == 0) {
79c44584
CM		 1137 parent_inode = dentry->d_parent->d_inode;
1138 parent_inode->i_size += dentry->d_name.len * 2;
1139 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
39279cc3
CM		 1140 ret = btrfs_update_inode(trans, root,
1141 dentry->d_parent->d_inode);
1142 }
1143 return ret;
1144}
1145
1146static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1147 struct dentry *dentry, struct inode *inode)
1148{
1149 int err = btrfs_add_link(trans, dentry, inode);
1150 if (!err) {
1151 d_instantiate(dentry, inode);
1152 return 0;
1153 }
1154 if (err > 0)
1155 err = -EEXIST;
1156 return err;
1157}
1158
618e21d5
JB		 1159static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1160 int mode, dev_t rdev)
1161{
1162 struct btrfs_trans_handle *trans;
1163 struct btrfs_root *root = BTRFS_I(dir)->root;
1164 struct inode *inode;
1165 int err;
1166 int drop_inode = 0;
1167 u64 objectid;
1168
1169 if (!new_valid_dev(rdev))
1170 return -EINVAL;
1171
1172 mutex_lock(&root->fs_info->fs_mutex);
1173 trans = btrfs_start_transaction(root, 1);
1174 btrfs_set_trans_block_group(trans, dir);
1175
1176 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1177 if (err) {
1178 err = -ENOSPC;
1179 goto out_unlock;
1180 }
1181
1182 inode = btrfs_new_inode(trans, root, objectid,
1183 BTRFS_I(dir)->block_group, mode);
1184 err = PTR_ERR(inode);
1185 if (IS_ERR(inode))
1186 goto out_unlock;
1187
1188 btrfs_set_trans_block_group(trans, inode);
1189 err = btrfs_add_nondir(trans, dentry, inode);
1190 if (err)
1191 drop_inode = 1;
1192 else {
1193 inode->i_op = &btrfs_special_inode_operations;
1194 init_special_inode(inode, inode->i_mode, rdev);
1b4ab1bb 1195 btrfs_update_inode(trans, root, inode);
618e21d5
JB		 1196 }
1197 dir->i_sb->s_dirt = 1;
1198 btrfs_update_inode_block_group(trans, inode);
1199 btrfs_update_inode_block_group(trans, dir);
1200out_unlock:
1201 btrfs_end_transaction(trans, root);
1202 mutex_unlock(&root->fs_info->fs_mutex);
1203
1204 if (drop_inode) {
1205 inode_dec_link_count(inode);
1206 iput(inode);
1207 }
1208 btrfs_btree_balance_dirty(root);
1209 return err;
1210}
1211
39279cc3
CM		 1212static int btrfs_create(struct inode *dir, struct dentry *dentry,
1213 int mode, struct nameidata *nd)
1214{
1215 struct btrfs_trans_handle *trans;
1216 struct btrfs_root *root = BTRFS_I(dir)->root;
1217 struct inode *inode;
1218 int err;
1219 int drop_inode = 0;
1220 u64 objectid;
1221
1222 mutex_lock(&root->fs_info->fs_mutex);
1223 trans = btrfs_start_transaction(root, 1);
1224 btrfs_set_trans_block_group(trans, dir);
1225
1226 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1227 if (err) {
1228 err = -ENOSPC;
1229 goto out_unlock;
1230 }
1231
1232 inode = btrfs_new_inode(trans, root, objectid,
1233 BTRFS_I(dir)->block_group, mode);
1234 err = PTR_ERR(inode);
1235 if (IS_ERR(inode))
1236 goto out_unlock;
1237
1238 btrfs_set_trans_block_group(trans, inode);
1239 err = btrfs_add_nondir(trans, dentry, inode);
1240 if (err)
1241 drop_inode = 1;
1242 else {
1243 inode->i_mapping->a_ops = &btrfs_aops;
1244 inode->i_fop = &btrfs_file_operations;
1245 inode->i_op = &btrfs_file_inode_operations;
a52d9a80
CM		 1246 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1247 inode->i_mapping, GFP_NOFS);
07157aac 1248 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
39279cc3
CM		 1249 }
1250 dir->i_sb->s_dirt = 1;
1251 btrfs_update_inode_block_group(trans, inode);
1252 btrfs_update_inode_block_group(trans, dir);
1253out_unlock:
1254 btrfs_end_transaction(trans, root);
1255 mutex_unlock(&root->fs_info->fs_mutex);
1256
1257 if (drop_inode) {
1258 inode_dec_link_count(inode);
1259 iput(inode);
1260 }
1261 btrfs_btree_balance_dirty(root);
1262 return err;
1263}
1264
1265static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1266 struct dentry *dentry)
1267{
1268 struct btrfs_trans_handle *trans;
1269 struct btrfs_root *root = BTRFS_I(dir)->root;
1270 struct inode *inode = old_dentry->d_inode;
1271 int err;
1272 int drop_inode = 0;
1273
1274 if (inode->i_nlink == 0)
1275 return -ENOENT;
1276
1277 inc_nlink(inode);
1278 mutex_lock(&root->fs_info->fs_mutex);
1279 trans = btrfs_start_transaction(root, 1);
1280 btrfs_set_trans_block_group(trans, dir);
1281 atomic_inc(&inode->i_count);
1282 err = btrfs_add_nondir(trans, dentry, inode);
1283 if (err)
1284 drop_inode = 1;
1285 dir->i_sb->s_dirt = 1;
1286 btrfs_update_inode_block_group(trans, dir);
54aa1f4d
CM		 1287 err = btrfs_update_inode(trans, root, inode);
1288 if (err)
1289 drop_inode = 1;
39279cc3
CM		 1290
1291 btrfs_end_transaction(trans, root);
1292 mutex_unlock(&root->fs_info->fs_mutex);
1293
1294 if (drop_inode) {
1295 inode_dec_link_count(inode);
1296 iput(inode);
1297 }
1298 btrfs_btree_balance_dirty(root);
1299 return err;
1300}
1301
1302static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
1303 struct btrfs_root *root,
1304 u64 objectid, u64 dirid)
1305{
1306 int ret;
1307 char buf[2];
1308 struct btrfs_key key;
1309
1310 buf[0] = '.';
1311 buf[1] = '.';
1312
1313 key.objectid = objectid;
1314 key.offset = 0;
1315 key.flags = 0;
1316 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1317
1318 ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid,
1319 &key, BTRFS_FT_DIR);
1320 if (ret)
1321 goto error;
1322 key.objectid = dirid;
1323 ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid,
1324 &key, BTRFS_FT_DIR);
1325 if (ret)
1326 goto error;
1327error:
1328 return ret;
1329}
1330
1331static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1332{
1333 struct inode *inode;
1334 struct btrfs_trans_handle *trans;
1335 struct btrfs_root *root = BTRFS_I(dir)->root;
1336 int err = 0;
1337 int drop_on_err = 0;
1338 u64 objectid;
1339
1340 mutex_lock(&root->fs_info->fs_mutex);
1341 trans = btrfs_start_transaction(root, 1);
1342 btrfs_set_trans_block_group(trans, dir);
1343 if (IS_ERR(trans)) {
1344 err = PTR_ERR(trans);
1345 goto out_unlock;
1346 }
1347
1348 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1349 if (err) {
1350 err = -ENOSPC;
1351 goto out_unlock;
1352 }
1353
1354 inode = btrfs_new_inode(trans, root, objectid,
1355 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1356 if (IS_ERR(inode)) {
1357 err = PTR_ERR(inode);
1358 goto out_fail;
1359 }
1360 drop_on_err = 1;
1361 inode->i_op = &btrfs_dir_inode_operations;
1362 inode->i_fop = &btrfs_dir_file_operations;
1363 btrfs_set_trans_block_group(trans, inode);
1364
1365 err = btrfs_make_empty_dir(trans, root, inode->i_ino, dir->i_ino);
1366 if (err)
1367 goto out_fail;
1368
1369 inode->i_size = 6;
1370 err = btrfs_update_inode(trans, root, inode);
1371 if (err)
1372 goto out_fail;
1373 err = btrfs_add_link(trans, dentry, inode);
1374 if (err)
1375 goto out_fail;
1376 d_instantiate(dentry, inode);
1377 drop_on_err = 0;
1378 dir->i_sb->s_dirt = 1;
1379 btrfs_update_inode_block_group(trans, inode);
1380 btrfs_update_inode_block_group(trans, dir);
1381
1382out_fail:
1383 btrfs_end_transaction(trans, root);
1384out_unlock:
1385 mutex_unlock(&root->fs_info->fs_mutex);
1386 if (drop_on_err)
1387 iput(inode);
1388 btrfs_btree_balance_dirty(root);
1389 return err;
1390}
1391
a52d9a80
CM		 1392struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
1393 size_t page_offset, u64 start, u64 end,
1394 int create)
1395{
1396 int ret;
1397 int err = 0;
1398 u64 blocknr;
1399 u64 extent_start = 0;
1400 u64 extent_end = 0;
1401 u64 objectid = inode->i_ino;
1402 u32 found_type;
1403 int failed_insert = 0;
1404 struct btrfs_path *path;
1405 struct btrfs_root *root = BTRFS_I(inode)->root;
1406 struct btrfs_file_extent_item *item;
1407 struct btrfs_leaf *leaf;
1408 struct btrfs_disk_key *found_key;
1409 struct extent_map *em = NULL;
1410 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1411 struct btrfs_trans_handle *trans = NULL;
1412
1413 path = btrfs_alloc_path();
1414 BUG_ON(!path);
1415 mutex_lock(&root->fs_info->fs_mutex);
1416
1417again:
1418 em = lookup_extent_mapping(em_tree, start, end);
1419 if (em) {
1420 goto out;
1421 }
1422 if (!em) {
1423 em = alloc_extent_map(GFP_NOFS);
1424 if (!em) {
1425 err = -ENOMEM;
1426 goto out;
1427 }
1428 em->start = 0;
1429 em->end = 0;
1430 }
1431 em->bdev = inode->i_sb->s_bdev;
1432 ret = btrfs_lookup_file_extent(NULL, root, path,
1433 objectid, start, 0);
1434 if (ret < 0) {
1435 err = ret;
1436 goto out;
1437 }
1438
1439 if (ret != 0) {
1440 if (path->slots[0] == 0)
1441 goto not_found;
1442 path->slots[0]--;
1443 }
1444
1445 item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
1446 struct btrfs_file_extent_item);
1447 leaf = btrfs_buffer_leaf(path->nodes[0]);
1448 blocknr = btrfs_file_extent_disk_blocknr(item);
1449 blocknr += btrfs_file_extent_offset(item);
1450
1451 /* are we inside the extent that was found? */
1452 found_key = &leaf->items[path->slots[0]].key;
1453 found_type = btrfs_disk_key_type(found_key);
1454 if (btrfs_disk_key_objectid(found_key) != objectid ||
1455 found_type != BTRFS_EXTENT_DATA_KEY) {
1456 goto not_found;
1457 }
1458
1459 found_type = btrfs_file_extent_type(item);
1460 extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
1461 if (found_type == BTRFS_FILE_EXTENT_REG) {
1462 extent_end = extent_start +
1463 (btrfs_file_extent_num_blocks(item) << inode->i_blkbits);
1464 err = 0;
b888db2b 1465 if (start < extent_start || start >= extent_end) {
a52d9a80
CM		 1466 em->start = start;
1467 if (start < extent_start) {
b888db2b
CM		 1468 if (end < extent_start)
1469 goto not_found;
a52d9a80
CM		 1470 em->end = extent_end - 1;
1471 } else {
1472 em->end = end;
1473 }
1474 goto not_found_em;
1475 }
1476 if (btrfs_file_extent_disk_blocknr(item) == 0) {
1477 em->start = extent_start;
1478 em->end = extent_end - 1;
1479 em->block_start = 0;
1480 em->block_end = 0;
1481 goto insert;
1482 }
1483 em->block_start = blocknr << inode->i_blkbits;
1484 em->block_end = em->block_start +
1485 (btrfs_file_extent_num_blocks(item) <<
1486 inode->i_blkbits) - 1;
1487 em->start = extent_start;
1488 em->end = extent_end - 1;
1489 goto insert;
1490 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1491 char *ptr;
1492 char *map;
1493 u32 size;
1494
1495 size = btrfs_file_extent_inline_len(leaf->items +
1496 path->slots[0]);
1497 extent_end = extent_start + size;
b888db2b 1498 if (start < extent_start || start >= extent_end) {
a52d9a80
CM		 1499 em->start = start;
1500 if (start < extent_start) {
b888db2b
CM		 1501 if (end < extent_start)
1502 goto not_found;
a52d9a80
CM		 1503 em->end = extent_end - 1;
1504 } else {
1505 em->end = end;
1506 }
1507 goto not_found_em;
1508 }
1509 em->block_start = EXTENT_MAP_INLINE;
1510 em->block_end = EXTENT_MAP_INLINE;
1511 em->start = extent_start;
1512 em->end = extent_end - 1;
1513 if (!page) {
1514 goto insert;
1515 }
1516 ptr = btrfs_file_extent_inline_start(item);
1517 map = kmap(page);
1518 memcpy(map + page_offset, ptr, size);
1519 flush_dcache_page(result->b_page);
1520 kunmap(page);
1521 set_extent_uptodate(em_tree, extent_start,
1522 extent_end, GFP_NOFS);
1523 goto insert;
1524 } else {
1525 printk("unkknown found_type %d\n", found_type);
1526 WARN_ON(1);
1527 }
1528not_found:
1529 em->start = start;
1530 em->end = end;
1531not_found_em:
1532 em->block_start = 0;
1533 em->block_end = 0;
1534insert:
1535 btrfs_release_path(root, path);
1536 if (em->start > start || em->end < start) {
b888db2b 1537 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->end, start, end);
a52d9a80
CM		 1538 err = -EIO;
1539 goto out;
1540 }
1541 ret = add_extent_mapping(em_tree, em);
1542 if (ret == -EEXIST) {
1543 free_extent_map(em);
2bf5a725 1544 em = NULL;
a52d9a80
CM		 1545 failed_insert++;
1546 if (failed_insert > 5) {
1547 printk("failing to insert %Lu %Lu\n", start, end);
1548 err = -EIO;
1549 goto out;
1550 }
a52d9a80
CM		 1551 goto again;
1552 }
1553 err = 0;
1554out:
1555 btrfs_free_path(path);
1556 if (trans) {
1557 ret = btrfs_end_transaction(trans, root);
1558 if (!err)
1559 err = ret;
1560 }
1561 mutex_unlock(&root->fs_info->fs_mutex);
1562 if (err) {
1563 free_extent_map(em);
1564 WARN_ON(1);
1565 return ERR_PTR(err);
1566 }
1567 return em;
1568}
1569
d396c6f5 1570static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
39279cc3 1571{
d396c6f5 1572 return extent_bmap(mapping, iblock, btrfs_get_extent);
39279cc3
CM		 1573}
1574
1575static int btrfs_prepare_write(struct file *file, struct page *page,
1576 unsigned from, unsigned to)
1577{
a52d9a80
CM		 1578 return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree,
1579 page->mapping->host, page, from, to,
1580 btrfs_get_extent);
39279cc3
CM		 1581}
1582
a52d9a80 1583int btrfs_readpage(struct file *file, struct page *page)
9ebefb18 1584{
a52d9a80
CM		 1585 struct extent_map_tree *tree;
1586 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1587 return extent_read_full_page(tree, page, btrfs_get_extent);
9ebefb18 1588}
a52d9a80 1589static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
39279cc3 1590{
a52d9a80 1591 struct extent_map_tree *tree;
b888db2b
CM		 1592
1593
1594 if (current->flags & PF_MEMALLOC) {
1595 redirty_page_for_writepage(wbc, page);
1596 unlock_page(page);
1597 return 0;
1598 }
a52d9a80
CM		 1599 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1600 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
9ebefb18
CM		 1601}
1602
a52d9a80 1603static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
9ebefb18 1604{
a52d9a80
CM		 1605 struct extent_map_tree *tree;
1606 int ret;
8c2383c3 1607
a52d9a80
CM		 1608 if (page->private != 1) {
1609 WARN_ON(1);
1610 return try_to_free_buffers(page);
39279cc3 1611 }
a52d9a80
CM		 1612 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1613 ret = try_release_extent_mapping(tree, page);
1614 if (ret == 1) {
1615 ClearPagePrivate(page);
1616 set_page_private(page, 0);
1617 page_cache_release(page);
39279cc3 1618 }
a52d9a80 1619 return ret;
39279cc3
CM		 1620}
1621
a52d9a80 1622static void btrfs_invalidatepage(struct page *page, unsigned long offset)
39279cc3 1623{
a52d9a80 1624 struct extent_map_tree *tree;
39279cc3 1625
a52d9a80
CM		 1626 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1627 extent_invalidatepage(tree, page, offset);
1628 btrfs_releasepage(page, GFP_NOFS);
39279cc3
CM		 1629}
1630
9ebefb18
CM		 1631/*
1632 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
1633 * called from a page fault handler when a page is first dirtied. Hence we must
1634 * be careful to check for EOF conditions here. We set the page up correctly
1635 * for a written page which means we get ENOSPC checking when writing into
1636 * holes and correct delalloc and unwritten extent mapping on filesystems that
1637 * support these features.
1638 *
1639 * We are not allowed to take the i_mutex here so we have to play games to
1640 * protect against truncate races as the page could now be beyond EOF. Because
1641 * vmtruncate() writes the inode size before removing pages, once we have the
1642 * page lock we can determine safely if the page is beyond EOF. If it is not
1643 * beyond EOF, then the page is guaranteed safe against truncation until we
1644 * unlock the page.
1645 */
1646int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1647{
1648 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
1649 unsigned long end;
1650 loff_t size;
1651 int ret = -EINVAL;
a52d9a80 1652 u64 page_start;
9ebefb18 1653
011410bd 1654 down_read(&BTRFS_I(inode)->root->snap_sem);
9ebefb18
CM		 1655 lock_page(page);
1656 wait_on_page_writeback(page);
9ebefb18 1657 size = i_size_read(inode);
a52d9a80
CM		 1658 page_start = page->index << PAGE_CACHE_SHIFT;
1659
9ebefb18 1660 if ((page->mapping != inode->i_mapping) ||
a52d9a80 1661 (page_start > size)) {
9ebefb18
CM		 1662 /* page got truncated out from underneath us */
1663 goto out_unlock;
1664 }
1665
1666 /* page is wholly or partially inside EOF */
a52d9a80 1667 if (page_start + PAGE_CACHE_SIZE > size)
9ebefb18
CM		 1668 end = size & ~PAGE_CACHE_MASK;
1669 else
1670 end = PAGE_CACHE_SIZE;
1671
b888db2b 1672 ret = btrfs_cow_one_page(inode, page, end);
9ebefb18
CM		 1673
1674out_unlock:
011410bd 1675 up_read(&BTRFS_I(inode)->root->snap_sem);
9ebefb18
CM		 1676 unlock_page(page);
1677 return ret;
1678}
1679
39279cc3
CM		 1680static void btrfs_truncate(struct inode *inode)
1681{
1682 struct btrfs_root *root = BTRFS_I(inode)->root;
1683 int ret;
1684 struct btrfs_trans_handle *trans;
1685
1686 if (!S_ISREG(inode->i_mode))
1687 return;
1688 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1689 return;
1690
1691 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1692
1693 mutex_lock(&root->fs_info->fs_mutex);
1694 trans = btrfs_start_transaction(root, 1);
1695 btrfs_set_trans_block_group(trans, inode);
1696
1697 /* FIXME, add redo link to tree so we don't leak on crash */
1698 ret = btrfs_truncate_in_trans(trans, root, inode);
39279cc3
CM		 1699 btrfs_update_inode(trans, root, inode);
1700 ret = btrfs_end_transaction(trans, root);
1701 BUG_ON(ret);
1702 mutex_unlock(&root->fs_info->fs_mutex);
1703 btrfs_btree_balance_dirty(root);
1704}
1705
1706int btrfs_commit_write(struct file *file, struct page *page,
1707 unsigned from, unsigned to)
1708{
a52d9a80
CM		 1709 return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree,
1710 page->mapping->host, page, from, to);
39279cc3
CM		 1711}
1712
1713static int create_subvol(struct btrfs_root *root, char *name, int namelen)
1714{
1715 struct btrfs_trans_handle *trans;
1716 struct btrfs_key key;
1717 struct btrfs_root_item root_item;
1718 struct btrfs_inode_item *inode_item;
1719 struct buffer_head *subvol;
1720 struct btrfs_leaf *leaf;
1721 struct btrfs_root *new_root;
1722 struct inode *inode;
1723 struct inode *dir;
1724 int ret;
54aa1f4d 1725 int err;
39279cc3
CM		 1726 u64 objectid;
1727 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
1728
1729 mutex_lock(&root->fs_info->fs_mutex);
1730 trans = btrfs_start_transaction(root, 1);
1731 BUG_ON(!trans);
1732
6702ed49 1733 subvol = btrfs_alloc_free_block(trans, root, 0, 0);
54aa1f4d
CM		 1734 if (IS_ERR(subvol))
1735 return PTR_ERR(subvol);
39279cc3
CM		 1736 leaf = btrfs_buffer_leaf(subvol);
1737 btrfs_set_header_nritems(&leaf->header, 0);
1738 btrfs_set_header_level(&leaf->header, 0);
1739 btrfs_set_header_blocknr(&leaf->header, bh_blocknr(subvol));
1740 btrfs_set_header_generation(&leaf->header, trans->transid);
1741 btrfs_set_header_owner(&leaf->header, root->root_key.objectid);
1742 memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid,
1743 sizeof(leaf->header.fsid));
ccd467d6 1744 btrfs_mark_buffer_dirty(subvol);
39279cc3
CM		 1745
1746 inode_item = &root_item.inode;
1747 memset(inode_item, 0, sizeof(*inode_item));
1748 btrfs_set_inode_generation(inode_item, 1);
1749 btrfs_set_inode_size(inode_item, 3);
1750 btrfs_set_inode_nlink(inode_item, 1);
1751 btrfs_set_inode_nblocks(inode_item, 1);
1752 btrfs_set_inode_mode(inode_item, S_IFDIR | 0755);
1753
1754 btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol));
1755 btrfs_set_root_refs(&root_item, 1);
58176a96 1756 btrfs_set_root_blocks_used(&root_item, 0);
5eda7b5e
CM		 1757 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
1758 root_item.drop_level = 0;
39279cc3
CM		 1759 brelse(subvol);
1760 subvol = NULL;
1761
1762 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1763 0, &objectid);
54aa1f4d
CM		 1764 if (ret)
1765 goto fail;
39279cc3
CM		 1766
1767 btrfs_set_root_dirid(&root_item, new_dirid);
1768
1769 key.objectid = objectid;
1770 key.offset = 1;
1771 key.flags = 0;
1772 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1773 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1774 &root_item);
54aa1f4d
CM		 1775 if (ret)
1776 goto fail;
39279cc3
CM		 1777
1778 /*
1779 * insert the directory item
1780 */
1781 key.offset = (u64)-1;
1782 dir = root->fs_info->sb->s_root->d_inode;
1783 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1784 name, namelen, dir->i_ino, &key,
1785 BTRFS_FT_DIR);
54aa1f4d
CM		 1786 if (ret)
1787 goto fail;
39279cc3
CM		 1788
1789 ret = btrfs_commit_transaction(trans, root);
54aa1f4d
CM		 1790 if (ret)
1791 goto fail_commit;
39279cc3 1792
58176a96 1793 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
39279cc3
CM		 1794 BUG_ON(!new_root);
1795
1796 trans = btrfs_start_transaction(new_root, 1);
1797 BUG_ON(!trans);
1798
1799 inode = btrfs_new_inode(trans, new_root, new_dirid,
1800 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
54aa1f4d
CM		 1801 if (IS_ERR(inode))
1802 goto fail;
39279cc3
CM		 1803 inode->i_op = &btrfs_dir_inode_operations;
1804 inode->i_fop = &btrfs_dir_file_operations;
34088780 1805 new_root->inode = inode;
39279cc3
CM		 1806
1807 ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid);
54aa1f4d
CM		 1808 if (ret)
1809 goto fail;
39279cc3
CM		 1810
1811 inode->i_nlink = 1;
1812 inode->i_size = 6;
1813 ret = btrfs_update_inode(trans, new_root, inode);
54aa1f4d
CM		 1814 if (ret)
1815 goto fail;
1816fail:
1817 err = btrfs_commit_transaction(trans, root);
1818 if (err && !ret)
1819 ret = err;
1820fail_commit:
39279cc3
CM		 1821 mutex_unlock(&root->fs_info->fs_mutex);
1822 btrfs_btree_balance_dirty(root);
54aa1f4d 1823 return ret;
39279cc3
CM		 1824}
1825
1826static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
1827{
1828 struct btrfs_trans_handle *trans;
1829 struct btrfs_key key;
1830 struct btrfs_root_item new_root_item;
83df7c1d 1831 struct buffer_head *tmp;
39279cc3 1832 int ret;
54aa1f4d 1833 int err;
39279cc3
CM		 1834 u64 objectid;
1835
1836 if (!root->ref_cows)
1837 return -EINVAL;
1838
011410bd
CM		 1839 down_write(&root->snap_sem);
1840 freeze_bdev(root->fs_info->sb->s_bdev);
1841 thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb);
1842
39279cc3
CM		 1843 mutex_lock(&root->fs_info->fs_mutex);
1844 trans = btrfs_start_transaction(root, 1);
1845 BUG_ON(!trans);
1846
1847 ret = btrfs_update_inode(trans, root, root->inode);
54aa1f4d
CM		 1848 if (ret)
1849 goto fail;
39279cc3
CM		 1850
1851 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1852 0, &objectid);
54aa1f4d
CM		 1853 if (ret)
1854 goto fail;
39279cc3
CM		 1855
1856 memcpy(&new_root_item, &root->root_item,
1857 sizeof(new_root_item));
1858
1859 key.objectid = objectid;
1860 key.offset = 1;
1861 key.flags = 0;
1862 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
83df7c1d 1863 btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
39279cc3
CM		 1864 btrfs_set_root_blocknr(&new_root_item, bh_blocknr(root->node));
1865
1866 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1867 &new_root_item);
54aa1f4d
CM		 1868 if (ret)
1869 goto fail;
39279cc3
CM		 1870
1871 /*
1872 * insert the directory item
1873 */
1874 key.offset = (u64)-1;
1875 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1876 name, namelen,
1877 root->fs_info->sb->s_root->d_inode->i_ino,
1878 &key, BTRFS_FT_DIR);
1879
54aa1f4d
CM		 1880 if (ret)
1881 goto fail;
39279cc3
CM		 1882
1883 ret = btrfs_inc_root_ref(trans, root);
54aa1f4d
CM		 1884 if (ret)
1885 goto fail;
54aa1f4d
CM		 1886fail:
1887 err = btrfs_commit_transaction(trans, root);
1888 if (err && !ret)
1889 ret = err;
39279cc3 1890 mutex_unlock(&root->fs_info->fs_mutex);
011410bd 1891 up_write(&root->snap_sem);
39279cc3 1892 btrfs_btree_balance_dirty(root);
54aa1f4d 1893 return ret;
39279cc3
CM		 1894}
1895
86479a04
CM		 1896static unsigned long force_ra(struct address_space *mapping,
1897 struct file_ra_state *ra, struct file *file,
1898 pgoff_t offset, pgoff_t last_index)
1899{
1900 pgoff_t req_size;
1901
1902#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
1903 req_size = last_index - offset + 1;
1904 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
1905 return offset;
1906#else
1907 req_size = min(last_index - offset + 1, (pgoff_t)128);
1908 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
1909 return offset + req_size;
1910#endif
1911}
1912
1913int btrfs_defrag_file(struct file *file) {
1914 struct inode *inode = file->f_path.dentry->d_inode;
1915 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1916 struct page *page;
1917 unsigned long last_index;
1918 unsigned long ra_index = 0;
1919 u64 page_start;
1920 u64 page_end;
1921 unsigned long i;
1922
1923 mutex_lock(&inode->i_mutex);
1924 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
1925 for (i = 0; i <= last_index; i++) {
1926 if (i == ra_index) {
1927 ra_index = force_ra(inode->i_mapping, &file->f_ra,
1928 file, ra_index, last_index);
1929 }
1930 page = grab_cache_page(inode->i_mapping, i);
1931 if (!page)
1932 goto out_unlock;
1933 if (!PageUptodate(page)) {
1934 btrfs_readpage(NULL, page);
1935 lock_page(page);
1936 if (!PageUptodate(page)) {
1937 unlock_page(page);
1938 page_cache_release(page);
1939 goto out_unlock;
1940 }
1941 }
1942 page_start = page->index << PAGE_CACHE_SHIFT;
1943 page_end = page_start + PAGE_CACHE_SIZE - 1;
1944
1945 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
1946 set_extent_delalloc(em_tree, page_start,
1947 page_end, GFP_NOFS);
1948 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
1949 set_page_dirty(page);
1950 unlock_page(page);
1951 page_cache_release(page);
1952 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
1953 }
1954
1955out_unlock:
1956 mutex_unlock(&inode->i_mutex);
1957 return 0;
1958}
1959
39279cc3
CM		 1960int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int
1961 cmd, unsigned long arg)
1962{
1963 struct btrfs_root *root = BTRFS_I(inode)->root;
1964 struct btrfs_ioctl_vol_args vol_args;
1965 int ret = 0;
1966 struct btrfs_dir_item *di;
1967 int namelen;
1968 struct btrfs_path *path;
1969 u64 root_dirid;
1970
1971 switch (cmd) {
1972 case BTRFS_IOC_SNAP_CREATE:
1973 if (copy_from_user(&vol_args,
1974 (struct btrfs_ioctl_vol_args __user *)arg,
1975 sizeof(vol_args)))
1976 return -EFAULT;
1977 namelen = strlen(vol_args.name);
1978 if (namelen > BTRFS_VOL_NAME_MAX)
1979 return -EINVAL;
8a712645
CM		 1980 if (strchr(vol_args.name, '/'))
1981 return -EINVAL;
39279cc3
CM		 1982 path = btrfs_alloc_path();
1983 if (!path)
1984 return -ENOMEM;
1985 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
1986 mutex_lock(&root->fs_info->fs_mutex);
1987 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
1988 path, root_dirid,
1989 vol_args.name, namelen, 0);
1990 mutex_unlock(&root->fs_info->fs_mutex);
1991 btrfs_free_path(path);
1992 if (di && !IS_ERR(di))
1993 return -EEXIST;
54aa1f4d
CM		 1994 if (IS_ERR(di))
1995 return PTR_ERR(di);
39279cc3
CM		 1996
1997 if (root == root->fs_info->tree_root)
1998 ret = create_subvol(root, vol_args.name, namelen);
1999 else
2000 ret = create_snapshot(root, vol_args.name, namelen);
39279cc3 2001 break;
6702ed49
CM		 2002
2003 case BTRFS_IOC_DEFRAG:
86479a04
CM		 2004 if (S_ISDIR(inode->i_mode)) {
2005 mutex_lock(&root->fs_info->fs_mutex);
2006 btrfs_defrag_root(root, 0);
2007 btrfs_defrag_root(root->fs_info->extent_root, 0);
2008 mutex_unlock(&root->fs_info->fs_mutex);
2009 } else if (S_ISREG(inode->i_mode)) {
2010 btrfs_defrag_file(filp);
2011 }
6702ed49
CM		 2012 ret = 0;
2013 break;
39279cc3
CM		 2014 default:
2015 return -ENOTTY;
2016 }
2017 return ret;
2018}
2019
2020#ifdef CONFIG_COMPAT
2021long btrfs_compat_ioctl(struct file *file, unsigned int cmd,
2022 unsigned long arg)
2023{
2024 struct inode *inode = file->f_path.dentry->d_inode;
2025 int ret;
2026 lock_kernel();
2027 ret = btrfs_ioctl(inode, file, cmd, (unsigned long) compat_ptr(arg));
2028 unlock_kernel();
2029 return ret;
2030
2031}
2032#endif
2033
2034/*
2035 * Called inside transaction, so use GFP_NOFS
2036 */
2037struct inode *btrfs_alloc_inode(struct super_block *sb)
2038{
2039 struct btrfs_inode *ei;
2040
2041 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2042 if (!ei)
2043 return NULL;
15ee9bc7 2044 ei->last_trans = 0;
39279cc3
CM		 2045 return &ei->vfs_inode;
2046}
2047
2048void btrfs_destroy_inode(struct inode *inode)
2049{
2050 WARN_ON(!list_empty(&inode->i_dentry));
2051 WARN_ON(inode->i_data.nrpages);
2052
2053 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2054}
2055
2056static void init_once(void * foo, struct kmem_cache * cachep,
2057 unsigned long flags)
2058{
2059 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2060
2061 inode_init_once(&ei->vfs_inode);
2062}
2063
2064void btrfs_destroy_cachep(void)
2065{
2066 if (btrfs_inode_cachep)
2067 kmem_cache_destroy(btrfs_inode_cachep);
2068 if (btrfs_trans_handle_cachep)
2069 kmem_cache_destroy(btrfs_trans_handle_cachep);
2070 if (btrfs_transaction_cachep)
2071 kmem_cache_destroy(btrfs_transaction_cachep);
2072 if (btrfs_bit_radix_cachep)
2073 kmem_cache_destroy(btrfs_bit_radix_cachep);
2074 if (btrfs_path_cachep)
2075 kmem_cache_destroy(btrfs_path_cachep);
2076}
2077
86479a04 2078struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
92fee66d
CM		 2079 unsigned long extra_flags,
2080 void (*ctor)(void *, struct kmem_cache *,
2081 unsigned long))
2082{
2083 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2084 SLAB_MEM_SPREAD | extra_flags), ctor
2085#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2086 ,NULL
2087#endif
2088 );
2089}
2090
39279cc3
CM		 2091int btrfs_init_cachep(void)
2092{
86479a04 2093 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
92fee66d
CM		 2094 sizeof(struct btrfs_inode),
2095 0, init_once);
39279cc3
CM		 2096 if (!btrfs_inode_cachep)
2097 goto fail;
86479a04
CM		 2098 btrfs_trans_handle_cachep =
2099 btrfs_cache_create("btrfs_trans_handle_cache",
2100 sizeof(struct btrfs_trans_handle),
2101 0, NULL);
39279cc3
CM		 2102 if (!btrfs_trans_handle_cachep)
2103 goto fail;
86479a04 2104 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
39279cc3 2105 sizeof(struct btrfs_transaction),
92fee66d 2106 0, NULL);
39279cc3
CM		 2107 if (!btrfs_transaction_cachep)
2108 goto fail;
86479a04 2109 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
92fee66d
CM		 2110 sizeof(struct btrfs_transaction),
2111 0, NULL);
39279cc3
CM		 2112 if (!btrfs_path_cachep)
2113 goto fail;
86479a04 2114 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
92fee66d 2115 SLAB_DESTROY_BY_RCU, NULL);
39279cc3
CM		 2116 if (!btrfs_bit_radix_cachep)
2117 goto fail;
2118 return 0;
2119fail:
2120 btrfs_destroy_cachep();
2121 return -ENOMEM;
2122}
2123
2124static int btrfs_getattr(struct vfsmount *mnt,
2125 struct dentry *dentry, struct kstat *stat)
2126{
2127 struct inode *inode = dentry->d_inode;
2128 generic_fillattr(inode, stat);
2129 stat->blksize = 256 * 1024;
2130 return 0;
2131}
2132
2133static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2134 struct inode * new_dir,struct dentry *new_dentry)
2135{
2136 struct btrfs_trans_handle *trans;
2137 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2138 struct inode *new_inode = new_dentry->d_inode;
2139 struct inode *old_inode = old_dentry->d_inode;
2140 struct timespec ctime = CURRENT_TIME;
2141 struct btrfs_path *path;
2142 struct btrfs_dir_item *di;
2143 int ret;
2144
2145 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2146 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2147 return -ENOTEMPTY;
2148 }
2149 mutex_lock(&root->fs_info->fs_mutex);
2150 trans = btrfs_start_transaction(root, 1);
2151 btrfs_set_trans_block_group(trans, new_dir);
2152 path = btrfs_alloc_path();
2153 if (!path) {
2154 ret = -ENOMEM;
2155 goto out_fail;
2156 }
2157
2158 old_dentry->d_inode->i_nlink++;
2159 old_dir->i_ctime = old_dir->i_mtime = ctime;
2160 new_dir->i_ctime = new_dir->i_mtime = ctime;
2161 old_inode->i_ctime = ctime;
2162 if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) {
2163 struct btrfs_key *location = &BTRFS_I(new_dir)->location;
2164 u64 old_parent_oid;
2165 di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino,
2166 "..", 2, -1);
2167 if (IS_ERR(di)) {
2168 ret = PTR_ERR(di);
2169 goto out_fail;
2170 }
2171 if (!di) {
2172 ret = -ENOENT;
2173 goto out_fail;
2174 }
2175 old_parent_oid = btrfs_disk_key_objectid(&di->location);
2176 ret = btrfs_del_item(trans, root, path);
2177 if (ret) {
39279cc3
CM		 2178 goto out_fail;
2179 }
2180 btrfs_release_path(root, path);
2181
2182 di = btrfs_lookup_dir_index_item(trans, root, path,
2183 old_inode->i_ino,
2184 old_parent_oid,
2185 "..", 2, -1);
2186 if (IS_ERR(di)) {
2187 ret = PTR_ERR(di);
2188 goto out_fail;
2189 }
2190 if (!di) {
2191 ret = -ENOENT;
2192 goto out_fail;
2193 }
2194 ret = btrfs_del_item(trans, root, path);
2195 if (ret) {
39279cc3
CM		 2196 goto out_fail;
2197 }
2198 btrfs_release_path(root, path);
2199
2200 ret = btrfs_insert_dir_item(trans, root, "..", 2,
2201 old_inode->i_ino, location,
2202 BTRFS_FT_DIR);
2203 if (ret)
2204 goto out_fail;
2205 }
2206
2207
2208 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2209 if (ret)
2210 goto out_fail;
2211
2212 if (new_inode) {
2213 new_inode->i_ctime = CURRENT_TIME;
2214 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2215 if (ret)
2216 goto out_fail;
39279cc3
CM		 2217 }
2218 ret = btrfs_add_link(trans, new_dentry, old_inode);
2219 if (ret)
2220 goto out_fail;
2221
2222out_fail:
2223 btrfs_free_path(path);
2224 btrfs_end_transaction(trans, root);
2225 mutex_unlock(&root->fs_info->fs_mutex);
2226 return ret;
2227}
2228
2229static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2230 const char *symname)
2231{
2232 struct btrfs_trans_handle *trans;
2233 struct btrfs_root *root = BTRFS_I(dir)->root;
2234 struct btrfs_path *path;
2235 struct btrfs_key key;
2236 struct inode *inode;
2237 int err;
2238 int drop_inode = 0;
2239 u64 objectid;
2240 int name_len;
2241 int datasize;
2242 char *ptr;
2243 struct btrfs_file_extent_item *ei;
2244
2245 name_len = strlen(symname) + 1;
2246 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2247 return -ENAMETOOLONG;
2248 mutex_lock(&root->fs_info->fs_mutex);
2249 trans = btrfs_start_transaction(root, 1);
2250 btrfs_set_trans_block_group(trans, dir);
2251
2252 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2253 if (err) {
2254 err = -ENOSPC;
2255 goto out_unlock;
2256 }
2257
2258 inode = btrfs_new_inode(trans, root, objectid,
2259 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2260 err = PTR_ERR(inode);
2261 if (IS_ERR(inode))
2262 goto out_unlock;
2263
2264 btrfs_set_trans_block_group(trans, inode);
2265 err = btrfs_add_nondir(trans, dentry, inode);
2266 if (err)
2267 drop_inode = 1;
2268 else {
2269 inode->i_mapping->a_ops = &btrfs_aops;
2270 inode->i_fop = &btrfs_file_operations;
2271 inode->i_op = &btrfs_file_inode_operations;
a52d9a80
CM		 2272 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
2273 inode->i_mapping, GFP_NOFS);
07157aac 2274 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
39279cc3
CM		 2275 }
2276 dir->i_sb->s_dirt = 1;
2277 btrfs_update_inode_block_group(trans, inode);
2278 btrfs_update_inode_block_group(trans, dir);
2279 if (drop_inode)
2280 goto out_unlock;
2281
2282 path = btrfs_alloc_path();
2283 BUG_ON(!path);
2284 key.objectid = inode->i_ino;
2285 key.offset = 0;
2286 key.flags = 0;
2287 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2288 datasize = btrfs_file_extent_calc_inline_size(name_len);
2289 err = btrfs_insert_empty_item(trans, root, path, &key,
2290 datasize);
54aa1f4d
CM		 2291 if (err) {
2292 drop_inode = 1;
2293 goto out_unlock;
2294 }
39279cc3
CM		 2295 ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
2296 path->slots[0], struct btrfs_file_extent_item);
2297 btrfs_set_file_extent_generation(ei, trans->transid);
2298 btrfs_set_file_extent_type(ei,
2299 BTRFS_FILE_EXTENT_INLINE);
2300 ptr = btrfs_file_extent_inline_start(ei);
2301 btrfs_memcpy(root, path->nodes[0]->b_data,
2302 ptr, symname, name_len);
ccd467d6 2303 btrfs_mark_buffer_dirty(path->nodes[0]);
39279cc3
CM		 2304 btrfs_free_path(path);
2305 inode->i_op = &btrfs_symlink_inode_operations;
2306 inode->i_mapping->a_ops = &btrfs_symlink_aops;
2307 inode->i_size = name_len - 1;
54aa1f4d
CM		 2308 err = btrfs_update_inode(trans, root, inode);
2309 if (err)
2310 drop_inode = 1;
39279cc3
CM		 2311
2312out_unlock:
2313 btrfs_end_transaction(trans, root);
2314 mutex_unlock(&root->fs_info->fs_mutex);
39279cc3
CM		 2315 if (drop_inode) {
2316 inode_dec_link_count(inode);
2317 iput(inode);
2318 }
2319 btrfs_btree_balance_dirty(root);
2320 return err;
2321}
2322
2323static struct inode_operations btrfs_dir_inode_operations = {
2324 .lookup = btrfs_lookup,
2325 .create = btrfs_create,
2326 .unlink = btrfs_unlink,
2327 .link = btrfs_link,
2328 .mkdir = btrfs_mkdir,
2329 .rmdir = btrfs_rmdir,
2330 .rename = btrfs_rename,
2331 .symlink = btrfs_symlink,
2332 .setattr = btrfs_setattr,
618e21d5 2333 .mknod = btrfs_mknod,
39279cc3
CM		 2334};
2335
2336static struct inode_operations btrfs_dir_ro_inode_operations = {
2337 .lookup = btrfs_lookup,
2338};
2339
2340static struct file_operations btrfs_dir_file_operations = {
2341 .llseek = generic_file_llseek,
2342 .read = generic_read_dir,
2343 .readdir = btrfs_readdir,
2344 .ioctl = btrfs_ioctl,
2345#ifdef CONFIG_COMPAT
2346 .compat_ioctl = btrfs_compat_ioctl,
2347#endif
2348};
2349
07157aac
CM		 2350static struct extent_map_ops btrfs_extent_map_ops = {
2351 .fill_delalloc = run_delalloc_range,
2352 .writepage_io_hook = btrfs_writepage_io_hook,
2353 .readpage_io_hook = btrfs_readpage_io_hook,
2354 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
2355};
2356
39279cc3
CM		 2357static struct address_space_operations btrfs_aops = {
2358 .readpage = btrfs_readpage,
2359 .writepage = btrfs_writepage,
2360 .sync_page = block_sync_page,
2361 .prepare_write = btrfs_prepare_write,
2362 .commit_write = btrfs_commit_write,
2363 .bmap = btrfs_bmap,
a52d9a80
CM		 2364 .invalidatepage = btrfs_invalidatepage,
2365 .releasepage = btrfs_releasepage,
2366 .set_page_dirty = __set_page_dirty_nobuffers,
39279cc3
CM		 2367};
2368
2369static struct address_space_operations btrfs_symlink_aops = {
2370 .readpage = btrfs_readpage,
2371 .writepage = btrfs_writepage,
2bf5a725
CM		 2372 .invalidatepage = btrfs_invalidatepage,
2373 .releasepage = btrfs_releasepage,
39279cc3
CM		 2374};
2375
2376static struct inode_operations btrfs_file_inode_operations = {
2377 .truncate = btrfs_truncate,
2378 .getattr = btrfs_getattr,
2379 .setattr = btrfs_setattr,
2380};
2381
618e21d5
JB		 2382static struct inode_operations btrfs_special_inode_operations = {
2383 .getattr = btrfs_getattr,
2384 .setattr = btrfs_setattr,
2385};
2386
39279cc3
CM		 2387static struct inode_operations btrfs_symlink_inode_operations = {
2388 .readlink = generic_readlink,
2389 .follow_link = page_follow_link_light,
2390 .put_link = page_put_link,
2391};
ubuntu-bionic-kernel mirror