]> 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: Retry metadata reads in the face of checksum failures
[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
065631f6 19#include <linux/bio.h>
39279cc3
CM		 20#include <linux/buffer_head.h>
21#include <linux/fs.h>
22#include <linux/pagemap.h>
23#include <linux/highmem.h>
24#include <linux/time.h>
25#include <linux/init.h>
26#include <linux/string.h>
27#include <linux/smp_lock.h>
28#include <linux/backing-dev.h>
29#include <linux/mpage.h>
30#include <linux/swap.h>
31#include <linux/writeback.h>
32#include <linux/statfs.h>
33#include <linux/compat.h>
9ebefb18 34#include <linux/bit_spinlock.h>
92fee66d 35#include <linux/version.h>
5103e947 36#include <linux/xattr.h>
39279cc3
CM		 37#include "ctree.h"
38#include "disk-io.h"
39#include "transaction.h"
40#include "btrfs_inode.h"
41#include "ioctl.h"
42#include "print-tree.h"
0b86a832 43#include "volumes.h"
39279cc3
CM		 44
45struct btrfs_iget_args {
46 u64 ino;
47 struct btrfs_root *root;
48};
49
50static struct inode_operations btrfs_dir_inode_operations;
51static struct inode_operations btrfs_symlink_inode_operations;
52static struct inode_operations btrfs_dir_ro_inode_operations;
618e21d5 53static struct inode_operations btrfs_special_inode_operations;
39279cc3
CM		 54static struct inode_operations btrfs_file_inode_operations;
55static struct address_space_operations btrfs_aops;
56static struct address_space_operations btrfs_symlink_aops;
57static struct file_operations btrfs_dir_file_operations;
d1310b2e 58static struct extent_io_ops btrfs_extent_io_ops;
39279cc3
CM		 59
60static struct kmem_cache *btrfs_inode_cachep;
61struct kmem_cache *btrfs_trans_handle_cachep;
62struct kmem_cache *btrfs_transaction_cachep;
63struct kmem_cache *btrfs_bit_radix_cachep;
64struct kmem_cache *btrfs_path_cachep;
65
66#define S_SHIFT 12
67static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
68 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
69 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
70 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
71 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
72 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
73 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
74 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
75};
76
1832a6d5
CM		 77int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
78 int for_del)
79{
80 u64 total = btrfs_super_total_bytes(&root->fs_info->super_copy);
81 u64 used = btrfs_super_bytes_used(&root->fs_info->super_copy);
82 u64 thresh;
83 int ret = 0;
84
85 if (for_del)
f9ef6604 86 thresh = total * 90;
1832a6d5 87 else
f9ef6604
CM		 88 thresh = total * 85;
89
90 do_div(thresh, 100);
1832a6d5
CM		 91
92 spin_lock(&root->fs_info->delalloc_lock);
93 if (used + root->fs_info->delalloc_bytes + num_required > thresh)
94 ret = -ENOSPC;
95 spin_unlock(&root->fs_info->delalloc_lock);
96 return ret;
97}
98
be20aa9d 99static int cow_file_range(struct inode *inode, u64 start, u64 end)
b888db2b
CM		 100{
101 struct btrfs_root *root = BTRFS_I(inode)->root;
102 struct btrfs_trans_handle *trans;
b888db2b 103 u64 alloc_hint = 0;
db94535d 104 u64 num_bytes;
c59f8951 105 u64 cur_alloc_size;
db94535d 106 u64 blocksize = root->sectorsize;
d1310b2e
CM		 107 u64 orig_start = start;
108 u64 orig_num_bytes;
be20aa9d
CM		 109 struct btrfs_key ins;
110 int ret;
b888db2b 111
b888db2b 112 trans = btrfs_start_transaction(root, 1);
b888db2b 113 BUG_ON(!trans);
be20aa9d
CM		 114 btrfs_set_trans_block_group(trans, inode);
115
db94535d 116 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
be20aa9d 117 num_bytes = max(blocksize, num_bytes);
b888db2b 118 ret = btrfs_drop_extents(trans, root, inode,
3326d1b0 119 start, start + num_bytes, start, &alloc_hint);
d1310b2e 120 orig_num_bytes = num_bytes;
db94535d 121
179e29e4
CM		 122 if (alloc_hint == EXTENT_MAP_INLINE)
123 goto out;
124
c59f8951
CM		 125 while(num_bytes > 0) {
126 cur_alloc_size = min(num_bytes, root->fs_info->max_extent);
127 ret = btrfs_alloc_extent(trans, root, cur_alloc_size,
128 root->root_key.objectid,
129 trans->transid,
130 inode->i_ino, start, 0,
131 alloc_hint, (u64)-1, &ins, 1);
132 if (ret) {
133 WARN_ON(1);
134 goto out;
135 }
136 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
137 start, ins.objectid, ins.offset,
138 ins.offset);
9069218d 139 inode->i_blocks += ins.offset >> 9;
5f56406a 140 btrfs_check_file(root, inode);
c59f8951
CM		 141 num_bytes -= cur_alloc_size;
142 alloc_hint = ins.objectid + ins.offset;
143 start += cur_alloc_size;
b888db2b 144 }
d1310b2e
CM		 145 btrfs_drop_extent_cache(inode, orig_start,
146 orig_start + orig_num_bytes - 1);
dc17ff8f 147 btrfs_add_ordered_inode(inode);
9069218d 148 btrfs_update_inode(trans, root, inode);
b888db2b
CM		 149out:
150 btrfs_end_transaction(trans, root);
be20aa9d
CM		 151 return ret;
152}
153
154static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end)
155{
156 u64 extent_start;
157 u64 extent_end;
158 u64 bytenr;
159 u64 cow_end;
1832a6d5 160 u64 loops = 0;
c31f8830 161 u64 total_fs_bytes;
be20aa9d
CM		 162 struct btrfs_root *root = BTRFS_I(inode)->root;
163 struct extent_buffer *leaf;
164 int found_type;
165 struct btrfs_path *path;
166 struct btrfs_file_extent_item *item;
167 int ret;
168 int err;
169 struct btrfs_key found_key;
170
c31f8830 171 total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
be20aa9d
CM		 172 path = btrfs_alloc_path();
173 BUG_ON(!path);
174again:
175 ret = btrfs_lookup_file_extent(NULL, root, path,
176 inode->i_ino, start, 0);
177 if (ret < 0) {
178 btrfs_free_path(path);
179 return ret;
180 }
181
182 cow_end = end;
183 if (ret != 0) {
184 if (path->slots[0] == 0)
185 goto not_found;
186 path->slots[0]--;
187 }
188
189 leaf = path->nodes[0];
190 item = btrfs_item_ptr(leaf, path->slots[0],
191 struct btrfs_file_extent_item);
192
193 /* are we inside the extent that was found? */
194 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
195 found_type = btrfs_key_type(&found_key);
196 if (found_key.objectid != inode->i_ino ||
197 found_type != BTRFS_EXTENT_DATA_KEY) {
198 goto not_found;
199 }
200
201 found_type = btrfs_file_extent_type(leaf, item);
202 extent_start = found_key.offset;
203 if (found_type == BTRFS_FILE_EXTENT_REG) {
c31f8830
CM		 204 u64 extent_num_bytes;
205
206 extent_num_bytes = btrfs_file_extent_num_bytes(leaf, item);
207 extent_end = extent_start + extent_num_bytes;
be20aa9d
CM		 208 err = 0;
209
1832a6d5
CM		 210 if (loops && start != extent_start)
211 goto not_found;
212
be20aa9d
CM		 213 if (start < extent_start || start >= extent_end)
214 goto not_found;
215
216 cow_end = min(end, extent_end - 1);
217 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
218 if (bytenr == 0)
219 goto not_found;
220
c31f8830
CM		 221 /*
222 * we may be called by the resizer, make sure we're inside
223 * the limits of the FS
224 */
225 if (bytenr + extent_num_bytes > total_fs_bytes)
226 goto not_found;
227
be20aa9d
CM		 228 if (btrfs_count_snapshots_in_path(root, path, bytenr) != 1) {
229 goto not_found;
230 }
231
232 start = extent_end;
bd09835d 233 } else {
be20aa9d
CM		 234 goto not_found;
235 }
236loop:
237 if (start > end) {
238 btrfs_free_path(path);
239 return 0;
240 }
241 btrfs_release_path(root, path);
1832a6d5 242 loops++;
be20aa9d
CM		 243 goto again;
244
245not_found:
246 cow_file_range(inode, start, cow_end);
247 start = cow_end + 1;
248 goto loop;
249}
250
251static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
252{
253 struct btrfs_root *root = BTRFS_I(inode)->root;
254 int ret;
be20aa9d 255 mutex_lock(&root->fs_info->fs_mutex);
b98b6767
Y		 256 if (btrfs_test_opt(root, NODATACOW) ||
257 btrfs_test_flag(inode, NODATACOW))
be20aa9d
CM		 258 ret = run_delalloc_nocow(inode, start, end);
259 else
260 ret = cow_file_range(inode, start, end);
1832a6d5 261
b888db2b
CM		 262 mutex_unlock(&root->fs_info->fs_mutex);
263 return ret;
264}
265
291d673e 266int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
b0c68f8b 267 unsigned long old, unsigned long bits)
291d673e 268{
b0c68f8b 269 if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
291d673e
CM		 270 struct btrfs_root *root = BTRFS_I(inode)->root;
271 spin_lock(&root->fs_info->delalloc_lock);
9069218d 272 BTRFS_I(inode)->delalloc_bytes += end - start + 1;
291d673e
CM		 273 root->fs_info->delalloc_bytes += end - start + 1;
274 spin_unlock(&root->fs_info->delalloc_lock);
275 }
276 return 0;
277}
278
279int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end,
b0c68f8b 280 unsigned long old, unsigned long bits)
291d673e 281{
b0c68f8b 282 if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
291d673e
CM		 283 struct btrfs_root *root = BTRFS_I(inode)->root;
284 spin_lock(&root->fs_info->delalloc_lock);
b0c68f8b
CM		 285 if (end - start + 1 > root->fs_info->delalloc_bytes) {
286 printk("warning: delalloc account %Lu %Lu\n",
287 end - start + 1, root->fs_info->delalloc_bytes);
288 root->fs_info->delalloc_bytes = 0;
9069218d 289 BTRFS_I(inode)->delalloc_bytes = 0;
b0c68f8b
CM		 290 } else {
291 root->fs_info->delalloc_bytes -= end - start + 1;
9069218d 292 BTRFS_I(inode)->delalloc_bytes -= end - start + 1;
b0c68f8b 293 }
291d673e
CM		 294 spin_unlock(&root->fs_info->delalloc_lock);
295 }
296 return 0;
297}
298
239b14b3
CM		 299int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
300 size_t size, struct bio *bio)
301{
302 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
303 struct btrfs_mapping_tree *map_tree;
239b14b3 304 u64 logical = bio->bi_sector << 9;
239b14b3
CM		 305 u64 length = 0;
306 u64 map_length;
307 struct bio_vec *bvec;
308 int i;
309 int ret;
310
311 bio_for_each_segment(bvec, bio, i) {
312 length += bvec->bv_len;
313 }
314 map_tree = &root->fs_info->mapping_tree;
315 map_length = length;
cea9e445 316 ret = btrfs_map_block(map_tree, READ, logical,
f188591e 317 &map_length, NULL, 0);
cea9e445 318
239b14b3 319 if (map_length < length + size) {
239b14b3
CM		 320 return 1;
321 }
322 return 0;
323}
324
f188591e
CM		 325int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
326 int mirror_num)
065631f6 327{
065631f6
CM		 328 struct btrfs_root *root = BTRFS_I(inode)->root;
329 struct btrfs_trans_handle *trans;
330 int ret = 0;
331
22c59948
CM		 332 if (!(rw & (1 << BIO_RW))) {
333 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
334 BUG_ON(ret);
0b86a832
CM		 335 goto mapit;
336 }
065631f6
CM		 337
338 if (btrfs_test_opt(root, NODATASUM) ||
0b86a832
CM		 339 btrfs_test_flag(inode, NODATASUM)) {
340 goto mapit;
341 }
065631f6
CM		 342
343 mutex_lock(&root->fs_info->fs_mutex);
344 trans = btrfs_start_transaction(root, 1);
345 btrfs_set_trans_block_group(trans, inode);
346 btrfs_csum_file_blocks(trans, root, inode, bio);
347 ret = btrfs_end_transaction(trans, root);
348 BUG_ON(ret);
349 mutex_unlock(&root->fs_info->fs_mutex);
0b86a832 350mapit:
f188591e 351 return btrfs_map_bio(root, rw, bio, mirror_num);
065631f6 352}
6885f308 353
07157aac
CM		 354int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
355{
356 int ret = 0;
357 struct inode *inode = page->mapping->host;
358 struct btrfs_root *root = BTRFS_I(inode)->root;
d1310b2e 359 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
07157aac
CM		 360 struct btrfs_csum_item *item;
361 struct btrfs_path *path = NULL;
ff79f819 362 u32 csum;
b98b6767
Y		 363 if (btrfs_test_opt(root, NODATASUM) ||
364 btrfs_test_flag(inode, NODATASUM))
b6cda9bc 365 return 0;
07157aac
CM		 366 mutex_lock(&root->fs_info->fs_mutex);
367 path = btrfs_alloc_path();
368 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
369 if (IS_ERR(item)) {
370 ret = PTR_ERR(item);
371 /* a csum that isn't present is a preallocated region. */
372 if (ret == -ENOENT || ret == -EFBIG)
373 ret = 0;
ff79f819 374 csum = 0;
aadfeb6e 375 printk("no csum found for inode %lu start %Lu\n", inode->i_ino, start);
07157aac
CM		 376 goto out;
377 }
ff79f819
CM		 378 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
379 BTRFS_CRC32_SIZE);
d1310b2e 380 set_state_private(io_tree, start, csum);
07157aac
CM		 381out:
382 if (path)
383 btrfs_free_path(path);
384 mutex_unlock(&root->fs_info->fs_mutex);
385 return ret;
386}
387
70dec807
CM		 388int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
389 struct extent_state *state)
07157aac 390{
35ebb934 391 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
07157aac 392 struct inode *inode = page->mapping->host;
d1310b2e 393 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
07157aac 394 char *kaddr;
aadfeb6e 395 u64 private = ~(u32)0;
07157aac 396 int ret;
ff79f819
CM		 397 struct btrfs_root *root = BTRFS_I(inode)->root;
398 u32 csum = ~(u32)0;
bbf0d006 399 unsigned long flags;
d1310b2e 400
b98b6767
Y		 401 if (btrfs_test_opt(root, NODATASUM) ||
402 btrfs_test_flag(inode, NODATASUM))
b6cda9bc 403 return 0;
c2e639f0 404 if (state && state->start == start) {
70dec807
CM		 405 private = state->private;
406 ret = 0;
407 } else {
408 ret = get_state_private(io_tree, start, &private);
409 }
bbf0d006 410 local_irq_save(flags);
07157aac
CM		 411 kaddr = kmap_atomic(page, KM_IRQ0);
412 if (ret) {
413 goto zeroit;
414 }
ff79f819
CM		 415 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
416 btrfs_csum_final(csum, (char *)&csum);
417 if (csum != private) {
07157aac
CM		 418 goto zeroit;
419 }
420 kunmap_atomic(kaddr, KM_IRQ0);
bbf0d006 421 local_irq_restore(flags);
07157aac
CM		 422 return 0;
423
424zeroit:
aadfeb6e
CM		 425 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
426 page->mapping->host->i_ino, (unsigned long long)start, csum,
427 private);
db94535d
CM		 428 memset(kaddr + offset, 1, end - start + 1);
429 flush_dcache_page(page);
07157aac 430 kunmap_atomic(kaddr, KM_IRQ0);
bbf0d006 431 local_irq_restore(flags);
07157aac
CM		 432 return 0;
433}
b888db2b 434
39279cc3
CM		 435void btrfs_read_locked_inode(struct inode *inode)
436{
437 struct btrfs_path *path;
5f39d397 438 struct extent_buffer *leaf;
39279cc3 439 struct btrfs_inode_item *inode_item;
0b86a832 440 struct btrfs_timespec *tspec;
39279cc3
CM		 441 struct btrfs_root *root = BTRFS_I(inode)->root;
442 struct btrfs_key location;
443 u64 alloc_group_block;
618e21d5 444 u32 rdev;
39279cc3
CM		 445 int ret;
446
447 path = btrfs_alloc_path();
448 BUG_ON(!path);
39279cc3 449 mutex_lock(&root->fs_info->fs_mutex);
39279cc3 450 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
dc17ff8f 451
39279cc3 452 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
5f39d397 453 if (ret)
39279cc3 454 goto make_bad;
39279cc3 455
5f39d397
CM		 456 leaf = path->nodes[0];
457 inode_item = btrfs_item_ptr(leaf, path->slots[0],
458 struct btrfs_inode_item);
459
460 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
461 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
462 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
463 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
464 inode->i_size = btrfs_inode_size(leaf, inode_item);
465
466 tspec = btrfs_inode_atime(inode_item);
467 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
468 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
469
470 tspec = btrfs_inode_mtime(inode_item);
471 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
472 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
473
474 tspec = btrfs_inode_ctime(inode_item);
475 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
476 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
477
478 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
479 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
618e21d5 480 inode->i_rdev = 0;
5f39d397
CM		 481 rdev = btrfs_inode_rdev(leaf, inode_item);
482
483 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
39279cc3
CM		 484 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
485 alloc_group_block);
b98b6767 486 BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
e52ec0eb
CM		 487 if (!BTRFS_I(inode)->block_group) {
488 BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
0b86a832
CM		 489 NULL, 0,
490 BTRFS_BLOCK_GROUP_METADATA, 0);
e52ec0eb 491 }
39279cc3
CM		 492 btrfs_free_path(path);
493 inode_item = NULL;
494
495 mutex_unlock(&root->fs_info->fs_mutex);
496
497 switch (inode->i_mode & S_IFMT) {
39279cc3
CM		 498 case S_IFREG:
499 inode->i_mapping->a_ops = &btrfs_aops;
04160088 500 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
d1310b2e 501 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
39279cc3
CM		 502 inode->i_fop = &btrfs_file_operations;
503 inode->i_op = &btrfs_file_inode_operations;
504 break;
505 case S_IFDIR:
506 inode->i_fop = &btrfs_dir_file_operations;
507 if (root == root->fs_info->tree_root)
508 inode->i_op = &btrfs_dir_ro_inode_operations;
509 else
510 inode->i_op = &btrfs_dir_inode_operations;
511 break;
512 case S_IFLNK:
513 inode->i_op = &btrfs_symlink_inode_operations;
514 inode->i_mapping->a_ops = &btrfs_symlink_aops;
04160088 515 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
39279cc3 516 break;
618e21d5
JB		 517 default:
518 init_special_inode(inode, inode->i_mode, rdev);
519 break;
39279cc3
CM		 520 }
521 return;
522
523make_bad:
524 btrfs_release_path(root, path);
525 btrfs_free_path(path);
526 mutex_unlock(&root->fs_info->fs_mutex);
527 make_bad_inode(inode);
528}
529
5f39d397
CM		 530static void fill_inode_item(struct extent_buffer *leaf,
531 struct btrfs_inode_item *item,
39279cc3
CM		 532 struct inode *inode)
533{
5f39d397
CM		 534 btrfs_set_inode_uid(leaf, item, inode->i_uid);
535 btrfs_set_inode_gid(leaf, item, inode->i_gid);
536 btrfs_set_inode_size(leaf, item, inode->i_size);
537 btrfs_set_inode_mode(leaf, item, inode->i_mode);
538 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
539
540 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
541 inode->i_atime.tv_sec);
542 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
543 inode->i_atime.tv_nsec);
544
545 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
546 inode->i_mtime.tv_sec);
547 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
548 inode->i_mtime.tv_nsec);
549
550 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
551 inode->i_ctime.tv_sec);
552 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
553 inode->i_ctime.tv_nsec);
554
555 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
556 btrfs_set_inode_generation(leaf, item, inode->i_generation);
557 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
b98b6767 558 btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
5f39d397 559 btrfs_set_inode_block_group(leaf, item,
39279cc3
CM		 560 BTRFS_I(inode)->block_group->key.objectid);
561}
562
a52d9a80 563int btrfs_update_inode(struct btrfs_trans_handle *trans,
39279cc3
CM		 564 struct btrfs_root *root,
565 struct inode *inode)
566{
567 struct btrfs_inode_item *inode_item;
568 struct btrfs_path *path;
5f39d397 569 struct extent_buffer *leaf;
39279cc3
CM		 570 int ret;
571
572 path = btrfs_alloc_path();
573 BUG_ON(!path);
39279cc3
CM		 574 ret = btrfs_lookup_inode(trans, root, path,
575 &BTRFS_I(inode)->location, 1);
576 if (ret) {
577 if (ret > 0)
578 ret = -ENOENT;
579 goto failed;
580 }
581
5f39d397
CM		 582 leaf = path->nodes[0];
583 inode_item = btrfs_item_ptr(leaf, path->slots[0],
39279cc3
CM		 584 struct btrfs_inode_item);
585
5f39d397
CM		 586 fill_inode_item(leaf, inode_item, inode);
587 btrfs_mark_buffer_dirty(leaf);
15ee9bc7 588 btrfs_set_inode_last_trans(trans, inode);
39279cc3
CM		 589 ret = 0;
590failed:
591 btrfs_release_path(root, path);
592 btrfs_free_path(path);
593 return ret;
594}
595
596
597static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
598 struct btrfs_root *root,
599 struct inode *dir,
600 struct dentry *dentry)
601{
602 struct btrfs_path *path;
603 const char *name = dentry->d_name.name;
604 int name_len = dentry->d_name.len;
605 int ret = 0;
5f39d397 606 struct extent_buffer *leaf;
39279cc3 607 struct btrfs_dir_item *di;
5f39d397 608 struct btrfs_key key;
39279cc3
CM		 609
610 path = btrfs_alloc_path();
54aa1f4d
CM		 611 if (!path) {
612 ret = -ENOMEM;
613 goto err;
614 }
615
39279cc3
CM		 616 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
617 name, name_len, -1);
618 if (IS_ERR(di)) {
619 ret = PTR_ERR(di);
620 goto err;
621 }
622 if (!di) {
623 ret = -ENOENT;
624 goto err;
625 }
5f39d397
CM		 626 leaf = path->nodes[0];
627 btrfs_dir_item_key_to_cpu(leaf, di, &key);
39279cc3 628 ret = btrfs_delete_one_dir_name(trans, root, path, di);
54aa1f4d
CM		 629 if (ret)
630 goto err;
39279cc3
CM		 631 btrfs_release_path(root, path);
632
633 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
5f39d397 634 key.objectid, name, name_len, -1);
39279cc3
CM		 635 if (IS_ERR(di)) {
636 ret = PTR_ERR(di);
637 goto err;
638 }
639 if (!di) {
640 ret = -ENOENT;
641 goto err;
642 }
643 ret = btrfs_delete_one_dir_name(trans, root, path, di);
39279cc3
CM		 644
645 dentry->d_inode->i_ctime = dir->i_ctime;
76fea00a
CM		 646 ret = btrfs_del_inode_ref(trans, root, name, name_len,
647 dentry->d_inode->i_ino,
648 dentry->d_parent->d_inode->i_ino);
649 if (ret) {
650 printk("failed to delete reference to %.*s, "
651 "inode %lu parent %lu\n", name_len, name,
652 dentry->d_inode->i_ino,
653 dentry->d_parent->d_inode->i_ino);
3954401f 654 }
39279cc3
CM		 655err:
656 btrfs_free_path(path);
657 if (!ret) {
658 dir->i_size -= name_len * 2;
79c44584 659 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
39279cc3 660 btrfs_update_inode(trans, root, dir);
6da6abae
CM		 661#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
662 dentry->d_inode->i_nlink--;
663#else
39279cc3 664 drop_nlink(dentry->d_inode);
6da6abae 665#endif
54aa1f4d 666 ret = btrfs_update_inode(trans, root, dentry->d_inode);
39279cc3
CM		 667 dir->i_sb->s_dirt = 1;
668 }
669 return ret;
670}
671
672static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
673{
674 struct btrfs_root *root;
675 struct btrfs_trans_handle *trans;
2da98f00 676 struct inode *inode = dentry->d_inode;
39279cc3 677 int ret;
1832a6d5 678 unsigned long nr = 0;
39279cc3
CM		 679
680 root = BTRFS_I(dir)->root;
681 mutex_lock(&root->fs_info->fs_mutex);
1832a6d5
CM		 682
683 ret = btrfs_check_free_space(root, 1, 1);
684 if (ret)
685 goto fail;
686
39279cc3 687 trans = btrfs_start_transaction(root, 1);
5f39d397 688
39279cc3
CM		 689 btrfs_set_trans_block_group(trans, dir);
690 ret = btrfs_unlink_trans(trans, root, dir, dentry);
d3c2fdcf 691 nr = trans->blocks_used;
5f39d397 692
2da98f00
CM		 693 if (inode->i_nlink == 0) {
694 int found;
695 /* if the inode isn't linked anywhere,
696 * we don't need to worry about
697 * data=ordered
698 */
699 found = btrfs_del_ordered_inode(inode);
700 if (found == 1) {
701 atomic_dec(&inode->i_count);
702 }
703 }
704
39279cc3 705 btrfs_end_transaction(trans, root);
1832a6d5 706fail:
39279cc3 707 mutex_unlock(&root->fs_info->fs_mutex);
d3c2fdcf 708 btrfs_btree_balance_dirty(root, nr);
e2008b61 709 btrfs_throttle(root);
39279cc3
CM		 710 return ret;
711}
712
713static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
714{
715 struct inode *inode = dentry->d_inode;
1832a6d5 716 int err = 0;
39279cc3
CM		 717 int ret;
718 struct btrfs_root *root = BTRFS_I(dir)->root;
39279cc3 719 struct btrfs_trans_handle *trans;
1832a6d5 720 unsigned long nr = 0;
39279cc3 721
134d4512
Y		 722 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
723 return -ENOTEMPTY;
724
39279cc3 725 mutex_lock(&root->fs_info->fs_mutex);
1832a6d5
CM		 726 ret = btrfs_check_free_space(root, 1, 1);
727 if (ret)
728 goto fail;
729
39279cc3
CM		 730 trans = btrfs_start_transaction(root, 1);
731 btrfs_set_trans_block_group(trans, dir);
39279cc3
CM		 732
733 /* now the directory is empty */
734 err = btrfs_unlink_trans(trans, root, dir, dentry);
735 if (!err) {
736 inode->i_size = 0;
737 }
3954401f 738
d3c2fdcf 739 nr = trans->blocks_used;
39279cc3 740 ret = btrfs_end_transaction(trans, root);
1832a6d5 741fail:
134d4512 742 mutex_unlock(&root->fs_info->fs_mutex);
d3c2fdcf 743 btrfs_btree_balance_dirty(root, nr);
e2008b61 744 btrfs_throttle(root);
3954401f 745
39279cc3
CM		 746 if (ret && !err)
747 err = ret;
748 return err;
749}
750
39279cc3
CM		 751/*
752 * this can truncate away extent items, csum items and directory items.
753 * It starts at a high offset and removes keys until it can't find
754 * any higher than i_size.
755 *
756 * csum items that cross the new i_size are truncated to the new size
757 * as well.
758 */
759static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
760 struct btrfs_root *root,
85e21bac
CM		 761 struct inode *inode,
762 u32 min_type)
39279cc3
CM		 763{
764 int ret;
765 struct btrfs_path *path;
766 struct btrfs_key key;
5f39d397 767 struct btrfs_key found_key;
39279cc3 768 u32 found_type;
5f39d397 769 struct extent_buffer *leaf;
39279cc3
CM		 770 struct btrfs_file_extent_item *fi;
771 u64 extent_start = 0;
db94535d 772 u64 extent_num_bytes = 0;
39279cc3 773 u64 item_end = 0;
7bb86316 774 u64 root_gen = 0;
d8d5f3e1 775 u64 root_owner = 0;
39279cc3
CM		 776 int found_extent;
777 int del_item;
85e21bac
CM		 778 int pending_del_nr = 0;
779 int pending_del_slot = 0;
179e29e4 780 int extent_type = -1;
39279cc3 781
a52d9a80 782 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
39279cc3 783 path = btrfs_alloc_path();
3c69faec 784 path->reada = -1;
39279cc3 785 BUG_ON(!path);
5f39d397 786
39279cc3
CM		 787 /* FIXME, add redo link to tree so we don't leak on crash */
788 key.objectid = inode->i_ino;
789 key.offset = (u64)-1;
5f39d397
CM		 790 key.type = (u8)-1;
791
85e21bac
CM		 792 btrfs_init_path(path);
793search_again:
794 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
795 if (ret < 0) {
796 goto error;
797 }
798 if (ret > 0) {
799 BUG_ON(path->slots[0] == 0);
800 path->slots[0]--;
801 }
802
39279cc3 803 while(1) {
39279cc3 804 fi = NULL;
5f39d397
CM		 805 leaf = path->nodes[0];
806 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
807 found_type = btrfs_key_type(&found_key);
39279cc3 808
5f39d397 809 if (found_key.objectid != inode->i_ino)
39279cc3 810 break;
5f39d397 811
85e21bac 812 if (found_type < min_type)
39279cc3
CM		 813 break;
814
5f39d397 815 item_end = found_key.offset;
39279cc3 816 if (found_type == BTRFS_EXTENT_DATA_KEY) {
5f39d397 817 fi = btrfs_item_ptr(leaf, path->slots[0],
39279cc3 818 struct btrfs_file_extent_item);
179e29e4
CM		 819 extent_type = btrfs_file_extent_type(leaf, fi);
820 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
5f39d397 821 item_end +=
db94535d 822 btrfs_file_extent_num_bytes(leaf, fi);
179e29e4
CM		 823 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
824 struct btrfs_item *item = btrfs_item_nr(leaf,
825 path->slots[0]);
826 item_end += btrfs_file_extent_inline_len(leaf,
827 item);
39279cc3 828 }
008630c1 829 item_end--;
39279cc3
CM		 830 }
831 if (found_type == BTRFS_CSUM_ITEM_KEY) {
832 ret = btrfs_csum_truncate(trans, root, path,
833 inode->i_size);
834 BUG_ON(ret);
835 }
008630c1 836 if (item_end < inode->i_size) {
b888db2b
CM		 837 if (found_type == BTRFS_DIR_ITEM_KEY) {
838 found_type = BTRFS_INODE_ITEM_KEY;
839 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
840 found_type = BTRFS_CSUM_ITEM_KEY;
85e21bac
CM		 841 } else if (found_type == BTRFS_EXTENT_DATA_KEY) {
842 found_type = BTRFS_XATTR_ITEM_KEY;
843 } else if (found_type == BTRFS_XATTR_ITEM_KEY) {
844 found_type = BTRFS_INODE_REF_KEY;
b888db2b
CM		 845 } else if (found_type) {
846 found_type--;
847 } else {
848 break;
39279cc3 849 }
a61721d5 850 btrfs_set_key_type(&key, found_type);
85e21bac 851 goto next;
39279cc3 852 }
5f39d397 853 if (found_key.offset >= inode->i_size)
39279cc3
CM		 854 del_item = 1;
855 else
856 del_item = 0;
857 found_extent = 0;
858
859 /* FIXME, shrink the extent if the ref count is only 1 */
179e29e4
CM		 860 if (found_type != BTRFS_EXTENT_DATA_KEY)
861 goto delete;
862
863 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
39279cc3 864 u64 num_dec;
db94535d 865 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
39279cc3 866 if (!del_item) {
db94535d
CM		 867 u64 orig_num_bytes =
868 btrfs_file_extent_num_bytes(leaf, fi);
869 extent_num_bytes = inode->i_size -
5f39d397 870 found_key.offset + root->sectorsize - 1;
b1632b10
Y		 871 extent_num_bytes = extent_num_bytes &
872 ~((u64)root->sectorsize - 1);
db94535d
CM		 873 btrfs_set_file_extent_num_bytes(leaf, fi,
874 extent_num_bytes);
875 num_dec = (orig_num_bytes -
9069218d
CM		 876 extent_num_bytes);
877 if (extent_start != 0)
878 dec_i_blocks(inode, num_dec);
5f39d397 879 btrfs_mark_buffer_dirty(leaf);
39279cc3 880 } else {
db94535d
CM		 881 extent_num_bytes =
882 btrfs_file_extent_disk_num_bytes(leaf,
883 fi);
39279cc3 884 /* FIXME blocksize != 4096 */
9069218d 885 num_dec = btrfs_file_extent_num_bytes(leaf, fi);
39279cc3
CM		 886 if (extent_start != 0) {
887 found_extent = 1;
9069218d 888 dec_i_blocks(inode, num_dec);
39279cc3 889 }
d8d5f3e1
CM		 890 root_gen = btrfs_header_generation(leaf);
891 root_owner = btrfs_header_owner(leaf);
39279cc3 892 }
9069218d
CM		 893 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
894 if (!del_item) {
895 u32 newsize = inode->i_size - found_key.offset;
896 dec_i_blocks(inode, item_end + 1 -
897 found_key.offset - newsize);
898 newsize =
899 btrfs_file_extent_calc_inline_size(newsize);
900 ret = btrfs_truncate_item(trans, root, path,
901 newsize, 1);
902 BUG_ON(ret);
903 } else {
904 dec_i_blocks(inode, item_end + 1 -
905 found_key.offset);
906 }
39279cc3 907 }
179e29e4 908delete:
39279cc3 909 if (del_item) {
85e21bac
CM		 910 if (!pending_del_nr) {
911 /* no pending yet, add ourselves */
912 pending_del_slot = path->slots[0];
913 pending_del_nr = 1;
914 } else if (pending_del_nr &&
915 path->slots[0] + 1 == pending_del_slot) {
916 /* hop on the pending chunk */
917 pending_del_nr++;
918 pending_del_slot = path->slots[0];
919 } else {
920 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot);
921 }
39279cc3
CM		 922 } else {
923 break;
924 }
39279cc3
CM		 925 if (found_extent) {
926 ret = btrfs_free_extent(trans, root, extent_start,
7bb86316 927 extent_num_bytes,
d8d5f3e1 928 root_owner,
7bb86316
CM		 929 root_gen, inode->i_ino,
930 found_key.offset, 0);
39279cc3
CM		 931 BUG_ON(ret);
932 }
85e21bac
CM		 933next:
934 if (path->slots[0] == 0) {
935 if (pending_del_nr)
936 goto del_pending;
937 btrfs_release_path(root, path);
938 goto search_again;
939 }
940
941 path->slots[0]--;
942 if (pending_del_nr &&
943 path->slots[0] + 1 != pending_del_slot) {
944 struct btrfs_key debug;
945del_pending:
946 btrfs_item_key_to_cpu(path->nodes[0], &debug,
947 pending_del_slot);
948 ret = btrfs_del_items(trans, root, path,
949 pending_del_slot,
950 pending_del_nr);
951 BUG_ON(ret);
952 pending_del_nr = 0;
953 btrfs_release_path(root, path);
954 goto search_again;
955 }
39279cc3
CM		 956 }
957 ret = 0;
958error:
85e21bac
CM		 959 if (pending_del_nr) {
960 ret = btrfs_del_items(trans, root, path, pending_del_slot,
961 pending_del_nr);
962 }
39279cc3
CM		 963 btrfs_release_path(root, path);
964 btrfs_free_path(path);
965 inode->i_sb->s_dirt = 1;
966 return ret;
967}
968
b888db2b 969static int btrfs_cow_one_page(struct inode *inode, struct page *page,
a52d9a80
CM		 970 size_t zero_start)
971{
972 char *kaddr;
d1310b2e 973 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
35ebb934 974 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
b888db2b 975 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
1832a6d5 976 int ret = 0;
a52d9a80 977
190662b2 978 WARN_ON(!PageLocked(page));
b3cfa35a 979 set_page_extent_mapped(page);
a52d9a80 980
d1310b2e 981 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
d1310b2e 982 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
b888db2b 983 page_end, GFP_NOFS);
1832a6d5 984
a52d9a80 985 if (zero_start != PAGE_CACHE_SIZE) {
b888db2b 986 kaddr = kmap(page);
a52d9a80
CM		 987 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
988 flush_dcache_page(page);
b888db2b 989 kunmap(page);
a52d9a80 990 }
b888db2b 991 set_page_dirty(page);
d1310b2e 992 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
a52d9a80 993
a52d9a80
CM		 994 return ret;
995}
996
39279cc3
CM		 997/*
998 * taken from block_truncate_page, but does cow as it zeros out
999 * any bytes left in the last page in the file.
1000 */
1001static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
1002{
1003 struct inode *inode = mapping->host;
db94535d
CM		 1004 struct btrfs_root *root = BTRFS_I(inode)->root;
1005 u32 blocksize = root->sectorsize;
39279cc3
CM		 1006 pgoff_t index = from >> PAGE_CACHE_SHIFT;
1007 unsigned offset = from & (PAGE_CACHE_SIZE-1);
1008 struct page *page;
39279cc3 1009 int ret = 0;
a52d9a80 1010 u64 page_start;
39279cc3
CM		 1011
1012 if ((offset & (blocksize - 1)) == 0)
1013 goto out;
1014
1015 ret = -ENOMEM;
1016 page = grab_cache_page(mapping, index);
1017 if (!page)
1018 goto out;
39279cc3 1019 if (!PageUptodate(page)) {
9ebefb18 1020 ret = btrfs_readpage(NULL, page);
39279cc3
CM		 1021 lock_page(page);
1022 if (!PageUptodate(page)) {
1023 ret = -EIO;
1024 goto out;
1025 }
1026 }
35ebb934 1027 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
a52d9a80 1028
b888db2b 1029 ret = btrfs_cow_one_page(inode, page, offset);
39279cc3 1030
39279cc3
CM		 1031 unlock_page(page);
1032 page_cache_release(page);
1033out:
1034 return ret;
1035}
1036
1037static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
1038{
1039 struct inode *inode = dentry->d_inode;
1040 int err;
1041
1042 err = inode_change_ok(inode, attr);
1043 if (err)
1044 return err;
1045
1046 if (S_ISREG(inode->i_mode) &&
1047 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
1048 struct btrfs_trans_handle *trans;
1049 struct btrfs_root *root = BTRFS_I(inode)->root;
d1310b2e 1050 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2bf5a725 1051
5f39d397 1052 u64 mask = root->sectorsize - 1;
1b0f7c29 1053 u64 hole_start = (inode->i_size + mask) & ~mask;
f392a938 1054 u64 block_end = (attr->ia_size + mask) & ~mask;
39279cc3 1055 u64 hole_size;
179e29e4 1056 u64 alloc_hint = 0;
39279cc3 1057
1b0f7c29 1058 if (attr->ia_size <= hole_start)
39279cc3
CM		 1059 goto out;
1060
1832a6d5
CM		 1061 mutex_lock(&root->fs_info->fs_mutex);
1062 err = btrfs_check_free_space(root, 1, 0);
1063 mutex_unlock(&root->fs_info->fs_mutex);
1064 if (err)
1065 goto fail;
1066
39279cc3
CM		 1067 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1068
1b0f7c29 1069 lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
5f56406a 1070 hole_size = block_end - hole_start;
39279cc3
CM		 1071
1072 mutex_lock(&root->fs_info->fs_mutex);
1073 trans = btrfs_start_transaction(root, 1);
1074 btrfs_set_trans_block_group(trans, inode);
2bf5a725 1075 err = btrfs_drop_extents(trans, root, inode,
1b0f7c29 1076 hole_start, block_end, hole_start,
3326d1b0 1077 &alloc_hint);
2bf5a725 1078
179e29e4
CM		 1079 if (alloc_hint != EXTENT_MAP_INLINE) {
1080 err = btrfs_insert_file_extent(trans, root,
1081 inode->i_ino,
5f56406a
CM		 1082 hole_start, 0, 0,
1083 hole_size);
d1310b2e
CM		 1084 btrfs_drop_extent_cache(inode, hole_start,
1085 hole_size - 1);
5f56406a 1086 btrfs_check_file(root, inode);
179e29e4 1087 }
39279cc3
CM		 1088 btrfs_end_transaction(trans, root);
1089 mutex_unlock(&root->fs_info->fs_mutex);
1b0f7c29 1090 unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
54aa1f4d
CM		 1091 if (err)
1092 return err;
39279cc3
CM		 1093 }
1094out:
1095 err = inode_setattr(inode, attr);
1832a6d5 1096fail:
39279cc3
CM		 1097 return err;
1098}
61295eb8 1099
2da98f00 1100void btrfs_put_inode(struct inode *inode)
61295eb8 1101{
2da98f00
CM		 1102 int ret;
1103
1104 if (!BTRFS_I(inode)->ordered_trans) {
61295eb8
CM		 1105 return;
1106 }
2da98f00
CM		 1107
1108 if (mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY) ||
1109 mapping_tagged(inode->i_mapping, PAGECACHE_TAG_WRITEBACK))
1110 return;
1111
1112 ret = btrfs_del_ordered_inode(inode);
1113 if (ret == 1) {
1114 atomic_dec(&inode->i_count);
1115 }
61295eb8
CM		 1116}
1117
39279cc3
CM		 1118void btrfs_delete_inode(struct inode *inode)
1119{
1120 struct btrfs_trans_handle *trans;
1121 struct btrfs_root *root = BTRFS_I(inode)->root;
d3c2fdcf 1122 unsigned long nr;
39279cc3
CM		 1123 int ret;
1124
1125 truncate_inode_pages(&inode->i_data, 0);
1126 if (is_bad_inode(inode)) {
1127 goto no_delete;
1128 }
5f39d397 1129
39279cc3
CM		 1130 inode->i_size = 0;
1131 mutex_lock(&root->fs_info->fs_mutex);
1132 trans = btrfs_start_transaction(root, 1);
5f39d397 1133
39279cc3 1134 btrfs_set_trans_block_group(trans, inode);
85e21bac 1135 ret = btrfs_truncate_in_trans(trans, root, inode, 0);
54aa1f4d
CM		 1136 if (ret)
1137 goto no_delete_lock;
85e21bac 1138
d3c2fdcf 1139 nr = trans->blocks_used;
85e21bac 1140 clear_inode(inode);
5f39d397 1141
39279cc3
CM		 1142 btrfs_end_transaction(trans, root);
1143 mutex_unlock(&root->fs_info->fs_mutex);
d3c2fdcf 1144 btrfs_btree_balance_dirty(root, nr);
e2008b61 1145 btrfs_throttle(root);
39279cc3 1146 return;
54aa1f4d
CM		 1147
1148no_delete_lock:
d3c2fdcf 1149 nr = trans->blocks_used;
54aa1f4d
CM		 1150 btrfs_end_transaction(trans, root);
1151 mutex_unlock(&root->fs_info->fs_mutex);
d3c2fdcf 1152 btrfs_btree_balance_dirty(root, nr);
e2008b61 1153 btrfs_throttle(root);
39279cc3
CM		 1154no_delete:
1155 clear_inode(inode);
1156}
1157
1158/*
1159 * this returns the key found in the dir entry in the location pointer.
1160 * If no dir entries were found, location->objectid is 0.
1161 */
1162static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
1163 struct btrfs_key *location)
1164{
1165 const char *name = dentry->d_name.name;
1166 int namelen = dentry->d_name.len;
1167 struct btrfs_dir_item *di;
1168 struct btrfs_path *path;
1169 struct btrfs_root *root = BTRFS_I(dir)->root;
0d9f7f3e 1170 int ret = 0;
39279cc3 1171
3954401f
CM		 1172 if (namelen == 1 && strcmp(name, ".") == 0) {
1173 location->objectid = dir->i_ino;
1174 location->type = BTRFS_INODE_ITEM_KEY;
1175 location->offset = 0;
1176 return 0;
1177 }
39279cc3
CM		 1178 path = btrfs_alloc_path();
1179 BUG_ON(!path);
3954401f 1180
7a720536 1181 if (namelen == 2 && strcmp(name, "..") == 0) {
3954401f
CM		 1182 struct btrfs_key key;
1183 struct extent_buffer *leaf;
1184 u32 nritems;
1185 int slot;
1186
1187 key.objectid = dir->i_ino;
1188 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1189 key.offset = 0;
1190 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1191 BUG_ON(ret == 0);
1192 ret = 0;
1193
1194 leaf = path->nodes[0];
1195 slot = path->slots[0];
1196 nritems = btrfs_header_nritems(leaf);
1197 if (slot >= nritems)
1198 goto out_err;
1199
1200 btrfs_item_key_to_cpu(leaf, &key, slot);
1201 if (key.objectid != dir->i_ino ||
1202 key.type != BTRFS_INODE_REF_KEY) {
1203 goto out_err;
1204 }
1205 location->objectid = key.offset;
1206 location->type = BTRFS_INODE_ITEM_KEY;
1207 location->offset = 0;
1208 goto out;
1209 }
1210
39279cc3
CM		 1211 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
1212 namelen, 0);
0d9f7f3e
Y		 1213 if (IS_ERR(di))
1214 ret = PTR_ERR(di);
39279cc3 1215 if (!di || IS_ERR(di)) {
3954401f 1216 goto out_err;
39279cc3 1217 }
5f39d397 1218 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
39279cc3 1219out:
39279cc3
CM		 1220 btrfs_free_path(path);
1221 return ret;
3954401f
CM		 1222out_err:
1223 location->objectid = 0;
1224 goto out;
39279cc3
CM		 1225}
1226
1227/*
1228 * when we hit a tree root in a directory, the btrfs part of the inode
1229 * needs to be changed to reflect the root directory of the tree root. This
1230 * is kind of like crossing a mount point.
1231 */
1232static int fixup_tree_root_location(struct btrfs_root *root,
1233 struct btrfs_key *location,
58176a96
JB		 1234 struct btrfs_root **sub_root,
1235 struct dentry *dentry)
39279cc3
CM		 1236{
1237 struct btrfs_path *path;
1238 struct btrfs_root_item *ri;
1239
1240 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
1241 return 0;
1242 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1243 return 0;
1244
1245 path = btrfs_alloc_path();
1246 BUG_ON(!path);
1247 mutex_lock(&root->fs_info->fs_mutex);
1248
58176a96
JB		 1249 *sub_root = btrfs_read_fs_root(root->fs_info, location,
1250 dentry->d_name.name,
1251 dentry->d_name.len);
39279cc3
CM		 1252 if (IS_ERR(*sub_root))
1253 return PTR_ERR(*sub_root);
1254
1255 ri = &(*sub_root)->root_item;
1256 location->objectid = btrfs_root_dirid(ri);
39279cc3
CM		 1257 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1258 location->offset = 0;
1259
1260 btrfs_free_path(path);
1261 mutex_unlock(&root->fs_info->fs_mutex);
1262 return 0;
1263}
1264
1265static int btrfs_init_locked_inode(struct inode *inode, void *p)
1266{
1267 struct btrfs_iget_args *args = p;
1268 inode->i_ino = args->ino;
1269 BTRFS_I(inode)->root = args->root;
9069218d 1270 BTRFS_I(inode)->delalloc_bytes = 0;
d1310b2e
CM		 1271 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1272 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
b888db2b 1273 inode->i_mapping, GFP_NOFS);
39279cc3
CM		 1274 return 0;
1275}
1276
1277static int btrfs_find_actor(struct inode *inode, void *opaque)
1278{
1279 struct btrfs_iget_args *args = opaque;
1280 return (args->ino == inode->i_ino &&
1281 args->root == BTRFS_I(inode)->root);
1282}
1283
dc17ff8f
CM		 1284struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
1285 u64 root_objectid)
1286{
1287 struct btrfs_iget_args args;
1288 args.ino = objectid;
1289 args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
1290
1291 if (!args.root)
1292 return NULL;
1293
1294 return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
1295}
1296
39279cc3
CM		 1297struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
1298 struct btrfs_root *root)
1299{
1300 struct inode *inode;
1301 struct btrfs_iget_args args;
1302 args.ino = objectid;
1303 args.root = root;
1304
1305 inode = iget5_locked(s, objectid, btrfs_find_actor,
1306 btrfs_init_locked_inode,
1307 (void *)&args);
1308 return inode;
1309}
1310
1311static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
1312 struct nameidata *nd)
1313{
1314 struct inode * inode;
1315 struct btrfs_inode *bi = BTRFS_I(dir);
1316 struct btrfs_root *root = bi->root;
1317 struct btrfs_root *sub_root = root;
1318 struct btrfs_key location;
1319 int ret;
1320
1321 if (dentry->d_name.len > BTRFS_NAME_LEN)
1322 return ERR_PTR(-ENAMETOOLONG);
5f39d397 1323
39279cc3
CM		 1324 mutex_lock(&root->fs_info->fs_mutex);
1325 ret = btrfs_inode_by_name(dir, dentry, &location);
1326 mutex_unlock(&root->fs_info->fs_mutex);
5f39d397 1327
39279cc3
CM		 1328 if (ret < 0)
1329 return ERR_PTR(ret);
5f39d397 1330
39279cc3
CM		 1331 inode = NULL;
1332 if (location.objectid) {
58176a96
JB		 1333 ret = fixup_tree_root_location(root, &location, &sub_root,
1334 dentry);
39279cc3
CM		 1335 if (ret < 0)
1336 return ERR_PTR(ret);
1337 if (ret > 0)
1338 return ERR_PTR(-ENOENT);
1339 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
1340 sub_root);
1341 if (!inode)
1342 return ERR_PTR(-EACCES);
1343 if (inode->i_state & I_NEW) {
1344 /* the inode and parent dir are two different roots */
1345 if (sub_root != root) {
1346 igrab(inode);
1347 sub_root->inode = inode;
1348 }
1349 BTRFS_I(inode)->root = sub_root;
1350 memcpy(&BTRFS_I(inode)->location, &location,
1351 sizeof(location));
1352 btrfs_read_locked_inode(inode);
1353 unlock_new_inode(inode);
1354 }
1355 }
1356 return d_splice_alias(inode, dentry);
1357}
1358
39279cc3
CM		 1359static unsigned char btrfs_filetype_table[] = {
1360 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1361};
1362
1363static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1364{
6da6abae 1365 struct inode *inode = filp->f_dentry->d_inode;
39279cc3
CM		 1366 struct btrfs_root *root = BTRFS_I(inode)->root;
1367 struct btrfs_item *item;
1368 struct btrfs_dir_item *di;
1369 struct btrfs_key key;
5f39d397 1370 struct btrfs_key found_key;
39279cc3
CM		 1371 struct btrfs_path *path;
1372 int ret;
1373 u32 nritems;
5f39d397 1374 struct extent_buffer *leaf;
39279cc3
CM		 1375 int slot;
1376 int advance;
1377 unsigned char d_type;
1378 int over = 0;
1379 u32 di_cur;
1380 u32 di_total;
1381 u32 di_len;
1382 int key_type = BTRFS_DIR_INDEX_KEY;
5f39d397
CM		 1383 char tmp_name[32];
1384 char *name_ptr;
1385 int name_len;
39279cc3
CM		 1386
1387 /* FIXME, use a real flag for deciding about the key type */
1388 if (root->fs_info->tree_root == root)
1389 key_type = BTRFS_DIR_ITEM_KEY;
5f39d397 1390
3954401f
CM		 1391 /* special case for "." */
1392 if (filp->f_pos == 0) {
1393 over = filldir(dirent, ".", 1,
1394 1, inode->i_ino,
1395 DT_DIR);
1396 if (over)
1397 return 0;
1398 filp->f_pos = 1;
1399 }
1400
39279cc3
CM		 1401 mutex_lock(&root->fs_info->fs_mutex);
1402 key.objectid = inode->i_ino;
3954401f
CM		 1403 path = btrfs_alloc_path();
1404 path->reada = 2;
1405
1406 /* special case for .., just use the back ref */
1407 if (filp->f_pos == 1) {
1408 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1409 key.offset = 0;
1410 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1411 BUG_ON(ret == 0);
1412 leaf = path->nodes[0];
1413 slot = path->slots[0];
1414 nritems = btrfs_header_nritems(leaf);
1415 if (slot >= nritems) {
1416 btrfs_release_path(root, path);
1417 goto read_dir_items;
1418 }
1419 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1420 btrfs_release_path(root, path);
1421 if (found_key.objectid != key.objectid ||
1422 found_key.type != BTRFS_INODE_REF_KEY)
1423 goto read_dir_items;
1424 over = filldir(dirent, "..", 2,
1425 2, found_key.offset, DT_DIR);
1426 if (over)
1427 goto nopos;
1428 filp->f_pos = 2;
1429 }
1430
1431read_dir_items:
39279cc3
CM		 1432 btrfs_set_key_type(&key, key_type);
1433 key.offset = filp->f_pos;
5f39d397 1434
39279cc3
CM		 1435 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1436 if (ret < 0)
1437 goto err;
1438 advance = 0;
39279cc3 1439 while(1) {
5f39d397
CM		 1440 leaf = path->nodes[0];
1441 nritems = btrfs_header_nritems(leaf);
39279cc3
CM		 1442 slot = path->slots[0];
1443 if (advance || slot >= nritems) {
1444 if (slot >= nritems -1) {
39279cc3
CM		 1445 ret = btrfs_next_leaf(root, path);
1446 if (ret)
1447 break;
5f39d397
CM		 1448 leaf = path->nodes[0];
1449 nritems = btrfs_header_nritems(leaf);
39279cc3
CM		 1450 slot = path->slots[0];
1451 } else {
1452 slot++;
1453 path->slots[0]++;
1454 }
1455 }
1456 advance = 1;
5f39d397
CM		 1457 item = btrfs_item_nr(leaf, slot);
1458 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1459
1460 if (found_key.objectid != key.objectid)
39279cc3 1461 break;
5f39d397 1462 if (btrfs_key_type(&found_key) != key_type)
39279cc3 1463 break;
5f39d397 1464 if (found_key.offset < filp->f_pos)
39279cc3 1465 continue;
5f39d397
CM		 1466
1467 filp->f_pos = found_key.offset;
39279cc3
CM		 1468 advance = 1;
1469 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1470 di_cur = 0;
5f39d397 1471 di_total = btrfs_item_size(leaf, item);
39279cc3 1472 while(di_cur < di_total) {
5f39d397
CM		 1473 struct btrfs_key location;
1474
1475 name_len = btrfs_dir_name_len(leaf, di);
1476 if (name_len < 32) {
1477 name_ptr = tmp_name;
1478 } else {
1479 name_ptr = kmalloc(name_len, GFP_NOFS);
1480 BUG_ON(!name_ptr);
1481 }
1482 read_extent_buffer(leaf, name_ptr,
1483 (unsigned long)(di + 1), name_len);
1484
1485 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1486 btrfs_dir_item_key_to_cpu(leaf, di, &location);
5f39d397
CM		 1487 over = filldir(dirent, name_ptr, name_len,
1488 found_key.offset,
1489 location.objectid,
39279cc3 1490 d_type);
5f39d397
CM		 1491
1492 if (name_ptr != tmp_name)
1493 kfree(name_ptr);
1494
39279cc3
CM		 1495 if (over)
1496 goto nopos;
5103e947
JB		 1497 di_len = btrfs_dir_name_len(leaf, di) +
1498 btrfs_dir_data_len(leaf, di) +sizeof(*di);
39279cc3
CM		 1499 di_cur += di_len;
1500 di = (struct btrfs_dir_item *)((char *)di + di_len);
1501 }
1502 }
5e591a07
YZ		 1503 if (key_type == BTRFS_DIR_INDEX_KEY)
1504 filp->f_pos = INT_LIMIT(typeof(filp->f_pos));
1505 else
1506 filp->f_pos++;
39279cc3
CM		 1507nopos:
1508 ret = 0;
1509err:
1510 btrfs_release_path(root, path);
1511 btrfs_free_path(path);
1512 mutex_unlock(&root->fs_info->fs_mutex);
1513 return ret;
1514}
1515
1516int btrfs_write_inode(struct inode *inode, int wait)
1517{
1518 struct btrfs_root *root = BTRFS_I(inode)->root;
1519 struct btrfs_trans_handle *trans;
1520 int ret = 0;
1521
1522 if (wait) {
1523 mutex_lock(&root->fs_info->fs_mutex);
1524 trans = btrfs_start_transaction(root, 1);
1525 btrfs_set_trans_block_group(trans, inode);
1526 ret = btrfs_commit_transaction(trans, root);
1527 mutex_unlock(&root->fs_info->fs_mutex);
1528 }
1529 return ret;
1530}
1531
1532/*
54aa1f4d 1533 * This is somewhat expensive, updating the tree every time the
39279cc3
CM		 1534 * inode changes. But, it is most likely to find the inode in cache.
1535 * FIXME, needs more benchmarking...there are no reasons other than performance
1536 * to keep or drop this code.
1537 */
1538void btrfs_dirty_inode(struct inode *inode)
1539{
1540 struct btrfs_root *root = BTRFS_I(inode)->root;
1541 struct btrfs_trans_handle *trans;
1542
1543 mutex_lock(&root->fs_info->fs_mutex);
1544 trans = btrfs_start_transaction(root, 1);
1545 btrfs_set_trans_block_group(trans, inode);
1546 btrfs_update_inode(trans, root, inode);
1547 btrfs_end_transaction(trans, root);
1548 mutex_unlock(&root->fs_info->fs_mutex);
39279cc3
CM		 1549}
1550
1551static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1552 struct btrfs_root *root,
9c58309d
CM		 1553 const char *name, int name_len,
1554 u64 ref_objectid,
39279cc3
CM		 1555 u64 objectid,
1556 struct btrfs_block_group_cache *group,
1557 int mode)
1558{
1559 struct inode *inode;
5f39d397 1560 struct btrfs_inode_item *inode_item;
6324fbf3 1561 struct btrfs_block_group_cache *new_inode_group;
39279cc3 1562 struct btrfs_key *location;
5f39d397 1563 struct btrfs_path *path;
9c58309d
CM		 1564 struct btrfs_inode_ref *ref;
1565 struct btrfs_key key[2];
1566 u32 sizes[2];
1567 unsigned long ptr;
39279cc3
CM		 1568 int ret;
1569 int owner;
1570
5f39d397
CM		 1571 path = btrfs_alloc_path();
1572 BUG_ON(!path);
1573
39279cc3
CM		 1574 inode = new_inode(root->fs_info->sb);
1575 if (!inode)
1576 return ERR_PTR(-ENOMEM);
1577
d1310b2e
CM		 1578 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1579 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
b888db2b 1580 inode->i_mapping, GFP_NOFS);
9069218d 1581 BTRFS_I(inode)->delalloc_bytes = 0;
39279cc3 1582 BTRFS_I(inode)->root = root;
b888db2b 1583
39279cc3
CM		 1584 if (mode & S_IFDIR)
1585 owner = 0;
1586 else
1587 owner = 1;
6324fbf3 1588 new_inode_group = btrfs_find_block_group(root, group, 0,
0b86a832 1589 BTRFS_BLOCK_GROUP_METADATA, owner);
6324fbf3
CM		 1590 if (!new_inode_group) {
1591 printk("find_block group failed\n");
1592 new_inode_group = group;
1593 }
1594 BTRFS_I(inode)->block_group = new_inode_group;
b98b6767 1595 BTRFS_I(inode)->flags = 0;
9c58309d
CM		 1596
1597 key[0].objectid = objectid;
1598 btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
1599 key[0].offset = 0;
1600
1601 key[1].objectid = objectid;
1602 btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
1603 key[1].offset = ref_objectid;
1604
1605 sizes[0] = sizeof(struct btrfs_inode_item);
1606 sizes[1] = name_len + sizeof(*ref);
1607
1608 ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
1609 if (ret != 0)
5f39d397
CM		 1610 goto fail;
1611
9c58309d
CM		 1612 if (objectid > root->highest_inode)
1613 root->highest_inode = objectid;
1614
39279cc3
CM		 1615 inode->i_uid = current->fsuid;
1616 inode->i_gid = current->fsgid;
1617 inode->i_mode = mode;
1618 inode->i_ino = objectid;
1619 inode->i_blocks = 0;
1620 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
5f39d397
CM		 1621 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1622 struct btrfs_inode_item);
1623 fill_inode_item(path->nodes[0], inode_item, inode);
9c58309d
CM		 1624
1625 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
1626 struct btrfs_inode_ref);
1627 btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
1628 ptr = (unsigned long)(ref + 1);
1629 write_extent_buffer(path->nodes[0], name, ptr, name_len);
1630
5f39d397
CM		 1631 btrfs_mark_buffer_dirty(path->nodes[0]);
1632 btrfs_free_path(path);
1633
39279cc3
CM		 1634 location = &BTRFS_I(inode)->location;
1635 location->objectid = objectid;
39279cc3
CM		 1636 location->offset = 0;
1637 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1638
39279cc3
CM		 1639 insert_inode_hash(inode);
1640 return inode;
5f39d397
CM		 1641fail:
1642 btrfs_free_path(path);
1643 return ERR_PTR(ret);
39279cc3
CM		 1644}
1645
1646static inline u8 btrfs_inode_type(struct inode *inode)
1647{
1648 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1649}
1650
1651static int btrfs_add_link(struct btrfs_trans_handle *trans,
9c58309d
CM		 1652 struct dentry *dentry, struct inode *inode,
1653 int add_backref)
39279cc3
CM		 1654{
1655 int ret;
1656 struct btrfs_key key;
1657 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
79c44584 1658 struct inode *parent_inode;
5f39d397 1659
39279cc3 1660 key.objectid = inode->i_ino;
39279cc3
CM		 1661 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1662 key.offset = 0;
1663
1664 ret = btrfs_insert_dir_item(trans, root,
1665 dentry->d_name.name, dentry->d_name.len,
1666 dentry->d_parent->d_inode->i_ino,
1667 &key, btrfs_inode_type(inode));
1668 if (ret == 0) {
9c58309d
CM		 1669 if (add_backref) {
1670 ret = btrfs_insert_inode_ref(trans, root,
1671 dentry->d_name.name,
1672 dentry->d_name.len,
1673 inode->i_ino,
1674 dentry->d_parent->d_inode->i_ino);
1675 }
79c44584
CM		 1676 parent_inode = dentry->d_parent->d_inode;
1677 parent_inode->i_size += dentry->d_name.len * 2;
1678 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
39279cc3
CM		 1679 ret = btrfs_update_inode(trans, root,
1680 dentry->d_parent->d_inode);
1681 }
1682 return ret;
1683}
1684
1685static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
9c58309d
CM		 1686 struct dentry *dentry, struct inode *inode,
1687 int backref)
39279cc3 1688{
9c58309d 1689 int err = btrfs_add_link(trans, dentry, inode, backref);
39279cc3
CM		 1690 if (!err) {
1691 d_instantiate(dentry, inode);
1692 return 0;
1693 }
1694 if (err > 0)
1695 err = -EEXIST;
1696 return err;
1697}
1698
618e21d5
JB		 1699static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1700 int mode, dev_t rdev)
1701{
1702 struct btrfs_trans_handle *trans;
1703 struct btrfs_root *root = BTRFS_I(dir)->root;
1832a6d5 1704 struct inode *inode = NULL;
618e21d5
JB		 1705 int err;
1706 int drop_inode = 0;
1707 u64 objectid;
1832a6d5 1708 unsigned long nr = 0;
618e21d5
JB		 1709
1710 if (!new_valid_dev(rdev))
1711 return -EINVAL;
1712
1713 mutex_lock(&root->fs_info->fs_mutex);
1832a6d5
CM		 1714 err = btrfs_check_free_space(root, 1, 0);
1715 if (err)
1716 goto fail;
1717
618e21d5
JB		 1718 trans = btrfs_start_transaction(root, 1);
1719 btrfs_set_trans_block_group(trans, dir);
1720
1721 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1722 if (err) {
1723 err = -ENOSPC;
1724 goto out_unlock;
1725 }
1726
9c58309d
CM		 1727 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1728 dentry->d_name.len,
1729 dentry->d_parent->d_inode->i_ino, objectid,
618e21d5
JB		 1730 BTRFS_I(dir)->block_group, mode);
1731 err = PTR_ERR(inode);
1732 if (IS_ERR(inode))
1733 goto out_unlock;
1734
1735 btrfs_set_trans_block_group(trans, inode);
9c58309d 1736 err = btrfs_add_nondir(trans, dentry, inode, 0);
618e21d5
JB		 1737 if (err)
1738 drop_inode = 1;
1739 else {
1740 inode->i_op = &btrfs_special_inode_operations;
1741 init_special_inode(inode, inode->i_mode, rdev);
1b4ab1bb 1742 btrfs_update_inode(trans, root, inode);
618e21d5
JB		 1743 }
1744 dir->i_sb->s_dirt = 1;
1745 btrfs_update_inode_block_group(trans, inode);
1746 btrfs_update_inode_block_group(trans, dir);
1747out_unlock:
d3c2fdcf 1748 nr = trans->blocks_used;
618e21d5 1749 btrfs_end_transaction(trans, root);
1832a6d5 1750fail:
618e21d5
JB		 1751 mutex_unlock(&root->fs_info->fs_mutex);
1752
1753 if (drop_inode) {
1754 inode_dec_link_count(inode);
1755 iput(inode);
1756 }
d3c2fdcf 1757 btrfs_btree_balance_dirty(root, nr);
e2008b61 1758 btrfs_throttle(root);
618e21d5
JB		 1759 return err;
1760}
1761
39279cc3
CM		 1762static int btrfs_create(struct inode *dir, struct dentry *dentry,
1763 int mode, struct nameidata *nd)
1764{
1765 struct btrfs_trans_handle *trans;
1766 struct btrfs_root *root = BTRFS_I(dir)->root;
1832a6d5 1767 struct inode *inode = NULL;
39279cc3
CM		 1768 int err;
1769 int drop_inode = 0;
1832a6d5 1770 unsigned long nr = 0;
39279cc3
CM		 1771 u64 objectid;
1772
1773 mutex_lock(&root->fs_info->fs_mutex);
1832a6d5
CM		 1774 err = btrfs_check_free_space(root, 1, 0);
1775 if (err)
1776 goto fail;
39279cc3
CM		 1777 trans = btrfs_start_transaction(root, 1);
1778 btrfs_set_trans_block_group(trans, dir);
1779
1780 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1781 if (err) {
1782 err = -ENOSPC;
1783 goto out_unlock;
1784 }
1785
9c58309d
CM		 1786 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1787 dentry->d_name.len,
1788 dentry->d_parent->d_inode->i_ino,
1789 objectid, BTRFS_I(dir)->block_group, mode);
39279cc3
CM		 1790 err = PTR_ERR(inode);
1791 if (IS_ERR(inode))
1792 goto out_unlock;
1793
1794 btrfs_set_trans_block_group(trans, inode);
9c58309d 1795 err = btrfs_add_nondir(trans, dentry, inode, 0);
39279cc3
CM		 1796 if (err)
1797 drop_inode = 1;
1798 else {
1799 inode->i_mapping->a_ops = &btrfs_aops;
04160088 1800 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
39279cc3
CM		 1801 inode->i_fop = &btrfs_file_operations;
1802 inode->i_op = &btrfs_file_inode_operations;
d1310b2e
CM		 1803 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1804 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
a52d9a80 1805 inode->i_mapping, GFP_NOFS);
9069218d 1806 BTRFS_I(inode)->delalloc_bytes = 0;
d1310b2e 1807 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
39279cc3
CM		 1808 }
1809 dir->i_sb->s_dirt = 1;
1810 btrfs_update_inode_block_group(trans, inode);
1811 btrfs_update_inode_block_group(trans, dir);
1812out_unlock:
d3c2fdcf 1813 nr = trans->blocks_used;
39279cc3 1814 btrfs_end_transaction(trans, root);
1832a6d5 1815fail:
39279cc3
CM		 1816 mutex_unlock(&root->fs_info->fs_mutex);
1817
1818 if (drop_inode) {
1819 inode_dec_link_count(inode);
1820 iput(inode);
1821 }
d3c2fdcf 1822 btrfs_btree_balance_dirty(root, nr);
e2008b61 1823 btrfs_throttle(root);
39279cc3
CM		 1824 return err;
1825}
1826
1827static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1828 struct dentry *dentry)
1829{
1830 struct btrfs_trans_handle *trans;
1831 struct btrfs_root *root = BTRFS_I(dir)->root;
1832 struct inode *inode = old_dentry->d_inode;
1832a6d5 1833 unsigned long nr = 0;
39279cc3
CM		 1834 int err;
1835 int drop_inode = 0;
1836
1837 if (inode->i_nlink == 0)
1838 return -ENOENT;
1839
6da6abae
CM		 1840#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1841 inode->i_nlink++;
1842#else
39279cc3 1843 inc_nlink(inode);
6da6abae 1844#endif
39279cc3 1845 mutex_lock(&root->fs_info->fs_mutex);
1832a6d5
CM		 1846 err = btrfs_check_free_space(root, 1, 0);
1847 if (err)
1848 goto fail;
39279cc3 1849 trans = btrfs_start_transaction(root, 1);
5f39d397 1850
39279cc3
CM		 1851 btrfs_set_trans_block_group(trans, dir);
1852 atomic_inc(&inode->i_count);
9c58309d 1853 err = btrfs_add_nondir(trans, dentry, inode, 1);
5f39d397 1854
39279cc3
CM		 1855 if (err)
1856 drop_inode = 1;
5f39d397 1857
39279cc3
CM		 1858 dir->i_sb->s_dirt = 1;
1859 btrfs_update_inode_block_group(trans, dir);
54aa1f4d 1860 err = btrfs_update_inode(trans, root, inode);
5f39d397 1861
54aa1f4d
CM		 1862 if (err)
1863 drop_inode = 1;
39279cc3 1864
d3c2fdcf 1865 nr = trans->blocks_used;
39279cc3 1866 btrfs_end_transaction(trans, root);
1832a6d5 1867fail:
39279cc3
CM		 1868 mutex_unlock(&root->fs_info->fs_mutex);
1869
1870 if (drop_inode) {
1871 inode_dec_link_count(inode);
1872 iput(inode);
1873 }
d3c2fdcf 1874 btrfs_btree_balance_dirty(root, nr);
e2008b61 1875 btrfs_throttle(root);
39279cc3
CM		 1876 return err;
1877}
1878
39279cc3
CM		 1879static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1880{
1881 struct inode *inode;
1882 struct btrfs_trans_handle *trans;
1883 struct btrfs_root *root = BTRFS_I(dir)->root;
1884 int err = 0;
1885 int drop_on_err = 0;
1886 u64 objectid;
d3c2fdcf 1887 unsigned long nr = 1;
39279cc3
CM		 1888
1889 mutex_lock(&root->fs_info->fs_mutex);
1832a6d5
CM		 1890 err = btrfs_check_free_space(root, 1, 0);
1891 if (err)
1892 goto out_unlock;
1893
39279cc3
CM		 1894 trans = btrfs_start_transaction(root, 1);
1895 btrfs_set_trans_block_group(trans, dir);
5f39d397 1896
39279cc3
CM		 1897 if (IS_ERR(trans)) {
1898 err = PTR_ERR(trans);
1899 goto out_unlock;
1900 }
1901
1902 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1903 if (err) {
1904 err = -ENOSPC;
1905 goto out_unlock;
1906 }
1907
9c58309d
CM		 1908 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1909 dentry->d_name.len,
1910 dentry->d_parent->d_inode->i_ino, objectid,
39279cc3
CM		 1911 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1912 if (IS_ERR(inode)) {
1913 err = PTR_ERR(inode);
1914 goto out_fail;
1915 }
5f39d397 1916
39279cc3
CM		 1917 drop_on_err = 1;
1918 inode->i_op = &btrfs_dir_inode_operations;
1919 inode->i_fop = &btrfs_dir_file_operations;
1920 btrfs_set_trans_block_group(trans, inode);
1921
3954401f 1922 inode->i_size = 0;
39279cc3
CM		 1923 err = btrfs_update_inode(trans, root, inode);
1924 if (err)
1925 goto out_fail;
5f39d397 1926
9c58309d 1927 err = btrfs_add_link(trans, dentry, inode, 0);
39279cc3
CM		 1928 if (err)
1929 goto out_fail;
5f39d397 1930
39279cc3
CM		 1931 d_instantiate(dentry, inode);
1932 drop_on_err = 0;
1933 dir->i_sb->s_dirt = 1;
1934 btrfs_update_inode_block_group(trans, inode);
1935 btrfs_update_inode_block_group(trans, dir);
1936
1937out_fail:
d3c2fdcf 1938 nr = trans->blocks_used;
39279cc3 1939 btrfs_end_transaction(trans, root);
5f39d397 1940
39279cc3
CM		 1941out_unlock:
1942 mutex_unlock(&root->fs_info->fs_mutex);
1943 if (drop_on_err)
1944 iput(inode);
d3c2fdcf 1945 btrfs_btree_balance_dirty(root, nr);
e2008b61 1946 btrfs_throttle(root);
39279cc3
CM		 1947 return err;
1948}
1949
a52d9a80 1950struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
70dec807 1951 size_t pg_offset, u64 start, u64 len,
a52d9a80
CM		 1952 int create)
1953{
1954 int ret;
1955 int err = 0;
db94535d 1956 u64 bytenr;
a52d9a80
CM		 1957 u64 extent_start = 0;
1958 u64 extent_end = 0;
1959 u64 objectid = inode->i_ino;
1960 u32 found_type;
a52d9a80
CM		 1961 struct btrfs_path *path;
1962 struct btrfs_root *root = BTRFS_I(inode)->root;
1963 struct btrfs_file_extent_item *item;
5f39d397
CM		 1964 struct extent_buffer *leaf;
1965 struct btrfs_key found_key;
a52d9a80
CM		 1966 struct extent_map *em = NULL;
1967 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
d1310b2e 1968 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
a52d9a80
CM		 1969 struct btrfs_trans_handle *trans = NULL;
1970
1971 path = btrfs_alloc_path();
1972 BUG_ON(!path);
1973 mutex_lock(&root->fs_info->fs_mutex);
1974
1975again:
d1310b2e
CM		 1976 spin_lock(&em_tree->lock);
1977 em = lookup_extent_mapping(em_tree, start, len);
1978 spin_unlock(&em_tree->lock);
1979
a52d9a80 1980 if (em) {
56b453c9 1981 if (em->start > start) {
d1310b2e
CM		 1982 printk("get_extent lookup [%Lu %Lu] em [%Lu %Lu]\n",
1983 start, len, em->start, em->len);
56b453c9
CM		 1984 WARN_ON(1);
1985 }
70dec807
CM		 1986 if (em->block_start == EXTENT_MAP_INLINE && page)
1987 free_extent_map(em);
1988 else
1989 goto out;
a52d9a80 1990 }
d1310b2e 1991 em = alloc_extent_map(GFP_NOFS);
a52d9a80 1992 if (!em) {
d1310b2e
CM		 1993 err = -ENOMEM;
1994 goto out;
a52d9a80 1995 }
d1310b2e
CM		 1996
1997 em->start = EXTENT_MAP_HOLE;
1998 em->len = (u64)-1;
a52d9a80 1999 em->bdev = inode->i_sb->s_bdev;
179e29e4
CM		 2000 ret = btrfs_lookup_file_extent(trans, root, path,
2001 objectid, start, trans != NULL);
a52d9a80
CM		 2002 if (ret < 0) {
2003 err = ret;
2004 goto out;
2005 }
2006
2007 if (ret != 0) {
2008 if (path->slots[0] == 0)
2009 goto not_found;
2010 path->slots[0]--;
2011 }
2012
5f39d397
CM		 2013 leaf = path->nodes[0];
2014 item = btrfs_item_ptr(leaf, path->slots[0],
a52d9a80 2015 struct btrfs_file_extent_item);
a52d9a80 2016 /* are we inside the extent that was found? */
5f39d397
CM		 2017 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2018 found_type = btrfs_key_type(&found_key);
2019 if (found_key.objectid != objectid ||
a52d9a80
CM		 2020 found_type != BTRFS_EXTENT_DATA_KEY) {
2021 goto not_found;
2022 }
2023
5f39d397
CM		 2024 found_type = btrfs_file_extent_type(leaf, item);
2025 extent_start = found_key.offset;
a52d9a80
CM		 2026 if (found_type == BTRFS_FILE_EXTENT_REG) {
2027 extent_end = extent_start +
db94535d 2028 btrfs_file_extent_num_bytes(leaf, item);
a52d9a80 2029 err = 0;
b888db2b 2030 if (start < extent_start || start >= extent_end) {
a52d9a80
CM		 2031 em->start = start;
2032 if (start < extent_start) {
d1310b2e 2033 if (start + len <= extent_start)
b888db2b 2034 goto not_found;
d1310b2e 2035 em->len = extent_end - extent_start;
a52d9a80 2036 } else {
d1310b2e 2037 em->len = len;
a52d9a80
CM		 2038 }
2039 goto not_found_em;
2040 }
db94535d
CM		 2041 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
2042 if (bytenr == 0) {
a52d9a80 2043 em->start = extent_start;
d1310b2e 2044 em->len = extent_end - extent_start;
5f39d397 2045 em->block_start = EXTENT_MAP_HOLE;
a52d9a80
CM		 2046 goto insert;
2047 }
db94535d
CM		 2048 bytenr += btrfs_file_extent_offset(leaf, item);
2049 em->block_start = bytenr;
a52d9a80 2050 em->start = extent_start;
d1310b2e 2051 em->len = extent_end - extent_start;
a52d9a80
CM		 2052 goto insert;
2053 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
70dec807 2054 u64 page_start;
5f39d397 2055 unsigned long ptr;
a52d9a80 2056 char *map;
3326d1b0
CM		 2057 size_t size;
2058 size_t extent_offset;
2059 size_t copy_size;
a52d9a80 2060
5f39d397
CM		 2061 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
2062 path->slots[0]));
d1310b2e
CM		 2063 extent_end = (extent_start + size + root->sectorsize - 1) &
2064 ~((u64)root->sectorsize - 1);
b888db2b 2065 if (start < extent_start || start >= extent_end) {
a52d9a80
CM		 2066 em->start = start;
2067 if (start < extent_start) {
d1310b2e 2068 if (start + len <= extent_start)
b888db2b 2069 goto not_found;
d1310b2e 2070 em->len = extent_end - extent_start;
a52d9a80 2071 } else {
d1310b2e 2072 em->len = len;
a52d9a80
CM		 2073 }
2074 goto not_found_em;
2075 }
689f9346 2076 em->block_start = EXTENT_MAP_INLINE;
689f9346
Y		 2077
2078 if (!page) {
2079 em->start = extent_start;
d1310b2e 2080 em->len = size;
689f9346
Y		 2081 goto out;
2082 }
5f39d397 2083
70dec807
CM		 2084 page_start = page_offset(page) + pg_offset;
2085 extent_offset = page_start - extent_start;
2086 copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
3326d1b0 2087 size - extent_offset);
3326d1b0 2088 em->start = extent_start + extent_offset;
70dec807
CM		 2089 em->len = (copy_size + root->sectorsize - 1) &
2090 ~((u64)root->sectorsize - 1);
689f9346
Y		 2091 map = kmap(page);
2092 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
179e29e4 2093 if (create == 0 && !PageUptodate(page)) {
70dec807 2094 read_extent_buffer(leaf, map + pg_offset, ptr,
179e29e4
CM		 2095 copy_size);
2096 flush_dcache_page(page);
2097 } else if (create && PageUptodate(page)) {
2098 if (!trans) {
2099 kunmap(page);
2100 free_extent_map(em);
2101 em = NULL;
2102 btrfs_release_path(root, path);
2103 trans = btrfs_start_transaction(root, 1);
2104 goto again;
2105 }
70dec807 2106 write_extent_buffer(leaf, map + pg_offset, ptr,
179e29e4
CM		 2107 copy_size);
2108 btrfs_mark_buffer_dirty(leaf);
a52d9a80 2109 }
a52d9a80 2110 kunmap(page);
d1310b2e
CM		 2111 set_extent_uptodate(io_tree, em->start,
2112 extent_map_end(em) - 1, GFP_NOFS);
a52d9a80
CM		 2113 goto insert;
2114 } else {
2115 printk("unkknown found_type %d\n", found_type);
2116 WARN_ON(1);
2117 }
2118not_found:
2119 em->start = start;
d1310b2e 2120 em->len = len;
a52d9a80 2121not_found_em:
5f39d397 2122 em->block_start = EXTENT_MAP_HOLE;
a52d9a80
CM		 2123insert:
2124 btrfs_release_path(root, path);
d1310b2e
CM		 2125 if (em->start > start || extent_map_end(em) <= start) {
2126 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len);
a52d9a80
CM		 2127 err = -EIO;
2128 goto out;
2129 }
d1310b2e
CM		 2130
2131 err = 0;
2132 spin_lock(&em_tree->lock);
a52d9a80
CM		 2133 ret = add_extent_mapping(em_tree, em);
2134 if (ret == -EEXIST) {
2135 free_extent_map(em);
d1310b2e
CM		 2136 em = lookup_extent_mapping(em_tree, start, len);
2137 if (!em) {
a52d9a80 2138 err = -EIO;
d1310b2e 2139 printk("failing to insert %Lu %Lu\n", start, len);
a52d9a80 2140 }
a52d9a80 2141 }
d1310b2e 2142 spin_unlock(&em_tree->lock);
a52d9a80
CM		 2143out:
2144 btrfs_free_path(path);
2145 if (trans) {
2146 ret = btrfs_end_transaction(trans, root);
2147 if (!err)
2148 err = ret;
2149 }
2150 mutex_unlock(&root->fs_info->fs_mutex);
2151 if (err) {
2152 free_extent_map(em);
2153 WARN_ON(1);
2154 return ERR_PTR(err);
2155 }
2156 return em;
2157}
2158
d396c6f5 2159static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
39279cc3 2160{
d396c6f5 2161 return extent_bmap(mapping, iblock, btrfs_get_extent);
39279cc3
CM		 2162}
2163
a52d9a80 2164int btrfs_readpage(struct file *file, struct page *page)
9ebefb18 2165{
d1310b2e
CM		 2166 struct extent_io_tree *tree;
2167 tree = &BTRFS_I(page->mapping->host)->io_tree;
a52d9a80 2168 return extent_read_full_page(tree, page, btrfs_get_extent);
9ebefb18 2169}
1832a6d5 2170
a52d9a80 2171static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
39279cc3 2172{
d1310b2e 2173 struct extent_io_tree *tree;
b888db2b
CM		 2174
2175
2176 if (current->flags & PF_MEMALLOC) {
2177 redirty_page_for_writepage(wbc, page);
2178 unlock_page(page);
2179 return 0;
2180 }
d1310b2e 2181 tree = &BTRFS_I(page->mapping->host)->io_tree;
a52d9a80 2182 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
9ebefb18
CM		 2183}
2184
b293f02e
CM		 2185static int btrfs_writepages(struct address_space *mapping,
2186 struct writeback_control *wbc)
2187{
d1310b2e
CM		 2188 struct extent_io_tree *tree;
2189 tree = &BTRFS_I(mapping->host)->io_tree;
b293f02e
CM		 2190 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
2191}
2192
3ab2fb5a
CM		 2193static int
2194btrfs_readpages(struct file *file, struct address_space *mapping,
2195 struct list_head *pages, unsigned nr_pages)
2196{
d1310b2e
CM		 2197 struct extent_io_tree *tree;
2198 tree = &BTRFS_I(mapping->host)->io_tree;
3ab2fb5a
CM		 2199 return extent_readpages(tree, mapping, pages, nr_pages,
2200 btrfs_get_extent);
2201}
2202
70dec807 2203static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
9ebefb18 2204{
d1310b2e
CM		 2205 struct extent_io_tree *tree;
2206 struct extent_map_tree *map;
a52d9a80 2207 int ret;
8c2383c3 2208
d1310b2e
CM		 2209 tree = &BTRFS_I(page->mapping->host)->io_tree;
2210 map = &BTRFS_I(page->mapping->host)->extent_tree;
70dec807 2211 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
a52d9a80
CM		 2212 if (ret == 1) {
2213 ClearPagePrivate(page);
2214 set_page_private(page, 0);
2215 page_cache_release(page);
39279cc3 2216 }
a52d9a80 2217 return ret;
39279cc3
CM		 2218}
2219
a52d9a80 2220static void btrfs_invalidatepage(struct page *page, unsigned long offset)
39279cc3 2221{
d1310b2e 2222 struct extent_io_tree *tree;
39279cc3 2223
d1310b2e 2224 tree = &BTRFS_I(page->mapping->host)->io_tree;
a52d9a80
CM		 2225 extent_invalidatepage(tree, page, offset);
2226 btrfs_releasepage(page, GFP_NOFS);
39279cc3
CM		 2227}
2228
9ebefb18
CM		 2229/*
2230 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2231 * called from a page fault handler when a page is first dirtied. Hence we must
2232 * be careful to check for EOF conditions here. We set the page up correctly
2233 * for a written page which means we get ENOSPC checking when writing into
2234 * holes and correct delalloc and unwritten extent mapping on filesystems that
2235 * support these features.
2236 *
2237 * We are not allowed to take the i_mutex here so we have to play games to
2238 * protect against truncate races as the page could now be beyond EOF. Because
2239 * vmtruncate() writes the inode size before removing pages, once we have the
2240 * page lock we can determine safely if the page is beyond EOF. If it is not
2241 * beyond EOF, then the page is guaranteed safe against truncation until we
2242 * unlock the page.
2243 */
2244int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
2245{
6da6abae 2246 struct inode *inode = fdentry(vma->vm_file)->d_inode;
1832a6d5 2247 struct btrfs_root *root = BTRFS_I(inode)->root;
9ebefb18
CM		 2248 unsigned long end;
2249 loff_t size;
1832a6d5 2250 int ret;
a52d9a80 2251 u64 page_start;
9ebefb18 2252
1832a6d5
CM		 2253 mutex_lock(&root->fs_info->fs_mutex);
2254 ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
8f662a76 2255 mutex_unlock(&root->fs_info->fs_mutex);
1832a6d5
CM		 2256 if (ret)
2257 goto out;
2258
2259 ret = -EINVAL;
2260
9ebefb18
CM		 2261 lock_page(page);
2262 wait_on_page_writeback(page);
9ebefb18 2263 size = i_size_read(inode);
35ebb934 2264 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
a52d9a80 2265
9ebefb18 2266 if ((page->mapping != inode->i_mapping) ||
a52d9a80 2267 (page_start > size)) {
9ebefb18
CM		 2268 /* page got truncated out from underneath us */
2269 goto out_unlock;
2270 }
2271
2272 /* page is wholly or partially inside EOF */
a52d9a80 2273 if (page_start + PAGE_CACHE_SIZE > size)
9ebefb18
CM		 2274 end = size & ~PAGE_CACHE_MASK;
2275 else
2276 end = PAGE_CACHE_SIZE;
2277
b888db2b 2278 ret = btrfs_cow_one_page(inode, page, end);
9ebefb18
CM		 2279
2280out_unlock:
2281 unlock_page(page);
1832a6d5 2282out:
9ebefb18
CM		 2283 return ret;
2284}
2285
39279cc3
CM		 2286static void btrfs_truncate(struct inode *inode)
2287{
2288 struct btrfs_root *root = BTRFS_I(inode)->root;
2289 int ret;
2290 struct btrfs_trans_handle *trans;
d3c2fdcf 2291 unsigned long nr;
39279cc3
CM		 2292
2293 if (!S_ISREG(inode->i_mode))
2294 return;
2295 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
2296 return;
2297
2298 btrfs_truncate_page(inode->i_mapping, inode->i_size);
2299
2300 mutex_lock(&root->fs_info->fs_mutex);
2301 trans = btrfs_start_transaction(root, 1);
2302 btrfs_set_trans_block_group(trans, inode);
2303
2304 /* FIXME, add redo link to tree so we don't leak on crash */
85e21bac
CM		 2305 ret = btrfs_truncate_in_trans(trans, root, inode,
2306 BTRFS_EXTENT_DATA_KEY);
39279cc3 2307 btrfs_update_inode(trans, root, inode);
d3c2fdcf 2308 nr = trans->blocks_used;
5f39d397 2309
39279cc3
CM		 2310 ret = btrfs_end_transaction(trans, root);
2311 BUG_ON(ret);
2312 mutex_unlock(&root->fs_info->fs_mutex);
d3c2fdcf 2313 btrfs_btree_balance_dirty(root, nr);
e2008b61 2314 btrfs_throttle(root);
39279cc3
CM		 2315}
2316
4313b399
CM		 2317static int noinline create_subvol(struct btrfs_root *root, char *name,
2318 int namelen)
39279cc3
CM		 2319{
2320 struct btrfs_trans_handle *trans;
2321 struct btrfs_key key;
2322 struct btrfs_root_item root_item;
2323 struct btrfs_inode_item *inode_item;
5f39d397 2324 struct extent_buffer *leaf;
dc17ff8f 2325 struct btrfs_root *new_root = root;
39279cc3
CM		 2326 struct inode *inode;
2327 struct inode *dir;
2328 int ret;
54aa1f4d 2329 int err;
39279cc3
CM		 2330 u64 objectid;
2331 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
d3c2fdcf 2332 unsigned long nr = 1;
39279cc3
CM		 2333
2334 mutex_lock(&root->fs_info->fs_mutex);
1832a6d5
CM		 2335 ret = btrfs_check_free_space(root, 1, 0);
2336 if (ret)
2337 goto fail_commit;
2338
39279cc3
CM		 2339 trans = btrfs_start_transaction(root, 1);
2340 BUG_ON(!trans);
2341
7bb86316
CM		 2342 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
2343 0, &objectid);
2344 if (ret)
2345 goto fail;
2346
2347 leaf = __btrfs_alloc_free_block(trans, root, root->leafsize,
2348 objectid, trans->transid, 0, 0,
2349 0, 0);
5f39d397
CM		 2350 if (IS_ERR(leaf))
2351 return PTR_ERR(leaf);
2352
2353 btrfs_set_header_nritems(leaf, 0);
2354 btrfs_set_header_level(leaf, 0);
db94535d 2355 btrfs_set_header_bytenr(leaf, leaf->start);
5f39d397 2356 btrfs_set_header_generation(leaf, trans->transid);
7bb86316
CM		 2357 btrfs_set_header_owner(leaf, objectid);
2358
5f39d397
CM		 2359 write_extent_buffer(leaf, root->fs_info->fsid,
2360 (unsigned long)btrfs_header_fsid(leaf),
2361 BTRFS_FSID_SIZE);
2362 btrfs_mark_buffer_dirty(leaf);
39279cc3
CM		 2363
2364 inode_item = &root_item.inode;
2365 memset(inode_item, 0, sizeof(*inode_item));
5f39d397
CM		 2366 inode_item->generation = cpu_to_le64(1);
2367 inode_item->size = cpu_to_le64(3);
2368 inode_item->nlink = cpu_to_le32(1);
2369 inode_item->nblocks = cpu_to_le64(1);
2370 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
39279cc3 2371
db94535d
CM		 2372 btrfs_set_root_bytenr(&root_item, leaf->start);
2373 btrfs_set_root_level(&root_item, 0);
39279cc3 2374 btrfs_set_root_refs(&root_item, 1);
5f39d397
CM		 2375 btrfs_set_root_used(&root_item, 0);
2376
5eda7b5e
CM		 2377 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
2378 root_item.drop_level = 0;
5f39d397
CM		 2379
2380 free_extent_buffer(leaf);
2381 leaf = NULL;
39279cc3 2382
39279cc3
CM		 2383 btrfs_set_root_dirid(&root_item, new_dirid);
2384
2385 key.objectid = objectid;
2386 key.offset = 1;
39279cc3
CM		 2387 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
2388 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
2389 &root_item);
54aa1f4d
CM		 2390 if (ret)
2391 goto fail;
39279cc3
CM		 2392
2393 /*
2394 * insert the directory item
2395 */
2396 key.offset = (u64)-1;
2397 dir = root->fs_info->sb->s_root->d_inode;
2398 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
2399 name, namelen, dir->i_ino, &key,
2400 BTRFS_FT_DIR);
54aa1f4d
CM		 2401 if (ret)
2402 goto fail;
39279cc3 2403
3954401f
CM		 2404 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
2405 name, namelen, objectid,
2406 root->fs_info->sb->s_root->d_inode->i_ino);
2407 if (ret)
2408 goto fail;
2409
39279cc3 2410 ret = btrfs_commit_transaction(trans, root);
54aa1f4d
CM		 2411 if (ret)
2412 goto fail_commit;
39279cc3 2413
58176a96 2414 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
39279cc3
CM		 2415 BUG_ON(!new_root);
2416
2417 trans = btrfs_start_transaction(new_root, 1);
2418 BUG_ON(!trans);
2419
9c58309d
CM		 2420 inode = btrfs_new_inode(trans, new_root, "..", 2, new_dirid,
2421 new_dirid,
39279cc3 2422 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
54aa1f4d
CM		 2423 if (IS_ERR(inode))
2424 goto fail;
39279cc3
CM		 2425 inode->i_op = &btrfs_dir_inode_operations;
2426 inode->i_fop = &btrfs_dir_file_operations;
34088780 2427 new_root->inode = inode;
39279cc3 2428
3954401f
CM		 2429 ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid,
2430 new_dirid);
39279cc3 2431 inode->i_nlink = 1;
3954401f 2432 inode->i_size = 0;
39279cc3 2433 ret = btrfs_update_inode(trans, new_root, inode);
54aa1f4d
CM		 2434 if (ret)
2435 goto fail;
2436fail:
d3c2fdcf 2437 nr = trans->blocks_used;
dc17ff8f 2438 err = btrfs_commit_transaction(trans, new_root);
54aa1f4d
CM		 2439 if (err && !ret)
2440 ret = err;
2441fail_commit:
39279cc3 2442 mutex_unlock(&root->fs_info->fs_mutex);
d3c2fdcf 2443 btrfs_btree_balance_dirty(root, nr);
e2008b61 2444 btrfs_throttle(root);
54aa1f4d 2445 return ret;
39279cc3
CM		 2446}
2447
2448static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
2449{
3063d29f 2450 struct btrfs_pending_snapshot *pending_snapshot;
39279cc3 2451 struct btrfs_trans_handle *trans;
39279cc3 2452 int ret;
54aa1f4d 2453 int err;
1832a6d5 2454 unsigned long nr = 0;
39279cc3
CM		 2455
2456 if (!root->ref_cows)
2457 return -EINVAL;
2458
2459 mutex_lock(&root->fs_info->fs_mutex);
1832a6d5
CM		 2460 ret = btrfs_check_free_space(root, 1, 0);
2461 if (ret)
2462 goto fail_unlock;
2463
3063d29f
CM		 2464 pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS);
2465 if (!pending_snapshot) {
2466 ret = -ENOMEM;
2467 goto fail_unlock;
2468 }
fb4bc1e0 2469 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
3063d29f
CM		 2470 if (!pending_snapshot->name) {
2471 ret = -ENOMEM;
2472 kfree(pending_snapshot);
2473 goto fail_unlock;
2474 }
fb4bc1e0
Y		 2475 memcpy(pending_snapshot->name, name, namelen);
2476 pending_snapshot->name[namelen] = '\0';
39279cc3
CM		 2477 trans = btrfs_start_transaction(root, 1);
2478 BUG_ON(!trans);
3063d29f
CM		 2479 pending_snapshot->root = root;
2480 list_add(&pending_snapshot->list,
2481 &trans->transaction->pending_snapshots);
39279cc3 2482 ret = btrfs_update_inode(trans, root, root->inode);
54aa1f4d 2483 err = btrfs_commit_transaction(trans, root);
5f39d397 2484
1832a6d5 2485fail_unlock:
39279cc3 2486 mutex_unlock(&root->fs_info->fs_mutex);
d3c2fdcf 2487 btrfs_btree_balance_dirty(root, nr);
e2008b61 2488 btrfs_throttle(root);
54aa1f4d 2489 return ret;
39279cc3
CM		 2490}
2491
edbd8d4e 2492unsigned long btrfs_force_ra(struct address_space *mapping,
86479a04
CM		 2493 struct file_ra_state *ra, struct file *file,
2494 pgoff_t offset, pgoff_t last_index)
2495{
2496 pgoff_t req_size;
2497
2498#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2499 req_size = last_index - offset + 1;
2500 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2501 return offset;
2502#else
2503 req_size = min(last_index - offset + 1, (pgoff_t)128);
2504 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2505 return offset + req_size;
2506#endif
2507}
2508
2509int btrfs_defrag_file(struct file *file) {
6da6abae 2510 struct inode *inode = fdentry(file)->d_inode;
1832a6d5 2511 struct btrfs_root *root = BTRFS_I(inode)->root;
d1310b2e 2512 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
86479a04
CM		 2513 struct page *page;
2514 unsigned long last_index;
2515 unsigned long ra_index = 0;
2516 u64 page_start;
2517 u64 page_end;
2518 unsigned long i;
1832a6d5
CM		 2519 int ret;
2520
2521 mutex_lock(&root->fs_info->fs_mutex);
2522 ret = btrfs_check_free_space(root, inode->i_size, 0);
2523 mutex_unlock(&root->fs_info->fs_mutex);
2524 if (ret)
2525 return -ENOSPC;
86479a04
CM		 2526
2527 mutex_lock(&inode->i_mutex);
2528 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
2529 for (i = 0; i <= last_index; i++) {
2530 if (i == ra_index) {
edbd8d4e
CM		 2531 ra_index = btrfs_force_ra(inode->i_mapping,
2532 &file->f_ra,
2533 file, ra_index, last_index);
86479a04
CM		 2534 }
2535 page = grab_cache_page(inode->i_mapping, i);
2536 if (!page)
2537 goto out_unlock;
2538 if (!PageUptodate(page)) {
2539 btrfs_readpage(NULL, page);
2540 lock_page(page);
2541 if (!PageUptodate(page)) {
2542 unlock_page(page);
2543 page_cache_release(page);
2544 goto out_unlock;
2545 }
2546 }
35ebb934 2547 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
86479a04
CM		 2548 page_end = page_start + PAGE_CACHE_SIZE - 1;
2549
d1310b2e 2550 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
d1310b2e 2551 set_extent_delalloc(io_tree, page_start,
86479a04 2552 page_end, GFP_NOFS);
edbd8d4e 2553
d1310b2e 2554 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
86479a04
CM		 2555 set_page_dirty(page);
2556 unlock_page(page);
2557 page_cache_release(page);
2558 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
2559 }
2560
2561out_unlock:
2562 mutex_unlock(&inode->i_mutex);
2563 return 0;
2564}
2565
edbd8d4e
CM		 2566static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
2567{
2568 u64 new_size;
2569 u64 old_size;
2570 struct btrfs_ioctl_vol_args *vol_args;
2571 struct btrfs_trans_handle *trans;
2572 char *sizestr;
2573 int ret = 0;
2574 int namelen;
2575 int mod = 0;
2576
2577 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
2578
2579 if (!vol_args)
2580 return -ENOMEM;
2581
2582 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
2583 ret = -EFAULT;
2584 goto out;
2585 }
2586 namelen = strlen(vol_args->name);
2587 if (namelen > BTRFS_VOL_NAME_MAX) {
2588 ret = -EINVAL;
2589 goto out;
2590 }
2591
2592 sizestr = vol_args->name;
2593 if (!strcmp(sizestr, "max"))
2594 new_size = root->fs_info->sb->s_bdev->bd_inode->i_size;
2595 else {
2596 if (sizestr[0] == '-') {
2597 mod = -1;
2598 sizestr++;
2599 } else if (sizestr[0] == '+') {
2600 mod = 1;
2601 sizestr++;
2602 }
2603 new_size = btrfs_parse_size(sizestr);
2604 if (new_size == 0) {
2605 ret = -EINVAL;
2606 goto out;
2607 }
2608 }
2609
2610 mutex_lock(&root->fs_info->fs_mutex);
2611 old_size = btrfs_super_total_bytes(&root->fs_info->super_copy);
2612
2613 if (mod < 0) {
2614 if (new_size > old_size) {
2615 ret = -EINVAL;
2616 goto out_unlock;
2617 }
2618 new_size = old_size - new_size;
2619 } else if (mod > 0) {
2620 new_size = old_size + new_size;
2621 }
2622
2623 if (new_size < 256 * 1024 * 1024) {
2624 ret = -EINVAL;
2625 goto out_unlock;
2626 }
2627 if (new_size > root->fs_info->sb->s_bdev->bd_inode->i_size) {
2628 ret = -EFBIG;
2629 goto out_unlock;
2630 }
f9ef6604
CM		 2631
2632 do_div(new_size, root->sectorsize);
2633 new_size *= root->sectorsize;
edbd8d4e
CM		 2634
2635printk("new size is %Lu\n", new_size);
2636 if (new_size > old_size) {
2637 trans = btrfs_start_transaction(root, 1);
2638 ret = btrfs_grow_extent_tree(trans, root, new_size);
2639 btrfs_commit_transaction(trans, root);
2640 } else {
2641 ret = btrfs_shrink_extent_tree(root, new_size);
2642 }
2643
2644out_unlock:
2645 mutex_unlock(&root->fs_info->fs_mutex);
2646out:
2647 kfree(vol_args);
2648 return ret;
2649}
2650
4313b399
CM		 2651static int noinline btrfs_ioctl_snap_create(struct btrfs_root *root,
2652 void __user *arg)
39279cc3 2653{
4aec2b52 2654 struct btrfs_ioctl_vol_args *vol_args;
39279cc3 2655 struct btrfs_dir_item *di;
39279cc3
CM		 2656 struct btrfs_path *path;
2657 u64 root_dirid;
4aec2b52
CM		 2658 int namelen;
2659 int ret;
39279cc3 2660
4aec2b52 2661 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
5f39d397 2662
4aec2b52
CM		 2663 if (!vol_args)
2664 return -ENOMEM;
2665
2666 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
2667 ret = -EFAULT;
2668 goto out;
2669 }
2670
2671 namelen = strlen(vol_args->name);
2672 if (namelen > BTRFS_VOL_NAME_MAX) {
2673 ret = -EINVAL;
2674 goto out;
2675 }
2676 if (strchr(vol_args->name, '/')) {
2677 ret = -EINVAL;
2678 goto out;
2679 }
d03581f4
CH		 2680
2681 path = btrfs_alloc_path();
4aec2b52
CM		 2682 if (!path) {
2683 ret = -ENOMEM;
2684 goto out;
2685 }
d03581f4
CH		 2686
2687 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2688 mutex_lock(&root->fs_info->fs_mutex);
2689 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2690 path, root_dirid,
4aec2b52 2691 vol_args->name, namelen, 0);
d03581f4
CH		 2692 mutex_unlock(&root->fs_info->fs_mutex);
2693 btrfs_free_path(path);
4aec2b52
CM		 2694
2695 if (di && !IS_ERR(di)) {
2696 ret = -EEXIST;
2697 goto out;
2698 }
2699
2700 if (IS_ERR(di)) {
2701 ret = PTR_ERR(di);
2702 goto out;
2703 }
d03581f4
CH		 2704
2705 if (root == root->fs_info->tree_root)
4aec2b52
CM		 2706 ret = create_subvol(root, vol_args->name, namelen);
2707 else
2708 ret = create_snapshot(root, vol_args->name, namelen);
2709out:
2710 kfree(vol_args);
2711 return ret;
d03581f4
CH		 2712}
2713
2714static int btrfs_ioctl_defrag(struct file *file)
2715{
6da6abae 2716 struct inode *inode = fdentry(file)->d_inode;
d03581f4
CH		 2717 struct btrfs_root *root = BTRFS_I(inode)->root;
2718
2719 switch (inode->i_mode & S_IFMT) {
2720 case S_IFDIR:
39279cc3 2721 mutex_lock(&root->fs_info->fs_mutex);
d03581f4
CH		 2722 btrfs_defrag_root(root, 0);
2723 btrfs_defrag_root(root->fs_info->extent_root, 0);
39279cc3 2724 mutex_unlock(&root->fs_info->fs_mutex);
39279cc3 2725 break;
d03581f4
CH		 2726 case S_IFREG:
2727 btrfs_defrag_file(file);
2728 break;
2729 }
2730
2731 return 0;
2732}
6702ed49 2733
d03581f4
CH		 2734long btrfs_ioctl(struct file *file, unsigned int
2735 cmd, unsigned long arg)
2736{
6da6abae 2737 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
d03581f4
CH		 2738
2739 switch (cmd) {
2740 case BTRFS_IOC_SNAP_CREATE:
2741 return btrfs_ioctl_snap_create(root, (void __user *)arg);
6702ed49 2742 case BTRFS_IOC_DEFRAG:
d03581f4 2743 return btrfs_ioctl_defrag(file);
edbd8d4e
CM		 2744 case BTRFS_IOC_RESIZE:
2745 return btrfs_ioctl_resize(root, (void __user *)arg);
39279cc3 2746 }
d03581f4
CH		 2747
2748 return -ENOTTY;
39279cc3
CM		 2749}
2750
39279cc3
CM		 2751/*
2752 * Called inside transaction, so use GFP_NOFS
2753 */
2754struct inode *btrfs_alloc_inode(struct super_block *sb)
2755{
2756 struct btrfs_inode *ei;
2757
2758 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2759 if (!ei)
2760 return NULL;
15ee9bc7 2761 ei->last_trans = 0;
dc17ff8f 2762 ei->ordered_trans = 0;
39279cc3
CM		 2763 return &ei->vfs_inode;
2764}
2765
2766void btrfs_destroy_inode(struct inode *inode)
2767{
2768 WARN_ON(!list_empty(&inode->i_dentry));
2769 WARN_ON(inode->i_data.nrpages);
2770
8c416c9e 2771 btrfs_drop_extent_cache(inode, 0, (u64)-1);
39279cc3
CM		 2772 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2773}
2774
44ec0b71
CM		 2775#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2776static void init_once(struct kmem_cache * cachep, void *foo)
2777#else
39279cc3
CM		 2778static void init_once(void * foo, struct kmem_cache * cachep,
2779 unsigned long flags)
44ec0b71 2780#endif
39279cc3
CM		 2781{
2782 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2783
2784 inode_init_once(&ei->vfs_inode);
2785}
2786
2787void btrfs_destroy_cachep(void)
2788{
2789 if (btrfs_inode_cachep)
2790 kmem_cache_destroy(btrfs_inode_cachep);
2791 if (btrfs_trans_handle_cachep)
2792 kmem_cache_destroy(btrfs_trans_handle_cachep);
2793 if (btrfs_transaction_cachep)
2794 kmem_cache_destroy(btrfs_transaction_cachep);
2795 if (btrfs_bit_radix_cachep)
2796 kmem_cache_destroy(btrfs_bit_radix_cachep);
2797 if (btrfs_path_cachep)
2798 kmem_cache_destroy(btrfs_path_cachep);
2799}
2800
86479a04 2801struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
92fee66d 2802 unsigned long extra_flags,
44ec0b71
CM		 2803#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2804 void (*ctor)(struct kmem_cache *, void *)
2805#else
92fee66d 2806 void (*ctor)(void *, struct kmem_cache *,
44ec0b71
CM		 2807 unsigned long)
2808#endif
2809 )
92fee66d
CM		 2810{
2811 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2812 SLAB_MEM_SPREAD | extra_flags), ctor
2813#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2814 ,NULL
2815#endif
2816 );
2817}
2818
39279cc3
CM		 2819int btrfs_init_cachep(void)
2820{
86479a04 2821 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
92fee66d
CM		 2822 sizeof(struct btrfs_inode),
2823 0, init_once);
39279cc3
CM		 2824 if (!btrfs_inode_cachep)
2825 goto fail;
86479a04
CM		 2826 btrfs_trans_handle_cachep =
2827 btrfs_cache_create("btrfs_trans_handle_cache",
2828 sizeof(struct btrfs_trans_handle),
2829 0, NULL);
39279cc3
CM		 2830 if (!btrfs_trans_handle_cachep)
2831 goto fail;
86479a04 2832 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
39279cc3 2833 sizeof(struct btrfs_transaction),
92fee66d 2834 0, NULL);
39279cc3
CM		 2835 if (!btrfs_transaction_cachep)
2836 goto fail;
86479a04 2837 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
23223584 2838 sizeof(struct btrfs_path),
92fee66d 2839 0, NULL);
39279cc3
CM		 2840 if (!btrfs_path_cachep)
2841 goto fail;
86479a04 2842 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
92fee66d 2843 SLAB_DESTROY_BY_RCU, NULL);
39279cc3
CM		 2844 if (!btrfs_bit_radix_cachep)
2845 goto fail;
2846 return 0;
2847fail:
2848 btrfs_destroy_cachep();
2849 return -ENOMEM;
2850}
2851
2852static int btrfs_getattr(struct vfsmount *mnt,
2853 struct dentry *dentry, struct kstat *stat)
2854{
2855 struct inode *inode = dentry->d_inode;
2856 generic_fillattr(inode, stat);
d6667462 2857 stat->blksize = PAGE_CACHE_SIZE;
9069218d 2858 stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9);
39279cc3
CM		 2859 return 0;
2860}
2861
2862static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2863 struct inode * new_dir,struct dentry *new_dentry)
2864{
2865 struct btrfs_trans_handle *trans;
2866 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2867 struct inode *new_inode = new_dentry->d_inode;
2868 struct inode *old_inode = old_dentry->d_inode;
2869 struct timespec ctime = CURRENT_TIME;
2870 struct btrfs_path *path;
39279cc3
CM		 2871 int ret;
2872
2873 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2874 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2875 return -ENOTEMPTY;
2876 }
5f39d397 2877
39279cc3 2878 mutex_lock(&root->fs_info->fs_mutex);
1832a6d5
CM		 2879 ret = btrfs_check_free_space(root, 1, 0);
2880 if (ret)
2881 goto out_unlock;
2882
39279cc3 2883 trans = btrfs_start_transaction(root, 1);
5f39d397 2884
39279cc3
CM		 2885 btrfs_set_trans_block_group(trans, new_dir);
2886 path = btrfs_alloc_path();
2887 if (!path) {
2888 ret = -ENOMEM;
2889 goto out_fail;
2890 }
2891
2892 old_dentry->d_inode->i_nlink++;
2893 old_dir->i_ctime = old_dir->i_mtime = ctime;
2894 new_dir->i_ctime = new_dir->i_mtime = ctime;
2895 old_inode->i_ctime = ctime;
5f39d397 2896
39279cc3
CM		 2897 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2898 if (ret)
2899 goto out_fail;
2900
2901 if (new_inode) {
2902 new_inode->i_ctime = CURRENT_TIME;
2903 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2904 if (ret)
2905 goto out_fail;
39279cc3 2906 }
9c58309d 2907 ret = btrfs_add_link(trans, new_dentry, old_inode, 1);
39279cc3
CM		 2908 if (ret)
2909 goto out_fail;
2910
2911out_fail:
2912 btrfs_free_path(path);
2913 btrfs_end_transaction(trans, root);
1832a6d5 2914out_unlock:
39279cc3
CM		 2915 mutex_unlock(&root->fs_info->fs_mutex);
2916 return ret;
2917}
2918
2919static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2920 const char *symname)
2921{
2922 struct btrfs_trans_handle *trans;
2923 struct btrfs_root *root = BTRFS_I(dir)->root;
2924 struct btrfs_path *path;
2925 struct btrfs_key key;
1832a6d5 2926 struct inode *inode = NULL;
39279cc3
CM		 2927 int err;
2928 int drop_inode = 0;
2929 u64 objectid;
2930 int name_len;
2931 int datasize;
5f39d397 2932 unsigned long ptr;
39279cc3 2933 struct btrfs_file_extent_item *ei;
5f39d397 2934 struct extent_buffer *leaf;
1832a6d5 2935 unsigned long nr = 0;
39279cc3
CM		 2936
2937 name_len = strlen(symname) + 1;
2938 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2939 return -ENAMETOOLONG;
1832a6d5 2940
39279cc3 2941 mutex_lock(&root->fs_info->fs_mutex);
1832a6d5
CM		 2942 err = btrfs_check_free_space(root, 1, 0);
2943 if (err)
2944 goto out_fail;
2945
39279cc3
CM		 2946 trans = btrfs_start_transaction(root, 1);
2947 btrfs_set_trans_block_group(trans, dir);
2948
2949 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2950 if (err) {
2951 err = -ENOSPC;
2952 goto out_unlock;
2953 }
2954
9c58309d
CM		 2955 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2956 dentry->d_name.len,
2957 dentry->d_parent->d_inode->i_ino, objectid,
39279cc3
CM		 2958 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2959 err = PTR_ERR(inode);
2960 if (IS_ERR(inode))
2961 goto out_unlock;
2962
2963 btrfs_set_trans_block_group(trans, inode);
9c58309d 2964 err = btrfs_add_nondir(trans, dentry, inode, 0);
39279cc3
CM		 2965 if (err)
2966 drop_inode = 1;
2967 else {
2968 inode->i_mapping->a_ops = &btrfs_aops;
04160088 2969 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
39279cc3
CM		 2970 inode->i_fop = &btrfs_file_operations;
2971 inode->i_op = &btrfs_file_inode_operations;
d1310b2e
CM		 2972 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2973 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
a52d9a80 2974 inode->i_mapping, GFP_NOFS);
9069218d 2975 BTRFS_I(inode)->delalloc_bytes = 0;
d1310b2e 2976 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
39279cc3
CM		 2977 }
2978 dir->i_sb->s_dirt = 1;
2979 btrfs_update_inode_block_group(trans, inode);
2980 btrfs_update_inode_block_group(trans, dir);
2981 if (drop_inode)
2982 goto out_unlock;
2983
2984 path = btrfs_alloc_path();
2985 BUG_ON(!path);
2986 key.objectid = inode->i_ino;
2987 key.offset = 0;
39279cc3
CM		 2988 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2989 datasize = btrfs_file_extent_calc_inline_size(name_len);
2990 err = btrfs_insert_empty_item(trans, root, path, &key,
2991 datasize);
54aa1f4d
CM		 2992 if (err) {
2993 drop_inode = 1;
2994 goto out_unlock;
2995 }
5f39d397
CM		 2996 leaf = path->nodes[0];
2997 ei = btrfs_item_ptr(leaf, path->slots[0],
2998 struct btrfs_file_extent_item);
2999 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
3000 btrfs_set_file_extent_type(leaf, ei,
39279cc3
CM		 3001 BTRFS_FILE_EXTENT_INLINE);
3002 ptr = btrfs_file_extent_inline_start(ei);
5f39d397
CM		 3003 write_extent_buffer(leaf, symname, ptr, name_len);
3004 btrfs_mark_buffer_dirty(leaf);
39279cc3 3005 btrfs_free_path(path);
5f39d397 3006
39279cc3
CM		 3007 inode->i_op = &btrfs_symlink_inode_operations;
3008 inode->i_mapping->a_ops = &btrfs_symlink_aops;
04160088 3009 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
39279cc3 3010 inode->i_size = name_len - 1;
54aa1f4d
CM		 3011 err = btrfs_update_inode(trans, root, inode);
3012 if (err)
3013 drop_inode = 1;
39279cc3
CM		 3014
3015out_unlock:
d3c2fdcf 3016 nr = trans->blocks_used;
39279cc3 3017 btrfs_end_transaction(trans, root);
1832a6d5 3018out_fail:
39279cc3 3019 mutex_unlock(&root->fs_info->fs_mutex);
39279cc3
CM		 3020 if (drop_inode) {
3021 inode_dec_link_count(inode);
3022 iput(inode);
3023 }
d3c2fdcf 3024 btrfs_btree_balance_dirty(root, nr);
e2008b61 3025 btrfs_throttle(root);
39279cc3
CM		 3026 return err;
3027}
fdebe2bd
Y		 3028static int btrfs_permission(struct inode *inode, int mask,
3029 struct nameidata *nd)
3030{
3031 if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
3032 return -EACCES;
3033 return generic_permission(inode, mask, NULL);
3034}
39279cc3
CM		 3035
3036static struct inode_operations btrfs_dir_inode_operations = {
3037 .lookup = btrfs_lookup,
3038 .create = btrfs_create,
3039 .unlink = btrfs_unlink,
3040 .link = btrfs_link,
3041 .mkdir = btrfs_mkdir,
3042 .rmdir = btrfs_rmdir,
3043 .rename = btrfs_rename,
3044 .symlink = btrfs_symlink,
3045 .setattr = btrfs_setattr,
618e21d5 3046 .mknod = btrfs_mknod,
5103e947
JB		 3047 .setxattr = generic_setxattr,
3048 .getxattr = generic_getxattr,
3049 .listxattr = btrfs_listxattr,
3050 .removexattr = generic_removexattr,
fdebe2bd 3051 .permission = btrfs_permission,
39279cc3 3052};
39279cc3
CM		 3053static struct inode_operations btrfs_dir_ro_inode_operations = {
3054 .lookup = btrfs_lookup,
fdebe2bd 3055 .permission = btrfs_permission,
39279cc3 3056};
39279cc3
CM		 3057static struct file_operations btrfs_dir_file_operations = {
3058 .llseek = generic_file_llseek,
3059 .read = generic_read_dir,
3060 .readdir = btrfs_readdir,
34287aa3 3061 .unlocked_ioctl = btrfs_ioctl,
39279cc3 3062#ifdef CONFIG_COMPAT
34287aa3 3063 .compat_ioctl = btrfs_ioctl,
39279cc3
CM		 3064#endif
3065};
3066
d1310b2e 3067static struct extent_io_ops btrfs_extent_io_ops = {
07157aac 3068 .fill_delalloc = run_delalloc_range,
065631f6 3069 .submit_bio_hook = btrfs_submit_bio_hook,
239b14b3 3070 .merge_bio_hook = btrfs_merge_bio_hook,
07157aac
CM		 3071 .readpage_io_hook = btrfs_readpage_io_hook,
3072 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
b0c68f8b
CM		 3073 .set_bit_hook = btrfs_set_bit_hook,
3074 .clear_bit_hook = btrfs_clear_bit_hook,
07157aac
CM		 3075};
3076
39279cc3
CM		 3077static struct address_space_operations btrfs_aops = {
3078 .readpage = btrfs_readpage,
3079 .writepage = btrfs_writepage,
b293f02e 3080 .writepages = btrfs_writepages,
3ab2fb5a 3081 .readpages = btrfs_readpages,
39279cc3 3082 .sync_page = block_sync_page,
39279cc3 3083 .bmap = btrfs_bmap,
a52d9a80
CM		 3084 .invalidatepage = btrfs_invalidatepage,
3085 .releasepage = btrfs_releasepage,
3086 .set_page_dirty = __set_page_dirty_nobuffers,
39279cc3
CM		 3087};
3088
3089static struct address_space_operations btrfs_symlink_aops = {
3090 .readpage = btrfs_readpage,
3091 .writepage = btrfs_writepage,
2bf5a725
CM		 3092 .invalidatepage = btrfs_invalidatepage,
3093 .releasepage = btrfs_releasepage,
39279cc3
CM		 3094};
3095
3096static struct inode_operations btrfs_file_inode_operations = {
3097 .truncate = btrfs_truncate,
3098 .getattr = btrfs_getattr,
3099 .setattr = btrfs_setattr,
5103e947
JB		 3100 .setxattr = generic_setxattr,
3101 .getxattr = generic_getxattr,
3102 .listxattr = btrfs_listxattr,
3103 .removexattr = generic_removexattr,
fdebe2bd 3104 .permission = btrfs_permission,
39279cc3 3105};
618e21d5
JB		 3106static struct inode_operations btrfs_special_inode_operations = {
3107 .getattr = btrfs_getattr,
3108 .setattr = btrfs_setattr,
fdebe2bd 3109 .permission = btrfs_permission,
618e21d5 3110};
39279cc3
CM		 3111static struct inode_operations btrfs_symlink_inode_operations = {
3112 .readlink = generic_readlink,
3113 .follow_link = page_follow_link_light,
3114 .put_link = page_put_link,
fdebe2bd 3115 .permission = btrfs_permission,
39279cc3 3116};
Clone of davem's net-next-2.6 tree