]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/btrfs/inode.c
Btrfs: Index extent buffers in an rbtree
[mirror_ubuntu-bionic-kernel.git] / fs / btrfs / inode.c
CommitLineData
6cbd5570
CM
1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
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>
39279cc3
CM
39#include "ctree.h"
40#include "disk-io.h"
41#include "transaction.h"
42#include "btrfs_inode.h"
43#include "ioctl.h"
44#include "print-tree.h"
0b86a832 45#include "volumes.h"
e6dcd2dc 46#include "ordered-data.h"
39279cc3
CM
47
48struct btrfs_iget_args {
49 u64 ino;
50 struct btrfs_root *root;
51};
52
53static struct inode_operations btrfs_dir_inode_operations;
54static struct inode_operations btrfs_symlink_inode_operations;
55static struct inode_operations btrfs_dir_ro_inode_operations;
618e21d5 56static struct inode_operations btrfs_special_inode_operations;
39279cc3
CM
57static struct inode_operations btrfs_file_inode_operations;
58static struct address_space_operations btrfs_aops;
59static struct address_space_operations btrfs_symlink_aops;
60static struct file_operations btrfs_dir_file_operations;
d1310b2e 61static struct extent_io_ops btrfs_extent_io_ops;
39279cc3
CM
62
63static struct kmem_cache *btrfs_inode_cachep;
64struct kmem_cache *btrfs_trans_handle_cachep;
65struct kmem_cache *btrfs_transaction_cachep;
66struct kmem_cache *btrfs_bit_radix_cachep;
67struct kmem_cache *btrfs_path_cachep;
68
69#define S_SHIFT 12
70static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
71 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
72 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
73 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
74 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
75 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
76 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
77 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
78};
79
1832a6d5
CM
80int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
81 int for_del)
82{
a2135011
CM
83 u64 total;
84 u64 used;
1832a6d5 85 u64 thresh;
bcbfce8a 86 unsigned long flags;
1832a6d5
CM
87 int ret = 0;
88
a2135011
CM
89 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
90 total = btrfs_super_total_bytes(&root->fs_info->super_copy);
91 used = btrfs_super_bytes_used(&root->fs_info->super_copy);
1832a6d5 92 if (for_del)
f9ef6604 93 thresh = total * 90;
1832a6d5 94 else
f9ef6604
CM
95 thresh = total * 85;
96
97 do_div(thresh, 100);
1832a6d5 98
1832a6d5
CM
99 if (used + root->fs_info->delalloc_bytes + num_required > thresh)
100 ret = -ENOSPC;
bcbfce8a 101 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
1832a6d5
CM
102 return ret;
103}
104
be20aa9d 105static int cow_file_range(struct inode *inode, u64 start, u64 end)
b888db2b
CM
106{
107 struct btrfs_root *root = BTRFS_I(inode)->root;
108 struct btrfs_trans_handle *trans;
b888db2b 109 u64 alloc_hint = 0;
db94535d 110 u64 num_bytes;
c59f8951 111 u64 cur_alloc_size;
db94535d 112 u64 blocksize = root->sectorsize;
d1310b2e 113 u64 orig_num_bytes;
be20aa9d 114 struct btrfs_key ins;
e6dcd2dc
CM
115 struct extent_map *em;
116 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
117 int ret = 0;
b888db2b 118
f9295749 119 trans = btrfs_join_transaction(root, 1);
b888db2b 120 BUG_ON(!trans);
be20aa9d
CM
121 btrfs_set_trans_block_group(trans, inode);
122
db94535d 123 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
be20aa9d 124 num_bytes = max(blocksize, num_bytes);
d1310b2e 125 orig_num_bytes = num_bytes;
db94535d 126
179e29e4
CM
127 if (alloc_hint == EXTENT_MAP_INLINE)
128 goto out;
129
3b951516 130 BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy));
e5a2217e 131 mutex_lock(&BTRFS_I(inode)->extent_mutex);
e6dcd2dc 132 btrfs_drop_extent_cache(inode, start, start + num_bytes - 1);
e5a2217e 133 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
3b951516 134
c59f8951
CM
135 while(num_bytes > 0) {
136 cur_alloc_size = min(num_bytes, root->fs_info->max_extent);
e6dcd2dc
CM
137 ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
138 root->sectorsize, 0, 0,
139 (u64)-1, &ins, 1);
c59f8951
CM
140 if (ret) {
141 WARN_ON(1);
142 goto out;
143 }
e6dcd2dc
CM
144 em = alloc_extent_map(GFP_NOFS);
145 em->start = start;
146 em->len = ins.offset;
147 em->block_start = ins.objectid;
148 em->bdev = root->fs_info->fs_devices->latest_bdev;
e5a2217e 149 mutex_lock(&BTRFS_I(inode)->extent_mutex);
7f3c74fb 150 set_bit(EXTENT_FLAG_PINNED, &em->flags);
e6dcd2dc
CM
151 while(1) {
152 spin_lock(&em_tree->lock);
153 ret = add_extent_mapping(em_tree, em);
154 spin_unlock(&em_tree->lock);
155 if (ret != -EEXIST) {
156 free_extent_map(em);
157 break;
158 }
159 btrfs_drop_extent_cache(inode, start,
160 start + ins.offset - 1);
161 }
e5a2217e 162 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
e6dcd2dc 163
98d20f67 164 cur_alloc_size = ins.offset;
e6dcd2dc
CM
165 ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
166 ins.offset);
167 BUG_ON(ret);
3b951516
CM
168 if (num_bytes < cur_alloc_size) {
169 printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes,
170 cur_alloc_size);
171 break;
172 }
c59f8951
CM
173 num_bytes -= cur_alloc_size;
174 alloc_hint = ins.objectid + ins.offset;
175 start += cur_alloc_size;
b888db2b 176 }
b888db2b
CM
177out:
178 btrfs_end_transaction(trans, root);
be20aa9d
CM
179 return ret;
180}
181
182static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end)
183{
184 u64 extent_start;
185 u64 extent_end;
186 u64 bytenr;
187 u64 cow_end;
1832a6d5 188 u64 loops = 0;
c31f8830 189 u64 total_fs_bytes;
be20aa9d 190 struct btrfs_root *root = BTRFS_I(inode)->root;
a68d5933 191 struct btrfs_block_group_cache *block_group;
be20aa9d
CM
192 struct extent_buffer *leaf;
193 int found_type;
194 struct btrfs_path *path;
195 struct btrfs_file_extent_item *item;
196 int ret;
197 int err;
198 struct btrfs_key found_key;
199
c31f8830 200 total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
be20aa9d
CM
201 path = btrfs_alloc_path();
202 BUG_ON(!path);
203again:
204 ret = btrfs_lookup_file_extent(NULL, root, path,
205 inode->i_ino, start, 0);
206 if (ret < 0) {
207 btrfs_free_path(path);
208 return ret;
209 }
210
211 cow_end = end;
212 if (ret != 0) {
213 if (path->slots[0] == 0)
214 goto not_found;
215 path->slots[0]--;
216 }
217
218 leaf = path->nodes[0];
219 item = btrfs_item_ptr(leaf, path->slots[0],
220 struct btrfs_file_extent_item);
221
222 /* are we inside the extent that was found? */
223 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
224 found_type = btrfs_key_type(&found_key);
225 if (found_key.objectid != inode->i_ino ||
bbaf549e 226 found_type != BTRFS_EXTENT_DATA_KEY)
be20aa9d 227 goto not_found;
be20aa9d
CM
228
229 found_type = btrfs_file_extent_type(leaf, item);
230 extent_start = found_key.offset;
231 if (found_type == BTRFS_FILE_EXTENT_REG) {
c31f8830
CM
232 u64 extent_num_bytes;
233
234 extent_num_bytes = btrfs_file_extent_num_bytes(leaf, item);
235 extent_end = extent_start + extent_num_bytes;
be20aa9d
CM
236 err = 0;
237
1832a6d5
CM
238 if (loops && start != extent_start)
239 goto not_found;
240
be20aa9d
CM
241 if (start < extent_start || start >= extent_end)
242 goto not_found;
243
244 cow_end = min(end, extent_end - 1);
245 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
246 if (bytenr == 0)
247 goto not_found;
248
a68d5933
CM
249 if (btrfs_count_snapshots_in_path(root, path, inode->i_ino,
250 bytenr) != 1) {
251 goto not_found;
252 }
253
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
be20aa9d 263 start = extent_end;
bd09835d 264 } else {
be20aa9d
CM
265 goto not_found;
266 }
267loop:
268 if (start > end) {
269 btrfs_free_path(path);
270 return 0;
271 }
272 btrfs_release_path(root, path);
1832a6d5 273 loops++;
be20aa9d
CM
274 goto again;
275
276not_found:
bbaf549e
CM
277 cow_file_range(inode, start, end);
278 start = end + 1;
be20aa9d
CM
279 goto loop;
280}
281
282static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
283{
284 struct btrfs_root *root = BTRFS_I(inode)->root;
285 int ret;
a2135011 286
b98b6767
Y
287 if (btrfs_test_opt(root, NODATACOW) ||
288 btrfs_test_flag(inode, NODATACOW))
be20aa9d
CM
289 ret = run_delalloc_nocow(inode, start, end);
290 else
291 ret = cow_file_range(inode, start, end);
1832a6d5 292
b888db2b
CM
293 return ret;
294}
295
291d673e 296int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
b0c68f8b 297 unsigned long old, unsigned long bits)
291d673e 298{
bcbfce8a 299 unsigned long flags;
b0c68f8b 300 if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
291d673e 301 struct btrfs_root *root = BTRFS_I(inode)->root;
bcbfce8a 302 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
9069218d 303 BTRFS_I(inode)->delalloc_bytes += end - start + 1;
291d673e 304 root->fs_info->delalloc_bytes += end - start + 1;
bcbfce8a 305 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
291d673e
CM
306 }
307 return 0;
308}
309
310int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end,
b0c68f8b 311 unsigned long old, unsigned long bits)
291d673e 312{
b0c68f8b 313 if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
291d673e 314 struct btrfs_root *root = BTRFS_I(inode)->root;
bcbfce8a
CM
315 unsigned long flags;
316
317 spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
b0c68f8b
CM
318 if (end - start + 1 > root->fs_info->delalloc_bytes) {
319 printk("warning: delalloc account %Lu %Lu\n",
320 end - start + 1, root->fs_info->delalloc_bytes);
321 root->fs_info->delalloc_bytes = 0;
9069218d 322 BTRFS_I(inode)->delalloc_bytes = 0;
b0c68f8b
CM
323 } else {
324 root->fs_info->delalloc_bytes -= end - start + 1;
9069218d 325 BTRFS_I(inode)->delalloc_bytes -= end - start + 1;
b0c68f8b 326 }
bcbfce8a 327 spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
291d673e
CM
328 }
329 return 0;
330}
331
239b14b3
CM
332int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
333 size_t size, struct bio *bio)
334{
335 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
336 struct btrfs_mapping_tree *map_tree;
239b14b3 337 u64 logical = bio->bi_sector << 9;
239b14b3
CM
338 u64 length = 0;
339 u64 map_length;
239b14b3
CM
340 int ret;
341
f2d8d74d 342 length = bio->bi_size;
239b14b3
CM
343 map_tree = &root->fs_info->mapping_tree;
344 map_length = length;
cea9e445 345 ret = btrfs_map_block(map_tree, READ, logical,
f188591e 346 &map_length, NULL, 0);
cea9e445 347
239b14b3 348 if (map_length < length + size) {
239b14b3
CM
349 return 1;
350 }
351 return 0;
352}
353
44b8bd7e 354int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
f188591e 355 int mirror_num)
065631f6 356{
065631f6 357 struct btrfs_root *root = BTRFS_I(inode)->root;
065631f6 358 int ret = 0;
e015640f 359
3edf7d33 360 ret = btrfs_csum_one_bio(root, inode, bio);
44b8bd7e 361 BUG_ON(ret);
e015640f 362
8b712842 363 return btrfs_map_bio(root, rw, bio, mirror_num, 1);
44b8bd7e
CM
364}
365
366int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
367 int mirror_num)
368{
369 struct btrfs_root *root = BTRFS_I(inode)->root;
370 int ret = 0;
371
e6dcd2dc
CM
372 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
373 BUG_ON(ret);
065631f6 374
e6dcd2dc 375 if (!(rw & (1 << BIO_RW))) {
0b86a832
CM
376 goto mapit;
377 }
065631f6 378
44b8bd7e
CM
379 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
380 inode, rw, bio, mirror_num,
381 __btrfs_submit_bio_hook);
0b86a832 382mapit:
8b712842 383 return btrfs_map_bio(root, rw, bio, mirror_num, 0);
065631f6 384}
6885f308 385
ba1da2f4 386static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
e6dcd2dc
CM
387 struct inode *inode, u64 file_offset,
388 struct list_head *list)
389{
390 struct list_head *cur;
391 struct btrfs_ordered_sum *sum;
392
393 btrfs_set_trans_block_group(trans, inode);
ba1da2f4 394 list_for_each(cur, list) {
e6dcd2dc
CM
395 sum = list_entry(cur, struct btrfs_ordered_sum, list);
396 mutex_lock(&BTRFS_I(inode)->csum_mutex);
397 btrfs_csum_file_blocks(trans, BTRFS_I(inode)->root,
398 inode, sum);
399 mutex_unlock(&BTRFS_I(inode)->csum_mutex);
e6dcd2dc
CM
400 }
401 return 0;
402}
403
247e743c
CM
404struct btrfs_writepage_fixup {
405 struct page *page;
406 struct btrfs_work work;
407};
408
409/* see btrfs_writepage_start_hook for details on why this is required */
410void btrfs_writepage_fixup_worker(struct btrfs_work *work)
411{
412 struct btrfs_writepage_fixup *fixup;
413 struct btrfs_ordered_extent *ordered;
414 struct page *page;
415 struct inode *inode;
416 u64 page_start;
417 u64 page_end;
418
419 fixup = container_of(work, struct btrfs_writepage_fixup, work);
420 page = fixup->page;
4a096752 421again:
247e743c
CM
422 lock_page(page);
423 if (!page->mapping || !PageDirty(page) || !PageChecked(page)) {
424 ClearPageChecked(page);
425 goto out_page;
426 }
427
428 inode = page->mapping->host;
429 page_start = page_offset(page);
430 page_end = page_offset(page) + PAGE_CACHE_SIZE - 1;
431
432 lock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
4a096752
CM
433
434 /* already ordered? We're done */
435 if (test_range_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
436 EXTENT_ORDERED, 0)) {
247e743c 437 goto out;
4a096752
CM
438 }
439
440 ordered = btrfs_lookup_ordered_extent(inode, page_start);
441 if (ordered) {
442 unlock_extent(&BTRFS_I(inode)->io_tree, page_start,
443 page_end, GFP_NOFS);
444 unlock_page(page);
445 btrfs_start_ordered_extent(inode, ordered, 1);
446 goto again;
447 }
247e743c
CM
448
449 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start, page_end,
450 GFP_NOFS);
451 ClearPageChecked(page);
452out:
453 unlock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS);
454out_page:
455 unlock_page(page);
456 page_cache_release(page);
457}
458
459/*
460 * There are a few paths in the higher layers of the kernel that directly
461 * set the page dirty bit without asking the filesystem if it is a
462 * good idea. This causes problems because we want to make sure COW
463 * properly happens and the data=ordered rules are followed.
464 *
465 * In our case any range that doesn't have the EXTENT_ORDERED bit set
466 * hasn't been properly setup for IO. We kick off an async process
467 * to fix it up. The async helper will wait for ordered extents, set
468 * the delalloc bit and make it safe to write the page.
469 */
470int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
471{
472 struct inode *inode = page->mapping->host;
473 struct btrfs_writepage_fixup *fixup;
474 struct btrfs_root *root = BTRFS_I(inode)->root;
475 int ret;
476
477 ret = test_range_bit(&BTRFS_I(inode)->io_tree, start, end,
478 EXTENT_ORDERED, 0);
479 if (ret)
480 return 0;
481
482 if (PageChecked(page))
483 return -EAGAIN;
484
485 fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
486 if (!fixup)
487 return -EAGAIN;
488printk("queueing worker to fixup page %lu %Lu\n", inode->i_ino, page_offset(page));
489 SetPageChecked(page);
490 page_cache_get(page);
491 fixup->work.func = btrfs_writepage_fixup_worker;
492 fixup->page = page;
493 btrfs_queue_worker(&root->fs_info->fixup_workers, &fixup->work);
494 return -EAGAIN;
495}
496
211f90e6 497static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
e6dcd2dc 498{
e6dcd2dc
CM
499 struct btrfs_root *root = BTRFS_I(inode)->root;
500 struct btrfs_trans_handle *trans;
501 struct btrfs_ordered_extent *ordered_extent;
502 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
7f3c74fb
CM
503 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
504 struct extent_map *em;
e6dcd2dc 505 u64 alloc_hint = 0;
e5a2217e
CM
506 u64 clear_start;
507 u64 clear_end;
e6dcd2dc
CM
508 struct list_head list;
509 struct btrfs_key ins;
510 int ret;
511
512 ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1);
ba1da2f4 513 if (!ret)
e6dcd2dc 514 return 0;
e6dcd2dc 515
f9295749 516 trans = btrfs_join_transaction(root, 1);
e6dcd2dc
CM
517
518 ordered_extent = btrfs_lookup_ordered_extent(inode, start);
519 BUG_ON(!ordered_extent);
520
521 lock_extent(io_tree, ordered_extent->file_offset,
522 ordered_extent->file_offset + ordered_extent->len - 1,
523 GFP_NOFS);
524
525 INIT_LIST_HEAD(&list);
526
527 ins.objectid = ordered_extent->start;
528 ins.offset = ordered_extent->len;
529 ins.type = BTRFS_EXTENT_ITEM_KEY;
e5a2217e 530
e6dcd2dc
CM
531 ret = btrfs_alloc_reserved_extent(trans, root, root->root_key.objectid,
532 trans->transid, inode->i_ino,
533 ordered_extent->file_offset, &ins);
534 BUG_ON(ret);
ee6e6504
CM
535
536 mutex_lock(&BTRFS_I(inode)->extent_mutex);
e5a2217e 537
e6dcd2dc
CM
538 ret = btrfs_drop_extents(trans, root, inode,
539 ordered_extent->file_offset,
540 ordered_extent->file_offset +
541 ordered_extent->len,
542 ordered_extent->file_offset, &alloc_hint);
543 BUG_ON(ret);
544 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
545 ordered_extent->file_offset,
546 ordered_extent->start,
547 ordered_extent->len,
548 ordered_extent->len, 0);
549 BUG_ON(ret);
7f3c74fb 550
7f3c74fb 551 spin_lock(&em_tree->lock);
e5a2217e
CM
552 clear_start = ordered_extent->file_offset;
553 clear_end = ordered_extent->file_offset + ordered_extent->len;
554 while(clear_start < clear_end) {
555 em = lookup_extent_mapping(em_tree, clear_start,
556 clear_end - clear_start);
557 if (em) {
558 clear_bit(EXTENT_FLAG_PINNED, &em->flags);
559 clear_start = em->start + em->len;
560 free_extent_map(em);
561 } else {
562 break;
563 }
7f3c74fb
CM
564 }
565 spin_unlock(&em_tree->lock);
566
e6dcd2dc
CM
567 btrfs_drop_extent_cache(inode, ordered_extent->file_offset,
568 ordered_extent->file_offset +
569 ordered_extent->len - 1);
ee6e6504
CM
570 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
571
e6dcd2dc
CM
572 inode->i_blocks += ordered_extent->len >> 9;
573 unlock_extent(io_tree, ordered_extent->file_offset,
574 ordered_extent->file_offset + ordered_extent->len - 1,
575 GFP_NOFS);
576 add_pending_csums(trans, inode, ordered_extent->file_offset,
577 &ordered_extent->list);
578
dbe674a9 579 btrfs_ordered_update_i_size(inode, ordered_extent);
e6dcd2dc 580 btrfs_remove_ordered_extent(inode, ordered_extent);
7f3c74fb 581
e6dcd2dc
CM
582 /* once for us */
583 btrfs_put_ordered_extent(ordered_extent);
584 /* once for the tree */
585 btrfs_put_ordered_extent(ordered_extent);
586
587 btrfs_update_inode(trans, root, inode);
588 btrfs_end_transaction(trans, root);
589 return 0;
590}
591
211f90e6
CM
592int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
593 struct extent_state *state, int uptodate)
594{
595 return btrfs_finish_ordered_io(page->mapping->host, start, end);
596}
597
07157aac
CM
598int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
599{
600 int ret = 0;
601 struct inode *inode = page->mapping->host;
602 struct btrfs_root *root = BTRFS_I(inode)->root;
d1310b2e 603 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
07157aac
CM
604 struct btrfs_csum_item *item;
605 struct btrfs_path *path = NULL;
ff79f819 606 u32 csum;
699122f5 607
b98b6767
Y
608 if (btrfs_test_opt(root, NODATASUM) ||
609 btrfs_test_flag(inode, NODATASUM))
b6cda9bc 610 return 0;
699122f5 611
07157aac
CM
612 path = btrfs_alloc_path();
613 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
614 if (IS_ERR(item)) {
ba1da2f4
CM
615 /*
616 * It is possible there is an ordered extent that has
617 * not yet finished for this range in the file. If so,
618 * that extent will have a csum cached, and it will insert
619 * the sum after all the blocks in the extent are fully
620 * on disk. So, look for an ordered extent and use the
621 * sum if found.
622 */
623 ret = btrfs_find_ordered_sum(inode, start, &csum);
624 if (ret == 0)
625 goto found;
626
07157aac
CM
627 ret = PTR_ERR(item);
628 /* a csum that isn't present is a preallocated region. */
629 if (ret == -ENOENT || ret == -EFBIG)
630 ret = 0;
ff79f819 631 csum = 0;
e6dcd2dc
CM
632 printk("no csum found for inode %lu start %Lu\n", inode->i_ino,
633 start);
07157aac
CM
634 goto out;
635 }
ff79f819
CM
636 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
637 BTRFS_CRC32_SIZE);
ba1da2f4 638found:
d1310b2e 639 set_state_private(io_tree, start, csum);
07157aac
CM
640out:
641 if (path)
642 btrfs_free_path(path);
07157aac
CM
643 return ret;
644}
645
7e38326f
CM
646struct io_failure_record {
647 struct page *page;
648 u64 start;
649 u64 len;
650 u64 logical;
651 int last_mirror;
652};
653
1259ab75
CM
654int btrfs_io_failed_hook(struct bio *failed_bio,
655 struct page *page, u64 start, u64 end,
656 struct extent_state *state)
7e38326f
CM
657{
658 struct io_failure_record *failrec = NULL;
659 u64 private;
660 struct extent_map *em;
661 struct inode *inode = page->mapping->host;
662 struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
3b951516 663 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
7e38326f
CM
664 struct bio *bio;
665 int num_copies;
666 int ret;
1259ab75 667 int rw;
7e38326f
CM
668 u64 logical;
669
670 ret = get_state_private(failure_tree, start, &private);
671 if (ret) {
7e38326f
CM
672 failrec = kmalloc(sizeof(*failrec), GFP_NOFS);
673 if (!failrec)
674 return -ENOMEM;
675 failrec->start = start;
676 failrec->len = end - start + 1;
677 failrec->last_mirror = 0;
678
3b951516
CM
679 spin_lock(&em_tree->lock);
680 em = lookup_extent_mapping(em_tree, start, failrec->len);
681 if (em->start > start || em->start + em->len < start) {
682 free_extent_map(em);
683 em = NULL;
684 }
685 spin_unlock(&em_tree->lock);
7e38326f
CM
686
687 if (!em || IS_ERR(em)) {
688 kfree(failrec);
689 return -EIO;
690 }
691 logical = start - em->start;
692 logical = em->block_start + logical;
693 failrec->logical = logical;
694 free_extent_map(em);
695 set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
696 EXTENT_DIRTY, GFP_NOFS);
587f7704
CM
697 set_state_private(failure_tree, start,
698 (u64)(unsigned long)failrec);
7e38326f 699 } else {
587f7704 700 failrec = (struct io_failure_record *)(unsigned long)private;
7e38326f
CM
701 }
702 num_copies = btrfs_num_copies(
703 &BTRFS_I(inode)->root->fs_info->mapping_tree,
704 failrec->logical, failrec->len);
705 failrec->last_mirror++;
706 if (!state) {
707 spin_lock_irq(&BTRFS_I(inode)->io_tree.lock);
708 state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
709 failrec->start,
710 EXTENT_LOCKED);
711 if (state && state->start != failrec->start)
712 state = NULL;
713 spin_unlock_irq(&BTRFS_I(inode)->io_tree.lock);
714 }
715 if (!state || failrec->last_mirror > num_copies) {
716 set_state_private(failure_tree, failrec->start, 0);
717 clear_extent_bits(failure_tree, failrec->start,
718 failrec->start + failrec->len - 1,
719 EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
720 kfree(failrec);
721 return -EIO;
722 }
723 bio = bio_alloc(GFP_NOFS, 1);
724 bio->bi_private = state;
725 bio->bi_end_io = failed_bio->bi_end_io;
726 bio->bi_sector = failrec->logical >> 9;
727 bio->bi_bdev = failed_bio->bi_bdev;
e1c4b745 728 bio->bi_size = 0;
7e38326f 729 bio_add_page(bio, page, failrec->len, start - page_offset(page));
1259ab75
CM
730 if (failed_bio->bi_rw & (1 << BIO_RW))
731 rw = WRITE;
732 else
733 rw = READ;
734
735 BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
736 failrec->last_mirror);
737 return 0;
738}
739
740int btrfs_clean_io_failures(struct inode *inode, u64 start)
741{
742 u64 private;
743 u64 private_failure;
744 struct io_failure_record *failure;
745 int ret;
746
747 private = 0;
748 if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
749 (u64)-1, 1, EXTENT_DIRTY)) {
750 ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
751 start, &private_failure);
752 if (ret == 0) {
753 failure = (struct io_failure_record *)(unsigned long)
754 private_failure;
755 set_state_private(&BTRFS_I(inode)->io_failure_tree,
756 failure->start, 0);
757 clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
758 failure->start,
759 failure->start + failure->len - 1,
760 EXTENT_DIRTY | EXTENT_LOCKED,
761 GFP_NOFS);
762 kfree(failure);
763 }
764 }
7e38326f
CM
765 return 0;
766}
767
70dec807
CM
768int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
769 struct extent_state *state)
07157aac 770{
35ebb934 771 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
07157aac 772 struct inode *inode = page->mapping->host;
d1310b2e 773 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
07157aac 774 char *kaddr;
aadfeb6e 775 u64 private = ~(u32)0;
07157aac 776 int ret;
ff79f819
CM
777 struct btrfs_root *root = BTRFS_I(inode)->root;
778 u32 csum = ~(u32)0;
bbf0d006 779 unsigned long flags;
d1310b2e 780
b98b6767
Y
781 if (btrfs_test_opt(root, NODATASUM) ||
782 btrfs_test_flag(inode, NODATASUM))
b6cda9bc 783 return 0;
c2e639f0 784 if (state && state->start == start) {
70dec807
CM
785 private = state->private;
786 ret = 0;
787 } else {
788 ret = get_state_private(io_tree, start, &private);
789 }
bbf0d006 790 local_irq_save(flags);
07157aac
CM
791 kaddr = kmap_atomic(page, KM_IRQ0);
792 if (ret) {
793 goto zeroit;
794 }
ff79f819
CM
795 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
796 btrfs_csum_final(csum, (char *)&csum);
797 if (csum != private) {
07157aac
CM
798 goto zeroit;
799 }
800 kunmap_atomic(kaddr, KM_IRQ0);
bbf0d006 801 local_irq_restore(flags);
7e38326f
CM
802
803 /* if the io failure tree for this inode is non-empty,
804 * check to see if we've recovered from a failed IO
805 */
1259ab75 806 btrfs_clean_io_failures(inode, start);
07157aac
CM
807 return 0;
808
809zeroit:
aadfeb6e
CM
810 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
811 page->mapping->host->i_ino, (unsigned long long)start, csum,
812 private);
db94535d
CM
813 memset(kaddr + offset, 1, end - start + 1);
814 flush_dcache_page(page);
07157aac 815 kunmap_atomic(kaddr, KM_IRQ0);
bbf0d006 816 local_irq_restore(flags);
3b951516
CM
817 if (private == 0)
818 return 0;
7e38326f 819 return -EIO;
07157aac 820}
b888db2b 821
39279cc3
CM
822void btrfs_read_locked_inode(struct inode *inode)
823{
824 struct btrfs_path *path;
5f39d397 825 struct extent_buffer *leaf;
39279cc3 826 struct btrfs_inode_item *inode_item;
0b86a832 827 struct btrfs_timespec *tspec;
39279cc3
CM
828 struct btrfs_root *root = BTRFS_I(inode)->root;
829 struct btrfs_key location;
830 u64 alloc_group_block;
618e21d5 831 u32 rdev;
39279cc3
CM
832 int ret;
833
834 path = btrfs_alloc_path();
835 BUG_ON(!path);
39279cc3 836 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
dc17ff8f 837
39279cc3 838 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
5f39d397 839 if (ret)
39279cc3 840 goto make_bad;
39279cc3 841
5f39d397
CM
842 leaf = path->nodes[0];
843 inode_item = btrfs_item_ptr(leaf, path->slots[0],
844 struct btrfs_inode_item);
845
846 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
847 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
848 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
849 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
dbe674a9 850 btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item));
5f39d397
CM
851
852 tspec = btrfs_inode_atime(inode_item);
853 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
854 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
855
856 tspec = btrfs_inode_mtime(inode_item);
857 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
858 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
859
860 tspec = btrfs_inode_ctime(inode_item);
861 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
862 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
863
864 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
865 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
618e21d5 866 inode->i_rdev = 0;
5f39d397
CM
867 rdev = btrfs_inode_rdev(leaf, inode_item);
868
869 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
39279cc3
CM
870 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
871 alloc_group_block);
b98b6767 872 BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
e52ec0eb
CM
873 if (!BTRFS_I(inode)->block_group) {
874 BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
0b86a832
CM
875 NULL, 0,
876 BTRFS_BLOCK_GROUP_METADATA, 0);
e52ec0eb 877 }
39279cc3
CM
878 btrfs_free_path(path);
879 inode_item = NULL;
880
39279cc3 881 switch (inode->i_mode & S_IFMT) {
39279cc3
CM
882 case S_IFREG:
883 inode->i_mapping->a_ops = &btrfs_aops;
04160088 884 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
d1310b2e 885 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
39279cc3
CM
886 inode->i_fop = &btrfs_file_operations;
887 inode->i_op = &btrfs_file_inode_operations;
888 break;
889 case S_IFDIR:
890 inode->i_fop = &btrfs_dir_file_operations;
891 if (root == root->fs_info->tree_root)
892 inode->i_op = &btrfs_dir_ro_inode_operations;
893 else
894 inode->i_op = &btrfs_dir_inode_operations;
895 break;
896 case S_IFLNK:
897 inode->i_op = &btrfs_symlink_inode_operations;
898 inode->i_mapping->a_ops = &btrfs_symlink_aops;
04160088 899 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
39279cc3 900 break;
618e21d5
JB
901 default:
902 init_special_inode(inode, inode->i_mode, rdev);
903 break;
39279cc3
CM
904 }
905 return;
906
907make_bad:
39279cc3 908 btrfs_free_path(path);
39279cc3
CM
909 make_bad_inode(inode);
910}
911
5f39d397
CM
912static void fill_inode_item(struct extent_buffer *leaf,
913 struct btrfs_inode_item *item,
39279cc3
CM
914 struct inode *inode)
915{
5f39d397
CM
916 btrfs_set_inode_uid(leaf, item, inode->i_uid);
917 btrfs_set_inode_gid(leaf, item, inode->i_gid);
dbe674a9 918 btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size);
5f39d397
CM
919 btrfs_set_inode_mode(leaf, item, inode->i_mode);
920 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
921
922 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
923 inode->i_atime.tv_sec);
924 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
925 inode->i_atime.tv_nsec);
926
927 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
928 inode->i_mtime.tv_sec);
929 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
930 inode->i_mtime.tv_nsec);
931
932 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
933 inode->i_ctime.tv_sec);
934 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
935 inode->i_ctime.tv_nsec);
936
937 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
938 btrfs_set_inode_generation(leaf, item, inode->i_generation);
939 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
b98b6767 940 btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
5f39d397 941 btrfs_set_inode_block_group(leaf, item,
39279cc3
CM
942 BTRFS_I(inode)->block_group->key.objectid);
943}
944
ba1da2f4 945int noinline btrfs_update_inode(struct btrfs_trans_handle *trans,
39279cc3
CM
946 struct btrfs_root *root,
947 struct inode *inode)
948{
949 struct btrfs_inode_item *inode_item;
950 struct btrfs_path *path;
5f39d397 951 struct extent_buffer *leaf;
39279cc3
CM
952 int ret;
953
954 path = btrfs_alloc_path();
955 BUG_ON(!path);
39279cc3
CM
956 ret = btrfs_lookup_inode(trans, root, path,
957 &BTRFS_I(inode)->location, 1);
958 if (ret) {
959 if (ret > 0)
960 ret = -ENOENT;
961 goto failed;
962 }
963
5f39d397
CM
964 leaf = path->nodes[0];
965 inode_item = btrfs_item_ptr(leaf, path->slots[0],
39279cc3
CM
966 struct btrfs_inode_item);
967
5f39d397
CM
968 fill_inode_item(leaf, inode_item, inode);
969 btrfs_mark_buffer_dirty(leaf);
15ee9bc7 970 btrfs_set_inode_last_trans(trans, inode);
39279cc3
CM
971 ret = 0;
972failed:
39279cc3
CM
973 btrfs_free_path(path);
974 return ret;
975}
976
977
978static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
979 struct btrfs_root *root,
980 struct inode *dir,
981 struct dentry *dentry)
982{
983 struct btrfs_path *path;
984 const char *name = dentry->d_name.name;
985 int name_len = dentry->d_name.len;
986 int ret = 0;
5f39d397 987 struct extent_buffer *leaf;
39279cc3 988 struct btrfs_dir_item *di;
5f39d397 989 struct btrfs_key key;
39279cc3
CM
990
991 path = btrfs_alloc_path();
54aa1f4d
CM
992 if (!path) {
993 ret = -ENOMEM;
994 goto err;
995 }
996
39279cc3
CM
997 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
998 name, name_len, -1);
999 if (IS_ERR(di)) {
1000 ret = PTR_ERR(di);
1001 goto err;
1002 }
1003 if (!di) {
1004 ret = -ENOENT;
1005 goto err;
1006 }
5f39d397
CM
1007 leaf = path->nodes[0];
1008 btrfs_dir_item_key_to_cpu(leaf, di, &key);
39279cc3 1009 ret = btrfs_delete_one_dir_name(trans, root, path, di);
54aa1f4d
CM
1010 if (ret)
1011 goto err;
39279cc3
CM
1012 btrfs_release_path(root, path);
1013
1014 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
5f39d397 1015 key.objectid, name, name_len, -1);
39279cc3
CM
1016 if (IS_ERR(di)) {
1017 ret = PTR_ERR(di);
1018 goto err;
1019 }
1020 if (!di) {
1021 ret = -ENOENT;
1022 goto err;
1023 }
1024 ret = btrfs_delete_one_dir_name(trans, root, path, di);
925baedd 1025 btrfs_release_path(root, path);
39279cc3
CM
1026
1027 dentry->d_inode->i_ctime = dir->i_ctime;
76fea00a
CM
1028 ret = btrfs_del_inode_ref(trans, root, name, name_len,
1029 dentry->d_inode->i_ino,
1030 dentry->d_parent->d_inode->i_ino);
1031 if (ret) {
1032 printk("failed to delete reference to %.*s, "
1033 "inode %lu parent %lu\n", name_len, name,
1034 dentry->d_inode->i_ino,
1035 dentry->d_parent->d_inode->i_ino);
3954401f 1036 }
39279cc3
CM
1037err:
1038 btrfs_free_path(path);
1039 if (!ret) {
dbe674a9 1040 btrfs_i_size_write(dir, dir->i_size - name_len * 2);
79c44584 1041 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
39279cc3 1042 btrfs_update_inode(trans, root, dir);
6da6abae
CM
1043#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1044 dentry->d_inode->i_nlink--;
1045#else
39279cc3 1046 drop_nlink(dentry->d_inode);
6da6abae 1047#endif
54aa1f4d 1048 ret = btrfs_update_inode(trans, root, dentry->d_inode);
39279cc3
CM
1049 dir->i_sb->s_dirt = 1;
1050 }
1051 return ret;
1052}
1053
1054static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
1055{
1056 struct btrfs_root *root;
1057 struct btrfs_trans_handle *trans;
1058 int ret;
1832a6d5 1059 unsigned long nr = 0;
39279cc3
CM
1060
1061 root = BTRFS_I(dir)->root;
1832a6d5
CM
1062
1063 ret = btrfs_check_free_space(root, 1, 1);
1064 if (ret)
1065 goto fail;
1066
39279cc3 1067 trans = btrfs_start_transaction(root, 1);
5f39d397 1068
39279cc3
CM
1069 btrfs_set_trans_block_group(trans, dir);
1070 ret = btrfs_unlink_trans(trans, root, dir, dentry);
d3c2fdcf 1071 nr = trans->blocks_used;
5f39d397 1072
89ce8a63 1073 btrfs_end_transaction_throttle(trans, root);
1832a6d5 1074fail:
d3c2fdcf 1075 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM
1076 return ret;
1077}
1078
1079static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
1080{
1081 struct inode *inode = dentry->d_inode;
1832a6d5 1082 int err = 0;
39279cc3
CM
1083 int ret;
1084 struct btrfs_root *root = BTRFS_I(dir)->root;
39279cc3 1085 struct btrfs_trans_handle *trans;
1832a6d5 1086 unsigned long nr = 0;
39279cc3 1087
925baedd 1088 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
134d4512 1089 return -ENOTEMPTY;
925baedd 1090 }
134d4512 1091
1832a6d5
CM
1092 ret = btrfs_check_free_space(root, 1, 1);
1093 if (ret)
1094 goto fail;
1095
39279cc3
CM
1096 trans = btrfs_start_transaction(root, 1);
1097 btrfs_set_trans_block_group(trans, dir);
39279cc3
CM
1098
1099 /* now the directory is empty */
1100 err = btrfs_unlink_trans(trans, root, dir, dentry);
1101 if (!err) {
dbe674a9 1102 btrfs_i_size_write(inode, 0);
39279cc3 1103 }
3954401f 1104
d3c2fdcf 1105 nr = trans->blocks_used;
89ce8a63 1106 ret = btrfs_end_transaction_throttle(trans, root);
1832a6d5 1107fail:
d3c2fdcf 1108 btrfs_btree_balance_dirty(root, nr);
3954401f 1109
39279cc3
CM
1110 if (ret && !err)
1111 err = ret;
1112 return err;
1113}
1114
39279cc3
CM
1115/*
1116 * this can truncate away extent items, csum items and directory items.
1117 * It starts at a high offset and removes keys until it can't find
1118 * any higher than i_size.
1119 *
1120 * csum items that cross the new i_size are truncated to the new size
1121 * as well.
1122 */
1123static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
1124 struct btrfs_root *root,
85e21bac
CM
1125 struct inode *inode,
1126 u32 min_type)
39279cc3
CM
1127{
1128 int ret;
1129 struct btrfs_path *path;
1130 struct btrfs_key key;
5f39d397 1131 struct btrfs_key found_key;
39279cc3 1132 u32 found_type;
5f39d397 1133 struct extent_buffer *leaf;
39279cc3
CM
1134 struct btrfs_file_extent_item *fi;
1135 u64 extent_start = 0;
db94535d 1136 u64 extent_num_bytes = 0;
39279cc3 1137 u64 item_end = 0;
7bb86316 1138 u64 root_gen = 0;
d8d5f3e1 1139 u64 root_owner = 0;
39279cc3
CM
1140 int found_extent;
1141 int del_item;
85e21bac
CM
1142 int pending_del_nr = 0;
1143 int pending_del_slot = 0;
179e29e4 1144 int extent_type = -1;
3b951516 1145 u64 mask = root->sectorsize - 1;
39279cc3 1146
3b951516 1147 btrfs_drop_extent_cache(inode, inode->i_size & (~mask), (u64)-1);
39279cc3 1148 path = btrfs_alloc_path();
3c69faec 1149 path->reada = -1;
39279cc3 1150 BUG_ON(!path);
5f39d397 1151
39279cc3
CM
1152 /* FIXME, add redo link to tree so we don't leak on crash */
1153 key.objectid = inode->i_ino;
1154 key.offset = (u64)-1;
5f39d397
CM
1155 key.type = (u8)-1;
1156
85e21bac
CM
1157 btrfs_init_path(path);
1158search_again:
1159 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1160 if (ret < 0) {
1161 goto error;
1162 }
1163 if (ret > 0) {
1164 BUG_ON(path->slots[0] == 0);
1165 path->slots[0]--;
1166 }
1167
39279cc3 1168 while(1) {
39279cc3 1169 fi = NULL;
5f39d397
CM
1170 leaf = path->nodes[0];
1171 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1172 found_type = btrfs_key_type(&found_key);
39279cc3 1173
5f39d397 1174 if (found_key.objectid != inode->i_ino)
39279cc3 1175 break;
5f39d397 1176
85e21bac 1177 if (found_type < min_type)
39279cc3
CM
1178 break;
1179
5f39d397 1180 item_end = found_key.offset;
39279cc3 1181 if (found_type == BTRFS_EXTENT_DATA_KEY) {
5f39d397 1182 fi = btrfs_item_ptr(leaf, path->slots[0],
39279cc3 1183 struct btrfs_file_extent_item);
179e29e4
CM
1184 extent_type = btrfs_file_extent_type(leaf, fi);
1185 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
5f39d397 1186 item_end +=
db94535d 1187 btrfs_file_extent_num_bytes(leaf, fi);
179e29e4
CM
1188 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1189 struct btrfs_item *item = btrfs_item_nr(leaf,
1190 path->slots[0]);
1191 item_end += btrfs_file_extent_inline_len(leaf,
1192 item);
39279cc3 1193 }
008630c1 1194 item_end--;
39279cc3
CM
1195 }
1196 if (found_type == BTRFS_CSUM_ITEM_KEY) {
1197 ret = btrfs_csum_truncate(trans, root, path,
1198 inode->i_size);
1199 BUG_ON(ret);
1200 }
008630c1 1201 if (item_end < inode->i_size) {
b888db2b
CM
1202 if (found_type == BTRFS_DIR_ITEM_KEY) {
1203 found_type = BTRFS_INODE_ITEM_KEY;
1204 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
1205 found_type = BTRFS_CSUM_ITEM_KEY;
85e21bac
CM
1206 } else if (found_type == BTRFS_EXTENT_DATA_KEY) {
1207 found_type = BTRFS_XATTR_ITEM_KEY;
1208 } else if (found_type == BTRFS_XATTR_ITEM_KEY) {
1209 found_type = BTRFS_INODE_REF_KEY;
b888db2b
CM
1210 } else if (found_type) {
1211 found_type--;
1212 } else {
1213 break;
39279cc3 1214 }
a61721d5 1215 btrfs_set_key_type(&key, found_type);
85e21bac 1216 goto next;
39279cc3 1217 }
5f39d397 1218 if (found_key.offset >= inode->i_size)
39279cc3
CM
1219 del_item = 1;
1220 else
1221 del_item = 0;
1222 found_extent = 0;
1223
1224 /* FIXME, shrink the extent if the ref count is only 1 */
179e29e4
CM
1225 if (found_type != BTRFS_EXTENT_DATA_KEY)
1226 goto delete;
1227
1228 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
39279cc3 1229 u64 num_dec;
db94535d 1230 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
39279cc3 1231 if (!del_item) {
db94535d
CM
1232 u64 orig_num_bytes =
1233 btrfs_file_extent_num_bytes(leaf, fi);
1234 extent_num_bytes = inode->i_size -
5f39d397 1235 found_key.offset + root->sectorsize - 1;
b1632b10
Y
1236 extent_num_bytes = extent_num_bytes &
1237 ~((u64)root->sectorsize - 1);
db94535d
CM
1238 btrfs_set_file_extent_num_bytes(leaf, fi,
1239 extent_num_bytes);
1240 num_dec = (orig_num_bytes -
9069218d
CM
1241 extent_num_bytes);
1242 if (extent_start != 0)
1243 dec_i_blocks(inode, num_dec);
5f39d397 1244 btrfs_mark_buffer_dirty(leaf);
39279cc3 1245 } else {
db94535d
CM
1246 extent_num_bytes =
1247 btrfs_file_extent_disk_num_bytes(leaf,
1248 fi);
39279cc3 1249 /* FIXME blocksize != 4096 */
9069218d 1250 num_dec = btrfs_file_extent_num_bytes(leaf, fi);
39279cc3
CM
1251 if (extent_start != 0) {
1252 found_extent = 1;
9069218d 1253 dec_i_blocks(inode, num_dec);
39279cc3 1254 }
d8d5f3e1
CM
1255 root_gen = btrfs_header_generation(leaf);
1256 root_owner = btrfs_header_owner(leaf);
39279cc3 1257 }
9069218d
CM
1258 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
1259 if (!del_item) {
1260 u32 newsize = inode->i_size - found_key.offset;
1261 dec_i_blocks(inode, item_end + 1 -
1262 found_key.offset - newsize);
1263 newsize =
1264 btrfs_file_extent_calc_inline_size(newsize);
1265 ret = btrfs_truncate_item(trans, root, path,
1266 newsize, 1);
1267 BUG_ON(ret);
1268 } else {
1269 dec_i_blocks(inode, item_end + 1 -
1270 found_key.offset);
1271 }
39279cc3 1272 }
179e29e4 1273delete:
39279cc3 1274 if (del_item) {
85e21bac
CM
1275 if (!pending_del_nr) {
1276 /* no pending yet, add ourselves */
1277 pending_del_slot = path->slots[0];
1278 pending_del_nr = 1;
1279 } else if (pending_del_nr &&
1280 path->slots[0] + 1 == pending_del_slot) {
1281 /* hop on the pending chunk */
1282 pending_del_nr++;
1283 pending_del_slot = path->slots[0];
1284 } else {
1285 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot);
1286 }
39279cc3
CM
1287 } else {
1288 break;
1289 }
39279cc3
CM
1290 if (found_extent) {
1291 ret = btrfs_free_extent(trans, root, extent_start,
7bb86316 1292 extent_num_bytes,
d8d5f3e1 1293 root_owner,
7bb86316
CM
1294 root_gen, inode->i_ino,
1295 found_key.offset, 0);
39279cc3
CM
1296 BUG_ON(ret);
1297 }
85e21bac
CM
1298next:
1299 if (path->slots[0] == 0) {
1300 if (pending_del_nr)
1301 goto del_pending;
1302 btrfs_release_path(root, path);
1303 goto search_again;
1304 }
1305
1306 path->slots[0]--;
1307 if (pending_del_nr &&
1308 path->slots[0] + 1 != pending_del_slot) {
1309 struct btrfs_key debug;
1310del_pending:
1311 btrfs_item_key_to_cpu(path->nodes[0], &debug,
1312 pending_del_slot);
1313 ret = btrfs_del_items(trans, root, path,
1314 pending_del_slot,
1315 pending_del_nr);
1316 BUG_ON(ret);
1317 pending_del_nr = 0;
1318 btrfs_release_path(root, path);
1319 goto search_again;
1320 }
39279cc3
CM
1321 }
1322 ret = 0;
1323error:
85e21bac
CM
1324 if (pending_del_nr) {
1325 ret = btrfs_del_items(trans, root, path, pending_del_slot,
1326 pending_del_nr);
1327 }
39279cc3
CM
1328 btrfs_free_path(path);
1329 inode->i_sb->s_dirt = 1;
1330 return ret;
1331}
1332
1333/*
1334 * taken from block_truncate_page, but does cow as it zeros out
1335 * any bytes left in the last page in the file.
1336 */
1337static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
1338{
1339 struct inode *inode = mapping->host;
db94535d 1340 struct btrfs_root *root = BTRFS_I(inode)->root;
e6dcd2dc
CM
1341 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1342 struct btrfs_ordered_extent *ordered;
1343 char *kaddr;
db94535d 1344 u32 blocksize = root->sectorsize;
39279cc3
CM
1345 pgoff_t index = from >> PAGE_CACHE_SHIFT;
1346 unsigned offset = from & (PAGE_CACHE_SIZE-1);
1347 struct page *page;
39279cc3 1348 int ret = 0;
a52d9a80 1349 u64 page_start;
e6dcd2dc 1350 u64 page_end;
39279cc3
CM
1351
1352 if ((offset & (blocksize - 1)) == 0)
1353 goto out;
1354
1355 ret = -ENOMEM;
211c17f5 1356again:
39279cc3
CM
1357 page = grab_cache_page(mapping, index);
1358 if (!page)
1359 goto out;
e6dcd2dc
CM
1360
1361 page_start = page_offset(page);
1362 page_end = page_start + PAGE_CACHE_SIZE - 1;
1363
39279cc3 1364 if (!PageUptodate(page)) {
9ebefb18 1365 ret = btrfs_readpage(NULL, page);
39279cc3 1366 lock_page(page);
211c17f5
CM
1367 if (page->mapping != mapping) {
1368 unlock_page(page);
1369 page_cache_release(page);
1370 goto again;
1371 }
39279cc3
CM
1372 if (!PageUptodate(page)) {
1373 ret = -EIO;
1374 goto out;
1375 }
1376 }
211c17f5 1377 wait_on_page_writeback(page);
e6dcd2dc
CM
1378
1379 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
1380 set_page_extent_mapped(page);
1381
1382 ordered = btrfs_lookup_ordered_extent(inode, page_start);
1383 if (ordered) {
1384 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
1385 unlock_page(page);
1386 page_cache_release(page);
eb84ae03 1387 btrfs_start_ordered_extent(inode, ordered, 1);
e6dcd2dc
CM
1388 btrfs_put_ordered_extent(ordered);
1389 goto again;
1390 }
1391
1392 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
1393 page_end, GFP_NOFS);
1394 ret = 0;
1395 if (offset != PAGE_CACHE_SIZE) {
1396 kaddr = kmap(page);
1397 memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
1398 flush_dcache_page(page);
1399 kunmap(page);
1400 }
247e743c 1401 ClearPageChecked(page);
e6dcd2dc
CM
1402 set_page_dirty(page);
1403 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
39279cc3 1404
39279cc3
CM
1405 unlock_page(page);
1406 page_cache_release(page);
1407out:
1408 return ret;
1409}
1410
1411static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
1412{
1413 struct inode *inode = dentry->d_inode;
1414 int err;
1415
1416 err = inode_change_ok(inode, attr);
1417 if (err)
1418 return err;
1419
1420 if (S_ISREG(inode->i_mode) &&
1421 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
1422 struct btrfs_trans_handle *trans;
1423 struct btrfs_root *root = BTRFS_I(inode)->root;
d1310b2e 1424 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2bf5a725 1425
5f39d397 1426 u64 mask = root->sectorsize - 1;
1b0f7c29 1427 u64 hole_start = (inode->i_size + mask) & ~mask;
f392a938 1428 u64 block_end = (attr->ia_size + mask) & ~mask;
39279cc3 1429 u64 hole_size;
179e29e4 1430 u64 alloc_hint = 0;
39279cc3 1431
1b0f7c29 1432 if (attr->ia_size <= hole_start)
39279cc3
CM
1433 goto out;
1434
1832a6d5 1435 err = btrfs_check_free_space(root, 1, 0);
1832a6d5
CM
1436 if (err)
1437 goto fail;
1438
39279cc3
CM
1439 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1440
5f56406a 1441 hole_size = block_end - hole_start;
e6dcd2dc
CM
1442 btrfs_wait_ordered_range(inode, hole_start, hole_size);
1443 lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
39279cc3 1444
39279cc3
CM
1445 trans = btrfs_start_transaction(root, 1);
1446 btrfs_set_trans_block_group(trans, inode);
ee6e6504 1447 mutex_lock(&BTRFS_I(inode)->extent_mutex);
2bf5a725 1448 err = btrfs_drop_extents(trans, root, inode,
1b0f7c29 1449 hole_start, block_end, hole_start,
3326d1b0 1450 &alloc_hint);
2bf5a725 1451
179e29e4
CM
1452 if (alloc_hint != EXTENT_MAP_INLINE) {
1453 err = btrfs_insert_file_extent(trans, root,
1454 inode->i_ino,
5f56406a 1455 hole_start, 0, 0,
f2eb0a24 1456 hole_size, 0);
d1310b2e 1457 btrfs_drop_extent_cache(inode, hole_start,
3b951516 1458 (u64)-1);
5f56406a 1459 btrfs_check_file(root, inode);
179e29e4 1460 }
ee6e6504 1461 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
39279cc3 1462 btrfs_end_transaction(trans, root);
1b0f7c29 1463 unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
54aa1f4d
CM
1464 if (err)
1465 return err;
39279cc3
CM
1466 }
1467out:
1468 err = inode_setattr(inode, attr);
1832a6d5 1469fail:
39279cc3
CM
1470 return err;
1471}
61295eb8 1472
39279cc3
CM
1473void btrfs_delete_inode(struct inode *inode)
1474{
1475 struct btrfs_trans_handle *trans;
1476 struct btrfs_root *root = BTRFS_I(inode)->root;
d3c2fdcf 1477 unsigned long nr;
39279cc3
CM
1478 int ret;
1479
1480 truncate_inode_pages(&inode->i_data, 0);
1481 if (is_bad_inode(inode)) {
1482 goto no_delete;
1483 }
4a096752 1484 btrfs_wait_ordered_range(inode, 0, (u64)-1);
5f39d397 1485
dbe674a9 1486 btrfs_i_size_write(inode, 0);
39279cc3 1487 trans = btrfs_start_transaction(root, 1);
5f39d397 1488
39279cc3 1489 btrfs_set_trans_block_group(trans, inode);
85e21bac 1490 ret = btrfs_truncate_in_trans(trans, root, inode, 0);
54aa1f4d
CM
1491 if (ret)
1492 goto no_delete_lock;
85e21bac 1493
d3c2fdcf 1494 nr = trans->blocks_used;
85e21bac 1495 clear_inode(inode);
5f39d397 1496
39279cc3 1497 btrfs_end_transaction(trans, root);
d3c2fdcf 1498 btrfs_btree_balance_dirty(root, nr);
39279cc3 1499 return;
54aa1f4d
CM
1500
1501no_delete_lock:
d3c2fdcf 1502 nr = trans->blocks_used;
54aa1f4d 1503 btrfs_end_transaction(trans, root);
d3c2fdcf 1504 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM
1505no_delete:
1506 clear_inode(inode);
1507}
1508
1509/*
1510 * this returns the key found in the dir entry in the location pointer.
1511 * If no dir entries were found, location->objectid is 0.
1512 */
1513static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
1514 struct btrfs_key *location)
1515{
1516 const char *name = dentry->d_name.name;
1517 int namelen = dentry->d_name.len;
1518 struct btrfs_dir_item *di;
1519 struct btrfs_path *path;
1520 struct btrfs_root *root = BTRFS_I(dir)->root;
0d9f7f3e 1521 int ret = 0;
39279cc3 1522
3954401f
CM
1523 if (namelen == 1 && strcmp(name, ".") == 0) {
1524 location->objectid = dir->i_ino;
1525 location->type = BTRFS_INODE_ITEM_KEY;
1526 location->offset = 0;
1527 return 0;
1528 }
39279cc3
CM
1529 path = btrfs_alloc_path();
1530 BUG_ON(!path);
3954401f 1531
7a720536 1532 if (namelen == 2 && strcmp(name, "..") == 0) {
3954401f
CM
1533 struct btrfs_key key;
1534 struct extent_buffer *leaf;
1535 u32 nritems;
1536 int slot;
1537
1538 key.objectid = dir->i_ino;
1539 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1540 key.offset = 0;
1541 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1542 BUG_ON(ret == 0);
1543 ret = 0;
1544
1545 leaf = path->nodes[0];
1546 slot = path->slots[0];
1547 nritems = btrfs_header_nritems(leaf);
1548 if (slot >= nritems)
1549 goto out_err;
1550
1551 btrfs_item_key_to_cpu(leaf, &key, slot);
1552 if (key.objectid != dir->i_ino ||
1553 key.type != BTRFS_INODE_REF_KEY) {
1554 goto out_err;
1555 }
1556 location->objectid = key.offset;
1557 location->type = BTRFS_INODE_ITEM_KEY;
1558 location->offset = 0;
1559 goto out;
1560 }
1561
39279cc3
CM
1562 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
1563 namelen, 0);
0d9f7f3e
Y
1564 if (IS_ERR(di))
1565 ret = PTR_ERR(di);
39279cc3 1566 if (!di || IS_ERR(di)) {
3954401f 1567 goto out_err;
39279cc3 1568 }
5f39d397 1569 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
39279cc3 1570out:
39279cc3
CM
1571 btrfs_free_path(path);
1572 return ret;
3954401f
CM
1573out_err:
1574 location->objectid = 0;
1575 goto out;
39279cc3
CM
1576}
1577
1578/*
1579 * when we hit a tree root in a directory, the btrfs part of the inode
1580 * needs to be changed to reflect the root directory of the tree root. This
1581 * is kind of like crossing a mount point.
1582 */
1583static int fixup_tree_root_location(struct btrfs_root *root,
1584 struct btrfs_key *location,
58176a96
JB
1585 struct btrfs_root **sub_root,
1586 struct dentry *dentry)
39279cc3
CM
1587{
1588 struct btrfs_path *path;
1589 struct btrfs_root_item *ri;
1590
1591 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
1592 return 0;
1593 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1594 return 0;
1595
1596 path = btrfs_alloc_path();
1597 BUG_ON(!path);
39279cc3 1598
58176a96
JB
1599 *sub_root = btrfs_read_fs_root(root->fs_info, location,
1600 dentry->d_name.name,
1601 dentry->d_name.len);
39279cc3
CM
1602 if (IS_ERR(*sub_root))
1603 return PTR_ERR(*sub_root);
1604
1605 ri = &(*sub_root)->root_item;
1606 location->objectid = btrfs_root_dirid(ri);
39279cc3
CM
1607 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1608 location->offset = 0;
1609
1610 btrfs_free_path(path);
39279cc3
CM
1611 return 0;
1612}
1613
1614static int btrfs_init_locked_inode(struct inode *inode, void *p)
1615{
1616 struct btrfs_iget_args *args = p;
1617 inode->i_ino = args->ino;
1618 BTRFS_I(inode)->root = args->root;
9069218d 1619 BTRFS_I(inode)->delalloc_bytes = 0;
dbe674a9 1620 BTRFS_I(inode)->disk_i_size = 0;
d1310b2e
CM
1621 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1622 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
b888db2b 1623 inode->i_mapping, GFP_NOFS);
7e38326f
CM
1624 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1625 inode->i_mapping, GFP_NOFS);
ba1da2f4 1626 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
1b1e2135 1627 mutex_init(&BTRFS_I(inode)->csum_mutex);
ee6e6504 1628 mutex_init(&BTRFS_I(inode)->extent_mutex);
39279cc3
CM
1629 return 0;
1630}
1631
1632static int btrfs_find_actor(struct inode *inode, void *opaque)
1633{
1634 struct btrfs_iget_args *args = opaque;
1635 return (args->ino == inode->i_ino &&
1636 args->root == BTRFS_I(inode)->root);
1637}
1638
dc17ff8f
CM
1639struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
1640 u64 root_objectid)
1641{
1642 struct btrfs_iget_args args;
1643 args.ino = objectid;
1644 args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
1645
1646 if (!args.root)
1647 return NULL;
1648
1649 return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
1650}
1651
39279cc3
CM
1652struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
1653 struct btrfs_root *root)
1654{
1655 struct inode *inode;
1656 struct btrfs_iget_args args;
1657 args.ino = objectid;
1658 args.root = root;
1659
1660 inode = iget5_locked(s, objectid, btrfs_find_actor,
1661 btrfs_init_locked_inode,
1662 (void *)&args);
1663 return inode;
1664}
1665
1666static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
1667 struct nameidata *nd)
1668{
1669 struct inode * inode;
1670 struct btrfs_inode *bi = BTRFS_I(dir);
1671 struct btrfs_root *root = bi->root;
1672 struct btrfs_root *sub_root = root;
1673 struct btrfs_key location;
1674 int ret;
1675
1676 if (dentry->d_name.len > BTRFS_NAME_LEN)
1677 return ERR_PTR(-ENAMETOOLONG);
5f39d397 1678
39279cc3 1679 ret = btrfs_inode_by_name(dir, dentry, &location);
5f39d397 1680
39279cc3
CM
1681 if (ret < 0)
1682 return ERR_PTR(ret);
5f39d397 1683
39279cc3
CM
1684 inode = NULL;
1685 if (location.objectid) {
58176a96
JB
1686 ret = fixup_tree_root_location(root, &location, &sub_root,
1687 dentry);
39279cc3
CM
1688 if (ret < 0)
1689 return ERR_PTR(ret);
1690 if (ret > 0)
1691 return ERR_PTR(-ENOENT);
1692 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
1693 sub_root);
1694 if (!inode)
1695 return ERR_PTR(-EACCES);
1696 if (inode->i_state & I_NEW) {
1697 /* the inode and parent dir are two different roots */
1698 if (sub_root != root) {
1699 igrab(inode);
1700 sub_root->inode = inode;
1701 }
1702 BTRFS_I(inode)->root = sub_root;
1703 memcpy(&BTRFS_I(inode)->location, &location,
1704 sizeof(location));
1705 btrfs_read_locked_inode(inode);
1706 unlock_new_inode(inode);
1707 }
1708 }
1709 return d_splice_alias(inode, dentry);
1710}
1711
39279cc3
CM
1712static unsigned char btrfs_filetype_table[] = {
1713 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1714};
1715
1716static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1717{
6da6abae 1718 struct inode *inode = filp->f_dentry->d_inode;
39279cc3
CM
1719 struct btrfs_root *root = BTRFS_I(inode)->root;
1720 struct btrfs_item *item;
1721 struct btrfs_dir_item *di;
1722 struct btrfs_key key;
5f39d397 1723 struct btrfs_key found_key;
39279cc3
CM
1724 struct btrfs_path *path;
1725 int ret;
1726 u32 nritems;
5f39d397 1727 struct extent_buffer *leaf;
39279cc3
CM
1728 int slot;
1729 int advance;
1730 unsigned char d_type;
1731 int over = 0;
1732 u32 di_cur;
1733 u32 di_total;
1734 u32 di_len;
1735 int key_type = BTRFS_DIR_INDEX_KEY;
5f39d397
CM
1736 char tmp_name[32];
1737 char *name_ptr;
1738 int name_len;
39279cc3
CM
1739
1740 /* FIXME, use a real flag for deciding about the key type */
1741 if (root->fs_info->tree_root == root)
1742 key_type = BTRFS_DIR_ITEM_KEY;
5f39d397 1743
3954401f
CM
1744 /* special case for "." */
1745 if (filp->f_pos == 0) {
1746 over = filldir(dirent, ".", 1,
1747 1, inode->i_ino,
1748 DT_DIR);
1749 if (over)
1750 return 0;
1751 filp->f_pos = 1;
1752 }
1753
39279cc3 1754 key.objectid = inode->i_ino;
3954401f
CM
1755 path = btrfs_alloc_path();
1756 path->reada = 2;
1757
1758 /* special case for .., just use the back ref */
1759 if (filp->f_pos == 1) {
1760 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1761 key.offset = 0;
1762 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1763 BUG_ON(ret == 0);
1764 leaf = path->nodes[0];
1765 slot = path->slots[0];
1766 nritems = btrfs_header_nritems(leaf);
1767 if (slot >= nritems) {
1768 btrfs_release_path(root, path);
1769 goto read_dir_items;
1770 }
1771 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1772 btrfs_release_path(root, path);
1773 if (found_key.objectid != key.objectid ||
1774 found_key.type != BTRFS_INODE_REF_KEY)
1775 goto read_dir_items;
1776 over = filldir(dirent, "..", 2,
1777 2, found_key.offset, DT_DIR);
1778 if (over)
1779 goto nopos;
1780 filp->f_pos = 2;
1781 }
1782
1783read_dir_items:
39279cc3
CM
1784 btrfs_set_key_type(&key, key_type);
1785 key.offset = filp->f_pos;
5f39d397 1786
39279cc3
CM
1787 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1788 if (ret < 0)
1789 goto err;
1790 advance = 0;
39279cc3 1791 while(1) {
5f39d397
CM
1792 leaf = path->nodes[0];
1793 nritems = btrfs_header_nritems(leaf);
39279cc3
CM
1794 slot = path->slots[0];
1795 if (advance || slot >= nritems) {
1796 if (slot >= nritems -1) {
39279cc3
CM
1797 ret = btrfs_next_leaf(root, path);
1798 if (ret)
1799 break;
5f39d397
CM
1800 leaf = path->nodes[0];
1801 nritems = btrfs_header_nritems(leaf);
39279cc3
CM
1802 slot = path->slots[0];
1803 } else {
1804 slot++;
1805 path->slots[0]++;
1806 }
1807 }
1808 advance = 1;
5f39d397
CM
1809 item = btrfs_item_nr(leaf, slot);
1810 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1811
1812 if (found_key.objectid != key.objectid)
39279cc3 1813 break;
5f39d397 1814 if (btrfs_key_type(&found_key) != key_type)
39279cc3 1815 break;
5f39d397 1816 if (found_key.offset < filp->f_pos)
39279cc3 1817 continue;
5f39d397
CM
1818
1819 filp->f_pos = found_key.offset;
39279cc3
CM
1820 advance = 1;
1821 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1822 di_cur = 0;
5f39d397 1823 di_total = btrfs_item_size(leaf, item);
39279cc3 1824 while(di_cur < di_total) {
5f39d397
CM
1825 struct btrfs_key location;
1826
1827 name_len = btrfs_dir_name_len(leaf, di);
1828 if (name_len < 32) {
1829 name_ptr = tmp_name;
1830 } else {
1831 name_ptr = kmalloc(name_len, GFP_NOFS);
1832 BUG_ON(!name_ptr);
1833 }
1834 read_extent_buffer(leaf, name_ptr,
1835 (unsigned long)(di + 1), name_len);
1836
1837 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1838 btrfs_dir_item_key_to_cpu(leaf, di, &location);
5f39d397
CM
1839 over = filldir(dirent, name_ptr, name_len,
1840 found_key.offset,
1841 location.objectid,
39279cc3 1842 d_type);
5f39d397
CM
1843
1844 if (name_ptr != tmp_name)
1845 kfree(name_ptr);
1846
39279cc3
CM
1847 if (over)
1848 goto nopos;
5103e947
JB
1849 di_len = btrfs_dir_name_len(leaf, di) +
1850 btrfs_dir_data_len(leaf, di) +sizeof(*di);
39279cc3
CM
1851 di_cur += di_len;
1852 di = (struct btrfs_dir_item *)((char *)di + di_len);
1853 }
1854 }
5e591a07
YZ
1855 if (key_type == BTRFS_DIR_INDEX_KEY)
1856 filp->f_pos = INT_LIMIT(typeof(filp->f_pos));
1857 else
1858 filp->f_pos++;
39279cc3
CM
1859nopos:
1860 ret = 0;
1861err:
39279cc3 1862 btrfs_free_path(path);
39279cc3
CM
1863 return ret;
1864}
1865
1866int btrfs_write_inode(struct inode *inode, int wait)
1867{
1868 struct btrfs_root *root = BTRFS_I(inode)->root;
1869 struct btrfs_trans_handle *trans;
1870 int ret = 0;
1871
1872 if (wait) {
f9295749 1873 trans = btrfs_join_transaction(root, 1);
39279cc3
CM
1874 btrfs_set_trans_block_group(trans, inode);
1875 ret = btrfs_commit_transaction(trans, root);
39279cc3
CM
1876 }
1877 return ret;
1878}
1879
1880/*
54aa1f4d 1881 * This is somewhat expensive, updating the tree every time the
39279cc3
CM
1882 * inode changes. But, it is most likely to find the inode in cache.
1883 * FIXME, needs more benchmarking...there are no reasons other than performance
1884 * to keep or drop this code.
1885 */
1886void btrfs_dirty_inode(struct inode *inode)
1887{
1888 struct btrfs_root *root = BTRFS_I(inode)->root;
1889 struct btrfs_trans_handle *trans;
1890
f9295749 1891 trans = btrfs_join_transaction(root, 1);
39279cc3
CM
1892 btrfs_set_trans_block_group(trans, inode);
1893 btrfs_update_inode(trans, root, inode);
1894 btrfs_end_transaction(trans, root);
39279cc3
CM
1895}
1896
1897static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1898 struct btrfs_root *root,
9c58309d
CM
1899 const char *name, int name_len,
1900 u64 ref_objectid,
39279cc3
CM
1901 u64 objectid,
1902 struct btrfs_block_group_cache *group,
1903 int mode)
1904{
1905 struct inode *inode;
5f39d397 1906 struct btrfs_inode_item *inode_item;
6324fbf3 1907 struct btrfs_block_group_cache *new_inode_group;
39279cc3 1908 struct btrfs_key *location;
5f39d397 1909 struct btrfs_path *path;
9c58309d
CM
1910 struct btrfs_inode_ref *ref;
1911 struct btrfs_key key[2];
1912 u32 sizes[2];
1913 unsigned long ptr;
39279cc3
CM
1914 int ret;
1915 int owner;
1916
5f39d397
CM
1917 path = btrfs_alloc_path();
1918 BUG_ON(!path);
1919
39279cc3
CM
1920 inode = new_inode(root->fs_info->sb);
1921 if (!inode)
1922 return ERR_PTR(-ENOMEM);
1923
d1310b2e
CM
1924 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1925 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
b888db2b 1926 inode->i_mapping, GFP_NOFS);
7e38326f
CM
1927 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
1928 inode->i_mapping, GFP_NOFS);
ba1da2f4 1929 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
1b1e2135 1930 mutex_init(&BTRFS_I(inode)->csum_mutex);
ee6e6504 1931 mutex_init(&BTRFS_I(inode)->extent_mutex);
9069218d 1932 BTRFS_I(inode)->delalloc_bytes = 0;
dbe674a9 1933 BTRFS_I(inode)->disk_i_size = 0;
39279cc3 1934 BTRFS_I(inode)->root = root;
b888db2b 1935
39279cc3
CM
1936 if (mode & S_IFDIR)
1937 owner = 0;
1938 else
1939 owner = 1;
6324fbf3 1940 new_inode_group = btrfs_find_block_group(root, group, 0,
0b86a832 1941 BTRFS_BLOCK_GROUP_METADATA, owner);
6324fbf3
CM
1942 if (!new_inode_group) {
1943 printk("find_block group failed\n");
1944 new_inode_group = group;
1945 }
1946 BTRFS_I(inode)->block_group = new_inode_group;
b98b6767 1947 BTRFS_I(inode)->flags = 0;
9c58309d
CM
1948
1949 key[0].objectid = objectid;
1950 btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
1951 key[0].offset = 0;
1952
1953 key[1].objectid = objectid;
1954 btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
1955 key[1].offset = ref_objectid;
1956
1957 sizes[0] = sizeof(struct btrfs_inode_item);
1958 sizes[1] = name_len + sizeof(*ref);
1959
1960 ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
1961 if (ret != 0)
5f39d397
CM
1962 goto fail;
1963
9c58309d
CM
1964 if (objectid > root->highest_inode)
1965 root->highest_inode = objectid;
1966
39279cc3
CM
1967 inode->i_uid = current->fsuid;
1968 inode->i_gid = current->fsgid;
1969 inode->i_mode = mode;
1970 inode->i_ino = objectid;
1971 inode->i_blocks = 0;
1972 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
5f39d397
CM
1973 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1974 struct btrfs_inode_item);
1975 fill_inode_item(path->nodes[0], inode_item, inode);
9c58309d
CM
1976
1977 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
1978 struct btrfs_inode_ref);
1979 btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
1980 ptr = (unsigned long)(ref + 1);
1981 write_extent_buffer(path->nodes[0], name, ptr, name_len);
1982
5f39d397
CM
1983 btrfs_mark_buffer_dirty(path->nodes[0]);
1984 btrfs_free_path(path);
1985
39279cc3
CM
1986 location = &BTRFS_I(inode)->location;
1987 location->objectid = objectid;
39279cc3
CM
1988 location->offset = 0;
1989 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1990
39279cc3
CM
1991 insert_inode_hash(inode);
1992 return inode;
5f39d397
CM
1993fail:
1994 btrfs_free_path(path);
1995 return ERR_PTR(ret);
39279cc3
CM
1996}
1997
1998static inline u8 btrfs_inode_type(struct inode *inode)
1999{
2000 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
2001}
2002
2003static int btrfs_add_link(struct btrfs_trans_handle *trans,
9c58309d
CM
2004 struct dentry *dentry, struct inode *inode,
2005 int add_backref)
39279cc3
CM
2006{
2007 int ret;
2008 struct btrfs_key key;
2009 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
79c44584 2010 struct inode *parent_inode;
5f39d397 2011
39279cc3 2012 key.objectid = inode->i_ino;
39279cc3
CM
2013 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
2014 key.offset = 0;
2015
2016 ret = btrfs_insert_dir_item(trans, root,
2017 dentry->d_name.name, dentry->d_name.len,
2018 dentry->d_parent->d_inode->i_ino,
2019 &key, btrfs_inode_type(inode));
2020 if (ret == 0) {
9c58309d
CM
2021 if (add_backref) {
2022 ret = btrfs_insert_inode_ref(trans, root,
2023 dentry->d_name.name,
2024 dentry->d_name.len,
2025 inode->i_ino,
2026 dentry->d_parent->d_inode->i_ino);
2027 }
79c44584 2028 parent_inode = dentry->d_parent->d_inode;
dbe674a9
CM
2029 btrfs_i_size_write(parent_inode, parent_inode->i_size +
2030 dentry->d_name.len * 2);
79c44584 2031 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
39279cc3
CM
2032 ret = btrfs_update_inode(trans, root,
2033 dentry->d_parent->d_inode);
2034 }
2035 return ret;
2036}
2037
2038static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
9c58309d
CM
2039 struct dentry *dentry, struct inode *inode,
2040 int backref)
39279cc3 2041{
9c58309d 2042 int err = btrfs_add_link(trans, dentry, inode, backref);
39279cc3
CM
2043 if (!err) {
2044 d_instantiate(dentry, inode);
2045 return 0;
2046 }
2047 if (err > 0)
2048 err = -EEXIST;
2049 return err;
2050}
2051
618e21d5
JB
2052static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
2053 int mode, dev_t rdev)
2054{
2055 struct btrfs_trans_handle *trans;
2056 struct btrfs_root *root = BTRFS_I(dir)->root;
1832a6d5 2057 struct inode *inode = NULL;
618e21d5
JB
2058 int err;
2059 int drop_inode = 0;
2060 u64 objectid;
1832a6d5 2061 unsigned long nr = 0;
618e21d5
JB
2062
2063 if (!new_valid_dev(rdev))
2064 return -EINVAL;
2065
1832a6d5
CM
2066 err = btrfs_check_free_space(root, 1, 0);
2067 if (err)
2068 goto fail;
2069
618e21d5
JB
2070 trans = btrfs_start_transaction(root, 1);
2071 btrfs_set_trans_block_group(trans, dir);
2072
2073 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2074 if (err) {
2075 err = -ENOSPC;
2076 goto out_unlock;
2077 }
2078
9c58309d
CM
2079 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2080 dentry->d_name.len,
2081 dentry->d_parent->d_inode->i_ino, objectid,
618e21d5
JB
2082 BTRFS_I(dir)->block_group, mode);
2083 err = PTR_ERR(inode);
2084 if (IS_ERR(inode))
2085 goto out_unlock;
2086
2087 btrfs_set_trans_block_group(trans, inode);
9c58309d 2088 err = btrfs_add_nondir(trans, dentry, inode, 0);
618e21d5
JB
2089 if (err)
2090 drop_inode = 1;
2091 else {
2092 inode->i_op = &btrfs_special_inode_operations;
2093 init_special_inode(inode, inode->i_mode, rdev);
1b4ab1bb 2094 btrfs_update_inode(trans, root, inode);
618e21d5
JB
2095 }
2096 dir->i_sb->s_dirt = 1;
2097 btrfs_update_inode_block_group(trans, inode);
2098 btrfs_update_inode_block_group(trans, dir);
2099out_unlock:
d3c2fdcf 2100 nr = trans->blocks_used;
89ce8a63 2101 btrfs_end_transaction_throttle(trans, root);
1832a6d5 2102fail:
618e21d5
JB
2103 if (drop_inode) {
2104 inode_dec_link_count(inode);
2105 iput(inode);
2106 }
d3c2fdcf 2107 btrfs_btree_balance_dirty(root, nr);
618e21d5
JB
2108 return err;
2109}
2110
39279cc3
CM
2111static int btrfs_create(struct inode *dir, struct dentry *dentry,
2112 int mode, struct nameidata *nd)
2113{
2114 struct btrfs_trans_handle *trans;
2115 struct btrfs_root *root = BTRFS_I(dir)->root;
1832a6d5 2116 struct inode *inode = NULL;
39279cc3
CM
2117 int err;
2118 int drop_inode = 0;
1832a6d5 2119 unsigned long nr = 0;
39279cc3
CM
2120 u64 objectid;
2121
1832a6d5
CM
2122 err = btrfs_check_free_space(root, 1, 0);
2123 if (err)
2124 goto fail;
39279cc3
CM
2125 trans = btrfs_start_transaction(root, 1);
2126 btrfs_set_trans_block_group(trans, dir);
2127
2128 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2129 if (err) {
2130 err = -ENOSPC;
2131 goto out_unlock;
2132 }
2133
9c58309d
CM
2134 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2135 dentry->d_name.len,
2136 dentry->d_parent->d_inode->i_ino,
2137 objectid, BTRFS_I(dir)->block_group, mode);
39279cc3
CM
2138 err = PTR_ERR(inode);
2139 if (IS_ERR(inode))
2140 goto out_unlock;
2141
2142 btrfs_set_trans_block_group(trans, inode);
9c58309d 2143 err = btrfs_add_nondir(trans, dentry, inode, 0);
39279cc3
CM
2144 if (err)
2145 drop_inode = 1;
2146 else {
2147 inode->i_mapping->a_ops = &btrfs_aops;
04160088 2148 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
39279cc3
CM
2149 inode->i_fop = &btrfs_file_operations;
2150 inode->i_op = &btrfs_file_inode_operations;
d1310b2e
CM
2151 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2152 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
a52d9a80 2153 inode->i_mapping, GFP_NOFS);
7e38326f
CM
2154 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
2155 inode->i_mapping, GFP_NOFS);
1b1e2135 2156 mutex_init(&BTRFS_I(inode)->csum_mutex);
ee6e6504 2157 mutex_init(&BTRFS_I(inode)->extent_mutex);
9069218d 2158 BTRFS_I(inode)->delalloc_bytes = 0;
dbe674a9 2159 BTRFS_I(inode)->disk_i_size = 0;
d1310b2e 2160 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
ba1da2f4 2161 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
39279cc3
CM
2162 }
2163 dir->i_sb->s_dirt = 1;
2164 btrfs_update_inode_block_group(trans, inode);
2165 btrfs_update_inode_block_group(trans, dir);
2166out_unlock:
d3c2fdcf 2167 nr = trans->blocks_used;
89ce8a63 2168 btrfs_end_transaction_throttle(trans, root);
1832a6d5 2169fail:
39279cc3
CM
2170 if (drop_inode) {
2171 inode_dec_link_count(inode);
2172 iput(inode);
2173 }
d3c2fdcf 2174 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM
2175 return err;
2176}
2177
2178static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
2179 struct dentry *dentry)
2180{
2181 struct btrfs_trans_handle *trans;
2182 struct btrfs_root *root = BTRFS_I(dir)->root;
2183 struct inode *inode = old_dentry->d_inode;
1832a6d5 2184 unsigned long nr = 0;
39279cc3
CM
2185 int err;
2186 int drop_inode = 0;
2187
2188 if (inode->i_nlink == 0)
2189 return -ENOENT;
2190
6da6abae
CM
2191#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2192 inode->i_nlink++;
2193#else
39279cc3 2194 inc_nlink(inode);
6da6abae 2195#endif
1832a6d5
CM
2196 err = btrfs_check_free_space(root, 1, 0);
2197 if (err)
2198 goto fail;
39279cc3 2199 trans = btrfs_start_transaction(root, 1);
5f39d397 2200
39279cc3
CM
2201 btrfs_set_trans_block_group(trans, dir);
2202 atomic_inc(&inode->i_count);
9c58309d 2203 err = btrfs_add_nondir(trans, dentry, inode, 1);
5f39d397 2204
39279cc3
CM
2205 if (err)
2206 drop_inode = 1;
5f39d397 2207
39279cc3
CM
2208 dir->i_sb->s_dirt = 1;
2209 btrfs_update_inode_block_group(trans, dir);
54aa1f4d 2210 err = btrfs_update_inode(trans, root, inode);
5f39d397 2211
54aa1f4d
CM
2212 if (err)
2213 drop_inode = 1;
39279cc3 2214
d3c2fdcf 2215 nr = trans->blocks_used;
89ce8a63 2216 btrfs_end_transaction_throttle(trans, root);
1832a6d5 2217fail:
39279cc3
CM
2218 if (drop_inode) {
2219 inode_dec_link_count(inode);
2220 iput(inode);
2221 }
d3c2fdcf 2222 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM
2223 return err;
2224}
2225
39279cc3
CM
2226static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
2227{
b9d86667 2228 struct inode *inode = NULL;
39279cc3
CM
2229 struct btrfs_trans_handle *trans;
2230 struct btrfs_root *root = BTRFS_I(dir)->root;
2231 int err = 0;
2232 int drop_on_err = 0;
b9d86667 2233 u64 objectid = 0;
d3c2fdcf 2234 unsigned long nr = 1;
39279cc3 2235
1832a6d5
CM
2236 err = btrfs_check_free_space(root, 1, 0);
2237 if (err)
2238 goto out_unlock;
2239
39279cc3
CM
2240 trans = btrfs_start_transaction(root, 1);
2241 btrfs_set_trans_block_group(trans, dir);
5f39d397 2242
39279cc3
CM
2243 if (IS_ERR(trans)) {
2244 err = PTR_ERR(trans);
2245 goto out_unlock;
2246 }
2247
2248 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2249 if (err) {
2250 err = -ENOSPC;
2251 goto out_unlock;
2252 }
2253
9c58309d
CM
2254 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2255 dentry->d_name.len,
2256 dentry->d_parent->d_inode->i_ino, objectid,
39279cc3
CM
2257 BTRFS_I(dir)->block_group, S_IFDIR | mode);
2258 if (IS_ERR(inode)) {
2259 err = PTR_ERR(inode);
2260 goto out_fail;
2261 }
5f39d397 2262
39279cc3
CM
2263 drop_on_err = 1;
2264 inode->i_op = &btrfs_dir_inode_operations;
2265 inode->i_fop = &btrfs_dir_file_operations;
2266 btrfs_set_trans_block_group(trans, inode);
2267
dbe674a9 2268 btrfs_i_size_write(inode, 0);
39279cc3
CM
2269 err = btrfs_update_inode(trans, root, inode);
2270 if (err)
2271 goto out_fail;
5f39d397 2272
9c58309d 2273 err = btrfs_add_link(trans, dentry, inode, 0);
39279cc3
CM
2274 if (err)
2275 goto out_fail;
5f39d397 2276
39279cc3
CM
2277 d_instantiate(dentry, inode);
2278 drop_on_err = 0;
2279 dir->i_sb->s_dirt = 1;
2280 btrfs_update_inode_block_group(trans, inode);
2281 btrfs_update_inode_block_group(trans, dir);
2282
2283out_fail:
d3c2fdcf 2284 nr = trans->blocks_used;
89ce8a63 2285 btrfs_end_transaction_throttle(trans, root);
5f39d397 2286
39279cc3 2287out_unlock:
39279cc3
CM
2288 if (drop_on_err)
2289 iput(inode);
d3c2fdcf 2290 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM
2291 return err;
2292}
2293
3b951516
CM
2294static int merge_extent_mapping(struct extent_map_tree *em_tree,
2295 struct extent_map *existing,
e6dcd2dc
CM
2296 struct extent_map *em,
2297 u64 map_start, u64 map_len)
3b951516
CM
2298{
2299 u64 start_diff;
3b951516 2300
e6dcd2dc
CM
2301 BUG_ON(map_start < em->start || map_start >= extent_map_end(em));
2302 start_diff = map_start - em->start;
2303 em->start = map_start;
2304 em->len = map_len;
2305 if (em->block_start < EXTENT_MAP_LAST_BYTE)
2306 em->block_start += start_diff;
2307 return add_extent_mapping(em_tree, em);
3b951516
CM
2308}
2309
a52d9a80 2310struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
70dec807 2311 size_t pg_offset, u64 start, u64 len,
a52d9a80
CM
2312 int create)
2313{
2314 int ret;
2315 int err = 0;
db94535d 2316 u64 bytenr;
a52d9a80
CM
2317 u64 extent_start = 0;
2318 u64 extent_end = 0;
2319 u64 objectid = inode->i_ino;
2320 u32 found_type;
a52d9a80
CM
2321 struct btrfs_path *path;
2322 struct btrfs_root *root = BTRFS_I(inode)->root;
2323 struct btrfs_file_extent_item *item;
5f39d397
CM
2324 struct extent_buffer *leaf;
2325 struct btrfs_key found_key;
a52d9a80
CM
2326 struct extent_map *em = NULL;
2327 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
d1310b2e 2328 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
a52d9a80
CM
2329 struct btrfs_trans_handle *trans = NULL;
2330
2331 path = btrfs_alloc_path();
2332 BUG_ON(!path);
a52d9a80
CM
2333
2334again:
d1310b2e
CM
2335 spin_lock(&em_tree->lock);
2336 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM
2337 if (em)
2338 em->bdev = root->fs_info->fs_devices->latest_bdev;
d1310b2e
CM
2339 spin_unlock(&em_tree->lock);
2340
a52d9a80 2341 if (em) {
e1c4b745
CM
2342 if (em->start > start || em->start + em->len <= start)
2343 free_extent_map(em);
2344 else if (em->block_start == EXTENT_MAP_INLINE && page)
70dec807
CM
2345 free_extent_map(em);
2346 else
2347 goto out;
a52d9a80 2348 }
d1310b2e 2349 em = alloc_extent_map(GFP_NOFS);
a52d9a80 2350 if (!em) {
d1310b2e
CM
2351 err = -ENOMEM;
2352 goto out;
a52d9a80 2353 }
e6dcd2dc 2354 em->bdev = root->fs_info->fs_devices->latest_bdev;
d1310b2e
CM
2355 em->start = EXTENT_MAP_HOLE;
2356 em->len = (u64)-1;
179e29e4
CM
2357 ret = btrfs_lookup_file_extent(trans, root, path,
2358 objectid, start, trans != NULL);
a52d9a80
CM
2359 if (ret < 0) {
2360 err = ret;
2361 goto out;
2362 }
2363
2364 if (ret != 0) {
2365 if (path->slots[0] == 0)
2366 goto not_found;
2367 path->slots[0]--;
2368 }
2369
5f39d397
CM
2370 leaf = path->nodes[0];
2371 item = btrfs_item_ptr(leaf, path->slots[0],
a52d9a80 2372 struct btrfs_file_extent_item);
a52d9a80 2373 /* are we inside the extent that was found? */
5f39d397
CM
2374 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2375 found_type = btrfs_key_type(&found_key);
2376 if (found_key.objectid != objectid ||
a52d9a80
CM
2377 found_type != BTRFS_EXTENT_DATA_KEY) {
2378 goto not_found;
2379 }
2380
5f39d397
CM
2381 found_type = btrfs_file_extent_type(leaf, item);
2382 extent_start = found_key.offset;
a52d9a80
CM
2383 if (found_type == BTRFS_FILE_EXTENT_REG) {
2384 extent_end = extent_start +
db94535d 2385 btrfs_file_extent_num_bytes(leaf, item);
a52d9a80 2386 err = 0;
b888db2b 2387 if (start < extent_start || start >= extent_end) {
a52d9a80
CM
2388 em->start = start;
2389 if (start < extent_start) {
d1310b2e 2390 if (start + len <= extent_start)
b888db2b 2391 goto not_found;
d1310b2e 2392 em->len = extent_end - extent_start;
a52d9a80 2393 } else {
d1310b2e 2394 em->len = len;
a52d9a80
CM
2395 }
2396 goto not_found_em;
2397 }
db94535d
CM
2398 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
2399 if (bytenr == 0) {
a52d9a80 2400 em->start = extent_start;
d1310b2e 2401 em->len = extent_end - extent_start;
5f39d397 2402 em->block_start = EXTENT_MAP_HOLE;
a52d9a80
CM
2403 goto insert;
2404 }
db94535d
CM
2405 bytenr += btrfs_file_extent_offset(leaf, item);
2406 em->block_start = bytenr;
a52d9a80 2407 em->start = extent_start;
d1310b2e 2408 em->len = extent_end - extent_start;
a52d9a80
CM
2409 goto insert;
2410 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
70dec807 2411 u64 page_start;
5f39d397 2412 unsigned long ptr;
a52d9a80 2413 char *map;
3326d1b0
CM
2414 size_t size;
2415 size_t extent_offset;
2416 size_t copy_size;
a52d9a80 2417
5f39d397
CM
2418 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
2419 path->slots[0]));
d1310b2e
CM
2420 extent_end = (extent_start + size + root->sectorsize - 1) &
2421 ~((u64)root->sectorsize - 1);
b888db2b 2422 if (start < extent_start || start >= extent_end) {
a52d9a80
CM
2423 em->start = start;
2424 if (start < extent_start) {
d1310b2e 2425 if (start + len <= extent_start)
b888db2b 2426 goto not_found;
d1310b2e 2427 em->len = extent_end - extent_start;
a52d9a80 2428 } else {
d1310b2e 2429 em->len = len;
a52d9a80
CM
2430 }
2431 goto not_found_em;
2432 }
689f9346 2433 em->block_start = EXTENT_MAP_INLINE;
689f9346
Y
2434
2435 if (!page) {
2436 em->start = extent_start;
d1310b2e 2437 em->len = size;
689f9346
Y
2438 goto out;
2439 }
5f39d397 2440
70dec807
CM
2441 page_start = page_offset(page) + pg_offset;
2442 extent_offset = page_start - extent_start;
2443 copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
3326d1b0 2444 size - extent_offset);
3326d1b0 2445 em->start = extent_start + extent_offset;
70dec807
CM
2446 em->len = (copy_size + root->sectorsize - 1) &
2447 ~((u64)root->sectorsize - 1);
689f9346
Y
2448 map = kmap(page);
2449 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
179e29e4 2450 if (create == 0 && !PageUptodate(page)) {
70dec807 2451 read_extent_buffer(leaf, map + pg_offset, ptr,
179e29e4
CM
2452 copy_size);
2453 flush_dcache_page(page);
2454 } else if (create && PageUptodate(page)) {
2455 if (!trans) {
2456 kunmap(page);
2457 free_extent_map(em);
2458 em = NULL;
2459 btrfs_release_path(root, path);
f9295749 2460 trans = btrfs_join_transaction(root, 1);
179e29e4
CM
2461 goto again;
2462 }
70dec807 2463 write_extent_buffer(leaf, map + pg_offset, ptr,
179e29e4
CM
2464 copy_size);
2465 btrfs_mark_buffer_dirty(leaf);
a52d9a80 2466 }
a52d9a80 2467 kunmap(page);
d1310b2e
CM
2468 set_extent_uptodate(io_tree, em->start,
2469 extent_map_end(em) - 1, GFP_NOFS);
a52d9a80
CM
2470 goto insert;
2471 } else {
2472 printk("unkknown found_type %d\n", found_type);
2473 WARN_ON(1);
2474 }
2475not_found:
2476 em->start = start;
d1310b2e 2477 em->len = len;
a52d9a80 2478not_found_em:
5f39d397 2479 em->block_start = EXTENT_MAP_HOLE;
a52d9a80
CM
2480insert:
2481 btrfs_release_path(root, path);
d1310b2e
CM
2482 if (em->start > start || extent_map_end(em) <= start) {
2483 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len);
a52d9a80
CM
2484 err = -EIO;
2485 goto out;
2486 }
d1310b2e
CM
2487
2488 err = 0;
2489 spin_lock(&em_tree->lock);
a52d9a80 2490 ret = add_extent_mapping(em_tree, em);
3b951516
CM
2491 /* it is possible that someone inserted the extent into the tree
2492 * while we had the lock dropped. It is also possible that
2493 * an overlapping map exists in the tree
2494 */
a52d9a80 2495 if (ret == -EEXIST) {
3b951516 2496 struct extent_map *existing;
e6dcd2dc
CM
2497
2498 ret = 0;
2499
3b951516 2500 existing = lookup_extent_mapping(em_tree, start, len);
e1c4b745
CM
2501 if (existing && (existing->start > start ||
2502 existing->start + existing->len <= start)) {
2503 free_extent_map(existing);
2504 existing = NULL;
2505 }
3b951516
CM
2506 if (!existing) {
2507 existing = lookup_extent_mapping(em_tree, em->start,
2508 em->len);
2509 if (existing) {
2510 err = merge_extent_mapping(em_tree, existing,
e6dcd2dc
CM
2511 em, start,
2512 root->sectorsize);
3b951516
CM
2513 free_extent_map(existing);
2514 if (err) {
2515 free_extent_map(em);
2516 em = NULL;
2517 }
2518 } else {
2519 err = -EIO;
2520 printk("failing to insert %Lu %Lu\n",
2521 start, len);
2522 free_extent_map(em);
2523 em = NULL;
2524 }
2525 } else {
2526 free_extent_map(em);
2527 em = existing;
e6dcd2dc 2528 err = 0;
a52d9a80 2529 }
a52d9a80 2530 }
d1310b2e 2531 spin_unlock(&em_tree->lock);
a52d9a80
CM
2532out:
2533 btrfs_free_path(path);
2534 if (trans) {
2535 ret = btrfs_end_transaction(trans, root);
e6dcd2dc 2536 if (!err) {
a52d9a80 2537 err = ret;
e6dcd2dc 2538 }
a52d9a80 2539 }
a52d9a80
CM
2540 if (err) {
2541 free_extent_map(em);
2542 WARN_ON(1);
2543 return ERR_PTR(err);
2544 }
2545 return em;
2546}
2547
e1c4b745 2548#if 0 /* waiting for O_DIRECT reads */
16432985
CM
2549static int btrfs_get_block(struct inode *inode, sector_t iblock,
2550 struct buffer_head *bh_result, int create)
2551{
2552 struct extent_map *em;
2553 u64 start = (u64)iblock << inode->i_blkbits;
2554 struct btrfs_multi_bio *multi = NULL;
2555 struct btrfs_root *root = BTRFS_I(inode)->root;
2556 u64 len;
2557 u64 logical;
2558 u64 map_length;
2559 int ret = 0;
2560
2561 em = btrfs_get_extent(inode, NULL, 0, start, bh_result->b_size, 0);
2562
2563 if (!em || IS_ERR(em))
2564 goto out;
2565
e1c4b745 2566 if (em->start > start || em->start + em->len <= start) {
16432985 2567 goto out;
e1c4b745 2568 }
16432985
CM
2569
2570 if (em->block_start == EXTENT_MAP_INLINE) {
2571 ret = -EINVAL;
2572 goto out;
2573 }
2574
e1c4b745
CM
2575 len = em->start + em->len - start;
2576 len = min_t(u64, len, INT_LIMIT(typeof(bh_result->b_size)));
2577
16432985
CM
2578 if (em->block_start == EXTENT_MAP_HOLE ||
2579 em->block_start == EXTENT_MAP_DELALLOC) {
e1c4b745 2580 bh_result->b_size = len;
16432985
CM
2581 goto out;
2582 }
2583
16432985
CM
2584 logical = start - em->start;
2585 logical = em->block_start + logical;
2586
2587 map_length = len;
2588 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
2589 logical, &map_length, &multi, 0);
2590 BUG_ON(ret);
2591 bh_result->b_blocknr = multi->stripes[0].physical >> inode->i_blkbits;
2592 bh_result->b_size = min(map_length, len);
e1c4b745 2593
16432985
CM
2594 bh_result->b_bdev = multi->stripes[0].dev->bdev;
2595 set_buffer_mapped(bh_result);
2596 kfree(multi);
2597out:
2598 free_extent_map(em);
2599 return ret;
2600}
e1c4b745 2601#endif
16432985
CM
2602
2603static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
2604 const struct iovec *iov, loff_t offset,
2605 unsigned long nr_segs)
2606{
e1c4b745
CM
2607 return -EINVAL;
2608#if 0
16432985
CM
2609 struct file *file = iocb->ki_filp;
2610 struct inode *inode = file->f_mapping->host;
2611
2612 if (rw == WRITE)
2613 return -EINVAL;
2614
2615 return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
2616 offset, nr_segs, btrfs_get_block, NULL);
e1c4b745 2617#endif
16432985
CM
2618}
2619
d396c6f5 2620static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
39279cc3 2621{
d396c6f5 2622 return extent_bmap(mapping, iblock, btrfs_get_extent);
39279cc3
CM
2623}
2624
a52d9a80 2625int btrfs_readpage(struct file *file, struct page *page)
9ebefb18 2626{
d1310b2e
CM
2627 struct extent_io_tree *tree;
2628 tree = &BTRFS_I(page->mapping->host)->io_tree;
a52d9a80 2629 return extent_read_full_page(tree, page, btrfs_get_extent);
9ebefb18 2630}
1832a6d5 2631
a52d9a80 2632static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
39279cc3 2633{
d1310b2e 2634 struct extent_io_tree *tree;
b888db2b
CM
2635
2636
2637 if (current->flags & PF_MEMALLOC) {
2638 redirty_page_for_writepage(wbc, page);
2639 unlock_page(page);
2640 return 0;
2641 }
d1310b2e 2642 tree = &BTRFS_I(page->mapping->host)->io_tree;
a52d9a80 2643 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
9ebefb18
CM
2644}
2645
b293f02e
CM
2646static int btrfs_writepages(struct address_space *mapping,
2647 struct writeback_control *wbc)
2648{
d1310b2e
CM
2649 struct extent_io_tree *tree;
2650 tree = &BTRFS_I(mapping->host)->io_tree;
b293f02e
CM
2651 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
2652}
2653
3ab2fb5a
CM
2654static int
2655btrfs_readpages(struct file *file, struct address_space *mapping,
2656 struct list_head *pages, unsigned nr_pages)
2657{
d1310b2e
CM
2658 struct extent_io_tree *tree;
2659 tree = &BTRFS_I(mapping->host)->io_tree;
3ab2fb5a
CM
2660 return extent_readpages(tree, mapping, pages, nr_pages,
2661 btrfs_get_extent);
2662}
e6dcd2dc 2663static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
9ebefb18 2664{
d1310b2e
CM
2665 struct extent_io_tree *tree;
2666 struct extent_map_tree *map;
a52d9a80 2667 int ret;
8c2383c3 2668
d1310b2e
CM
2669 tree = &BTRFS_I(page->mapping->host)->io_tree;
2670 map = &BTRFS_I(page->mapping->host)->extent_tree;
70dec807 2671 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
a52d9a80
CM
2672 if (ret == 1) {
2673 ClearPagePrivate(page);
2674 set_page_private(page, 0);
2675 page_cache_release(page);
39279cc3 2676 }
a52d9a80 2677 return ret;
39279cc3
CM
2678}
2679
e6dcd2dc
CM
2680static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
2681{
e6dcd2dc
CM
2682 return __btrfs_releasepage(page, gfp_flags);
2683}
2684
a52d9a80 2685static void btrfs_invalidatepage(struct page *page, unsigned long offset)
39279cc3 2686{
d1310b2e 2687 struct extent_io_tree *tree;
e6dcd2dc
CM
2688 struct btrfs_ordered_extent *ordered;
2689 u64 page_start = page_offset(page);
2690 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
39279cc3 2691
e6dcd2dc 2692 wait_on_page_writeback(page);
d1310b2e 2693 tree = &BTRFS_I(page->mapping->host)->io_tree;
e6dcd2dc
CM
2694 if (offset) {
2695 btrfs_releasepage(page, GFP_NOFS);
2696 return;
2697 }
2698
2699 lock_extent(tree, page_start, page_end, GFP_NOFS);
2700 ordered = btrfs_lookup_ordered_extent(page->mapping->host,
2701 page_offset(page));
2702 if (ordered) {
eb84ae03
CM
2703 /*
2704 * IO on this page will never be started, so we need
2705 * to account for any ordered extents now
2706 */
e6dcd2dc
CM
2707 clear_extent_bit(tree, page_start, page_end,
2708 EXTENT_DIRTY | EXTENT_DELALLOC |
2709 EXTENT_LOCKED, 1, 0, GFP_NOFS);
211f90e6
CM
2710 btrfs_finish_ordered_io(page->mapping->host,
2711 page_start, page_end);
e6dcd2dc
CM
2712 btrfs_put_ordered_extent(ordered);
2713 lock_extent(tree, page_start, page_end, GFP_NOFS);
2714 }
2715 clear_extent_bit(tree, page_start, page_end,
2716 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
2717 EXTENT_ORDERED,
2718 1, 1, GFP_NOFS);
2719 __btrfs_releasepage(page, GFP_NOFS);
2720
4a096752 2721 ClearPageChecked(page);
9ad6b7bc 2722 if (PagePrivate(page)) {
9ad6b7bc
CM
2723 ClearPagePrivate(page);
2724 set_page_private(page, 0);
2725 page_cache_release(page);
2726 }
39279cc3
CM
2727}
2728
9ebefb18
CM
2729/*
2730 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2731 * called from a page fault handler when a page is first dirtied. Hence we must
2732 * be careful to check for EOF conditions here. We set the page up correctly
2733 * for a written page which means we get ENOSPC checking when writing into
2734 * holes and correct delalloc and unwritten extent mapping on filesystems that
2735 * support these features.
2736 *
2737 * We are not allowed to take the i_mutex here so we have to play games to
2738 * protect against truncate races as the page could now be beyond EOF. Because
2739 * vmtruncate() writes the inode size before removing pages, once we have the
2740 * page lock we can determine safely if the page is beyond EOF. If it is not
2741 * beyond EOF, then the page is guaranteed safe against truncation until we
2742 * unlock the page.
2743 */
2744int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
2745{
6da6abae 2746 struct inode *inode = fdentry(vma->vm_file)->d_inode;
1832a6d5 2747 struct btrfs_root *root = BTRFS_I(inode)->root;
e6dcd2dc
CM
2748 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2749 struct btrfs_ordered_extent *ordered;
2750 char *kaddr;
2751 unsigned long zero_start;
9ebefb18 2752 loff_t size;
1832a6d5 2753 int ret;
a52d9a80 2754 u64 page_start;
e6dcd2dc 2755 u64 page_end;
9ebefb18 2756
1832a6d5 2757 ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
1832a6d5
CM
2758 if (ret)
2759 goto out;
2760
2761 ret = -EINVAL;
e6dcd2dc 2762again:
9ebefb18 2763 lock_page(page);
9ebefb18 2764 size = i_size_read(inode);
e6dcd2dc
CM
2765 page_start = page_offset(page);
2766 page_end = page_start + PAGE_CACHE_SIZE - 1;
a52d9a80 2767
9ebefb18 2768 if ((page->mapping != inode->i_mapping) ||
e6dcd2dc 2769 (page_start >= size)) {
9ebefb18
CM
2770 /* page got truncated out from underneath us */
2771 goto out_unlock;
2772 }
e6dcd2dc
CM
2773 wait_on_page_writeback(page);
2774
2775 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
2776 set_page_extent_mapped(page);
2777
eb84ae03
CM
2778 /*
2779 * we can't set the delalloc bits if there are pending ordered
2780 * extents. Drop our locks and wait for them to finish
2781 */
e6dcd2dc
CM
2782 ordered = btrfs_lookup_ordered_extent(inode, page_start);
2783 if (ordered) {
2784 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
2785 unlock_page(page);
eb84ae03 2786 btrfs_start_ordered_extent(inode, ordered, 1);
e6dcd2dc
CM
2787 btrfs_put_ordered_extent(ordered);
2788 goto again;
2789 }
2790
2791 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
2792 page_end, GFP_NOFS);
2793 ret = 0;
9ebefb18
CM
2794
2795 /* page is wholly or partially inside EOF */
a52d9a80 2796 if (page_start + PAGE_CACHE_SIZE > size)
e6dcd2dc 2797 zero_start = size & ~PAGE_CACHE_MASK;
9ebefb18 2798 else
e6dcd2dc 2799 zero_start = PAGE_CACHE_SIZE;
9ebefb18 2800
e6dcd2dc
CM
2801 if (zero_start != PAGE_CACHE_SIZE) {
2802 kaddr = kmap(page);
2803 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
2804 flush_dcache_page(page);
2805 kunmap(page);
2806 }
247e743c 2807 ClearPageChecked(page);
e6dcd2dc
CM
2808 set_page_dirty(page);
2809 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
9ebefb18
CM
2810
2811out_unlock:
2812 unlock_page(page);
1832a6d5 2813out:
9ebefb18
CM
2814 return ret;
2815}
2816
39279cc3
CM
2817static void btrfs_truncate(struct inode *inode)
2818{
2819 struct btrfs_root *root = BTRFS_I(inode)->root;
2820 int ret;
2821 struct btrfs_trans_handle *trans;
d3c2fdcf 2822 unsigned long nr;
dbe674a9 2823 u64 mask = root->sectorsize - 1;
39279cc3
CM
2824
2825 if (!S_ISREG(inode->i_mode))
2826 return;
2827 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
2828 return;
2829
2830 btrfs_truncate_page(inode->i_mapping, inode->i_size);
4a096752 2831 btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
39279cc3 2832
39279cc3
CM
2833 trans = btrfs_start_transaction(root, 1);
2834 btrfs_set_trans_block_group(trans, inode);
dbe674a9 2835 btrfs_i_size_write(inode, inode->i_size);
39279cc3
CM
2836
2837 /* FIXME, add redo link to tree so we don't leak on crash */
85e21bac
CM
2838 ret = btrfs_truncate_in_trans(trans, root, inode,
2839 BTRFS_EXTENT_DATA_KEY);
39279cc3 2840 btrfs_update_inode(trans, root, inode);
d3c2fdcf 2841 nr = trans->blocks_used;
5f39d397 2842
89ce8a63 2843 ret = btrfs_end_transaction_throttle(trans, root);
39279cc3 2844 BUG_ON(ret);
d3c2fdcf 2845 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM
2846}
2847
3b96362c
SW
2848/*
2849 * Invalidate a single dcache entry at the root of the filesystem.
2850 * Needed after creation of snapshot or subvolume.
2851 */
2852void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name,
2853 int namelen)
2854{
2855 struct dentry *alias, *entry;
2856 struct qstr qstr;
2857
2858 alias = d_find_alias(root->fs_info->sb->s_root->d_inode);
2859 if (alias) {
2860 qstr.name = name;
2861 qstr.len = namelen;
2862 /* change me if btrfs ever gets a d_hash operation */
2863 qstr.hash = full_name_hash(qstr.name, qstr.len);
2864 entry = d_lookup(alias, &qstr);
2865 dput(alias);
2866 if (entry) {
2867 d_invalidate(entry);
2868 dput(entry);
2869 }
2870 }
2871}
2872
f46b5a66
CH
2873int btrfs_create_subvol_root(struct btrfs_root *new_root,
2874 struct btrfs_trans_handle *trans, u64 new_dirid,
2875 struct btrfs_block_group_cache *block_group)
39279cc3 2876{
39279cc3 2877 struct inode *inode;
39279cc3 2878 int ret;
39279cc3 2879
9c58309d 2880 inode = btrfs_new_inode(trans, new_root, "..", 2, new_dirid,
f46b5a66 2881 new_dirid, block_group, S_IFDIR | 0700);
54aa1f4d 2882 if (IS_ERR(inode))
f46b5a66 2883 return PTR_ERR(inode);
39279cc3
CM
2884 inode->i_op = &btrfs_dir_inode_operations;
2885 inode->i_fop = &btrfs_dir_file_operations;
34088780 2886 new_root->inode = inode;
39279cc3 2887
3954401f
CM
2888 ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid,
2889 new_dirid);
39279cc3 2890 inode->i_nlink = 1;
dbe674a9 2891 btrfs_i_size_write(inode, 0);
3b96362c 2892
f46b5a66 2893 return btrfs_update_inode(trans, new_root, inode);
39279cc3
CM
2894}
2895
edbd8d4e 2896unsigned long btrfs_force_ra(struct address_space *mapping,
86479a04
CM
2897 struct file_ra_state *ra, struct file *file,
2898 pgoff_t offset, pgoff_t last_index)
2899{
8e7bf94f 2900 pgoff_t req_size = last_index - offset + 1;
86479a04
CM
2901
2902#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
86479a04
CM
2903 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2904 return offset;
2905#else
86479a04
CM
2906 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2907 return offset + req_size;
2908#endif
2909}
2910
39279cc3
CM
2911struct inode *btrfs_alloc_inode(struct super_block *sb)
2912{
2913 struct btrfs_inode *ei;
2914
2915 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2916 if (!ei)
2917 return NULL;
15ee9bc7 2918 ei->last_trans = 0;
e6dcd2dc 2919 btrfs_ordered_inode_tree_init(&ei->ordered_tree);
39279cc3
CM
2920 return &ei->vfs_inode;
2921}
2922
2923void btrfs_destroy_inode(struct inode *inode)
2924{
e6dcd2dc 2925 struct btrfs_ordered_extent *ordered;
39279cc3
CM
2926 WARN_ON(!list_empty(&inode->i_dentry));
2927 WARN_ON(inode->i_data.nrpages);
2928
e6dcd2dc
CM
2929 while(1) {
2930 ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
2931 if (!ordered)
2932 break;
2933 else {
2934 printk("found ordered extent %Lu %Lu\n",
2935 ordered->file_offset, ordered->len);
2936 btrfs_remove_ordered_extent(inode, ordered);
2937 btrfs_put_ordered_extent(ordered);
2938 btrfs_put_ordered_extent(ordered);
2939 }
2940 }
8c416c9e 2941 btrfs_drop_extent_cache(inode, 0, (u64)-1);
39279cc3
CM
2942 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2943}
2944
44ec0b71
CM
2945#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2946static void init_once(struct kmem_cache * cachep, void *foo)
2947#else
39279cc3
CM
2948static void init_once(void * foo, struct kmem_cache * cachep,
2949 unsigned long flags)
44ec0b71 2950#endif
39279cc3
CM
2951{
2952 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2953
2954 inode_init_once(&ei->vfs_inode);
2955}
2956
2957void btrfs_destroy_cachep(void)
2958{
2959 if (btrfs_inode_cachep)
2960 kmem_cache_destroy(btrfs_inode_cachep);
2961 if (btrfs_trans_handle_cachep)
2962 kmem_cache_destroy(btrfs_trans_handle_cachep);
2963 if (btrfs_transaction_cachep)
2964 kmem_cache_destroy(btrfs_transaction_cachep);
2965 if (btrfs_bit_radix_cachep)
2966 kmem_cache_destroy(btrfs_bit_radix_cachep);
2967 if (btrfs_path_cachep)
2968 kmem_cache_destroy(btrfs_path_cachep);
2969}
2970
86479a04 2971struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
92fee66d 2972 unsigned long extra_flags,
44ec0b71
CM
2973#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2974 void (*ctor)(struct kmem_cache *, void *)
2975#else
92fee66d 2976 void (*ctor)(void *, struct kmem_cache *,
44ec0b71
CM
2977 unsigned long)
2978#endif
2979 )
92fee66d
CM
2980{
2981 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2982 SLAB_MEM_SPREAD | extra_flags), ctor
2983#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2984 ,NULL
2985#endif
2986 );
2987}
2988
39279cc3
CM
2989int btrfs_init_cachep(void)
2990{
86479a04 2991 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
92fee66d
CM
2992 sizeof(struct btrfs_inode),
2993 0, init_once);
39279cc3
CM
2994 if (!btrfs_inode_cachep)
2995 goto fail;
86479a04
CM
2996 btrfs_trans_handle_cachep =
2997 btrfs_cache_create("btrfs_trans_handle_cache",
2998 sizeof(struct btrfs_trans_handle),
2999 0, NULL);
39279cc3
CM
3000 if (!btrfs_trans_handle_cachep)
3001 goto fail;
86479a04 3002 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
39279cc3 3003 sizeof(struct btrfs_transaction),
92fee66d 3004 0, NULL);
39279cc3
CM
3005 if (!btrfs_transaction_cachep)
3006 goto fail;
86479a04 3007 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
23223584 3008 sizeof(struct btrfs_path),
92fee66d 3009 0, NULL);
39279cc3
CM
3010 if (!btrfs_path_cachep)
3011 goto fail;
86479a04 3012 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
92fee66d 3013 SLAB_DESTROY_BY_RCU, NULL);
39279cc3
CM
3014 if (!btrfs_bit_radix_cachep)
3015 goto fail;
3016 return 0;
3017fail:
3018 btrfs_destroy_cachep();
3019 return -ENOMEM;
3020}
3021
3022static int btrfs_getattr(struct vfsmount *mnt,
3023 struct dentry *dentry, struct kstat *stat)
3024{
3025 struct inode *inode = dentry->d_inode;
3026 generic_fillattr(inode, stat);
d6667462 3027 stat->blksize = PAGE_CACHE_SIZE;
9069218d 3028 stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9);
39279cc3
CM
3029 return 0;
3030}
3031
3032static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
3033 struct inode * new_dir,struct dentry *new_dentry)
3034{
3035 struct btrfs_trans_handle *trans;
3036 struct btrfs_root *root = BTRFS_I(old_dir)->root;
3037 struct inode *new_inode = new_dentry->d_inode;
3038 struct inode *old_inode = old_dentry->d_inode;
3039 struct timespec ctime = CURRENT_TIME;
39279cc3
CM
3040 int ret;
3041
3042 if (S_ISDIR(old_inode->i_mode) && new_inode &&
3043 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
3044 return -ENOTEMPTY;
3045 }
5f39d397 3046
1832a6d5
CM
3047 ret = btrfs_check_free_space(root, 1, 0);
3048 if (ret)
3049 goto out_unlock;
3050
39279cc3 3051 trans = btrfs_start_transaction(root, 1);
5f39d397 3052
39279cc3 3053 btrfs_set_trans_block_group(trans, new_dir);
39279cc3
CM
3054
3055 old_dentry->d_inode->i_nlink++;
3056 old_dir->i_ctime = old_dir->i_mtime = ctime;
3057 new_dir->i_ctime = new_dir->i_mtime = ctime;
3058 old_inode->i_ctime = ctime;
5f39d397 3059
39279cc3
CM
3060 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
3061 if (ret)
3062 goto out_fail;
3063
3064 if (new_inode) {
3065 new_inode->i_ctime = CURRENT_TIME;
3066 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
3067 if (ret)
3068 goto out_fail;
39279cc3 3069 }
9c58309d 3070 ret = btrfs_add_link(trans, new_dentry, old_inode, 1);
39279cc3
CM
3071 if (ret)
3072 goto out_fail;
3073
3074out_fail:
39279cc3 3075 btrfs_end_transaction(trans, root);
1832a6d5 3076out_unlock:
39279cc3
CM
3077 return ret;
3078}
3079
3080static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
3081 const char *symname)
3082{
3083 struct btrfs_trans_handle *trans;
3084 struct btrfs_root *root = BTRFS_I(dir)->root;
3085 struct btrfs_path *path;
3086 struct btrfs_key key;
1832a6d5 3087 struct inode *inode = NULL;
39279cc3
CM
3088 int err;
3089 int drop_inode = 0;
3090 u64 objectid;
3091 int name_len;
3092 int datasize;
5f39d397 3093 unsigned long ptr;
39279cc3 3094 struct btrfs_file_extent_item *ei;
5f39d397 3095 struct extent_buffer *leaf;
1832a6d5 3096 unsigned long nr = 0;
39279cc3
CM
3097
3098 name_len = strlen(symname) + 1;
3099 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
3100 return -ENAMETOOLONG;
1832a6d5 3101
1832a6d5
CM
3102 err = btrfs_check_free_space(root, 1, 0);
3103 if (err)
3104 goto out_fail;
3105
39279cc3
CM
3106 trans = btrfs_start_transaction(root, 1);
3107 btrfs_set_trans_block_group(trans, dir);
3108
3109 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
3110 if (err) {
3111 err = -ENOSPC;
3112 goto out_unlock;
3113 }
3114
9c58309d
CM
3115 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
3116 dentry->d_name.len,
3117 dentry->d_parent->d_inode->i_ino, objectid,
39279cc3
CM
3118 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
3119 err = PTR_ERR(inode);
3120 if (IS_ERR(inode))
3121 goto out_unlock;
3122
3123 btrfs_set_trans_block_group(trans, inode);
9c58309d 3124 err = btrfs_add_nondir(trans, dentry, inode, 0);
39279cc3
CM
3125 if (err)
3126 drop_inode = 1;
3127 else {
3128 inode->i_mapping->a_ops = &btrfs_aops;
04160088 3129 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
39279cc3
CM
3130 inode->i_fop = &btrfs_file_operations;
3131 inode->i_op = &btrfs_file_inode_operations;
d1310b2e
CM
3132 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
3133 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
a52d9a80 3134 inode->i_mapping, GFP_NOFS);
7e38326f
CM
3135 extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
3136 inode->i_mapping, GFP_NOFS);
1b1e2135 3137 mutex_init(&BTRFS_I(inode)->csum_mutex);
ee6e6504 3138 mutex_init(&BTRFS_I(inode)->extent_mutex);
9069218d 3139 BTRFS_I(inode)->delalloc_bytes = 0;
dbe674a9 3140 BTRFS_I(inode)->disk_i_size = 0;
d1310b2e 3141 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
ba1da2f4 3142 btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
39279cc3
CM
3143 }
3144 dir->i_sb->s_dirt = 1;
3145 btrfs_update_inode_block_group(trans, inode);
3146 btrfs_update_inode_block_group(trans, dir);
3147 if (drop_inode)
3148 goto out_unlock;
3149
3150 path = btrfs_alloc_path();
3151 BUG_ON(!path);
3152 key.objectid = inode->i_ino;
3153 key.offset = 0;
39279cc3
CM
3154 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
3155 datasize = btrfs_file_extent_calc_inline_size(name_len);
3156 err = btrfs_insert_empty_item(trans, root, path, &key,
3157 datasize);
54aa1f4d
CM
3158 if (err) {
3159 drop_inode = 1;
3160 goto out_unlock;
3161 }
5f39d397
CM
3162 leaf = path->nodes[0];
3163 ei = btrfs_item_ptr(leaf, path->slots[0],
3164 struct btrfs_file_extent_item);
3165 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
3166 btrfs_set_file_extent_type(leaf, ei,
39279cc3
CM
3167 BTRFS_FILE_EXTENT_INLINE);
3168 ptr = btrfs_file_extent_inline_start(ei);
5f39d397
CM
3169 write_extent_buffer(leaf, symname, ptr, name_len);
3170 btrfs_mark_buffer_dirty(leaf);
39279cc3 3171 btrfs_free_path(path);
5f39d397 3172
39279cc3
CM
3173 inode->i_op = &btrfs_symlink_inode_operations;
3174 inode->i_mapping->a_ops = &btrfs_symlink_aops;
04160088 3175 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
dbe674a9 3176 btrfs_i_size_write(inode, name_len - 1);
54aa1f4d
CM
3177 err = btrfs_update_inode(trans, root, inode);
3178 if (err)
3179 drop_inode = 1;
39279cc3
CM
3180
3181out_unlock:
d3c2fdcf 3182 nr = trans->blocks_used;
89ce8a63 3183 btrfs_end_transaction_throttle(trans, root);
1832a6d5 3184out_fail:
39279cc3
CM
3185 if (drop_inode) {
3186 inode_dec_link_count(inode);
3187 iput(inode);
3188 }
d3c2fdcf 3189 btrfs_btree_balance_dirty(root, nr);
39279cc3
CM
3190 return err;
3191}
16432985 3192
e6dcd2dc
CM
3193static int btrfs_set_page_dirty(struct page *page)
3194{
e6dcd2dc
CM
3195 return __set_page_dirty_nobuffers(page);
3196}
3197
fdebe2bd
Y
3198static int btrfs_permission(struct inode *inode, int mask,
3199 struct nameidata *nd)
3200{
3201 if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
3202 return -EACCES;
3203 return generic_permission(inode, mask, NULL);
3204}
39279cc3
CM
3205
3206static struct inode_operations btrfs_dir_inode_operations = {
3207 .lookup = btrfs_lookup,
3208 .create = btrfs_create,
3209 .unlink = btrfs_unlink,
3210 .link = btrfs_link,
3211 .mkdir = btrfs_mkdir,
3212 .rmdir = btrfs_rmdir,
3213 .rename = btrfs_rename,
3214 .symlink = btrfs_symlink,
3215 .setattr = btrfs_setattr,
618e21d5 3216 .mknod = btrfs_mknod,
5103e947
JB
3217 .setxattr = generic_setxattr,
3218 .getxattr = generic_getxattr,
3219 .listxattr = btrfs_listxattr,
3220 .removexattr = generic_removexattr,
fdebe2bd 3221 .permission = btrfs_permission,
39279cc3 3222};
39279cc3
CM
3223static struct inode_operations btrfs_dir_ro_inode_operations = {
3224 .lookup = btrfs_lookup,
fdebe2bd 3225 .permission = btrfs_permission,
39279cc3 3226};
39279cc3
CM
3227static struct file_operations btrfs_dir_file_operations = {
3228 .llseek = generic_file_llseek,
3229 .read = generic_read_dir,
3230 .readdir = btrfs_readdir,
34287aa3 3231 .unlocked_ioctl = btrfs_ioctl,
39279cc3 3232#ifdef CONFIG_COMPAT
34287aa3 3233 .compat_ioctl = btrfs_ioctl,
39279cc3 3234#endif
6bf13c0c 3235 .release = btrfs_release_file,
39279cc3
CM
3236};
3237
d1310b2e 3238static struct extent_io_ops btrfs_extent_io_ops = {
07157aac 3239 .fill_delalloc = run_delalloc_range,
065631f6 3240 .submit_bio_hook = btrfs_submit_bio_hook,
239b14b3 3241 .merge_bio_hook = btrfs_merge_bio_hook,
07157aac
CM
3242 .readpage_io_hook = btrfs_readpage_io_hook,
3243 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
e6dcd2dc 3244 .writepage_end_io_hook = btrfs_writepage_end_io_hook,
247e743c 3245 .writepage_start_hook = btrfs_writepage_start_hook,
1259ab75 3246 .readpage_io_failed_hook = btrfs_io_failed_hook,
b0c68f8b
CM
3247 .set_bit_hook = btrfs_set_bit_hook,
3248 .clear_bit_hook = btrfs_clear_bit_hook,
07157aac
CM
3249};
3250
39279cc3
CM
3251static struct address_space_operations btrfs_aops = {
3252 .readpage = btrfs_readpage,
3253 .writepage = btrfs_writepage,
b293f02e 3254 .writepages = btrfs_writepages,
3ab2fb5a 3255 .readpages = btrfs_readpages,
39279cc3 3256 .sync_page = block_sync_page,
39279cc3 3257 .bmap = btrfs_bmap,
16432985 3258 .direct_IO = btrfs_direct_IO,
a52d9a80
CM
3259 .invalidatepage = btrfs_invalidatepage,
3260 .releasepage = btrfs_releasepage,
e6dcd2dc 3261 .set_page_dirty = btrfs_set_page_dirty,
39279cc3
CM
3262};
3263
3264static struct address_space_operations btrfs_symlink_aops = {
3265 .readpage = btrfs_readpage,
3266 .writepage = btrfs_writepage,
2bf5a725
CM
3267 .invalidatepage = btrfs_invalidatepage,
3268 .releasepage = btrfs_releasepage,
39279cc3
CM
3269};
3270
3271static struct inode_operations btrfs_file_inode_operations = {
3272 .truncate = btrfs_truncate,
3273 .getattr = btrfs_getattr,
3274 .setattr = btrfs_setattr,
5103e947
JB
3275 .setxattr = generic_setxattr,
3276 .getxattr = generic_getxattr,
3277 .listxattr = btrfs_listxattr,
3278 .removexattr = generic_removexattr,
fdebe2bd 3279 .permission = btrfs_permission,
39279cc3 3280};
618e21d5
JB
3281static struct inode_operations btrfs_special_inode_operations = {
3282 .getattr = btrfs_getattr,
3283 .setattr = btrfs_setattr,
fdebe2bd 3284 .permission = btrfs_permission,
618e21d5 3285};
39279cc3
CM
3286static struct inode_operations btrfs_symlink_inode_operations = {
3287 .readlink = generic_readlink,
3288 .follow_link = page_follow_link_light,
3289 .put_link = page_put_link,
fdebe2bd 3290 .permission = btrfs_permission,
39279cc3 3291};