projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: Wait for async bio submissions to make some progress at queue time
[linux-2.6-block.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
8f18cf13 19#include <linux/kernel.h>
065631f6 20#include <linux/bio.h>
39279cc3 21#include <linux/buffer_head.h>
f2eb0a24 22#include <linux/file.h>
39279cc3
CM		 23#include <linux/fs.h>
24#include <linux/pagemap.h>
25#include <linux/highmem.h>
26#include <linux/time.h>
27#include <linux/init.h>
28#include <linux/string.h>
29#include <linux/smp_lock.h>
30#include <linux/backing-dev.h>
31#include <linux/mpage.h>
32#include <linux/swap.h>
33#include <linux/writeback.h>
34#include <linux/statfs.h>
35#include <linux/compat.h>
9ebefb18 36#include <linux/bit_spinlock.h>
92fee66d 37#include <linux/version.h>
5103e947 38#include <linux/xattr.h>
33268eaf 39#include <linux/posix_acl.h>
39279cc3
CM		 40#include "ctree.h"
41#include "disk-io.h"
42#include "transaction.h"
43#include "btrfs_inode.h"
44#include "ioctl.h"
45#include "print-tree.h"
0b86a832 46#include "volumes.h"
e6dcd2dc 47#include "ordered-data.h"
39279cc3
CM		 48
49struct btrfs_iget_args {
50 u64 ino;
51 struct btrfs_root *root;
52};
53
54static struct inode_operations btrfs_dir_inode_operations;
55static struct inode_operations btrfs_symlink_inode_operations;
56static struct inode_operations btrfs_dir_ro_inode_operations;
618e21d5 57static struct inode_operations btrfs_special_inode_operations;
39279cc3
CM		 58static struct inode_operations btrfs_file_inode_operations;
59static struct address_space_operations btrfs_aops;
60static struct address_space_operations btrfs_symlink_aops;
61static struct file_operations btrfs_dir_file_operations;
d1310b2e 62static struct extent_io_ops btrfs_extent_io_ops;
39279cc3
CM		 63
64static struct kmem_cache *btrfs_inode_cachep;
65struct kmem_cache *btrfs_trans_handle_cachep;
66struct kmem_cache *btrfs_transaction_cachep;
67struct kmem_cache *btrfs_bit_radix_cachep;
68struct kmem_cache *btrfs_path_cachep;
69
70#define S_SHIFT 12
71static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
72 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
73 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
74 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
75 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
76 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
77 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
78 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
79};
80
7b128766
JB		 81static void btrfs_truncate(struct inode *inode);
82
1832a6d5
CM		 83int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
84 int for_del)
85{
a2135011
CM		 86 u64 total;
87 u64 used;
1832a6d5 88 u64 thresh;
bcbfce8a 89 unsigned long flags;
1832a6d5
CM		 90 int ret = 0;
91
a2135011
CM		 92 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
93 total = btrfs_super_total_bytes(&root->fs_info->super_copy);
94 used = btrfs_super_bytes_used(&root->fs_info->super_copy);
1832a6d5 95 if (for_del)
f9ef6604 96 thresh = total * 90;
1832a6d5 97 else
f9ef6604
CM		 98 thresh = total * 85;
99
100 do_div(thresh, 100);
1832a6d5 101
1832a6d5
CM		 102 if (used + root->fs_info->delalloc_bytes + num_required > thresh)
103 ret = -ENOSPC;
bcbfce8a 104 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
1832a6d5
CM		 105 return ret;
106}
107
be20aa9d 108static int cow_file_range(struct inode *inode, u64 start, u64 end)
b888db2b
CM		 109{
110 struct btrfs_root *root = BTRFS_I(inode)->root;
111 struct btrfs_trans_handle *trans;
b888db2b 112 u64 alloc_hint = 0;
db94535d 113 u64 num_bytes;
c59f8951 114 u64 cur_alloc_size;
db94535d 115 u64 blocksize = root->sectorsize;
d1310b2e 116 u64 orig_num_bytes;
be20aa9d 117 struct btrfs_key ins;
e6dcd2dc
CM		 118 struct extent_map *em;
119 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
120 int ret = 0;
b888db2b 121
f9295749 122 trans = btrfs_join_transaction(root, 1);
b888db2b 123 BUG_ON(!trans);
be20aa9d
CM		 124 btrfs_set_trans_block_group(trans, inode);
125
db94535d 126 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
be20aa9d 127 num_bytes = max(blocksize, num_bytes);
d1310b2e 128 orig_num_bytes = num_bytes;
db94535d 129
179e29e4
CM		 130 if (alloc_hint == EXTENT_MAP_INLINE)
131 goto out;
132
3b951516 133 BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy));
e5a2217e 134 mutex_lock(&BTRFS_I(inode)->extent_mutex);
e6dcd2dc 135 btrfs_drop_extent_cache(inode, start, start + num_bytes - 1);
e5a2217e 136 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
3b951516 137
c59f8951
CM		 138 while(num_bytes > 0) {
139 cur_alloc_size = min(num_bytes, root->fs_info->max_extent);
e6dcd2dc
CM		 140 ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
141 root->sectorsize, 0, 0,
142 (u64)-1, &ins, 1);
c59f8951
CM		 143 if (ret) {
144 WARN_ON(1);
145 goto out;
146 }
e6dcd2dc
CM		 147 em = alloc_extent_map(GFP_NOFS);
148 em->start = start;
149 em->len = ins.offset;
150 em->block_start = ins.objectid;
151 em->bdev = root->fs_info->fs_devices->latest_bdev;
e5a2217e 152 mutex_lock(&BTRFS_I(inode)->extent_mutex);
7f3c74fb 153 set_bit(EXTENT_FLAG_PINNED, &em->flags);
e6dcd2dc
CM		 154 while(1) {
155 spin_lock(&em_tree->lock);
156 ret = add_extent_mapping(em_tree, em);
157 spin_unlock(&em_tree->lock);
158 if (ret != -EEXIST) {
159 free_extent_map(em);
160 break;
161 }
162 btrfs_drop_extent_cache(inode, start,
163 start + ins.offset - 1);
164 }
e5a2217e 165 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
e6dcd2dc 166
98d20f67 167 cur_alloc_size = ins.offset;
e6dcd2dc 168 ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
7ea394f1 169 ins.offset, 0);
e6dcd2dc 170 BUG_ON(ret);
3b951516
CM		 171 if (num_bytes < cur_alloc_size) {
172 printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes,
173 cur_alloc_size);
174 break;
175 }
c59f8951
CM		 176 num_bytes -= cur_alloc_size;
177 alloc_hint = ins.objectid + ins.offset;
178 start += cur_alloc_size;
b888db2b 179 }
b888db2b
CM		 180out:
181 btrfs_end_transaction(trans, root);
be20aa9d
CM		 182 return ret;
183}
184
185static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end)
186{
187 u64 extent_start;
188 u64 extent_end;
189 u64 bytenr;
1832a6d5 190 u64 loops = 0;
c31f8830 191 u64 total_fs_bytes;
be20aa9d 192 struct btrfs_root *root = BTRFS_I(inode)->root;
a68d5933 193 struct btrfs_block_group_cache *block_group;
7ea394f1 194 struct btrfs_trans_handle *trans;
be20aa9d
CM		 195 struct extent_buffer *leaf;
196 int found_type;
197 struct btrfs_path *path;
198 struct btrfs_file_extent_item *item;
199 int ret;
7ea394f1 200 int err = 0;
be20aa9d
CM		 201 struct btrfs_key found_key;
202
c31f8830 203 total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
be20aa9d
CM		 204 path = btrfs_alloc_path();
205 BUG_ON(!path);
7ea394f1
YZ		 206 trans = btrfs_join_transaction(root, 1);
207 BUG_ON(!trans);
be20aa9d
CM		 208again:
209 ret = btrfs_lookup_file_extent(NULL, root, path,
210 inode->i_ino, start, 0);
211 if (ret < 0) {
7ea394f1
YZ		 212 err = ret;
213 goto out;
be20aa9d
CM		 214 }
215
be20aa9d
CM		 216 if (ret != 0) {
217 if (path->slots[0] == 0)
218 goto not_found;
219 path->slots[0]--;
220 }
221
222 leaf = path->nodes[0];
223 item = btrfs_item_ptr(leaf, path->slots[0],
224 struct btrfs_file_extent_item);
225
226 /* are we inside the extent that was found? */
227 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
228 found_type = btrfs_key_type(&found_key);
229 if (found_key.objectid != inode->i_ino ||
bbaf549e 230 found_type != BTRFS_EXTENT_DATA_KEY)
be20aa9d 231 goto not_found;
be20aa9d
CM		 232
233 found_type = btrfs_file_extent_type(leaf, item);
234 extent_start = found_key.offset;
235 if (found_type == BTRFS_FILE_EXTENT_REG) {
c31f8830
CM		 236 u64 extent_num_bytes;
237
238 extent_num_bytes = btrfs_file_extent_num_bytes(leaf, item);
239 extent_end = extent_start + extent_num_bytes;
be20aa9d
CM		 240 err = 0;
241
1832a6d5
CM		 242 if (loops && start != extent_start)
243 goto not_found;
244
be20aa9d
CM		 245 if (start < extent_start || start >= extent_end)
246 goto not_found;
247
be20aa9d
CM		 248 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
249 if (bytenr == 0)
250 goto not_found;
251
7ea394f1 252 if (btrfs_cross_ref_exists(trans, root, &found_key, bytenr))
a68d5933 253 goto not_found;
c31f8830
CM		 254 /*
255 * we may be called by the resizer, make sure we're inside
256 * the limits of the FS
257 */
a68d5933
CM		 258 block_group = btrfs_lookup_block_group(root->fs_info,
259 bytenr);
260 if (!block_group || block_group->ro)
c31f8830
CM		 261 goto not_found;
262
7ea394f1
YZ		 263 bytenr += btrfs_file_extent_offset(leaf, item);
264 extent_num_bytes = min(end + 1, extent_end) - start;
265 ret = btrfs_add_ordered_extent(inode, start, bytenr,
266 extent_num_bytes, 1);
267 if (ret) {
268 err = ret;
269 goto out;
270 }
271
272 btrfs_release_path(root, path);
be20aa9d 273 start = extent_end;
7ea394f1
YZ		 274 if (start <= end) {
275 loops++;
276 goto again;
277 }
bd09835d 278 } else {
7ea394f1
YZ		 279not_found:
280 btrfs_end_transaction(trans, root);
be20aa9d 281 btrfs_free_path(path);
7ea394f1 282 return cow_file_range(inode, start, end);
be20aa9d 283 }
7ea394f1
YZ		 284out:
285 WARN_ON(err);
286 btrfs_end_transaction(trans, root);
287 btrfs_free_path(path);
288 return err;
be20aa9d
CM		 289}
290
291static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
292{
293 struct btrfs_root *root = BTRFS_I(inode)->root;
294 int ret;
a2135011 295
b98b6767
Y		 296 if (btrfs_test_opt(root, NODATACOW) ||
297 btrfs_test_flag(inode, NODATACOW))
be20aa9d
CM		 298 ret = run_delalloc_nocow(inode, start, end);
299 else
300 ret = cow_file_range(inode, start, end);
1832a6d5 301
b888db2b
CM		 302 return ret;
303}
304
291d673e 305int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
b0c68f8b 306 unsigned long old, unsigned long bits)
291d673e 307{
bcbfce8a 308 unsigned long flags;
b0c68f8b 309 if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
291d673e 310 struct btrfs_root *root = BTRFS_I(inode)->root;
bcbfce8a 311 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
9069218d 312 BTRFS_I(inode)->delalloc_bytes += end - start + 1;
291d673e 313 root->fs_info->delalloc_bytes += end - start + 1;
ea8c2819
CM		 314 if (list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
315 list_add_tail(&BTRFS_I(inode)->delalloc_inodes,
316 &root->fs_info->delalloc_inodes);
317 }
bcbfce8a 318 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
291d673e
CM		 319 }
320 return 0;
321}
322
323int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end,
b0c68f8b 324 unsigned long old, unsigned long bits)
291d673e 325{
b0c68f8b 326 if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
291d673e 327 struct btrfs_root *root = BTRFS_I(inode)->root;
bcbfce8a
CM		 328 unsigned long flags;
329
330 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
b0c68f8b
CM		 331 if (end - start + 1 > root->fs_info->delalloc_bytes) {
332 printk("warning: delalloc account %Lu %Lu\n",
333 end - start + 1, root->fs_info->delalloc_bytes);
334 root->fs_info->delalloc_bytes = 0;
9069218d 335 BTRFS_I(inode)->delalloc_bytes = 0;
b0c68f8b
CM		 336 } else {
337 root->fs_info->delalloc_bytes -= end - start + 1;
9069218d 338 BTRFS_I(inode)->delalloc_bytes -= end - start + 1;
b0c68f8b 339 }
ea8c2819
CM		 340 if (BTRFS_I(inode)->delalloc_bytes == 0 &&
341 !list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
342 list_del_init(&BTRFS_I(inode)->delalloc_inodes);
343 }
bcbfce8a 344 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
291d673e
CM		 345 }
346 return 0;
347}
348
239b14b3
CM		 349int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
350 size_t size, struct bio *bio)
351{
352 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
353 struct btrfs_mapping_tree *map_tree;
239b14b3 354 u64 logical = bio->bi_sector << 9;
239b14b3
CM		 355 u64 length = 0;
356 u64 map_length;
239b14b3
CM		 357 int ret;
358
f2d8d74d 359 length = bio->bi_size;
239b14b3
CM		 360 map_tree = &root->fs_info->mapping_tree;
361 map_length = length;
cea9e445 362 ret = btrfs_map_block(map_tree, READ, logical,
f188591e 363 &map_length, NULL, 0);
cea9e445 364
239b14b3 365 if (map_length < length + size) {
239b14b3
CM		 366 return 1;
367 }
368 return 0;
369}
370
44b8bd7e 371int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
f188591e 372 int mirror_num)
065631f6 373{
065631f6 374 struct btrfs_root *root = BTRFS_I(inode)->root;
065631f6 375 int ret = 0;
e015640f 376
3edf7d33 377 ret = btrfs_csum_one_bio(root, inode, bio);
44b8bd7e 378 BUG_ON(ret);
e015640f 379
8b712842 380 return btrfs_map_bio(root, rw, bio, mirror_num, 1);
44b8bd7e
CM		 381}
382
383int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
384 int mirror_num)
385{
386 struct btrfs_root *root = BTRFS_I(inode)->root;
387 int ret = 0;
388
e6dcd2dc
CM		 389 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
390 BUG_ON(ret);
065631f6 391
7ea394f1
YZ		 392 if (btrfs_test_opt(root, NODATASUM) ||
393 btrfs_test_flag(inode, NODATASUM)) {
394 goto mapit;
395 }
396
4d1b5fb4
CM		 397 if (!(rw & (1 << BIO_RW))) {
398 btrfs_lookup_bio_sums(root, inode, bio);
399 goto mapit;
400 }
44b8bd7e
CM		 401 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
402 inode, rw, bio, mirror_num,
403 __btrfs_submit_bio_hook);
0b86a832 404mapit:
8b712842 405 return btrfs_map_bio(root, rw, bio, mirror_num, 0);
065631f6 406}
6885f308 407
ba1da2f4 408static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
e6dcd2dc
CM		 409 struct inode *inode, u64 file_offset,
410 struct list_head *list)
411{
412 struct list_head *cur;
413 struct btrfs_ordered_sum *sum;
414
415 btrfs_set_trans_block_group(trans, inode);
ba1da2f4 416 list_for_each(cur, list) {
e6dcd2dc 417 sum = list_entry(cur, struct btrfs_ordered_sum, list);
e6dcd2dc
CM		 418 btrfs_csum_file_blocks(trans, BTRFS_I(inode)->root,
419 inode, sum);
e6dcd2dc
CM		 420 }
421 return 0;
422}
423
ea8c2819
CM		 424int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end)
425{
426 return set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end,
427 GFP_NOFS);
428}
429
247e743c
CM		 430struct btrfs_writepage_fixup {
431 struct page *page;
432 struct btrfs_work work;
433};
434
435/* see btrfs_writepage_start_hook for details on why this is required */
436void btrfs_writepage_fixup_worker(struct btrfs_work *work)
437{
438 struct btrfs_writepage_fixup *fixup;
439 struct btrfs_ordered_extent *ordered;
440 struct page *page;
441 struct inode *inode;
442 u64 page_start;
443 u64 page_end;
444
445 fixup = container_of(work, struct btrfs_writepage_fixup, work);
446 page = fixup->page;
4a096752 447again:
247e743c
CM		 448 lock_page(page);
449 if (!page->mapping || !PageDirty(page) || !PageChecked(page)) {
450 ClearPageChecked(page);
451 goto out_page;
452 }
453
454 inode = page->mapping->host;
455 page_start = page_offset(page);
456 page_end = page_offset(page) + PAGE_CACHE_SIZE - 1;
457
458 lock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
4a096752
CM		 459
460 /* already ordered? We're done */
461 if (test_range_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
462 EXTENT_ORDERED, 0)) {
247e743c 463 goto out;
4a096752
CM		 464 }
465
466 ordered = btrfs_lookup_ordered_extent(inode, page_start);
467 if (ordered) {
468 unlock_extent(&BTRFS_I(inode)->io_tree, page_start,
469 page_end, GFP_NOFS);
470 unlock_page(page);
471 btrfs_start_ordered_extent(inode, ordered, 1);
472 goto again;
473 }
247e743c 474
ea8c2819 475 btrfs_set_extent_delalloc(inode, page_start, page_end);
247e743c
CM		 476 ClearPageChecked(page);
477out:
478 unlock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
479out_page:
480 unlock_page(page);
481 page_cache_release(page);
482}
483
484/*
485 * There are a few paths in the higher layers of the kernel that directly
486 * set the page dirty bit without asking the filesystem if it is a
487 * good idea. This causes problems because we want to make sure COW
488 * properly happens and the data=ordered rules are followed.
489 *
490 * In our case any range that doesn't have the EXTENT_ORDERED bit set
491 * hasn't been properly setup for IO. We kick off an async process
492 * to fix it up. The async helper will wait for ordered extents, set
493 * the delalloc bit and make it safe to write the page.
494 */
495int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
496{
497 struct inode *inode = page->mapping->host;
498 struct btrfs_writepage_fixup *fixup;
499 struct btrfs_root *root = BTRFS_I(inode)->root;
500 int ret;
501
502 ret = test_range_bit(&BTRFS_I(inode)->io_tree, start, end,
503 EXTENT_ORDERED, 0);
504 if (ret)
505 return 0;
506
507 if (PageChecked(page))
508 return -EAGAIN;
509
510 fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
511 if (!fixup)
512 return -EAGAIN;
f421950f 513
247e743c
CM		 514 SetPageChecked(page);
515 page_cache_get(page);
516 fixup->work.func = btrfs_writepage_fixup_worker;
517 fixup->page = page;
518 btrfs_queue_worker(&root->fs_info->fixup_workers, &fixup->work);
519 return -EAGAIN;
520}
521
211f90e6 522static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
e6dcd2dc 523{
e6dcd2dc
CM		 524 struct btrfs_root *root = BTRFS_I(inode)->root;
525 struct btrfs_trans_handle *trans;
526 struct btrfs_ordered_extent *ordered_extent;
527 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
528 u64 alloc_hint = 0;
529 struct list_head list;
530 struct btrfs_key ins;
531 int ret;
532
533 ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1);
ba1da2f4 534 if (!ret)
e6dcd2dc 535 return 0;
e6dcd2dc 536
f9295749 537 trans = btrfs_join_transaction(root, 1);
e6dcd2dc
CM		 538
539 ordered_extent = btrfs_lookup_ordered_extent(inode, start);
540 BUG_ON(!ordered_extent);
7ea394f1
YZ		 541 if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags))
542 goto nocow;
e6dcd2dc
CM		 543
544 lock_extent(io_tree, ordered_extent->file_offset,
545 ordered_extent->file_offset + ordered_extent->len - 1,
546 GFP_NOFS);
547
548 INIT_LIST_HEAD(&list);
549
550 ins.objectid = ordered_extent->start;
551 ins.offset = ordered_extent->len;
552 ins.type = BTRFS_EXTENT_ITEM_KEY;
e5a2217e 553
e6dcd2dc
CM		 554 ret = btrfs_alloc_reserved_extent(trans, root, root->root_key.objectid,
555 trans->transid, inode->i_ino,
556 ordered_extent->file_offset, &ins);
557 BUG_ON(ret);
ee6e6504
CM		 558
559 mutex_lock(&BTRFS_I(inode)->extent_mutex);
e5a2217e 560
e6dcd2dc
CM		 561 ret = btrfs_drop_extents(trans, root, inode,
562 ordered_extent->file_offset,
563 ordered_extent->file_offset +
564 ordered_extent->len,
565 ordered_extent->file_offset, &alloc_hint);
566 BUG_ON(ret);
567 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
568 ordered_extent->file_offset,
569 ordered_extent->start,
570 ordered_extent->len,
571 ordered_extent->len, 0);
572 BUG_ON(ret);
7f3c74fb 573
e6dcd2dc
CM		 574 btrfs_drop_extent_cache(inode, ordered_extent->file_offset,
575 ordered_extent->file_offset +
576 ordered_extent->len - 1);
ee6e6504
CM		 577 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
578
e6dcd2dc
CM		 579 inode->i_blocks += ordered_extent->len >> 9;
580 unlock_extent(io_tree, ordered_extent->file_offset,
581 ordered_extent->file_offset + ordered_extent->len - 1,
582 GFP_NOFS);
7ea394f1 583nocow:
e6dcd2dc
CM		 584 add_pending_csums(trans, inode, ordered_extent->file_offset,
585 &ordered_extent->list);
586
dbe674a9 587 btrfs_ordered_update_i_size(inode, ordered_extent);
e6dcd2dc 588 btrfs_remove_ordered_extent(inode, ordered_extent);
7f3c74fb 589
e6dcd2dc
CM		 590 /* once for us */
591 btrfs_put_ordered_extent(ordered_extent);
592 /* once for the tree */
593 btrfs_put_ordered_extent(ordered_extent);
594
595 btrfs_update_inode(trans, root, inode);
596 btrfs_end_transaction(trans, root);
597 return 0;
598}
599
211f90e6
CM		 600int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
601 struct extent_state *state, int uptodate)
602{
603 return btrfs_finish_ordered_io(page->mapping->host, start, end);
604}
605
7e38326f
CM		 606struct io_failure_record {
607 struct page *page;
608 u64 start;
609 u64 len;
610 u64 logical;
611 int last_mirror;
612};
613
1259ab75
CM		 614int btrfs_io_failed_hook(struct bio *failed_bio,
615 struct page *page, u64 start, u64 end,
616 struct extent_state *state)
7e38326f
CM		 617{
618 struct io_failure_record *failrec = NULL;
619 u64 private;
620 struct extent_map *em;
621 struct inode *inode = page->mapping->host;
622 struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
3b951516 623 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
7e38326f
CM		 624 struct bio *bio;
625 int num_copies;
626 int ret;
1259ab75 627 int rw;
7e38326f
CM		 628 u64 logical;
629
630 ret = get_state_private(failure_tree, start, &private);
631 if (ret) {
7e38326f
CM		 632 failrec = kmalloc(sizeof(*failrec), GFP_NOFS);
633 if (!failrec)
634 return -ENOMEM;
635 failrec->start = start;
636 failrec->len = end - start + 1;
637 failrec->last_mirror = 0;
638
3b951516
CM		 639 spin_lock(&em_tree->lock);
640 em = lookup_extent_mapping(em_tree, start, failrec->len);
641 if (em->start > start || em->start + em->len < start) {
642 free_extent_map(em);
643 em = NULL;
644 }
645 spin_unlock(&em_tree->lock);
7e38326f
CM		 646
647 if (!em || IS_ERR(em)) {
648 kfree(failrec);
649 return -EIO;
650 }
651 logical = start - em->start;
652 logical = em->block_start + logical;
653 failrec->logical = logical;
654 free_extent_map(em);
655 set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
656 EXTENT_DIRTY, GFP_NOFS);
587f7704
CM		 657 set_state_private(failure_tree, start,
658 (u64)(unsigned long)failrec);
7e38326f 659 } else {
587f7704 660 failrec = (struct io_failure_record *)(unsigned long)private;
7e38326f
CM		 661 }
662 num_copies = btrfs_num_copies(
663 &BTRFS_I(inode)->root->fs_info->mapping_tree,
664 failrec->logical, failrec->len);
665 failrec->last_mirror++;
666 if (!state) {
667 spin_lock_irq(&BTRFS_I(inode)->io_tree.lock);
668 state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
669 failrec->start,
670 EXTENT_LOCKED);
671 if (state && state->start != failrec->start)
672 state = NULL;
673 spin_unlock_irq(&BTRFS_I(inode)->io_tree.lock);
674 }
675 if (!state || failrec->last_mirror > num_copies) {
676 set_state_private(failure_tree, failrec->start, 0);
677 clear_extent_bits(failure_tree, failrec->start,
678 failrec->start + failrec->len - 1,
679 EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
680 kfree(failrec);
681 return -EIO;
682 }
683 bio = bio_alloc(GFP_NOFS, 1);
684 bio->bi_private = state;
685 bio->bi_end_io = failed_bio->bi_end_io;
686 bio->bi_sector = failrec->logical >> 9;
687 bio->bi_bdev = failed_bio->bi_bdev;
e1c4b745 688 bio->bi_size = 0;
7e38326f 689 bio_add_page(bio, page, failrec->len, start - page_offset(page));
1259ab75
CM		 690 if (failed_bio->bi_rw & (1 << BIO_RW))
691 rw = WRITE;
692 else
693 rw = READ;
694
695 BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
696 failrec->last_mirror);
697 return 0;
698}
699
700int btrfs_clean_io_failures(struct inode *inode, u64 start)
701{
702 u64 private;
703 u64 private_failure;
704 struct io_failure_record *failure;
705 int ret;
706
707 private = 0;
708 if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
709 (u64)-1, 1, EXTENT_DIRTY)) {
710 ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
711 start, &private_failure);
712 if (ret == 0) {
713 failure = (struct io_failure_record *)(unsigned long)
714 private_failure;
715 set_state_private(&BTRFS_I(inode)->io_failure_tree,
716 failure->start, 0);
717 clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
718 failure->start,
719 failure->start + failure->len - 1,
720 EXTENT_DIRTY | EXTENT_LOCKED,
721 GFP_NOFS);
722 kfree(failure);
723 }
724 }
7e38326f
CM		 725 return 0;
726}
727
70dec807
CM		 728int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
729 struct extent_state *state)
07157aac 730{
35ebb934 731 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
07157aac 732 struct inode *inode = page->mapping->host;
d1310b2e 733 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
07157aac 734 char *kaddr;
aadfeb6e 735 u64 private = ~(u32)0;
07157aac 736 int ret;
ff79f819
CM		 737 struct btrfs_root *root = BTRFS_I(inode)->root;
738 u32 csum = ~(u32)0;
bbf0d006 739 unsigned long flags;
d1310b2e 740
b98b6767
Y		 741 if (btrfs_test_opt(root, NODATASUM) ||
742 btrfs_test_flag(inode, NODATASUM))
b6cda9bc 743 return 0;
c2e639f0 744 if (state && state->start == start) {
70dec807
CM		 745 private = state->private;
746 ret = 0;
747 } else {
748 ret = get_state_private(io_tree, start, &private);
749 }
bbf0d006 750 local_irq_save(flags);
07157aac
CM		 751 kaddr = kmap_atomic(page, KM_IRQ0);
752 if (ret) {
753 goto zeroit;
754 }
ff79f819
CM		 755 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
756 btrfs_csum_final(csum, (char *)&csum);
757 if (csum != private) {
07157aac
CM		 758 goto zeroit;
759 }
760 kunmap_atomic(kaddr, KM_IRQ0);
bbf0d006 761 local_irq_restore(flags);
7e38326f
CM		 762
763 /* if the io failure tree for this inode is non-empty,
764 * check to see if we've recovered from a failed IO
765 */
1259ab75 766 btrfs_clean_io_failures(inode, start);
07157aac
CM		 767 return 0;
768
769zeroit:
aadfeb6e
CM		 770 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
771 page->mapping->host->i_ino, (unsigned long long)start, csum,
772 private);
db94535d
CM		 773 memset(kaddr + offset, 1, end - start + 1);
774 flush_dcache_page(page);
07157aac 775 kunmap_atomic(kaddr, KM_IRQ0);
bbf0d006 776 local_irq_restore(flags);
3b951516
CM		 777 if (private == 0)
778 return 0;
7e38326f 779 return -EIO;
07157aac 780}
b888db2b 781
7b128766
JB		 782/*
783 * This creates an orphan entry for the given inode in case something goes
784 * wrong in the middle of an unlink/truncate.
785 */
786int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode)
787{
788 struct btrfs_root *root = BTRFS_I(inode)->root;
789 int ret = 0;
790
bcc63abb 791 spin_lock(&root->list_lock);
7b128766
JB		 792
793 /* already on the orphan list, we're good */
794 if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
bcc63abb 795 spin_unlock(&root->list_lock);
7b128766
JB		 796 return 0;
797 }
798
799 list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);
800
bcc63abb 801 spin_unlock(&root->list_lock);
7b128766
JB		 802
803 /*
804 * insert an orphan item to track this unlinked/truncated file
805 */
806 ret = btrfs_insert_orphan_item(trans, root, inode->i_ino);
807
808 return ret;
809}
810
811/*
812 * We have done the truncate/delete so we can go ahead and remove the orphan
813 * item for this particular inode.
814 */
815int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode)
816{
817 struct btrfs_root *root = BTRFS_I(inode)->root;
818 int ret = 0;
819
bcc63abb 820 spin_lock(&root->list_lock);
7b128766
JB		 821
822 if (list_empty(&BTRFS_I(inode)->i_orphan)) {
bcc63abb 823 spin_unlock(&root->list_lock);
7b128766
JB		 824 return 0;
825 }
826
827 list_del_init(&BTRFS_I(inode)->i_orphan);
828 if (!trans) {
bcc63abb 829 spin_unlock(&root->list_lock);
7b128766
JB		 830 return 0;
831 }
832
bcc63abb 833 spin_unlock(&root->list_lock);
7b128766
JB		 834
835 ret = btrfs_del_orphan_item(trans, root, inode->i_ino);
836
837 return ret;
838}
839
840/*
841 * this cleans up any orphans that may be left on the list from the last use
842 * of this root.
843 */
844void btrfs_orphan_cleanup(struct btrfs_root *root)
845{
846 struct btrfs_path *path;
847 struct extent_buffer *leaf;
848 struct btrfs_item *item;
849 struct btrfs_key key, found_key;
850 struct btrfs_trans_handle *trans;
851 struct inode *inode;
852 int ret = 0, nr_unlink = 0, nr_truncate = 0;
853
854 /* don't do orphan cleanup if the fs is readonly. */
855 if (root->inode->i_sb->s_flags & MS_RDONLY)
856 return;
857
858 path = btrfs_alloc_path();
859 if (!path)
860 return;
861 path->reada = -1;
862
863 key.objectid = BTRFS_ORPHAN_OBJECTID;
864 btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY);
865 key.offset = (u64)-1;
866
867 trans = btrfs_start_transaction(root, 1);
868 btrfs_set_trans_block_group(trans, root->inode);
869
870 while (1) {
871 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
872 if (ret < 0) {
873 printk(KERN_ERR "Error searching slot for orphan: %d"
874 "\n", ret);
875 break;
876 }
877
878 /*
879 * if ret == 0 means we found what we were searching for, which
880 * is weird, but possible, so only screw with path if we didnt
881 * find the key and see if we have stuff that matches
882 */
883 if (ret > 0) {
884 if (path->slots[0] == 0)
885 break;
886 path->slots[0]--;
887 }
888
889 /* pull out the item */
890 leaf = path->nodes[0];
891 item = btrfs_item_nr(leaf, path->slots[0]);
892 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
893
894 /* make sure the item matches what we want */
895 if (found_key.objectid != BTRFS_ORPHAN_OBJECTID)
896 break;
897 if (btrfs_key_type(&found_key) != BTRFS_ORPHAN_ITEM_KEY)
898 break;
899
900 /* release the path since we're done with it */
901 btrfs_release_path(root, path);
902
903 /*
904 * this is where we are basically btrfs_lookup, without the
905 * crossing root thing. we store the inode number in the
906 * offset of the orphan item.
907 */
908 inode = btrfs_iget_locked(root->inode->i_sb,
909 found_key.offset, root);
910 if (!inode)
911 break;
912
913 if (inode->i_state & I_NEW) {
914 BTRFS_I(inode)->root = root;
915
916 /* have to set the location manually */
917 BTRFS_I(inode)->location.objectid = inode->i_ino;
918 BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
919 BTRFS_I(inode)->location.offset = 0;
920
921 btrfs_read_locked_inode(inode);
922 unlock_new_inode(inode);
923 }
924
925 /*
926 * add this inode to the orphan list so btrfs_orphan_del does
927 * the proper thing when we hit it
928 */
bcc63abb 929 spin_lock(&root->list_lock);
7b128766 930 list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);
bcc63abb 931 spin_unlock(&root->list_lock);
7b128766
JB		 932
933 /*
934 * if this is a bad inode, means we actually succeeded in
935 * removing the inode, but not the orphan record, which means
936 * we need to manually delete the orphan since iput will just
937 * do a destroy_inode
938 */
939 if (is_bad_inode(inode)) {
940 btrfs_orphan_del(trans, inode);
941 iput(inode);
942 continue;
943 }
944
945 /* if we have links, this was a truncate, lets do that */
946 if (inode->i_nlink) {
947 nr_truncate++;
948 btrfs_truncate(inode);
949 } else {
950 nr_unlink++;
951 }
952
953 /* this will do delete_inode and everything for us */
954 iput(inode);
955 }
956
957 if (nr_unlink)
958 printk(KERN_INFO "btrfs: unlinked %d orphans\n", nr_unlink);
959 if (nr_truncate)
960 printk(KERN_INFO "btrfs: truncated %d orphans\n", nr_truncate);
961
962 btrfs_free_path(path);
963 btrfs_end_transaction(trans, root);
964}
965
39279cc3
CM		 966void btrfs_read_locked_inode(struct inode *inode)
967{
968 struct btrfs_path *path;
5f39d397 969 struct extent_buffer *leaf;
39279cc3 970 struct btrfs_inode_item *inode_item;
0b86a832 971 struct btrfs_timespec *tspec;
39279cc3
CM		 972 struct btrfs_root *root = BTRFS_I(inode)->root;
973 struct btrfs_key location;
974 u64 alloc_group_block;
618e21d5 975 u32 rdev;
39279cc3
CM		 976 int ret;
977
978 path = btrfs_alloc_path();
979 BUG_ON(!path);
39279cc3 980 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
dc17ff8f 981
39279cc3 982 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
5f39d397 983 if (ret)
39279cc3 984 goto make_bad;
39279cc3 985
5f39d397
CM		 986 leaf = path->nodes[0];
987 inode_item = btrfs_item_ptr(leaf, path->slots[0],
988 struct btrfs_inode_item);
989
990 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
991 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
992 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
993 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
dbe674a9 994 btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item));
5f39d397
CM		 995
996 tspec = btrfs_inode_atime(inode_item);
997 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
998 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
999
1000 tspec = btrfs_inode_mtime(inode_item);
1001 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
1002 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
1003
1004 tspec = btrfs_inode_ctime(inode_item);
1005 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
1006 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
1007
1008 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
1009 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
618e21d5 1010 inode->i_rdev = 0;
5f39d397
CM		 1011 rdev = btrfs_inode_rdev(leaf, inode_item);
1012
aec7477b
JB		 1013 BTRFS_I(inode)->index_cnt = (u64)-1;
1014
5f39d397 1015 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
39279cc3
CM		 1016 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
1017 alloc_group_block);
b98b6767 1018 BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
e52ec0eb
CM		 1019 if (!BTRFS_I(inode)->block_group) {
1020 BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
0b86a832
CM		 1021 NULL, 0,
1022 BTRFS_BLOCK_GROUP_METADATA, 0);
e52ec0eb 1023 }
39279cc3
CM		 1024 btrfs_free_path(path);
1025 inode_item = NULL;
1026
39279cc3 1027 switch (inode->i_mode & S_IFMT) {
39279cc3
CM		 1028 case S_IFREG:
1029 inode->i_mapping->a_ops = &btrfs_aops;
04160088 1030 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
d1310b2e 1031 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
39279cc3
CM		 1032 inode->i_fop = &btrfs_file_operations;
1033 inode->i_op = &btrfs_file_inode_operations;
1034 break;
1035 case S_IFDIR:
1036 inode->i_fop = &btrfs_dir_file_operations;
1037 if (root == root->fs_info->tree_root)
1038 inode->i_op = &btrfs_dir_ro_inode_operations;
1039 else
1040 inode->i_op = &btrfs_dir_inode_operations;
1041 break;
1042 case S_IFLNK:
1043 inode->i_op = &btrfs_symlink_inode_operations;
1044 inode->i_mapping->a_ops = &btrfs_symlink_aops;
04160088 1045 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
39279cc3 1046 break;
618e21d5
JB		 1047 default:
1048 init_special_inode(inode, inode->i_mode, rdev);
1049 break;
39279cc3
CM		 1050 }
1051 return;
1052
1053make_bad:
39279cc3 1054 btrfs_free_path(path);
39279cc3
CM		 1055 make_bad_inode(inode);
1056}
1057
5f39d397
CM		 1058static void fill_inode_item(struct extent_buffer *leaf,
1059 struct btrfs_inode_item *item,
39279cc3
CM		 1060 struct inode *inode)
1061{
5f39d397
CM		 1062 btrfs_set_inode_uid(leaf, item, inode->i_uid);
1063 btrfs_set_inode_gid(leaf, item, inode->i_gid);
dbe674a9 1064 btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size);
5f39d397
CM		 1065 btrfs_set_inode_mode(leaf, item, inode->i_mode);
1066 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
1067
1068 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
1069 inode->i_atime.tv_sec);
1070 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
1071 inode->i_atime.tv_nsec);
1072
1073 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
1074 inode->i_mtime.tv_sec);
1075 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
1076 inode->i_mtime.tv_nsec);
1077
1078 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
1079 inode->i_ctime.tv_sec);
1080 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
1081 inode->i_ctime.tv_nsec);
1082
1083 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
1084 btrfs_set_inode_generation(leaf, item, inode->i_generation);
1085 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
b98b6767 1086 btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
5f39d397 1087 btrfs_set_inode_block_group(leaf, item,
39279cc3
CM		 1088 BTRFS_I(inode)->block_group->key.objectid);
1089}
1090
ba1da2f4 1091int noinline btrfs_update_inode(struct btrfs_trans_handle *trans,
39279cc3
CM		 1092 struct btrfs_root *root,
1093 struct inode *inode)
1094{
1095 struct btrfs_inode_item *inode_item;
1096 struct btrfs_path *path;
5f39d397 1097 struct extent_buffer *leaf;
39279cc3
CM		 1098 int ret;
1099
1100 path = btrfs_alloc_path();
1101 BUG_ON(!path);
39279cc3
CM		 1102 ret = btrfs_lookup_inode(trans, root, path,
1103 &BTRFS_I(inode)->location, 1);
1104 if (ret) {
1105 if (ret > 0)
1106 ret = -ENOENT;
1107 goto failed;
1108 }
1109
5f39d397
CM		 1110 leaf = path->nodes[0];
1111 inode_item = btrfs_item_ptr(leaf, path->slots[0],
39279cc3
CM		 1112 struct btrfs_inode_item);
1113
5f39d397
CM		 1114 fill_inode_item(leaf, inode_item, inode);
1115 btrfs_mark_buffer_dirty(leaf);
15ee9bc7 1116 btrfs_set_inode_last_trans(trans, inode);
39279cc3
CM		 1117 ret = 0;
1118failed:
39279cc3
CM		 1119 btrfs_free_path(path);
1120 return ret;
1121}
1122
1123
1124static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
1125 struct btrfs_root *root,
1126 struct inode *dir,
1127 struct dentry *dentry)
1128{
1129 struct btrfs_path *path;
1130 const char *name = dentry->d_name.name;
1131 int name_len = dentry->d_name.len;
1132 int ret = 0;
5f39d397 1133 struct extent_buffer *leaf;
39279cc3 1134 struct btrfs_dir_item *di;
5f39d397 1135 struct btrfs_key key;
aec7477b 1136 u64 index;
39279cc3
CM		 1137
1138 path = btrfs_alloc_path();
54aa1f4d
CM		 1139 if (!path) {
1140 ret = -ENOMEM;
1141 goto err;
1142 }
1143
39279cc3
CM		 1144 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
1145 name, name_len, -1);
1146 if (IS_ERR(di)) {
1147 ret = PTR_ERR(di);
1148 goto err;
1149 }
1150 if (!di) {
1151 ret = -ENOENT;
1152 goto err;
1153 }
5f39d397
CM		 1154 leaf = path->nodes[0];
1155 btrfs_dir_item_key_to_cpu(leaf, di, &key);
39279cc3 1156 ret = btrfs_delete_one_dir_name(trans, root, path, di);
54aa1f4d
CM		 1157 if (ret)
1158 goto err;
39279cc3
CM		 1159 btrfs_release_path(root, path);
1160
aec7477b
JB		 1161 ret = btrfs_del_inode_ref(trans, root, name, name_len,
1162 dentry->d_inode->i_ino,
1163 dentry->d_parent->d_inode->i_ino, &index);
1164 if (ret) {
1165 printk("failed to delete reference to %.*s, "
1166 "inode %lu parent %lu\n", name_len, name,
1167 dentry->d_inode->i_ino,
1168 dentry->d_parent->d_inode->i_ino);
1169 goto err;
1170 }
1171
39279cc3 1172 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
aec7477b 1173 index, name, name_len, -1);
39279cc3
CM		 1174 if (IS_ERR(di)) {
1175 ret = PTR_ERR(di);
1176 goto err;
1177 }
1178 if (!di) {
1179 ret = -ENOENT;
1180 goto err;
1181 }
1182 ret = btrfs_delete_one_dir_name(trans, root, path, di);
925baedd 1183 btrfs_release_path(root, path);
39279cc3
CM		 1184
1185 dentry->d_inode->i_ctime = dir->i_ctime;
1186err:
1187 btrfs_free_path(path);
1188 if (!ret) {
dbe674a9 1189 btrfs_i_size_write(dir, dir->i_size - name_len * 2);
79c44584 1190 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
39279cc3 1191 btrfs_update_inode(trans, root, dir);
6da6abae
CM		 1192#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1193 dentry->d_inode->i_nlink--;
1194#else
39279cc3 1195 drop_nlink(dentry->d_inode);
6da6abae 1196#endif
54aa1f4d 1197 ret = btrfs_update_inode(trans, root, dentry->d_inode);
39279cc3
CM		 1198 dir->i_sb->s_dirt = 1;
1199 }
1200 return ret;
1201}
1202
1203static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
1204{
1205 struct btrfs_root *root;
1206 struct btrfs_trans_handle *trans;
7b128766 1207 struct inode *inode = dentry->d_inode;
39279cc3 1208 int ret;
1832a6d5 1209 unsigned long nr = 0;
39279cc3
CM		 1210
1211 root = BTRFS_I(dir)->root;
1832a6d5
CM		 1212
1213 ret = btrfs_check_free_space(root, 1, 1);
1214 if (ret)
1215 goto fail;
1216
39279cc3 1217 trans = btrfs_start_transaction(root, 1);
5f39d397 1218
39279cc3
CM		 1219 btrfs_set_trans_block_group(trans, dir);
1220 ret = btrfs_unlink_trans(trans, root, dir, dentry);
7b128766
JB		 1221
1222 if (inode->i_nlink == 0)
1223 ret = btrfs_orphan_add(trans, inode);
1224
d3c2fdcf 1225 nr = trans->blocks_used;
5f39d397 1226
89ce8a63 1227 btrfs_end_transaction_throttle(trans, root);
1832a6d5 1228fail:
d3c2fdcf 1229 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM		 1230 return ret;
1231}
1232
1233static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
1234{
1235 struct inode *inode = dentry->d_inode;
1832a6d5 1236 int err = 0;
39279cc3
CM		 1237 int ret;
1238 struct btrfs_root *root = BTRFS_I(dir)->root;
39279cc3 1239 struct btrfs_trans_handle *trans;
1832a6d5 1240 unsigned long nr = 0;
39279cc3 1241
925baedd 1242 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
134d4512 1243 return -ENOTEMPTY;
925baedd 1244 }
134d4512 1245
1832a6d5
CM		 1246 ret = btrfs_check_free_space(root, 1, 1);
1247 if (ret)
1248 goto fail;
1249
39279cc3
CM		 1250 trans = btrfs_start_transaction(root, 1);
1251 btrfs_set_trans_block_group(trans, dir);
39279cc3 1252
7b128766
JB		 1253 err = btrfs_orphan_add(trans, inode);
1254 if (err)
1255 goto fail_trans;
1256
39279cc3
CM		 1257 /* now the directory is empty */
1258 err = btrfs_unlink_trans(trans, root, dir, dentry);
1259 if (!err) {
dbe674a9 1260 btrfs_i_size_write(inode, 0);
39279cc3 1261 }
3954401f 1262
7b128766 1263fail_trans:
d3c2fdcf 1264 nr = trans->blocks_used;
89ce8a63 1265 ret = btrfs_end_transaction_throttle(trans, root);
1832a6d5 1266fail:
d3c2fdcf 1267 btrfs_btree_balance_dirty(root, nr);
3954401f 1268
39279cc3
CM		 1269 if (ret && !err)
1270 err = ret;
1271 return err;
1272}
1273
39279cc3
CM		 1274/*
1275 * this can truncate away extent items, csum items and directory items.
1276 * It starts at a high offset and removes keys until it can't find
1277 * any higher than i_size.
1278 *
1279 * csum items that cross the new i_size are truncated to the new size
1280 * as well.
7b128766
JB		 1281 *
1282 * min_type is the minimum key type to truncate down to. If set to 0, this
1283 * will kill all the items on this inode, including the INODE_ITEM_KEY.
39279cc3
CM		 1284 */
1285static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
1286 struct btrfs_root *root,
85e21bac
CM		 1287 struct inode *inode,
1288 u32 min_type)
39279cc3
CM		 1289{
1290 int ret;
1291 struct btrfs_path *path;
1292 struct btrfs_key key;
5f39d397 1293 struct btrfs_key found_key;
39279cc3 1294 u32 found_type;
5f39d397 1295 struct extent_buffer *leaf;
39279cc3
CM		 1296 struct btrfs_file_extent_item *fi;
1297 u64 extent_start = 0;
db94535d 1298 u64 extent_num_bytes = 0;
39279cc3 1299 u64 item_end = 0;
7bb86316 1300 u64 root_gen = 0;
d8d5f3e1 1301 u64 root_owner = 0;
39279cc3
CM		 1302 int found_extent;
1303 int del_item;
85e21bac
CM		 1304 int pending_del_nr = 0;
1305 int pending_del_slot = 0;
179e29e4 1306 int extent_type = -1;
3b951516 1307 u64 mask = root->sectorsize - 1;
39279cc3 1308
3b951516 1309 btrfs_drop_extent_cache(inode, inode->i_size & (~mask), (u64)-1);
39279cc3 1310 path = btrfs_alloc_path();
3c69faec 1311 path->reada = -1;
39279cc3 1312 BUG_ON(!path);
5f39d397 1313
39279cc3
CM		 1314 /* FIXME, add redo link to tree so we don't leak on crash */
1315 key.objectid = inode->i_ino;
1316 key.offset = (u64)-1;
5f39d397
CM		 1317 key.type = (u8)-1;
1318
85e21bac
CM		 1319 btrfs_init_path(path);
1320search_again:
1321 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1322 if (ret < 0) {
1323 goto error;
1324 }
1325 if (ret > 0) {
1326 BUG_ON(path->slots[0] == 0);
1327 path->slots[0]--;
1328 }
1329
39279cc3 1330 while(1) {
39279cc3 1331 fi = NULL;
5f39d397
CM		 1332 leaf = path->nodes[0];
1333 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1334 found_type = btrfs_key_type(&found_key);
39279cc3 1335
5f39d397 1336 if (found_key.objectid != inode->i_ino)
39279cc3 1337 break;
5f39d397 1338
85e21bac 1339 if (found_type < min_type)
39279cc3
CM		 1340 break;
1341
5f39d397 1342 item_end = found_key.offset;
39279cc3 1343 if (found_type == BTRFS_EXTENT_DATA_KEY) {
5f39d397 1344 fi = btrfs_item_ptr(leaf, path->slots[0],
39279cc3 1345 struct btrfs_file_extent_item);
179e29e4
CM		 1346 extent_type = btrfs_file_extent_type(leaf, fi);
1347 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
5f39d397 1348 item_end +=
db94535d 1349 btrfs_file_extent_num_bytes(leaf, fi);
179e29e4
CM		 1350 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1351 struct btrfs_item *item = btrfs_item_nr(leaf,
1352 path->slots[0]);
1353 item_end += btrfs_file_extent_inline_len(leaf,
1354 item);
39279cc3 1355 }
008630c1 1356 item_end--;
39279cc3
CM		 1357 }
1358 if (found_type == BTRFS_CSUM_ITEM_KEY) {
1359 ret = btrfs_csum_truncate(trans, root, path,
1360 inode->i_size);
1361 BUG_ON(ret);
1362 }
008630c1 1363 if (item_end < inode->i_size) {
b888db2b
CM		 1364 if (found_type == BTRFS_DIR_ITEM_KEY) {
1365 found_type = BTRFS_INODE_ITEM_KEY;
1366 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
1367 found_type = BTRFS_CSUM_ITEM_KEY;
85e21bac
CM		 1368 } else if (found_type == BTRFS_EXTENT_DATA_KEY) {
1369 found_type = BTRFS_XATTR_ITEM_KEY;
1370 } else if (found_type == BTRFS_XATTR_ITEM_KEY) {
1371 found_type = BTRFS_INODE_REF_KEY;
b888db2b
CM		 1372 } else if (found_type) {
1373 found_type--;
1374 } else {
1375 break;
39279cc3 1376 }
a61721d5 1377 btrfs_set_key_type(&key, found_type);
85e21bac 1378 goto next;
39279cc3 1379 }
5f39d397 1380 if (found_key.offset >= inode->i_size)
39279cc3
CM		 1381 del_item = 1;
1382 else
1383 del_item = 0;
1384 found_extent = 0;
1385
1386 /* FIXME, shrink the extent if the ref count is only 1 */
179e29e4
CM		 1387 if (found_type != BTRFS_EXTENT_DATA_KEY)
1388 goto delete;
1389
1390 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
39279cc3 1391 u64 num_dec;
db94535d 1392 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
39279cc3 1393 if (!del_item) {
db94535d
CM		 1394 u64 orig_num_bytes =
1395 btrfs_file_extent_num_bytes(leaf, fi);
1396 extent_num_bytes = inode->i_size -
5f39d397 1397 found_key.offset + root->sectorsize - 1;
b1632b10
Y		 1398 extent_num_bytes = extent_num_bytes &
1399 ~((u64)root->sectorsize - 1);
db94535d
CM		 1400 btrfs_set_file_extent_num_bytes(leaf, fi,
1401 extent_num_bytes);
1402 num_dec = (orig_num_bytes -
9069218d
CM		 1403 extent_num_bytes);
1404 if (extent_start != 0)
1405 dec_i_blocks(inode, num_dec);
5f39d397 1406 btrfs_mark_buffer_dirty(leaf);
39279cc3 1407 } else {
db94535d
CM		 1408 extent_num_bytes =
1409 btrfs_file_extent_disk_num_bytes(leaf,
1410 fi);
39279cc3 1411 /* FIXME blocksize != 4096 */
9069218d 1412 num_dec = btrfs_file_extent_num_bytes(leaf, fi);
39279cc3
CM		 1413 if (extent_start != 0) {
1414 found_extent = 1;
9069218d 1415 dec_i_blocks(inode, num_dec);
39279cc3 1416 }
d8d5f3e1
CM		 1417 root_gen = btrfs_header_generation(leaf);
1418 root_owner = btrfs_header_owner(leaf);
39279cc3 1419 }
9069218d
CM		 1420 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1421 if (!del_item) {
1422 u32 newsize = inode->i_size - found_key.offset;
1423 dec_i_blocks(inode, item_end + 1 -
1424 found_key.offset - newsize);
1425 newsize =
1426 btrfs_file_extent_calc_inline_size(newsize);
1427 ret = btrfs_truncate_item(trans, root, path,
1428 newsize, 1);
1429 BUG_ON(ret);
1430 } else {
1431 dec_i_blocks(inode, item_end + 1 -
1432 found_key.offset);
1433 }
39279cc3 1434 }
179e29e4 1435delete:
39279cc3 1436 if (del_item) {
85e21bac
CM		 1437 if (!pending_del_nr) {
1438 /* no pending yet, add ourselves */
1439 pending_del_slot = path->slots[0];
1440 pending_del_nr = 1;
1441 } else if (pending_del_nr &&
1442 path->slots[0] + 1 == pending_del_slot) {
1443 /* hop on the pending chunk */
1444 pending_del_nr++;
1445 pending_del_slot = path->slots[0];
1446 } else {
1447 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot);
1448 }
39279cc3
CM		 1449 } else {
1450 break;
1451 }
39279cc3
CM		 1452 if (found_extent) {
1453 ret = btrfs_free_extent(trans, root, extent_start,
7bb86316 1454 extent_num_bytes,
d8d5f3e1 1455 root_owner,
7bb86316
CM		 1456 root_gen, inode->i_ino,
1457 found_key.offset, 0);
39279cc3
CM		 1458 BUG_ON(ret);
1459 }
85e21bac
CM		 1460next:
1461 if (path->slots[0] == 0) {
1462 if (pending_del_nr)
1463 goto del_pending;
1464 btrfs_release_path(root, path);
1465 goto search_again;
1466 }
1467
1468 path->slots[0]--;
1469 if (pending_del_nr &&
1470 path->slots[0] + 1 != pending_del_slot) {
1471 struct btrfs_key debug;
1472del_pending:
1473 btrfs_item_key_to_cpu(path->nodes[0], &debug,
1474 pending_del_slot);
1475 ret = btrfs_del_items(trans, root, path,
1476 pending_del_slot,
1477 pending_del_nr);
1478 BUG_ON(ret);
1479 pending_del_nr = 0;
1480 btrfs_release_path(root, path);
1481 goto search_again;
1482 }
39279cc3
CM		 1483 }
1484 ret = 0;
1485error:
85e21bac
CM		 1486 if (pending_del_nr) {
1487 ret = btrfs_del_items(trans, root, path, pending_del_slot,
1488 pending_del_nr);
1489 }
39279cc3
CM		 1490 btrfs_free_path(path);
1491 inode->i_sb->s_dirt = 1;
1492 return ret;
1493}
1494
1495/*
1496 * taken from block_truncate_page, but does cow as it zeros out
1497 * any bytes left in the last page in the file.
1498 */
1499static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
1500{
1501 struct inode *inode = mapping->host;
db94535d 1502 struct btrfs_root *root = BTRFS_I(inode)->root;
e6dcd2dc
CM		 1503 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1504 struct btrfs_ordered_extent *ordered;
1505 char *kaddr;
db94535d 1506 u32 blocksize = root->sectorsize;
39279cc3
CM		 1507 pgoff_t index = from >> PAGE_CACHE_SHIFT;
1508 unsigned offset = from & (PAGE_CACHE_SIZE-1);
1509 struct page *page;
39279cc3 1510 int ret = 0;
a52d9a80 1511 u64 page_start;
e6dcd2dc 1512 u64 page_end;
39279cc3
CM		 1513
1514 if ((offset & (blocksize - 1)) == 0)
1515 goto out;
1516
1517 ret = -ENOMEM;
211c17f5 1518again:
39279cc3
CM		 1519 page = grab_cache_page(mapping, index);
1520 if (!page)
1521 goto out;
e6dcd2dc
CM		 1522
1523 page_start = page_offset(page);
1524 page_end = page_start + PAGE_CACHE_SIZE - 1;
1525
39279cc3 1526 if (!PageUptodate(page)) {
9ebefb18 1527 ret = btrfs_readpage(NULL, page);
39279cc3 1528 lock_page(page);
211c17f5
CM		 1529 if (page->mapping != mapping) {
1530 unlock_page(page);
1531 page_cache_release(page);
1532 goto again;
1533 }
39279cc3
CM		 1534 if (!PageUptodate(page)) {
1535 ret = -EIO;
89642229 1536 goto out_unlock;
39279cc3
CM		 1537 }
1538 }
211c17f5 1539 wait_on_page_writeback(page);
e6dcd2dc
CM		 1540
1541 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
1542 set_page_extent_mapped(page);
1543
1544 ordered = btrfs_lookup_ordered_extent(inode, page_start);
1545 if (ordered) {
1546 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
1547 unlock_page(page);
1548 page_cache_release(page);
eb84ae03 1549 btrfs_start_ordered_extent(inode, ordered, 1);
e6dcd2dc
CM		 1550 btrfs_put_ordered_extent(ordered);
1551 goto again;
1552 }
1553
ea8c2819 1554 btrfs_set_extent_delalloc(inode, page_start, page_end);
e6dcd2dc
CM		 1555 ret = 0;
1556 if (offset != PAGE_CACHE_SIZE) {
1557 kaddr = kmap(page);
1558 memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
1559 flush_dcache_page(page);
1560 kunmap(page);
1561 }
247e743c 1562 ClearPageChecked(page);
e6dcd2dc
CM		 1563 set_page_dirty(page);
1564 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
39279cc3 1565
89642229 1566out_unlock:
39279cc3
CM		 1567 unlock_page(page);
1568 page_cache_release(page);
1569out:
1570 return ret;
1571}
1572
1573static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
1574{
1575 struct inode *inode = dentry->d_inode;
1576 int err;
1577
1578 err = inode_change_ok(inode, attr);
1579 if (err)
1580 return err;
1581
1582 if (S_ISREG(inode->i_mode) &&
1583 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
1584 struct btrfs_trans_handle *trans;
1585 struct btrfs_root *root = BTRFS_I(inode)->root;
d1310b2e 1586 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2bf5a725 1587
5f39d397 1588 u64 mask = root->sectorsize - 1;
1b0f7c29 1589 u64 hole_start = (inode->i_size + mask) & ~mask;
f392a938 1590 u64 block_end = (attr->ia_size + mask) & ~mask;
39279cc3 1591 u64 hole_size;
179e29e4 1592 u64 alloc_hint = 0;
39279cc3 1593
1b0f7c29 1594 if (attr->ia_size <= hole_start)
39279cc3
CM		 1595 goto out;
1596
1832a6d5 1597 err = btrfs_check_free_space(root, 1, 0);
1832a6d5
CM		 1598 if (err)
1599 goto fail;
1600
39279cc3
CM		 1601 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1602
5f56406a 1603 hole_size = block_end - hole_start;
7c2fe32a
CM		 1604 while(1) {
1605 struct btrfs_ordered_extent *ordered;
1606 btrfs_wait_ordered_range(inode, hole_start, hole_size);
1607
1608 lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1609 ordered = btrfs_lookup_ordered_extent(inode, hole_start);
1610 if (ordered) {
1611 unlock_extent(io_tree, hole_start,
1612 block_end - 1, GFP_NOFS);
1613 btrfs_put_ordered_extent(ordered);
1614 } else {
1615 break;
1616 }
1617 }
39279cc3 1618
39279cc3
CM		 1619 trans = btrfs_start_transaction(root, 1);
1620 btrfs_set_trans_block_group(trans, inode);
ee6e6504 1621 mutex_lock(&BTRFS_I(inode)->extent_mutex);
2bf5a725 1622 err = btrfs_drop_extents(trans, root, inode,
1b0f7c29 1623 hole_start, block_end, hole_start,
3326d1b0 1624 &alloc_hint);
2bf5a725 1625
179e29e4
CM		 1626 if (alloc_hint != EXTENT_MAP_INLINE) {
1627 err = btrfs_insert_file_extent(trans, root,
1628 inode->i_ino,
5f56406a 1629 hole_start, 0, 0,
f2eb0a24 1630 hole_size, 0);
d1310b2e 1631 btrfs_drop_extent_cache(inode, hole_start,
3b951516 1632 (u64)-1);
5f56406a 1633 btrfs_check_file(root, inode);
179e29e4 1634 }
ee6e6504 1635 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
39279cc3 1636 btrfs_end_transaction(trans, root);
1b0f7c29 1637 unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
54aa1f4d
CM		 1638 if (err)
1639 return err;
39279cc3
CM		 1640 }
1641out:
1642 err = inode_setattr(inode, attr);
33268eaf
JB		 1643
1644 if (!err && ((attr->ia_valid & ATTR_MODE)))
1645 err = btrfs_acl_chmod(inode);
1832a6d5 1646fail:
39279cc3
CM		 1647 return err;
1648}
61295eb8 1649
39279cc3
CM		 1650void btrfs_delete_inode(struct inode *inode)
1651{
1652 struct btrfs_trans_handle *trans;
1653 struct btrfs_root *root = BTRFS_I(inode)->root;
d3c2fdcf 1654 unsigned long nr;
39279cc3
CM		 1655 int ret;
1656
1657 truncate_inode_pages(&inode->i_data, 0);
1658 if (is_bad_inode(inode)) {
7b128766 1659 btrfs_orphan_del(NULL, inode);
39279cc3
CM		 1660 goto no_delete;
1661 }
4a096752 1662 btrfs_wait_ordered_range(inode, 0, (u64)-1);
5f39d397 1663
dbe674a9 1664 btrfs_i_size_write(inode, 0);
39279cc3 1665 trans = btrfs_start_transaction(root, 1);
5f39d397 1666
39279cc3 1667 btrfs_set_trans_block_group(trans, inode);
85e21bac 1668 ret = btrfs_truncate_in_trans(trans, root, inode, 0);
7b128766
JB		 1669 if (ret) {
1670 btrfs_orphan_del(NULL, inode);
54aa1f4d 1671 goto no_delete_lock;
7b128766
JB		 1672 }
1673
1674 btrfs_orphan_del(trans, inode);
85e21bac 1675
d3c2fdcf 1676 nr = trans->blocks_used;
85e21bac 1677 clear_inode(inode);
5f39d397 1678
39279cc3 1679 btrfs_end_transaction(trans, root);
d3c2fdcf 1680 btrfs_btree_balance_dirty(root, nr);
39279cc3 1681 return;
54aa1f4d
CM		 1682
1683no_delete_lock:
d3c2fdcf 1684 nr = trans->blocks_used;
54aa1f4d 1685 btrfs_end_transaction(trans, root);
d3c2fdcf 1686 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM		 1687no_delete:
1688 clear_inode(inode);
1689}
1690
1691/*
1692 * this returns the key found in the dir entry in the location pointer.
1693 * If no dir entries were found, location->objectid is 0.
1694 */
1695static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
1696 struct btrfs_key *location)
1697{
1698 const char *name = dentry->d_name.name;
1699 int namelen = dentry->d_name.len;
1700 struct btrfs_dir_item *di;
1701 struct btrfs_path *path;
1702 struct btrfs_root *root = BTRFS_I(dir)->root;
0d9f7f3e 1703 int ret = 0;
39279cc3 1704
3954401f
CM		 1705 if (namelen == 1 && strcmp(name, ".") == 0) {
1706 location->objectid = dir->i_ino;
1707 location->type = BTRFS_INODE_ITEM_KEY;
1708 location->offset = 0;
1709 return 0;
1710 }
39279cc3
CM		 1711 path = btrfs_alloc_path();
1712 BUG_ON(!path);
3954401f 1713
7a720536 1714 if (namelen == 2 && strcmp(name, "..") == 0) {
3954401f
CM		 1715 struct btrfs_key key;
1716 struct extent_buffer *leaf;
3954401f
CM		 1717 int slot;
1718
1719 key.objectid = dir->i_ino;
445dceb7 1720 key.offset = (u64)-1;
3954401f 1721 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
445dceb7
Y		 1722 if (ret < 0 || path->slots[0] == 0)
1723 goto out_err;
3954401f
CM		 1724 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1725 BUG_ON(ret == 0);
1726 ret = 0;
3954401f 1727 leaf = path->nodes[0];
445dceb7 1728 slot = path->slots[0] - 1;
3954401f
CM		 1729
1730 btrfs_item_key_to_cpu(leaf, &key, slot);
1731 if (key.objectid != dir->i_ino ||
1732 key.type != BTRFS_INODE_REF_KEY) {
1733 goto out_err;
1734 }
1735 location->objectid = key.offset;
1736 location->type = BTRFS_INODE_ITEM_KEY;
1737 location->offset = 0;
1738 goto out;
1739 }
1740
39279cc3
CM		 1741 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
1742 namelen, 0);
0d9f7f3e
Y		 1743 if (IS_ERR(di))
1744 ret = PTR_ERR(di);
39279cc3 1745 if (!di || IS_ERR(di)) {
3954401f 1746 goto out_err;
39279cc3 1747 }
5f39d397 1748 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
39279cc3 1749out:
39279cc3
CM		 1750 btrfs_free_path(path);
1751 return ret;
3954401f
CM		 1752out_err:
1753 location->objectid = 0;
1754 goto out;
39279cc3
CM		 1755}
1756
1757/*
1758 * when we hit a tree root in a directory, the btrfs part of the inode
1759 * needs to be changed to reflect the root directory of the tree root. This
1760 * is kind of like crossing a mount point.
1761 */
1762static int fixup_tree_root_location(struct btrfs_root *root,
1763 struct btrfs_key *location,
58176a96
JB		 1764 struct btrfs_root **sub_root,
1765 struct dentry *dentry)
39279cc3 1766{
39279cc3
CM		 1767 struct btrfs_root_item *ri;
1768
1769 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
1770 return 0;
1771 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1772 return 0;
1773
58176a96
JB		 1774 *sub_root = btrfs_read_fs_root(root->fs_info, location,
1775 dentry->d_name.name,
1776 dentry->d_name.len);
39279cc3
CM		 1777 if (IS_ERR(*sub_root))
1778 return PTR_ERR(*sub_root);
1779
1780 ri = &(*sub_root)->root_item;
1781 location->objectid = btrfs_root_dirid(ri);
39279cc3
CM		 1782 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1783 location->offset = 0;
1784
39279cc3
CM		 1785 return 0;
1786}
1787
1788static int btrfs_init_locked_inode(struct inode *inode, void *p)
1789{
1790 struct btrfs_iget_args *args = p;
1791 inode->i_ino = args->ino;
1792 BTRFS_I(inode)->root = args->root;
9069218d 1793 BTRFS_I(inode)->delalloc_bytes = 0;
db69e0eb 1794 inode->i_mapping->writeback_index = 0;
dbe674a9 1795 BTRFS_I(inode)->disk_i_size = 0;
aec7477b 1796 BTRFS_I(inode)->index_cnt = (u64)-1;
d1310b2e
CM		 1797 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1798 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
b888db2b 1799 inode->i_mapping, GFP_NOFS);
7e38326f
CM		 1800 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1801 inode->i_mapping, GFP_NOFS);
ea8c2819 1802 INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
ba1da2f4 1803 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
1b1e2135 1804 mutex_init(&BTRFS_I(inode)->csum_mutex);
ee6e6504 1805 mutex_init(&BTRFS_I(inode)->extent_mutex);
39279cc3
CM		 1806 return 0;
1807}
1808
1809static int btrfs_find_actor(struct inode *inode, void *opaque)
1810{
1811 struct btrfs_iget_args *args = opaque;
1812 return (args->ino == inode->i_ino &&
1813 args->root == BTRFS_I(inode)->root);
1814}
1815
dc17ff8f
CM		 1816struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
1817 u64 root_objectid)
1818{
1819 struct btrfs_iget_args args;
1820 args.ino = objectid;
1821 args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
1822
1823 if (!args.root)
1824 return NULL;
1825
1826 return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
1827}
1828
39279cc3
CM		 1829struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
1830 struct btrfs_root *root)
1831{
1832 struct inode *inode;
1833 struct btrfs_iget_args args;
1834 args.ino = objectid;
1835 args.root = root;
1836
1837 inode = iget5_locked(s, objectid, btrfs_find_actor,
1838 btrfs_init_locked_inode,
1839 (void *)&args);
1840 return inode;
1841}
1842
1843static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
1844 struct nameidata *nd)
1845{
1846 struct inode * inode;
1847 struct btrfs_inode *bi = BTRFS_I(dir);
1848 struct btrfs_root *root = bi->root;
1849 struct btrfs_root *sub_root = root;
1850 struct btrfs_key location;
7b128766 1851 int ret, do_orphan = 0;
39279cc3
CM		 1852
1853 if (dentry->d_name.len > BTRFS_NAME_LEN)
1854 return ERR_PTR(-ENAMETOOLONG);
5f39d397 1855
39279cc3 1856 ret = btrfs_inode_by_name(dir, dentry, &location);
5f39d397 1857
39279cc3
CM		 1858 if (ret < 0)
1859 return ERR_PTR(ret);
5f39d397 1860
39279cc3
CM		 1861 inode = NULL;
1862 if (location.objectid) {
58176a96
JB		 1863 ret = fixup_tree_root_location(root, &location, &sub_root,
1864 dentry);
39279cc3
CM		 1865 if (ret < 0)
1866 return ERR_PTR(ret);
1867 if (ret > 0)
1868 return ERR_PTR(-ENOENT);
7b128766 1869
39279cc3
CM		 1870 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
1871 sub_root);
1872 if (!inode)
1873 return ERR_PTR(-EACCES);
1874 if (inode->i_state & I_NEW) {
1875 /* the inode and parent dir are two different roots */
1876 if (sub_root != root) {
1877 igrab(inode);
1878 sub_root->inode = inode;
7b128766 1879 do_orphan = 1;
39279cc3
CM		 1880 }
1881 BTRFS_I(inode)->root = sub_root;
1882 memcpy(&BTRFS_I(inode)->location, &location,
1883 sizeof(location));
1884 btrfs_read_locked_inode(inode);
1885 unlock_new_inode(inode);
1886 }
1887 }
7b128766
JB		 1888
1889 if (unlikely(do_orphan))
1890 btrfs_orphan_cleanup(sub_root);
1891
39279cc3
CM		 1892 return d_splice_alias(inode, dentry);
1893}
1894
39279cc3
CM		 1895static unsigned char btrfs_filetype_table[] = {
1896 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1897};
1898
1899static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1900{
6da6abae 1901 struct inode *inode = filp->f_dentry->d_inode;
39279cc3
CM		 1902 struct btrfs_root *root = BTRFS_I(inode)->root;
1903 struct btrfs_item *item;
1904 struct btrfs_dir_item *di;
1905 struct btrfs_key key;
5f39d397 1906 struct btrfs_key found_key;
39279cc3
CM		 1907 struct btrfs_path *path;
1908 int ret;
1909 u32 nritems;
5f39d397 1910 struct extent_buffer *leaf;
39279cc3
CM		 1911 int slot;
1912 int advance;
1913 unsigned char d_type;
1914 int over = 0;
1915 u32 di_cur;
1916 u32 di_total;
1917 u32 di_len;
1918 int key_type = BTRFS_DIR_INDEX_KEY;
5f39d397
CM		 1919 char tmp_name[32];
1920 char *name_ptr;
1921 int name_len;
39279cc3
CM		 1922
1923 /* FIXME, use a real flag for deciding about the key type */
1924 if (root->fs_info->tree_root == root)
1925 key_type = BTRFS_DIR_ITEM_KEY;
5f39d397 1926
3954401f
CM		 1927 /* special case for "." */
1928 if (filp->f_pos == 0) {
1929 over = filldir(dirent, ".", 1,
1930 1, inode->i_ino,
1931 DT_DIR);
1932 if (over)
1933 return 0;
1934 filp->f_pos = 1;
1935 }
1936
39279cc3 1937 key.objectid = inode->i_ino;
3954401f
CM		 1938 path = btrfs_alloc_path();
1939 path->reada = 2;
1940
1941 /* special case for .., just use the back ref */
1942 if (filp->f_pos == 1) {
1943 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
445dceb7 1944 key.offset = (u64)-1;
3954401f 1945 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
445dceb7 1946 if (ret < 0 || path->slots[0] == 0) {
3954401f
CM		 1947 btrfs_release_path(root, path);
1948 goto read_dir_items;
1949 }
445dceb7
Y		 1950 BUG_ON(ret == 0);
1951 leaf = path->nodes[0];
1952 slot = path->slots[0] - 1;
3954401f
CM		 1953 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1954 btrfs_release_path(root, path);
1955 if (found_key.objectid != key.objectid ||
1956 found_key.type != BTRFS_INODE_REF_KEY)
1957 goto read_dir_items;
1958 over = filldir(dirent, "..", 2,
1959 2, found_key.offset, DT_DIR);
1960 if (over)
1961 goto nopos;
1962 filp->f_pos = 2;
1963 }
1964
1965read_dir_items:
39279cc3
CM		 1966 btrfs_set_key_type(&key, key_type);
1967 key.offset = filp->f_pos;
5f39d397 1968
39279cc3
CM		 1969 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1970 if (ret < 0)
1971 goto err;
1972 advance = 0;
39279cc3 1973 while(1) {
5f39d397
CM		 1974 leaf = path->nodes[0];
1975 nritems = btrfs_header_nritems(leaf);
39279cc3
CM		 1976 slot = path->slots[0];
1977 if (advance || slot >= nritems) {
1978 if (slot >= nritems -1) {
39279cc3
CM		 1979 ret = btrfs_next_leaf(root, path);
1980 if (ret)
1981 break;
5f39d397
CM		 1982 leaf = path->nodes[0];
1983 nritems = btrfs_header_nritems(leaf);
39279cc3
CM		 1984 slot = path->slots[0];
1985 } else {
1986 slot++;
1987 path->slots[0]++;
1988 }
1989 }
1990 advance = 1;
5f39d397
CM		 1991 item = btrfs_item_nr(leaf, slot);
1992 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1993
1994 if (found_key.objectid != key.objectid)
39279cc3 1995 break;
5f39d397 1996 if (btrfs_key_type(&found_key) != key_type)
39279cc3 1997 break;
5f39d397 1998 if (found_key.offset < filp->f_pos)
39279cc3 1999 continue;
5f39d397
CM		 2000
2001 filp->f_pos = found_key.offset;
39279cc3
CM		 2002 advance = 1;
2003 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
2004 di_cur = 0;
5f39d397 2005 di_total = btrfs_item_size(leaf, item);
39279cc3 2006 while(di_cur < di_total) {
5f39d397
CM		 2007 struct btrfs_key location;
2008
2009 name_len = btrfs_dir_name_len(leaf, di);
2010 if (name_len < 32) {
2011 name_ptr = tmp_name;
2012 } else {
2013 name_ptr = kmalloc(name_len, GFP_NOFS);
2014 BUG_ON(!name_ptr);
2015 }
2016 read_extent_buffer(leaf, name_ptr,
2017 (unsigned long)(di + 1), name_len);
2018
2019 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
2020 btrfs_dir_item_key_to_cpu(leaf, di, &location);
5f39d397
CM		 2021 over = filldir(dirent, name_ptr, name_len,
2022 found_key.offset,
2023 location.objectid,
39279cc3 2024 d_type);
5f39d397
CM		 2025
2026 if (name_ptr != tmp_name)
2027 kfree(name_ptr);
2028
39279cc3
CM		 2029 if (over)
2030 goto nopos;
5103e947
JB		 2031 di_len = btrfs_dir_name_len(leaf, di) +
2032 btrfs_dir_data_len(leaf, di) +sizeof(*di);
39279cc3
CM		 2033 di_cur += di_len;
2034 di = (struct btrfs_dir_item *)((char *)di + di_len);
2035 }
2036 }
5e591a07
YZ		 2037 if (key_type == BTRFS_DIR_INDEX_KEY)
2038 filp->f_pos = INT_LIMIT(typeof(filp->f_pos));
2039 else
2040 filp->f_pos++;
39279cc3
CM		 2041nopos:
2042 ret = 0;
2043err:
39279cc3 2044 btrfs_free_path(path);
39279cc3
CM		 2045 return ret;
2046}
2047
2048int btrfs_write_inode(struct inode *inode, int wait)
2049{
2050 struct btrfs_root *root = BTRFS_I(inode)->root;
2051 struct btrfs_trans_handle *trans;
2052 int ret = 0;
2053
4ca8b41e
CM		 2054 if (root->fs_info->closing > 1)
2055 return 0;
2056
39279cc3 2057 if (wait) {
f9295749 2058 trans = btrfs_join_transaction(root, 1);
39279cc3
CM		 2059 btrfs_set_trans_block_group(trans, inode);
2060 ret = btrfs_commit_transaction(trans, root);
39279cc3
CM		 2061 }
2062 return ret;
2063}
2064
2065/*
54aa1f4d 2066 * This is somewhat expensive, updating the tree every time the
39279cc3
CM		 2067 * inode changes. But, it is most likely to find the inode in cache.
2068 * FIXME, needs more benchmarking...there are no reasons other than performance
2069 * to keep or drop this code.
2070 */
2071void btrfs_dirty_inode(struct inode *inode)
2072{
2073 struct btrfs_root *root = BTRFS_I(inode)->root;
2074 struct btrfs_trans_handle *trans;
2075
f9295749 2076 trans = btrfs_join_transaction(root, 1);
39279cc3
CM		 2077 btrfs_set_trans_block_group(trans, inode);
2078 btrfs_update_inode(trans, root, inode);
2079 btrfs_end_transaction(trans, root);
39279cc3
CM		 2080}
2081
aec7477b
JB		 2082static int btrfs_set_inode_index_count(struct inode *inode)
2083{
2084 struct btrfs_root *root = BTRFS_I(inode)->root;
2085 struct btrfs_key key, found_key;
2086 struct btrfs_path *path;
2087 struct extent_buffer *leaf;
2088 int ret;
2089
2090 key.objectid = inode->i_ino;
2091 btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY);
2092 key.offset = (u64)-1;
2093
2094 path = btrfs_alloc_path();
2095 if (!path)
2096 return -ENOMEM;
2097
2098 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2099 if (ret < 0)
2100 goto out;
2101 /* FIXME: we should be able to handle this */
2102 if (ret == 0)
2103 goto out;
2104 ret = 0;
2105
2106 /*
2107 * MAGIC NUMBER EXPLANATION:
2108 * since we search a directory based on f_pos we have to start at 2
2109 * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody
2110 * else has to start at 2
2111 */
2112 if (path->slots[0] == 0) {
2113 BTRFS_I(inode)->index_cnt = 2;
2114 goto out;
2115 }
2116
2117 path->slots[0]--;
2118
2119 leaf = path->nodes[0];
2120 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2121
2122 if (found_key.objectid != inode->i_ino ||
2123 btrfs_key_type(&found_key) != BTRFS_DIR_INDEX_KEY) {
2124 BTRFS_I(inode)->index_cnt = 2;
2125 goto out;
2126 }
2127
2128 BTRFS_I(inode)->index_cnt = found_key.offset + 1;
2129out:
2130 btrfs_free_path(path);
2131 return ret;
2132}
2133
00e4e6b3
CM		 2134static int btrfs_set_inode_index(struct inode *dir, struct inode *inode,
2135 u64 *index)
aec7477b
JB		 2136{
2137 int ret = 0;
2138
2139 if (BTRFS_I(dir)->index_cnt == (u64)-1) {
2140 ret = btrfs_set_inode_index_count(dir);
2141 if (ret)
2142 return ret;
2143 }
2144
00e4e6b3 2145 *index = BTRFS_I(dir)->index_cnt;
aec7477b
JB		 2146 BTRFS_I(dir)->index_cnt++;
2147
2148 return ret;
2149}
2150
39279cc3
CM		 2151static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
2152 struct btrfs_root *root,
aec7477b 2153 struct inode *dir,
9c58309d
CM		 2154 const char *name, int name_len,
2155 u64 ref_objectid,
39279cc3
CM		 2156 u64 objectid,
2157 struct btrfs_block_group_cache *group,
00e4e6b3 2158 int mode, u64 *index)
39279cc3
CM		 2159{
2160 struct inode *inode;
5f39d397 2161 struct btrfs_inode_item *inode_item;
6324fbf3 2162 struct btrfs_block_group_cache *new_inode_group;
39279cc3 2163 struct btrfs_key *location;
5f39d397 2164 struct btrfs_path *path;
9c58309d
CM		 2165 struct btrfs_inode_ref *ref;
2166 struct btrfs_key key[2];
2167 u32 sizes[2];
2168 unsigned long ptr;
39279cc3
CM		 2169 int ret;
2170 int owner;
2171
5f39d397
CM		 2172 path = btrfs_alloc_path();
2173 BUG_ON(!path);
2174
39279cc3
CM		 2175 inode = new_inode(root->fs_info->sb);
2176 if (!inode)
2177 return ERR_PTR(-ENOMEM);
2178
aec7477b 2179 if (dir) {
00e4e6b3 2180 ret = btrfs_set_inode_index(dir, inode, index);
aec7477b
JB		 2181 if (ret)
2182 return ERR_PTR(ret);
aec7477b
JB		 2183 }
2184 /*
2185 * index_cnt is ignored for everything but a dir,
2186 * btrfs_get_inode_index_count has an explanation for the magic
2187 * number
2188 */
2189 BTRFS_I(inode)->index_cnt = 2;
2190
d1310b2e
CM		 2191 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2192 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
b888db2b 2193 inode->i_mapping, GFP_NOFS);
7e38326f
CM		 2194 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
2195 inode->i_mapping, GFP_NOFS);
ba1da2f4 2196 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
ea8c2819 2197 INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
1b1e2135 2198 mutex_init(&BTRFS_I(inode)->csum_mutex);
ee6e6504 2199 mutex_init(&BTRFS_I(inode)->extent_mutex);
9069218d 2200 BTRFS_I(inode)->delalloc_bytes = 0;
db69e0eb 2201 inode->i_mapping->writeback_index = 0;
dbe674a9 2202 BTRFS_I(inode)->disk_i_size = 0;
39279cc3 2203 BTRFS_I(inode)->root = root;
b888db2b 2204
39279cc3
CM		 2205 if (mode & S_IFDIR)
2206 owner = 0;
2207 else
2208 owner = 1;
6324fbf3 2209 new_inode_group = btrfs_find_block_group(root, group, 0,
0b86a832 2210 BTRFS_BLOCK_GROUP_METADATA, owner);
6324fbf3
CM		 2211 if (!new_inode_group) {
2212 printk("find_block group failed\n");
2213 new_inode_group = group;
2214 }
2215 BTRFS_I(inode)->block_group = new_inode_group;
b98b6767 2216 BTRFS_I(inode)->flags = 0;
9c58309d
CM		 2217
2218 key[0].objectid = objectid;
2219 btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
2220 key[0].offset = 0;
2221
2222 key[1].objectid = objectid;
2223 btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
2224 key[1].offset = ref_objectid;
2225
2226 sizes[0] = sizeof(struct btrfs_inode_item);
2227 sizes[1] = name_len + sizeof(*ref);
2228
2229 ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
2230 if (ret != 0)
5f39d397
CM		 2231 goto fail;
2232
9c58309d
CM		 2233 if (objectid > root->highest_inode)
2234 root->highest_inode = objectid;
2235
39279cc3
CM		 2236 inode->i_uid = current->fsuid;
2237 inode->i_gid = current->fsgid;
2238 inode->i_mode = mode;
2239 inode->i_ino = objectid;
2240 inode->i_blocks = 0;
2241 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
5f39d397
CM		 2242 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
2243 struct btrfs_inode_item);
2244 fill_inode_item(path->nodes[0], inode_item, inode);
9c58309d
CM		 2245
2246 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
2247 struct btrfs_inode_ref);
2248 btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
00e4e6b3 2249 btrfs_set_inode_ref_index(path->nodes[0], ref, *index);
9c58309d
CM		 2250 ptr = (unsigned long)(ref + 1);
2251 write_extent_buffer(path->nodes[0], name, ptr, name_len);
2252
5f39d397
CM		 2253 btrfs_mark_buffer_dirty(path->nodes[0]);
2254 btrfs_free_path(path);
2255
39279cc3
CM		 2256 location = &BTRFS_I(inode)->location;
2257 location->objectid = objectid;
39279cc3
CM		 2258 location->offset = 0;
2259 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
2260
39279cc3
CM		 2261 insert_inode_hash(inode);
2262 return inode;
5f39d397 2263fail:
aec7477b
JB		 2264 if (dir)
2265 BTRFS_I(dir)->index_cnt--;
5f39d397
CM		 2266 btrfs_free_path(path);
2267 return ERR_PTR(ret);
39279cc3
CM		 2268}
2269
2270static inline u8 btrfs_inode_type(struct inode *inode)
2271{
2272 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
2273}
2274
2275static int btrfs_add_link(struct btrfs_trans_handle *trans,
9c58309d 2276 struct dentry *dentry, struct inode *inode,
00e4e6b3 2277 int add_backref, u64 index)
39279cc3
CM		 2278{
2279 int ret;
2280 struct btrfs_key key;
2281 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
aec7477b 2282 struct inode *parent_inode = dentry->d_parent->d_inode;
5f39d397 2283
39279cc3 2284 key.objectid = inode->i_ino;
39279cc3
CM		 2285 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
2286 key.offset = 0;
2287
2288 ret = btrfs_insert_dir_item(trans, root,
2289 dentry->d_name.name, dentry->d_name.len,
2290 dentry->d_parent->d_inode->i_ino,
aec7477b 2291 &key, btrfs_inode_type(inode),
00e4e6b3 2292 index);
39279cc3 2293 if (ret == 0) {
9c58309d
CM		 2294 if (add_backref) {
2295 ret = btrfs_insert_inode_ref(trans, root,
2296 dentry->d_name.name,
2297 dentry->d_name.len,
2298 inode->i_ino,
aec7477b 2299 parent_inode->i_ino,
00e4e6b3 2300 index);
9c58309d 2301 }
dbe674a9
CM		 2302 btrfs_i_size_write(parent_inode, parent_inode->i_size +
2303 dentry->d_name.len * 2);
79c44584 2304 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
39279cc3
CM		 2305 ret = btrfs_update_inode(trans, root,
2306 dentry->d_parent->d_inode);
2307 }
2308 return ret;
2309}
2310
2311static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
9c58309d 2312 struct dentry *dentry, struct inode *inode,
00e4e6b3 2313 int backref, u64 index)
39279cc3 2314{
00e4e6b3 2315 int err = btrfs_add_link(trans, dentry, inode, backref, index);
39279cc3
CM		 2316 if (!err) {
2317 d_instantiate(dentry, inode);
2318 return 0;
2319 }
2320 if (err > 0)
2321 err = -EEXIST;
2322 return err;
2323}
2324
618e21d5
JB		 2325static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
2326 int mode, dev_t rdev)
2327{
2328 struct btrfs_trans_handle *trans;
2329 struct btrfs_root *root = BTRFS_I(dir)->root;
1832a6d5 2330 struct inode *inode = NULL;
618e21d5
JB		 2331 int err;
2332 int drop_inode = 0;
2333 u64 objectid;
1832a6d5 2334 unsigned long nr = 0;
00e4e6b3 2335 u64 index = 0;
618e21d5
JB		 2336
2337 if (!new_valid_dev(rdev))
2338 return -EINVAL;
2339
1832a6d5
CM		 2340 err = btrfs_check_free_space(root, 1, 0);
2341 if (err)
2342 goto fail;
2343
618e21d5
JB		 2344 trans = btrfs_start_transaction(root, 1);
2345 btrfs_set_trans_block_group(trans, dir);
2346
2347 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2348 if (err) {
2349 err = -ENOSPC;
2350 goto out_unlock;
2351 }
2352
aec7477b 2353 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
9c58309d
CM		 2354 dentry->d_name.len,
2355 dentry->d_parent->d_inode->i_ino, objectid,
00e4e6b3 2356 BTRFS_I(dir)->block_group, mode, &index);
618e21d5
JB		 2357 err = PTR_ERR(inode);
2358 if (IS_ERR(inode))
2359 goto out_unlock;
2360
33268eaf
JB		 2361 err = btrfs_init_acl(inode, dir);
2362 if (err) {
2363 drop_inode = 1;
2364 goto out_unlock;
2365 }
2366
618e21d5 2367 btrfs_set_trans_block_group(trans, inode);
00e4e6b3 2368 err = btrfs_add_nondir(trans, dentry, inode, 0, index);
618e21d5
JB		 2369 if (err)
2370 drop_inode = 1;
2371 else {
2372 inode->i_op = &btrfs_special_inode_operations;
2373 init_special_inode(inode, inode->i_mode, rdev);
1b4ab1bb 2374 btrfs_update_inode(trans, root, inode);
618e21d5
JB		 2375 }
2376 dir->i_sb->s_dirt = 1;
2377 btrfs_update_inode_block_group(trans, inode);
2378 btrfs_update_inode_block_group(trans, dir);
2379out_unlock:
d3c2fdcf 2380 nr = trans->blocks_used;
89ce8a63 2381 btrfs_end_transaction_throttle(trans, root);
1832a6d5 2382fail:
618e21d5
JB		 2383 if (drop_inode) {
2384 inode_dec_link_count(inode);
2385 iput(inode);
2386 }
d3c2fdcf 2387 btrfs_btree_balance_dirty(root, nr);
618e21d5
JB		 2388 return err;
2389}
2390
39279cc3
CM		 2391static int btrfs_create(struct inode *dir, struct dentry *dentry,
2392 int mode, struct nameidata *nd)
2393{
2394 struct btrfs_trans_handle *trans;
2395 struct btrfs_root *root = BTRFS_I(dir)->root;
1832a6d5 2396 struct inode *inode = NULL;
39279cc3
CM		 2397 int err;
2398 int drop_inode = 0;
1832a6d5 2399 unsigned long nr = 0;
39279cc3 2400 u64 objectid;
00e4e6b3 2401 u64 index = 0;
39279cc3 2402
1832a6d5
CM		 2403 err = btrfs_check_free_space(root, 1, 0);
2404 if (err)
2405 goto fail;
39279cc3
CM		 2406 trans = btrfs_start_transaction(root, 1);
2407 btrfs_set_trans_block_group(trans, dir);
2408
2409 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2410 if (err) {
2411 err = -ENOSPC;
2412 goto out_unlock;
2413 }
2414
aec7477b 2415 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
9c58309d
CM		 2416 dentry->d_name.len,
2417 dentry->d_parent->d_inode->i_ino,
00e4e6b3
CM		 2418 objectid, BTRFS_I(dir)->block_group, mode,
2419 &index);
39279cc3
CM		 2420 err = PTR_ERR(inode);
2421 if (IS_ERR(inode))
2422 goto out_unlock;
2423
33268eaf
JB		 2424 err = btrfs_init_acl(inode, dir);
2425 if (err) {
2426 drop_inode = 1;
2427 goto out_unlock;
2428 }
2429
39279cc3 2430 btrfs_set_trans_block_group(trans, inode);
00e4e6b3 2431 err = btrfs_add_nondir(trans, dentry, inode, 0, index);
39279cc3
CM		 2432 if (err)
2433 drop_inode = 1;
2434 else {
2435 inode->i_mapping->a_ops = &btrfs_aops;
04160088 2436 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
39279cc3
CM		 2437 inode->i_fop = &btrfs_file_operations;
2438 inode->i_op = &btrfs_file_inode_operations;
d1310b2e
CM		 2439 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2440 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
a52d9a80 2441 inode->i_mapping, GFP_NOFS);
7e38326f
CM		 2442 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
2443 inode->i_mapping, GFP_NOFS);
ea8c2819 2444 INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
1b1e2135 2445 mutex_init(&BTRFS_I(inode)->csum_mutex);
ee6e6504 2446 mutex_init(&BTRFS_I(inode)->extent_mutex);
9069218d 2447 BTRFS_I(inode)->delalloc_bytes = 0;
dbe674a9 2448 BTRFS_I(inode)->disk_i_size = 0;
db69e0eb 2449 inode->i_mapping->writeback_index = 0;
d1310b2e 2450 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
ba1da2f4 2451 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
39279cc3
CM		 2452 }
2453 dir->i_sb->s_dirt = 1;
2454 btrfs_update_inode_block_group(trans, inode);
2455 btrfs_update_inode_block_group(trans, dir);
2456out_unlock:
d3c2fdcf 2457 nr = trans->blocks_used;
ab78c84d 2458 btrfs_end_transaction_throttle(trans, root);
1832a6d5 2459fail:
39279cc3
CM		 2460 if (drop_inode) {
2461 inode_dec_link_count(inode);
2462 iput(inode);
2463 }
d3c2fdcf 2464 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM		 2465 return err;
2466}
2467
2468static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
2469 struct dentry *dentry)
2470{
2471 struct btrfs_trans_handle *trans;
2472 struct btrfs_root *root = BTRFS_I(dir)->root;
2473 struct inode *inode = old_dentry->d_inode;
00e4e6b3 2474 u64 index;
1832a6d5 2475 unsigned long nr = 0;
39279cc3
CM		 2476 int err;
2477 int drop_inode = 0;
2478
2479 if (inode->i_nlink == 0)
2480 return -ENOENT;
2481
6da6abae
CM		 2482#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2483 inode->i_nlink++;
2484#else
39279cc3 2485 inc_nlink(inode);
6da6abae 2486#endif
1832a6d5
CM		 2487 err = btrfs_check_free_space(root, 1, 0);
2488 if (err)
2489 goto fail;
00e4e6b3 2490 err = btrfs_set_inode_index(dir, inode, &index);
aec7477b
JB		 2491 if (err)
2492 goto fail;
2493
39279cc3 2494 trans = btrfs_start_transaction(root, 1);
5f39d397 2495
39279cc3
CM		 2496 btrfs_set_trans_block_group(trans, dir);
2497 atomic_inc(&inode->i_count);
aec7477b 2498
00e4e6b3 2499 err = btrfs_add_nondir(trans, dentry, inode, 1, index);
5f39d397 2500
39279cc3
CM		 2501 if (err)
2502 drop_inode = 1;
5f39d397 2503
39279cc3
CM		 2504 dir->i_sb->s_dirt = 1;
2505 btrfs_update_inode_block_group(trans, dir);
54aa1f4d 2506 err = btrfs_update_inode(trans, root, inode);
5f39d397 2507
54aa1f4d
CM		 2508 if (err)
2509 drop_inode = 1;
39279cc3 2510
d3c2fdcf 2511 nr = trans->blocks_used;
ab78c84d 2512 btrfs_end_transaction_throttle(trans, root);
1832a6d5 2513fail:
39279cc3
CM		 2514 if (drop_inode) {
2515 inode_dec_link_count(inode);
2516 iput(inode);
2517 }
d3c2fdcf 2518 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM		 2519 return err;
2520}
2521
39279cc3
CM		 2522static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
2523{
b9d86667 2524 struct inode *inode = NULL;
39279cc3
CM		 2525 struct btrfs_trans_handle *trans;
2526 struct btrfs_root *root = BTRFS_I(dir)->root;
2527 int err = 0;
2528 int drop_on_err = 0;
b9d86667 2529 u64 objectid = 0;
00e4e6b3 2530 u64 index = 0;
d3c2fdcf 2531 unsigned long nr = 1;
39279cc3 2532
1832a6d5
CM		 2533 err = btrfs_check_free_space(root, 1, 0);
2534 if (err)
2535 goto out_unlock;
2536
39279cc3
CM		 2537 trans = btrfs_start_transaction(root, 1);
2538 btrfs_set_trans_block_group(trans, dir);
5f39d397 2539
39279cc3
CM		 2540 if (IS_ERR(trans)) {
2541 err = PTR_ERR(trans);
2542 goto out_unlock;
2543 }
2544
2545 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2546 if (err) {
2547 err = -ENOSPC;
2548 goto out_unlock;
2549 }
2550
aec7477b 2551 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
9c58309d
CM		 2552 dentry->d_name.len,
2553 dentry->d_parent->d_inode->i_ino, objectid,
00e4e6b3
CM		 2554 BTRFS_I(dir)->block_group, S_IFDIR | mode,
2555 &index);
39279cc3
CM		 2556 if (IS_ERR(inode)) {
2557 err = PTR_ERR(inode);
2558 goto out_fail;
2559 }
5f39d397 2560
39279cc3 2561 drop_on_err = 1;
33268eaf
JB		 2562
2563 err = btrfs_init_acl(inode, dir);
2564 if (err)
2565 goto out_fail;
2566
39279cc3
CM		 2567 inode->i_op = &btrfs_dir_inode_operations;
2568 inode->i_fop = &btrfs_dir_file_operations;
2569 btrfs_set_trans_block_group(trans, inode);
2570
dbe674a9 2571 btrfs_i_size_write(inode, 0);
39279cc3
CM		 2572 err = btrfs_update_inode(trans, root, inode);
2573 if (err)
2574 goto out_fail;
5f39d397 2575
00e4e6b3 2576 err = btrfs_add_link(trans, dentry, inode, 0, index);
39279cc3
CM		 2577 if (err)
2578 goto out_fail;
5f39d397 2579
39279cc3
CM		 2580 d_instantiate(dentry, inode);
2581 drop_on_err = 0;
2582 dir->i_sb->s_dirt = 1;
2583 btrfs_update_inode_block_group(trans, inode);
2584 btrfs_update_inode_block_group(trans, dir);
2585
2586out_fail:
d3c2fdcf 2587 nr = trans->blocks_used;
ab78c84d 2588 btrfs_end_transaction_throttle(trans, root);
5f39d397 2589
39279cc3 2590out_unlock:
39279cc3
CM		 2591 if (drop_on_err)
2592 iput(inode);
d3c2fdcf 2593 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM		 2594 return err;
2595}
2596
3b951516
CM		 2597static int merge_extent_mapping(struct extent_map_tree *em_tree,
2598 struct extent_map *existing,
e6dcd2dc
CM		 2599 struct extent_map *em,
2600 u64 map_start, u64 map_len)
3b951516
CM		 2601{
2602 u64 start_diff;
3b951516 2603
e6dcd2dc
CM		 2604 BUG_ON(map_start < em->start || map_start >= extent_map_end(em));
2605 start_diff = map_start - em->start;
2606 em->start = map_start;
2607 em->len = map_len;
2608 if (em->block_start < EXTENT_MAP_LAST_BYTE)
2609 em->block_start += start_diff;
2610 return add_extent_mapping(em_tree, em);
3b951516
CM		 2611}
2612
a52d9a80 2613struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
70dec807 2614 size_t pg_offset, u64 start, u64 len,
a52d9a80
CM		 2615 int create)
2616{
2617 int ret;
2618 int err = 0;
db94535d 2619 u64 bytenr;
a52d9a80
CM		 2620 u64 extent_start = 0;
2621 u64 extent_end = 0;
2622 u64 objectid = inode->i_ino;
2623 u32 found_type;
f421950f 2624 struct btrfs_path *path = NULL;
a52d9a80
CM		 2625 struct btrfs_root *root = BTRFS_I(inode)->root;
2626 struct btrfs_file_extent_item *item;
5f39d397
CM		 2627 struct extent_buffer *leaf;
2628 struct btrfs_key found_key;
a52d9a80
CM		 2629 struct extent_map *em = NULL;
2630 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
d1310b2e 2631 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
a52d9a80
CM		 2632 struct btrfs_trans_handle *trans = NULL;
2633
a52d9a80 2634again:
d1310b2e
CM		 2635 spin_lock(&em_tree->lock);
2636 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 2637 if (em)
2638 em->bdev = root->fs_info->fs_devices->latest_bdev;
d1310b2e
CM		 2639 spin_unlock(&em_tree->lock);
2640
a52d9a80 2641 if (em) {
e1c4b745
CM		 2642 if (em->start > start || em->start + em->len <= start)
2643 free_extent_map(em);
2644 else if (em->block_start == EXTENT_MAP_INLINE && page)
70dec807
CM		 2645 free_extent_map(em);
2646 else
2647 goto out;
a52d9a80 2648 }
d1310b2e 2649 em = alloc_extent_map(GFP_NOFS);
a52d9a80 2650 if (!em) {
d1310b2e
CM		 2651 err = -ENOMEM;
2652 goto out;
a52d9a80 2653 }
e6dcd2dc 2654 em->bdev = root->fs_info->fs_devices->latest_bdev;
d1310b2e
CM		 2655 em->start = EXTENT_MAP_HOLE;
2656 em->len = (u64)-1;
f421950f
CM		 2657
2658 if (!path) {
2659 path = btrfs_alloc_path();
2660 BUG_ON(!path);
2661 }
2662
179e29e4
CM		 2663 ret = btrfs_lookup_file_extent(trans, root, path,
2664 objectid, start, trans != NULL);
a52d9a80
CM		 2665 if (ret < 0) {
2666 err = ret;
2667 goto out;
2668 }
2669
2670 if (ret != 0) {
2671 if (path->slots[0] == 0)
2672 goto not_found;
2673 path->slots[0]--;
2674 }
2675
5f39d397
CM		 2676 leaf = path->nodes[0];
2677 item = btrfs_item_ptr(leaf, path->slots[0],
a52d9a80 2678 struct btrfs_file_extent_item);
a52d9a80 2679 /* are we inside the extent that was found? */
5f39d397
CM		 2680 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2681 found_type = btrfs_key_type(&found_key);
2682 if (found_key.objectid != objectid ||
a52d9a80
CM		 2683 found_type != BTRFS_EXTENT_DATA_KEY) {
2684 goto not_found;
2685 }
2686
5f39d397
CM		 2687 found_type = btrfs_file_extent_type(leaf, item);
2688 extent_start = found_key.offset;
a52d9a80
CM		 2689 if (found_type == BTRFS_FILE_EXTENT_REG) {
2690 extent_end = extent_start +
db94535d 2691 btrfs_file_extent_num_bytes(leaf, item);
a52d9a80 2692 err = 0;
b888db2b 2693 if (start < extent_start || start >= extent_end) {
a52d9a80
CM		 2694 em->start = start;
2695 if (start < extent_start) {
d1310b2e 2696 if (start + len <= extent_start)
b888db2b 2697 goto not_found;
d1310b2e 2698 em->len = extent_end - extent_start;
a52d9a80 2699 } else {
d1310b2e 2700 em->len = len;
a52d9a80
CM		 2701 }
2702 goto not_found_em;
2703 }
db94535d
CM		 2704 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
2705 if (bytenr == 0) {
a52d9a80 2706 em->start = extent_start;
d1310b2e 2707 em->len = extent_end - extent_start;
5f39d397 2708 em->block_start = EXTENT_MAP_HOLE;
a52d9a80
CM		 2709 goto insert;
2710 }
db94535d
CM		 2711 bytenr += btrfs_file_extent_offset(leaf, item);
2712 em->block_start = bytenr;
a52d9a80 2713 em->start = extent_start;
d1310b2e 2714 em->len = extent_end - extent_start;
a52d9a80
CM		 2715 goto insert;
2716 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
70dec807 2717 u64 page_start;
5f39d397 2718 unsigned long ptr;
a52d9a80 2719 char *map;
3326d1b0
CM		 2720 size_t size;
2721 size_t extent_offset;
2722 size_t copy_size;
a52d9a80 2723
5f39d397
CM		 2724 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
2725 path->slots[0]));
d1310b2e
CM		 2726 extent_end = (extent_start + size + root->sectorsize - 1) &
2727 ~((u64)root->sectorsize - 1);
b888db2b 2728 if (start < extent_start || start >= extent_end) {
a52d9a80
CM		 2729 em->start = start;
2730 if (start < extent_start) {
d1310b2e 2731 if (start + len <= extent_start)
b888db2b 2732 goto not_found;
d1310b2e 2733 em->len = extent_end - extent_start;
a52d9a80 2734 } else {
d1310b2e 2735 em->len = len;
a52d9a80
CM		 2736 }
2737 goto not_found_em;
2738 }
689f9346 2739 em->block_start = EXTENT_MAP_INLINE;
689f9346
Y		 2740
2741 if (!page) {
2742 em->start = extent_start;
d1310b2e 2743 em->len = size;
689f9346
Y		 2744 goto out;
2745 }
5f39d397 2746
70dec807
CM		 2747 page_start = page_offset(page) + pg_offset;
2748 extent_offset = page_start - extent_start;
2749 copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
3326d1b0 2750 size - extent_offset);
3326d1b0 2751 em->start = extent_start + extent_offset;
70dec807
CM		 2752 em->len = (copy_size + root->sectorsize - 1) &
2753 ~((u64)root->sectorsize - 1);
689f9346
Y		 2754 map = kmap(page);
2755 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
179e29e4 2756 if (create == 0 && !PageUptodate(page)) {
70dec807 2757 read_extent_buffer(leaf, map + pg_offset, ptr,
179e29e4
CM		 2758 copy_size);
2759 flush_dcache_page(page);
2760 } else if (create && PageUptodate(page)) {
2761 if (!trans) {
2762 kunmap(page);
2763 free_extent_map(em);
2764 em = NULL;
2765 btrfs_release_path(root, path);
f9295749 2766 trans = btrfs_join_transaction(root, 1);
179e29e4
CM		 2767 goto again;
2768 }
70dec807 2769 write_extent_buffer(leaf, map + pg_offset, ptr,
179e29e4
CM		 2770 copy_size);
2771 btrfs_mark_buffer_dirty(leaf);
a52d9a80 2772 }
a52d9a80 2773 kunmap(page);
d1310b2e
CM		 2774 set_extent_uptodate(io_tree, em->start,
2775 extent_map_end(em) - 1, GFP_NOFS);
a52d9a80
CM		 2776 goto insert;
2777 } else {
2778 printk("unkknown found_type %d\n", found_type);
2779 WARN_ON(1);
2780 }
2781not_found:
2782 em->start = start;
d1310b2e 2783 em->len = len;
a52d9a80 2784not_found_em:
5f39d397 2785 em->block_start = EXTENT_MAP_HOLE;
a52d9a80
CM		 2786insert:
2787 btrfs_release_path(root, path);
d1310b2e
CM		 2788 if (em->start > start || extent_map_end(em) <= start) {
2789 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len);
a52d9a80
CM		 2790 err = -EIO;
2791 goto out;
2792 }
d1310b2e
CM		 2793
2794 err = 0;
2795 spin_lock(&em_tree->lock);
a52d9a80 2796 ret = add_extent_mapping(em_tree, em);
3b951516
CM		 2797 /* it is possible that someone inserted the extent into the tree
2798 * while we had the lock dropped. It is also possible that
2799 * an overlapping map exists in the tree
2800 */
a52d9a80 2801 if (ret == -EEXIST) {
3b951516 2802 struct extent_map *existing;
e6dcd2dc
CM		 2803
2804 ret = 0;
2805
3b951516 2806 existing = lookup_extent_mapping(em_tree, start, len);
e1c4b745
CM		 2807 if (existing && (existing->start > start ||
2808 existing->start + existing->len <= start)) {
2809 free_extent_map(existing);
2810 existing = NULL;
2811 }
3b951516
CM		 2812 if (!existing) {
2813 existing = lookup_extent_mapping(em_tree, em->start,
2814 em->len);
2815 if (existing) {
2816 err = merge_extent_mapping(em_tree, existing,
e6dcd2dc
CM		 2817 em, start,
2818 root->sectorsize);
3b951516
CM		 2819 free_extent_map(existing);
2820 if (err) {
2821 free_extent_map(em);
2822 em = NULL;
2823 }
2824 } else {
2825 err = -EIO;
2826 printk("failing to insert %Lu %Lu\n",
2827 start, len);
2828 free_extent_map(em);
2829 em = NULL;
2830 }
2831 } else {
2832 free_extent_map(em);
2833 em = existing;
e6dcd2dc 2834 err = 0;
a52d9a80 2835 }
a52d9a80 2836 }
d1310b2e 2837 spin_unlock(&em_tree->lock);
a52d9a80 2838out:
f421950f
CM		 2839 if (path)
2840 btrfs_free_path(path);
a52d9a80
CM		 2841 if (trans) {
2842 ret = btrfs_end_transaction(trans, root);
e6dcd2dc 2843 if (!err) {
a52d9a80 2844 err = ret;
e6dcd2dc 2845 }
a52d9a80 2846 }
a52d9a80
CM		 2847 if (err) {
2848 free_extent_map(em);
2849 WARN_ON(1);
2850 return ERR_PTR(err);
2851 }
2852 return em;
2853}
2854
e1c4b745 2855#if 0 /* waiting for O_DIRECT reads */
16432985
CM		 2856static int btrfs_get_block(struct inode *inode, sector_t iblock,
2857 struct buffer_head *bh_result, int create)
2858{
2859 struct extent_map *em;
2860 u64 start = (u64)iblock << inode->i_blkbits;
2861 struct btrfs_multi_bio *multi = NULL;
2862 struct btrfs_root *root = BTRFS_I(inode)->root;
2863 u64 len;
2864 u64 logical;
2865 u64 map_length;
2866 int ret = 0;
2867
2868 em = btrfs_get_extent(inode, NULL, 0, start, bh_result->b_size, 0);
2869
2870 if (!em || IS_ERR(em))
2871 goto out;
2872
e1c4b745 2873 if (em->start > start || em->start + em->len <= start) {
16432985 2874 goto out;
e1c4b745 2875 }
16432985
CM		 2876
2877 if (em->block_start == EXTENT_MAP_INLINE) {
2878 ret = -EINVAL;
2879 goto out;
2880 }
2881
e1c4b745
CM		 2882 len = em->start + em->len - start;
2883 len = min_t(u64, len, INT_LIMIT(typeof(bh_result->b_size)));
2884
16432985
CM		 2885 if (em->block_start == EXTENT_MAP_HOLE ||
2886 em->block_start == EXTENT_MAP_DELALLOC) {
e1c4b745 2887 bh_result->b_size = len;
16432985
CM		 2888 goto out;
2889 }
2890
16432985
CM		 2891 logical = start - em->start;
2892 logical = em->block_start + logical;
2893
2894 map_length = len;
2895 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
2896 logical, &map_length, &multi, 0);
2897 BUG_ON(ret);
2898 bh_result->b_blocknr = multi->stripes[0].physical >> inode->i_blkbits;
2899 bh_result->b_size = min(map_length, len);
e1c4b745 2900
16432985
CM		 2901 bh_result->b_bdev = multi->stripes[0].dev->bdev;
2902 set_buffer_mapped(bh_result);
2903 kfree(multi);
2904out:
2905 free_extent_map(em);
2906 return ret;
2907}
e1c4b745 2908#endif
16432985
CM		 2909
2910static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
2911 const struct iovec *iov, loff_t offset,
2912 unsigned long nr_segs)
2913{
e1c4b745
CM		 2914 return -EINVAL;
2915#if 0
16432985
CM		 2916 struct file *file = iocb->ki_filp;
2917 struct inode *inode = file->f_mapping->host;
2918
2919 if (rw == WRITE)
2920 return -EINVAL;
2921
2922 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2923 offset, nr_segs, btrfs_get_block, NULL);
e1c4b745 2924#endif
16432985
CM		 2925}
2926
d396c6f5 2927static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
39279cc3 2928{
d396c6f5 2929 return extent_bmap(mapping, iblock, btrfs_get_extent);
39279cc3
CM		 2930}
2931
a52d9a80 2932int btrfs_readpage(struct file *file, struct page *page)
9ebefb18 2933{
d1310b2e
CM		 2934 struct extent_io_tree *tree;
2935 tree = &BTRFS_I(page->mapping->host)->io_tree;
a52d9a80 2936 return extent_read_full_page(tree, page, btrfs_get_extent);
9ebefb18 2937}
1832a6d5 2938
a52d9a80 2939static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
39279cc3 2940{
d1310b2e 2941 struct extent_io_tree *tree;
b888db2b
CM		 2942
2943
2944 if (current->flags & PF_MEMALLOC) {
2945 redirty_page_for_writepage(wbc, page);
2946 unlock_page(page);
2947 return 0;
2948 }
d1310b2e 2949 tree = &BTRFS_I(page->mapping->host)->io_tree;
a52d9a80 2950 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
9ebefb18
CM		 2951}
2952
f421950f
CM		 2953int btrfs_writepages(struct address_space *mapping,
2954 struct writeback_control *wbc)
b293f02e 2955{
d1310b2e
CM		 2956 struct extent_io_tree *tree;
2957 tree = &BTRFS_I(mapping->host)->io_tree;
b293f02e
CM		 2958 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
2959}
2960
3ab2fb5a
CM		 2961static int
2962btrfs_readpages(struct file *file, struct address_space *mapping,
2963 struct list_head *pages, unsigned nr_pages)
2964{
d1310b2e
CM		 2965 struct extent_io_tree *tree;
2966 tree = &BTRFS_I(mapping->host)->io_tree;
3ab2fb5a
CM		 2967 return extent_readpages(tree, mapping, pages, nr_pages,
2968 btrfs_get_extent);
2969}
e6dcd2dc 2970static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
9ebefb18 2971{
d1310b2e
CM		 2972 struct extent_io_tree *tree;
2973 struct extent_map_tree *map;
a52d9a80 2974 int ret;
8c2383c3 2975
d1310b2e
CM		 2976 tree = &BTRFS_I(page->mapping->host)->io_tree;
2977 map = &BTRFS_I(page->mapping->host)->extent_tree;
70dec807 2978 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
a52d9a80
CM		 2979 if (ret == 1) {
2980 ClearPagePrivate(page);
2981 set_page_private(page, 0);
2982 page_cache_release(page);
39279cc3 2983 }
a52d9a80 2984 return ret;
39279cc3
CM		 2985}
2986
e6dcd2dc
CM		 2987static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
2988{
e6dcd2dc
CM		 2989 return __btrfs_releasepage(page, gfp_flags);
2990}
2991
a52d9a80 2992static void btrfs_invalidatepage(struct page *page, unsigned long offset)
39279cc3 2993{
d1310b2e 2994 struct extent_io_tree *tree;
e6dcd2dc
CM		 2995 struct btrfs_ordered_extent *ordered;
2996 u64 page_start = page_offset(page);
2997 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
39279cc3 2998
e6dcd2dc 2999 wait_on_page_writeback(page);
d1310b2e 3000 tree = &BTRFS_I(page->mapping->host)->io_tree;
e6dcd2dc
CM		 3001 if (offset) {
3002 btrfs_releasepage(page, GFP_NOFS);
3003 return;
3004 }
3005
3006 lock_extent(tree, page_start, page_end, GFP_NOFS);
3007 ordered = btrfs_lookup_ordered_extent(page->mapping->host,
3008 page_offset(page));
3009 if (ordered) {
eb84ae03
CM		 3010 /*
3011 * IO on this page will never be started, so we need
3012 * to account for any ordered extents now
3013 */
e6dcd2dc
CM		 3014 clear_extent_bit(tree, page_start, page_end,
3015 EXTENT_DIRTY | EXTENT_DELALLOC |
3016 EXTENT_LOCKED, 1, 0, GFP_NOFS);
211f90e6
CM		 3017 btrfs_finish_ordered_io(page->mapping->host,
3018 page_start, page_end);
e6dcd2dc
CM		 3019 btrfs_put_ordered_extent(ordered);
3020 lock_extent(tree, page_start, page_end, GFP_NOFS);
3021 }
3022 clear_extent_bit(tree, page_start, page_end,
3023 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
3024 EXTENT_ORDERED,
3025 1, 1, GFP_NOFS);
3026 __btrfs_releasepage(page, GFP_NOFS);
3027
4a096752 3028 ClearPageChecked(page);
9ad6b7bc 3029 if (PagePrivate(page)) {
9ad6b7bc
CM		 3030 ClearPagePrivate(page);
3031 set_page_private(page, 0);
3032 page_cache_release(page);
3033 }
39279cc3
CM		 3034}
3035
9ebefb18
CM		 3036/*
3037 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
3038 * called from a page fault handler when a page is first dirtied. Hence we must
3039 * be careful to check for EOF conditions here. We set the page up correctly
3040 * for a written page which means we get ENOSPC checking when writing into
3041 * holes and correct delalloc and unwritten extent mapping on filesystems that
3042 * support these features.
3043 *
3044 * We are not allowed to take the i_mutex here so we have to play games to
3045 * protect against truncate races as the page could now be beyond EOF. Because
3046 * vmtruncate() writes the inode size before removing pages, once we have the
3047 * page lock we can determine safely if the page is beyond EOF. If it is not
3048 * beyond EOF, then the page is guaranteed safe against truncation until we
3049 * unlock the page.
3050 */
3051int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
3052{
6da6abae 3053 struct inode *inode = fdentry(vma->vm_file)->d_inode;
1832a6d5 3054 struct btrfs_root *root = BTRFS_I(inode)->root;
e6dcd2dc
CM		 3055 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
3056 struct btrfs_ordered_extent *ordered;
3057 char *kaddr;
3058 unsigned long zero_start;
9ebefb18 3059 loff_t size;
1832a6d5 3060 int ret;
a52d9a80 3061 u64 page_start;
e6dcd2dc 3062 u64 page_end;
9ebefb18 3063
1832a6d5 3064 ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
1832a6d5
CM		 3065 if (ret)
3066 goto out;
3067
3068 ret = -EINVAL;
e6dcd2dc 3069again:
9ebefb18 3070 lock_page(page);
9ebefb18 3071 size = i_size_read(inode);
e6dcd2dc
CM		 3072 page_start = page_offset(page);
3073 page_end = page_start + PAGE_CACHE_SIZE - 1;
a52d9a80 3074
9ebefb18 3075 if ((page->mapping != inode->i_mapping) ||
e6dcd2dc 3076 (page_start >= size)) {
9ebefb18
CM		 3077 /* page got truncated out from underneath us */
3078 goto out_unlock;
3079 }
e6dcd2dc
CM		 3080 wait_on_page_writeback(page);
3081
3082 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
3083 set_page_extent_mapped(page);
3084
eb84ae03
CM		 3085 /*
3086 * we can't set the delalloc bits if there are pending ordered
3087 * extents. Drop our locks and wait for them to finish
3088 */
e6dcd2dc
CM		 3089 ordered = btrfs_lookup_ordered_extent(inode, page_start);
3090 if (ordered) {
3091 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
3092 unlock_page(page);
eb84ae03 3093 btrfs_start_ordered_extent(inode, ordered, 1);
e6dcd2dc
CM		 3094 btrfs_put_ordered_extent(ordered);
3095 goto again;
3096 }
3097
ea8c2819 3098 btrfs_set_extent_delalloc(inode, page_start, page_end);
e6dcd2dc 3099 ret = 0;
9ebefb18
CM		 3100
3101 /* page is wholly or partially inside EOF */
a52d9a80 3102 if (page_start + PAGE_CACHE_SIZE > size)
e6dcd2dc 3103 zero_start = size & ~PAGE_CACHE_MASK;
9ebefb18 3104 else
e6dcd2dc 3105 zero_start = PAGE_CACHE_SIZE;
9ebefb18 3106
e6dcd2dc
CM		 3107 if (zero_start != PAGE_CACHE_SIZE) {
3108 kaddr = kmap(page);
3109 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
3110 flush_dcache_page(page);
3111 kunmap(page);
3112 }
247e743c 3113 ClearPageChecked(page);
e6dcd2dc
CM		 3114 set_page_dirty(page);
3115 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
9ebefb18
CM		 3116
3117out_unlock:
3118 unlock_page(page);
1832a6d5 3119out:
9ebefb18
CM		 3120 return ret;
3121}
3122
39279cc3
CM		 3123static void btrfs_truncate(struct inode *inode)
3124{
3125 struct btrfs_root *root = BTRFS_I(inode)->root;
3126 int ret;
3127 struct btrfs_trans_handle *trans;
d3c2fdcf 3128 unsigned long nr;
dbe674a9 3129 u64 mask = root->sectorsize - 1;
39279cc3
CM		 3130
3131 if (!S_ISREG(inode->i_mode))
3132 return;
3133 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
3134 return;
3135
3136 btrfs_truncate_page(inode->i_mapping, inode->i_size);
4a096752 3137 btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
39279cc3 3138
39279cc3
CM		 3139 trans = btrfs_start_transaction(root, 1);
3140 btrfs_set_trans_block_group(trans, inode);
dbe674a9 3141 btrfs_i_size_write(inode, inode->i_size);
39279cc3 3142
7b128766
JB		 3143 ret = btrfs_orphan_add(trans, inode);
3144 if (ret)
3145 goto out;
39279cc3 3146 /* FIXME, add redo link to tree so we don't leak on crash */
85e21bac
CM		 3147 ret = btrfs_truncate_in_trans(trans, root, inode,
3148 BTRFS_EXTENT_DATA_KEY);
39279cc3 3149 btrfs_update_inode(trans, root, inode);
5f39d397 3150
7b128766
JB		 3151 ret = btrfs_orphan_del(trans, inode);
3152 BUG_ON(ret);
3153
3154out:
3155 nr = trans->blocks_used;
89ce8a63 3156 ret = btrfs_end_transaction_throttle(trans, root);
39279cc3 3157 BUG_ON(ret);
d3c2fdcf 3158 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM		 3159}
3160
3b96362c
SW		 3161/*
3162 * Invalidate a single dcache entry at the root of the filesystem.
3163 * Needed after creation of snapshot or subvolume.
3164 */
3165void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name,
3166 int namelen)
3167{
3168 struct dentry *alias, *entry;
3169 struct qstr qstr;
3170
3171 alias = d_find_alias(root->fs_info->sb->s_root->d_inode);
3172 if (alias) {
3173 qstr.name = name;
3174 qstr.len = namelen;
3175 /* change me if btrfs ever gets a d_hash operation */
3176 qstr.hash = full_name_hash(qstr.name, qstr.len);
3177 entry = d_lookup(alias, &qstr);
3178 dput(alias);
3179 if (entry) {
3180 d_invalidate(entry);
3181 dput(entry);
3182 }
3183 }
3184}
3185
f46b5a66
CH		 3186int btrfs_create_subvol_root(struct btrfs_root *new_root,
3187 struct btrfs_trans_handle *trans, u64 new_dirid,
3188 struct btrfs_block_group_cache *block_group)
39279cc3 3189{
39279cc3 3190 struct inode *inode;
00e4e6b3 3191 u64 index = 0;
39279cc3 3192
aec7477b 3193 inode = btrfs_new_inode(trans, new_root, NULL, "..", 2, new_dirid,
00e4e6b3 3194 new_dirid, block_group, S_IFDIR | 0700, &index);
54aa1f4d 3195 if (IS_ERR(inode))
f46b5a66 3196 return PTR_ERR(inode);
39279cc3
CM		 3197 inode->i_op = &btrfs_dir_inode_operations;
3198 inode->i_fop = &btrfs_dir_file_operations;
34088780 3199 new_root->inode = inode;
39279cc3 3200
39279cc3 3201 inode->i_nlink = 1;
dbe674a9 3202 btrfs_i_size_write(inode, 0);
3b96362c 3203
f46b5a66 3204 return btrfs_update_inode(trans, new_root, inode);
39279cc3
CM		 3205}
3206
edbd8d4e 3207unsigned long btrfs_force_ra(struct address_space *mapping,
86479a04
CM		 3208 struct file_ra_state *ra, struct file *file,
3209 pgoff_t offset, pgoff_t last_index)
3210{
8e7bf94f 3211 pgoff_t req_size = last_index - offset + 1;
86479a04
CM		 3212
3213#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
86479a04
CM		 3214 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
3215 return offset;
3216#else
86479a04
CM		 3217 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
3218 return offset + req_size;
3219#endif
3220}
3221
39279cc3
CM		 3222struct inode *btrfs_alloc_inode(struct super_block *sb)
3223{
3224 struct btrfs_inode *ei;
3225
3226 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
3227 if (!ei)
3228 return NULL;
15ee9bc7 3229 ei->last_trans = 0;
e6dcd2dc 3230 btrfs_ordered_inode_tree_init(&ei->ordered_tree);
33268eaf
JB		 3231 ei->i_acl = BTRFS_ACL_NOT_CACHED;
3232 ei->i_default_acl = BTRFS_ACL_NOT_CACHED;
7b128766 3233 INIT_LIST_HEAD(&ei->i_orphan);
39279cc3
CM		 3234 return &ei->vfs_inode;
3235}
3236
3237void btrfs_destroy_inode(struct inode *inode)
3238{
e6dcd2dc 3239 struct btrfs_ordered_extent *ordered;
39279cc3
CM		 3240 WARN_ON(!list_empty(&inode->i_dentry));
3241 WARN_ON(inode->i_data.nrpages);
3242
33268eaf
JB		 3243 if (BTRFS_I(inode)->i_acl &&
3244 BTRFS_I(inode)->i_acl != BTRFS_ACL_NOT_CACHED)
3245 posix_acl_release(BTRFS_I(inode)->i_acl);
3246 if (BTRFS_I(inode)->i_default_acl &&
3247 BTRFS_I(inode)->i_default_acl != BTRFS_ACL_NOT_CACHED)
3248 posix_acl_release(BTRFS_I(inode)->i_default_acl);
3249
bcc63abb 3250 spin_lock(&BTRFS_I(inode)->root->list_lock);
7b128766
JB		 3251 if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
3252 printk(KERN_ERR "BTRFS: inode %lu: inode still on the orphan"
3253 " list\n", inode->i_ino);
3254 dump_stack();
3255 }
bcc63abb 3256 spin_unlock(&BTRFS_I(inode)->root->list_lock);
7b128766 3257
e6dcd2dc
CM		 3258 while(1) {
3259 ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
3260 if (!ordered)
3261 break;
3262 else {
3263 printk("found ordered extent %Lu %Lu\n",
3264 ordered->file_offset, ordered->len);
3265 btrfs_remove_ordered_extent(inode, ordered);
3266 btrfs_put_ordered_extent(ordered);
3267 btrfs_put_ordered_extent(ordered);
3268 }
3269 }
8c416c9e 3270 btrfs_drop_extent_cache(inode, 0, (u64)-1);
39279cc3
CM		 3271 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
3272}
3273
0ee0fda0
SW		 3274#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3275static void init_once(void *foo)
3276#elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
44ec0b71
CM		 3277static void init_once(struct kmem_cache * cachep, void *foo)
3278#else
39279cc3
CM		 3279static void init_once(void * foo, struct kmem_cache * cachep,
3280 unsigned long flags)
44ec0b71 3281#endif
39279cc3
CM		 3282{
3283 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
3284
3285 inode_init_once(&ei->vfs_inode);
3286}
3287
3288void btrfs_destroy_cachep(void)
3289{
3290 if (btrfs_inode_cachep)
3291 kmem_cache_destroy(btrfs_inode_cachep);
3292 if (btrfs_trans_handle_cachep)
3293 kmem_cache_destroy(btrfs_trans_handle_cachep);
3294 if (btrfs_transaction_cachep)
3295 kmem_cache_destroy(btrfs_transaction_cachep);
3296 if (btrfs_bit_radix_cachep)
3297 kmem_cache_destroy(btrfs_bit_radix_cachep);
3298 if (btrfs_path_cachep)
3299 kmem_cache_destroy(btrfs_path_cachep);
3300}
3301
86479a04 3302struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
92fee66d 3303 unsigned long extra_flags,
0ee0fda0
SW		 3304#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3305 void (*ctor)(void *)
3306#elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
44ec0b71
CM		 3307 void (*ctor)(struct kmem_cache *, void *)
3308#else
92fee66d 3309 void (*ctor)(void *, struct kmem_cache *,
44ec0b71
CM		 3310 unsigned long)
3311#endif
3312 )
92fee66d
CM		 3313{
3314 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
3315 SLAB_MEM_SPREAD | extra_flags), ctor
3316#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
3317 ,NULL
3318#endif
3319 );
3320}
3321
39279cc3
CM		 3322int btrfs_init_cachep(void)
3323{
86479a04 3324 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
92fee66d
CM		 3325 sizeof(struct btrfs_inode),
3326 0, init_once);
39279cc3
CM		 3327 if (!btrfs_inode_cachep)
3328 goto fail;
86479a04
CM		 3329 btrfs_trans_handle_cachep =
3330 btrfs_cache_create("btrfs_trans_handle_cache",
3331 sizeof(struct btrfs_trans_handle),
3332 0, NULL);
39279cc3
CM		 3333 if (!btrfs_trans_handle_cachep)
3334 goto fail;
86479a04 3335 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
39279cc3 3336 sizeof(struct btrfs_transaction),
92fee66d 3337 0, NULL);
39279cc3
CM		 3338 if (!btrfs_transaction_cachep)
3339 goto fail;
86479a04 3340 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
23223584 3341 sizeof(struct btrfs_path),
92fee66d 3342 0, NULL);
39279cc3
CM		 3343 if (!btrfs_path_cachep)
3344 goto fail;
86479a04 3345 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
92fee66d 3346 SLAB_DESTROY_BY_RCU, NULL);
39279cc3
CM		 3347 if (!btrfs_bit_radix_cachep)
3348 goto fail;
3349 return 0;
3350fail:
3351 btrfs_destroy_cachep();
3352 return -ENOMEM;
3353}
3354
3355static int btrfs_getattr(struct vfsmount *mnt,
3356 struct dentry *dentry, struct kstat *stat)
3357{
3358 struct inode *inode = dentry->d_inode;
3359 generic_fillattr(inode, stat);
d6667462 3360 stat->blksize = PAGE_CACHE_SIZE;
9069218d 3361 stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9);
39279cc3
CM		 3362 return 0;
3363}
3364
3365static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
3366 struct inode * new_dir,struct dentry *new_dentry)
3367{
3368 struct btrfs_trans_handle *trans;
3369 struct btrfs_root *root = BTRFS_I(old_dir)->root;
3370 struct inode *new_inode = new_dentry->d_inode;
3371 struct inode *old_inode = old_dentry->d_inode;
3372 struct timespec ctime = CURRENT_TIME;
00e4e6b3 3373 u64 index = 0;
39279cc3
CM		 3374 int ret;
3375
3376 if (S_ISDIR(old_inode->i_mode) && new_inode &&
3377 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
3378 return -ENOTEMPTY;
3379 }
5f39d397 3380
1832a6d5
CM		 3381 ret = btrfs_check_free_space(root, 1, 0);
3382 if (ret)
3383 goto out_unlock;
3384
39279cc3 3385 trans = btrfs_start_transaction(root, 1);
5f39d397 3386
39279cc3 3387 btrfs_set_trans_block_group(trans, new_dir);
39279cc3
CM		 3388
3389 old_dentry->d_inode->i_nlink++;
3390 old_dir->i_ctime = old_dir->i_mtime = ctime;
3391 new_dir->i_ctime = new_dir->i_mtime = ctime;
3392 old_inode->i_ctime = ctime;
5f39d397 3393
39279cc3
CM		 3394 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
3395 if (ret)
3396 goto out_fail;
3397
3398 if (new_inode) {
3399 new_inode->i_ctime = CURRENT_TIME;
3400 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
3401 if (ret)
3402 goto out_fail;
7b128766
JB		 3403 if (new_inode->i_nlink == 0) {
3404 ret = btrfs_orphan_add(trans, new_inode);
3405 if (ret)
3406 goto out_fail;
3407 }
39279cc3 3408 }
00e4e6b3 3409 ret = btrfs_set_inode_index(new_dir, old_inode, &index);
aec7477b
JB		 3410 if (ret)
3411 goto out_fail;
3412
00e4e6b3 3413 ret = btrfs_add_link(trans, new_dentry, old_inode, 1, index);
39279cc3
CM		 3414 if (ret)
3415 goto out_fail;
3416
3417out_fail:
ab78c84d 3418 btrfs_end_transaction_throttle(trans, root);
1832a6d5 3419out_unlock:
39279cc3
CM		 3420 return ret;
3421}
3422
ea8c2819
CM		 3423int btrfs_start_delalloc_inodes(struct btrfs_root *root)
3424{
3425 struct list_head *head = &root->fs_info->delalloc_inodes;
3426 struct btrfs_inode *binode;
3427 unsigned long flags;
3428
3429 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
3430 while(!list_empty(head)) {
3431 binode = list_entry(head->next, struct btrfs_inode,
3432 delalloc_inodes);
3433 atomic_inc(&binode->vfs_inode.i_count);
3434 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
3435 filemap_write_and_wait(binode->vfs_inode.i_mapping);
3436 iput(&binode->vfs_inode);
3437 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
3438 }
3439 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
3440 return 0;
3441}
3442
39279cc3
CM		 3443static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
3444 const char *symname)
3445{
3446 struct btrfs_trans_handle *trans;
3447 struct btrfs_root *root = BTRFS_I(dir)->root;
3448 struct btrfs_path *path;
3449 struct btrfs_key key;
1832a6d5 3450 struct inode *inode = NULL;
39279cc3
CM		 3451 int err;
3452 int drop_inode = 0;
3453 u64 objectid;
00e4e6b3 3454 u64 index = 0 ;
39279cc3
CM		 3455 int name_len;
3456 int datasize;
5f39d397 3457 unsigned long ptr;
39279cc3 3458 struct btrfs_file_extent_item *ei;
5f39d397 3459 struct extent_buffer *leaf;
1832a6d5 3460 unsigned long nr = 0;
39279cc3
CM		 3461
3462 name_len = strlen(symname) + 1;
3463 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
3464 return -ENAMETOOLONG;
1832a6d5 3465
1832a6d5
CM		 3466 err = btrfs_check_free_space(root, 1, 0);
3467 if (err)
3468 goto out_fail;
3469
39279cc3
CM		 3470 trans = btrfs_start_transaction(root, 1);
3471 btrfs_set_trans_block_group(trans, dir);
3472
3473 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
3474 if (err) {
3475 err = -ENOSPC;
3476 goto out_unlock;
3477 }
3478
aec7477b 3479 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
9c58309d
CM		 3480 dentry->d_name.len,
3481 dentry->d_parent->d_inode->i_ino, objectid,
00e4e6b3
CM		 3482 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO,
3483 &index);
39279cc3
CM		 3484 err = PTR_ERR(inode);
3485 if (IS_ERR(inode))
3486 goto out_unlock;
3487
33268eaf
JB		 3488 err = btrfs_init_acl(inode, dir);
3489 if (err) {
3490 drop_inode = 1;
3491 goto out_unlock;
3492 }
3493
39279cc3 3494 btrfs_set_trans_block_group(trans, inode);
00e4e6b3 3495 err = btrfs_add_nondir(trans, dentry, inode, 0, index);
39279cc3
CM		 3496 if (err)
3497 drop_inode = 1;
3498 else {
3499 inode->i_mapping->a_ops = &btrfs_aops;
04160088 3500 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
39279cc3
CM		 3501 inode->i_fop = &btrfs_file_operations;
3502 inode->i_op = &btrfs_file_inode_operations;
d1310b2e
CM		 3503 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
3504 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
a52d9a80 3505 inode->i_mapping, GFP_NOFS);
7e38326f
CM		 3506 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
3507 inode->i_mapping, GFP_NOFS);
ea8c2819 3508 INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
1b1e2135 3509 mutex_init(&BTRFS_I(inode)->csum_mutex);
ee6e6504 3510 mutex_init(&BTRFS_I(inode)->extent_mutex);
9069218d 3511 BTRFS_I(inode)->delalloc_bytes = 0;
dbe674a9 3512 BTRFS_I(inode)->disk_i_size = 0;
db69e0eb 3513 inode->i_mapping->writeback_index = 0;
d1310b2e 3514 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
ba1da2f4 3515 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
39279cc3
CM		 3516 }
3517 dir->i_sb->s_dirt = 1;
3518 btrfs_update_inode_block_group(trans, inode);
3519 btrfs_update_inode_block_group(trans, dir);
3520 if (drop_inode)
3521 goto out_unlock;
3522
3523 path = btrfs_alloc_path();
3524 BUG_ON(!path);
3525 key.objectid = inode->i_ino;
3526 key.offset = 0;
39279cc3
CM		 3527 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
3528 datasize = btrfs_file_extent_calc_inline_size(name_len);
3529 err = btrfs_insert_empty_item(trans, root, path, &key,
3530 datasize);
54aa1f4d
CM		 3531 if (err) {
3532 drop_inode = 1;
3533 goto out_unlock;
3534 }
5f39d397
CM		 3535 leaf = path->nodes[0];
3536 ei = btrfs_item_ptr(leaf, path->slots[0],
3537 struct btrfs_file_extent_item);
3538 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
3539 btrfs_set_file_extent_type(leaf, ei,
39279cc3
CM		 3540 BTRFS_FILE_EXTENT_INLINE);
3541 ptr = btrfs_file_extent_inline_start(ei);
5f39d397
CM		 3542 write_extent_buffer(leaf, symname, ptr, name_len);
3543 btrfs_mark_buffer_dirty(leaf);
39279cc3 3544 btrfs_free_path(path);
5f39d397 3545
39279cc3
CM		 3546 inode->i_op = &btrfs_symlink_inode_operations;
3547 inode->i_mapping->a_ops = &btrfs_symlink_aops;
04160088 3548 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
dbe674a9 3549 btrfs_i_size_write(inode, name_len - 1);
54aa1f4d
CM		 3550 err = btrfs_update_inode(trans, root, inode);
3551 if (err)
3552 drop_inode = 1;
39279cc3
CM		 3553
3554out_unlock:
d3c2fdcf 3555 nr = trans->blocks_used;
ab78c84d 3556 btrfs_end_transaction_throttle(trans, root);
1832a6d5 3557out_fail:
39279cc3
CM		 3558 if (drop_inode) {
3559 inode_dec_link_count(inode);
3560 iput(inode);
3561 }
d3c2fdcf 3562 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM		 3563 return err;
3564}
16432985 3565
e6dcd2dc
CM		 3566static int btrfs_set_page_dirty(struct page *page)
3567{
e6dcd2dc
CM		 3568 return __set_page_dirty_nobuffers(page);
3569}
3570
0ee0fda0
SW		 3571#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26)
3572static int btrfs_permission(struct inode *inode, int mask)
3573#else
fdebe2bd
Y		 3574static int btrfs_permission(struct inode *inode, int mask,
3575 struct nameidata *nd)
0ee0fda0 3576#endif
fdebe2bd
Y		 3577{
3578 if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
3579 return -EACCES;
33268eaf 3580 return generic_permission(inode, mask, btrfs_check_acl);
fdebe2bd 3581}
39279cc3
CM		 3582
3583static struct inode_operations btrfs_dir_inode_operations = {
3584 .lookup = btrfs_lookup,
3585 .create = btrfs_create,
3586 .unlink = btrfs_unlink,
3587 .link = btrfs_link,
3588 .mkdir = btrfs_mkdir,
3589 .rmdir = btrfs_rmdir,
3590 .rename = btrfs_rename,
3591 .symlink = btrfs_symlink,
3592 .setattr = btrfs_setattr,
618e21d5 3593 .mknod = btrfs_mknod,
5103e947
JB		 3594 .setxattr = generic_setxattr,
3595 .getxattr = generic_getxattr,
3596 .listxattr = btrfs_listxattr,
3597 .removexattr = generic_removexattr,
fdebe2bd 3598 .permission = btrfs_permission,
39279cc3 3599};
39279cc3
CM		 3600static struct inode_operations btrfs_dir_ro_inode_operations = {
3601 .lookup = btrfs_lookup,
fdebe2bd 3602 .permission = btrfs_permission,
39279cc3 3603};
39279cc3
CM		 3604static struct file_operations btrfs_dir_file_operations = {
3605 .llseek = generic_file_llseek,
3606 .read = generic_read_dir,
3607 .readdir = btrfs_readdir,
34287aa3 3608 .unlocked_ioctl = btrfs_ioctl,
39279cc3 3609#ifdef CONFIG_COMPAT
34287aa3 3610 .compat_ioctl = btrfs_ioctl,
39279cc3 3611#endif
6bf13c0c 3612 .release = btrfs_release_file,
39279cc3
CM		 3613};
3614
d1310b2e 3615static struct extent_io_ops btrfs_extent_io_ops = {
07157aac 3616 .fill_delalloc = run_delalloc_range,
065631f6 3617 .submit_bio_hook = btrfs_submit_bio_hook,
239b14b3 3618 .merge_bio_hook = btrfs_merge_bio_hook,
07157aac 3619 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
e6dcd2dc 3620 .writepage_end_io_hook = btrfs_writepage_end_io_hook,
247e743c 3621 .writepage_start_hook = btrfs_writepage_start_hook,
1259ab75 3622 .readpage_io_failed_hook = btrfs_io_failed_hook,
b0c68f8b
CM		 3623 .set_bit_hook = btrfs_set_bit_hook,
3624 .clear_bit_hook = btrfs_clear_bit_hook,
07157aac
CM		 3625};
3626
39279cc3
CM		 3627static struct address_space_operations btrfs_aops = {
3628 .readpage = btrfs_readpage,
3629 .writepage = btrfs_writepage,
b293f02e 3630 .writepages = btrfs_writepages,
3ab2fb5a 3631 .readpages = btrfs_readpages,
39279cc3 3632 .sync_page = block_sync_page,
39279cc3 3633 .bmap = btrfs_bmap,
16432985 3634 .direct_IO = btrfs_direct_IO,
a52d9a80
CM		 3635 .invalidatepage = btrfs_invalidatepage,
3636 .releasepage = btrfs_releasepage,
e6dcd2dc 3637 .set_page_dirty = btrfs_set_page_dirty,
39279cc3
CM		 3638};
3639
3640static struct address_space_operations btrfs_symlink_aops = {
3641 .readpage = btrfs_readpage,
3642 .writepage = btrfs_writepage,
2bf5a725
CM		 3643 .invalidatepage = btrfs_invalidatepage,
3644 .releasepage = btrfs_releasepage,
39279cc3
CM		 3645};
3646
3647static struct inode_operations btrfs_file_inode_operations = {
3648 .truncate = btrfs_truncate,
3649 .getattr = btrfs_getattr,
3650 .setattr = btrfs_setattr,
5103e947
JB		 3651 .setxattr = generic_setxattr,
3652 .getxattr = generic_getxattr,
3653 .listxattr = btrfs_listxattr,
3654 .removexattr = generic_removexattr,
fdebe2bd 3655 .permission = btrfs_permission,
39279cc3 3656};
618e21d5
JB		 3657static struct inode_operations btrfs_special_inode_operations = {
3658 .getattr = btrfs_getattr,
3659 .setattr = btrfs_setattr,
fdebe2bd 3660 .permission = btrfs_permission,
33268eaf
JB		 3661 .setxattr = generic_setxattr,
3662 .getxattr = generic_getxattr,
3663 .listxattr = btrfs_listxattr,
3664 .removexattr = generic_removexattr,
618e21d5 3665};
39279cc3
CM		 3666static struct inode_operations btrfs_symlink_inode_operations = {
3667 .readlink = generic_readlink,
3668 .follow_link = page_follow_link_light,
3669 .put_link = page_put_link,
fdebe2bd 3670 .permission = btrfs_permission,
39279cc3 3671};
Linux 6.x block layer and io_uring tree(s)