projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Btrfs: add compat ioctl
[deliverable/linux.git] / fs / btrfs / super.c
1 #include <linux/module.h>
2 #include <linux/buffer_head.h>
3 #include <linux/fs.h>
4 #include <linux/pagemap.h>
5 #include <linux/highmem.h>
6 #include <linux/time.h>
7 #include <linux/init.h>
8 #include <linux/string.h>
9 #include <linux/smp_lock.h>
10 #include <linux/backing-dev.h>
11 #include <linux/mpage.h>
12 #include <linux/swap.h>
13 #include <linux/writeback.h>
14 #include <linux/statfs.h>
15 #include <linux/compat.h>
16 #include "ctree.h"
17 #include "disk-io.h"
18 #include "transaction.h"
19 #include "btrfs_inode.h"
20 #include "ioctl.h"
21 #include "print-tree.h"
22
23 struct btrfs_iget_args {
24 u64 ino;
25 struct btrfs_root *root;
26 };
27
28 #define BTRFS_SUPER_MAGIC 0x9123682E
29
30 static struct inode_operations btrfs_dir_inode_operations;
31 static struct inode_operations btrfs_symlink_inode_operations;
32 static struct inode_operations btrfs_dir_ro_inode_operations;
33 static struct super_operations btrfs_super_ops;
34 static struct file_operations btrfs_dir_file_operations;
35 static struct inode_operations btrfs_file_inode_operations;
36 static struct address_space_operations btrfs_aops;
37 static struct address_space_operations btrfs_symlink_aops;
38 static struct file_operations btrfs_file_operations;
39
40 static int drop_extents(struct btrfs_trans_handle *trans,
41 struct btrfs_root *root,
42 struct inode *inode,
43 u64 start, u64 end, u64 *hint_block);
44 static int btrfs_get_block(struct inode *inode, sector_t iblock,
45 struct buffer_head *result, int create);
46
47
48 #define S_SHIFT 12
49 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
50 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
51 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
52 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
53 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
54 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
55 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
56 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
57 };
58
59 static void btrfs_read_locked_inode(struct inode *inode)
60 {
61 struct btrfs_path *path;
62 struct btrfs_inode_item *inode_item;
63 struct btrfs_root *root = BTRFS_I(inode)->root;
64 struct btrfs_key location;
65 struct btrfs_block_group_cache *alloc_group;
66 u64 alloc_group_block;
67 int ret;
68
69 path = btrfs_alloc_path();
70 BUG_ON(!path);
71 btrfs_init_path(path);
72 mutex_lock(&root->fs_info->fs_mutex);
73
74 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
75 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
76 if (ret) {
77 btrfs_free_path(path);
78 goto make_bad;
79 }
80 inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
81 path->slots[0],
82 struct btrfs_inode_item);
83
84 inode->i_mode = btrfs_inode_mode(inode_item);
85 inode->i_nlink = btrfs_inode_nlink(inode_item);
86 inode->i_uid = btrfs_inode_uid(inode_item);
87 inode->i_gid = btrfs_inode_gid(inode_item);
88 inode->i_size = btrfs_inode_size(inode_item);
89 inode->i_atime.tv_sec = btrfs_timespec_sec(&inode_item->atime);
90 inode->i_atime.tv_nsec = btrfs_timespec_nsec(&inode_item->atime);
91 inode->i_mtime.tv_sec = btrfs_timespec_sec(&inode_item->mtime);
92 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(&inode_item->mtime);
93 inode->i_ctime.tv_sec = btrfs_timespec_sec(&inode_item->ctime);
94 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime);
95 inode->i_blocks = btrfs_inode_nblocks(inode_item);
96 inode->i_generation = btrfs_inode_generation(inode_item);
97 alloc_group_block = btrfs_inode_block_group(inode_item);
98 ret = radix_tree_gang_lookup(&root->fs_info->block_group_radix,
99 (void **)&alloc_group,
100 alloc_group_block, 1);
101 BUG_ON(!ret);
102 BTRFS_I(inode)->block_group = alloc_group;
103
104 btrfs_free_path(path);
105 inode_item = NULL;
106
107 mutex_unlock(&root->fs_info->fs_mutex);
108
109 switch (inode->i_mode & S_IFMT) {
110 #if 0
111 default:
112 init_special_inode(inode, inode->i_mode,
113 btrfs_inode_rdev(inode_item));
114 break;
115 #endif
116 case S_IFREG:
117 inode->i_mapping->a_ops = &btrfs_aops;
118 inode->i_fop = &btrfs_file_operations;
119 inode->i_op = &btrfs_file_inode_operations;
120 break;
121 case S_IFDIR:
122 inode->i_fop = &btrfs_dir_file_operations;
123 if (root == root->fs_info->tree_root)
124 inode->i_op = &btrfs_dir_ro_inode_operations;
125 else
126 inode->i_op = &btrfs_dir_inode_operations;
127 break;
128 case S_IFLNK:
129 inode->i_op = &btrfs_symlink_inode_operations;
130 inode->i_mapping->a_ops = &btrfs_symlink_aops;
131 break;
132 }
133 return;
134
135 make_bad:
136 btrfs_release_path(root, path);
137 btrfs_free_path(path);
138 mutex_unlock(&root->fs_info->fs_mutex);
139 make_bad_inode(inode);
140 }
141
142 static void fill_inode_item(struct btrfs_inode_item *item,
143 struct inode *inode)
144 {
145 btrfs_set_inode_uid(item, inode->i_uid);
146 btrfs_set_inode_gid(item, inode->i_gid);
147 btrfs_set_inode_size(item, inode->i_size);
148 btrfs_set_inode_mode(item, inode->i_mode);
149 btrfs_set_inode_nlink(item, inode->i_nlink);
150 btrfs_set_timespec_sec(&item->atime, inode->i_atime.tv_sec);
151 btrfs_set_timespec_nsec(&item->atime, inode->i_atime.tv_nsec);
152 btrfs_set_timespec_sec(&item->mtime, inode->i_mtime.tv_sec);
153 btrfs_set_timespec_nsec(&item->mtime, inode->i_mtime.tv_nsec);
154 btrfs_set_timespec_sec(&item->ctime, inode->i_ctime.tv_sec);
155 btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec);
156 btrfs_set_inode_nblocks(item, inode->i_blocks);
157 btrfs_set_inode_generation(item, inode->i_generation);
158 btrfs_set_inode_block_group(item,
159 BTRFS_I(inode)->block_group->key.objectid);
160 }
161
162 static int btrfs_update_inode(struct btrfs_trans_handle *trans,
163 struct btrfs_root *root,
164 struct inode *inode)
165 {
166 struct btrfs_inode_item *inode_item;
167 struct btrfs_path *path;
168 int ret;
169
170 path = btrfs_alloc_path();
171 BUG_ON(!path);
172 btrfs_init_path(path);
173 ret = btrfs_lookup_inode(trans, root, path,
174 &BTRFS_I(inode)->location, 1);
175 if (ret) {
176 if (ret > 0)
177 ret = -ENOENT;
178 goto failed;
179 }
180
181 inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
182 path->slots[0],
183 struct btrfs_inode_item);
184
185 fill_inode_item(inode_item, inode);
186 btrfs_mark_buffer_dirty(path->nodes[0]);
187 ret = 0;
188 failed:
189 btrfs_release_path(root, path);
190 btrfs_free_path(path);
191 return ret;
192 }
193
194
195 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
196 struct btrfs_root *root,
197 struct inode *dir,
198 struct dentry *dentry)
199 {
200 struct btrfs_path *path;
201 const char *name = dentry->d_name.name;
202 int name_len = dentry->d_name.len;
203 int ret = 0;
204 u64 objectid;
205 struct btrfs_dir_item *di;
206
207 path = btrfs_alloc_path();
208 BUG_ON(!path);
209 btrfs_init_path(path);
210 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
211 name, name_len, -1);
212 if (IS_ERR(di)) {
213 ret = PTR_ERR(di);
214 goto err;
215 }
216 if (!di) {
217 ret = -ENOENT;
218 goto err;
219 }
220 objectid = btrfs_disk_key_objectid(&di->location);
221 ret = btrfs_delete_one_dir_name(trans, root, path, di);
222 BUG_ON(ret);
223 btrfs_release_path(root, path);
224
225 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
226 objectid, name, name_len, -1);
227 if (IS_ERR(di)) {
228 ret = PTR_ERR(di);
229 goto err;
230 }
231 if (!di) {
232 ret = -ENOENT;
233 goto err;
234 }
235 ret = btrfs_delete_one_dir_name(trans, root, path, di);
236 BUG_ON(ret);
237
238 dentry->d_inode->i_ctime = dir->i_ctime;
239 err:
240 btrfs_free_path(path);
241 if (!ret) {
242 dir->i_size -= name_len * 2;
243 btrfs_update_inode(trans, root, dir);
244 drop_nlink(dentry->d_inode);
245 btrfs_update_inode(trans, root, dentry->d_inode);
246 dir->i_sb->s_dirt = 1;
247 }
248 return ret;
249 }
250
251 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
252 {
253 struct btrfs_root *root;
254 struct btrfs_trans_handle *trans;
255 int ret;
256
257 root = BTRFS_I(dir)->root;
258 mutex_lock(&root->fs_info->fs_mutex);
259 trans = btrfs_start_transaction(root, 1);
260 btrfs_set_trans_block_group(trans, dir);
261 ret = btrfs_unlink_trans(trans, root, dir, dentry);
262 btrfs_end_transaction(trans, root);
263 mutex_unlock(&root->fs_info->fs_mutex);
264 btrfs_btree_balance_dirty(root);
265 return ret;
266 }
267
268 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
269 {
270 struct inode *inode = dentry->d_inode;
271 int err;
272 int ret;
273 struct btrfs_root *root = BTRFS_I(dir)->root;
274 struct btrfs_path *path;
275 struct btrfs_key key;
276 struct btrfs_trans_handle *trans;
277 struct btrfs_key found_key;
278 int found_type;
279 struct btrfs_leaf *leaf;
280 char *goodnames = "..";
281
282 path = btrfs_alloc_path();
283 BUG_ON(!path);
284 btrfs_init_path(path);
285 mutex_lock(&root->fs_info->fs_mutex);
286 trans = btrfs_start_transaction(root, 1);
287 btrfs_set_trans_block_group(trans, dir);
288 key.objectid = inode->i_ino;
289 key.offset = (u64)-1;
290 key.flags = (u32)-1;
291 while(1) {
292 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
293 if (ret < 0) {
294 err = ret;
295 goto out;
296 }
297 BUG_ON(ret == 0);
298 if (path->slots[0] == 0) {
299 err = -ENOENT;
300 goto out;
301 }
302 path->slots[0]--;
303 leaf = btrfs_buffer_leaf(path->nodes[0]);
304 btrfs_disk_key_to_cpu(&found_key,
305 &leaf->items[path->slots[0]].key);
306 found_type = btrfs_key_type(&found_key);
307 if (found_key.objectid != inode->i_ino) {
308 err = -ENOENT;
309 goto out;
310 }
311 if ((found_type != BTRFS_DIR_ITEM_KEY &&
312 found_type != BTRFS_DIR_INDEX_KEY) ||
313 (!btrfs_match_dir_item_name(root, path, goodnames, 2) &&
314 !btrfs_match_dir_item_name(root, path, goodnames, 1))) {
315 err = -ENOTEMPTY;
316 goto out;
317 }
318 ret = btrfs_del_item(trans, root, path);
319 BUG_ON(ret);
320
321 if (found_type == BTRFS_DIR_ITEM_KEY && found_key.offset == 1)
322 break;
323 btrfs_release_path(root, path);
324 }
325 ret = 0;
326 btrfs_release_path(root, path);
327
328 /* now the directory is empty */
329 err = btrfs_unlink_trans(trans, root, dir, dentry);
330 if (!err) {
331 inode->i_size = 0;
332 }
333 out:
334 btrfs_release_path(root, path);
335 btrfs_free_path(path);
336 mutex_unlock(&root->fs_info->fs_mutex);
337 ret = btrfs_end_transaction(trans, root);
338 btrfs_btree_balance_dirty(root);
339 if (ret && !err)
340 err = ret;
341 return err;
342 }
343
344 static int btrfs_free_inode(struct btrfs_trans_handle *trans,
345 struct btrfs_root *root,
346 struct inode *inode)
347 {
348 struct btrfs_path *path;
349 int ret;
350
351 clear_inode(inode);
352
353 path = btrfs_alloc_path();
354 BUG_ON(!path);
355 btrfs_init_path(path);
356 ret = btrfs_lookup_inode(trans, root, path,
357 &BTRFS_I(inode)->location, -1);
358 BUG_ON(ret);
359 ret = btrfs_del_item(trans, root, path);
360 BUG_ON(ret);
361 btrfs_free_path(path);
362 return ret;
363 }
364
365 static void reada_truncate(struct btrfs_root *root, struct btrfs_path *path,
366 u64 objectid)
367 {
368 struct btrfs_node *node;
369 int i;
370 int nritems;
371 u64 item_objectid;
372 u64 blocknr;
373 int slot;
374 int ret;
375
376 if (!path->nodes[1])
377 return;
378 node = btrfs_buffer_node(path->nodes[1]);
379 slot = path->slots[1];
380 if (slot == 0)
381 return;
382 nritems = btrfs_header_nritems(&node->header);
383 for (i = slot - 1; i >= 0; i--) {
384 item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key);
385 if (item_objectid != objectid)
386 break;
387 blocknr = btrfs_node_blockptr(node, i);
388 ret = readahead_tree_block(root, blocknr);
389 if (ret)
390 break;
391 }
392 }
393
394 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
395 struct btrfs_root *root,
396 struct inode *inode)
397 {
398 int ret;
399 struct btrfs_path *path;
400 struct btrfs_key key;
401 struct btrfs_disk_key *found_key;
402 u32 found_type;
403 struct btrfs_leaf *leaf;
404 struct btrfs_file_extent_item *fi;
405 u64 extent_start = 0;
406 u64 extent_num_blocks = 0;
407 u64 item_end = 0;
408 int found_extent;
409 int del_item;
410
411 path = btrfs_alloc_path();
412 BUG_ON(!path);
413 /* FIXME, add redo link to tree so we don't leak on crash */
414 key.objectid = inode->i_ino;
415 key.offset = (u64)-1;
416 key.flags = (u32)-1;
417 while(1) {
418 btrfs_init_path(path);
419 fi = NULL;
420 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
421 if (ret < 0) {
422 goto error;
423 }
424 if (ret > 0) {
425 BUG_ON(path->slots[0] == 0);
426 path->slots[0]--;
427 }
428 reada_truncate(root, path, inode->i_ino);
429 leaf = btrfs_buffer_leaf(path->nodes[0]);
430 found_key = &leaf->items[path->slots[0]].key;
431 found_type = btrfs_disk_key_type(found_key);
432 if (btrfs_disk_key_objectid(found_key) != inode->i_ino)
433 break;
434 if (found_type != BTRFS_CSUM_ITEM_KEY &&
435 found_type != BTRFS_DIR_ITEM_KEY &&
436 found_type != BTRFS_DIR_INDEX_KEY &&
437 found_type != BTRFS_EXTENT_DATA_KEY)
438 break;
439 item_end = btrfs_disk_key_offset(found_key);
440 if (found_type == BTRFS_EXTENT_DATA_KEY) {
441 fi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
442 path->slots[0],
443 struct btrfs_file_extent_item);
444 if (btrfs_file_extent_type(fi) !=
445 BTRFS_FILE_EXTENT_INLINE) {
446 item_end += btrfs_file_extent_num_blocks(fi) <<
447 inode->i_blkbits;
448 }
449 }
450 if (found_type == BTRFS_CSUM_ITEM_KEY) {
451 ret = btrfs_csum_truncate(trans, root, path,
452 inode->i_size);
453 BUG_ON(ret);
454 }
455 if (item_end < inode->i_size) {
456 if (found_type) {
457 btrfs_set_key_type(&key, found_type - 1);
458 continue;
459 }
460 break;
461 }
462 if (btrfs_disk_key_offset(found_key) >= inode->i_size)
463 del_item = 1;
464 else
465 del_item = 0;
466 found_extent = 0;
467
468 if (found_type == BTRFS_EXTENT_DATA_KEY &&
469 btrfs_file_extent_type(fi) !=
470 BTRFS_FILE_EXTENT_INLINE) {
471 u64 num_dec;
472 if (!del_item) {
473 u64 orig_num_blocks =
474 btrfs_file_extent_num_blocks(fi);
475 extent_num_blocks = inode->i_size -
476 btrfs_disk_key_offset(found_key) +
477 root->blocksize - 1;
478 extent_num_blocks >>= inode->i_blkbits;
479 btrfs_set_file_extent_num_blocks(fi,
480 extent_num_blocks);
481 inode->i_blocks -= (orig_num_blocks -
482 extent_num_blocks) << 3;
483 mark_buffer_dirty(path->nodes[0]);
484 } else {
485 extent_start =
486 btrfs_file_extent_disk_blocknr(fi);
487 extent_num_blocks =
488 btrfs_file_extent_disk_num_blocks(fi);
489 /* FIXME blocksize != 4096 */
490 num_dec = btrfs_file_extent_num_blocks(fi) << 3;
491 if (extent_start != 0) {
492 found_extent = 1;
493 inode->i_blocks -= num_dec;
494 }
495 }
496 }
497 if (del_item) {
498 ret = btrfs_del_item(trans, root, path);
499 BUG_ON(ret);
500 } else {
501 break;
502 }
503 btrfs_release_path(root, path);
504 if (found_extent) {
505 ret = btrfs_free_extent(trans, root, extent_start,
506 extent_num_blocks, 0);
507 BUG_ON(ret);
508 }
509 }
510 ret = 0;
511 error:
512 btrfs_release_path(root, path);
513 btrfs_free_path(path);
514 inode->i_sb->s_dirt = 1;
515 return ret;
516 }
517
518 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
519 {
520 struct inode *inode = mapping->host;
521 unsigned blocksize = 1 << inode->i_blkbits;
522 pgoff_t index = from >> PAGE_CACHE_SHIFT;
523 unsigned offset = from & (PAGE_CACHE_SIZE-1);
524 struct page *page;
525 char *kaddr;
526 int ret = 0;
527 struct btrfs_root *root = BTRFS_I(inode)->root;
528 u64 alloc_hint;
529 struct btrfs_key ins;
530 struct btrfs_trans_handle *trans;
531
532 if ((offset & (blocksize - 1)) == 0)
533 goto out;
534
535 ret = -ENOMEM;
536 page = grab_cache_page(mapping, index);
537 if (!page)
538 goto out;
539
540 if (!PageUptodate(page)) {
541 ret = mpage_readpage(page, btrfs_get_block);
542 lock_page(page);
543 if (!PageUptodate(page)) {
544 ret = -EIO;
545 goto out;
546 }
547 }
548 mutex_lock(&root->fs_info->fs_mutex);
549 trans = btrfs_start_transaction(root, 1);
550 btrfs_set_trans_block_group(trans, inode);
551
552 ret = drop_extents(trans, root, inode, page->index << PAGE_CACHE_SHIFT,
553 (page->index + 1) << PAGE_CACHE_SHIFT, &alloc_hint);
554 BUG_ON(ret);
555 ret = btrfs_alloc_extent(trans, root, inode->i_ino, 1,
556 alloc_hint, (u64)-1, &ins, 1);
557 BUG_ON(ret);
558 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
559 page->index << PAGE_CACHE_SHIFT,
560 ins.objectid, 1, 1);
561 BUG_ON(ret);
562 SetPageChecked(page);
563 kaddr = kmap(page);
564 memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
565 flush_dcache_page(page);
566 btrfs_csum_file_block(trans, root, inode->i_ino,
567 page->index << PAGE_CACHE_SHIFT,
568 kaddr, PAGE_CACHE_SIZE);
569 kunmap(page);
570 btrfs_end_transaction(trans, root);
571 mutex_unlock(&root->fs_info->fs_mutex);
572
573 set_page_dirty(page);
574 unlock_page(page);
575 page_cache_release(page);
576 out:
577 return ret;
578 }
579
580 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
581 {
582 struct inode *inode = dentry->d_inode;
583 int err;
584
585 err = inode_change_ok(inode, attr);
586 if (err)
587 return err;
588
589 if (S_ISREG(inode->i_mode) &&
590 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
591 struct btrfs_trans_handle *trans;
592 struct btrfs_root *root = BTRFS_I(inode)->root;
593 u64 mask = root->blocksize - 1;
594 u64 pos = (inode->i_size + mask) & ~mask;
595 u64 hole_size;
596
597 if (attr->ia_size <= pos)
598 goto out;
599
600 btrfs_truncate_page(inode->i_mapping, inode->i_size);
601
602 hole_size = (attr->ia_size - pos + mask) & ~mask;
603 hole_size >>= inode->i_blkbits;
604
605 mutex_lock(&root->fs_info->fs_mutex);
606 trans = btrfs_start_transaction(root, 1);
607 btrfs_set_trans_block_group(trans, inode);
608 err = btrfs_insert_file_extent(trans, root, inode->i_ino,
609 pos, 0, 0, hole_size);
610 BUG_ON(err);
611 btrfs_end_transaction(trans, root);
612 mutex_unlock(&root->fs_info->fs_mutex);
613 }
614 out:
615 err = inode_setattr(inode, attr);
616
617 return err;
618 }
619 static void btrfs_delete_inode(struct inode *inode)
620 {
621 struct btrfs_trans_handle *trans;
622 struct btrfs_root *root = BTRFS_I(inode)->root;
623 int ret;
624
625 truncate_inode_pages(&inode->i_data, 0);
626 if (is_bad_inode(inode)) {
627 goto no_delete;
628 }
629 inode->i_size = 0;
630 mutex_lock(&root->fs_info->fs_mutex);
631 trans = btrfs_start_transaction(root, 1);
632 btrfs_set_trans_block_group(trans, inode);
633 ret = btrfs_truncate_in_trans(trans, root, inode);
634 BUG_ON(ret);
635 btrfs_free_inode(trans, root, inode);
636 btrfs_end_transaction(trans, root);
637 mutex_unlock(&root->fs_info->fs_mutex);
638 btrfs_btree_balance_dirty(root);
639 return;
640 no_delete:
641 clear_inode(inode);
642 }
643
644 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
645 struct btrfs_key *location)
646 {
647 const char *name = dentry->d_name.name;
648 int namelen = dentry->d_name.len;
649 struct btrfs_dir_item *di;
650 struct btrfs_path *path;
651 struct btrfs_root *root = BTRFS_I(dir)->root;
652 int ret;
653
654 path = btrfs_alloc_path();
655 BUG_ON(!path);
656 btrfs_init_path(path);
657 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
658 namelen, 0);
659 if (!di || IS_ERR(di)) {
660 location->objectid = 0;
661 ret = 0;
662 goto out;
663 }
664 btrfs_disk_key_to_cpu(location, &di->location);
665 out:
666 btrfs_release_path(root, path);
667 btrfs_free_path(path);
668 return ret;
669 }
670
671 static int fixup_tree_root_location(struct btrfs_root *root,
672 struct btrfs_key *location,
673 struct btrfs_root **sub_root)
674 {
675 struct btrfs_path *path;
676 struct btrfs_root_item *ri;
677
678 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
679 return 0;
680 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
681 return 0;
682
683 path = btrfs_alloc_path();
684 BUG_ON(!path);
685 mutex_lock(&root->fs_info->fs_mutex);
686
687 *sub_root = btrfs_read_fs_root(root->fs_info, location);
688 if (IS_ERR(*sub_root))
689 return PTR_ERR(*sub_root);
690
691 ri = &(*sub_root)->root_item;
692 location->objectid = btrfs_root_dirid(ri);
693 location->flags = 0;
694 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
695 location->offset = 0;
696
697 btrfs_free_path(path);
698 mutex_unlock(&root->fs_info->fs_mutex);
699 return 0;
700 }
701
702 static int btrfs_init_locked_inode(struct inode *inode, void *p)
703 {
704 struct btrfs_iget_args *args = p;
705 inode->i_ino = args->ino;
706 BTRFS_I(inode)->root = args->root;
707 return 0;
708 }
709
710 static int btrfs_find_actor(struct inode *inode, void *opaque)
711 {
712 struct btrfs_iget_args *args = opaque;
713 return (args->ino == inode->i_ino &&
714 args->root == BTRFS_I(inode)->root);
715 }
716
717 static struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
718 struct btrfs_root *root)
719 {
720 struct inode *inode;
721 struct btrfs_iget_args args;
722 args.ino = objectid;
723 args.root = root;
724
725 inode = iget5_locked(s, objectid, btrfs_find_actor,
726 btrfs_init_locked_inode,
727 (void *)&args);
728 return inode;
729 }
730
731 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
732 struct nameidata *nd)
733 {
734 struct inode * inode;
735 struct btrfs_inode *bi = BTRFS_I(dir);
736 struct btrfs_root *root = bi->root;
737 struct btrfs_root *sub_root = root;
738 struct btrfs_key location;
739 int ret;
740
741 if (dentry->d_name.len > BTRFS_NAME_LEN)
742 return ERR_PTR(-ENAMETOOLONG);
743 mutex_lock(&root->fs_info->fs_mutex);
744 ret = btrfs_inode_by_name(dir, dentry, &location);
745 mutex_unlock(&root->fs_info->fs_mutex);
746 if (ret < 0)
747 return ERR_PTR(ret);
748 inode = NULL;
749 if (location.objectid) {
750 ret = fixup_tree_root_location(root, &location, &sub_root);
751 if (ret < 0)
752 return ERR_PTR(ret);
753 if (ret > 0)
754 return ERR_PTR(-ENOENT);
755 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
756 sub_root);
757 if (!inode)
758 return ERR_PTR(-EACCES);
759 if (inode->i_state & I_NEW) {
760 if (sub_root != root) {
761 printk("adding new root for inode %lu root %p (found %p)\n", inode->i_ino, sub_root, BTRFS_I(inode)->root);
762 igrab(inode);
763 sub_root->inode = inode;
764 }
765 BTRFS_I(inode)->root = sub_root;
766 memcpy(&BTRFS_I(inode)->location, &location,
767 sizeof(location));
768 btrfs_read_locked_inode(inode);
769 unlock_new_inode(inode);
770 }
771 }
772 return d_splice_alias(inode, dentry);
773 }
774
775 static void reada_leaves(struct btrfs_root *root, struct btrfs_path *path,
776 u64 objectid)
777 {
778 struct btrfs_node *node;
779 int i;
780 u32 nritems;
781 u64 item_objectid;
782 u64 blocknr;
783 int slot;
784 int ret;
785
786 if (!path->nodes[1])
787 return;
788 node = btrfs_buffer_node(path->nodes[1]);
789 slot = path->slots[1];
790 nritems = btrfs_header_nritems(&node->header);
791 for (i = slot + 1; i < nritems; i++) {
792 item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key);
793 if (item_objectid != objectid)
794 break;
795 blocknr = btrfs_node_blockptr(node, i);
796 ret = readahead_tree_block(root, blocknr);
797 if (ret)
798 break;
799 }
800 }
801 static unsigned char btrfs_filetype_table[] = {
802 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
803 };
804
805 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
806 {
807 struct inode *inode = filp->f_path.dentry->d_inode;
808 struct btrfs_root *root = BTRFS_I(inode)->root;
809 struct btrfs_item *item;
810 struct btrfs_dir_item *di;
811 struct btrfs_key key;
812 struct btrfs_path *path;
813 int ret;
814 u32 nritems;
815 struct btrfs_leaf *leaf;
816 int slot;
817 int advance;
818 unsigned char d_type;
819 int over = 0;
820 u32 di_cur;
821 u32 di_total;
822 u32 di_len;
823 int key_type = BTRFS_DIR_INDEX_KEY;
824
825 /* FIXME, use a real flag for deciding about the key type */
826 if (root->fs_info->tree_root == root)
827 key_type = BTRFS_DIR_ITEM_KEY;
828 mutex_lock(&root->fs_info->fs_mutex);
829 key.objectid = inode->i_ino;
830 key.flags = 0;
831 btrfs_set_key_type(&key, key_type);
832 key.offset = filp->f_pos;
833 path = btrfs_alloc_path();
834 btrfs_init_path(path);
835 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
836 if (ret < 0)
837 goto err;
838 advance = 0;
839 reada_leaves(root, path, inode->i_ino);
840 while(1) {
841 leaf = btrfs_buffer_leaf(path->nodes[0]);
842 nritems = btrfs_header_nritems(&leaf->header);
843 slot = path->slots[0];
844 if (advance || slot >= nritems) {
845 if (slot >= nritems -1) {
846 reada_leaves(root, path, inode->i_ino);
847 ret = btrfs_next_leaf(root, path);
848 if (ret)
849 break;
850 leaf = btrfs_buffer_leaf(path->nodes[0]);
851 nritems = btrfs_header_nritems(&leaf->header);
852 slot = path->slots[0];
853 } else {
854 slot++;
855 path->slots[0]++;
856 }
857 }
858 advance = 1;
859 item = leaf->items + slot;
860 if (btrfs_disk_key_objectid(&item->key) != key.objectid)
861 break;
862 if (btrfs_disk_key_type(&item->key) != key_type)
863 break;
864 if (btrfs_disk_key_offset(&item->key) < filp->f_pos)
865 continue;
866 filp->f_pos = btrfs_disk_key_offset(&item->key);
867 advance = 1;
868 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
869 di_cur = 0;
870 di_total = btrfs_item_size(leaf->items + slot);
871 while(di_cur < di_total) {
872 d_type = btrfs_filetype_table[btrfs_dir_type(di)];
873 over = filldir(dirent, (const char *)(di + 1),
874 btrfs_dir_name_len(di),
875 btrfs_disk_key_offset(&item->key),
876 btrfs_disk_key_objectid(&di->location),
877 d_type);
878 if (over)
879 goto nopos;
880 di_len = btrfs_dir_name_len(di) + sizeof(*di);
881 di_cur += di_len;
882 di = (struct btrfs_dir_item *)((char *)di + di_len);
883 }
884 }
885 filp->f_pos++;
886 nopos:
887 ret = 0;
888 err:
889 btrfs_release_path(root, path);
890 btrfs_free_path(path);
891 mutex_unlock(&root->fs_info->fs_mutex);
892 return ret;
893 }
894
895 static void btrfs_put_super (struct super_block * sb)
896 {
897 struct btrfs_root *root = btrfs_sb(sb);
898 int ret;
899
900 ret = close_ctree(root);
901 if (ret) {
902 printk("close ctree returns %d\n", ret);
903 }
904 sb->s_fs_info = NULL;
905 }
906
907 static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
908 {
909 struct inode * inode;
910 struct dentry * root_dentry;
911 struct btrfs_super_block *disk_super;
912 struct btrfs_root *tree_root;
913 struct btrfs_inode *bi;
914
915 sb->s_maxbytes = MAX_LFS_FILESIZE;
916 sb->s_magic = BTRFS_SUPER_MAGIC;
917 sb->s_op = &btrfs_super_ops;
918 sb->s_time_gran = 1;
919
920 tree_root = open_ctree(sb);
921
922 if (!tree_root) {
923 printk("btrfs: open_ctree failed\n");
924 return -EIO;
925 }
926 sb->s_fs_info = tree_root;
927 disk_super = tree_root->fs_info->disk_super;
928 printk("read in super total blocks %Lu root %Lu\n",
929 btrfs_super_total_blocks(disk_super),
930 btrfs_super_root_dir(disk_super));
931
932 inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super),
933 tree_root);
934 bi = BTRFS_I(inode);
935 bi->location.objectid = inode->i_ino;
936 bi->location.offset = 0;
937 bi->location.flags = 0;
938 bi->root = tree_root;
939 btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
940
941 if (!inode)
942 return -ENOMEM;
943 if (inode->i_state & I_NEW) {
944 btrfs_read_locked_inode(inode);
945 unlock_new_inode(inode);
946 }
947
948 root_dentry = d_alloc_root(inode);
949 if (!root_dentry) {
950 iput(inode);
951 return -ENOMEM;
952 }
953 sb->s_root = root_dentry;
954 btrfs_transaction_queue_work(tree_root, HZ * 30);
955 return 0;
956 }
957
958 static int btrfs_write_inode(struct inode *inode, int wait)
959 {
960 struct btrfs_root *root = BTRFS_I(inode)->root;
961 struct btrfs_trans_handle *trans;
962 int ret = 0;
963
964 if (wait) {
965 mutex_lock(&root->fs_info->fs_mutex);
966 trans = btrfs_start_transaction(root, 1);
967 btrfs_set_trans_block_group(trans, inode);
968 ret = btrfs_commit_transaction(trans, root);
969 mutex_unlock(&root->fs_info->fs_mutex);
970 }
971 return ret;
972 }
973
974 static void btrfs_dirty_inode(struct inode *inode)
975 {
976 struct btrfs_root *root = BTRFS_I(inode)->root;
977 struct btrfs_trans_handle *trans;
978
979 mutex_lock(&root->fs_info->fs_mutex);
980 trans = btrfs_start_transaction(root, 1);
981 btrfs_set_trans_block_group(trans, inode);
982 btrfs_update_inode(trans, root, inode);
983 btrfs_end_transaction(trans, root);
984 mutex_unlock(&root->fs_info->fs_mutex);
985 btrfs_btree_balance_dirty(root);
986 }
987
988 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
989 struct btrfs_root *root,
990 u64 objectid,
991 struct btrfs_block_group_cache *group,
992 int mode)
993 {
994 struct inode *inode;
995 struct btrfs_inode_item inode_item;
996 struct btrfs_key *location;
997 int ret;
998 int owner;
999
1000 inode = new_inode(root->fs_info->sb);
1001 if (!inode)
1002 return ERR_PTR(-ENOMEM);
1003
1004 BTRFS_I(inode)->root = root;
1005 if (mode & S_IFDIR)
1006 owner = 0;
1007 else
1008 owner = 1;
1009 group = btrfs_find_block_group(root, group, 0, 0, owner);
1010 BTRFS_I(inode)->block_group = group;
1011
1012 inode->i_uid = current->fsuid;
1013 inode->i_gid = current->fsgid;
1014 inode->i_mode = mode;
1015 inode->i_ino = objectid;
1016 inode->i_blocks = 0;
1017 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1018 fill_inode_item(&inode_item, inode);
1019 location = &BTRFS_I(inode)->location;
1020 location->objectid = objectid;
1021 location->flags = 0;
1022 location->offset = 0;
1023 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1024
1025 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1026 BUG_ON(ret);
1027
1028 insert_inode_hash(inode);
1029 return inode;
1030 }
1031
1032 static inline u8 btrfs_inode_type(struct inode *inode)
1033 {
1034 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1035 }
1036
1037 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1038 struct dentry *dentry, struct inode *inode)
1039 {
1040 int ret;
1041 struct btrfs_key key;
1042 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1043 key.objectid = inode->i_ino;
1044 key.flags = 0;
1045 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1046 key.offset = 0;
1047
1048 ret = btrfs_insert_dir_item(trans, root,
1049 dentry->d_name.name, dentry->d_name.len,
1050 dentry->d_parent->d_inode->i_ino,
1051 &key, btrfs_inode_type(inode));
1052 if (ret == 0) {
1053 dentry->d_parent->d_inode->i_size += dentry->d_name.len * 2;
1054 ret = btrfs_update_inode(trans, root,
1055 dentry->d_parent->d_inode);
1056 }
1057 return ret;
1058 }
1059
1060 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1061 struct dentry *dentry, struct inode *inode)
1062 {
1063 int err = btrfs_add_link(trans, dentry, inode);
1064 if (!err) {
1065 d_instantiate(dentry, inode);
1066 return 0;
1067 }
1068 if (err > 0)
1069 err = -EEXIST;
1070 return err;
1071 }
1072
1073 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1074 int mode, struct nameidata *nd)
1075 {
1076 struct btrfs_trans_handle *trans;
1077 struct btrfs_root *root = BTRFS_I(dir)->root;
1078 struct inode *inode;
1079 int err;
1080 int drop_inode = 0;
1081 u64 objectid;
1082
1083 mutex_lock(&root->fs_info->fs_mutex);
1084 trans = btrfs_start_transaction(root, 1);
1085 btrfs_set_trans_block_group(trans, dir);
1086
1087 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1088 if (err) {
1089 err = -ENOSPC;
1090 goto out_unlock;
1091 }
1092
1093 inode = btrfs_new_inode(trans, root, objectid,
1094 BTRFS_I(dir)->block_group, mode);
1095 err = PTR_ERR(inode);
1096 if (IS_ERR(inode))
1097 goto out_unlock;
1098
1099 btrfs_set_trans_block_group(trans, inode);
1100 err = btrfs_add_nondir(trans, dentry, inode);
1101 if (err)
1102 drop_inode = 1;
1103 else {
1104 inode->i_mapping->a_ops = &btrfs_aops;
1105 inode->i_fop = &btrfs_file_operations;
1106 inode->i_op = &btrfs_file_inode_operations;
1107 }
1108 dir->i_sb->s_dirt = 1;
1109 btrfs_update_inode_block_group(trans, inode);
1110 btrfs_update_inode_block_group(trans, dir);
1111 out_unlock:
1112 btrfs_end_transaction(trans, root);
1113 mutex_unlock(&root->fs_info->fs_mutex);
1114
1115 if (drop_inode) {
1116 inode_dec_link_count(inode);
1117 iput(inode);
1118 }
1119 btrfs_btree_balance_dirty(root);
1120 return err;
1121 }
1122
1123 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1124 struct dentry *dentry)
1125 {
1126 struct btrfs_trans_handle *trans;
1127 struct btrfs_root *root = BTRFS_I(dir)->root;
1128 struct inode *inode = old_dentry->d_inode;
1129 int err;
1130 int drop_inode = 0;
1131
1132 if (inode->i_nlink == 0)
1133 return -ENOENT;
1134
1135 inc_nlink(inode);
1136 mutex_lock(&root->fs_info->fs_mutex);
1137 trans = btrfs_start_transaction(root, 1);
1138 btrfs_set_trans_block_group(trans, dir);
1139 atomic_inc(&inode->i_count);
1140 err = btrfs_add_nondir(trans, dentry, inode);
1141 if (err)
1142 drop_inode = 1;
1143 dir->i_sb->s_dirt = 1;
1144 btrfs_update_inode_block_group(trans, dir);
1145 btrfs_update_inode(trans, root, inode);
1146
1147 btrfs_end_transaction(trans, root);
1148 mutex_unlock(&root->fs_info->fs_mutex);
1149
1150 if (drop_inode) {
1151 inode_dec_link_count(inode);
1152 iput(inode);
1153 }
1154 btrfs_btree_balance_dirty(root);
1155 return err;
1156 }
1157
1158 static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
1159 struct btrfs_root *root,
1160 u64 objectid, u64 dirid)
1161 {
1162 int ret;
1163 char buf[2];
1164 struct btrfs_key key;
1165
1166 buf[0] = '.';
1167 buf[1] = '.';
1168
1169 key.objectid = objectid;
1170 key.offset = 0;
1171 key.flags = 0;
1172 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1173
1174 ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid,
1175 &key, BTRFS_FT_DIR);
1176 if (ret)
1177 goto error;
1178 key.objectid = dirid;
1179 ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid,
1180 &key, BTRFS_FT_DIR);
1181 if (ret)
1182 goto error;
1183 error:
1184 return ret;
1185 }
1186
1187 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1188 {
1189 struct inode *inode;
1190 struct btrfs_trans_handle *trans;
1191 struct btrfs_root *root = BTRFS_I(dir)->root;
1192 int err = 0;
1193 int drop_on_err = 0;
1194 u64 objectid;
1195
1196 mutex_lock(&root->fs_info->fs_mutex);
1197 trans = btrfs_start_transaction(root, 1);
1198 btrfs_set_trans_block_group(trans, dir);
1199 if (IS_ERR(trans)) {
1200 err = PTR_ERR(trans);
1201 goto out_unlock;
1202 }
1203
1204 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1205 if (err) {
1206 err = -ENOSPC;
1207 goto out_unlock;
1208 }
1209
1210 inode = btrfs_new_inode(trans, root, objectid,
1211 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1212 if (IS_ERR(inode)) {
1213 err = PTR_ERR(inode);
1214 goto out_fail;
1215 }
1216 drop_on_err = 1;
1217 inode->i_op = &btrfs_dir_inode_operations;
1218 inode->i_fop = &btrfs_dir_file_operations;
1219 btrfs_set_trans_block_group(trans, inode);
1220
1221 err = btrfs_make_empty_dir(trans, root, inode->i_ino, dir->i_ino);
1222 if (err)
1223 goto out_fail;
1224
1225 inode->i_size = 6;
1226 err = btrfs_update_inode(trans, root, inode);
1227 if (err)
1228 goto out_fail;
1229 err = btrfs_add_link(trans, dentry, inode);
1230 if (err)
1231 goto out_fail;
1232 d_instantiate(dentry, inode);
1233 drop_on_err = 0;
1234 dir->i_sb->s_dirt = 1;
1235 btrfs_update_inode_block_group(trans, inode);
1236 btrfs_update_inode_block_group(trans, dir);
1237
1238 out_fail:
1239 btrfs_end_transaction(trans, root);
1240 out_unlock:
1241 mutex_unlock(&root->fs_info->fs_mutex);
1242 if (drop_on_err)
1243 iput(inode);
1244 btrfs_btree_balance_dirty(root);
1245 return err;
1246 }
1247
1248 static int btrfs_sync_file(struct file *file,
1249 struct dentry *dentry, int datasync)
1250 {
1251 struct inode *inode = dentry->d_inode;
1252 struct btrfs_root *root = BTRFS_I(inode)->root;
1253 int ret;
1254 struct btrfs_trans_handle *trans;
1255
1256 mutex_lock(&root->fs_info->fs_mutex);
1257 trans = btrfs_start_transaction(root, 1);
1258 if (!trans) {
1259 ret = -ENOMEM;
1260 goto out;
1261 }
1262 ret = btrfs_commit_transaction(trans, root);
1263 mutex_unlock(&root->fs_info->fs_mutex);
1264 out:
1265 return ret > 0 ? EIO : ret;
1266 }
1267
1268 static int btrfs_sync_fs(struct super_block *sb, int wait)
1269 {
1270 struct btrfs_trans_handle *trans;
1271 struct btrfs_root *root;
1272 int ret;
1273 root = btrfs_sb(sb);
1274
1275 sb->s_dirt = 0;
1276 if (!wait) {
1277 filemap_flush(root->fs_info->btree_inode->i_mapping);
1278 return 0;
1279 }
1280 mutex_lock(&root->fs_info->fs_mutex);
1281 trans = btrfs_start_transaction(root, 1);
1282 ret = btrfs_commit_transaction(trans, root);
1283 sb->s_dirt = 0;
1284 BUG_ON(ret);
1285 printk("btrfs sync_fs\n");
1286 mutex_unlock(&root->fs_info->fs_mutex);
1287 return 0;
1288 }
1289
1290 #define BTRFS_GET_BLOCK_NO_CREATE 0
1291 #define BTRFS_GET_BLOCK_CREATE 1
1292 #define BTRFS_GET_BLOCK_NO_DIRECT 2
1293
1294 static int btrfs_get_block_lock(struct inode *inode, sector_t iblock,
1295 struct buffer_head *result, int create)
1296 {
1297 int ret;
1298 int err = 0;
1299 u64 blocknr;
1300 u64 extent_start = 0;
1301 u64 extent_end = 0;
1302 u64 objectid = inode->i_ino;
1303 u32 found_type;
1304 u64 alloc_hint = 0;
1305 struct btrfs_path *path;
1306 struct btrfs_root *root = BTRFS_I(inode)->root;
1307 struct btrfs_file_extent_item *item;
1308 struct btrfs_leaf *leaf;
1309 struct btrfs_disk_key *found_key;
1310 struct btrfs_trans_handle *trans = NULL;
1311
1312 path = btrfs_alloc_path();
1313 BUG_ON(!path);
1314 btrfs_init_path(path);
1315 if (create & BTRFS_GET_BLOCK_CREATE) {
1316 WARN_ON(1);
1317 /* this almost but not quite works */
1318 trans = btrfs_start_transaction(root, 1);
1319 if (!trans) {
1320 err = -ENOMEM;
1321 goto out;
1322 }
1323 ret = drop_extents(trans, root, inode,
1324 iblock << inode->i_blkbits,
1325 (iblock + 1) << inode->i_blkbits,
1326 &alloc_hint);
1327 BUG_ON(ret);
1328 }
1329
1330 ret = btrfs_lookup_file_extent(NULL, root, path,
1331 inode->i_ino,
1332 iblock << inode->i_blkbits, 0);
1333 if (ret < 0) {
1334 err = ret;
1335 goto out;
1336 }
1337
1338 if (ret != 0) {
1339 if (path->slots[0] == 0) {
1340 btrfs_release_path(root, path);
1341 goto not_found;
1342 }
1343 path->slots[0]--;
1344 }
1345
1346 item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
1347 struct btrfs_file_extent_item);
1348 leaf = btrfs_buffer_leaf(path->nodes[0]);
1349 blocknr = btrfs_file_extent_disk_blocknr(item);
1350 blocknr += btrfs_file_extent_offset(item);
1351
1352 /* are we inside the extent that was found? */
1353 found_key = &leaf->items[path->slots[0]].key;
1354 found_type = btrfs_disk_key_type(found_key);
1355 if (btrfs_disk_key_objectid(found_key) != objectid ||
1356 found_type != BTRFS_EXTENT_DATA_KEY) {
1357 extent_end = 0;
1358 extent_start = 0;
1359 goto not_found;
1360 }
1361 found_type = btrfs_file_extent_type(item);
1362 extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
1363 if (found_type == BTRFS_FILE_EXTENT_REG) {
1364 extent_start = extent_start >> inode->i_blkbits;
1365 extent_end = extent_start + btrfs_file_extent_num_blocks(item);
1366 err = 0;
1367 if (btrfs_file_extent_disk_blocknr(item) == 0)
1368 goto out;
1369 if (iblock >= extent_start && iblock < extent_end) {
1370 btrfs_map_bh_to_logical(root, result, blocknr +
1371 iblock - extent_start);
1372 goto out;
1373 }
1374 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1375 char *ptr;
1376 char *map;
1377 u32 size;
1378
1379 if (create & BTRFS_GET_BLOCK_NO_DIRECT) {
1380 err = -EINVAL;
1381 goto out;
1382 }
1383 size = btrfs_file_extent_inline_len(leaf->items +
1384 path->slots[0]);
1385 extent_end = (extent_start + size) >> inode->i_blkbits;
1386 extent_start >>= inode->i_blkbits;
1387 if (iblock < extent_start || iblock > extent_end) {
1388 goto not_found;
1389 }
1390 ptr = btrfs_file_extent_inline_start(item);
1391 map = kmap(result->b_page);
1392 memcpy(map, ptr, size);
1393 memset(map + size, 0, PAGE_CACHE_SIZE - size);
1394 flush_dcache_page(result->b_page);
1395 kunmap(result->b_page);
1396 set_buffer_uptodate(result);
1397 SetPageChecked(result->b_page);
1398 btrfs_map_bh_to_logical(root, result, 0);
1399 }
1400 not_found:
1401 if (create & BTRFS_GET_BLOCK_CREATE) {
1402 struct btrfs_key ins;
1403 ret = btrfs_alloc_extent(trans, root, inode->i_ino,
1404 1, alloc_hint, (u64)-1,
1405 &ins, 1);
1406 BUG_ON(ret);
1407 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
1408 iblock << inode->i_blkbits,
1409 ins.objectid, ins.offset,
1410 ins.offset);
1411 BUG_ON(ret);
1412 SetPageChecked(result->b_page);
1413 btrfs_map_bh_to_logical(root, result, ins.objectid);
1414 }
1415 out:
1416 if (trans)
1417 err = btrfs_end_transaction(trans, root);
1418 btrfs_free_path(path);
1419 return err;
1420 }
1421
1422 static int btrfs_get_block(struct inode *inode, sector_t iblock,
1423 struct buffer_head *result, int create)
1424 {
1425 int err;
1426 struct btrfs_root *root = BTRFS_I(inode)->root;
1427 mutex_lock(&root->fs_info->fs_mutex);
1428 err = btrfs_get_block_lock(inode, iblock, result, create);
1429 mutex_unlock(&root->fs_info->fs_mutex);
1430 return err;
1431 }
1432
1433 static int btrfs_get_block_bmap(struct inode *inode, sector_t iblock,
1434 struct buffer_head *result, int create)
1435 {
1436 struct btrfs_root *root = BTRFS_I(inode)->root;
1437 mutex_lock(&root->fs_info->fs_mutex);
1438 btrfs_get_block_lock(inode, iblock, result, BTRFS_GET_BLOCK_NO_DIRECT);
1439 mutex_unlock(&root->fs_info->fs_mutex);
1440 return 0;
1441 }
1442
1443 static sector_t btrfs_bmap(struct address_space *as, sector_t block)
1444 {
1445 return generic_block_bmap(as, block, btrfs_get_block_bmap);
1446 }
1447
1448 static int btrfs_prepare_write(struct file *file, struct page *page,
1449 unsigned from, unsigned to)
1450 {
1451 return block_prepare_write(page, from, to, btrfs_get_block);
1452 }
1453
1454 static void btrfs_write_super(struct super_block *sb)
1455 {
1456 sb->s_dirt = 0;
1457 }
1458
1459 static int btrfs_readpage(struct file *file, struct page *page)
1460 {
1461 return mpage_readpage(page, btrfs_get_block);
1462 }
1463
1464 /*
1465 * While block_write_full_page is writing back the dirty buffers under
1466 * the page lock, whoever dirtied the buffers may decide to clean them
1467 * again at any time. We handle that by only looking at the buffer
1468 * state inside lock_buffer().
1469 *
1470 * If block_write_full_page() is called for regular writeback
1471 * (wbc->sync_mode == WB_SYNC_NONE) then it will redirty a page which has a
1472 * locked buffer. This only can happen if someone has written the buffer
1473 * directly, with submit_bh(). At the address_space level PageWriteback
1474 * prevents this contention from occurring.
1475 */
1476 static int __btrfs_write_full_page(struct inode *inode, struct page *page,
1477 struct writeback_control *wbc)
1478 {
1479 int err;
1480 sector_t block;
1481 sector_t last_block;
1482 struct buffer_head *bh, *head;
1483 const unsigned blocksize = 1 << inode->i_blkbits;
1484 int nr_underway = 0;
1485
1486 BUG_ON(!PageLocked(page));
1487
1488 last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
1489
1490 if (!page_has_buffers(page)) {
1491 create_empty_buffers(page, blocksize,
1492 (1 << BH_Dirty)|(1 << BH_Uptodate));
1493 }
1494
1495 /*
1496 * Be very careful. We have no exclusion from __set_page_dirty_buffers
1497 * here, and the (potentially unmapped) buffers may become dirty at
1498 * any time. If a buffer becomes dirty here after we've inspected it
1499 * then we just miss that fact, and the page stays dirty.
1500 *
1501 * Buffers outside i_size may be dirtied by __set_page_dirty_buffers;
1502 * handle that here by just cleaning them.
1503 */
1504
1505 block = (sector_t)page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
1506 head = page_buffers(page);
1507 bh = head;
1508
1509 /*
1510 * Get all the dirty buffers mapped to disk addresses and
1511 * handle any aliases from the underlying blockdev's mapping.
1512 */
1513 do {
1514 if (block > last_block) {
1515 /*
1516 * mapped buffers outside i_size will occur, because
1517 * this page can be outside i_size when there is a
1518 * truncate in progress.
1519 */
1520 /*
1521 * The buffer was zeroed by block_write_full_page()
1522 */
1523 clear_buffer_dirty(bh);
1524 set_buffer_uptodate(bh);
1525 } else if (!buffer_mapped(bh) && buffer_dirty(bh)) {
1526 WARN_ON(bh->b_size != blocksize);
1527 err = btrfs_get_block(inode, block, bh, 0);
1528 if (err) {
1529 printk("writepage going to recovery err %d\n", err);
1530 goto recover;
1531 }
1532 if (buffer_new(bh)) {
1533 /* blockdev mappings never come here */
1534 clear_buffer_new(bh);
1535 }
1536 }
1537 bh = bh->b_this_page;
1538 block++;
1539 } while (bh != head);
1540
1541 do {
1542 if (!buffer_mapped(bh))
1543 continue;
1544 /*
1545 * If it's a fully non-blocking write attempt and we cannot
1546 * lock the buffer then redirty the page. Note that this can
1547 * potentially cause a busy-wait loop from pdflush and kswapd
1548 * activity, but those code paths have their own higher-level
1549 * throttling.
1550 */
1551 if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
1552 lock_buffer(bh);
1553 } else if (test_set_buffer_locked(bh)) {
1554 redirty_page_for_writepage(wbc, page);
1555 continue;
1556 }
1557 if (test_clear_buffer_dirty(bh) && bh->b_blocknr != 0) {
1558 mark_buffer_async_write(bh);
1559 } else {
1560 unlock_buffer(bh);
1561 }
1562 } while ((bh = bh->b_this_page) != head);
1563
1564 /*
1565 * The page and its buffers are protected by PageWriteback(), so we can
1566 * drop the bh refcounts early.
1567 */
1568 BUG_ON(PageWriteback(page));
1569 set_page_writeback(page);
1570
1571 do {
1572 struct buffer_head *next = bh->b_this_page;
1573 if (buffer_async_write(bh)) {
1574 submit_bh(WRITE, bh);
1575 nr_underway++;
1576 }
1577 bh = next;
1578 } while (bh != head);
1579 unlock_page(page);
1580
1581 err = 0;
1582 done:
1583 if (nr_underway == 0) {
1584 /*
1585 * The page was marked dirty, but the buffers were
1586 * clean. Someone wrote them back by hand with
1587 * ll_rw_block/submit_bh. A rare case.
1588 */
1589 int uptodate = 1;
1590 do {
1591 if (!buffer_uptodate(bh)) {
1592 uptodate = 0;
1593 break;
1594 }
1595 bh = bh->b_this_page;
1596 } while (bh != head);
1597 if (uptodate)
1598 SetPageUptodate(page);
1599 end_page_writeback(page);
1600 }
1601 return err;
1602
1603 recover:
1604 /*
1605 * ENOSPC, or some other error. We may already have added some
1606 * blocks to the file, so we need to write these out to avoid
1607 * exposing stale data.
1608 * The page is currently locked and not marked for writeback
1609 */
1610 bh = head;
1611 /* Recovery: lock and submit the mapped buffers */
1612 do {
1613 if (buffer_mapped(bh) && buffer_dirty(bh)) {
1614 lock_buffer(bh);
1615 mark_buffer_async_write(bh);
1616 } else {
1617 /*
1618 * The buffer may have been set dirty during
1619 * attachment to a dirty page.
1620 */
1621 clear_buffer_dirty(bh);
1622 }
1623 } while ((bh = bh->b_this_page) != head);
1624 SetPageError(page);
1625 BUG_ON(PageWriteback(page));
1626 set_page_writeback(page);
1627 do {
1628 struct buffer_head *next = bh->b_this_page;
1629 if (buffer_async_write(bh)) {
1630 clear_buffer_dirty(bh);
1631 submit_bh(WRITE, bh);
1632 nr_underway++;
1633 }
1634 bh = next;
1635 } while (bh != head);
1636 unlock_page(page);
1637 goto done;
1638 }
1639
1640 /*
1641 * The generic ->writepage function for buffer-backed address_spaces
1642 */
1643 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
1644 {
1645 struct inode * const inode = page->mapping->host;
1646 loff_t i_size = i_size_read(inode);
1647 const pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
1648 unsigned offset;
1649 void *kaddr;
1650
1651 /* Is the page fully inside i_size? */
1652 if (page->index < end_index)
1653 return __btrfs_write_full_page(inode, page, wbc);
1654
1655 /* Is the page fully outside i_size? (truncate in progress) */
1656 offset = i_size & (PAGE_CACHE_SIZE-1);
1657 if (page->index >= end_index+1 || !offset) {
1658 /*
1659 * The page may have dirty, unmapped buffers. For example,
1660 * they may have been added in ext3_writepage(). Make them
1661 * freeable here, so the page does not leak.
1662 */
1663 block_invalidatepage(page, 0);
1664 unlock_page(page);
1665 return 0; /* don't care */
1666 }
1667
1668 /*
1669 * The page straddles i_size. It must be zeroed out on each and every
1670 * writepage invokation because it may be mmapped. "A file is mapped
1671 * in multiples of the page size. For a file that is not a multiple of
1672 * the page size, the remaining memory is zeroed when mapped, and
1673 * writes to that region are not written out to the file."
1674 */
1675 kaddr = kmap_atomic(page, KM_USER0);
1676 memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
1677 flush_dcache_page(page);
1678 kunmap_atomic(kaddr, KM_USER0);
1679 return __btrfs_write_full_page(inode, page, wbc);
1680 }
1681
1682 static void btrfs_truncate(struct inode *inode)
1683 {
1684 struct btrfs_root *root = BTRFS_I(inode)->root;
1685 int ret;
1686 struct btrfs_trans_handle *trans;
1687
1688 if (!S_ISREG(inode->i_mode))
1689 return;
1690 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1691 return;
1692
1693 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1694
1695 mutex_lock(&root->fs_info->fs_mutex);
1696 trans = btrfs_start_transaction(root, 1);
1697 btrfs_set_trans_block_group(trans, inode);
1698
1699 /* FIXME, add redo link to tree so we don't leak on crash */
1700 ret = btrfs_truncate_in_trans(trans, root, inode);
1701 BUG_ON(ret);
1702 btrfs_update_inode(trans, root, inode);
1703 ret = btrfs_end_transaction(trans, root);
1704 BUG_ON(ret);
1705 mutex_unlock(&root->fs_info->fs_mutex);
1706 btrfs_btree_balance_dirty(root);
1707 }
1708
1709 static int btrfs_commit_write(struct file *file, struct page *page,
1710 unsigned from, unsigned to)
1711 {
1712 struct inode *inode = page->mapping->host;
1713 struct buffer_head *bh;
1714 loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
1715
1716 SetPageUptodate(page);
1717 bh = page_buffers(page);
1718 set_buffer_uptodate(bh);
1719 if (buffer_mapped(bh) && bh->b_blocknr != 0) {
1720 set_page_dirty(page);
1721 }
1722 if (pos > inode->i_size) {
1723 i_size_write(inode, pos);
1724 mark_inode_dirty(inode);
1725 }
1726 return 0;
1727 }
1728
1729 static int btrfs_copy_from_user(loff_t pos, int num_pages, int write_bytes,
1730 struct page **prepared_pages,
1731 const char __user * buf)
1732 {
1733 long page_fault = 0;
1734 int i;
1735 int offset = pos & (PAGE_CACHE_SIZE - 1);
1736
1737 for (i = 0; i < num_pages && write_bytes > 0; i++, offset = 0) {
1738 size_t count = min_t(size_t,
1739 PAGE_CACHE_SIZE - offset, write_bytes);
1740 struct page *page = prepared_pages[i];
1741 fault_in_pages_readable(buf, count);
1742
1743 /* Copy data from userspace to the current page */
1744 kmap(page);
1745 page_fault = __copy_from_user(page_address(page) + offset,
1746 buf, count);
1747 /* Flush processor's dcache for this page */
1748 flush_dcache_page(page);
1749 kunmap(page);
1750 buf += count;
1751 write_bytes -= count;
1752
1753 if (page_fault)
1754 break;
1755 }
1756 return page_fault ? -EFAULT : 0;
1757 }
1758
1759 static void btrfs_drop_pages(struct page **pages, size_t num_pages)
1760 {
1761 size_t i;
1762 for (i = 0; i < num_pages; i++) {
1763 if (!pages[i])
1764 break;
1765 unlock_page(pages[i]);
1766 mark_page_accessed(pages[i]);
1767 page_cache_release(pages[i]);
1768 }
1769 }
1770 static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
1771 struct btrfs_root *root,
1772 struct file *file,
1773 struct page **pages,
1774 size_t num_pages,
1775 loff_t pos,
1776 size_t write_bytes)
1777 {
1778 int i;
1779 int offset;
1780 int err = 0;
1781 int ret;
1782 int this_write;
1783 struct inode *inode = file->f_path.dentry->d_inode;
1784 struct buffer_head *bh;
1785 struct btrfs_file_extent_item *ei;
1786
1787 for (i = 0; i < num_pages; i++) {
1788 offset = pos & (PAGE_CACHE_SIZE -1);
1789 this_write = min(PAGE_CACHE_SIZE - offset, write_bytes);
1790 /* FIXME, one block at a time */
1791
1792 mutex_lock(&root->fs_info->fs_mutex);
1793 trans = btrfs_start_transaction(root, 1);
1794 btrfs_set_trans_block_group(trans, inode);
1795
1796 bh = page_buffers(pages[i]);
1797 if (buffer_mapped(bh) && bh->b_blocknr == 0) {
1798 struct btrfs_key key;
1799 struct btrfs_path *path;
1800 char *ptr;
1801 u32 datasize;
1802
1803 path = btrfs_alloc_path();
1804 BUG_ON(!path);
1805 key.objectid = inode->i_ino;
1806 key.offset = pages[i]->index << PAGE_CACHE_SHIFT;
1807 key.flags = 0;
1808 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
1809 BUG_ON(write_bytes >= PAGE_CACHE_SIZE);
1810 datasize = offset +
1811 btrfs_file_extent_calc_inline_size(write_bytes);
1812 ret = btrfs_insert_empty_item(trans, root, path, &key,
1813 datasize);
1814 BUG_ON(ret);
1815 ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
1816 path->slots[0], struct btrfs_file_extent_item);
1817 btrfs_set_file_extent_generation(ei, trans->transid);
1818 btrfs_set_file_extent_type(ei,
1819 BTRFS_FILE_EXTENT_INLINE);
1820 ptr = btrfs_file_extent_inline_start(ei);
1821 btrfs_memcpy(root, path->nodes[0]->b_data,
1822 ptr, bh->b_data, offset + write_bytes);
1823 mark_buffer_dirty(path->nodes[0]);
1824 btrfs_free_path(path);
1825 } else if (buffer_mapped(bh)) {
1826 btrfs_csum_file_block(trans, root, inode->i_ino,
1827 pages[i]->index << PAGE_CACHE_SHIFT,
1828 kmap(pages[i]), PAGE_CACHE_SIZE);
1829 kunmap(pages[i]);
1830 }
1831 SetPageChecked(pages[i]);
1832 // btrfs_update_inode_block_group(trans, inode);
1833 ret = btrfs_end_transaction(trans, root);
1834 BUG_ON(ret);
1835 mutex_unlock(&root->fs_info->fs_mutex);
1836
1837 ret = btrfs_commit_write(file, pages[i], offset,
1838 offset + this_write);
1839 pos += this_write;
1840 if (ret) {
1841 err = ret;
1842 goto failed;
1843 }
1844 WARN_ON(this_write > write_bytes);
1845 write_bytes -= this_write;
1846 }
1847 failed:
1848 return err;
1849 }
1850
1851 static int drop_extents(struct btrfs_trans_handle *trans,
1852 struct btrfs_root *root,
1853 struct inode *inode,
1854 u64 start, u64 end, u64 *hint_block)
1855 {
1856 int ret;
1857 struct btrfs_key key;
1858 struct btrfs_leaf *leaf;
1859 int slot;
1860 struct btrfs_file_extent_item *extent;
1861 u64 extent_end = 0;
1862 int keep;
1863 struct btrfs_file_extent_item old;
1864 struct btrfs_path *path;
1865 u64 search_start = start;
1866 int bookend;
1867 int found_type;
1868 int found_extent;
1869 int found_inline;
1870
1871 path = btrfs_alloc_path();
1872 if (!path)
1873 return -ENOMEM;
1874 while(1) {
1875 btrfs_release_path(root, path);
1876 ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
1877 search_start, -1);
1878 if (ret < 0)
1879 goto out;
1880 if (ret > 0) {
1881 if (path->slots[0] == 0) {
1882 ret = 0;
1883 goto out;
1884 }
1885 path->slots[0]--;
1886 }
1887 keep = 0;
1888 bookend = 0;
1889 found_extent = 0;
1890 found_inline = 0;
1891 extent = NULL;
1892 leaf = btrfs_buffer_leaf(path->nodes[0]);
1893 slot = path->slots[0];
1894 btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key);
1895 if (key.offset >= end || key.objectid != inode->i_ino) {
1896 ret = 0;
1897 goto out;
1898 }
1899 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) {
1900 ret = 0;
1901 goto out;
1902 }
1903 extent = btrfs_item_ptr(leaf, slot,
1904 struct btrfs_file_extent_item);
1905 found_type = btrfs_file_extent_type(extent);
1906 if (found_type == BTRFS_FILE_EXTENT_REG) {
1907 extent_end = key.offset +
1908 (btrfs_file_extent_num_blocks(extent) <<
1909 inode->i_blkbits);
1910 found_extent = 1;
1911 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1912 found_inline = 1;
1913 extent_end = key.offset +
1914 btrfs_file_extent_inline_len(leaf->items + slot);
1915 }
1916
1917 if (!found_extent && !found_inline) {
1918 ret = 0;
1919 goto out;
1920 }
1921
1922 if (search_start >= extent_end) {
1923 ret = 0;
1924 goto out;
1925 }
1926
1927 if (found_inline) {
1928 u64 mask = root->blocksize - 1;
1929 search_start = (extent_end + mask) & ~mask;
1930 } else
1931 search_start = extent_end;
1932
1933 if (end < extent_end && end >= key.offset) {
1934 if (found_extent) {
1935 u64 disk_blocknr =
1936 btrfs_file_extent_disk_blocknr(extent);
1937 u64 disk_num_blocks =
1938 btrfs_file_extent_disk_num_blocks(extent);
1939 memcpy(&old, extent, sizeof(old));
1940 if (disk_blocknr != 0) {
1941 ret = btrfs_inc_extent_ref(trans, root,
1942 disk_blocknr, disk_num_blocks);
1943 BUG_ON(ret);
1944 }
1945 }
1946 WARN_ON(found_inline);
1947 bookend = 1;
1948 }
1949
1950 if (start > key.offset) {
1951 u64 new_num;
1952 u64 old_num;
1953 /* truncate existing extent */
1954 keep = 1;
1955 WARN_ON(start & (root->blocksize - 1));
1956 if (found_extent) {
1957 new_num = (start - key.offset) >>
1958 inode->i_blkbits;
1959 old_num = btrfs_file_extent_num_blocks(extent);
1960 *hint_block =
1961 btrfs_file_extent_disk_blocknr(extent);
1962 if (btrfs_file_extent_disk_blocknr(extent)) {
1963 inode->i_blocks -=
1964 (old_num - new_num) << 3;
1965 }
1966 btrfs_set_file_extent_num_blocks(extent,
1967 new_num);
1968 mark_buffer_dirty(path->nodes[0]);
1969 } else {
1970 WARN_ON(1);
1971 }
1972 }
1973 if (!keep) {
1974 u64 disk_blocknr = 0;
1975 u64 disk_num_blocks = 0;
1976 u64 extent_num_blocks = 0;
1977 if (found_extent) {
1978 disk_blocknr =
1979 btrfs_file_extent_disk_blocknr(extent);
1980 disk_num_blocks =
1981 btrfs_file_extent_disk_num_blocks(extent);
1982 extent_num_blocks =
1983 btrfs_file_extent_num_blocks(extent);
1984 *hint_block =
1985 btrfs_file_extent_disk_blocknr(extent);
1986 }
1987 ret = btrfs_del_item(trans, root, path);
1988 BUG_ON(ret);
1989 btrfs_release_path(root, path);
1990 extent = NULL;
1991 if (found_extent && disk_blocknr != 0) {
1992 inode->i_blocks -= extent_num_blocks << 3;
1993 ret = btrfs_free_extent(trans, root,
1994 disk_blocknr,
1995 disk_num_blocks, 0);
1996 }
1997
1998 BUG_ON(ret);
1999 if (!bookend && search_start >= end) {
2000 ret = 0;
2001 goto out;
2002 }
2003 if (!bookend)
2004 continue;
2005 }
2006 if (bookend && found_extent) {
2007 /* create bookend */
2008 struct btrfs_key ins;
2009 ins.objectid = inode->i_ino;
2010 ins.offset = end;
2011 ins.flags = 0;
2012 btrfs_set_key_type(&ins, BTRFS_EXTENT_DATA_KEY);
2013
2014 btrfs_release_path(root, path);
2015 ret = btrfs_insert_empty_item(trans, root, path, &ins,
2016 sizeof(*extent));
2017 BUG_ON(ret);
2018 extent = btrfs_item_ptr(
2019 btrfs_buffer_leaf(path->nodes[0]),
2020 path->slots[0],
2021 struct btrfs_file_extent_item);
2022 btrfs_set_file_extent_disk_blocknr(extent,
2023 btrfs_file_extent_disk_blocknr(&old));
2024 btrfs_set_file_extent_disk_num_blocks(extent,
2025 btrfs_file_extent_disk_num_blocks(&old));
2026
2027 btrfs_set_file_extent_offset(extent,
2028 btrfs_file_extent_offset(&old) +
2029 ((end - key.offset) >> inode->i_blkbits));
2030 WARN_ON(btrfs_file_extent_num_blocks(&old) <
2031 (extent_end - end) >> inode->i_blkbits);
2032 btrfs_set_file_extent_num_blocks(extent,
2033 (extent_end - end) >> inode->i_blkbits);
2034
2035 btrfs_set_file_extent_type(extent,
2036 BTRFS_FILE_EXTENT_REG);
2037 btrfs_set_file_extent_generation(extent,
2038 btrfs_file_extent_generation(&old));
2039 btrfs_mark_buffer_dirty(path->nodes[0]);
2040 if (btrfs_file_extent_disk_blocknr(&old) != 0) {
2041 inode->i_blocks +=
2042 btrfs_file_extent_num_blocks(extent) << 3;
2043 }
2044 ret = 0;
2045 goto out;
2046 }
2047 }
2048 out:
2049 btrfs_free_path(path);
2050 return ret;
2051 }
2052
2053 static int prepare_pages(struct btrfs_root *root,
2054 struct file *file,
2055 struct page **pages,
2056 size_t num_pages,
2057 loff_t pos,
2058 unsigned long first_index,
2059 unsigned long last_index,
2060 size_t write_bytes,
2061 u64 alloc_extent_start)
2062 {
2063 int i;
2064 unsigned long index = pos >> PAGE_CACHE_SHIFT;
2065 struct inode *inode = file->f_path.dentry->d_inode;
2066 int offset;
2067 int err = 0;
2068 int this_write;
2069 struct buffer_head *bh;
2070 struct buffer_head *head;
2071 loff_t isize = i_size_read(inode);
2072
2073 memset(pages, 0, num_pages * sizeof(struct page *));
2074
2075 for (i = 0; i < num_pages; i++) {
2076 pages[i] = grab_cache_page(inode->i_mapping, index + i);
2077 if (!pages[i]) {
2078 err = -ENOMEM;
2079 goto failed_release;
2080 }
2081 cancel_dirty_page(pages[i], PAGE_CACHE_SIZE);
2082 wait_on_page_writeback(pages[i]);
2083 offset = pos & (PAGE_CACHE_SIZE -1);
2084 this_write = min(PAGE_CACHE_SIZE - offset, write_bytes);
2085 if (!page_has_buffers(pages[i])) {
2086 create_empty_buffers(pages[i],
2087 root->fs_info->sb->s_blocksize,
2088 (1 << BH_Uptodate));
2089 }
2090 head = page_buffers(pages[i]);
2091 bh = head;
2092 do {
2093 err = btrfs_map_bh_to_logical(root, bh,
2094 alloc_extent_start);
2095 BUG_ON(err);
2096 if (err)
2097 goto failed_truncate;
2098 bh = bh->b_this_page;
2099 if (alloc_extent_start)
2100 alloc_extent_start++;
2101 } while (bh != head);
2102 pos += this_write;
2103 WARN_ON(this_write > write_bytes);
2104 write_bytes -= this_write;
2105 }
2106 return 0;
2107
2108 failed_release:
2109 btrfs_drop_pages(pages, num_pages);
2110 return err;
2111
2112 failed_truncate:
2113 btrfs_drop_pages(pages, num_pages);
2114 if (pos > isize)
2115 vmtruncate(inode, isize);
2116 return err;
2117 }
2118
2119 static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
2120 size_t count, loff_t *ppos)
2121 {
2122 loff_t pos;
2123 size_t num_written = 0;
2124 int err = 0;
2125 int ret = 0;
2126 struct inode *inode = file->f_path.dentry->d_inode;
2127 struct btrfs_root *root = BTRFS_I(inode)->root;
2128 struct page *pages[8];
2129 struct page *pinned[2];
2130 unsigned long first_index;
2131 unsigned long last_index;
2132 u64 start_pos;
2133 u64 num_blocks;
2134 u64 alloc_extent_start;
2135 u64 hint_block;
2136 struct btrfs_trans_handle *trans;
2137 struct btrfs_key ins;
2138 pinned[0] = NULL;
2139 pinned[1] = NULL;
2140 if (file->f_flags & O_DIRECT)
2141 return -EINVAL;
2142 pos = *ppos;
2143 vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
2144 current->backing_dev_info = inode->i_mapping->backing_dev_info;
2145 err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
2146 if (err)
2147 goto out;
2148 if (count == 0)
2149 goto out;
2150 err = remove_suid(file->f_path.dentry);
2151 if (err)
2152 goto out;
2153 file_update_time(file);
2154
2155 start_pos = pos & ~((u64)PAGE_CACHE_SIZE - 1);
2156 num_blocks = (count + pos - start_pos + root->blocksize - 1) >>
2157 inode->i_blkbits;
2158
2159 mutex_lock(&inode->i_mutex);
2160 first_index = pos >> PAGE_CACHE_SHIFT;
2161 last_index = (pos + count) >> PAGE_CACHE_SHIFT;
2162
2163 if ((pos & (PAGE_CACHE_SIZE - 1))) {
2164 pinned[0] = grab_cache_page(inode->i_mapping, first_index);
2165 if (!PageUptodate(pinned[0])) {
2166 ret = mpage_readpage(pinned[0], btrfs_get_block);
2167 BUG_ON(ret);
2168 wait_on_page_locked(pinned[0]);
2169 } else {
2170 unlock_page(pinned[0]);
2171 }
2172 }
2173 if ((pos + count) & (PAGE_CACHE_SIZE - 1)) {
2174 pinned[1] = grab_cache_page(inode->i_mapping, last_index);
2175 if (!PageUptodate(pinned[1])) {
2176 ret = mpage_readpage(pinned[1], btrfs_get_block);
2177 BUG_ON(ret);
2178 wait_on_page_locked(pinned[1]);
2179 } else {
2180 unlock_page(pinned[1]);
2181 }
2182 }
2183
2184 mutex_lock(&root->fs_info->fs_mutex);
2185 trans = btrfs_start_transaction(root, 1);
2186 if (!trans) {
2187 err = -ENOMEM;
2188 mutex_unlock(&root->fs_info->fs_mutex);
2189 goto out_unlock;
2190 }
2191 btrfs_set_trans_block_group(trans, inode);
2192 /* FIXME blocksize != 4096 */
2193 inode->i_blocks += num_blocks << 3;
2194 hint_block = 0;
2195 if (start_pos < inode->i_size) {
2196 /* FIXME blocksize != pagesize */
2197 ret = drop_extents(trans, root, inode,
2198 start_pos,
2199 (pos + count + root->blocksize -1) &
2200 ~((u64)root->blocksize - 1), &hint_block);
2201 BUG_ON(ret);
2202 }
2203 if (inode->i_size < start_pos) {
2204 u64 last_pos_in_file;
2205 u64 hole_size;
2206 u64 mask = root->blocksize - 1;
2207 last_pos_in_file = (inode->i_size + mask) & ~mask;
2208 hole_size = (start_pos - last_pos_in_file + mask) & ~mask;
2209 hole_size >>= inode->i_blkbits;
2210 if (last_pos_in_file < start_pos) {
2211 ret = btrfs_insert_file_extent(trans, root,
2212 inode->i_ino,
2213 last_pos_in_file,
2214 0, 0, hole_size);
2215 }
2216 BUG_ON(ret);
2217 }
2218 if (inode->i_size >= PAGE_CACHE_SIZE || pos + count < inode->i_size ||
2219 pos + count - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) {
2220 ret = btrfs_alloc_extent(trans, root, inode->i_ino,
2221 num_blocks, hint_block, (u64)-1,
2222 &ins, 1);
2223 BUG_ON(ret);
2224 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
2225 start_pos, ins.objectid, ins.offset,
2226 ins.offset);
2227 BUG_ON(ret);
2228 } else {
2229 ins.offset = 0;
2230 ins.objectid = 0;
2231 }
2232 BUG_ON(ret);
2233 alloc_extent_start = ins.objectid;
2234 // btrfs_update_inode_block_group(trans, inode);
2235 ret = btrfs_end_transaction(trans, root);
2236 mutex_unlock(&root->fs_info->fs_mutex);
2237
2238 while(count > 0) {
2239 size_t offset = pos & (PAGE_CACHE_SIZE - 1);
2240 size_t write_bytes = min(count, PAGE_CACHE_SIZE - offset);
2241 size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >>
2242 PAGE_CACHE_SHIFT;
2243
2244 memset(pages, 0, sizeof(pages));
2245 ret = prepare_pages(root, file, pages, num_pages,
2246 pos, first_index, last_index,
2247 write_bytes, alloc_extent_start);
2248 BUG_ON(ret);
2249
2250 /* FIXME blocks != pagesize */
2251 if (alloc_extent_start)
2252 alloc_extent_start += num_pages;
2253 ret = btrfs_copy_from_user(pos, num_pages,
2254 write_bytes, pages, buf);
2255 BUG_ON(ret);
2256
2257 ret = dirty_and_release_pages(NULL, root, file, pages,
2258 num_pages, pos, write_bytes);
2259 BUG_ON(ret);
2260 btrfs_drop_pages(pages, num_pages);
2261
2262 buf += write_bytes;
2263 count -= write_bytes;
2264 pos += write_bytes;
2265 num_written += write_bytes;
2266
2267 balance_dirty_pages_ratelimited(inode->i_mapping);
2268 btrfs_btree_balance_dirty(root);
2269 cond_resched();
2270 }
2271 out_unlock:
2272 mutex_unlock(&inode->i_mutex);
2273 out:
2274 if (pinned[0])
2275 page_cache_release(pinned[0]);
2276 if (pinned[1])
2277 page_cache_release(pinned[1]);
2278 *ppos = pos;
2279 current->backing_dev_info = NULL;
2280 mark_inode_dirty(inode);
2281 return num_written ? num_written : err;
2282 }
2283
2284 static int btrfs_read_actor(read_descriptor_t *desc, struct page *page,
2285 unsigned long offset, unsigned long size)
2286 {
2287 char *kaddr;
2288 unsigned long left, count = desc->count;
2289 struct inode *inode = page->mapping->host;
2290
2291 if (size > count)
2292 size = count;
2293
2294 if (!PageChecked(page)) {
2295 /* FIXME, do it per block */
2296 struct btrfs_root *root = BTRFS_I(inode)->root;
2297 int ret;
2298 struct buffer_head *bh;
2299
2300 if (page_has_buffers(page)) {
2301 bh = page_buffers(page);
2302 if (!buffer_mapped(bh)) {
2303 SetPageChecked(page);
2304 goto checked;
2305 }
2306 }
2307
2308 ret = btrfs_csum_verify_file_block(root,
2309 page->mapping->host->i_ino,
2310 page->index << PAGE_CACHE_SHIFT,
2311 kmap(page), PAGE_CACHE_SIZE);
2312 if (ret) {
2313 if (ret != -ENOENT) {
2314 printk("failed to verify ino %lu page %lu ret %d\n",
2315 page->mapping->host->i_ino,
2316 page->index, ret);
2317 memset(page_address(page), 1, PAGE_CACHE_SIZE);
2318 flush_dcache_page(page);
2319 }
2320 }
2321 SetPageChecked(page);
2322 kunmap(page);
2323 }
2324 checked:
2325 /*
2326 * Faults on the destination of a read are common, so do it before
2327 * taking the kmap.
2328 */
2329 if (!fault_in_pages_writeable(desc->arg.buf, size)) {
2330 kaddr = kmap_atomic(page, KM_USER0);
2331 left = __copy_to_user_inatomic(desc->arg.buf,
2332 kaddr + offset, size);
2333 kunmap_atomic(kaddr, KM_USER0);
2334 if (left == 0)
2335 goto success;
2336 }
2337
2338 /* Do it the slow way */
2339 kaddr = kmap(page);
2340 left = __copy_to_user(desc->arg.buf, kaddr + offset, size);
2341 kunmap(page);
2342
2343 if (left) {
2344 size -= left;
2345 desc->error = -EFAULT;
2346 }
2347 success:
2348 desc->count = count - size;
2349 desc->written += size;
2350 desc->arg.buf += size;
2351 return size;
2352 }
2353
2354 /**
2355 * btrfs_file_aio_read - filesystem read routine
2356 * @iocb: kernel I/O control block
2357 * @iov: io vector request
2358 * @nr_segs: number of segments in the iovec
2359 * @pos: current file position
2360 */
2361 static ssize_t btrfs_file_aio_read(struct kiocb *iocb, const struct iovec *iov,
2362 unsigned long nr_segs, loff_t pos)
2363 {
2364 struct file *filp = iocb->ki_filp;
2365 ssize_t retval;
2366 unsigned long seg;
2367 size_t count;
2368 loff_t *ppos = &iocb->ki_pos;
2369
2370 count = 0;
2371 for (seg = 0; seg < nr_segs; seg++) {
2372 const struct iovec *iv = &iov[seg];
2373
2374 /*
2375 * If any segment has a negative length, or the cumulative
2376 * length ever wraps negative then return -EINVAL.
2377 */
2378 count += iv->iov_len;
2379 if (unlikely((ssize_t)(count|iv->iov_len) < 0))
2380 return -EINVAL;
2381 if (access_ok(VERIFY_WRITE, iv->iov_base, iv->iov_len))
2382 continue;
2383 if (seg == 0)
2384 return -EFAULT;
2385 nr_segs = seg;
2386 count -= iv->iov_len; /* This segment is no good */
2387 break;
2388 }
2389 retval = 0;
2390 if (count) {
2391 for (seg = 0; seg < nr_segs; seg++) {
2392 read_descriptor_t desc;
2393
2394 desc.written = 0;
2395 desc.arg.buf = iov[seg].iov_base;
2396 desc.count = iov[seg].iov_len;
2397 if (desc.count == 0)
2398 continue;
2399 desc.error = 0;
2400 do_generic_file_read(filp, ppos, &desc,
2401 btrfs_read_actor);
2402 retval += desc.written;
2403 if (desc.error) {
2404 retval = retval ?: desc.error;
2405 break;
2406 }
2407 }
2408 }
2409 return retval;
2410 }
2411
2412 static int create_subvol(struct btrfs_root *root, char *name, int namelen)
2413 {
2414 struct btrfs_trans_handle *trans;
2415 struct btrfs_key key;
2416 struct btrfs_root_item root_item;
2417 struct btrfs_inode_item *inode_item;
2418 struct buffer_head *subvol;
2419 struct btrfs_leaf *leaf;
2420 struct btrfs_root *new_root;
2421 struct inode *inode;
2422 struct inode *dir;
2423 int ret;
2424 u64 objectid;
2425 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
2426
2427 mutex_lock(&root->fs_info->fs_mutex);
2428 trans = btrfs_start_transaction(root, 1);
2429 BUG_ON(!trans);
2430
2431 subvol = btrfs_alloc_free_block(trans, root, 0);
2432 if (subvol == NULL)
2433 return -ENOSPC;
2434 leaf = btrfs_buffer_leaf(subvol);
2435 btrfs_set_header_nritems(&leaf->header, 0);
2436 btrfs_set_header_level(&leaf->header, 0);
2437 btrfs_set_header_blocknr(&leaf->header, bh_blocknr(subvol));
2438 btrfs_set_header_generation(&leaf->header, trans->transid);
2439 btrfs_set_header_owner(&leaf->header, root->root_key.objectid);
2440 memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid,
2441 sizeof(leaf->header.fsid));
2442 mark_buffer_dirty(subvol);
2443
2444 inode_item = &root_item.inode;
2445 memset(inode_item, 0, sizeof(*inode_item));
2446 btrfs_set_inode_generation(inode_item, 1);
2447 btrfs_set_inode_size(inode_item, 3);
2448 btrfs_set_inode_nlink(inode_item, 1);
2449 btrfs_set_inode_nblocks(inode_item, 1);
2450 btrfs_set_inode_mode(inode_item, S_IFDIR | 0755);
2451
2452 btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol));
2453 btrfs_set_root_refs(&root_item, 1);
2454 brelse(subvol);
2455 subvol = NULL;
2456
2457 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
2458 0, &objectid);
2459 BUG_ON(ret);
2460
2461 btrfs_set_root_dirid(&root_item, new_dirid);
2462
2463 key.objectid = objectid;
2464 key.offset = 1;
2465 key.flags = 0;
2466 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
2467 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
2468 &root_item);
2469 BUG_ON(ret);
2470
2471 /*
2472 * insert the directory item
2473 */
2474 key.offset = (u64)-1;
2475 dir = root->fs_info->sb->s_root->d_inode;
2476 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
2477 name, namelen, dir->i_ino, &key,
2478 BTRFS_FT_DIR);
2479 BUG_ON(ret);
2480
2481 ret = btrfs_commit_transaction(trans, root);
2482 BUG_ON(ret);
2483
2484 new_root = btrfs_read_fs_root(root->fs_info, &key);
2485 BUG_ON(!new_root);
2486
2487 trans = btrfs_start_transaction(new_root, 1);
2488 BUG_ON(!trans);
2489
2490 inode = btrfs_new_inode(trans, new_root, new_dirid,
2491 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
2492 inode->i_op = &btrfs_dir_inode_operations;
2493 inode->i_fop = &btrfs_dir_file_operations;
2494
2495 ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid);
2496 BUG_ON(ret);
2497
2498 inode->i_nlink = 1;
2499 inode->i_size = 6;
2500 ret = btrfs_update_inode(trans, new_root, inode);
2501 BUG_ON(ret);
2502
2503 ret = btrfs_commit_transaction(trans, new_root);
2504 BUG_ON(ret);
2505
2506 iput(inode);
2507
2508 mutex_unlock(&root->fs_info->fs_mutex);
2509 btrfs_btree_balance_dirty(root);
2510 return 0;
2511 }
2512
2513 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
2514 {
2515 struct btrfs_trans_handle *trans;
2516 struct btrfs_key key;
2517 struct btrfs_root_item new_root_item;
2518 int ret;
2519 u64 objectid;
2520
2521 if (!root->ref_cows)
2522 return -EINVAL;
2523
2524 mutex_lock(&root->fs_info->fs_mutex);
2525 trans = btrfs_start_transaction(root, 1);
2526 BUG_ON(!trans);
2527
2528 ret = btrfs_update_inode(trans, root, root->inode);
2529 BUG_ON(ret);
2530
2531 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
2532 0, &objectid);
2533 BUG_ON(ret);
2534
2535 memcpy(&new_root_item, &root->root_item,
2536 sizeof(new_root_item));
2537
2538 key.objectid = objectid;
2539 key.offset = 1;
2540 key.flags = 0;
2541 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
2542 btrfs_set_root_blocknr(&new_root_item, bh_blocknr(root->node));
2543
2544 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
2545 &new_root_item);
2546 BUG_ON(ret);
2547
2548 /*
2549 * insert the directory item
2550 */
2551 key.offset = (u64)-1;
2552 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
2553 name, namelen,
2554 root->fs_info->sb->s_root->d_inode->i_ino,
2555 &key, BTRFS_FT_DIR);
2556
2557 BUG_ON(ret);
2558
2559 ret = btrfs_inc_root_ref(trans, root);
2560 BUG_ON(ret);
2561
2562 ret = btrfs_commit_transaction(trans, root);
2563 BUG_ON(ret);
2564 mutex_unlock(&root->fs_info->fs_mutex);
2565 btrfs_btree_balance_dirty(root);
2566 return 0;
2567 }
2568
2569 static int add_disk(struct btrfs_root *root, char *name, int namelen)
2570 {
2571 struct block_device *bdev;
2572 struct btrfs_path *path;
2573 struct super_block *sb = root->fs_info->sb;
2574 struct btrfs_root *dev_root = root->fs_info->dev_root;
2575 struct btrfs_trans_handle *trans;
2576 struct btrfs_device_item *dev_item;
2577 struct btrfs_key key;
2578 u16 item_size;
2579 u64 num_blocks;
2580 u64 new_blocks;
2581 u64 device_id;
2582 int ret;
2583
2584 printk("adding disk %s\n", name);
2585 path = btrfs_alloc_path();
2586 if (!path)
2587 return -ENOMEM;
2588 num_blocks = btrfs_super_total_blocks(root->fs_info->disk_super);
2589 bdev = open_bdev_excl(name, O_RDWR, sb);
2590 if (IS_ERR(bdev)) {
2591 ret = PTR_ERR(bdev);
2592 printk("open bdev excl failed ret %d\n", ret);
2593 goto out_nolock;
2594 }
2595 set_blocksize(bdev, sb->s_blocksize);
2596 new_blocks = bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2597 key.objectid = num_blocks;
2598 key.offset = new_blocks;
2599 key.flags = 0;
2600 btrfs_set_key_type(&key, BTRFS_DEV_ITEM_KEY);
2601
2602 mutex_lock(&dev_root->fs_info->fs_mutex);
2603 trans = btrfs_start_transaction(dev_root, 1);
2604 item_size = sizeof(*dev_item) + namelen;
2605 printk("insert empty on %Lu %Lu %u size %d\n", num_blocks, new_blocks, key.flags, item_size);
2606 ret = btrfs_insert_empty_item(trans, dev_root, path, &key, item_size);
2607 if (ret) {
2608 printk("insert failed %d\n", ret);
2609 close_bdev_excl(bdev);
2610 if (ret > 0)
2611 ret = -EEXIST;
2612 goto out;
2613 }
2614 dev_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
2615 path->slots[0], struct btrfs_device_item);
2616 btrfs_set_device_pathlen(dev_item, namelen);
2617 memcpy(dev_item + 1, name, namelen);
2618
2619 device_id = btrfs_super_last_device_id(root->fs_info->disk_super) + 1;
2620 btrfs_set_super_last_device_id(root->fs_info->disk_super, device_id);
2621 btrfs_set_device_id(dev_item, device_id);
2622 mark_buffer_dirty(path->nodes[0]);
2623
2624 ret = btrfs_insert_dev_radix(root, bdev, device_id, num_blocks,
2625 new_blocks);
2626
2627 if (!ret) {
2628 btrfs_set_super_total_blocks(root->fs_info->disk_super,
2629 num_blocks + new_blocks);
2630 i_size_write(root->fs_info->btree_inode,
2631 (num_blocks + new_blocks) <<
2632 root->fs_info->btree_inode->i_blkbits);
2633 }
2634
2635 out:
2636 ret = btrfs_commit_transaction(trans, dev_root);
2637 BUG_ON(ret);
2638 mutex_unlock(&root->fs_info->fs_mutex);
2639 out_nolock:
2640 btrfs_free_path(path);
2641 btrfs_btree_balance_dirty(root);
2642
2643 return ret;
2644 }
2645
2646 static int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int
2647 cmd, unsigned long arg)
2648 {
2649 struct btrfs_root *root = BTRFS_I(inode)->root;
2650 struct btrfs_ioctl_vol_args vol_args;
2651 int ret = 0;
2652 struct btrfs_dir_item *di;
2653 int namelen;
2654 struct btrfs_path *path;
2655 u64 root_dirid;
2656
2657 switch (cmd) {
2658 case BTRFS_IOC_SNAP_CREATE:
2659 if (copy_from_user(&vol_args,
2660 (struct btrfs_ioctl_vol_args __user *)arg,
2661 sizeof(vol_args)))
2662 return -EFAULT;
2663 namelen = strlen(vol_args.name);
2664 if (namelen > BTRFS_VOL_NAME_MAX)
2665 return -EINVAL;
2666 path = btrfs_alloc_path();
2667 if (!path)
2668 return -ENOMEM;
2669 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2670 mutex_lock(&root->fs_info->fs_mutex);
2671 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2672 path, root_dirid,
2673 vol_args.name, namelen, 0);
2674 mutex_unlock(&root->fs_info->fs_mutex);
2675 btrfs_free_path(path);
2676 if (di && !IS_ERR(di))
2677 return -EEXIST;
2678
2679 if (root == root->fs_info->tree_root)
2680 ret = create_subvol(root, vol_args.name, namelen);
2681 else
2682 ret = create_snapshot(root, vol_args.name, namelen);
2683 WARN_ON(ret);
2684 break;
2685 case BTRFS_IOC_ADD_DISK:
2686 if (copy_from_user(&vol_args,
2687 (struct btrfs_ioctl_vol_args __user *)arg,
2688 sizeof(vol_args)))
2689 return -EFAULT;
2690 namelen = strlen(vol_args.name);
2691 if (namelen > BTRFS_VOL_NAME_MAX)
2692 return -EINVAL;
2693 vol_args.name[namelen] = '\0';
2694 ret = add_disk(root, vol_args.name, namelen);
2695 break;
2696 default:
2697 return -ENOTTY;
2698 }
2699 return ret;
2700 }
2701
2702 #ifdef CONFIG_COMPAT
2703 static long btrfs_compat_ioctl(struct file *file, unsigned int cmd,
2704 unsigned long arg)
2705 {
2706 struct inode *inode = file->f_path.dentry->d_inode;
2707 int ret;
2708 lock_kernel();
2709 ret = btrfs_ioctl(inode, file, cmd, (unsigned long) compat_ptr(arg));
2710 unlock_kernel();
2711 return ret;
2712
2713 }
2714 #endif
2715
2716 static struct kmem_cache *btrfs_inode_cachep;
2717 struct kmem_cache *btrfs_trans_handle_cachep;
2718 struct kmem_cache *btrfs_transaction_cachep;
2719 struct kmem_cache *btrfs_bit_radix_cachep;
2720 struct kmem_cache *btrfs_path_cachep;
2721
2722 /*
2723 * Called inside transaction, so use GFP_NOFS
2724 */
2725 static struct inode *btrfs_alloc_inode(struct super_block *sb)
2726 {
2727 struct btrfs_inode *ei;
2728
2729 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2730 if (!ei)
2731 return NULL;
2732 return &ei->vfs_inode;
2733 }
2734
2735 static void btrfs_destroy_inode(struct inode *inode)
2736 {
2737 WARN_ON(!list_empty(&inode->i_dentry));
2738 WARN_ON(inode->i_data.nrpages);
2739
2740 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2741 }
2742
2743 static void init_once(void * foo, struct kmem_cache * cachep,
2744 unsigned long flags)
2745 {
2746 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2747
2748 if ((flags & (SLAB_CTOR_CONSTRUCTOR)) ==
2749 SLAB_CTOR_CONSTRUCTOR) {
2750 inode_init_once(&ei->vfs_inode);
2751 }
2752 }
2753
2754 static int init_inodecache(void)
2755 {
2756 btrfs_inode_cachep = kmem_cache_create("btrfs_inode_cache",
2757 sizeof(struct btrfs_inode),
2758 0, (SLAB_RECLAIM_ACCOUNT|
2759 SLAB_MEM_SPREAD),
2760 init_once, NULL);
2761 btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle_cache",
2762 sizeof(struct btrfs_trans_handle),
2763 0, (SLAB_RECLAIM_ACCOUNT|
2764 SLAB_MEM_SPREAD),
2765 NULL, NULL);
2766 btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction_cache",
2767 sizeof(struct btrfs_transaction),
2768 0, (SLAB_RECLAIM_ACCOUNT|
2769 SLAB_MEM_SPREAD),
2770 NULL, NULL);
2771 btrfs_path_cachep = kmem_cache_create("btrfs_path_cache",
2772 sizeof(struct btrfs_transaction),
2773 0, (SLAB_RECLAIM_ACCOUNT|
2774 SLAB_MEM_SPREAD),
2775 NULL, NULL);
2776 btrfs_bit_radix_cachep = kmem_cache_create("btrfs_radix",
2777 256,
2778 0, (SLAB_RECLAIM_ACCOUNT|
2779 SLAB_MEM_SPREAD |
2780 SLAB_DESTROY_BY_RCU),
2781 NULL, NULL);
2782 if (btrfs_inode_cachep == NULL || btrfs_trans_handle_cachep == NULL ||
2783 btrfs_transaction_cachep == NULL || btrfs_bit_radix_cachep == NULL)
2784 return -ENOMEM;
2785 return 0;
2786 }
2787
2788 static void destroy_inodecache(void)
2789 {
2790 kmem_cache_destroy(btrfs_inode_cachep);
2791 kmem_cache_destroy(btrfs_trans_handle_cachep);
2792 kmem_cache_destroy(btrfs_transaction_cachep);
2793 kmem_cache_destroy(btrfs_bit_radix_cachep);
2794 kmem_cache_destroy(btrfs_path_cachep);
2795 }
2796
2797 static int btrfs_get_sb(struct file_system_type *fs_type,
2798 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2799 {
2800 return get_sb_bdev(fs_type, flags, dev_name, data,
2801 btrfs_fill_super, mnt);
2802 }
2803
2804 static int btrfs_getattr(struct vfsmount *mnt,
2805 struct dentry *dentry, struct kstat *stat)
2806 {
2807 struct inode *inode = dentry->d_inode;
2808 generic_fillattr(inode, stat);
2809 stat->blksize = 256 * 1024;
2810 return 0;
2811 }
2812
2813 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2814 {
2815 struct btrfs_root *root = btrfs_sb(dentry->d_sb);
2816 struct btrfs_super_block *disk_super = root->fs_info->disk_super;
2817
2818 buf->f_namelen = BTRFS_NAME_LEN;
2819 buf->f_blocks = btrfs_super_total_blocks(disk_super);
2820 buf->f_bfree = buf->f_blocks - btrfs_super_blocks_used(disk_super);
2821 buf->f_bavail = buf->f_bfree;
2822 buf->f_bsize = dentry->d_sb->s_blocksize;
2823 buf->f_type = BTRFS_SUPER_MAGIC;
2824 return 0;
2825 }
2826
2827 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2828 struct inode * new_dir,struct dentry *new_dentry)
2829 {
2830 struct btrfs_trans_handle *trans;
2831 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2832 struct inode *new_inode = new_dentry->d_inode;
2833 struct inode *old_inode = old_dentry->d_inode;
2834 struct timespec ctime = CURRENT_TIME;
2835 struct btrfs_path *path;
2836 struct btrfs_dir_item *di;
2837 int ret;
2838
2839 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2840 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2841 return -ENOTEMPTY;
2842 }
2843 mutex_lock(&root->fs_info->fs_mutex);
2844 trans = btrfs_start_transaction(root, 1);
2845 btrfs_set_trans_block_group(trans, new_dir);
2846 path = btrfs_alloc_path();
2847 if (!path) {
2848 ret = -ENOMEM;
2849 goto out_fail;
2850 }
2851
2852 old_dentry->d_inode->i_nlink++;
2853 old_dir->i_ctime = old_dir->i_mtime = ctime;
2854 new_dir->i_ctime = new_dir->i_mtime = ctime;
2855 old_inode->i_ctime = ctime;
2856 if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) {
2857 struct btrfs_key *location = &BTRFS_I(new_dir)->location;
2858 u64 old_parent_oid;
2859 di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino,
2860 "..", 2, -1);
2861 if (IS_ERR(di)) {
2862 ret = PTR_ERR(di);
2863 goto out_fail;
2864 }
2865 if (!di) {
2866 ret = -ENOENT;
2867 goto out_fail;
2868 }
2869 old_parent_oid = btrfs_disk_key_objectid(&di->location);
2870 ret = btrfs_del_item(trans, root, path);
2871 if (ret) {
2872 ret = -EIO;
2873 goto out_fail;
2874 }
2875 btrfs_release_path(root, path);
2876
2877 di = btrfs_lookup_dir_index_item(trans, root, path,
2878 old_inode->i_ino,
2879 old_parent_oid,
2880 "..", 2, -1);
2881 if (IS_ERR(di)) {
2882 ret = PTR_ERR(di);
2883 goto out_fail;
2884 }
2885 if (!di) {
2886 ret = -ENOENT;
2887 goto out_fail;
2888 }
2889 ret = btrfs_del_item(trans, root, path);
2890 if (ret) {
2891 ret = -EIO;
2892 goto out_fail;
2893 }
2894 btrfs_release_path(root, path);
2895
2896 ret = btrfs_insert_dir_item(trans, root, "..", 2,
2897 old_inode->i_ino, location,
2898 BTRFS_FT_DIR);
2899 if (ret)
2900 goto out_fail;
2901 }
2902
2903
2904 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2905 if (ret)
2906 goto out_fail;
2907
2908 if (new_inode) {
2909 new_inode->i_ctime = CURRENT_TIME;
2910 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2911 if (ret)
2912 goto out_fail;
2913 if (S_ISDIR(new_inode->i_mode))
2914 clear_nlink(new_inode);
2915 else
2916 drop_nlink(new_inode);
2917 btrfs_update_inode(trans, root, new_inode);
2918 }
2919 ret = btrfs_add_link(trans, new_dentry, old_inode);
2920 if (ret)
2921 goto out_fail;
2922
2923 out_fail:
2924 btrfs_free_path(path);
2925 btrfs_end_transaction(trans, root);
2926 mutex_unlock(&root->fs_info->fs_mutex);
2927 return ret;
2928 }
2929
2930 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2931 const char *symname)
2932 {
2933 struct btrfs_trans_handle *trans;
2934 struct btrfs_root *root = BTRFS_I(dir)->root;
2935 struct btrfs_path *path;
2936 struct btrfs_key key;
2937 struct inode *inode;
2938 int err;
2939 int drop_inode = 0;
2940 u64 objectid;
2941 int name_len;
2942 int datasize;
2943 char *ptr;
2944 struct btrfs_file_extent_item *ei;
2945
2946 name_len = strlen(symname) + 1;
2947 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2948 return -ENAMETOOLONG;
2949 mutex_lock(&root->fs_info->fs_mutex);
2950 trans = btrfs_start_transaction(root, 1);
2951 btrfs_set_trans_block_group(trans, dir);
2952
2953 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2954 if (err) {
2955 err = -ENOSPC;
2956 goto out_unlock;
2957 }
2958
2959 inode = btrfs_new_inode(trans, root, objectid,
2960 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2961 err = PTR_ERR(inode);
2962 if (IS_ERR(inode))
2963 goto out_unlock;
2964
2965 btrfs_set_trans_block_group(trans, inode);
2966 err = btrfs_add_nondir(trans, dentry, inode);
2967 if (err)
2968 drop_inode = 1;
2969 else {
2970 inode->i_mapping->a_ops = &btrfs_aops;
2971 inode->i_fop = &btrfs_file_operations;
2972 inode->i_op = &btrfs_file_inode_operations;
2973 }
2974 dir->i_sb->s_dirt = 1;
2975 btrfs_update_inode_block_group(trans, inode);
2976 btrfs_update_inode_block_group(trans, dir);
2977 if (drop_inode)
2978 goto out_unlock;
2979
2980 path = btrfs_alloc_path();
2981 BUG_ON(!path);
2982 key.objectid = inode->i_ino;
2983 key.offset = 0;
2984 key.flags = 0;
2985 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2986 datasize = btrfs_file_extent_calc_inline_size(name_len);
2987 err = btrfs_insert_empty_item(trans, root, path, &key,
2988 datasize);
2989 BUG_ON(err);
2990 ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
2991 path->slots[0], struct btrfs_file_extent_item);
2992 btrfs_set_file_extent_generation(ei, trans->transid);
2993 btrfs_set_file_extent_type(ei,
2994 BTRFS_FILE_EXTENT_INLINE);
2995 ptr = btrfs_file_extent_inline_start(ei);
2996 btrfs_memcpy(root, path->nodes[0]->b_data,
2997 ptr, symname, name_len);
2998 mark_buffer_dirty(path->nodes[0]);
2999 btrfs_free_path(path);
3000 inode->i_op = &btrfs_symlink_inode_operations;
3001 inode->i_mapping->a_ops = &btrfs_symlink_aops;
3002 inode->i_size = name_len - 1;
3003 btrfs_update_inode(trans, root, inode);
3004 err = 0;
3005
3006 out_unlock:
3007 btrfs_end_transaction(trans, root);
3008 mutex_unlock(&root->fs_info->fs_mutex);
3009
3010 if (drop_inode) {
3011 inode_dec_link_count(inode);
3012 iput(inode);
3013 }
3014 btrfs_btree_balance_dirty(root);
3015 return err;
3016 }
3017
3018 static struct file_system_type btrfs_fs_type = {
3019 .owner = THIS_MODULE,
3020 .name = "btrfs",
3021 .get_sb = btrfs_get_sb,
3022 .kill_sb = kill_block_super,
3023 .fs_flags = FS_REQUIRES_DEV,
3024 };
3025
3026 static struct super_operations btrfs_super_ops = {
3027 .delete_inode = btrfs_delete_inode,
3028 .put_super = btrfs_put_super,
3029 .read_inode = btrfs_read_locked_inode,
3030 .write_super = btrfs_write_super,
3031 .sync_fs = btrfs_sync_fs,
3032 .write_inode = btrfs_write_inode,
3033 .dirty_inode = btrfs_dirty_inode,
3034 .alloc_inode = btrfs_alloc_inode,
3035 .destroy_inode = btrfs_destroy_inode,
3036 .statfs = btrfs_statfs,
3037 };
3038
3039 static struct inode_operations btrfs_dir_inode_operations = {
3040 .lookup = btrfs_lookup,
3041 .create = btrfs_create,
3042 .unlink = btrfs_unlink,
3043 .link = btrfs_link,
3044 .mkdir = btrfs_mkdir,
3045 .rmdir = btrfs_rmdir,
3046 .rename = btrfs_rename,
3047 .symlink = btrfs_symlink,
3048 .setattr = btrfs_setattr,
3049 };
3050
3051 static struct inode_operations btrfs_dir_ro_inode_operations = {
3052 .lookup = btrfs_lookup,
3053 };
3054
3055 static struct file_operations btrfs_dir_file_operations = {
3056 .llseek = generic_file_llseek,
3057 .read = generic_read_dir,
3058 .readdir = btrfs_readdir,
3059 .ioctl = btrfs_ioctl,
3060 #ifdef CONFIG_COMPAT
3061 .compat_ioctl = btrfs_compat_ioctl,
3062 #endif
3063 };
3064
3065 static struct address_space_operations btrfs_aops = {
3066 .readpage = btrfs_readpage,
3067 .writepage = btrfs_writepage,
3068 .sync_page = block_sync_page,
3069 .prepare_write = btrfs_prepare_write,
3070 .commit_write = btrfs_commit_write,
3071 .bmap = btrfs_bmap,
3072 };
3073
3074 static struct address_space_operations btrfs_symlink_aops = {
3075 .readpage = btrfs_readpage,
3076 .writepage = btrfs_writepage,
3077 };
3078
3079 static struct inode_operations btrfs_file_inode_operations = {
3080 .truncate = btrfs_truncate,
3081 .getattr = btrfs_getattr,
3082 .setattr = btrfs_setattr,
3083 };
3084
3085 static struct file_operations btrfs_file_operations = {
3086 .llseek = generic_file_llseek,
3087 .read = do_sync_read,
3088 .aio_read = btrfs_file_aio_read,
3089 .write = btrfs_file_write,
3090 .mmap = generic_file_mmap,
3091 .open = generic_file_open,
3092 .ioctl = btrfs_ioctl,
3093 .fsync = btrfs_sync_file,
3094 #ifdef CONFIG_COMPAT
3095 .compat_ioctl = btrfs_compat_ioctl,
3096 #endif
3097 };
3098
3099 static struct inode_operations btrfs_symlink_inode_operations = {
3100 .readlink = generic_readlink,
3101 .follow_link = page_follow_link_light,
3102 .put_link = page_put_link,
3103 };
3104
3105 static int __init init_btrfs_fs(void)
3106 {
3107 int err;
3108 printk("btrfs loaded!\n");
3109 btrfs_init_transaction_sys();
3110 err = init_inodecache();
3111 if (err)
3112 return err;
3113 return register_filesystem(&btrfs_fs_type);
3114 destroy_inodecache();
3115 return err;
3116 }
3117
3118 static void __exit exit_btrfs_fs(void)
3119 {
3120 btrfs_exit_transaction_sys();
3121 destroy_inodecache();
3122 unregister_filesystem(&btrfs_fs_type);
3123 printk("btrfs unloaded\n");
3124 }
3125
3126 module_init(init_btrfs_fs)
3127 module_exit(exit_btrfs_fs)
3128
3129 MODULE_LICENSE("GPL");
Linux kernel - Deliverables
This page took 0.140518 seconds and 6 git commands to generate.