]>
Commit | Line | Data |
---|---|---|
6cbd5570 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
8f18cf13 | 19 | #include <linux/kernel.h> |
065631f6 | 20 | #include <linux/bio.h> |
39279cc3 | 21 | #include <linux/buffer_head.h> |
f2eb0a24 | 22 | #include <linux/file.h> |
39279cc3 CM |
23 | #include <linux/fs.h> |
24 | #include <linux/pagemap.h> | |
25 | #include <linux/highmem.h> | |
26 | #include <linux/time.h> | |
27 | #include <linux/init.h> | |
28 | #include <linux/string.h> | |
29 | #include <linux/smp_lock.h> | |
30 | #include <linux/backing-dev.h> | |
31 | #include <linux/mpage.h> | |
32 | #include <linux/swap.h> | |
33 | #include <linux/writeback.h> | |
34 | #include <linux/statfs.h> | |
35 | #include <linux/compat.h> | |
9ebefb18 | 36 | #include <linux/bit_spinlock.h> |
92fee66d | 37 | #include <linux/version.h> |
5103e947 | 38 | #include <linux/xattr.h> |
33268eaf | 39 | #include <linux/posix_acl.h> |
39279cc3 CM |
40 | #include "ctree.h" |
41 | #include "disk-io.h" | |
42 | #include "transaction.h" | |
43 | #include "btrfs_inode.h" | |
44 | #include "ioctl.h" | |
45 | #include "print-tree.h" | |
0b86a832 | 46 | #include "volumes.h" |
e6dcd2dc | 47 | #include "ordered-data.h" |
95819c05 | 48 | #include "xattr.h" |
e02119d5 CM |
49 | #include "compat.h" |
50 | #include "tree-log.h" | |
39279cc3 CM |
51 | |
52 | struct btrfs_iget_args { | |
53 | u64 ino; | |
54 | struct btrfs_root *root; | |
55 | }; | |
56 | ||
57 | static struct inode_operations btrfs_dir_inode_operations; | |
58 | static struct inode_operations btrfs_symlink_inode_operations; | |
59 | static struct inode_operations btrfs_dir_ro_inode_operations; | |
618e21d5 | 60 | static struct inode_operations btrfs_special_inode_operations; |
39279cc3 CM |
61 | static struct inode_operations btrfs_file_inode_operations; |
62 | static struct address_space_operations btrfs_aops; | |
63 | static struct address_space_operations btrfs_symlink_aops; | |
64 | static struct file_operations btrfs_dir_file_operations; | |
d1310b2e | 65 | static struct extent_io_ops btrfs_extent_io_ops; |
39279cc3 CM |
66 | |
67 | static struct kmem_cache *btrfs_inode_cachep; | |
68 | struct kmem_cache *btrfs_trans_handle_cachep; | |
69 | struct kmem_cache *btrfs_transaction_cachep; | |
70 | struct kmem_cache *btrfs_bit_radix_cachep; | |
71 | struct kmem_cache *btrfs_path_cachep; | |
72 | ||
73 | #define S_SHIFT 12 | |
74 | static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = { | |
75 | [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE, | |
76 | [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR, | |
77 | [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV, | |
78 | [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV, | |
79 | [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO, | |
80 | [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK, | |
81 | [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK, | |
82 | }; | |
83 | ||
7b128766 JB |
84 | static void btrfs_truncate(struct inode *inode); |
85 | ||
1832a6d5 CM |
86 | int btrfs_check_free_space(struct btrfs_root *root, u64 num_required, |
87 | int for_del) | |
88 | { | |
a2135011 CM |
89 | u64 total; |
90 | u64 used; | |
1832a6d5 | 91 | u64 thresh; |
bcbfce8a | 92 | unsigned long flags; |
1832a6d5 CM |
93 | int ret = 0; |
94 | ||
a2135011 CM |
95 | spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); |
96 | total = btrfs_super_total_bytes(&root->fs_info->super_copy); | |
97 | used = btrfs_super_bytes_used(&root->fs_info->super_copy); | |
1832a6d5 | 98 | if (for_del) |
f9ef6604 | 99 | thresh = total * 90; |
1832a6d5 | 100 | else |
f9ef6604 CM |
101 | thresh = total * 85; |
102 | ||
103 | do_div(thresh, 100); | |
1832a6d5 | 104 | |
1832a6d5 CM |
105 | if (used + root->fs_info->delalloc_bytes + num_required > thresh) |
106 | ret = -ENOSPC; | |
bcbfce8a | 107 | spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); |
1832a6d5 CM |
108 | return ret; |
109 | } | |
110 | ||
be20aa9d | 111 | static int cow_file_range(struct inode *inode, u64 start, u64 end) |
b888db2b CM |
112 | { |
113 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
114 | struct btrfs_trans_handle *trans; | |
b888db2b | 115 | u64 alloc_hint = 0; |
db94535d | 116 | u64 num_bytes; |
c59f8951 | 117 | u64 cur_alloc_size; |
db94535d | 118 | u64 blocksize = root->sectorsize; |
d1310b2e | 119 | u64 orig_num_bytes; |
be20aa9d | 120 | struct btrfs_key ins; |
e6dcd2dc CM |
121 | struct extent_map *em; |
122 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
123 | int ret = 0; | |
b888db2b | 124 | |
f9295749 | 125 | trans = btrfs_join_transaction(root, 1); |
b888db2b | 126 | BUG_ON(!trans); |
be20aa9d CM |
127 | btrfs_set_trans_block_group(trans, inode); |
128 | ||
db94535d | 129 | num_bytes = (end - start + blocksize) & ~(blocksize - 1); |
be20aa9d | 130 | num_bytes = max(blocksize, num_bytes); |
d1310b2e | 131 | orig_num_bytes = num_bytes; |
db94535d | 132 | |
179e29e4 CM |
133 | if (alloc_hint == EXTENT_MAP_INLINE) |
134 | goto out; | |
135 | ||
3b951516 | 136 | BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy)); |
e5a2217e | 137 | mutex_lock(&BTRFS_I(inode)->extent_mutex); |
e6dcd2dc | 138 | btrfs_drop_extent_cache(inode, start, start + num_bytes - 1); |
e5a2217e | 139 | mutex_unlock(&BTRFS_I(inode)->extent_mutex); |
3b951516 | 140 | |
c59f8951 CM |
141 | while(num_bytes > 0) { |
142 | cur_alloc_size = min(num_bytes, root->fs_info->max_extent); | |
e6dcd2dc CM |
143 | ret = btrfs_reserve_extent(trans, root, cur_alloc_size, |
144 | root->sectorsize, 0, 0, | |
145 | (u64)-1, &ins, 1); | |
c59f8951 CM |
146 | if (ret) { |
147 | WARN_ON(1); | |
148 | goto out; | |
149 | } | |
e6dcd2dc CM |
150 | em = alloc_extent_map(GFP_NOFS); |
151 | em->start = start; | |
152 | em->len = ins.offset; | |
153 | em->block_start = ins.objectid; | |
154 | em->bdev = root->fs_info->fs_devices->latest_bdev; | |
e5a2217e | 155 | mutex_lock(&BTRFS_I(inode)->extent_mutex); |
7f3c74fb | 156 | set_bit(EXTENT_FLAG_PINNED, &em->flags); |
e6dcd2dc CM |
157 | while(1) { |
158 | spin_lock(&em_tree->lock); | |
159 | ret = add_extent_mapping(em_tree, em); | |
160 | spin_unlock(&em_tree->lock); | |
161 | if (ret != -EEXIST) { | |
162 | free_extent_map(em); | |
163 | break; | |
164 | } | |
165 | btrfs_drop_extent_cache(inode, start, | |
166 | start + ins.offset - 1); | |
167 | } | |
e5a2217e | 168 | mutex_unlock(&BTRFS_I(inode)->extent_mutex); |
e6dcd2dc | 169 | |
98d20f67 | 170 | cur_alloc_size = ins.offset; |
e6dcd2dc | 171 | ret = btrfs_add_ordered_extent(inode, start, ins.objectid, |
7ea394f1 | 172 | ins.offset, 0); |
e6dcd2dc | 173 | BUG_ON(ret); |
3b951516 CM |
174 | if (num_bytes < cur_alloc_size) { |
175 | printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes, | |
176 | cur_alloc_size); | |
177 | break; | |
178 | } | |
c59f8951 CM |
179 | num_bytes -= cur_alloc_size; |
180 | alloc_hint = ins.objectid + ins.offset; | |
181 | start += cur_alloc_size; | |
b888db2b | 182 | } |
b888db2b CM |
183 | out: |
184 | btrfs_end_transaction(trans, root); | |
be20aa9d CM |
185 | return ret; |
186 | } | |
187 | ||
188 | static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end) | |
189 | { | |
190 | u64 extent_start; | |
191 | u64 extent_end; | |
192 | u64 bytenr; | |
1832a6d5 | 193 | u64 loops = 0; |
c31f8830 | 194 | u64 total_fs_bytes; |
be20aa9d | 195 | struct btrfs_root *root = BTRFS_I(inode)->root; |
a68d5933 | 196 | struct btrfs_block_group_cache *block_group; |
7ea394f1 | 197 | struct btrfs_trans_handle *trans; |
be20aa9d CM |
198 | struct extent_buffer *leaf; |
199 | int found_type; | |
200 | struct btrfs_path *path; | |
201 | struct btrfs_file_extent_item *item; | |
202 | int ret; | |
7ea394f1 | 203 | int err = 0; |
be20aa9d CM |
204 | struct btrfs_key found_key; |
205 | ||
c31f8830 | 206 | total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); |
be20aa9d CM |
207 | path = btrfs_alloc_path(); |
208 | BUG_ON(!path); | |
7ea394f1 YZ |
209 | trans = btrfs_join_transaction(root, 1); |
210 | BUG_ON(!trans); | |
be20aa9d CM |
211 | again: |
212 | ret = btrfs_lookup_file_extent(NULL, root, path, | |
213 | inode->i_ino, start, 0); | |
214 | if (ret < 0) { | |
7ea394f1 YZ |
215 | err = ret; |
216 | goto out; | |
be20aa9d CM |
217 | } |
218 | ||
be20aa9d CM |
219 | if (ret != 0) { |
220 | if (path->slots[0] == 0) | |
221 | goto not_found; | |
222 | path->slots[0]--; | |
223 | } | |
224 | ||
225 | leaf = path->nodes[0]; | |
226 | item = btrfs_item_ptr(leaf, path->slots[0], | |
227 | struct btrfs_file_extent_item); | |
228 | ||
229 | /* are we inside the extent that was found? */ | |
230 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
231 | found_type = btrfs_key_type(&found_key); | |
232 | if (found_key.objectid != inode->i_ino || | |
bbaf549e | 233 | found_type != BTRFS_EXTENT_DATA_KEY) |
be20aa9d | 234 | goto not_found; |
be20aa9d CM |
235 | |
236 | found_type = btrfs_file_extent_type(leaf, item); | |
237 | extent_start = found_key.offset; | |
238 | if (found_type == BTRFS_FILE_EXTENT_REG) { | |
c31f8830 CM |
239 | u64 extent_num_bytes; |
240 | ||
241 | extent_num_bytes = btrfs_file_extent_num_bytes(leaf, item); | |
242 | extent_end = extent_start + extent_num_bytes; | |
be20aa9d CM |
243 | err = 0; |
244 | ||
1832a6d5 CM |
245 | if (loops && start != extent_start) |
246 | goto not_found; | |
247 | ||
be20aa9d CM |
248 | if (start < extent_start || start >= extent_end) |
249 | goto not_found; | |
250 | ||
be20aa9d CM |
251 | bytenr = btrfs_file_extent_disk_bytenr(leaf, item); |
252 | if (bytenr == 0) | |
253 | goto not_found; | |
254 | ||
7ea394f1 | 255 | if (btrfs_cross_ref_exists(trans, root, &found_key, bytenr)) |
a68d5933 | 256 | goto not_found; |
c31f8830 CM |
257 | /* |
258 | * we may be called by the resizer, make sure we're inside | |
259 | * the limits of the FS | |
260 | */ | |
a68d5933 CM |
261 | block_group = btrfs_lookup_block_group(root->fs_info, |
262 | bytenr); | |
263 | if (!block_group || block_group->ro) | |
c31f8830 CM |
264 | goto not_found; |
265 | ||
7ea394f1 YZ |
266 | bytenr += btrfs_file_extent_offset(leaf, item); |
267 | extent_num_bytes = min(end + 1, extent_end) - start; | |
268 | ret = btrfs_add_ordered_extent(inode, start, bytenr, | |
269 | extent_num_bytes, 1); | |
270 | if (ret) { | |
271 | err = ret; | |
272 | goto out; | |
273 | } | |
274 | ||
275 | btrfs_release_path(root, path); | |
be20aa9d | 276 | start = extent_end; |
7ea394f1 YZ |
277 | if (start <= end) { |
278 | loops++; | |
279 | goto again; | |
280 | } | |
bd09835d | 281 | } else { |
7ea394f1 YZ |
282 | not_found: |
283 | btrfs_end_transaction(trans, root); | |
be20aa9d | 284 | btrfs_free_path(path); |
7ea394f1 | 285 | return cow_file_range(inode, start, end); |
be20aa9d | 286 | } |
7ea394f1 YZ |
287 | out: |
288 | WARN_ON(err); | |
289 | btrfs_end_transaction(trans, root); | |
290 | btrfs_free_path(path); | |
291 | return err; | |
be20aa9d CM |
292 | } |
293 | ||
294 | static int run_delalloc_range(struct inode *inode, u64 start, u64 end) | |
295 | { | |
296 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
297 | int ret; | |
a2135011 | 298 | |
b98b6767 Y |
299 | if (btrfs_test_opt(root, NODATACOW) || |
300 | btrfs_test_flag(inode, NODATACOW)) | |
be20aa9d CM |
301 | ret = run_delalloc_nocow(inode, start, end); |
302 | else | |
303 | ret = cow_file_range(inode, start, end); | |
1832a6d5 | 304 | |
b888db2b CM |
305 | return ret; |
306 | } | |
307 | ||
291d673e | 308 | int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end, |
b0c68f8b | 309 | unsigned long old, unsigned long bits) |
291d673e | 310 | { |
bcbfce8a | 311 | unsigned long flags; |
b0c68f8b | 312 | if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) { |
291d673e | 313 | struct btrfs_root *root = BTRFS_I(inode)->root; |
bcbfce8a | 314 | spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); |
9069218d | 315 | BTRFS_I(inode)->delalloc_bytes += end - start + 1; |
291d673e | 316 | root->fs_info->delalloc_bytes += end - start + 1; |
ea8c2819 CM |
317 | if (list_empty(&BTRFS_I(inode)->delalloc_inodes)) { |
318 | list_add_tail(&BTRFS_I(inode)->delalloc_inodes, | |
319 | &root->fs_info->delalloc_inodes); | |
320 | } | |
bcbfce8a | 321 | spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); |
291d673e CM |
322 | } |
323 | return 0; | |
324 | } | |
325 | ||
326 | int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end, | |
b0c68f8b | 327 | unsigned long old, unsigned long bits) |
291d673e | 328 | { |
b0c68f8b | 329 | if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) { |
291d673e | 330 | struct btrfs_root *root = BTRFS_I(inode)->root; |
bcbfce8a CM |
331 | unsigned long flags; |
332 | ||
333 | spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); | |
b0c68f8b CM |
334 | if (end - start + 1 > root->fs_info->delalloc_bytes) { |
335 | printk("warning: delalloc account %Lu %Lu\n", | |
336 | end - start + 1, root->fs_info->delalloc_bytes); | |
337 | root->fs_info->delalloc_bytes = 0; | |
9069218d | 338 | BTRFS_I(inode)->delalloc_bytes = 0; |
b0c68f8b CM |
339 | } else { |
340 | root->fs_info->delalloc_bytes -= end - start + 1; | |
9069218d | 341 | BTRFS_I(inode)->delalloc_bytes -= end - start + 1; |
b0c68f8b | 342 | } |
ea8c2819 CM |
343 | if (BTRFS_I(inode)->delalloc_bytes == 0 && |
344 | !list_empty(&BTRFS_I(inode)->delalloc_inodes)) { | |
345 | list_del_init(&BTRFS_I(inode)->delalloc_inodes); | |
346 | } | |
bcbfce8a | 347 | spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); |
291d673e CM |
348 | } |
349 | return 0; | |
350 | } | |
351 | ||
239b14b3 CM |
352 | int btrfs_merge_bio_hook(struct page *page, unsigned long offset, |
353 | size_t size, struct bio *bio) | |
354 | { | |
355 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
356 | struct btrfs_mapping_tree *map_tree; | |
239b14b3 | 357 | u64 logical = bio->bi_sector << 9; |
239b14b3 CM |
358 | u64 length = 0; |
359 | u64 map_length; | |
239b14b3 CM |
360 | int ret; |
361 | ||
f2d8d74d | 362 | length = bio->bi_size; |
239b14b3 CM |
363 | map_tree = &root->fs_info->mapping_tree; |
364 | map_length = length; | |
cea9e445 | 365 | ret = btrfs_map_block(map_tree, READ, logical, |
f188591e | 366 | &map_length, NULL, 0); |
cea9e445 | 367 | |
239b14b3 | 368 | if (map_length < length + size) { |
239b14b3 CM |
369 | return 1; |
370 | } | |
371 | return 0; | |
372 | } | |
373 | ||
44b8bd7e | 374 | int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
f188591e | 375 | int mirror_num) |
065631f6 | 376 | { |
065631f6 | 377 | struct btrfs_root *root = BTRFS_I(inode)->root; |
065631f6 | 378 | int ret = 0; |
e015640f | 379 | |
3edf7d33 | 380 | ret = btrfs_csum_one_bio(root, inode, bio); |
44b8bd7e | 381 | BUG_ON(ret); |
e015640f | 382 | |
8b712842 | 383 | return btrfs_map_bio(root, rw, bio, mirror_num, 1); |
44b8bd7e CM |
384 | } |
385 | ||
386 | int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, | |
387 | int mirror_num) | |
388 | { | |
389 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
390 | int ret = 0; | |
391 | ||
e6dcd2dc CM |
392 | ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); |
393 | BUG_ON(ret); | |
065631f6 | 394 | |
7ea394f1 YZ |
395 | if (btrfs_test_opt(root, NODATASUM) || |
396 | btrfs_test_flag(inode, NODATASUM)) { | |
397 | goto mapit; | |
398 | } | |
399 | ||
4d1b5fb4 CM |
400 | if (!(rw & (1 << BIO_RW))) { |
401 | btrfs_lookup_bio_sums(root, inode, bio); | |
402 | goto mapit; | |
403 | } | |
44b8bd7e CM |
404 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, |
405 | inode, rw, bio, mirror_num, | |
406 | __btrfs_submit_bio_hook); | |
0b86a832 | 407 | mapit: |
8b712842 | 408 | return btrfs_map_bio(root, rw, bio, mirror_num, 0); |
065631f6 | 409 | } |
6885f308 | 410 | |
ba1da2f4 | 411 | static noinline int add_pending_csums(struct btrfs_trans_handle *trans, |
e6dcd2dc CM |
412 | struct inode *inode, u64 file_offset, |
413 | struct list_head *list) | |
414 | { | |
415 | struct list_head *cur; | |
416 | struct btrfs_ordered_sum *sum; | |
417 | ||
418 | btrfs_set_trans_block_group(trans, inode); | |
ba1da2f4 | 419 | list_for_each(cur, list) { |
e6dcd2dc | 420 | sum = list_entry(cur, struct btrfs_ordered_sum, list); |
e6dcd2dc CM |
421 | btrfs_csum_file_blocks(trans, BTRFS_I(inode)->root, |
422 | inode, sum); | |
e6dcd2dc CM |
423 | } |
424 | return 0; | |
425 | } | |
426 | ||
ea8c2819 CM |
427 | int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end) |
428 | { | |
429 | return set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end, | |
430 | GFP_NOFS); | |
431 | } | |
432 | ||
247e743c CM |
433 | struct btrfs_writepage_fixup { |
434 | struct page *page; | |
435 | struct btrfs_work work; | |
436 | }; | |
437 | ||
438 | /* see btrfs_writepage_start_hook for details on why this is required */ | |
439 | void btrfs_writepage_fixup_worker(struct btrfs_work *work) | |
440 | { | |
441 | struct btrfs_writepage_fixup *fixup; | |
442 | struct btrfs_ordered_extent *ordered; | |
443 | struct page *page; | |
444 | struct inode *inode; | |
445 | u64 page_start; | |
446 | u64 page_end; | |
447 | ||
448 | fixup = container_of(work, struct btrfs_writepage_fixup, work); | |
449 | page = fixup->page; | |
4a096752 | 450 | again: |
247e743c CM |
451 | lock_page(page); |
452 | if (!page->mapping || !PageDirty(page) || !PageChecked(page)) { | |
453 | ClearPageChecked(page); | |
454 | goto out_page; | |
455 | } | |
456 | ||
457 | inode = page->mapping->host; | |
458 | page_start = page_offset(page); | |
459 | page_end = page_offset(page) + PAGE_CACHE_SIZE - 1; | |
460 | ||
461 | lock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS); | |
4a096752 CM |
462 | |
463 | /* already ordered? We're done */ | |
464 | if (test_range_bit(&BTRFS_I(inode)->io_tree, page_start, page_end, | |
465 | EXTENT_ORDERED, 0)) { | |
247e743c | 466 | goto out; |
4a096752 CM |
467 | } |
468 | ||
469 | ordered = btrfs_lookup_ordered_extent(inode, page_start); | |
470 | if (ordered) { | |
471 | unlock_extent(&BTRFS_I(inode)->io_tree, page_start, | |
472 | page_end, GFP_NOFS); | |
473 | unlock_page(page); | |
474 | btrfs_start_ordered_extent(inode, ordered, 1); | |
475 | goto again; | |
476 | } | |
247e743c | 477 | |
ea8c2819 | 478 | btrfs_set_extent_delalloc(inode, page_start, page_end); |
247e743c CM |
479 | ClearPageChecked(page); |
480 | out: | |
481 | unlock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end, GFP_NOFS); | |
482 | out_page: | |
483 | unlock_page(page); | |
484 | page_cache_release(page); | |
485 | } | |
486 | ||
487 | /* | |
488 | * There are a few paths in the higher layers of the kernel that directly | |
489 | * set the page dirty bit without asking the filesystem if it is a | |
490 | * good idea. This causes problems because we want to make sure COW | |
491 | * properly happens and the data=ordered rules are followed. | |
492 | * | |
493 | * In our case any range that doesn't have the EXTENT_ORDERED bit set | |
494 | * hasn't been properly setup for IO. We kick off an async process | |
495 | * to fix it up. The async helper will wait for ordered extents, set | |
496 | * the delalloc bit and make it safe to write the page. | |
497 | */ | |
498 | int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end) | |
499 | { | |
500 | struct inode *inode = page->mapping->host; | |
501 | struct btrfs_writepage_fixup *fixup; | |
502 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
503 | int ret; | |
504 | ||
505 | ret = test_range_bit(&BTRFS_I(inode)->io_tree, start, end, | |
506 | EXTENT_ORDERED, 0); | |
507 | if (ret) | |
508 | return 0; | |
509 | ||
510 | if (PageChecked(page)) | |
511 | return -EAGAIN; | |
512 | ||
513 | fixup = kzalloc(sizeof(*fixup), GFP_NOFS); | |
514 | if (!fixup) | |
515 | return -EAGAIN; | |
f421950f | 516 | |
247e743c CM |
517 | SetPageChecked(page); |
518 | page_cache_get(page); | |
519 | fixup->work.func = btrfs_writepage_fixup_worker; | |
520 | fixup->page = page; | |
521 | btrfs_queue_worker(&root->fs_info->fixup_workers, &fixup->work); | |
522 | return -EAGAIN; | |
523 | } | |
524 | ||
211f90e6 | 525 | static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) |
e6dcd2dc | 526 | { |
e6dcd2dc CM |
527 | struct btrfs_root *root = BTRFS_I(inode)->root; |
528 | struct btrfs_trans_handle *trans; | |
529 | struct btrfs_ordered_extent *ordered_extent; | |
530 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
531 | u64 alloc_hint = 0; | |
532 | struct list_head list; | |
533 | struct btrfs_key ins; | |
534 | int ret; | |
535 | ||
536 | ret = btrfs_dec_test_ordered_pending(inode, start, end - start + 1); | |
ba1da2f4 | 537 | if (!ret) |
e6dcd2dc | 538 | return 0; |
e6dcd2dc | 539 | |
f9295749 | 540 | trans = btrfs_join_transaction(root, 1); |
e6dcd2dc CM |
541 | |
542 | ordered_extent = btrfs_lookup_ordered_extent(inode, start); | |
543 | BUG_ON(!ordered_extent); | |
7ea394f1 YZ |
544 | if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags)) |
545 | goto nocow; | |
e6dcd2dc CM |
546 | |
547 | lock_extent(io_tree, ordered_extent->file_offset, | |
548 | ordered_extent->file_offset + ordered_extent->len - 1, | |
549 | GFP_NOFS); | |
550 | ||
551 | INIT_LIST_HEAD(&list); | |
552 | ||
553 | ins.objectid = ordered_extent->start; | |
554 | ins.offset = ordered_extent->len; | |
555 | ins.type = BTRFS_EXTENT_ITEM_KEY; | |
e5a2217e | 556 | |
e6dcd2dc CM |
557 | ret = btrfs_alloc_reserved_extent(trans, root, root->root_key.objectid, |
558 | trans->transid, inode->i_ino, | |
559 | ordered_extent->file_offset, &ins); | |
560 | BUG_ON(ret); | |
ee6e6504 CM |
561 | |
562 | mutex_lock(&BTRFS_I(inode)->extent_mutex); | |
e5a2217e | 563 | |
e6dcd2dc CM |
564 | ret = btrfs_drop_extents(trans, root, inode, |
565 | ordered_extent->file_offset, | |
566 | ordered_extent->file_offset + | |
567 | ordered_extent->len, | |
568 | ordered_extent->file_offset, &alloc_hint); | |
569 | BUG_ON(ret); | |
570 | ret = btrfs_insert_file_extent(trans, root, inode->i_ino, | |
571 | ordered_extent->file_offset, | |
572 | ordered_extent->start, | |
573 | ordered_extent->len, | |
574 | ordered_extent->len, 0); | |
575 | BUG_ON(ret); | |
7f3c74fb | 576 | |
e6dcd2dc CM |
577 | btrfs_drop_extent_cache(inode, ordered_extent->file_offset, |
578 | ordered_extent->file_offset + | |
579 | ordered_extent->len - 1); | |
ee6e6504 CM |
580 | mutex_unlock(&BTRFS_I(inode)->extent_mutex); |
581 | ||
e6dcd2dc CM |
582 | inode->i_blocks += ordered_extent->len >> 9; |
583 | unlock_extent(io_tree, ordered_extent->file_offset, | |
584 | ordered_extent->file_offset + ordered_extent->len - 1, | |
585 | GFP_NOFS); | |
7ea394f1 | 586 | nocow: |
e6dcd2dc CM |
587 | add_pending_csums(trans, inode, ordered_extent->file_offset, |
588 | &ordered_extent->list); | |
589 | ||
dbe674a9 | 590 | btrfs_ordered_update_i_size(inode, ordered_extent); |
e02119d5 | 591 | btrfs_update_inode(trans, root, inode); |
e6dcd2dc | 592 | btrfs_remove_ordered_extent(inode, ordered_extent); |
7f3c74fb | 593 | |
e6dcd2dc CM |
594 | /* once for us */ |
595 | btrfs_put_ordered_extent(ordered_extent); | |
596 | /* once for the tree */ | |
597 | btrfs_put_ordered_extent(ordered_extent); | |
598 | ||
e6dcd2dc CM |
599 | btrfs_end_transaction(trans, root); |
600 | return 0; | |
601 | } | |
602 | ||
211f90e6 CM |
603 | int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end, |
604 | struct extent_state *state, int uptodate) | |
605 | { | |
606 | return btrfs_finish_ordered_io(page->mapping->host, start, end); | |
607 | } | |
608 | ||
7e38326f CM |
609 | struct io_failure_record { |
610 | struct page *page; | |
611 | u64 start; | |
612 | u64 len; | |
613 | u64 logical; | |
614 | int last_mirror; | |
615 | }; | |
616 | ||
1259ab75 CM |
617 | int btrfs_io_failed_hook(struct bio *failed_bio, |
618 | struct page *page, u64 start, u64 end, | |
619 | struct extent_state *state) | |
7e38326f CM |
620 | { |
621 | struct io_failure_record *failrec = NULL; | |
622 | u64 private; | |
623 | struct extent_map *em; | |
624 | struct inode *inode = page->mapping->host; | |
625 | struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree; | |
3b951516 | 626 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
7e38326f CM |
627 | struct bio *bio; |
628 | int num_copies; | |
629 | int ret; | |
1259ab75 | 630 | int rw; |
7e38326f CM |
631 | u64 logical; |
632 | ||
633 | ret = get_state_private(failure_tree, start, &private); | |
634 | if (ret) { | |
7e38326f CM |
635 | failrec = kmalloc(sizeof(*failrec), GFP_NOFS); |
636 | if (!failrec) | |
637 | return -ENOMEM; | |
638 | failrec->start = start; | |
639 | failrec->len = end - start + 1; | |
640 | failrec->last_mirror = 0; | |
641 | ||
3b951516 CM |
642 | spin_lock(&em_tree->lock); |
643 | em = lookup_extent_mapping(em_tree, start, failrec->len); | |
644 | if (em->start > start || em->start + em->len < start) { | |
645 | free_extent_map(em); | |
646 | em = NULL; | |
647 | } | |
648 | spin_unlock(&em_tree->lock); | |
7e38326f CM |
649 | |
650 | if (!em || IS_ERR(em)) { | |
651 | kfree(failrec); | |
652 | return -EIO; | |
653 | } | |
654 | logical = start - em->start; | |
655 | logical = em->block_start + logical; | |
656 | failrec->logical = logical; | |
657 | free_extent_map(em); | |
658 | set_extent_bits(failure_tree, start, end, EXTENT_LOCKED | | |
659 | EXTENT_DIRTY, GFP_NOFS); | |
587f7704 CM |
660 | set_state_private(failure_tree, start, |
661 | (u64)(unsigned long)failrec); | |
7e38326f | 662 | } else { |
587f7704 | 663 | failrec = (struct io_failure_record *)(unsigned long)private; |
7e38326f CM |
664 | } |
665 | num_copies = btrfs_num_copies( | |
666 | &BTRFS_I(inode)->root->fs_info->mapping_tree, | |
667 | failrec->logical, failrec->len); | |
668 | failrec->last_mirror++; | |
669 | if (!state) { | |
670 | spin_lock_irq(&BTRFS_I(inode)->io_tree.lock); | |
671 | state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree, | |
672 | failrec->start, | |
673 | EXTENT_LOCKED); | |
674 | if (state && state->start != failrec->start) | |
675 | state = NULL; | |
676 | spin_unlock_irq(&BTRFS_I(inode)->io_tree.lock); | |
677 | } | |
678 | if (!state || failrec->last_mirror > num_copies) { | |
679 | set_state_private(failure_tree, failrec->start, 0); | |
680 | clear_extent_bits(failure_tree, failrec->start, | |
681 | failrec->start + failrec->len - 1, | |
682 | EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS); | |
683 | kfree(failrec); | |
684 | return -EIO; | |
685 | } | |
686 | bio = bio_alloc(GFP_NOFS, 1); | |
687 | bio->bi_private = state; | |
688 | bio->bi_end_io = failed_bio->bi_end_io; | |
689 | bio->bi_sector = failrec->logical >> 9; | |
690 | bio->bi_bdev = failed_bio->bi_bdev; | |
e1c4b745 | 691 | bio->bi_size = 0; |
7e38326f | 692 | bio_add_page(bio, page, failrec->len, start - page_offset(page)); |
1259ab75 CM |
693 | if (failed_bio->bi_rw & (1 << BIO_RW)) |
694 | rw = WRITE; | |
695 | else | |
696 | rw = READ; | |
697 | ||
698 | BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio, | |
699 | failrec->last_mirror); | |
700 | return 0; | |
701 | } | |
702 | ||
703 | int btrfs_clean_io_failures(struct inode *inode, u64 start) | |
704 | { | |
705 | u64 private; | |
706 | u64 private_failure; | |
707 | struct io_failure_record *failure; | |
708 | int ret; | |
709 | ||
710 | private = 0; | |
711 | if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private, | |
712 | (u64)-1, 1, EXTENT_DIRTY)) { | |
713 | ret = get_state_private(&BTRFS_I(inode)->io_failure_tree, | |
714 | start, &private_failure); | |
715 | if (ret == 0) { | |
716 | failure = (struct io_failure_record *)(unsigned long) | |
717 | private_failure; | |
718 | set_state_private(&BTRFS_I(inode)->io_failure_tree, | |
719 | failure->start, 0); | |
720 | clear_extent_bits(&BTRFS_I(inode)->io_failure_tree, | |
721 | failure->start, | |
722 | failure->start + failure->len - 1, | |
723 | EXTENT_DIRTY | EXTENT_LOCKED, | |
724 | GFP_NOFS); | |
725 | kfree(failure); | |
726 | } | |
727 | } | |
7e38326f CM |
728 | return 0; |
729 | } | |
730 | ||
70dec807 CM |
731 | int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
732 | struct extent_state *state) | |
07157aac | 733 | { |
35ebb934 | 734 | size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT); |
07157aac | 735 | struct inode *inode = page->mapping->host; |
d1310b2e | 736 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
07157aac | 737 | char *kaddr; |
aadfeb6e | 738 | u64 private = ~(u32)0; |
07157aac | 739 | int ret; |
ff79f819 CM |
740 | struct btrfs_root *root = BTRFS_I(inode)->root; |
741 | u32 csum = ~(u32)0; | |
bbf0d006 | 742 | unsigned long flags; |
d1310b2e | 743 | |
b98b6767 Y |
744 | if (btrfs_test_opt(root, NODATASUM) || |
745 | btrfs_test_flag(inode, NODATASUM)) | |
b6cda9bc | 746 | return 0; |
c2e639f0 | 747 | if (state && state->start == start) { |
70dec807 CM |
748 | private = state->private; |
749 | ret = 0; | |
750 | } else { | |
751 | ret = get_state_private(io_tree, start, &private); | |
752 | } | |
bbf0d006 | 753 | local_irq_save(flags); |
07157aac CM |
754 | kaddr = kmap_atomic(page, KM_IRQ0); |
755 | if (ret) { | |
756 | goto zeroit; | |
757 | } | |
ff79f819 CM |
758 | csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1); |
759 | btrfs_csum_final(csum, (char *)&csum); | |
760 | if (csum != private) { | |
07157aac CM |
761 | goto zeroit; |
762 | } | |
763 | kunmap_atomic(kaddr, KM_IRQ0); | |
bbf0d006 | 764 | local_irq_restore(flags); |
7e38326f CM |
765 | |
766 | /* if the io failure tree for this inode is non-empty, | |
767 | * check to see if we've recovered from a failed IO | |
768 | */ | |
1259ab75 | 769 | btrfs_clean_io_failures(inode, start); |
07157aac CM |
770 | return 0; |
771 | ||
772 | zeroit: | |
aadfeb6e CM |
773 | printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n", |
774 | page->mapping->host->i_ino, (unsigned long long)start, csum, | |
775 | private); | |
db94535d CM |
776 | memset(kaddr + offset, 1, end - start + 1); |
777 | flush_dcache_page(page); | |
07157aac | 778 | kunmap_atomic(kaddr, KM_IRQ0); |
bbf0d006 | 779 | local_irq_restore(flags); |
3b951516 CM |
780 | if (private == 0) |
781 | return 0; | |
7e38326f | 782 | return -EIO; |
07157aac | 783 | } |
b888db2b | 784 | |
7b128766 JB |
785 | /* |
786 | * This creates an orphan entry for the given inode in case something goes | |
787 | * wrong in the middle of an unlink/truncate. | |
788 | */ | |
789 | int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode) | |
790 | { | |
791 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
792 | int ret = 0; | |
793 | ||
bcc63abb | 794 | spin_lock(&root->list_lock); |
7b128766 JB |
795 | |
796 | /* already on the orphan list, we're good */ | |
797 | if (!list_empty(&BTRFS_I(inode)->i_orphan)) { | |
bcc63abb | 798 | spin_unlock(&root->list_lock); |
7b128766 JB |
799 | return 0; |
800 | } | |
801 | ||
802 | list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list); | |
803 | ||
bcc63abb | 804 | spin_unlock(&root->list_lock); |
7b128766 JB |
805 | |
806 | /* | |
807 | * insert an orphan item to track this unlinked/truncated file | |
808 | */ | |
809 | ret = btrfs_insert_orphan_item(trans, root, inode->i_ino); | |
810 | ||
811 | return ret; | |
812 | } | |
813 | ||
814 | /* | |
815 | * We have done the truncate/delete so we can go ahead and remove the orphan | |
816 | * item for this particular inode. | |
817 | */ | |
818 | int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode) | |
819 | { | |
820 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
821 | int ret = 0; | |
822 | ||
bcc63abb | 823 | spin_lock(&root->list_lock); |
7b128766 JB |
824 | |
825 | if (list_empty(&BTRFS_I(inode)->i_orphan)) { | |
bcc63abb | 826 | spin_unlock(&root->list_lock); |
7b128766 JB |
827 | return 0; |
828 | } | |
829 | ||
830 | list_del_init(&BTRFS_I(inode)->i_orphan); | |
831 | if (!trans) { | |
bcc63abb | 832 | spin_unlock(&root->list_lock); |
7b128766 JB |
833 | return 0; |
834 | } | |
835 | ||
bcc63abb | 836 | spin_unlock(&root->list_lock); |
7b128766 JB |
837 | |
838 | ret = btrfs_del_orphan_item(trans, root, inode->i_ino); | |
839 | ||
840 | return ret; | |
841 | } | |
842 | ||
843 | /* | |
844 | * this cleans up any orphans that may be left on the list from the last use | |
845 | * of this root. | |
846 | */ | |
847 | void btrfs_orphan_cleanup(struct btrfs_root *root) | |
848 | { | |
849 | struct btrfs_path *path; | |
850 | struct extent_buffer *leaf; | |
851 | struct btrfs_item *item; | |
852 | struct btrfs_key key, found_key; | |
853 | struct btrfs_trans_handle *trans; | |
854 | struct inode *inode; | |
855 | int ret = 0, nr_unlink = 0, nr_truncate = 0; | |
856 | ||
857 | /* don't do orphan cleanup if the fs is readonly. */ | |
858 | if (root->inode->i_sb->s_flags & MS_RDONLY) | |
859 | return; | |
860 | ||
861 | path = btrfs_alloc_path(); | |
862 | if (!path) | |
863 | return; | |
864 | path->reada = -1; | |
865 | ||
866 | key.objectid = BTRFS_ORPHAN_OBJECTID; | |
867 | btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY); | |
868 | key.offset = (u64)-1; | |
869 | ||
870 | trans = btrfs_start_transaction(root, 1); | |
871 | btrfs_set_trans_block_group(trans, root->inode); | |
872 | ||
873 | while (1) { | |
874 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
875 | if (ret < 0) { | |
876 | printk(KERN_ERR "Error searching slot for orphan: %d" | |
877 | "\n", ret); | |
878 | break; | |
879 | } | |
880 | ||
881 | /* | |
882 | * if ret == 0 means we found what we were searching for, which | |
883 | * is weird, but possible, so only screw with path if we didnt | |
884 | * find the key and see if we have stuff that matches | |
885 | */ | |
886 | if (ret > 0) { | |
887 | if (path->slots[0] == 0) | |
888 | break; | |
889 | path->slots[0]--; | |
890 | } | |
891 | ||
892 | /* pull out the item */ | |
893 | leaf = path->nodes[0]; | |
894 | item = btrfs_item_nr(leaf, path->slots[0]); | |
895 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
896 | ||
897 | /* make sure the item matches what we want */ | |
898 | if (found_key.objectid != BTRFS_ORPHAN_OBJECTID) | |
899 | break; | |
900 | if (btrfs_key_type(&found_key) != BTRFS_ORPHAN_ITEM_KEY) | |
901 | break; | |
902 | ||
903 | /* release the path since we're done with it */ | |
904 | btrfs_release_path(root, path); | |
905 | ||
906 | /* | |
907 | * this is where we are basically btrfs_lookup, without the | |
908 | * crossing root thing. we store the inode number in the | |
909 | * offset of the orphan item. | |
910 | */ | |
911 | inode = btrfs_iget_locked(root->inode->i_sb, | |
912 | found_key.offset, root); | |
913 | if (!inode) | |
914 | break; | |
915 | ||
916 | if (inode->i_state & I_NEW) { | |
917 | BTRFS_I(inode)->root = root; | |
918 | ||
919 | /* have to set the location manually */ | |
920 | BTRFS_I(inode)->location.objectid = inode->i_ino; | |
921 | BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY; | |
922 | BTRFS_I(inode)->location.offset = 0; | |
923 | ||
924 | btrfs_read_locked_inode(inode); | |
925 | unlock_new_inode(inode); | |
926 | } | |
927 | ||
928 | /* | |
929 | * add this inode to the orphan list so btrfs_orphan_del does | |
930 | * the proper thing when we hit it | |
931 | */ | |
bcc63abb | 932 | spin_lock(&root->list_lock); |
7b128766 | 933 | list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list); |
bcc63abb | 934 | spin_unlock(&root->list_lock); |
7b128766 JB |
935 | |
936 | /* | |
937 | * if this is a bad inode, means we actually succeeded in | |
938 | * removing the inode, but not the orphan record, which means | |
939 | * we need to manually delete the orphan since iput will just | |
940 | * do a destroy_inode | |
941 | */ | |
942 | if (is_bad_inode(inode)) { | |
943 | btrfs_orphan_del(trans, inode); | |
944 | iput(inode); | |
945 | continue; | |
946 | } | |
947 | ||
948 | /* if we have links, this was a truncate, lets do that */ | |
949 | if (inode->i_nlink) { | |
950 | nr_truncate++; | |
951 | btrfs_truncate(inode); | |
952 | } else { | |
953 | nr_unlink++; | |
954 | } | |
955 | ||
956 | /* this will do delete_inode and everything for us */ | |
957 | iput(inode); | |
958 | } | |
959 | ||
960 | if (nr_unlink) | |
961 | printk(KERN_INFO "btrfs: unlinked %d orphans\n", nr_unlink); | |
962 | if (nr_truncate) | |
963 | printk(KERN_INFO "btrfs: truncated %d orphans\n", nr_truncate); | |
964 | ||
965 | btrfs_free_path(path); | |
966 | btrfs_end_transaction(trans, root); | |
967 | } | |
968 | ||
39279cc3 CM |
969 | void btrfs_read_locked_inode(struct inode *inode) |
970 | { | |
971 | struct btrfs_path *path; | |
5f39d397 | 972 | struct extent_buffer *leaf; |
39279cc3 | 973 | struct btrfs_inode_item *inode_item; |
0b86a832 | 974 | struct btrfs_timespec *tspec; |
39279cc3 CM |
975 | struct btrfs_root *root = BTRFS_I(inode)->root; |
976 | struct btrfs_key location; | |
977 | u64 alloc_group_block; | |
618e21d5 | 978 | u32 rdev; |
39279cc3 CM |
979 | int ret; |
980 | ||
981 | path = btrfs_alloc_path(); | |
982 | BUG_ON(!path); | |
39279cc3 | 983 | memcpy(&location, &BTRFS_I(inode)->location, sizeof(location)); |
dc17ff8f | 984 | |
39279cc3 | 985 | ret = btrfs_lookup_inode(NULL, root, path, &location, 0); |
5f39d397 | 986 | if (ret) |
39279cc3 | 987 | goto make_bad; |
39279cc3 | 988 | |
5f39d397 CM |
989 | leaf = path->nodes[0]; |
990 | inode_item = btrfs_item_ptr(leaf, path->slots[0], | |
991 | struct btrfs_inode_item); | |
992 | ||
993 | inode->i_mode = btrfs_inode_mode(leaf, inode_item); | |
994 | inode->i_nlink = btrfs_inode_nlink(leaf, inode_item); | |
995 | inode->i_uid = btrfs_inode_uid(leaf, inode_item); | |
996 | inode->i_gid = btrfs_inode_gid(leaf, inode_item); | |
dbe674a9 | 997 | btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item)); |
5f39d397 CM |
998 | |
999 | tspec = btrfs_inode_atime(inode_item); | |
1000 | inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec); | |
1001 | inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec); | |
1002 | ||
1003 | tspec = btrfs_inode_mtime(inode_item); | |
1004 | inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec); | |
1005 | inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec); | |
1006 | ||
1007 | tspec = btrfs_inode_ctime(inode_item); | |
1008 | inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec); | |
1009 | inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec); | |
1010 | ||
1011 | inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item); | |
e02119d5 CM |
1012 | BTRFS_I(inode)->generation = btrfs_inode_generation(leaf, inode_item); |
1013 | inode->i_generation = BTRFS_I(inode)->generation; | |
618e21d5 | 1014 | inode->i_rdev = 0; |
5f39d397 CM |
1015 | rdev = btrfs_inode_rdev(leaf, inode_item); |
1016 | ||
aec7477b JB |
1017 | BTRFS_I(inode)->index_cnt = (u64)-1; |
1018 | ||
5f39d397 | 1019 | alloc_group_block = btrfs_inode_block_group(leaf, inode_item); |
39279cc3 CM |
1020 | BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info, |
1021 | alloc_group_block); | |
b98b6767 | 1022 | BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item); |
e52ec0eb CM |
1023 | if (!BTRFS_I(inode)->block_group) { |
1024 | BTRFS_I(inode)->block_group = btrfs_find_block_group(root, | |
0b86a832 CM |
1025 | NULL, 0, |
1026 | BTRFS_BLOCK_GROUP_METADATA, 0); | |
e52ec0eb | 1027 | } |
39279cc3 CM |
1028 | btrfs_free_path(path); |
1029 | inode_item = NULL; | |
1030 | ||
39279cc3 | 1031 | switch (inode->i_mode & S_IFMT) { |
39279cc3 CM |
1032 | case S_IFREG: |
1033 | inode->i_mapping->a_ops = &btrfs_aops; | |
04160088 | 1034 | inode->i_mapping->backing_dev_info = &root->fs_info->bdi; |
d1310b2e | 1035 | BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; |
39279cc3 CM |
1036 | inode->i_fop = &btrfs_file_operations; |
1037 | inode->i_op = &btrfs_file_inode_operations; | |
1038 | break; | |
1039 | case S_IFDIR: | |
1040 | inode->i_fop = &btrfs_dir_file_operations; | |
1041 | if (root == root->fs_info->tree_root) | |
1042 | inode->i_op = &btrfs_dir_ro_inode_operations; | |
1043 | else | |
1044 | inode->i_op = &btrfs_dir_inode_operations; | |
1045 | break; | |
1046 | case S_IFLNK: | |
1047 | inode->i_op = &btrfs_symlink_inode_operations; | |
1048 | inode->i_mapping->a_ops = &btrfs_symlink_aops; | |
04160088 | 1049 | inode->i_mapping->backing_dev_info = &root->fs_info->bdi; |
39279cc3 | 1050 | break; |
618e21d5 JB |
1051 | default: |
1052 | init_special_inode(inode, inode->i_mode, rdev); | |
1053 | break; | |
39279cc3 CM |
1054 | } |
1055 | return; | |
1056 | ||
1057 | make_bad: | |
39279cc3 | 1058 | btrfs_free_path(path); |
39279cc3 CM |
1059 | make_bad_inode(inode); |
1060 | } | |
1061 | ||
e02119d5 CM |
1062 | static void fill_inode_item(struct btrfs_trans_handle *trans, |
1063 | struct extent_buffer *leaf, | |
5f39d397 | 1064 | struct btrfs_inode_item *item, |
39279cc3 CM |
1065 | struct inode *inode) |
1066 | { | |
5f39d397 CM |
1067 | btrfs_set_inode_uid(leaf, item, inode->i_uid); |
1068 | btrfs_set_inode_gid(leaf, item, inode->i_gid); | |
dbe674a9 | 1069 | btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size); |
5f39d397 CM |
1070 | btrfs_set_inode_mode(leaf, item, inode->i_mode); |
1071 | btrfs_set_inode_nlink(leaf, item, inode->i_nlink); | |
1072 | ||
1073 | btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item), | |
1074 | inode->i_atime.tv_sec); | |
1075 | btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item), | |
1076 | inode->i_atime.tv_nsec); | |
1077 | ||
1078 | btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item), | |
1079 | inode->i_mtime.tv_sec); | |
1080 | btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item), | |
1081 | inode->i_mtime.tv_nsec); | |
1082 | ||
1083 | btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item), | |
1084 | inode->i_ctime.tv_sec); | |
1085 | btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item), | |
1086 | inode->i_ctime.tv_nsec); | |
1087 | ||
1088 | btrfs_set_inode_nblocks(leaf, item, inode->i_blocks); | |
e02119d5 CM |
1089 | btrfs_set_inode_generation(leaf, item, BTRFS_I(inode)->generation); |
1090 | btrfs_set_inode_transid(leaf, item, trans->transid); | |
5f39d397 | 1091 | btrfs_set_inode_rdev(leaf, item, inode->i_rdev); |
b98b6767 | 1092 | btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags); |
5f39d397 | 1093 | btrfs_set_inode_block_group(leaf, item, |
39279cc3 CM |
1094 | BTRFS_I(inode)->block_group->key.objectid); |
1095 | } | |
1096 | ||
ba1da2f4 | 1097 | int noinline btrfs_update_inode(struct btrfs_trans_handle *trans, |
39279cc3 CM |
1098 | struct btrfs_root *root, |
1099 | struct inode *inode) | |
1100 | { | |
1101 | struct btrfs_inode_item *inode_item; | |
1102 | struct btrfs_path *path; | |
5f39d397 | 1103 | struct extent_buffer *leaf; |
39279cc3 CM |
1104 | int ret; |
1105 | ||
1106 | path = btrfs_alloc_path(); | |
1107 | BUG_ON(!path); | |
39279cc3 CM |
1108 | ret = btrfs_lookup_inode(trans, root, path, |
1109 | &BTRFS_I(inode)->location, 1); | |
1110 | if (ret) { | |
1111 | if (ret > 0) | |
1112 | ret = -ENOENT; | |
1113 | goto failed; | |
1114 | } | |
1115 | ||
5f39d397 CM |
1116 | leaf = path->nodes[0]; |
1117 | inode_item = btrfs_item_ptr(leaf, path->slots[0], | |
39279cc3 CM |
1118 | struct btrfs_inode_item); |
1119 | ||
e02119d5 | 1120 | fill_inode_item(trans, leaf, inode_item, inode); |
5f39d397 | 1121 | btrfs_mark_buffer_dirty(leaf); |
15ee9bc7 | 1122 | btrfs_set_inode_last_trans(trans, inode); |
39279cc3 CM |
1123 | ret = 0; |
1124 | failed: | |
39279cc3 CM |
1125 | btrfs_free_path(path); |
1126 | return ret; | |
1127 | } | |
1128 | ||
1129 | ||
e02119d5 CM |
1130 | int btrfs_unlink_inode(struct btrfs_trans_handle *trans, |
1131 | struct btrfs_root *root, | |
1132 | struct inode *dir, struct inode *inode, | |
1133 | const char *name, int name_len) | |
39279cc3 CM |
1134 | { |
1135 | struct btrfs_path *path; | |
39279cc3 | 1136 | int ret = 0; |
5f39d397 | 1137 | struct extent_buffer *leaf; |
39279cc3 | 1138 | struct btrfs_dir_item *di; |
5f39d397 | 1139 | struct btrfs_key key; |
aec7477b | 1140 | u64 index; |
39279cc3 CM |
1141 | |
1142 | path = btrfs_alloc_path(); | |
54aa1f4d CM |
1143 | if (!path) { |
1144 | ret = -ENOMEM; | |
1145 | goto err; | |
1146 | } | |
1147 | ||
39279cc3 CM |
1148 | di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino, |
1149 | name, name_len, -1); | |
1150 | if (IS_ERR(di)) { | |
1151 | ret = PTR_ERR(di); | |
1152 | goto err; | |
1153 | } | |
1154 | if (!di) { | |
1155 | ret = -ENOENT; | |
1156 | goto err; | |
1157 | } | |
5f39d397 CM |
1158 | leaf = path->nodes[0]; |
1159 | btrfs_dir_item_key_to_cpu(leaf, di, &key); | |
39279cc3 | 1160 | ret = btrfs_delete_one_dir_name(trans, root, path, di); |
54aa1f4d CM |
1161 | if (ret) |
1162 | goto err; | |
39279cc3 CM |
1163 | btrfs_release_path(root, path); |
1164 | ||
aec7477b | 1165 | ret = btrfs_del_inode_ref(trans, root, name, name_len, |
e02119d5 CM |
1166 | inode->i_ino, |
1167 | dir->i_ino, &index); | |
aec7477b JB |
1168 | if (ret) { |
1169 | printk("failed to delete reference to %.*s, " | |
1170 | "inode %lu parent %lu\n", name_len, name, | |
e02119d5 | 1171 | inode->i_ino, dir->i_ino); |
aec7477b JB |
1172 | goto err; |
1173 | } | |
1174 | ||
39279cc3 | 1175 | di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino, |
aec7477b | 1176 | index, name, name_len, -1); |
39279cc3 CM |
1177 | if (IS_ERR(di)) { |
1178 | ret = PTR_ERR(di); | |
1179 | goto err; | |
1180 | } | |
1181 | if (!di) { | |
1182 | ret = -ENOENT; | |
1183 | goto err; | |
1184 | } | |
1185 | ret = btrfs_delete_one_dir_name(trans, root, path, di); | |
925baedd | 1186 | btrfs_release_path(root, path); |
39279cc3 | 1187 | |
e02119d5 CM |
1188 | ret = btrfs_del_inode_ref_in_log(trans, root, name, name_len, |
1189 | inode, dir->i_ino); | |
1190 | BUG_ON(ret); | |
1191 | ||
1192 | ret = btrfs_del_dir_entries_in_log(trans, root, name, name_len, | |
1193 | dir, index); | |
1194 | BUG_ON(ret); | |
39279cc3 CM |
1195 | err: |
1196 | btrfs_free_path(path); | |
e02119d5 CM |
1197 | if (ret) |
1198 | goto out; | |
1199 | ||
1200 | btrfs_i_size_write(dir, dir->i_size - name_len * 2); | |
1201 | inode->i_ctime = dir->i_mtime = dir->i_ctime = CURRENT_TIME; | |
1202 | btrfs_update_inode(trans, root, dir); | |
1203 | btrfs_drop_nlink(inode); | |
1204 | ret = btrfs_update_inode(trans, root, inode); | |
1205 | dir->i_sb->s_dirt = 1; | |
1206 | out: | |
39279cc3 CM |
1207 | return ret; |
1208 | } | |
1209 | ||
1210 | static int btrfs_unlink(struct inode *dir, struct dentry *dentry) | |
1211 | { | |
1212 | struct btrfs_root *root; | |
1213 | struct btrfs_trans_handle *trans; | |
7b128766 | 1214 | struct inode *inode = dentry->d_inode; |
39279cc3 | 1215 | int ret; |
1832a6d5 | 1216 | unsigned long nr = 0; |
39279cc3 CM |
1217 | |
1218 | root = BTRFS_I(dir)->root; | |
1832a6d5 CM |
1219 | |
1220 | ret = btrfs_check_free_space(root, 1, 1); | |
1221 | if (ret) | |
1222 | goto fail; | |
1223 | ||
39279cc3 | 1224 | trans = btrfs_start_transaction(root, 1); |
5f39d397 | 1225 | |
39279cc3 | 1226 | btrfs_set_trans_block_group(trans, dir); |
e02119d5 CM |
1227 | ret = btrfs_unlink_inode(trans, root, dir, dentry->d_inode, |
1228 | dentry->d_name.name, dentry->d_name.len); | |
7b128766 JB |
1229 | |
1230 | if (inode->i_nlink == 0) | |
1231 | ret = btrfs_orphan_add(trans, inode); | |
1232 | ||
d3c2fdcf | 1233 | nr = trans->blocks_used; |
5f39d397 | 1234 | |
89ce8a63 | 1235 | btrfs_end_transaction_throttle(trans, root); |
1832a6d5 | 1236 | fail: |
d3c2fdcf | 1237 | btrfs_btree_balance_dirty(root, nr); |
39279cc3 CM |
1238 | return ret; |
1239 | } | |
1240 | ||
1241 | static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) | |
1242 | { | |
1243 | struct inode *inode = dentry->d_inode; | |
1832a6d5 | 1244 | int err = 0; |
39279cc3 CM |
1245 | int ret; |
1246 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
39279cc3 | 1247 | struct btrfs_trans_handle *trans; |
1832a6d5 | 1248 | unsigned long nr = 0; |
39279cc3 | 1249 | |
925baedd | 1250 | if (inode->i_size > BTRFS_EMPTY_DIR_SIZE) { |
134d4512 | 1251 | return -ENOTEMPTY; |
925baedd | 1252 | } |
134d4512 | 1253 | |
1832a6d5 CM |
1254 | ret = btrfs_check_free_space(root, 1, 1); |
1255 | if (ret) | |
1256 | goto fail; | |
1257 | ||
39279cc3 CM |
1258 | trans = btrfs_start_transaction(root, 1); |
1259 | btrfs_set_trans_block_group(trans, dir); | |
39279cc3 | 1260 | |
7b128766 JB |
1261 | err = btrfs_orphan_add(trans, inode); |
1262 | if (err) | |
1263 | goto fail_trans; | |
1264 | ||
39279cc3 | 1265 | /* now the directory is empty */ |
e02119d5 CM |
1266 | err = btrfs_unlink_inode(trans, root, dir, dentry->d_inode, |
1267 | dentry->d_name.name, dentry->d_name.len); | |
39279cc3 | 1268 | if (!err) { |
dbe674a9 | 1269 | btrfs_i_size_write(inode, 0); |
39279cc3 | 1270 | } |
3954401f | 1271 | |
7b128766 | 1272 | fail_trans: |
d3c2fdcf | 1273 | nr = trans->blocks_used; |
89ce8a63 | 1274 | ret = btrfs_end_transaction_throttle(trans, root); |
1832a6d5 | 1275 | fail: |
d3c2fdcf | 1276 | btrfs_btree_balance_dirty(root, nr); |
3954401f | 1277 | |
39279cc3 CM |
1278 | if (ret && !err) |
1279 | err = ret; | |
1280 | return err; | |
1281 | } | |
1282 | ||
39279cc3 CM |
1283 | /* |
1284 | * this can truncate away extent items, csum items and directory items. | |
1285 | * It starts at a high offset and removes keys until it can't find | |
1286 | * any higher than i_size. | |
1287 | * | |
1288 | * csum items that cross the new i_size are truncated to the new size | |
1289 | * as well. | |
7b128766 JB |
1290 | * |
1291 | * min_type is the minimum key type to truncate down to. If set to 0, this | |
1292 | * will kill all the items on this inode, including the INODE_ITEM_KEY. | |
39279cc3 | 1293 | */ |
e02119d5 CM |
1294 | noinline int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans, |
1295 | struct btrfs_root *root, | |
1296 | struct inode *inode, | |
1297 | u64 new_size, u32 min_type) | |
39279cc3 CM |
1298 | { |
1299 | int ret; | |
1300 | struct btrfs_path *path; | |
1301 | struct btrfs_key key; | |
5f39d397 | 1302 | struct btrfs_key found_key; |
39279cc3 | 1303 | u32 found_type; |
5f39d397 | 1304 | struct extent_buffer *leaf; |
39279cc3 CM |
1305 | struct btrfs_file_extent_item *fi; |
1306 | u64 extent_start = 0; | |
db94535d | 1307 | u64 extent_num_bytes = 0; |
39279cc3 | 1308 | u64 item_end = 0; |
7bb86316 | 1309 | u64 root_gen = 0; |
d8d5f3e1 | 1310 | u64 root_owner = 0; |
39279cc3 CM |
1311 | int found_extent; |
1312 | int del_item; | |
85e21bac CM |
1313 | int pending_del_nr = 0; |
1314 | int pending_del_slot = 0; | |
179e29e4 | 1315 | int extent_type = -1; |
3b951516 | 1316 | u64 mask = root->sectorsize - 1; |
39279cc3 | 1317 | |
e02119d5 CM |
1318 | if (root->ref_cows) |
1319 | btrfs_drop_extent_cache(inode, | |
1320 | new_size & (~mask), (u64)-1); | |
39279cc3 | 1321 | path = btrfs_alloc_path(); |
3c69faec | 1322 | path->reada = -1; |
39279cc3 | 1323 | BUG_ON(!path); |
5f39d397 | 1324 | |
39279cc3 CM |
1325 | /* FIXME, add redo link to tree so we don't leak on crash */ |
1326 | key.objectid = inode->i_ino; | |
1327 | key.offset = (u64)-1; | |
5f39d397 CM |
1328 | key.type = (u8)-1; |
1329 | ||
85e21bac CM |
1330 | btrfs_init_path(path); |
1331 | search_again: | |
1332 | ret = btrfs_search_slot(trans, root, &key, path, -1, 1); | |
1333 | if (ret < 0) { | |
1334 | goto error; | |
1335 | } | |
1336 | if (ret > 0) { | |
e02119d5 CM |
1337 | /* there are no items in the tree for us to truncate, we're |
1338 | * done | |
1339 | */ | |
1340 | if (path->slots[0] == 0) { | |
1341 | ret = 0; | |
1342 | goto error; | |
1343 | } | |
85e21bac CM |
1344 | path->slots[0]--; |
1345 | } | |
1346 | ||
39279cc3 | 1347 | while(1) { |
39279cc3 | 1348 | fi = NULL; |
5f39d397 CM |
1349 | leaf = path->nodes[0]; |
1350 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
1351 | found_type = btrfs_key_type(&found_key); | |
39279cc3 | 1352 | |
5f39d397 | 1353 | if (found_key.objectid != inode->i_ino) |
39279cc3 | 1354 | break; |
5f39d397 | 1355 | |
85e21bac | 1356 | if (found_type < min_type) |
39279cc3 CM |
1357 | break; |
1358 | ||
5f39d397 | 1359 | item_end = found_key.offset; |
39279cc3 | 1360 | if (found_type == BTRFS_EXTENT_DATA_KEY) { |
5f39d397 | 1361 | fi = btrfs_item_ptr(leaf, path->slots[0], |
39279cc3 | 1362 | struct btrfs_file_extent_item); |
179e29e4 CM |
1363 | extent_type = btrfs_file_extent_type(leaf, fi); |
1364 | if (extent_type != BTRFS_FILE_EXTENT_INLINE) { | |
5f39d397 | 1365 | item_end += |
db94535d | 1366 | btrfs_file_extent_num_bytes(leaf, fi); |
179e29e4 CM |
1367 | } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) { |
1368 | struct btrfs_item *item = btrfs_item_nr(leaf, | |
1369 | path->slots[0]); | |
1370 | item_end += btrfs_file_extent_inline_len(leaf, | |
1371 | item); | |
39279cc3 | 1372 | } |
008630c1 | 1373 | item_end--; |
39279cc3 CM |
1374 | } |
1375 | if (found_type == BTRFS_CSUM_ITEM_KEY) { | |
1376 | ret = btrfs_csum_truncate(trans, root, path, | |
e02119d5 | 1377 | new_size); |
39279cc3 CM |
1378 | BUG_ON(ret); |
1379 | } | |
e02119d5 | 1380 | if (item_end < new_size) { |
b888db2b CM |
1381 | if (found_type == BTRFS_DIR_ITEM_KEY) { |
1382 | found_type = BTRFS_INODE_ITEM_KEY; | |
1383 | } else if (found_type == BTRFS_EXTENT_ITEM_KEY) { | |
1384 | found_type = BTRFS_CSUM_ITEM_KEY; | |
85e21bac CM |
1385 | } else if (found_type == BTRFS_EXTENT_DATA_KEY) { |
1386 | found_type = BTRFS_XATTR_ITEM_KEY; | |
1387 | } else if (found_type == BTRFS_XATTR_ITEM_KEY) { | |
1388 | found_type = BTRFS_INODE_REF_KEY; | |
b888db2b CM |
1389 | } else if (found_type) { |
1390 | found_type--; | |
1391 | } else { | |
1392 | break; | |
39279cc3 | 1393 | } |
a61721d5 | 1394 | btrfs_set_key_type(&key, found_type); |
85e21bac | 1395 | goto next; |
39279cc3 | 1396 | } |
e02119d5 | 1397 | if (found_key.offset >= new_size) |
39279cc3 CM |
1398 | del_item = 1; |
1399 | else | |
1400 | del_item = 0; | |
1401 | found_extent = 0; | |
1402 | ||
1403 | /* FIXME, shrink the extent if the ref count is only 1 */ | |
179e29e4 CM |
1404 | if (found_type != BTRFS_EXTENT_DATA_KEY) |
1405 | goto delete; | |
1406 | ||
1407 | if (extent_type != BTRFS_FILE_EXTENT_INLINE) { | |
39279cc3 | 1408 | u64 num_dec; |
db94535d | 1409 | extent_start = btrfs_file_extent_disk_bytenr(leaf, fi); |
39279cc3 | 1410 | if (!del_item) { |
db94535d CM |
1411 | u64 orig_num_bytes = |
1412 | btrfs_file_extent_num_bytes(leaf, fi); | |
e02119d5 | 1413 | extent_num_bytes = new_size - |
5f39d397 | 1414 | found_key.offset + root->sectorsize - 1; |
b1632b10 Y |
1415 | extent_num_bytes = extent_num_bytes & |
1416 | ~((u64)root->sectorsize - 1); | |
db94535d CM |
1417 | btrfs_set_file_extent_num_bytes(leaf, fi, |
1418 | extent_num_bytes); | |
1419 | num_dec = (orig_num_bytes - | |
9069218d | 1420 | extent_num_bytes); |
e02119d5 | 1421 | if (root->ref_cows && extent_start != 0) |
9069218d | 1422 | dec_i_blocks(inode, num_dec); |
5f39d397 | 1423 | btrfs_mark_buffer_dirty(leaf); |
39279cc3 | 1424 | } else { |
db94535d CM |
1425 | extent_num_bytes = |
1426 | btrfs_file_extent_disk_num_bytes(leaf, | |
1427 | fi); | |
39279cc3 | 1428 | /* FIXME blocksize != 4096 */ |
9069218d | 1429 | num_dec = btrfs_file_extent_num_bytes(leaf, fi); |
39279cc3 CM |
1430 | if (extent_start != 0) { |
1431 | found_extent = 1; | |
e02119d5 CM |
1432 | if (root->ref_cows) |
1433 | dec_i_blocks(inode, num_dec); | |
1434 | } | |
1435 | if (root->ref_cows) { | |
1436 | root_gen = | |
1437 | btrfs_header_generation(leaf); | |
39279cc3 | 1438 | } |
d8d5f3e1 | 1439 | root_owner = btrfs_header_owner(leaf); |
39279cc3 | 1440 | } |
9069218d CM |
1441 | } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) { |
1442 | if (!del_item) { | |
e02119d5 CM |
1443 | u32 size = new_size - found_key.offset; |
1444 | ||
1445 | if (root->ref_cows) { | |
1446 | dec_i_blocks(inode, item_end + 1 - | |
1447 | found_key.offset - size); | |
1448 | } | |
1449 | size = | |
1450 | btrfs_file_extent_calc_inline_size(size); | |
9069218d | 1451 | ret = btrfs_truncate_item(trans, root, path, |
e02119d5 | 1452 | size, 1); |
9069218d | 1453 | BUG_ON(ret); |
e02119d5 | 1454 | } else if (root->ref_cows) { |
9069218d CM |
1455 | dec_i_blocks(inode, item_end + 1 - |
1456 | found_key.offset); | |
1457 | } | |
39279cc3 | 1458 | } |
179e29e4 | 1459 | delete: |
39279cc3 | 1460 | if (del_item) { |
85e21bac CM |
1461 | if (!pending_del_nr) { |
1462 | /* no pending yet, add ourselves */ | |
1463 | pending_del_slot = path->slots[0]; | |
1464 | pending_del_nr = 1; | |
1465 | } else if (pending_del_nr && | |
1466 | path->slots[0] + 1 == pending_del_slot) { | |
1467 | /* hop on the pending chunk */ | |
1468 | pending_del_nr++; | |
1469 | pending_del_slot = path->slots[0]; | |
1470 | } else { | |
1471 | printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot); | |
1472 | } | |
39279cc3 CM |
1473 | } else { |
1474 | break; | |
1475 | } | |
39279cc3 CM |
1476 | if (found_extent) { |
1477 | ret = btrfs_free_extent(trans, root, extent_start, | |
7bb86316 | 1478 | extent_num_bytes, |
d8d5f3e1 | 1479 | root_owner, |
7bb86316 CM |
1480 | root_gen, inode->i_ino, |
1481 | found_key.offset, 0); | |
39279cc3 CM |
1482 | BUG_ON(ret); |
1483 | } | |
85e21bac CM |
1484 | next: |
1485 | if (path->slots[0] == 0) { | |
1486 | if (pending_del_nr) | |
1487 | goto del_pending; | |
1488 | btrfs_release_path(root, path); | |
1489 | goto search_again; | |
1490 | } | |
1491 | ||
1492 | path->slots[0]--; | |
1493 | if (pending_del_nr && | |
1494 | path->slots[0] + 1 != pending_del_slot) { | |
1495 | struct btrfs_key debug; | |
1496 | del_pending: | |
1497 | btrfs_item_key_to_cpu(path->nodes[0], &debug, | |
1498 | pending_del_slot); | |
1499 | ret = btrfs_del_items(trans, root, path, | |
1500 | pending_del_slot, | |
1501 | pending_del_nr); | |
1502 | BUG_ON(ret); | |
1503 | pending_del_nr = 0; | |
1504 | btrfs_release_path(root, path); | |
1505 | goto search_again; | |
1506 | } | |
39279cc3 CM |
1507 | } |
1508 | ret = 0; | |
1509 | error: | |
85e21bac CM |
1510 | if (pending_del_nr) { |
1511 | ret = btrfs_del_items(trans, root, path, pending_del_slot, | |
1512 | pending_del_nr); | |
1513 | } | |
39279cc3 CM |
1514 | btrfs_free_path(path); |
1515 | inode->i_sb->s_dirt = 1; | |
1516 | return ret; | |
1517 | } | |
1518 | ||
1519 | /* | |
1520 | * taken from block_truncate_page, but does cow as it zeros out | |
1521 | * any bytes left in the last page in the file. | |
1522 | */ | |
1523 | static int btrfs_truncate_page(struct address_space *mapping, loff_t from) | |
1524 | { | |
1525 | struct inode *inode = mapping->host; | |
db94535d | 1526 | struct btrfs_root *root = BTRFS_I(inode)->root; |
e6dcd2dc CM |
1527 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
1528 | struct btrfs_ordered_extent *ordered; | |
1529 | char *kaddr; | |
db94535d | 1530 | u32 blocksize = root->sectorsize; |
39279cc3 CM |
1531 | pgoff_t index = from >> PAGE_CACHE_SHIFT; |
1532 | unsigned offset = from & (PAGE_CACHE_SIZE-1); | |
1533 | struct page *page; | |
39279cc3 | 1534 | int ret = 0; |
a52d9a80 | 1535 | u64 page_start; |
e6dcd2dc | 1536 | u64 page_end; |
39279cc3 CM |
1537 | |
1538 | if ((offset & (blocksize - 1)) == 0) | |
1539 | goto out; | |
1540 | ||
1541 | ret = -ENOMEM; | |
211c17f5 | 1542 | again: |
39279cc3 CM |
1543 | page = grab_cache_page(mapping, index); |
1544 | if (!page) | |
1545 | goto out; | |
e6dcd2dc CM |
1546 | |
1547 | page_start = page_offset(page); | |
1548 | page_end = page_start + PAGE_CACHE_SIZE - 1; | |
1549 | ||
39279cc3 | 1550 | if (!PageUptodate(page)) { |
9ebefb18 | 1551 | ret = btrfs_readpage(NULL, page); |
39279cc3 | 1552 | lock_page(page); |
211c17f5 CM |
1553 | if (page->mapping != mapping) { |
1554 | unlock_page(page); | |
1555 | page_cache_release(page); | |
1556 | goto again; | |
1557 | } | |
39279cc3 CM |
1558 | if (!PageUptodate(page)) { |
1559 | ret = -EIO; | |
89642229 | 1560 | goto out_unlock; |
39279cc3 CM |
1561 | } |
1562 | } | |
211c17f5 | 1563 | wait_on_page_writeback(page); |
e6dcd2dc CM |
1564 | |
1565 | lock_extent(io_tree, page_start, page_end, GFP_NOFS); | |
1566 | set_page_extent_mapped(page); | |
1567 | ||
1568 | ordered = btrfs_lookup_ordered_extent(inode, page_start); | |
1569 | if (ordered) { | |
1570 | unlock_extent(io_tree, page_start, page_end, GFP_NOFS); | |
1571 | unlock_page(page); | |
1572 | page_cache_release(page); | |
eb84ae03 | 1573 | btrfs_start_ordered_extent(inode, ordered, 1); |
e6dcd2dc CM |
1574 | btrfs_put_ordered_extent(ordered); |
1575 | goto again; | |
1576 | } | |
1577 | ||
ea8c2819 | 1578 | btrfs_set_extent_delalloc(inode, page_start, page_end); |
e6dcd2dc CM |
1579 | ret = 0; |
1580 | if (offset != PAGE_CACHE_SIZE) { | |
1581 | kaddr = kmap(page); | |
1582 | memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset); | |
1583 | flush_dcache_page(page); | |
1584 | kunmap(page); | |
1585 | } | |
247e743c | 1586 | ClearPageChecked(page); |
e6dcd2dc CM |
1587 | set_page_dirty(page); |
1588 | unlock_extent(io_tree, page_start, page_end, GFP_NOFS); | |
39279cc3 | 1589 | |
89642229 | 1590 | out_unlock: |
39279cc3 CM |
1591 | unlock_page(page); |
1592 | page_cache_release(page); | |
1593 | out: | |
1594 | return ret; | |
1595 | } | |
1596 | ||
1597 | static int btrfs_setattr(struct dentry *dentry, struct iattr *attr) | |
1598 | { | |
1599 | struct inode *inode = dentry->d_inode; | |
1600 | int err; | |
1601 | ||
1602 | err = inode_change_ok(inode, attr); | |
1603 | if (err) | |
1604 | return err; | |
1605 | ||
1606 | if (S_ISREG(inode->i_mode) && | |
1607 | attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) { | |
1608 | struct btrfs_trans_handle *trans; | |
1609 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
d1310b2e | 1610 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
2bf5a725 | 1611 | |
5f39d397 | 1612 | u64 mask = root->sectorsize - 1; |
1b0f7c29 | 1613 | u64 hole_start = (inode->i_size + mask) & ~mask; |
f392a938 | 1614 | u64 block_end = (attr->ia_size + mask) & ~mask; |
39279cc3 | 1615 | u64 hole_size; |
179e29e4 | 1616 | u64 alloc_hint = 0; |
39279cc3 | 1617 | |
1b0f7c29 | 1618 | if (attr->ia_size <= hole_start) |
39279cc3 CM |
1619 | goto out; |
1620 | ||
1832a6d5 | 1621 | err = btrfs_check_free_space(root, 1, 0); |
1832a6d5 CM |
1622 | if (err) |
1623 | goto fail; | |
1624 | ||
39279cc3 CM |
1625 | btrfs_truncate_page(inode->i_mapping, inode->i_size); |
1626 | ||
5f56406a | 1627 | hole_size = block_end - hole_start; |
7c2fe32a CM |
1628 | while(1) { |
1629 | struct btrfs_ordered_extent *ordered; | |
1630 | btrfs_wait_ordered_range(inode, hole_start, hole_size); | |
1631 | ||
1632 | lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); | |
1633 | ordered = btrfs_lookup_ordered_extent(inode, hole_start); | |
1634 | if (ordered) { | |
1635 | unlock_extent(io_tree, hole_start, | |
1636 | block_end - 1, GFP_NOFS); | |
1637 | btrfs_put_ordered_extent(ordered); | |
1638 | } else { | |
1639 | break; | |
1640 | } | |
1641 | } | |
39279cc3 | 1642 | |
39279cc3 CM |
1643 | trans = btrfs_start_transaction(root, 1); |
1644 | btrfs_set_trans_block_group(trans, inode); | |
ee6e6504 | 1645 | mutex_lock(&BTRFS_I(inode)->extent_mutex); |
2bf5a725 | 1646 | err = btrfs_drop_extents(trans, root, inode, |
1b0f7c29 | 1647 | hole_start, block_end, hole_start, |
3326d1b0 | 1648 | &alloc_hint); |
2bf5a725 | 1649 | |
179e29e4 CM |
1650 | if (alloc_hint != EXTENT_MAP_INLINE) { |
1651 | err = btrfs_insert_file_extent(trans, root, | |
1652 | inode->i_ino, | |
5f56406a | 1653 | hole_start, 0, 0, |
f2eb0a24 | 1654 | hole_size, 0); |
d1310b2e | 1655 | btrfs_drop_extent_cache(inode, hole_start, |
3b951516 | 1656 | (u64)-1); |
5f56406a | 1657 | btrfs_check_file(root, inode); |
179e29e4 | 1658 | } |
ee6e6504 | 1659 | mutex_unlock(&BTRFS_I(inode)->extent_mutex); |
39279cc3 | 1660 | btrfs_end_transaction(trans, root); |
1b0f7c29 | 1661 | unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS); |
54aa1f4d CM |
1662 | if (err) |
1663 | return err; | |
39279cc3 CM |
1664 | } |
1665 | out: | |
1666 | err = inode_setattr(inode, attr); | |
33268eaf JB |
1667 | |
1668 | if (!err && ((attr->ia_valid & ATTR_MODE))) | |
1669 | err = btrfs_acl_chmod(inode); | |
1832a6d5 | 1670 | fail: |
39279cc3 CM |
1671 | return err; |
1672 | } | |
61295eb8 | 1673 | |
39279cc3 CM |
1674 | void btrfs_delete_inode(struct inode *inode) |
1675 | { | |
1676 | struct btrfs_trans_handle *trans; | |
1677 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
d3c2fdcf | 1678 | unsigned long nr; |
39279cc3 CM |
1679 | int ret; |
1680 | ||
1681 | truncate_inode_pages(&inode->i_data, 0); | |
1682 | if (is_bad_inode(inode)) { | |
7b128766 | 1683 | btrfs_orphan_del(NULL, inode); |
39279cc3 CM |
1684 | goto no_delete; |
1685 | } | |
4a096752 | 1686 | btrfs_wait_ordered_range(inode, 0, (u64)-1); |
5f39d397 | 1687 | |
dbe674a9 | 1688 | btrfs_i_size_write(inode, 0); |
39279cc3 | 1689 | trans = btrfs_start_transaction(root, 1); |
5f39d397 | 1690 | |
39279cc3 | 1691 | btrfs_set_trans_block_group(trans, inode); |
e02119d5 | 1692 | ret = btrfs_truncate_inode_items(trans, root, inode, inode->i_size, 0); |
7b128766 JB |
1693 | if (ret) { |
1694 | btrfs_orphan_del(NULL, inode); | |
54aa1f4d | 1695 | goto no_delete_lock; |
7b128766 JB |
1696 | } |
1697 | ||
1698 | btrfs_orphan_del(trans, inode); | |
85e21bac | 1699 | |
d3c2fdcf | 1700 | nr = trans->blocks_used; |
85e21bac | 1701 | clear_inode(inode); |
5f39d397 | 1702 | |
39279cc3 | 1703 | btrfs_end_transaction(trans, root); |
d3c2fdcf | 1704 | btrfs_btree_balance_dirty(root, nr); |
39279cc3 | 1705 | return; |
54aa1f4d CM |
1706 | |
1707 | no_delete_lock: | |
d3c2fdcf | 1708 | nr = trans->blocks_used; |
54aa1f4d | 1709 | btrfs_end_transaction(trans, root); |
d3c2fdcf | 1710 | btrfs_btree_balance_dirty(root, nr); |
39279cc3 CM |
1711 | no_delete: |
1712 | clear_inode(inode); | |
1713 | } | |
1714 | ||
1715 | /* | |
1716 | * this returns the key found in the dir entry in the location pointer. | |
1717 | * If no dir entries were found, location->objectid is 0. | |
1718 | */ | |
1719 | static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry, | |
1720 | struct btrfs_key *location) | |
1721 | { | |
1722 | const char *name = dentry->d_name.name; | |
1723 | int namelen = dentry->d_name.len; | |
1724 | struct btrfs_dir_item *di; | |
1725 | struct btrfs_path *path; | |
1726 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
0d9f7f3e | 1727 | int ret = 0; |
39279cc3 CM |
1728 | |
1729 | path = btrfs_alloc_path(); | |
1730 | BUG_ON(!path); | |
3954401f | 1731 | |
39279cc3 CM |
1732 | di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name, |
1733 | namelen, 0); | |
0d9f7f3e Y |
1734 | if (IS_ERR(di)) |
1735 | ret = PTR_ERR(di); | |
39279cc3 | 1736 | if (!di || IS_ERR(di)) { |
3954401f | 1737 | goto out_err; |
39279cc3 | 1738 | } |
5f39d397 | 1739 | btrfs_dir_item_key_to_cpu(path->nodes[0], di, location); |
39279cc3 | 1740 | out: |
39279cc3 CM |
1741 | btrfs_free_path(path); |
1742 | return ret; | |
3954401f CM |
1743 | out_err: |
1744 | location->objectid = 0; | |
1745 | goto out; | |
39279cc3 CM |
1746 | } |
1747 | ||
1748 | /* | |
1749 | * when we hit a tree root in a directory, the btrfs part of the inode | |
1750 | * needs to be changed to reflect the root directory of the tree root. This | |
1751 | * is kind of like crossing a mount point. | |
1752 | */ | |
1753 | static int fixup_tree_root_location(struct btrfs_root *root, | |
1754 | struct btrfs_key *location, | |
58176a96 JB |
1755 | struct btrfs_root **sub_root, |
1756 | struct dentry *dentry) | |
39279cc3 | 1757 | { |
39279cc3 CM |
1758 | struct btrfs_root_item *ri; |
1759 | ||
1760 | if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY) | |
1761 | return 0; | |
1762 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) | |
1763 | return 0; | |
1764 | ||
58176a96 JB |
1765 | *sub_root = btrfs_read_fs_root(root->fs_info, location, |
1766 | dentry->d_name.name, | |
1767 | dentry->d_name.len); | |
39279cc3 CM |
1768 | if (IS_ERR(*sub_root)) |
1769 | return PTR_ERR(*sub_root); | |
1770 | ||
1771 | ri = &(*sub_root)->root_item; | |
1772 | location->objectid = btrfs_root_dirid(ri); | |
39279cc3 CM |
1773 | btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY); |
1774 | location->offset = 0; | |
1775 | ||
39279cc3 CM |
1776 | return 0; |
1777 | } | |
1778 | ||
e02119d5 | 1779 | static noinline void init_btrfs_i(struct inode *inode) |
39279cc3 | 1780 | { |
e02119d5 CM |
1781 | struct btrfs_inode *bi = BTRFS_I(inode); |
1782 | ||
1783 | bi->i_acl = NULL; | |
1784 | bi->i_default_acl = NULL; | |
1785 | ||
1786 | bi->generation = 0; | |
1787 | bi->last_trans = 0; | |
1788 | bi->logged_trans = 0; | |
1789 | bi->delalloc_bytes = 0; | |
1790 | bi->disk_i_size = 0; | |
1791 | bi->flags = 0; | |
1792 | bi->index_cnt = (u64)-1; | |
d1310b2e CM |
1793 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS); |
1794 | extent_io_tree_init(&BTRFS_I(inode)->io_tree, | |
b888db2b | 1795 | inode->i_mapping, GFP_NOFS); |
7e38326f CM |
1796 | extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree, |
1797 | inode->i_mapping, GFP_NOFS); | |
ea8c2819 | 1798 | INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes); |
ba1da2f4 | 1799 | btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree); |
1b1e2135 | 1800 | mutex_init(&BTRFS_I(inode)->csum_mutex); |
ee6e6504 | 1801 | mutex_init(&BTRFS_I(inode)->extent_mutex); |
e02119d5 CM |
1802 | mutex_init(&BTRFS_I(inode)->log_mutex); |
1803 | } | |
1804 | ||
1805 | static int btrfs_init_locked_inode(struct inode *inode, void *p) | |
1806 | { | |
1807 | struct btrfs_iget_args *args = p; | |
1808 | inode->i_ino = args->ino; | |
1809 | init_btrfs_i(inode); | |
1810 | BTRFS_I(inode)->root = args->root; | |
39279cc3 CM |
1811 | return 0; |
1812 | } | |
1813 | ||
1814 | static int btrfs_find_actor(struct inode *inode, void *opaque) | |
1815 | { | |
1816 | struct btrfs_iget_args *args = opaque; | |
1817 | return (args->ino == inode->i_ino && | |
1818 | args->root == BTRFS_I(inode)->root); | |
1819 | } | |
1820 | ||
1821 | struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid, | |
1822 | struct btrfs_root *root) | |
1823 | { | |
1824 | struct inode *inode; | |
1825 | struct btrfs_iget_args args; | |
1826 | args.ino = objectid; | |
1827 | args.root = root; | |
1828 | ||
1829 | inode = iget5_locked(s, objectid, btrfs_find_actor, | |
1830 | btrfs_init_locked_inode, | |
1831 | (void *)&args); | |
1832 | return inode; | |
1833 | } | |
1834 | ||
1a54ef8c BR |
1835 | /* Get an inode object given its location and corresponding root. |
1836 | * Returns in *is_new if the inode was read from disk | |
1837 | */ | |
1838 | struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location, | |
1839 | struct btrfs_root *root, int *is_new) | |
1840 | { | |
1841 | struct inode *inode; | |
1842 | ||
1843 | inode = btrfs_iget_locked(s, location->objectid, root); | |
1844 | if (!inode) | |
1845 | return ERR_PTR(-EACCES); | |
1846 | ||
1847 | if (inode->i_state & I_NEW) { | |
1848 | BTRFS_I(inode)->root = root; | |
1849 | memcpy(&BTRFS_I(inode)->location, location, sizeof(*location)); | |
1850 | btrfs_read_locked_inode(inode); | |
1851 | unlock_new_inode(inode); | |
1852 | if (is_new) | |
1853 | *is_new = 1; | |
1854 | } else { | |
1855 | if (is_new) | |
1856 | *is_new = 0; | |
1857 | } | |
1858 | ||
1859 | return inode; | |
1860 | } | |
1861 | ||
39279cc3 CM |
1862 | static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry, |
1863 | struct nameidata *nd) | |
1864 | { | |
1865 | struct inode * inode; | |
1866 | struct btrfs_inode *bi = BTRFS_I(dir); | |
1867 | struct btrfs_root *root = bi->root; | |
1868 | struct btrfs_root *sub_root = root; | |
1869 | struct btrfs_key location; | |
1a54ef8c | 1870 | int ret, new, do_orphan = 0; |
39279cc3 CM |
1871 | |
1872 | if (dentry->d_name.len > BTRFS_NAME_LEN) | |
1873 | return ERR_PTR(-ENAMETOOLONG); | |
5f39d397 | 1874 | |
39279cc3 | 1875 | ret = btrfs_inode_by_name(dir, dentry, &location); |
5f39d397 | 1876 | |
39279cc3 CM |
1877 | if (ret < 0) |
1878 | return ERR_PTR(ret); | |
5f39d397 | 1879 | |
39279cc3 CM |
1880 | inode = NULL; |
1881 | if (location.objectid) { | |
58176a96 JB |
1882 | ret = fixup_tree_root_location(root, &location, &sub_root, |
1883 | dentry); | |
39279cc3 CM |
1884 | if (ret < 0) |
1885 | return ERR_PTR(ret); | |
1886 | if (ret > 0) | |
1887 | return ERR_PTR(-ENOENT); | |
1a54ef8c BR |
1888 | inode = btrfs_iget(dir->i_sb, &location, sub_root, &new); |
1889 | if (IS_ERR(inode)) | |
1890 | return ERR_CAST(inode); | |
1891 | ||
1892 | /* the inode and parent dir are two different roots */ | |
1893 | if (new && root != sub_root) { | |
1894 | igrab(inode); | |
1895 | sub_root->inode = inode; | |
1896 | do_orphan = 1; | |
39279cc3 CM |
1897 | } |
1898 | } | |
7b128766 JB |
1899 | |
1900 | if (unlikely(do_orphan)) | |
1901 | btrfs_orphan_cleanup(sub_root); | |
1902 | ||
39279cc3 CM |
1903 | return d_splice_alias(inode, dentry); |
1904 | } | |
1905 | ||
39279cc3 CM |
1906 | static unsigned char btrfs_filetype_table[] = { |
1907 | DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK | |
1908 | }; | |
1909 | ||
cbdf5a24 DW |
1910 | static int btrfs_real_readdir(struct file *filp, void *dirent, |
1911 | filldir_t filldir) | |
39279cc3 | 1912 | { |
6da6abae | 1913 | struct inode *inode = filp->f_dentry->d_inode; |
39279cc3 CM |
1914 | struct btrfs_root *root = BTRFS_I(inode)->root; |
1915 | struct btrfs_item *item; | |
1916 | struct btrfs_dir_item *di; | |
1917 | struct btrfs_key key; | |
5f39d397 | 1918 | struct btrfs_key found_key; |
39279cc3 CM |
1919 | struct btrfs_path *path; |
1920 | int ret; | |
1921 | u32 nritems; | |
5f39d397 | 1922 | struct extent_buffer *leaf; |
39279cc3 CM |
1923 | int slot; |
1924 | int advance; | |
1925 | unsigned char d_type; | |
1926 | int over = 0; | |
1927 | u32 di_cur; | |
1928 | u32 di_total; | |
1929 | u32 di_len; | |
1930 | int key_type = BTRFS_DIR_INDEX_KEY; | |
5f39d397 CM |
1931 | char tmp_name[32]; |
1932 | char *name_ptr; | |
1933 | int name_len; | |
39279cc3 CM |
1934 | |
1935 | /* FIXME, use a real flag for deciding about the key type */ | |
1936 | if (root->fs_info->tree_root == root) | |
1937 | key_type = BTRFS_DIR_ITEM_KEY; | |
5f39d397 | 1938 | |
3954401f CM |
1939 | /* special case for "." */ |
1940 | if (filp->f_pos == 0) { | |
1941 | over = filldir(dirent, ".", 1, | |
1942 | 1, inode->i_ino, | |
1943 | DT_DIR); | |
1944 | if (over) | |
1945 | return 0; | |
1946 | filp->f_pos = 1; | |
1947 | } | |
3954401f CM |
1948 | /* special case for .., just use the back ref */ |
1949 | if (filp->f_pos == 1) { | |
5ecc7e5d | 1950 | u64 pino = parent_ino(filp->f_path.dentry); |
3954401f | 1951 | over = filldir(dirent, "..", 2, |
5ecc7e5d | 1952 | 2, pino, DT_DIR); |
3954401f | 1953 | if (over) |
49593bfa | 1954 | return 0; |
3954401f CM |
1955 | filp->f_pos = 2; |
1956 | } | |
1957 | ||
49593bfa DW |
1958 | path = btrfs_alloc_path(); |
1959 | path->reada = 2; | |
1960 | ||
39279cc3 CM |
1961 | btrfs_set_key_type(&key, key_type); |
1962 | key.offset = filp->f_pos; | |
49593bfa | 1963 | key.objectid = inode->i_ino; |
5f39d397 | 1964 | |
39279cc3 CM |
1965 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
1966 | if (ret < 0) | |
1967 | goto err; | |
1968 | advance = 0; | |
49593bfa DW |
1969 | |
1970 | while (1) { | |
5f39d397 CM |
1971 | leaf = path->nodes[0]; |
1972 | nritems = btrfs_header_nritems(leaf); | |
39279cc3 CM |
1973 | slot = path->slots[0]; |
1974 | if (advance || slot >= nritems) { | |
49593bfa | 1975 | if (slot >= nritems - 1) { |
39279cc3 CM |
1976 | ret = btrfs_next_leaf(root, path); |
1977 | if (ret) | |
1978 | break; | |
5f39d397 CM |
1979 | leaf = path->nodes[0]; |
1980 | nritems = btrfs_header_nritems(leaf); | |
39279cc3 CM |
1981 | slot = path->slots[0]; |
1982 | } else { | |
1983 | slot++; | |
1984 | path->slots[0]++; | |
1985 | } | |
1986 | } | |
1987 | advance = 1; | |
5f39d397 CM |
1988 | item = btrfs_item_nr(leaf, slot); |
1989 | btrfs_item_key_to_cpu(leaf, &found_key, slot); | |
1990 | ||
1991 | if (found_key.objectid != key.objectid) | |
39279cc3 | 1992 | break; |
5f39d397 | 1993 | if (btrfs_key_type(&found_key) != key_type) |
39279cc3 | 1994 | break; |
5f39d397 | 1995 | if (found_key.offset < filp->f_pos) |
39279cc3 | 1996 | continue; |
5f39d397 CM |
1997 | |
1998 | filp->f_pos = found_key.offset; | |
49593bfa | 1999 | |
39279cc3 CM |
2000 | di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item); |
2001 | di_cur = 0; | |
5f39d397 | 2002 | di_total = btrfs_item_size(leaf, item); |
49593bfa DW |
2003 | |
2004 | while (di_cur < di_total) { | |
5f39d397 CM |
2005 | struct btrfs_key location; |
2006 | ||
2007 | name_len = btrfs_dir_name_len(leaf, di); | |
49593bfa | 2008 | if (name_len <= sizeof(tmp_name)) { |
5f39d397 CM |
2009 | name_ptr = tmp_name; |
2010 | } else { | |
2011 | name_ptr = kmalloc(name_len, GFP_NOFS); | |
49593bfa DW |
2012 | if (!name_ptr) { |
2013 | ret = -ENOMEM; | |
2014 | goto err; | |
2015 | } | |
5f39d397 CM |
2016 | } |
2017 | read_extent_buffer(leaf, name_ptr, | |
2018 | (unsigned long)(di + 1), name_len); | |
2019 | ||
2020 | d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)]; | |
2021 | btrfs_dir_item_key_to_cpu(leaf, di, &location); | |
5f39d397 | 2022 | over = filldir(dirent, name_ptr, name_len, |
49593bfa | 2023 | found_key.offset, location.objectid, |
39279cc3 | 2024 | d_type); |
5f39d397 CM |
2025 | |
2026 | if (name_ptr != tmp_name) | |
2027 | kfree(name_ptr); | |
2028 | ||
39279cc3 CM |
2029 | if (over) |
2030 | goto nopos; | |
49593bfa | 2031 | |
5103e947 | 2032 | di_len = btrfs_dir_name_len(leaf, di) + |
49593bfa | 2033 | btrfs_dir_data_len(leaf, di) + sizeof(*di); |
39279cc3 CM |
2034 | di_cur += di_len; |
2035 | di = (struct btrfs_dir_item *)((char *)di + di_len); | |
2036 | } | |
2037 | } | |
49593bfa DW |
2038 | |
2039 | /* Reached end of directory/root. Bump pos past the last item. */ | |
5e591a07 YZ |
2040 | if (key_type == BTRFS_DIR_INDEX_KEY) |
2041 | filp->f_pos = INT_LIMIT(typeof(filp->f_pos)); | |
2042 | else | |
2043 | filp->f_pos++; | |
39279cc3 CM |
2044 | nopos: |
2045 | ret = 0; | |
2046 | err: | |
39279cc3 | 2047 | btrfs_free_path(path); |
39279cc3 CM |
2048 | return ret; |
2049 | } | |
2050 | ||
cbdf5a24 DW |
2051 | /* Kernels earlier than 2.6.28 still have the NFS deadlock where nfsd |
2052 | will call the file system's ->lookup() method from within its | |
2053 | filldir callback, which in turn was called from the file system's | |
2054 | ->readdir() method. And will deadlock for many file systems. */ | |
2055 | #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28) | |
2056 | ||
2057 | struct nfshack_dirent { | |
2058 | u64 ino; | |
2059 | loff_t offset; | |
2060 | int namlen; | |
2061 | unsigned int d_type; | |
2062 | char name[]; | |
2063 | }; | |
2064 | ||
2065 | struct nfshack_readdir { | |
2066 | char *dirent; | |
2067 | size_t used; | |
f2322b1c | 2068 | int full; |
cbdf5a24 DW |
2069 | }; |
2070 | ||
2071 | ||
2072 | ||
2073 | static int btrfs_nfshack_filldir(void *__buf, const char *name, int namlen, | |
2074 | loff_t offset, u64 ino, unsigned int d_type) | |
2075 | { | |
2076 | struct nfshack_readdir *buf = __buf; | |
2077 | struct nfshack_dirent *de = (void *)(buf->dirent + buf->used); | |
2078 | unsigned int reclen; | |
2079 | ||
2080 | reclen = ALIGN(sizeof(struct nfshack_dirent) + namlen, sizeof(u64)); | |
f2322b1c DW |
2081 | if (buf->used + reclen > PAGE_SIZE) { |
2082 | buf->full = 1; | |
cbdf5a24 | 2083 | return -EINVAL; |
f2322b1c | 2084 | } |
cbdf5a24 DW |
2085 | |
2086 | de->namlen = namlen; | |
2087 | de->offset = offset; | |
2088 | de->ino = ino; | |
2089 | de->d_type = d_type; | |
2090 | memcpy(de->name, name, namlen); | |
2091 | buf->used += reclen; | |
2092 | ||
2093 | return 0; | |
2094 | } | |
2095 | ||
2096 | static int btrfs_nfshack_readdir(struct file *file, void *dirent, | |
2097 | filldir_t filldir) | |
2098 | { | |
2099 | struct nfshack_readdir buf; | |
2100 | struct nfshack_dirent *de; | |
2101 | int err; | |
2102 | int size; | |
2103 | loff_t offset; | |
2104 | ||
2105 | buf.dirent = (void *)__get_free_page(GFP_KERNEL); | |
2106 | if (!buf.dirent) | |
2107 | return -ENOMEM; | |
2108 | ||
2109 | offset = file->f_pos; | |
2110 | ||
f2322b1c | 2111 | do { |
cbdf5a24 DW |
2112 | unsigned int reclen; |
2113 | ||
2114 | buf.used = 0; | |
f2322b1c | 2115 | buf.full = 0; |
cbdf5a24 DW |
2116 | err = btrfs_real_readdir(file, &buf, btrfs_nfshack_filldir); |
2117 | if (err) | |
2118 | break; | |
2119 | ||
2120 | size = buf.used; | |
2121 | ||
2122 | if (!size) | |
2123 | break; | |
2124 | ||
2125 | de = (struct nfshack_dirent *)buf.dirent; | |
2126 | while (size > 0) { | |
2127 | offset = de->offset; | |
2128 | ||
2129 | if (filldir(dirent, de->name, de->namlen, de->offset, | |
2130 | de->ino, de->d_type)) | |
2131 | goto done; | |
2132 | offset = file->f_pos; | |
2133 | ||
2134 | reclen = ALIGN(sizeof(*de) + de->namlen, | |
2135 | sizeof(u64)); | |
2136 | size -= reclen; | |
2137 | de = (struct nfshack_dirent *)((char *)de + reclen); | |
2138 | } | |
f2322b1c | 2139 | } while (buf.full); |
cbdf5a24 DW |
2140 | |
2141 | done: | |
2142 | free_page((unsigned long)buf.dirent); | |
2143 | file->f_pos = offset; | |
2144 | ||
2145 | return err; | |
2146 | } | |
2147 | #endif | |
2148 | ||
39279cc3 CM |
2149 | int btrfs_write_inode(struct inode *inode, int wait) |
2150 | { | |
2151 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
2152 | struct btrfs_trans_handle *trans; | |
2153 | int ret = 0; | |
2154 | ||
4ca8b41e CM |
2155 | if (root->fs_info->closing > 1) |
2156 | return 0; | |
2157 | ||
39279cc3 | 2158 | if (wait) { |
f9295749 | 2159 | trans = btrfs_join_transaction(root, 1); |
39279cc3 CM |
2160 | btrfs_set_trans_block_group(trans, inode); |
2161 | ret = btrfs_commit_transaction(trans, root); | |
39279cc3 CM |
2162 | } |
2163 | return ret; | |
2164 | } | |
2165 | ||
2166 | /* | |
54aa1f4d | 2167 | * This is somewhat expensive, updating the tree every time the |
39279cc3 CM |
2168 | * inode changes. But, it is most likely to find the inode in cache. |
2169 | * FIXME, needs more benchmarking...there are no reasons other than performance | |
2170 | * to keep or drop this code. | |
2171 | */ | |
2172 | void btrfs_dirty_inode(struct inode *inode) | |
2173 | { | |
2174 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
2175 | struct btrfs_trans_handle *trans; | |
2176 | ||
f9295749 | 2177 | trans = btrfs_join_transaction(root, 1); |
39279cc3 CM |
2178 | btrfs_set_trans_block_group(trans, inode); |
2179 | btrfs_update_inode(trans, root, inode); | |
2180 | btrfs_end_transaction(trans, root); | |
39279cc3 CM |
2181 | } |
2182 | ||
aec7477b JB |
2183 | static int btrfs_set_inode_index_count(struct inode *inode) |
2184 | { | |
2185 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
2186 | struct btrfs_key key, found_key; | |
2187 | struct btrfs_path *path; | |
2188 | struct extent_buffer *leaf; | |
2189 | int ret; | |
2190 | ||
2191 | key.objectid = inode->i_ino; | |
2192 | btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY); | |
2193 | key.offset = (u64)-1; | |
2194 | ||
2195 | path = btrfs_alloc_path(); | |
2196 | if (!path) | |
2197 | return -ENOMEM; | |
2198 | ||
2199 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
2200 | if (ret < 0) | |
2201 | goto out; | |
2202 | /* FIXME: we should be able to handle this */ | |
2203 | if (ret == 0) | |
2204 | goto out; | |
2205 | ret = 0; | |
2206 | ||
2207 | /* | |
2208 | * MAGIC NUMBER EXPLANATION: | |
2209 | * since we search a directory based on f_pos we have to start at 2 | |
2210 | * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody | |
2211 | * else has to start at 2 | |
2212 | */ | |
2213 | if (path->slots[0] == 0) { | |
2214 | BTRFS_I(inode)->index_cnt = 2; | |
2215 | goto out; | |
2216 | } | |
2217 | ||
2218 | path->slots[0]--; | |
2219 | ||
2220 | leaf = path->nodes[0]; | |
2221 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
2222 | ||
2223 | if (found_key.objectid != inode->i_ino || | |
2224 | btrfs_key_type(&found_key) != BTRFS_DIR_INDEX_KEY) { | |
2225 | BTRFS_I(inode)->index_cnt = 2; | |
2226 | goto out; | |
2227 | } | |
2228 | ||
2229 | BTRFS_I(inode)->index_cnt = found_key.offset + 1; | |
2230 | out: | |
2231 | btrfs_free_path(path); | |
2232 | return ret; | |
2233 | } | |
2234 | ||
00e4e6b3 CM |
2235 | static int btrfs_set_inode_index(struct inode *dir, struct inode *inode, |
2236 | u64 *index) | |
aec7477b JB |
2237 | { |
2238 | int ret = 0; | |
2239 | ||
2240 | if (BTRFS_I(dir)->index_cnt == (u64)-1) { | |
2241 | ret = btrfs_set_inode_index_count(dir); | |
8d5bf1cb | 2242 | if (ret) { |
aec7477b | 2243 | return ret; |
8d5bf1cb | 2244 | } |
aec7477b JB |
2245 | } |
2246 | ||
00e4e6b3 | 2247 | *index = BTRFS_I(dir)->index_cnt; |
aec7477b JB |
2248 | BTRFS_I(dir)->index_cnt++; |
2249 | ||
2250 | return ret; | |
2251 | } | |
2252 | ||
39279cc3 CM |
2253 | static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, |
2254 | struct btrfs_root *root, | |
aec7477b | 2255 | struct inode *dir, |
9c58309d CM |
2256 | const char *name, int name_len, |
2257 | u64 ref_objectid, | |
39279cc3 CM |
2258 | u64 objectid, |
2259 | struct btrfs_block_group_cache *group, | |
00e4e6b3 | 2260 | int mode, u64 *index) |
39279cc3 CM |
2261 | { |
2262 | struct inode *inode; | |
5f39d397 | 2263 | struct btrfs_inode_item *inode_item; |
6324fbf3 | 2264 | struct btrfs_block_group_cache *new_inode_group; |
39279cc3 | 2265 | struct btrfs_key *location; |
5f39d397 | 2266 | struct btrfs_path *path; |
9c58309d CM |
2267 | struct btrfs_inode_ref *ref; |
2268 | struct btrfs_key key[2]; | |
2269 | u32 sizes[2]; | |
2270 | unsigned long ptr; | |
39279cc3 CM |
2271 | int ret; |
2272 | int owner; | |
2273 | ||
5f39d397 CM |
2274 | path = btrfs_alloc_path(); |
2275 | BUG_ON(!path); | |
2276 | ||
39279cc3 CM |
2277 | inode = new_inode(root->fs_info->sb); |
2278 | if (!inode) | |
2279 | return ERR_PTR(-ENOMEM); | |
2280 | ||
aec7477b | 2281 | if (dir) { |
00e4e6b3 | 2282 | ret = btrfs_set_inode_index(dir, inode, index); |
aec7477b JB |
2283 | if (ret) |
2284 | return ERR_PTR(ret); | |
aec7477b JB |
2285 | } |
2286 | /* | |
2287 | * index_cnt is ignored for everything but a dir, | |
2288 | * btrfs_get_inode_index_count has an explanation for the magic | |
2289 | * number | |
2290 | */ | |
e02119d5 | 2291 | init_btrfs_i(inode); |
aec7477b | 2292 | BTRFS_I(inode)->index_cnt = 2; |
39279cc3 | 2293 | BTRFS_I(inode)->root = root; |
e02119d5 | 2294 | BTRFS_I(inode)->generation = trans->transid; |
b888db2b | 2295 | |
39279cc3 CM |
2296 | if (mode & S_IFDIR) |
2297 | owner = 0; | |
2298 | else | |
2299 | owner = 1; | |
6324fbf3 | 2300 | new_inode_group = btrfs_find_block_group(root, group, 0, |
0b86a832 | 2301 | BTRFS_BLOCK_GROUP_METADATA, owner); |
6324fbf3 CM |
2302 | if (!new_inode_group) { |
2303 | printk("find_block group failed\n"); | |
2304 | new_inode_group = group; | |
2305 | } | |
2306 | BTRFS_I(inode)->block_group = new_inode_group; | |
9c58309d CM |
2307 | |
2308 | key[0].objectid = objectid; | |
2309 | btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY); | |
2310 | key[0].offset = 0; | |
2311 | ||
2312 | key[1].objectid = objectid; | |
2313 | btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY); | |
2314 | key[1].offset = ref_objectid; | |
2315 | ||
2316 | sizes[0] = sizeof(struct btrfs_inode_item); | |
2317 | sizes[1] = name_len + sizeof(*ref); | |
2318 | ||
2319 | ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2); | |
2320 | if (ret != 0) | |
5f39d397 CM |
2321 | goto fail; |
2322 | ||
9c58309d CM |
2323 | if (objectid > root->highest_inode) |
2324 | root->highest_inode = objectid; | |
2325 | ||
39279cc3 CM |
2326 | inode->i_uid = current->fsuid; |
2327 | inode->i_gid = current->fsgid; | |
2328 | inode->i_mode = mode; | |
2329 | inode->i_ino = objectid; | |
2330 | inode->i_blocks = 0; | |
2331 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; | |
5f39d397 CM |
2332 | inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0], |
2333 | struct btrfs_inode_item); | |
e02119d5 | 2334 | fill_inode_item(trans, path->nodes[0], inode_item, inode); |
9c58309d CM |
2335 | |
2336 | ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1, | |
2337 | struct btrfs_inode_ref); | |
2338 | btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len); | |
00e4e6b3 | 2339 | btrfs_set_inode_ref_index(path->nodes[0], ref, *index); |
9c58309d CM |
2340 | ptr = (unsigned long)(ref + 1); |
2341 | write_extent_buffer(path->nodes[0], name, ptr, name_len); | |
2342 | ||
5f39d397 CM |
2343 | btrfs_mark_buffer_dirty(path->nodes[0]); |
2344 | btrfs_free_path(path); | |
2345 | ||
39279cc3 CM |
2346 | location = &BTRFS_I(inode)->location; |
2347 | location->objectid = objectid; | |
39279cc3 CM |
2348 | location->offset = 0; |
2349 | btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY); | |
2350 | ||
39279cc3 CM |
2351 | insert_inode_hash(inode); |
2352 | return inode; | |
5f39d397 | 2353 | fail: |
aec7477b JB |
2354 | if (dir) |
2355 | BTRFS_I(dir)->index_cnt--; | |
5f39d397 CM |
2356 | btrfs_free_path(path); |
2357 | return ERR_PTR(ret); | |
39279cc3 CM |
2358 | } |
2359 | ||
2360 | static inline u8 btrfs_inode_type(struct inode *inode) | |
2361 | { | |
2362 | return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT]; | |
2363 | } | |
2364 | ||
e02119d5 CM |
2365 | int btrfs_add_link(struct btrfs_trans_handle *trans, |
2366 | struct inode *parent_inode, struct inode *inode, | |
2367 | const char *name, int name_len, int add_backref, u64 index) | |
39279cc3 CM |
2368 | { |
2369 | int ret; | |
2370 | struct btrfs_key key; | |
e02119d5 | 2371 | struct btrfs_root *root = BTRFS_I(parent_inode)->root; |
5f39d397 | 2372 | |
39279cc3 | 2373 | key.objectid = inode->i_ino; |
39279cc3 CM |
2374 | btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY); |
2375 | key.offset = 0; | |
2376 | ||
e02119d5 CM |
2377 | ret = btrfs_insert_dir_item(trans, root, name, name_len, |
2378 | parent_inode->i_ino, | |
aec7477b | 2379 | &key, btrfs_inode_type(inode), |
00e4e6b3 | 2380 | index); |
39279cc3 | 2381 | if (ret == 0) { |
9c58309d CM |
2382 | if (add_backref) { |
2383 | ret = btrfs_insert_inode_ref(trans, root, | |
e02119d5 CM |
2384 | name, name_len, |
2385 | inode->i_ino, | |
2386 | parent_inode->i_ino, | |
2387 | index); | |
9c58309d | 2388 | } |
dbe674a9 | 2389 | btrfs_i_size_write(parent_inode, parent_inode->i_size + |
e02119d5 | 2390 | name_len * 2); |
79c44584 | 2391 | parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME; |
e02119d5 | 2392 | ret = btrfs_update_inode(trans, root, parent_inode); |
39279cc3 CM |
2393 | } |
2394 | return ret; | |
2395 | } | |
2396 | ||
2397 | static int btrfs_add_nondir(struct btrfs_trans_handle *trans, | |
9c58309d | 2398 | struct dentry *dentry, struct inode *inode, |
00e4e6b3 | 2399 | int backref, u64 index) |
39279cc3 | 2400 | { |
e02119d5 CM |
2401 | int err = btrfs_add_link(trans, dentry->d_parent->d_inode, |
2402 | inode, dentry->d_name.name, | |
2403 | dentry->d_name.len, backref, index); | |
39279cc3 CM |
2404 | if (!err) { |
2405 | d_instantiate(dentry, inode); | |
2406 | return 0; | |
2407 | } | |
2408 | if (err > 0) | |
2409 | err = -EEXIST; | |
2410 | return err; | |
2411 | } | |
2412 | ||
618e21d5 JB |
2413 | static int btrfs_mknod(struct inode *dir, struct dentry *dentry, |
2414 | int mode, dev_t rdev) | |
2415 | { | |
2416 | struct btrfs_trans_handle *trans; | |
2417 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
1832a6d5 | 2418 | struct inode *inode = NULL; |
618e21d5 JB |
2419 | int err; |
2420 | int drop_inode = 0; | |
2421 | u64 objectid; | |
1832a6d5 | 2422 | unsigned long nr = 0; |
00e4e6b3 | 2423 | u64 index = 0; |
618e21d5 JB |
2424 | |
2425 | if (!new_valid_dev(rdev)) | |
2426 | return -EINVAL; | |
2427 | ||
1832a6d5 CM |
2428 | err = btrfs_check_free_space(root, 1, 0); |
2429 | if (err) | |
2430 | goto fail; | |
2431 | ||
618e21d5 JB |
2432 | trans = btrfs_start_transaction(root, 1); |
2433 | btrfs_set_trans_block_group(trans, dir); | |
2434 | ||
2435 | err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid); | |
2436 | if (err) { | |
2437 | err = -ENOSPC; | |
2438 | goto out_unlock; | |
2439 | } | |
2440 | ||
aec7477b | 2441 | inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, |
9c58309d CM |
2442 | dentry->d_name.len, |
2443 | dentry->d_parent->d_inode->i_ino, objectid, | |
00e4e6b3 | 2444 | BTRFS_I(dir)->block_group, mode, &index); |
618e21d5 JB |
2445 | err = PTR_ERR(inode); |
2446 | if (IS_ERR(inode)) | |
2447 | goto out_unlock; | |
2448 | ||
33268eaf JB |
2449 | err = btrfs_init_acl(inode, dir); |
2450 | if (err) { | |
2451 | drop_inode = 1; | |
2452 | goto out_unlock; | |
2453 | } | |
2454 | ||
618e21d5 | 2455 | btrfs_set_trans_block_group(trans, inode); |
00e4e6b3 | 2456 | err = btrfs_add_nondir(trans, dentry, inode, 0, index); |
618e21d5 JB |
2457 | if (err) |
2458 | drop_inode = 1; | |
2459 | else { | |
2460 | inode->i_op = &btrfs_special_inode_operations; | |
2461 | init_special_inode(inode, inode->i_mode, rdev); | |
1b4ab1bb | 2462 | btrfs_update_inode(trans, root, inode); |
618e21d5 JB |
2463 | } |
2464 | dir->i_sb->s_dirt = 1; | |
2465 | btrfs_update_inode_block_group(trans, inode); | |
2466 | btrfs_update_inode_block_group(trans, dir); | |
2467 | out_unlock: | |
d3c2fdcf | 2468 | nr = trans->blocks_used; |
89ce8a63 | 2469 | btrfs_end_transaction_throttle(trans, root); |
1832a6d5 | 2470 | fail: |
618e21d5 JB |
2471 | if (drop_inode) { |
2472 | inode_dec_link_count(inode); | |
2473 | iput(inode); | |
2474 | } | |
d3c2fdcf | 2475 | btrfs_btree_balance_dirty(root, nr); |
618e21d5 JB |
2476 | return err; |
2477 | } | |
2478 | ||
39279cc3 CM |
2479 | static int btrfs_create(struct inode *dir, struct dentry *dentry, |
2480 | int mode, struct nameidata *nd) | |
2481 | { | |
2482 | struct btrfs_trans_handle *trans; | |
2483 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
1832a6d5 | 2484 | struct inode *inode = NULL; |
39279cc3 CM |
2485 | int err; |
2486 | int drop_inode = 0; | |
1832a6d5 | 2487 | unsigned long nr = 0; |
39279cc3 | 2488 | u64 objectid; |
00e4e6b3 | 2489 | u64 index = 0; |
39279cc3 | 2490 | |
1832a6d5 CM |
2491 | err = btrfs_check_free_space(root, 1, 0); |
2492 | if (err) | |
2493 | goto fail; | |
39279cc3 CM |
2494 | trans = btrfs_start_transaction(root, 1); |
2495 | btrfs_set_trans_block_group(trans, dir); | |
2496 | ||
2497 | err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid); | |
2498 | if (err) { | |
2499 | err = -ENOSPC; | |
2500 | goto out_unlock; | |
2501 | } | |
2502 | ||
aec7477b | 2503 | inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, |
9c58309d CM |
2504 | dentry->d_name.len, |
2505 | dentry->d_parent->d_inode->i_ino, | |
00e4e6b3 CM |
2506 | objectid, BTRFS_I(dir)->block_group, mode, |
2507 | &index); | |
39279cc3 CM |
2508 | err = PTR_ERR(inode); |
2509 | if (IS_ERR(inode)) | |
2510 | goto out_unlock; | |
2511 | ||
33268eaf JB |
2512 | err = btrfs_init_acl(inode, dir); |
2513 | if (err) { | |
2514 | drop_inode = 1; | |
2515 | goto out_unlock; | |
2516 | } | |
2517 | ||
39279cc3 | 2518 | btrfs_set_trans_block_group(trans, inode); |
00e4e6b3 | 2519 | err = btrfs_add_nondir(trans, dentry, inode, 0, index); |
39279cc3 CM |
2520 | if (err) |
2521 | drop_inode = 1; | |
2522 | else { | |
2523 | inode->i_mapping->a_ops = &btrfs_aops; | |
04160088 | 2524 | inode->i_mapping->backing_dev_info = &root->fs_info->bdi; |
39279cc3 CM |
2525 | inode->i_fop = &btrfs_file_operations; |
2526 | inode->i_op = &btrfs_file_inode_operations; | |
d1310b2e | 2527 | BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; |
39279cc3 CM |
2528 | } |
2529 | dir->i_sb->s_dirt = 1; | |
2530 | btrfs_update_inode_block_group(trans, inode); | |
2531 | btrfs_update_inode_block_group(trans, dir); | |
2532 | out_unlock: | |
d3c2fdcf | 2533 | nr = trans->blocks_used; |
ab78c84d | 2534 | btrfs_end_transaction_throttle(trans, root); |
1832a6d5 | 2535 | fail: |
39279cc3 CM |
2536 | if (drop_inode) { |
2537 | inode_dec_link_count(inode); | |
2538 | iput(inode); | |
2539 | } | |
d3c2fdcf | 2540 | btrfs_btree_balance_dirty(root, nr); |
39279cc3 CM |
2541 | return err; |
2542 | } | |
2543 | ||
2544 | static int btrfs_link(struct dentry *old_dentry, struct inode *dir, | |
2545 | struct dentry *dentry) | |
2546 | { | |
2547 | struct btrfs_trans_handle *trans; | |
2548 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
2549 | struct inode *inode = old_dentry->d_inode; | |
00e4e6b3 | 2550 | u64 index; |
1832a6d5 | 2551 | unsigned long nr = 0; |
39279cc3 CM |
2552 | int err; |
2553 | int drop_inode = 0; | |
2554 | ||
2555 | if (inode->i_nlink == 0) | |
2556 | return -ENOENT; | |
2557 | ||
e02119d5 | 2558 | btrfs_inc_nlink(inode); |
1832a6d5 CM |
2559 | err = btrfs_check_free_space(root, 1, 0); |
2560 | if (err) | |
2561 | goto fail; | |
00e4e6b3 | 2562 | err = btrfs_set_inode_index(dir, inode, &index); |
aec7477b JB |
2563 | if (err) |
2564 | goto fail; | |
2565 | ||
39279cc3 | 2566 | trans = btrfs_start_transaction(root, 1); |
5f39d397 | 2567 | |
39279cc3 CM |
2568 | btrfs_set_trans_block_group(trans, dir); |
2569 | atomic_inc(&inode->i_count); | |
aec7477b | 2570 | |
00e4e6b3 | 2571 | err = btrfs_add_nondir(trans, dentry, inode, 1, index); |
5f39d397 | 2572 | |
39279cc3 CM |
2573 | if (err) |
2574 | drop_inode = 1; | |
5f39d397 | 2575 | |
39279cc3 CM |
2576 | dir->i_sb->s_dirt = 1; |
2577 | btrfs_update_inode_block_group(trans, dir); | |
54aa1f4d | 2578 | err = btrfs_update_inode(trans, root, inode); |
5f39d397 | 2579 | |
54aa1f4d CM |
2580 | if (err) |
2581 | drop_inode = 1; | |
39279cc3 | 2582 | |
d3c2fdcf | 2583 | nr = trans->blocks_used; |
ab78c84d | 2584 | btrfs_end_transaction_throttle(trans, root); |
1832a6d5 | 2585 | fail: |
39279cc3 CM |
2586 | if (drop_inode) { |
2587 | inode_dec_link_count(inode); | |
2588 | iput(inode); | |
2589 | } | |
d3c2fdcf | 2590 | btrfs_btree_balance_dirty(root, nr); |
39279cc3 CM |
2591 | return err; |
2592 | } | |
2593 | ||
39279cc3 CM |
2594 | static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) |
2595 | { | |
b9d86667 | 2596 | struct inode *inode = NULL; |
39279cc3 CM |
2597 | struct btrfs_trans_handle *trans; |
2598 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
2599 | int err = 0; | |
2600 | int drop_on_err = 0; | |
b9d86667 | 2601 | u64 objectid = 0; |
00e4e6b3 | 2602 | u64 index = 0; |
d3c2fdcf | 2603 | unsigned long nr = 1; |
39279cc3 | 2604 | |
1832a6d5 CM |
2605 | err = btrfs_check_free_space(root, 1, 0); |
2606 | if (err) | |
2607 | goto out_unlock; | |
2608 | ||
39279cc3 CM |
2609 | trans = btrfs_start_transaction(root, 1); |
2610 | btrfs_set_trans_block_group(trans, dir); | |
5f39d397 | 2611 | |
39279cc3 CM |
2612 | if (IS_ERR(trans)) { |
2613 | err = PTR_ERR(trans); | |
2614 | goto out_unlock; | |
2615 | } | |
2616 | ||
2617 | err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid); | |
2618 | if (err) { | |
2619 | err = -ENOSPC; | |
2620 | goto out_unlock; | |
2621 | } | |
2622 | ||
aec7477b | 2623 | inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, |
9c58309d CM |
2624 | dentry->d_name.len, |
2625 | dentry->d_parent->d_inode->i_ino, objectid, | |
00e4e6b3 CM |
2626 | BTRFS_I(dir)->block_group, S_IFDIR | mode, |
2627 | &index); | |
39279cc3 CM |
2628 | if (IS_ERR(inode)) { |
2629 | err = PTR_ERR(inode); | |
2630 | goto out_fail; | |
2631 | } | |
5f39d397 | 2632 | |
39279cc3 | 2633 | drop_on_err = 1; |
33268eaf JB |
2634 | |
2635 | err = btrfs_init_acl(inode, dir); | |
2636 | if (err) | |
2637 | goto out_fail; | |
2638 | ||
39279cc3 CM |
2639 | inode->i_op = &btrfs_dir_inode_operations; |
2640 | inode->i_fop = &btrfs_dir_file_operations; | |
2641 | btrfs_set_trans_block_group(trans, inode); | |
2642 | ||
dbe674a9 | 2643 | btrfs_i_size_write(inode, 0); |
39279cc3 CM |
2644 | err = btrfs_update_inode(trans, root, inode); |
2645 | if (err) | |
2646 | goto out_fail; | |
5f39d397 | 2647 | |
e02119d5 CM |
2648 | err = btrfs_add_link(trans, dentry->d_parent->d_inode, |
2649 | inode, dentry->d_name.name, | |
2650 | dentry->d_name.len, 0, index); | |
39279cc3 CM |
2651 | if (err) |
2652 | goto out_fail; | |
5f39d397 | 2653 | |
39279cc3 CM |
2654 | d_instantiate(dentry, inode); |
2655 | drop_on_err = 0; | |
2656 | dir->i_sb->s_dirt = 1; | |
2657 | btrfs_update_inode_block_group(trans, inode); | |
2658 | btrfs_update_inode_block_group(trans, dir); | |
2659 | ||
2660 | out_fail: | |
d3c2fdcf | 2661 | nr = trans->blocks_used; |
ab78c84d | 2662 | btrfs_end_transaction_throttle(trans, root); |
5f39d397 | 2663 | |
39279cc3 | 2664 | out_unlock: |
39279cc3 CM |
2665 | if (drop_on_err) |
2666 | iput(inode); | |
d3c2fdcf | 2667 | btrfs_btree_balance_dirty(root, nr); |
39279cc3 CM |
2668 | return err; |
2669 | } | |
2670 | ||
3b951516 CM |
2671 | static int merge_extent_mapping(struct extent_map_tree *em_tree, |
2672 | struct extent_map *existing, | |
e6dcd2dc CM |
2673 | struct extent_map *em, |
2674 | u64 map_start, u64 map_len) | |
3b951516 CM |
2675 | { |
2676 | u64 start_diff; | |
3b951516 | 2677 | |
e6dcd2dc CM |
2678 | BUG_ON(map_start < em->start || map_start >= extent_map_end(em)); |
2679 | start_diff = map_start - em->start; | |
2680 | em->start = map_start; | |
2681 | em->len = map_len; | |
2682 | if (em->block_start < EXTENT_MAP_LAST_BYTE) | |
2683 | em->block_start += start_diff; | |
2684 | return add_extent_mapping(em_tree, em); | |
3b951516 CM |
2685 | } |
2686 | ||
a52d9a80 | 2687 | struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, |
70dec807 | 2688 | size_t pg_offset, u64 start, u64 len, |
a52d9a80 CM |
2689 | int create) |
2690 | { | |
2691 | int ret; | |
2692 | int err = 0; | |
db94535d | 2693 | u64 bytenr; |
a52d9a80 CM |
2694 | u64 extent_start = 0; |
2695 | u64 extent_end = 0; | |
2696 | u64 objectid = inode->i_ino; | |
2697 | u32 found_type; | |
f421950f | 2698 | struct btrfs_path *path = NULL; |
a52d9a80 CM |
2699 | struct btrfs_root *root = BTRFS_I(inode)->root; |
2700 | struct btrfs_file_extent_item *item; | |
5f39d397 CM |
2701 | struct extent_buffer *leaf; |
2702 | struct btrfs_key found_key; | |
a52d9a80 CM |
2703 | struct extent_map *em = NULL; |
2704 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
d1310b2e | 2705 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
a52d9a80 CM |
2706 | struct btrfs_trans_handle *trans = NULL; |
2707 | ||
a52d9a80 | 2708 | again: |
d1310b2e CM |
2709 | spin_lock(&em_tree->lock); |
2710 | em = lookup_extent_mapping(em_tree, start, len); | |
a061fc8d CM |
2711 | if (em) |
2712 | em->bdev = root->fs_info->fs_devices->latest_bdev; | |
d1310b2e CM |
2713 | spin_unlock(&em_tree->lock); |
2714 | ||
a52d9a80 | 2715 | if (em) { |
e1c4b745 CM |
2716 | if (em->start > start || em->start + em->len <= start) |
2717 | free_extent_map(em); | |
2718 | else if (em->block_start == EXTENT_MAP_INLINE && page) | |
70dec807 CM |
2719 | free_extent_map(em); |
2720 | else | |
2721 | goto out; | |
a52d9a80 | 2722 | } |
d1310b2e | 2723 | em = alloc_extent_map(GFP_NOFS); |
a52d9a80 | 2724 | if (!em) { |
d1310b2e CM |
2725 | err = -ENOMEM; |
2726 | goto out; | |
a52d9a80 | 2727 | } |
e6dcd2dc | 2728 | em->bdev = root->fs_info->fs_devices->latest_bdev; |
d1310b2e CM |
2729 | em->start = EXTENT_MAP_HOLE; |
2730 | em->len = (u64)-1; | |
f421950f CM |
2731 | |
2732 | if (!path) { | |
2733 | path = btrfs_alloc_path(); | |
2734 | BUG_ON(!path); | |
2735 | } | |
2736 | ||
179e29e4 CM |
2737 | ret = btrfs_lookup_file_extent(trans, root, path, |
2738 | objectid, start, trans != NULL); | |
a52d9a80 CM |
2739 | if (ret < 0) { |
2740 | err = ret; | |
2741 | goto out; | |
2742 | } | |
2743 | ||
2744 | if (ret != 0) { | |
2745 | if (path->slots[0] == 0) | |
2746 | goto not_found; | |
2747 | path->slots[0]--; | |
2748 | } | |
2749 | ||
5f39d397 CM |
2750 | leaf = path->nodes[0]; |
2751 | item = btrfs_item_ptr(leaf, path->slots[0], | |
a52d9a80 | 2752 | struct btrfs_file_extent_item); |
a52d9a80 | 2753 | /* are we inside the extent that was found? */ |
5f39d397 CM |
2754 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); |
2755 | found_type = btrfs_key_type(&found_key); | |
2756 | if (found_key.objectid != objectid || | |
a52d9a80 CM |
2757 | found_type != BTRFS_EXTENT_DATA_KEY) { |
2758 | goto not_found; | |
2759 | } | |
2760 | ||
5f39d397 CM |
2761 | found_type = btrfs_file_extent_type(leaf, item); |
2762 | extent_start = found_key.offset; | |
a52d9a80 CM |
2763 | if (found_type == BTRFS_FILE_EXTENT_REG) { |
2764 | extent_end = extent_start + | |
db94535d | 2765 | btrfs_file_extent_num_bytes(leaf, item); |
a52d9a80 | 2766 | err = 0; |
b888db2b | 2767 | if (start < extent_start || start >= extent_end) { |
a52d9a80 CM |
2768 | em->start = start; |
2769 | if (start < extent_start) { | |
d1310b2e | 2770 | if (start + len <= extent_start) |
b888db2b | 2771 | goto not_found; |
d1310b2e | 2772 | em->len = extent_end - extent_start; |
a52d9a80 | 2773 | } else { |
d1310b2e | 2774 | em->len = len; |
a52d9a80 CM |
2775 | } |
2776 | goto not_found_em; | |
2777 | } | |
db94535d CM |
2778 | bytenr = btrfs_file_extent_disk_bytenr(leaf, item); |
2779 | if (bytenr == 0) { | |
a52d9a80 | 2780 | em->start = extent_start; |
d1310b2e | 2781 | em->len = extent_end - extent_start; |
5f39d397 | 2782 | em->block_start = EXTENT_MAP_HOLE; |
a52d9a80 CM |
2783 | goto insert; |
2784 | } | |
db94535d CM |
2785 | bytenr += btrfs_file_extent_offset(leaf, item); |
2786 | em->block_start = bytenr; | |
a52d9a80 | 2787 | em->start = extent_start; |
d1310b2e | 2788 | em->len = extent_end - extent_start; |
a52d9a80 CM |
2789 | goto insert; |
2790 | } else if (found_type == BTRFS_FILE_EXTENT_INLINE) { | |
70dec807 | 2791 | u64 page_start; |
5f39d397 | 2792 | unsigned long ptr; |
a52d9a80 | 2793 | char *map; |
3326d1b0 CM |
2794 | size_t size; |
2795 | size_t extent_offset; | |
2796 | size_t copy_size; | |
a52d9a80 | 2797 | |
5f39d397 CM |
2798 | size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf, |
2799 | path->slots[0])); | |
d1310b2e CM |
2800 | extent_end = (extent_start + size + root->sectorsize - 1) & |
2801 | ~((u64)root->sectorsize - 1); | |
b888db2b | 2802 | if (start < extent_start || start >= extent_end) { |
a52d9a80 CM |
2803 | em->start = start; |
2804 | if (start < extent_start) { | |
d1310b2e | 2805 | if (start + len <= extent_start) |
b888db2b | 2806 | goto not_found; |
d1310b2e | 2807 | em->len = extent_end - extent_start; |
a52d9a80 | 2808 | } else { |
d1310b2e | 2809 | em->len = len; |
a52d9a80 CM |
2810 | } |
2811 | goto not_found_em; | |
2812 | } | |
689f9346 | 2813 | em->block_start = EXTENT_MAP_INLINE; |
689f9346 Y |
2814 | |
2815 | if (!page) { | |
2816 | em->start = extent_start; | |
d1310b2e | 2817 | em->len = size; |
689f9346 Y |
2818 | goto out; |
2819 | } | |
5f39d397 | 2820 | |
70dec807 CM |
2821 | page_start = page_offset(page) + pg_offset; |
2822 | extent_offset = page_start - extent_start; | |
2823 | copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset, | |
3326d1b0 | 2824 | size - extent_offset); |
3326d1b0 | 2825 | em->start = extent_start + extent_offset; |
70dec807 CM |
2826 | em->len = (copy_size + root->sectorsize - 1) & |
2827 | ~((u64)root->sectorsize - 1); | |
689f9346 Y |
2828 | map = kmap(page); |
2829 | ptr = btrfs_file_extent_inline_start(item) + extent_offset; | |
179e29e4 | 2830 | if (create == 0 && !PageUptodate(page)) { |
70dec807 | 2831 | read_extent_buffer(leaf, map + pg_offset, ptr, |
179e29e4 CM |
2832 | copy_size); |
2833 | flush_dcache_page(page); | |
2834 | } else if (create && PageUptodate(page)) { | |
2835 | if (!trans) { | |
2836 | kunmap(page); | |
2837 | free_extent_map(em); | |
2838 | em = NULL; | |
2839 | btrfs_release_path(root, path); | |
f9295749 | 2840 | trans = btrfs_join_transaction(root, 1); |
179e29e4 CM |
2841 | goto again; |
2842 | } | |
70dec807 | 2843 | write_extent_buffer(leaf, map + pg_offset, ptr, |
179e29e4 CM |
2844 | copy_size); |
2845 | btrfs_mark_buffer_dirty(leaf); | |
a52d9a80 | 2846 | } |
a52d9a80 | 2847 | kunmap(page); |
d1310b2e CM |
2848 | set_extent_uptodate(io_tree, em->start, |
2849 | extent_map_end(em) - 1, GFP_NOFS); | |
a52d9a80 CM |
2850 | goto insert; |
2851 | } else { | |
2852 | printk("unkknown found_type %d\n", found_type); | |
2853 | WARN_ON(1); | |
2854 | } | |
2855 | not_found: | |
2856 | em->start = start; | |
d1310b2e | 2857 | em->len = len; |
a52d9a80 | 2858 | not_found_em: |
5f39d397 | 2859 | em->block_start = EXTENT_MAP_HOLE; |
a52d9a80 CM |
2860 | insert: |
2861 | btrfs_release_path(root, path); | |
d1310b2e CM |
2862 | if (em->start > start || extent_map_end(em) <= start) { |
2863 | printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len); | |
a52d9a80 CM |
2864 | err = -EIO; |
2865 | goto out; | |
2866 | } | |
d1310b2e CM |
2867 | |
2868 | err = 0; | |
2869 | spin_lock(&em_tree->lock); | |
a52d9a80 | 2870 | ret = add_extent_mapping(em_tree, em); |
3b951516 CM |
2871 | /* it is possible that someone inserted the extent into the tree |
2872 | * while we had the lock dropped. It is also possible that | |
2873 | * an overlapping map exists in the tree | |
2874 | */ | |
a52d9a80 | 2875 | if (ret == -EEXIST) { |
3b951516 | 2876 | struct extent_map *existing; |
e6dcd2dc CM |
2877 | |
2878 | ret = 0; | |
2879 | ||
3b951516 | 2880 | existing = lookup_extent_mapping(em_tree, start, len); |
e1c4b745 CM |
2881 | if (existing && (existing->start > start || |
2882 | existing->start + existing->len <= start)) { | |
2883 | free_extent_map(existing); | |
2884 | existing = NULL; | |
2885 | } | |
3b951516 CM |
2886 | if (!existing) { |
2887 | existing = lookup_extent_mapping(em_tree, em->start, | |
2888 | em->len); | |
2889 | if (existing) { | |
2890 | err = merge_extent_mapping(em_tree, existing, | |
e6dcd2dc CM |
2891 | em, start, |
2892 | root->sectorsize); | |
3b951516 CM |
2893 | free_extent_map(existing); |
2894 | if (err) { | |
2895 | free_extent_map(em); | |
2896 | em = NULL; | |
2897 | } | |
2898 | } else { | |
2899 | err = -EIO; | |
2900 | printk("failing to insert %Lu %Lu\n", | |
2901 | start, len); | |
2902 | free_extent_map(em); | |
2903 | em = NULL; | |
2904 | } | |
2905 | } else { | |
2906 | free_extent_map(em); | |
2907 | em = existing; | |
e6dcd2dc | 2908 | err = 0; |
a52d9a80 | 2909 | } |
a52d9a80 | 2910 | } |
d1310b2e | 2911 | spin_unlock(&em_tree->lock); |
a52d9a80 | 2912 | out: |
f421950f CM |
2913 | if (path) |
2914 | btrfs_free_path(path); | |
a52d9a80 CM |
2915 | if (trans) { |
2916 | ret = btrfs_end_transaction(trans, root); | |
e6dcd2dc | 2917 | if (!err) { |
a52d9a80 | 2918 | err = ret; |
e6dcd2dc | 2919 | } |
a52d9a80 | 2920 | } |
a52d9a80 CM |
2921 | if (err) { |
2922 | free_extent_map(em); | |
2923 | WARN_ON(1); | |
2924 | return ERR_PTR(err); | |
2925 | } | |
2926 | return em; | |
2927 | } | |
2928 | ||
e1c4b745 | 2929 | #if 0 /* waiting for O_DIRECT reads */ |
16432985 CM |
2930 | static int btrfs_get_block(struct inode *inode, sector_t iblock, |
2931 | struct buffer_head *bh_result, int create) | |
2932 | { | |
2933 | struct extent_map *em; | |
2934 | u64 start = (u64)iblock << inode->i_blkbits; | |
2935 | struct btrfs_multi_bio *multi = NULL; | |
2936 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
2937 | u64 len; | |
2938 | u64 logical; | |
2939 | u64 map_length; | |
2940 | int ret = 0; | |
2941 | ||
2942 | em = btrfs_get_extent(inode, NULL, 0, start, bh_result->b_size, 0); | |
2943 | ||
2944 | if (!em || IS_ERR(em)) | |
2945 | goto out; | |
2946 | ||
e1c4b745 | 2947 | if (em->start > start || em->start + em->len <= start) { |
16432985 | 2948 | goto out; |
e1c4b745 | 2949 | } |
16432985 CM |
2950 | |
2951 | if (em->block_start == EXTENT_MAP_INLINE) { | |
2952 | ret = -EINVAL; | |
2953 | goto out; | |
2954 | } | |
2955 | ||
e1c4b745 CM |
2956 | len = em->start + em->len - start; |
2957 | len = min_t(u64, len, INT_LIMIT(typeof(bh_result->b_size))); | |
2958 | ||
16432985 CM |
2959 | if (em->block_start == EXTENT_MAP_HOLE || |
2960 | em->block_start == EXTENT_MAP_DELALLOC) { | |
e1c4b745 | 2961 | bh_result->b_size = len; |
16432985 CM |
2962 | goto out; |
2963 | } | |
2964 | ||
16432985 CM |
2965 | logical = start - em->start; |
2966 | logical = em->block_start + logical; | |
2967 | ||
2968 | map_length = len; | |
2969 | ret = btrfs_map_block(&root->fs_info->mapping_tree, READ, | |
2970 | logical, &map_length, &multi, 0); | |
2971 | BUG_ON(ret); | |
2972 | bh_result->b_blocknr = multi->stripes[0].physical >> inode->i_blkbits; | |
2973 | bh_result->b_size = min(map_length, len); | |
e1c4b745 | 2974 | |
16432985 CM |
2975 | bh_result->b_bdev = multi->stripes[0].dev->bdev; |
2976 | set_buffer_mapped(bh_result); | |
2977 | kfree(multi); | |
2978 | out: | |
2979 | free_extent_map(em); | |
2980 | return ret; | |
2981 | } | |
e1c4b745 | 2982 | #endif |
16432985 CM |
2983 | |
2984 | static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb, | |
2985 | const struct iovec *iov, loff_t offset, | |
2986 | unsigned long nr_segs) | |
2987 | { | |
e1c4b745 CM |
2988 | return -EINVAL; |
2989 | #if 0 | |
16432985 CM |
2990 | struct file *file = iocb->ki_filp; |
2991 | struct inode *inode = file->f_mapping->host; | |
2992 | ||
2993 | if (rw == WRITE) | |
2994 | return -EINVAL; | |
2995 | ||
2996 | return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov, | |
2997 | offset, nr_segs, btrfs_get_block, NULL); | |
e1c4b745 | 2998 | #endif |
16432985 CM |
2999 | } |
3000 | ||
d396c6f5 | 3001 | static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock) |
39279cc3 | 3002 | { |
d396c6f5 | 3003 | return extent_bmap(mapping, iblock, btrfs_get_extent); |
39279cc3 CM |
3004 | } |
3005 | ||
a52d9a80 | 3006 | int btrfs_readpage(struct file *file, struct page *page) |
9ebefb18 | 3007 | { |
d1310b2e CM |
3008 | struct extent_io_tree *tree; |
3009 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
a52d9a80 | 3010 | return extent_read_full_page(tree, page, btrfs_get_extent); |
9ebefb18 | 3011 | } |
1832a6d5 | 3012 | |
a52d9a80 | 3013 | static int btrfs_writepage(struct page *page, struct writeback_control *wbc) |
39279cc3 | 3014 | { |
d1310b2e | 3015 | struct extent_io_tree *tree; |
b888db2b CM |
3016 | |
3017 | ||
3018 | if (current->flags & PF_MEMALLOC) { | |
3019 | redirty_page_for_writepage(wbc, page); | |
3020 | unlock_page(page); | |
3021 | return 0; | |
3022 | } | |
d1310b2e | 3023 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
a52d9a80 | 3024 | return extent_write_full_page(tree, page, btrfs_get_extent, wbc); |
9ebefb18 CM |
3025 | } |
3026 | ||
f421950f CM |
3027 | int btrfs_writepages(struct address_space *mapping, |
3028 | struct writeback_control *wbc) | |
b293f02e | 3029 | { |
d1310b2e CM |
3030 | struct extent_io_tree *tree; |
3031 | tree = &BTRFS_I(mapping->host)->io_tree; | |
b293f02e CM |
3032 | return extent_writepages(tree, mapping, btrfs_get_extent, wbc); |
3033 | } | |
3034 | ||
3ab2fb5a CM |
3035 | static int |
3036 | btrfs_readpages(struct file *file, struct address_space *mapping, | |
3037 | struct list_head *pages, unsigned nr_pages) | |
3038 | { | |
d1310b2e CM |
3039 | struct extent_io_tree *tree; |
3040 | tree = &BTRFS_I(mapping->host)->io_tree; | |
3ab2fb5a CM |
3041 | return extent_readpages(tree, mapping, pages, nr_pages, |
3042 | btrfs_get_extent); | |
3043 | } | |
e6dcd2dc | 3044 | static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags) |
9ebefb18 | 3045 | { |
d1310b2e CM |
3046 | struct extent_io_tree *tree; |
3047 | struct extent_map_tree *map; | |
a52d9a80 | 3048 | int ret; |
8c2383c3 | 3049 | |
d1310b2e CM |
3050 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
3051 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
70dec807 | 3052 | ret = try_release_extent_mapping(map, tree, page, gfp_flags); |
a52d9a80 CM |
3053 | if (ret == 1) { |
3054 | ClearPagePrivate(page); | |
3055 | set_page_private(page, 0); | |
3056 | page_cache_release(page); | |
39279cc3 | 3057 | } |
a52d9a80 | 3058 | return ret; |
39279cc3 CM |
3059 | } |
3060 | ||
e6dcd2dc CM |
3061 | static int btrfs_releasepage(struct page *page, gfp_t gfp_flags) |
3062 | { | |
98509cfc CM |
3063 | if (PageWriteback(page) || PageDirty(page)) |
3064 | return 0; | |
e6dcd2dc CM |
3065 | return __btrfs_releasepage(page, gfp_flags); |
3066 | } | |
3067 | ||
a52d9a80 | 3068 | static void btrfs_invalidatepage(struct page *page, unsigned long offset) |
39279cc3 | 3069 | { |
d1310b2e | 3070 | struct extent_io_tree *tree; |
e6dcd2dc CM |
3071 | struct btrfs_ordered_extent *ordered; |
3072 | u64 page_start = page_offset(page); | |
3073 | u64 page_end = page_start + PAGE_CACHE_SIZE - 1; | |
39279cc3 | 3074 | |
e6dcd2dc | 3075 | wait_on_page_writeback(page); |
d1310b2e | 3076 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
e6dcd2dc CM |
3077 | if (offset) { |
3078 | btrfs_releasepage(page, GFP_NOFS); | |
3079 | return; | |
3080 | } | |
3081 | ||
3082 | lock_extent(tree, page_start, page_end, GFP_NOFS); | |
3083 | ordered = btrfs_lookup_ordered_extent(page->mapping->host, | |
3084 | page_offset(page)); | |
3085 | if (ordered) { | |
eb84ae03 CM |
3086 | /* |
3087 | * IO on this page will never be started, so we need | |
3088 | * to account for any ordered extents now | |
3089 | */ | |
e6dcd2dc CM |
3090 | clear_extent_bit(tree, page_start, page_end, |
3091 | EXTENT_DIRTY | EXTENT_DELALLOC | | |
3092 | EXTENT_LOCKED, 1, 0, GFP_NOFS); | |
211f90e6 CM |
3093 | btrfs_finish_ordered_io(page->mapping->host, |
3094 | page_start, page_end); | |
e6dcd2dc CM |
3095 | btrfs_put_ordered_extent(ordered); |
3096 | lock_extent(tree, page_start, page_end, GFP_NOFS); | |
3097 | } | |
3098 | clear_extent_bit(tree, page_start, page_end, | |
3099 | EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC | | |
3100 | EXTENT_ORDERED, | |
3101 | 1, 1, GFP_NOFS); | |
3102 | __btrfs_releasepage(page, GFP_NOFS); | |
3103 | ||
4a096752 | 3104 | ClearPageChecked(page); |
9ad6b7bc | 3105 | if (PagePrivate(page)) { |
9ad6b7bc CM |
3106 | ClearPagePrivate(page); |
3107 | set_page_private(page, 0); | |
3108 | page_cache_release(page); | |
3109 | } | |
39279cc3 CM |
3110 | } |
3111 | ||
9ebefb18 CM |
3112 | /* |
3113 | * btrfs_page_mkwrite() is not allowed to change the file size as it gets | |
3114 | * called from a page fault handler when a page is first dirtied. Hence we must | |
3115 | * be careful to check for EOF conditions here. We set the page up correctly | |
3116 | * for a written page which means we get ENOSPC checking when writing into | |
3117 | * holes and correct delalloc and unwritten extent mapping on filesystems that | |
3118 | * support these features. | |
3119 | * | |
3120 | * We are not allowed to take the i_mutex here so we have to play games to | |
3121 | * protect against truncate races as the page could now be beyond EOF. Because | |
3122 | * vmtruncate() writes the inode size before removing pages, once we have the | |
3123 | * page lock we can determine safely if the page is beyond EOF. If it is not | |
3124 | * beyond EOF, then the page is guaranteed safe against truncation until we | |
3125 | * unlock the page. | |
3126 | */ | |
3127 | int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page) | |
3128 | { | |
6da6abae | 3129 | struct inode *inode = fdentry(vma->vm_file)->d_inode; |
1832a6d5 | 3130 | struct btrfs_root *root = BTRFS_I(inode)->root; |
e6dcd2dc CM |
3131 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
3132 | struct btrfs_ordered_extent *ordered; | |
3133 | char *kaddr; | |
3134 | unsigned long zero_start; | |
9ebefb18 | 3135 | loff_t size; |
1832a6d5 | 3136 | int ret; |
a52d9a80 | 3137 | u64 page_start; |
e6dcd2dc | 3138 | u64 page_end; |
9ebefb18 | 3139 | |
1832a6d5 | 3140 | ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0); |
1832a6d5 CM |
3141 | if (ret) |
3142 | goto out; | |
3143 | ||
3144 | ret = -EINVAL; | |
e6dcd2dc | 3145 | again: |
9ebefb18 | 3146 | lock_page(page); |
9ebefb18 | 3147 | size = i_size_read(inode); |
e6dcd2dc CM |
3148 | page_start = page_offset(page); |
3149 | page_end = page_start + PAGE_CACHE_SIZE - 1; | |
a52d9a80 | 3150 | |
9ebefb18 | 3151 | if ((page->mapping != inode->i_mapping) || |
e6dcd2dc | 3152 | (page_start >= size)) { |
9ebefb18 CM |
3153 | /* page got truncated out from underneath us */ |
3154 | goto out_unlock; | |
3155 | } | |
e6dcd2dc CM |
3156 | wait_on_page_writeback(page); |
3157 | ||
3158 | lock_extent(io_tree, page_start, page_end, GFP_NOFS); | |
3159 | set_page_extent_mapped(page); | |
3160 | ||
eb84ae03 CM |
3161 | /* |
3162 | * we can't set the delalloc bits if there are pending ordered | |
3163 | * extents. Drop our locks and wait for them to finish | |
3164 | */ | |
e6dcd2dc CM |
3165 | ordered = btrfs_lookup_ordered_extent(inode, page_start); |
3166 | if (ordered) { | |
3167 | unlock_extent(io_tree, page_start, page_end, GFP_NOFS); | |
3168 | unlock_page(page); | |
eb84ae03 | 3169 | btrfs_start_ordered_extent(inode, ordered, 1); |
e6dcd2dc CM |
3170 | btrfs_put_ordered_extent(ordered); |
3171 | goto again; | |
3172 | } | |
3173 | ||
ea8c2819 | 3174 | btrfs_set_extent_delalloc(inode, page_start, page_end); |
e6dcd2dc | 3175 | ret = 0; |
9ebefb18 CM |
3176 | |
3177 | /* page is wholly or partially inside EOF */ | |
a52d9a80 | 3178 | if (page_start + PAGE_CACHE_SIZE > size) |
e6dcd2dc | 3179 | zero_start = size & ~PAGE_CACHE_MASK; |
9ebefb18 | 3180 | else |
e6dcd2dc | 3181 | zero_start = PAGE_CACHE_SIZE; |
9ebefb18 | 3182 | |
e6dcd2dc CM |
3183 | if (zero_start != PAGE_CACHE_SIZE) { |
3184 | kaddr = kmap(page); | |
3185 | memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start); | |
3186 | flush_dcache_page(page); | |
3187 | kunmap(page); | |
3188 | } | |
247e743c | 3189 | ClearPageChecked(page); |
e6dcd2dc CM |
3190 | set_page_dirty(page); |
3191 | unlock_extent(io_tree, page_start, page_end, GFP_NOFS); | |
9ebefb18 CM |
3192 | |
3193 | out_unlock: | |
3194 | unlock_page(page); | |
1832a6d5 | 3195 | out: |
9ebefb18 CM |
3196 | return ret; |
3197 | } | |
3198 | ||
39279cc3 CM |
3199 | static void btrfs_truncate(struct inode *inode) |
3200 | { | |
3201 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
3202 | int ret; | |
3203 | struct btrfs_trans_handle *trans; | |
d3c2fdcf | 3204 | unsigned long nr; |
dbe674a9 | 3205 | u64 mask = root->sectorsize - 1; |
39279cc3 CM |
3206 | |
3207 | if (!S_ISREG(inode->i_mode)) | |
3208 | return; | |
3209 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) | |
3210 | return; | |
3211 | ||
3212 | btrfs_truncate_page(inode->i_mapping, inode->i_size); | |
4a096752 | 3213 | btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1); |
39279cc3 | 3214 | |
39279cc3 CM |
3215 | trans = btrfs_start_transaction(root, 1); |
3216 | btrfs_set_trans_block_group(trans, inode); | |
dbe674a9 | 3217 | btrfs_i_size_write(inode, inode->i_size); |
39279cc3 | 3218 | |
7b128766 JB |
3219 | ret = btrfs_orphan_add(trans, inode); |
3220 | if (ret) | |
3221 | goto out; | |
39279cc3 | 3222 | /* FIXME, add redo link to tree so we don't leak on crash */ |
e02119d5 | 3223 | ret = btrfs_truncate_inode_items(trans, root, inode, inode->i_size, |
85e21bac | 3224 | BTRFS_EXTENT_DATA_KEY); |
39279cc3 | 3225 | btrfs_update_inode(trans, root, inode); |
5f39d397 | 3226 | |
7b128766 JB |
3227 | ret = btrfs_orphan_del(trans, inode); |
3228 | BUG_ON(ret); | |
3229 | ||
3230 | out: | |
3231 | nr = trans->blocks_used; | |
89ce8a63 | 3232 | ret = btrfs_end_transaction_throttle(trans, root); |
39279cc3 | 3233 | BUG_ON(ret); |
d3c2fdcf | 3234 | btrfs_btree_balance_dirty(root, nr); |
39279cc3 CM |
3235 | } |
3236 | ||
3b96362c SW |
3237 | /* |
3238 | * Invalidate a single dcache entry at the root of the filesystem. | |
3239 | * Needed after creation of snapshot or subvolume. | |
3240 | */ | |
3241 | void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name, | |
3242 | int namelen) | |
3243 | { | |
3244 | struct dentry *alias, *entry; | |
3245 | struct qstr qstr; | |
3246 | ||
3247 | alias = d_find_alias(root->fs_info->sb->s_root->d_inode); | |
3248 | if (alias) { | |
3249 | qstr.name = name; | |
3250 | qstr.len = namelen; | |
3251 | /* change me if btrfs ever gets a d_hash operation */ | |
3252 | qstr.hash = full_name_hash(qstr.name, qstr.len); | |
3253 | entry = d_lookup(alias, &qstr); | |
3254 | dput(alias); | |
3255 | if (entry) { | |
3256 | d_invalidate(entry); | |
3257 | dput(entry); | |
3258 | } | |
3259 | } | |
3260 | } | |
3261 | ||
f46b5a66 CH |
3262 | int btrfs_create_subvol_root(struct btrfs_root *new_root, |
3263 | struct btrfs_trans_handle *trans, u64 new_dirid, | |
3264 | struct btrfs_block_group_cache *block_group) | |
39279cc3 | 3265 | { |
39279cc3 | 3266 | struct inode *inode; |
00e4e6b3 | 3267 | u64 index = 0; |
39279cc3 | 3268 | |
aec7477b | 3269 | inode = btrfs_new_inode(trans, new_root, NULL, "..", 2, new_dirid, |
00e4e6b3 | 3270 | new_dirid, block_group, S_IFDIR | 0700, &index); |
54aa1f4d | 3271 | if (IS_ERR(inode)) |
f46b5a66 | 3272 | return PTR_ERR(inode); |
39279cc3 CM |
3273 | inode->i_op = &btrfs_dir_inode_operations; |
3274 | inode->i_fop = &btrfs_dir_file_operations; | |
34088780 | 3275 | new_root->inode = inode; |
39279cc3 | 3276 | |
39279cc3 | 3277 | inode->i_nlink = 1; |
dbe674a9 | 3278 | btrfs_i_size_write(inode, 0); |
3b96362c | 3279 | |
f46b5a66 | 3280 | return btrfs_update_inode(trans, new_root, inode); |
39279cc3 CM |
3281 | } |
3282 | ||
edbd8d4e | 3283 | unsigned long btrfs_force_ra(struct address_space *mapping, |
86479a04 CM |
3284 | struct file_ra_state *ra, struct file *file, |
3285 | pgoff_t offset, pgoff_t last_index) | |
3286 | { | |
8e7bf94f | 3287 | pgoff_t req_size = last_index - offset + 1; |
86479a04 CM |
3288 | |
3289 | #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) | |
86479a04 CM |
3290 | offset = page_cache_readahead(mapping, ra, file, offset, req_size); |
3291 | return offset; | |
3292 | #else | |
86479a04 CM |
3293 | page_cache_sync_readahead(mapping, ra, file, offset, req_size); |
3294 | return offset + req_size; | |
3295 | #endif | |
3296 | } | |
3297 | ||
39279cc3 CM |
3298 | struct inode *btrfs_alloc_inode(struct super_block *sb) |
3299 | { | |
3300 | struct btrfs_inode *ei; | |
3301 | ||
3302 | ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS); | |
3303 | if (!ei) | |
3304 | return NULL; | |
15ee9bc7 | 3305 | ei->last_trans = 0; |
e02119d5 | 3306 | ei->logged_trans = 0; |
e6dcd2dc | 3307 | btrfs_ordered_inode_tree_init(&ei->ordered_tree); |
33268eaf JB |
3308 | ei->i_acl = BTRFS_ACL_NOT_CACHED; |
3309 | ei->i_default_acl = BTRFS_ACL_NOT_CACHED; | |
7b128766 | 3310 | INIT_LIST_HEAD(&ei->i_orphan); |
39279cc3 CM |
3311 | return &ei->vfs_inode; |
3312 | } | |
3313 | ||
3314 | void btrfs_destroy_inode(struct inode *inode) | |
3315 | { | |
e6dcd2dc | 3316 | struct btrfs_ordered_extent *ordered; |
39279cc3 CM |
3317 | WARN_ON(!list_empty(&inode->i_dentry)); |
3318 | WARN_ON(inode->i_data.nrpages); | |
3319 | ||
33268eaf JB |
3320 | if (BTRFS_I(inode)->i_acl && |
3321 | BTRFS_I(inode)->i_acl != BTRFS_ACL_NOT_CACHED) | |
3322 | posix_acl_release(BTRFS_I(inode)->i_acl); | |
3323 | if (BTRFS_I(inode)->i_default_acl && | |
3324 | BTRFS_I(inode)->i_default_acl != BTRFS_ACL_NOT_CACHED) | |
3325 | posix_acl_release(BTRFS_I(inode)->i_default_acl); | |
3326 | ||
bcc63abb | 3327 | spin_lock(&BTRFS_I(inode)->root->list_lock); |
7b128766 JB |
3328 | if (!list_empty(&BTRFS_I(inode)->i_orphan)) { |
3329 | printk(KERN_ERR "BTRFS: inode %lu: inode still on the orphan" | |
3330 | " list\n", inode->i_ino); | |
3331 | dump_stack(); | |
3332 | } | |
bcc63abb | 3333 | spin_unlock(&BTRFS_I(inode)->root->list_lock); |
7b128766 | 3334 | |
e6dcd2dc CM |
3335 | while(1) { |
3336 | ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1); | |
3337 | if (!ordered) | |
3338 | break; | |
3339 | else { | |
3340 | printk("found ordered extent %Lu %Lu\n", | |
3341 | ordered->file_offset, ordered->len); | |
3342 | btrfs_remove_ordered_extent(inode, ordered); | |
3343 | btrfs_put_ordered_extent(ordered); | |
3344 | btrfs_put_ordered_extent(ordered); | |
3345 | } | |
3346 | } | |
8c416c9e | 3347 | btrfs_drop_extent_cache(inode, 0, (u64)-1); |
39279cc3 CM |
3348 | kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode)); |
3349 | } | |
3350 | ||
0ee0fda0 SW |
3351 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) |
3352 | static void init_once(void *foo) | |
3353 | #elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) | |
44ec0b71 CM |
3354 | static void init_once(struct kmem_cache * cachep, void *foo) |
3355 | #else | |
39279cc3 CM |
3356 | static void init_once(void * foo, struct kmem_cache * cachep, |
3357 | unsigned long flags) | |
44ec0b71 | 3358 | #endif |
39279cc3 CM |
3359 | { |
3360 | struct btrfs_inode *ei = (struct btrfs_inode *) foo; | |
3361 | ||
3362 | inode_init_once(&ei->vfs_inode); | |
3363 | } | |
3364 | ||
3365 | void btrfs_destroy_cachep(void) | |
3366 | { | |
3367 | if (btrfs_inode_cachep) | |
3368 | kmem_cache_destroy(btrfs_inode_cachep); | |
3369 | if (btrfs_trans_handle_cachep) | |
3370 | kmem_cache_destroy(btrfs_trans_handle_cachep); | |
3371 | if (btrfs_transaction_cachep) | |
3372 | kmem_cache_destroy(btrfs_transaction_cachep); | |
3373 | if (btrfs_bit_radix_cachep) | |
3374 | kmem_cache_destroy(btrfs_bit_radix_cachep); | |
3375 | if (btrfs_path_cachep) | |
3376 | kmem_cache_destroy(btrfs_path_cachep); | |
3377 | } | |
3378 | ||
86479a04 | 3379 | struct kmem_cache *btrfs_cache_create(const char *name, size_t size, |
92fee66d | 3380 | unsigned long extra_flags, |
0ee0fda0 SW |
3381 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) |
3382 | void (*ctor)(void *) | |
3383 | #elif LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) | |
44ec0b71 CM |
3384 | void (*ctor)(struct kmem_cache *, void *) |
3385 | #else | |
92fee66d | 3386 | void (*ctor)(void *, struct kmem_cache *, |
44ec0b71 CM |
3387 | unsigned long) |
3388 | #endif | |
3389 | ) | |
92fee66d CM |
3390 | { |
3391 | return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT | | |
3392 | SLAB_MEM_SPREAD | extra_flags), ctor | |
3393 | #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23) | |
3394 | ,NULL | |
3395 | #endif | |
3396 | ); | |
3397 | } | |
3398 | ||
39279cc3 CM |
3399 | int btrfs_init_cachep(void) |
3400 | { | |
86479a04 | 3401 | btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache", |
92fee66d CM |
3402 | sizeof(struct btrfs_inode), |
3403 | 0, init_once); | |
39279cc3 CM |
3404 | if (!btrfs_inode_cachep) |
3405 | goto fail; | |
86479a04 CM |
3406 | btrfs_trans_handle_cachep = |
3407 | btrfs_cache_create("btrfs_trans_handle_cache", | |
3408 | sizeof(struct btrfs_trans_handle), | |
3409 | 0, NULL); | |
39279cc3 CM |
3410 | if (!btrfs_trans_handle_cachep) |
3411 | goto fail; | |
86479a04 | 3412 | btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache", |
39279cc3 | 3413 | sizeof(struct btrfs_transaction), |
92fee66d | 3414 | 0, NULL); |
39279cc3 CM |
3415 | if (!btrfs_transaction_cachep) |
3416 | goto fail; | |
86479a04 | 3417 | btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache", |
23223584 | 3418 | sizeof(struct btrfs_path), |
92fee66d | 3419 | 0, NULL); |
39279cc3 CM |
3420 | if (!btrfs_path_cachep) |
3421 | goto fail; | |
86479a04 | 3422 | btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256, |
92fee66d | 3423 | SLAB_DESTROY_BY_RCU, NULL); |
39279cc3 CM |
3424 | if (!btrfs_bit_radix_cachep) |
3425 | goto fail; | |
3426 | return 0; | |
3427 | fail: | |
3428 | btrfs_destroy_cachep(); | |
3429 | return -ENOMEM; | |
3430 | } | |
3431 | ||
3432 | static int btrfs_getattr(struct vfsmount *mnt, | |
3433 | struct dentry *dentry, struct kstat *stat) | |
3434 | { | |
3435 | struct inode *inode = dentry->d_inode; | |
3436 | generic_fillattr(inode, stat); | |
d6667462 | 3437 | stat->blksize = PAGE_CACHE_SIZE; |
9069218d | 3438 | stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9); |
39279cc3 CM |
3439 | return 0; |
3440 | } | |
3441 | ||
3442 | static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry, | |
3443 | struct inode * new_dir,struct dentry *new_dentry) | |
3444 | { | |
3445 | struct btrfs_trans_handle *trans; | |
3446 | struct btrfs_root *root = BTRFS_I(old_dir)->root; | |
3447 | struct inode *new_inode = new_dentry->d_inode; | |
3448 | struct inode *old_inode = old_dentry->d_inode; | |
3449 | struct timespec ctime = CURRENT_TIME; | |
00e4e6b3 | 3450 | u64 index = 0; |
39279cc3 CM |
3451 | int ret; |
3452 | ||
3453 | if (S_ISDIR(old_inode->i_mode) && new_inode && | |
3454 | new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) { | |
3455 | return -ENOTEMPTY; | |
3456 | } | |
5f39d397 | 3457 | |
1832a6d5 CM |
3458 | ret = btrfs_check_free_space(root, 1, 0); |
3459 | if (ret) | |
3460 | goto out_unlock; | |
3461 | ||
39279cc3 | 3462 | trans = btrfs_start_transaction(root, 1); |
5f39d397 | 3463 | |
39279cc3 | 3464 | btrfs_set_trans_block_group(trans, new_dir); |
39279cc3 | 3465 | |
e02119d5 | 3466 | btrfs_inc_nlink(old_dentry->d_inode); |
39279cc3 CM |
3467 | old_dir->i_ctime = old_dir->i_mtime = ctime; |
3468 | new_dir->i_ctime = new_dir->i_mtime = ctime; | |
3469 | old_inode->i_ctime = ctime; | |
5f39d397 | 3470 | |
e02119d5 CM |
3471 | ret = btrfs_unlink_inode(trans, root, old_dir, old_dentry->d_inode, |
3472 | old_dentry->d_name.name, | |
3473 | old_dentry->d_name.len); | |
39279cc3 CM |
3474 | if (ret) |
3475 | goto out_fail; | |
3476 | ||
3477 | if (new_inode) { | |
3478 | new_inode->i_ctime = CURRENT_TIME; | |
e02119d5 CM |
3479 | ret = btrfs_unlink_inode(trans, root, new_dir, |
3480 | new_dentry->d_inode, | |
3481 | new_dentry->d_name.name, | |
3482 | new_dentry->d_name.len); | |
39279cc3 CM |
3483 | if (ret) |
3484 | goto out_fail; | |
7b128766 | 3485 | if (new_inode->i_nlink == 0) { |
e02119d5 | 3486 | ret = btrfs_orphan_add(trans, new_dentry->d_inode); |
7b128766 JB |
3487 | if (ret) |
3488 | goto out_fail; | |
3489 | } | |
e02119d5 | 3490 | |
39279cc3 | 3491 | } |
00e4e6b3 | 3492 | ret = btrfs_set_inode_index(new_dir, old_inode, &index); |
aec7477b JB |
3493 | if (ret) |
3494 | goto out_fail; | |
3495 | ||
e02119d5 CM |
3496 | ret = btrfs_add_link(trans, new_dentry->d_parent->d_inode, |
3497 | old_inode, new_dentry->d_name.name, | |
3498 | new_dentry->d_name.len, 1, index); | |
39279cc3 CM |
3499 | if (ret) |
3500 | goto out_fail; | |
3501 | ||
3502 | out_fail: | |
ab78c84d | 3503 | btrfs_end_transaction_throttle(trans, root); |
1832a6d5 | 3504 | out_unlock: |
39279cc3 CM |
3505 | return ret; |
3506 | } | |
3507 | ||
ea8c2819 CM |
3508 | int btrfs_start_delalloc_inodes(struct btrfs_root *root) |
3509 | { | |
3510 | struct list_head *head = &root->fs_info->delalloc_inodes; | |
3511 | struct btrfs_inode *binode; | |
3512 | unsigned long flags; | |
3513 | ||
3514 | spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); | |
3515 | while(!list_empty(head)) { | |
3516 | binode = list_entry(head->next, struct btrfs_inode, | |
3517 | delalloc_inodes); | |
3518 | atomic_inc(&binode->vfs_inode.i_count); | |
3519 | spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); | |
3520 | filemap_write_and_wait(binode->vfs_inode.i_mapping); | |
3521 | iput(&binode->vfs_inode); | |
3522 | spin_lock_irqsave(&root->fs_info->delalloc_lock, flags); | |
3523 | } | |
3524 | spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags); | |
3525 | return 0; | |
3526 | } | |
3527 | ||
39279cc3 CM |
3528 | static int btrfs_symlink(struct inode *dir, struct dentry *dentry, |
3529 | const char *symname) | |
3530 | { | |
3531 | struct btrfs_trans_handle *trans; | |
3532 | struct btrfs_root *root = BTRFS_I(dir)->root; | |
3533 | struct btrfs_path *path; | |
3534 | struct btrfs_key key; | |
1832a6d5 | 3535 | struct inode *inode = NULL; |
39279cc3 CM |
3536 | int err; |
3537 | int drop_inode = 0; | |
3538 | u64 objectid; | |
00e4e6b3 | 3539 | u64 index = 0 ; |
39279cc3 CM |
3540 | int name_len; |
3541 | int datasize; | |
5f39d397 | 3542 | unsigned long ptr; |
39279cc3 | 3543 | struct btrfs_file_extent_item *ei; |
5f39d397 | 3544 | struct extent_buffer *leaf; |
1832a6d5 | 3545 | unsigned long nr = 0; |
39279cc3 CM |
3546 | |
3547 | name_len = strlen(symname) + 1; | |
3548 | if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root)) | |
3549 | return -ENAMETOOLONG; | |
1832a6d5 | 3550 | |
1832a6d5 CM |
3551 | err = btrfs_check_free_space(root, 1, 0); |
3552 | if (err) | |
3553 | goto out_fail; | |
3554 | ||
39279cc3 CM |
3555 | trans = btrfs_start_transaction(root, 1); |
3556 | btrfs_set_trans_block_group(trans, dir); | |
3557 | ||
3558 | err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid); | |
3559 | if (err) { | |
3560 | err = -ENOSPC; | |
3561 | goto out_unlock; | |
3562 | } | |
3563 | ||
aec7477b | 3564 | inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, |
9c58309d CM |
3565 | dentry->d_name.len, |
3566 | dentry->d_parent->d_inode->i_ino, objectid, | |
00e4e6b3 CM |
3567 | BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO, |
3568 | &index); | |
39279cc3 CM |
3569 | err = PTR_ERR(inode); |
3570 | if (IS_ERR(inode)) | |
3571 | goto out_unlock; | |
3572 | ||
33268eaf JB |
3573 | err = btrfs_init_acl(inode, dir); |
3574 | if (err) { | |
3575 | drop_inode = 1; | |
3576 | goto out_unlock; | |
3577 | } | |
3578 | ||
39279cc3 | 3579 | btrfs_set_trans_block_group(trans, inode); |
00e4e6b3 | 3580 | err = btrfs_add_nondir(trans, dentry, inode, 0, index); |
39279cc3 CM |
3581 | if (err) |
3582 | drop_inode = 1; | |
3583 | else { | |
3584 | inode->i_mapping->a_ops = &btrfs_aops; | |
04160088 | 3585 | inode->i_mapping->backing_dev_info = &root->fs_info->bdi; |
39279cc3 CM |
3586 | inode->i_fop = &btrfs_file_operations; |
3587 | inode->i_op = &btrfs_file_inode_operations; | |
d1310b2e | 3588 | BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; |
39279cc3 CM |
3589 | } |
3590 | dir->i_sb->s_dirt = 1; | |
3591 | btrfs_update_inode_block_group(trans, inode); | |
3592 | btrfs_update_inode_block_group(trans, dir); | |
3593 | if (drop_inode) | |
3594 | goto out_unlock; | |
3595 | ||
3596 | path = btrfs_alloc_path(); | |
3597 | BUG_ON(!path); | |
3598 | key.objectid = inode->i_ino; | |
3599 | key.offset = 0; | |
39279cc3 CM |
3600 | btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY); |
3601 | datasize = btrfs_file_extent_calc_inline_size(name_len); | |
3602 | err = btrfs_insert_empty_item(trans, root, path, &key, | |
3603 | datasize); | |
54aa1f4d CM |
3604 | if (err) { |
3605 | drop_inode = 1; | |
3606 | goto out_unlock; | |
3607 | } | |
5f39d397 CM |
3608 | leaf = path->nodes[0]; |
3609 | ei = btrfs_item_ptr(leaf, path->slots[0], | |
3610 | struct btrfs_file_extent_item); | |
3611 | btrfs_set_file_extent_generation(leaf, ei, trans->transid); | |
3612 | btrfs_set_file_extent_type(leaf, ei, | |
39279cc3 CM |
3613 | BTRFS_FILE_EXTENT_INLINE); |
3614 | ptr = btrfs_file_extent_inline_start(ei); | |
5f39d397 CM |
3615 | write_extent_buffer(leaf, symname, ptr, name_len); |
3616 | btrfs_mark_buffer_dirty(leaf); | |
39279cc3 | 3617 | btrfs_free_path(path); |
5f39d397 | 3618 | |
39279cc3 CM |
3619 | inode->i_op = &btrfs_symlink_inode_operations; |
3620 | inode->i_mapping->a_ops = &btrfs_symlink_aops; | |
04160088 | 3621 | inode->i_mapping->backing_dev_info = &root->fs_info->bdi; |
dbe674a9 | 3622 | btrfs_i_size_write(inode, name_len - 1); |
54aa1f4d CM |
3623 | err = btrfs_update_inode(trans, root, inode); |
3624 | if (err) | |
3625 | drop_inode = 1; | |
39279cc3 CM |
3626 | |
3627 | out_unlock: | |
d3c2fdcf | 3628 | nr = trans->blocks_used; |
ab78c84d | 3629 | btrfs_end_transaction_throttle(trans, root); |
1832a6d5 | 3630 | out_fail: |
39279cc3 CM |
3631 | if (drop_inode) { |
3632 | inode_dec_link_count(inode); | |
3633 | iput(inode); | |
3634 | } | |
d3c2fdcf | 3635 | btrfs_btree_balance_dirty(root, nr); |
39279cc3 CM |
3636 | return err; |
3637 | } | |
16432985 | 3638 | |
e6dcd2dc CM |
3639 | static int btrfs_set_page_dirty(struct page *page) |
3640 | { | |
e6dcd2dc CM |
3641 | return __set_page_dirty_nobuffers(page); |
3642 | } | |
3643 | ||
0ee0fda0 SW |
3644 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,26) |
3645 | static int btrfs_permission(struct inode *inode, int mask) | |
3646 | #else | |
fdebe2bd Y |
3647 | static int btrfs_permission(struct inode *inode, int mask, |
3648 | struct nameidata *nd) | |
0ee0fda0 | 3649 | #endif |
fdebe2bd Y |
3650 | { |
3651 | if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE)) | |
3652 | return -EACCES; | |
33268eaf | 3653 | return generic_permission(inode, mask, btrfs_check_acl); |
fdebe2bd | 3654 | } |
39279cc3 CM |
3655 | |
3656 | static struct inode_operations btrfs_dir_inode_operations = { | |
3657 | .lookup = btrfs_lookup, | |
3658 | .create = btrfs_create, | |
3659 | .unlink = btrfs_unlink, | |
3660 | .link = btrfs_link, | |
3661 | .mkdir = btrfs_mkdir, | |
3662 | .rmdir = btrfs_rmdir, | |
3663 | .rename = btrfs_rename, | |
3664 | .symlink = btrfs_symlink, | |
3665 | .setattr = btrfs_setattr, | |
618e21d5 | 3666 | .mknod = btrfs_mknod, |
95819c05 CH |
3667 | .setxattr = btrfs_setxattr, |
3668 | .getxattr = btrfs_getxattr, | |
5103e947 | 3669 | .listxattr = btrfs_listxattr, |
95819c05 | 3670 | .removexattr = btrfs_removexattr, |
fdebe2bd | 3671 | .permission = btrfs_permission, |
39279cc3 | 3672 | }; |
39279cc3 CM |
3673 | static struct inode_operations btrfs_dir_ro_inode_operations = { |
3674 | .lookup = btrfs_lookup, | |
fdebe2bd | 3675 | .permission = btrfs_permission, |
39279cc3 | 3676 | }; |
39279cc3 CM |
3677 | static struct file_operations btrfs_dir_file_operations = { |
3678 | .llseek = generic_file_llseek, | |
3679 | .read = generic_read_dir, | |
cbdf5a24 DW |
3680 | #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28) |
3681 | .readdir = btrfs_nfshack_readdir, | |
3682 | #else /* NFSd readdir/lookup deadlock is fixed */ | |
3683 | .readdir = btrfs_real_readdir, | |
3684 | #endif | |
34287aa3 | 3685 | .unlocked_ioctl = btrfs_ioctl, |
39279cc3 | 3686 | #ifdef CONFIG_COMPAT |
34287aa3 | 3687 | .compat_ioctl = btrfs_ioctl, |
39279cc3 | 3688 | #endif |
6bf13c0c | 3689 | .release = btrfs_release_file, |
e02119d5 | 3690 | .fsync = btrfs_sync_file, |
39279cc3 CM |
3691 | }; |
3692 | ||
d1310b2e | 3693 | static struct extent_io_ops btrfs_extent_io_ops = { |
07157aac | 3694 | .fill_delalloc = run_delalloc_range, |
065631f6 | 3695 | .submit_bio_hook = btrfs_submit_bio_hook, |
239b14b3 | 3696 | .merge_bio_hook = btrfs_merge_bio_hook, |
07157aac | 3697 | .readpage_end_io_hook = btrfs_readpage_end_io_hook, |
e6dcd2dc | 3698 | .writepage_end_io_hook = btrfs_writepage_end_io_hook, |
247e743c | 3699 | .writepage_start_hook = btrfs_writepage_start_hook, |
1259ab75 | 3700 | .readpage_io_failed_hook = btrfs_io_failed_hook, |
b0c68f8b CM |
3701 | .set_bit_hook = btrfs_set_bit_hook, |
3702 | .clear_bit_hook = btrfs_clear_bit_hook, | |
07157aac CM |
3703 | }; |
3704 | ||
39279cc3 CM |
3705 | static struct address_space_operations btrfs_aops = { |
3706 | .readpage = btrfs_readpage, | |
3707 | .writepage = btrfs_writepage, | |
b293f02e | 3708 | .writepages = btrfs_writepages, |
3ab2fb5a | 3709 | .readpages = btrfs_readpages, |
39279cc3 | 3710 | .sync_page = block_sync_page, |
39279cc3 | 3711 | .bmap = btrfs_bmap, |
16432985 | 3712 | .direct_IO = btrfs_direct_IO, |
a52d9a80 CM |
3713 | .invalidatepage = btrfs_invalidatepage, |
3714 | .releasepage = btrfs_releasepage, | |
e6dcd2dc | 3715 | .set_page_dirty = btrfs_set_page_dirty, |
39279cc3 CM |
3716 | }; |
3717 | ||
3718 | static struct address_space_operations btrfs_symlink_aops = { | |
3719 | .readpage = btrfs_readpage, | |
3720 | .writepage = btrfs_writepage, | |
2bf5a725 CM |
3721 | .invalidatepage = btrfs_invalidatepage, |
3722 | .releasepage = btrfs_releasepage, | |
39279cc3 CM |
3723 | }; |
3724 | ||
3725 | static struct inode_operations btrfs_file_inode_operations = { | |
3726 | .truncate = btrfs_truncate, | |
3727 | .getattr = btrfs_getattr, | |
3728 | .setattr = btrfs_setattr, | |
95819c05 CH |
3729 | .setxattr = btrfs_setxattr, |
3730 | .getxattr = btrfs_getxattr, | |
5103e947 | 3731 | .listxattr = btrfs_listxattr, |
95819c05 | 3732 | .removexattr = btrfs_removexattr, |
fdebe2bd | 3733 | .permission = btrfs_permission, |
39279cc3 | 3734 | }; |
618e21d5 JB |
3735 | static struct inode_operations btrfs_special_inode_operations = { |
3736 | .getattr = btrfs_getattr, | |
3737 | .setattr = btrfs_setattr, | |
fdebe2bd | 3738 | .permission = btrfs_permission, |
95819c05 CH |
3739 | .setxattr = btrfs_setxattr, |
3740 | .getxattr = btrfs_getxattr, | |
33268eaf | 3741 | .listxattr = btrfs_listxattr, |
95819c05 | 3742 | .removexattr = btrfs_removexattr, |
618e21d5 | 3743 | }; |
39279cc3 CM |
3744 | static struct inode_operations btrfs_symlink_inode_operations = { |
3745 | .readlink = generic_readlink, | |
3746 | .follow_link = page_follow_link_light, | |
3747 | .put_link = page_put_link, | |
fdebe2bd | 3748 | .permission = btrfs_permission, |
39279cc3 | 3749 | }; |