Btrfs: Dynamic chunk and block group allocation
[deliverable/linux.git] / fs / btrfs / disk-io.c
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
19 #include <linux/fs.h>
20 #include <linux/blkdev.h>
21 #include <linux/crc32c.h>
22 #include <linux/scatterlist.h>
23 #include <linux/swap.h>
24 #include <linux/radix-tree.h>
25 #include <linux/writeback.h>
26 #include <linux/buffer_head.h> // for block_sync_page
27 #include "ctree.h"
28 #include "disk-io.h"
29 #include "transaction.h"
30 #include "btrfs_inode.h"
31 #include "volumes.h"
32 #include "print-tree.h"
33
34 #if 0
35 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
36 {
37 if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) {
38 printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n",
39 (unsigned long long)extent_buffer_blocknr(buf),
40 (unsigned long long)btrfs_header_blocknr(buf));
41 return 1;
42 }
43 return 0;
44 }
45 #endif
46
47 static struct extent_io_ops btree_extent_io_ops;
48
49 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
50 u64 bytenr, u32 blocksize)
51 {
52 struct inode *btree_inode = root->fs_info->btree_inode;
53 struct extent_buffer *eb;
54 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
55 bytenr, blocksize, GFP_NOFS);
56 return eb;
57 }
58
59 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
60 u64 bytenr, u32 blocksize)
61 {
62 struct inode *btree_inode = root->fs_info->btree_inode;
63 struct extent_buffer *eb;
64
65 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
66 bytenr, blocksize, NULL, GFP_NOFS);
67 return eb;
68 }
69
70 struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
71 size_t page_offset, u64 start, u64 len,
72 int create)
73 {
74 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
75 struct extent_map *em;
76 int ret;
77
78 again:
79 spin_lock(&em_tree->lock);
80 em = lookup_extent_mapping(em_tree, start, len);
81 spin_unlock(&em_tree->lock);
82 if (em) {
83 goto out;
84 }
85 em = alloc_extent_map(GFP_NOFS);
86 if (!em) {
87 em = ERR_PTR(-ENOMEM);
88 goto out;
89 }
90 em->start = 0;
91 em->len = i_size_read(inode);
92 em->block_start = 0;
93 em->bdev = inode->i_sb->s_bdev;
94
95 spin_lock(&em_tree->lock);
96 ret = add_extent_mapping(em_tree, em);
97 spin_unlock(&em_tree->lock);
98
99 if (ret == -EEXIST) {
100 free_extent_map(em);
101 em = NULL;
102 goto again;
103 } else if (ret) {
104 em = ERR_PTR(ret);
105 }
106 out:
107 return em;
108 }
109
110 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
111 {
112 return crc32c(seed, data, len);
113 }
114
115 void btrfs_csum_final(u32 crc, char *result)
116 {
117 *(__le32 *)result = ~cpu_to_le32(crc);
118 }
119
120 static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
121 int verify)
122 {
123 char result[BTRFS_CRC32_SIZE];
124 unsigned long len;
125 unsigned long cur_len;
126 unsigned long offset = BTRFS_CSUM_SIZE;
127 char *map_token = NULL;
128 char *kaddr;
129 unsigned long map_start;
130 unsigned long map_len;
131 int err;
132 u32 crc = ~(u32)0;
133
134 len = buf->len - offset;
135 while(len > 0) {
136 err = map_private_extent_buffer(buf, offset, 32,
137 &map_token, &kaddr,
138 &map_start, &map_len, KM_USER0);
139 if (err) {
140 printk("failed to map extent buffer! %lu\n",
141 offset);
142 return 1;
143 }
144 cur_len = min(len, map_len - (offset - map_start));
145 crc = btrfs_csum_data(root, kaddr + offset - map_start,
146 crc, cur_len);
147 len -= cur_len;
148 offset += cur_len;
149 unmap_extent_buffer(buf, map_token, KM_USER0);
150 }
151 btrfs_csum_final(crc, result);
152
153 if (verify) {
154 int from_this_trans = 0;
155
156 if (root->fs_info->running_transaction &&
157 btrfs_header_generation(buf) ==
158 root->fs_info->running_transaction->transid)
159 from_this_trans = 1;
160
161 /* FIXME, this is not good */
162 if (from_this_trans == 0 &&
163 memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) {
164 u32 val;
165 u32 found = 0;
166 memcpy(&found, result, BTRFS_CRC32_SIZE);
167
168 read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE);
169 printk("btrfs: %s checksum verify failed on %llu "
170 "wanted %X found %X from_this_trans %d\n",
171 root->fs_info->sb->s_id,
172 buf->start, val, found, from_this_trans);
173 return 1;
174 }
175 } else {
176 write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE);
177 }
178 return 0;
179 }
180
181
182 int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
183 {
184 struct extent_io_tree *tree;
185 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
186 u64 found_start;
187 int found_level;
188 unsigned long len;
189 struct extent_buffer *eb;
190 tree = &BTRFS_I(page->mapping->host)->io_tree;
191
192 if (page->private == EXTENT_PAGE_PRIVATE)
193 goto out;
194 if (!page->private)
195 goto out;
196 len = page->private >> 2;
197 if (len == 0) {
198 WARN_ON(1);
199 }
200 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
201 read_extent_buffer_pages(tree, eb, start + PAGE_CACHE_SIZE, 1,
202 btree_get_extent);
203 btrfs_clear_buffer_defrag(eb);
204 found_start = btrfs_header_bytenr(eb);
205 if (found_start != start) {
206 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
207 start, found_start, len);
208 WARN_ON(1);
209 goto err;
210 }
211 if (eb->first_page != page) {
212 printk("bad first page %lu %lu\n", eb->first_page->index,
213 page->index);
214 WARN_ON(1);
215 goto err;
216 }
217 if (!PageUptodate(page)) {
218 printk("csum not up to date page %lu\n", page->index);
219 WARN_ON(1);
220 goto err;
221 }
222 found_level = btrfs_header_level(eb);
223 csum_tree_block(root, eb, 0);
224 err:
225 free_extent_buffer(eb);
226 out:
227 return 0;
228 }
229
230 static int btree_writepage_io_hook(struct page *page, u64 start, u64 end)
231 {
232 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
233
234 csum_dirty_buffer(root, page);
235 return 0;
236 }
237
238 static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio)
239 {
240 struct btrfs_root *root = BTRFS_I(inode)->root;
241 u64 offset;
242 offset = bio->bi_sector << 9;
243 if (offset == BTRFS_SUPER_INFO_OFFSET) {
244 bio->bi_bdev = root->fs_info->sb->s_bdev;
245 submit_bio(rw, bio);
246 return 0;
247 }
248 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio);
249 }
250
251 static int btree_writepage(struct page *page, struct writeback_control *wbc)
252 {
253 struct extent_io_tree *tree;
254 tree = &BTRFS_I(page->mapping->host)->io_tree;
255 return extent_write_full_page(tree, page, btree_get_extent, wbc);
256 }
257
258 static int btree_writepages(struct address_space *mapping,
259 struct writeback_control *wbc)
260 {
261 struct extent_io_tree *tree;
262 tree = &BTRFS_I(mapping->host)->io_tree;
263 if (wbc->sync_mode == WB_SYNC_NONE) {
264 u64 num_dirty;
265 u64 start = 0;
266 unsigned long thresh = 96 * 1024 * 1024;
267
268 if (wbc->for_kupdate)
269 return 0;
270
271 if (current_is_pdflush()) {
272 thresh = 96 * 1024 * 1024;
273 } else {
274 thresh = 8 * 1024 * 1024;
275 }
276 num_dirty = count_range_bits(tree, &start, (u64)-1,
277 thresh, EXTENT_DIRTY);
278 if (num_dirty < thresh) {
279 return 0;
280 }
281 }
282 return extent_writepages(tree, mapping, btree_get_extent, wbc);
283 }
284
285 int btree_readpage(struct file *file, struct page *page)
286 {
287 struct extent_io_tree *tree;
288 tree = &BTRFS_I(page->mapping->host)->io_tree;
289 return extent_read_full_page(tree, page, btree_get_extent);
290 }
291
292 static int btree_releasepage(struct page *page, gfp_t gfp_flags)
293 {
294 struct extent_io_tree *tree;
295 struct extent_map_tree *map;
296 int ret;
297
298 tree = &BTRFS_I(page->mapping->host)->io_tree;
299 map = &BTRFS_I(page->mapping->host)->extent_tree;
300 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
301 if (ret == 1) {
302 ClearPagePrivate(page);
303 set_page_private(page, 0);
304 page_cache_release(page);
305 }
306 return ret;
307 }
308
309 static void btree_invalidatepage(struct page *page, unsigned long offset)
310 {
311 struct extent_io_tree *tree;
312 tree = &BTRFS_I(page->mapping->host)->io_tree;
313 extent_invalidatepage(tree, page, offset);
314 btree_releasepage(page, GFP_NOFS);
315 }
316
317 #if 0
318 static int btree_writepage(struct page *page, struct writeback_control *wbc)
319 {
320 struct buffer_head *bh;
321 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
322 struct buffer_head *head;
323 if (!page_has_buffers(page)) {
324 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
325 (1 << BH_Dirty)|(1 << BH_Uptodate));
326 }
327 head = page_buffers(page);
328 bh = head;
329 do {
330 if (buffer_dirty(bh))
331 csum_tree_block(root, bh, 0);
332 bh = bh->b_this_page;
333 } while (bh != head);
334 return block_write_full_page(page, btree_get_block, wbc);
335 }
336 #endif
337
338 static struct address_space_operations btree_aops = {
339 .readpage = btree_readpage,
340 .writepage = btree_writepage,
341 .writepages = btree_writepages,
342 .releasepage = btree_releasepage,
343 .invalidatepage = btree_invalidatepage,
344 .sync_page = block_sync_page,
345 };
346
347 int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
348 {
349 struct extent_buffer *buf = NULL;
350 struct inode *btree_inode = root->fs_info->btree_inode;
351 int ret = 0;
352
353 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
354 if (!buf)
355 return 0;
356 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
357 buf, 0, 0, btree_get_extent);
358 free_extent_buffer(buf);
359 return ret;
360 }
361
362 static int close_all_devices(struct btrfs_fs_info *fs_info)
363 {
364 struct list_head *list;
365 struct list_head *next;
366 struct btrfs_device *device;
367
368 list = &fs_info->devices;
369 while(!list_empty(list)) {
370 next = list->next;
371 list_del(next);
372 device = list_entry(next, struct btrfs_device, dev_list);
373 kfree(device->name);
374 kfree(device);
375 }
376 return 0;
377 }
378
379 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
380 u32 blocksize)
381 {
382 struct extent_buffer *buf = NULL;
383 struct inode *btree_inode = root->fs_info->btree_inode;
384 struct extent_io_tree *io_tree;
385 u64 end;
386 int ret;
387
388 io_tree = &BTRFS_I(btree_inode)->io_tree;
389
390 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
391 if (!buf)
392 return NULL;
393 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, buf, 0, 1,
394 btree_get_extent);
395
396 if (buf->flags & EXTENT_CSUM)
397 return buf;
398
399 end = buf->start + PAGE_CACHE_SIZE - 1;
400 if (test_range_bit(io_tree, buf->start, end, EXTENT_CSUM, 1)) {
401 buf->flags |= EXTENT_CSUM;
402 return buf;
403 }
404
405 lock_extent(io_tree, buf->start, end, GFP_NOFS);
406
407 if (test_range_bit(io_tree, buf->start, end, EXTENT_CSUM, 1)) {
408 buf->flags |= EXTENT_CSUM;
409 goto out_unlock;
410 }
411
412 ret = csum_tree_block(root, buf, 1);
413 set_extent_bits(io_tree, buf->start, end, EXTENT_CSUM, GFP_NOFS);
414 buf->flags |= EXTENT_CSUM;
415
416 out_unlock:
417 unlock_extent(io_tree, buf->start, end, GFP_NOFS);
418 return buf;
419 }
420
421 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
422 struct extent_buffer *buf)
423 {
424 struct inode *btree_inode = root->fs_info->btree_inode;
425 if (btrfs_header_generation(buf) ==
426 root->fs_info->running_transaction->transid)
427 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
428 buf);
429 return 0;
430 }
431
432 int wait_on_tree_block_writeback(struct btrfs_root *root,
433 struct extent_buffer *buf)
434 {
435 struct inode *btree_inode = root->fs_info->btree_inode;
436 wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree,
437 buf);
438 return 0;
439 }
440
441 static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
442 u32 stripesize, struct btrfs_root *root,
443 struct btrfs_fs_info *fs_info,
444 u64 objectid)
445 {
446 root->node = NULL;
447 root->inode = NULL;
448 root->commit_root = NULL;
449 root->sectorsize = sectorsize;
450 root->nodesize = nodesize;
451 root->leafsize = leafsize;
452 root->stripesize = stripesize;
453 root->ref_cows = 0;
454 root->track_dirty = 0;
455
456 root->fs_info = fs_info;
457 root->objectid = objectid;
458 root->last_trans = 0;
459 root->highest_inode = 0;
460 root->last_inode_alloc = 0;
461 root->name = NULL;
462 root->in_sysfs = 0;
463
464 INIT_LIST_HEAD(&root->dirty_list);
465 memset(&root->root_key, 0, sizeof(root->root_key));
466 memset(&root->root_item, 0, sizeof(root->root_item));
467 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
468 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
469 init_completion(&root->kobj_unregister);
470 root->defrag_running = 0;
471 root->defrag_level = 0;
472 root->root_key.objectid = objectid;
473 return 0;
474 }
475
476 static int find_and_setup_root(struct btrfs_root *tree_root,
477 struct btrfs_fs_info *fs_info,
478 u64 objectid,
479 struct btrfs_root *root)
480 {
481 int ret;
482 u32 blocksize;
483
484 __setup_root(tree_root->nodesize, tree_root->leafsize,
485 tree_root->sectorsize, tree_root->stripesize,
486 root, fs_info, objectid);
487 ret = btrfs_find_last_root(tree_root, objectid,
488 &root->root_item, &root->root_key);
489 BUG_ON(ret);
490
491 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
492 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
493 blocksize);
494 BUG_ON(!root->node);
495 return 0;
496 }
497
498 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
499 struct btrfs_key *location)
500 {
501 struct btrfs_root *root;
502 struct btrfs_root *tree_root = fs_info->tree_root;
503 struct btrfs_path *path;
504 struct extent_buffer *l;
505 u64 highest_inode;
506 u32 blocksize;
507 int ret = 0;
508
509 root = kzalloc(sizeof(*root), GFP_NOFS);
510 if (!root)
511 return ERR_PTR(-ENOMEM);
512 if (location->offset == (u64)-1) {
513 ret = find_and_setup_root(tree_root, fs_info,
514 location->objectid, root);
515 if (ret) {
516 kfree(root);
517 return ERR_PTR(ret);
518 }
519 goto insert;
520 }
521
522 __setup_root(tree_root->nodesize, tree_root->leafsize,
523 tree_root->sectorsize, tree_root->stripesize,
524 root, fs_info, location->objectid);
525
526 path = btrfs_alloc_path();
527 BUG_ON(!path);
528 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
529 if (ret != 0) {
530 if (ret > 0)
531 ret = -ENOENT;
532 goto out;
533 }
534 l = path->nodes[0];
535 read_extent_buffer(l, &root->root_item,
536 btrfs_item_ptr_offset(l, path->slots[0]),
537 sizeof(root->root_item));
538 memcpy(&root->root_key, location, sizeof(*location));
539 ret = 0;
540 out:
541 btrfs_release_path(root, path);
542 btrfs_free_path(path);
543 if (ret) {
544 kfree(root);
545 return ERR_PTR(ret);
546 }
547 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
548 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
549 blocksize);
550 BUG_ON(!root->node);
551 insert:
552 root->ref_cows = 1;
553 ret = btrfs_find_highest_inode(root, &highest_inode);
554 if (ret == 0) {
555 root->highest_inode = highest_inode;
556 root->last_inode_alloc = highest_inode;
557 }
558 return root;
559 }
560
561 struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
562 u64 root_objectid)
563 {
564 struct btrfs_root *root;
565
566 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
567 return fs_info->tree_root;
568 if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
569 return fs_info->extent_root;
570
571 root = radix_tree_lookup(&fs_info->fs_roots_radix,
572 (unsigned long)root_objectid);
573 return root;
574 }
575
576 struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
577 struct btrfs_key *location)
578 {
579 struct btrfs_root *root;
580 int ret;
581
582 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
583 return fs_info->tree_root;
584 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
585 return fs_info->extent_root;
586
587 root = radix_tree_lookup(&fs_info->fs_roots_radix,
588 (unsigned long)location->objectid);
589 if (root)
590 return root;
591
592 root = btrfs_read_fs_root_no_radix(fs_info, location);
593 if (IS_ERR(root))
594 return root;
595 ret = radix_tree_insert(&fs_info->fs_roots_radix,
596 (unsigned long)root->root_key.objectid,
597 root);
598 if (ret) {
599 free_extent_buffer(root->node);
600 kfree(root);
601 return ERR_PTR(ret);
602 }
603 ret = btrfs_find_dead_roots(fs_info->tree_root,
604 root->root_key.objectid, root);
605 BUG_ON(ret);
606
607 return root;
608 }
609
610 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
611 struct btrfs_key *location,
612 const char *name, int namelen)
613 {
614 struct btrfs_root *root;
615 int ret;
616
617 root = btrfs_read_fs_root_no_name(fs_info, location);
618 if (!root)
619 return NULL;
620
621 if (root->in_sysfs)
622 return root;
623
624 ret = btrfs_set_root_name(root, name, namelen);
625 if (ret) {
626 free_extent_buffer(root->node);
627 kfree(root);
628 return ERR_PTR(ret);
629 }
630
631 ret = btrfs_sysfs_add_root(root);
632 if (ret) {
633 free_extent_buffer(root->node);
634 kfree(root->name);
635 kfree(root);
636 return ERR_PTR(ret);
637 }
638 root->in_sysfs = 1;
639 return root;
640 }
641 #if 0
642 static int add_hasher(struct btrfs_fs_info *info, char *type) {
643 struct btrfs_hasher *hasher;
644
645 hasher = kmalloc(sizeof(*hasher), GFP_NOFS);
646 if (!hasher)
647 return -ENOMEM;
648 hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC);
649 if (!hasher->hash_tfm) {
650 kfree(hasher);
651 return -EINVAL;
652 }
653 spin_lock(&info->hash_lock);
654 list_add(&hasher->list, &info->hashers);
655 spin_unlock(&info->hash_lock);
656 return 0;
657 }
658 #endif
659 struct btrfs_root *open_ctree(struct super_block *sb)
660 {
661 u32 sectorsize;
662 u32 nodesize;
663 u32 leafsize;
664 u32 blocksize;
665 u32 stripesize;
666 struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
667 GFP_NOFS);
668 struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
669 GFP_NOFS);
670 struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
671 GFP_NOFS);
672 struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root),
673 GFP_NOFS);
674 struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root),
675 GFP_NOFS);
676 int ret;
677 int err = -EIO;
678 struct btrfs_super_block *disk_super;
679
680 if (!extent_root || !tree_root || !fs_info) {
681 err = -ENOMEM;
682 goto fail;
683 }
684 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
685 INIT_LIST_HEAD(&fs_info->trans_list);
686 INIT_LIST_HEAD(&fs_info->dead_roots);
687 INIT_LIST_HEAD(&fs_info->hashers);
688 spin_lock_init(&fs_info->hash_lock);
689 spin_lock_init(&fs_info->delalloc_lock);
690 spin_lock_init(&fs_info->new_trans_lock);
691
692 memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj));
693 init_completion(&fs_info->kobj_unregister);
694 sb_set_blocksize(sb, 4096);
695 fs_info->running_transaction = NULL;
696 fs_info->last_trans_committed = 0;
697 fs_info->tree_root = tree_root;
698 fs_info->extent_root = extent_root;
699 fs_info->chunk_root = chunk_root;
700 fs_info->dev_root = dev_root;
701 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
702 INIT_LIST_HEAD(&fs_info->devices);
703 INIT_LIST_HEAD(&fs_info->space_info);
704 btrfs_mapping_init(&fs_info->mapping_tree);
705 fs_info->sb = sb;
706 fs_info->throttles = 0;
707 fs_info->mount_opt = 0;
708 fs_info->max_extent = (u64)-1;
709 fs_info->max_inline = 8192 * 1024;
710 fs_info->delalloc_bytes = 0;
711 fs_info->btree_inode = new_inode(sb);
712 fs_info->btree_inode->i_ino = 1;
713 fs_info->btree_inode->i_nlink = 1;
714 fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
715 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
716 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
717 fs_info->btree_inode->i_mapping,
718 GFP_NOFS);
719 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
720 GFP_NOFS);
721
722 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
723
724 extent_io_tree_init(&fs_info->free_space_cache,
725 fs_info->btree_inode->i_mapping, GFP_NOFS);
726 extent_io_tree_init(&fs_info->block_group_cache,
727 fs_info->btree_inode->i_mapping, GFP_NOFS);
728 extent_io_tree_init(&fs_info->pinned_extents,
729 fs_info->btree_inode->i_mapping, GFP_NOFS);
730 extent_io_tree_init(&fs_info->pending_del,
731 fs_info->btree_inode->i_mapping, GFP_NOFS);
732 extent_io_tree_init(&fs_info->extent_ins,
733 fs_info->btree_inode->i_mapping, GFP_NOFS);
734 fs_info->do_barriers = 1;
735 fs_info->closing = 0;
736 fs_info->total_pinned = 0;
737 fs_info->last_alloc = 0;
738 fs_info->last_data_alloc = 0;
739
740 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
741 INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info);
742 #else
743 INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
744 #endif
745 BTRFS_I(fs_info->btree_inode)->root = tree_root;
746 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
747 sizeof(struct btrfs_key));
748 insert_inode_hash(fs_info->btree_inode);
749 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
750
751 mutex_init(&fs_info->trans_mutex);
752 mutex_init(&fs_info->fs_mutex);
753
754 #if 0
755 ret = add_hasher(fs_info, "crc32c");
756 if (ret) {
757 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
758 err = -ENOMEM;
759 goto fail_iput;
760 }
761 #endif
762 __setup_root(4096, 4096, 4096, 4096, tree_root,
763 fs_info, BTRFS_ROOT_TREE_OBJECTID);
764
765 fs_info->sb_buffer = read_tree_block(tree_root,
766 BTRFS_SUPER_INFO_OFFSET,
767 4096);
768
769 if (!fs_info->sb_buffer)
770 goto fail_iput;
771
772 read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0,
773 sizeof(fs_info->super_copy));
774
775 read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
776 (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
777 BTRFS_FSID_SIZE);
778
779 disk_super = &fs_info->super_copy;
780 if (!btrfs_super_root(disk_super))
781 goto fail_sb_buffer;
782
783 nodesize = btrfs_super_nodesize(disk_super);
784 leafsize = btrfs_super_leafsize(disk_super);
785 sectorsize = btrfs_super_sectorsize(disk_super);
786 stripesize = btrfs_super_stripesize(disk_super);
787 tree_root->nodesize = nodesize;
788 tree_root->leafsize = leafsize;
789 tree_root->sectorsize = sectorsize;
790 tree_root->stripesize = stripesize;
791 sb_set_blocksize(sb, sectorsize);
792
793 i_size_write(fs_info->btree_inode,
794 btrfs_super_total_bytes(disk_super));
795
796 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
797 sizeof(disk_super->magic))) {
798 printk("btrfs: valid FS not found on %s\n", sb->s_id);
799 goto fail_sb_buffer;
800 }
801
802 mutex_lock(&fs_info->fs_mutex);
803 ret = btrfs_read_sys_array(tree_root);
804 BUG_ON(ret);
805
806 blocksize = btrfs_level_size(tree_root,
807 btrfs_super_chunk_root_level(disk_super));
808
809 __setup_root(nodesize, leafsize, sectorsize, stripesize,
810 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
811
812 chunk_root->node = read_tree_block(chunk_root,
813 btrfs_super_chunk_root(disk_super),
814 blocksize);
815 BUG_ON(!chunk_root->node);
816
817 ret = btrfs_read_chunk_tree(chunk_root);
818 BUG_ON(ret);
819
820 blocksize = btrfs_level_size(tree_root,
821 btrfs_super_root_level(disk_super));
822
823
824 tree_root->node = read_tree_block(tree_root,
825 btrfs_super_root(disk_super),
826 blocksize);
827 if (!tree_root->node)
828 goto fail_sb_buffer;
829
830
831 ret = find_and_setup_root(tree_root, fs_info,
832 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
833 if (ret)
834 goto fail_tree_root;
835 extent_root->track_dirty = 1;
836
837 ret = find_and_setup_root(tree_root, fs_info,
838 BTRFS_DEV_TREE_OBJECTID, dev_root);
839 dev_root->track_dirty = 1;
840
841 if (ret)
842 goto fail_extent_root;
843
844 btrfs_read_block_groups(extent_root);
845
846 fs_info->generation = btrfs_super_generation(disk_super) + 1;
847 mutex_unlock(&fs_info->fs_mutex);
848 return tree_root;
849
850 fail_extent_root:
851 free_extent_buffer(extent_root->node);
852 fail_tree_root:
853 mutex_unlock(&fs_info->fs_mutex);
854 free_extent_buffer(tree_root->node);
855 fail_sb_buffer:
856 free_extent_buffer(fs_info->sb_buffer);
857 fail_iput:
858 iput(fs_info->btree_inode);
859 fail:
860 kfree(extent_root);
861 kfree(tree_root);
862 kfree(fs_info);
863 return ERR_PTR(err);
864 }
865
866 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
867 *root)
868 {
869 int ret;
870 struct extent_buffer *super = root->fs_info->sb_buffer;
871 struct inode *btree_inode = root->fs_info->btree_inode;
872 struct super_block *sb = root->fs_info->sb;
873
874 if (!btrfs_test_opt(root, NOBARRIER))
875 blkdev_issue_flush(sb->s_bdev, NULL);
876 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, super);
877 ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping,
878 super->start, super->len);
879 if (!btrfs_test_opt(root, NOBARRIER))
880 blkdev_issue_flush(sb->s_bdev, NULL);
881 return ret;
882 }
883
884 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
885 {
886 radix_tree_delete(&fs_info->fs_roots_radix,
887 (unsigned long)root->root_key.objectid);
888 if (root->in_sysfs)
889 btrfs_sysfs_del_root(root);
890 if (root->inode)
891 iput(root->inode);
892 if (root->node)
893 free_extent_buffer(root->node);
894 if (root->commit_root)
895 free_extent_buffer(root->commit_root);
896 if (root->name)
897 kfree(root->name);
898 kfree(root);
899 return 0;
900 }
901
902 static int del_fs_roots(struct btrfs_fs_info *fs_info)
903 {
904 int ret;
905 struct btrfs_root *gang[8];
906 int i;
907
908 while(1) {
909 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
910 (void **)gang, 0,
911 ARRAY_SIZE(gang));
912 if (!ret)
913 break;
914 for (i = 0; i < ret; i++)
915 btrfs_free_fs_root(fs_info, gang[i]);
916 }
917 return 0;
918 }
919
920 int close_ctree(struct btrfs_root *root)
921 {
922 int ret;
923 struct btrfs_trans_handle *trans;
924 struct btrfs_fs_info *fs_info = root->fs_info;
925
926 fs_info->closing = 1;
927 btrfs_transaction_flush_work(root);
928 mutex_lock(&fs_info->fs_mutex);
929 btrfs_defrag_dirty_roots(root->fs_info);
930 trans = btrfs_start_transaction(root, 1);
931 ret = btrfs_commit_transaction(trans, root);
932 /* run commit again to drop the original snapshot */
933 trans = btrfs_start_transaction(root, 1);
934 btrfs_commit_transaction(trans, root);
935 ret = btrfs_write_and_wait_transaction(NULL, root);
936 BUG_ON(ret);
937 write_ctree_super(NULL, root);
938 mutex_unlock(&fs_info->fs_mutex);
939
940 if (fs_info->delalloc_bytes) {
941 printk("btrfs: at unmount delalloc count %Lu\n",
942 fs_info->delalloc_bytes);
943 }
944 if (fs_info->extent_root->node)
945 free_extent_buffer(fs_info->extent_root->node);
946
947 if (fs_info->tree_root->node)
948 free_extent_buffer(fs_info->tree_root->node);
949
950 if (root->fs_info->chunk_root->node);
951 free_extent_buffer(root->fs_info->chunk_root->node);
952
953 if (root->fs_info->dev_root->node);
954 free_extent_buffer(root->fs_info->dev_root->node);
955
956 free_extent_buffer(fs_info->sb_buffer);
957
958 btrfs_free_block_groups(root->fs_info);
959 del_fs_roots(fs_info);
960
961 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
962
963 extent_io_tree_empty_lru(&fs_info->free_space_cache);
964 extent_io_tree_empty_lru(&fs_info->block_group_cache);
965 extent_io_tree_empty_lru(&fs_info->pinned_extents);
966 extent_io_tree_empty_lru(&fs_info->pending_del);
967 extent_io_tree_empty_lru(&fs_info->extent_ins);
968 extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
969
970 truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
971
972 iput(fs_info->btree_inode);
973 #if 0
974 while(!list_empty(&fs_info->hashers)) {
975 struct btrfs_hasher *hasher;
976 hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher,
977 hashers);
978 list_del(&hasher->hashers);
979 crypto_free_hash(&fs_info->hash_tfm);
980 kfree(hasher);
981 }
982 #endif
983 close_all_devices(fs_info);
984 btrfs_mapping_tree_free(&fs_info->mapping_tree);
985
986 kfree(fs_info->extent_root);
987 kfree(fs_info->tree_root);
988 kfree(fs_info->chunk_root);
989 kfree(fs_info->dev_root);
990 return 0;
991 }
992
993 int btrfs_buffer_uptodate(struct extent_buffer *buf)
994 {
995 struct inode *btree_inode = buf->first_page->mapping->host;
996 return extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf);
997 }
998
999 int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
1000 {
1001 struct inode *btree_inode = buf->first_page->mapping->host;
1002 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
1003 buf);
1004 }
1005
1006 void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
1007 {
1008 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1009 u64 transid = btrfs_header_generation(buf);
1010 struct inode *btree_inode = root->fs_info->btree_inode;
1011
1012 if (transid != root->fs_info->generation) {
1013 printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
1014 (unsigned long long)buf->start,
1015 transid, root->fs_info->generation);
1016 WARN_ON(1);
1017 }
1018 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf);
1019 }
1020
1021 void btrfs_throttle(struct btrfs_root *root)
1022 {
1023 struct backing_dev_info *bdi;
1024
1025 bdi = root->fs_info->sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
1026 if (root->fs_info->throttles && bdi_write_congested(bdi)) {
1027 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18)
1028 congestion_wait(WRITE, HZ/20);
1029 #else
1030 blk_congestion_wait(WRITE, HZ/20);
1031 #endif
1032 }
1033 }
1034
1035 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
1036 {
1037 balance_dirty_pages_ratelimited_nr(
1038 root->fs_info->btree_inode->i_mapping, 1);
1039 }
1040
1041 void btrfs_set_buffer_defrag(struct extent_buffer *buf)
1042 {
1043 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1044 struct inode *btree_inode = root->fs_info->btree_inode;
1045 set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
1046 buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS);
1047 }
1048
1049 void btrfs_set_buffer_defrag_done(struct extent_buffer *buf)
1050 {
1051 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1052 struct inode *btree_inode = root->fs_info->btree_inode;
1053 set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
1054 buf->start + buf->len - 1, EXTENT_DEFRAG_DONE,
1055 GFP_NOFS);
1056 }
1057
1058 int btrfs_buffer_defrag(struct extent_buffer *buf)
1059 {
1060 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1061 struct inode *btree_inode = root->fs_info->btree_inode;
1062 return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
1063 buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0);
1064 }
1065
1066 int btrfs_buffer_defrag_done(struct extent_buffer *buf)
1067 {
1068 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1069 struct inode *btree_inode = root->fs_info->btree_inode;
1070 return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
1071 buf->start, buf->start + buf->len - 1,
1072 EXTENT_DEFRAG_DONE, 0);
1073 }
1074
1075 int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf)
1076 {
1077 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1078 struct inode *btree_inode = root->fs_info->btree_inode;
1079 return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
1080 buf->start, buf->start + buf->len - 1,
1081 EXTENT_DEFRAG_DONE, GFP_NOFS);
1082 }
1083
1084 int btrfs_clear_buffer_defrag(struct extent_buffer *buf)
1085 {
1086 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1087 struct inode *btree_inode = root->fs_info->btree_inode;
1088 return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
1089 buf->start, buf->start + buf->len - 1,
1090 EXTENT_DEFRAG, GFP_NOFS);
1091 }
1092
1093 int btrfs_read_buffer(struct extent_buffer *buf)
1094 {
1095 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1096 struct inode *btree_inode = root->fs_info->btree_inode;
1097 return read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
1098 buf, 0, 1, btree_get_extent);
1099 }
1100
1101 static struct extent_io_ops btree_extent_io_ops = {
1102 .writepage_io_hook = btree_writepage_io_hook,
1103 .submit_bio_hook = btree_submit_bio_hook,
1104 };
This page took 0.072126 seconds and 5 git commands to generate.