xref: /openbmc/linux/fs/btrfs/disk-io.c (revision 6da6abae027e2dbc59bca5f4168b0760f25068c7)
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 "print-tree.h"
32 
33 #if 0
34 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
35 {
36 	if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) {
37 		printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n",
38 		       (unsigned long long)extent_buffer_blocknr(buf),
39 		       (unsigned long long)btrfs_header_blocknr(buf));
40 		return 1;
41 	}
42 	return 0;
43 }
44 #endif
45 
46 static struct extent_map_ops btree_extent_map_ops;
47 
48 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
49 					    u64 bytenr, u32 blocksize)
50 {
51 	struct inode *btree_inode = root->fs_info->btree_inode;
52 	struct extent_buffer *eb;
53 	eb = find_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
54 				bytenr, blocksize, GFP_NOFS);
55 	return eb;
56 }
57 
58 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
59 						 u64 bytenr, u32 blocksize)
60 {
61 	struct inode *btree_inode = root->fs_info->btree_inode;
62 	struct extent_buffer *eb;
63 
64 	eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
65 				 bytenr, blocksize, NULL, GFP_NOFS);
66 	return eb;
67 }
68 
69 struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
70 				    size_t page_offset, u64 start, u64 end,
71 				    int create)
72 {
73 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
74 	struct extent_map *em;
75 	int ret;
76 
77 again:
78 	em = lookup_extent_mapping(em_tree, start, end);
79 	if (em) {
80 		goto out;
81 	}
82 	em = alloc_extent_map(GFP_NOFS);
83 	if (!em) {
84 		em = ERR_PTR(-ENOMEM);
85 		goto out;
86 	}
87 	em->start = 0;
88 	em->end = (i_size_read(inode) & ~((u64)PAGE_CACHE_SIZE -1)) - 1;
89 	em->block_start = 0;
90 	em->block_end = em->end;
91 	em->bdev = inode->i_sb->s_bdev;
92 	ret = add_extent_mapping(em_tree, em);
93 	if (ret == -EEXIST) {
94 		free_extent_map(em);
95 		em = NULL;
96 		goto again;
97 	} else if (ret) {
98 		em = ERR_PTR(ret);
99 	}
100 out:
101 	return em;
102 }
103 
104 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
105 {
106 	return crc32c(seed, data, len);
107 }
108 
109 void btrfs_csum_final(u32 crc, char *result)
110 {
111 	*(__le32 *)result = ~cpu_to_le32(crc);
112 }
113 
114 static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
115 			   int verify)
116 {
117 	char result[BTRFS_CRC32_SIZE];
118 	unsigned long len;
119 	unsigned long cur_len;
120 	unsigned long offset = BTRFS_CSUM_SIZE;
121 	char *map_token = NULL;
122 	char *kaddr;
123 	unsigned long map_start;
124 	unsigned long map_len;
125 	int err;
126 	u32 crc = ~(u32)0;
127 
128 	len = buf->len - offset;
129 	while(len > 0) {
130 		err = map_private_extent_buffer(buf, offset, 32,
131 					&map_token, &kaddr,
132 					&map_start, &map_len, KM_USER0);
133 		if (err) {
134 			printk("failed to map extent buffer! %lu\n",
135 			       offset);
136 			return 1;
137 		}
138 		cur_len = min(len, map_len - (offset - map_start));
139 		crc = btrfs_csum_data(root, kaddr + offset - map_start,
140 				      crc, cur_len);
141 		len -= cur_len;
142 		offset += cur_len;
143 		unmap_extent_buffer(buf, map_token, KM_USER0);
144 	}
145 	btrfs_csum_final(crc, result);
146 
147 	if (verify) {
148 		if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) {
149 			printk("btrfs: %s checksum verify failed on %llu\n",
150 			       root->fs_info->sb->s_id,
151 			       buf->start);
152 			return 1;
153 		}
154 	} else {
155 		write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE);
156 	}
157 	return 0;
158 }
159 
160 
161 int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
162 {
163 	struct extent_map_tree *tree;
164 	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
165 	u64 found_start;
166 	int found_level;
167 	unsigned long len;
168 	struct extent_buffer *eb;
169 	tree = &BTRFS_I(page->mapping->host)->extent_tree;
170 
171 	if (page->private == EXTENT_PAGE_PRIVATE)
172 		goto out;
173 	if (!page->private)
174 		goto out;
175 	len = page->private >> 2;
176 	if (len == 0) {
177 		WARN_ON(1);
178 	}
179 	eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
180 	read_extent_buffer_pages(tree, eb, start + PAGE_CACHE_SIZE, 1);
181 	found_start = btrfs_header_bytenr(eb);
182 	if (found_start != start) {
183 		printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
184 		       start, found_start, len);
185 	}
186 	found_level = btrfs_header_level(eb);
187 	csum_tree_block(root, eb, 0);
188 	free_extent_buffer(eb);
189 out:
190 	return 0;
191 }
192 
193 static int btree_writepage_io_hook(struct page *page, u64 start, u64 end)
194 {
195 	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
196 
197 	csum_dirty_buffer(root, page);
198 	return 0;
199 }
200 
201 static int btree_writepage(struct page *page, struct writeback_control *wbc)
202 {
203 	struct extent_map_tree *tree;
204 	tree = &BTRFS_I(page->mapping->host)->extent_tree;
205 	return extent_write_full_page(tree, page, btree_get_extent, wbc);
206 }
207 
208 static int btree_writepages(struct address_space *mapping,
209 			    struct writeback_control *wbc)
210 {
211 	struct extent_map_tree *tree;
212 	tree = &BTRFS_I(mapping->host)->extent_tree;
213 	if (wbc->sync_mode == WB_SYNC_NONE) {
214 		u64 num_dirty;
215 		u64 start = 0;
216 		unsigned long thresh = 96 * 1024 * 1024;
217 
218 		if (wbc->for_kupdate)
219 			return 0;
220 
221 		if (current_is_pdflush()) {
222 			thresh = 96 * 1024 * 1024;
223 		} else {
224 			thresh = 8 * 1024 * 1024;
225 		}
226 		num_dirty = count_range_bits(tree, &start, thresh, EXTENT_DIRTY);
227 		if (num_dirty < thresh) {
228 			return 0;
229 		}
230 	}
231 	return extent_writepages(tree, mapping, btree_get_extent, wbc);
232 }
233 
234 int btree_readpage(struct file *file, struct page *page)
235 {
236 	struct extent_map_tree *tree;
237 	tree = &BTRFS_I(page->mapping->host)->extent_tree;
238 	return extent_read_full_page(tree, page, btree_get_extent);
239 }
240 
241 static int btree_releasepage(struct page *page, gfp_t unused_gfp_flags)
242 {
243 	struct extent_map_tree *tree;
244 	int ret;
245 
246 	tree = &BTRFS_I(page->mapping->host)->extent_tree;
247 	ret = try_release_extent_mapping(tree, page);
248 	if (ret == 1) {
249 		ClearPagePrivate(page);
250 		set_page_private(page, 0);
251 		page_cache_release(page);
252 	}
253 	return ret;
254 }
255 
256 static void btree_invalidatepage(struct page *page, unsigned long offset)
257 {
258 	struct extent_map_tree *tree;
259 	tree = &BTRFS_I(page->mapping->host)->extent_tree;
260 	extent_invalidatepage(tree, page, offset);
261 	btree_releasepage(page, GFP_NOFS);
262 }
263 
264 #if 0
265 static int btree_writepage(struct page *page, struct writeback_control *wbc)
266 {
267 	struct buffer_head *bh;
268 	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
269 	struct buffer_head *head;
270 	if (!page_has_buffers(page)) {
271 		create_empty_buffers(page, root->fs_info->sb->s_blocksize,
272 					(1 << BH_Dirty)|(1 << BH_Uptodate));
273 	}
274 	head = page_buffers(page);
275 	bh = head;
276 	do {
277 		if (buffer_dirty(bh))
278 			csum_tree_block(root, bh, 0);
279 		bh = bh->b_this_page;
280 	} while (bh != head);
281 	return block_write_full_page(page, btree_get_block, wbc);
282 }
283 #endif
284 
285 static struct address_space_operations btree_aops = {
286 	.readpage	= btree_readpage,
287 	.writepage	= btree_writepage,
288 	.writepages	= btree_writepages,
289 	.releasepage	= btree_releasepage,
290 	.invalidatepage = btree_invalidatepage,
291 	.sync_page	= block_sync_page,
292 };
293 
294 int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
295 {
296 	struct extent_buffer *buf = NULL;
297 	struct inode *btree_inode = root->fs_info->btree_inode;
298 	int ret = 0;
299 
300 	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
301 	if (!buf)
302 		return 0;
303 	read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
304 				 buf, 0, 0);
305 	free_extent_buffer(buf);
306 	return ret;
307 }
308 
309 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
310 				      u32 blocksize)
311 {
312 	struct extent_buffer *buf = NULL;
313 	struct inode *btree_inode = root->fs_info->btree_inode;
314 	struct extent_map_tree *extent_tree;
315 	int ret;
316 
317 	extent_tree = &BTRFS_I(btree_inode)->extent_tree;
318 
319 	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
320 	if (!buf)
321 		return NULL;
322 	read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
323 				 buf, 0, 1);
324 	if (buf->flags & EXTENT_CSUM) {
325 		return buf;
326 	}
327 	if (test_range_bit(extent_tree, buf->start, buf->start + buf->len - 1,
328 			   EXTENT_CSUM, 1)) {
329 		buf->flags |= EXTENT_CSUM;
330 		return buf;
331 	}
332 	ret = csum_tree_block(root, buf, 1);
333 	set_extent_bits(extent_tree, buf->start,
334 			buf->start + buf->len - 1,
335 			EXTENT_CSUM, GFP_NOFS);
336 	buf->flags |= EXTENT_CSUM;
337 	return buf;
338 }
339 
340 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
341 		     struct extent_buffer *buf)
342 {
343 	struct inode *btree_inode = root->fs_info->btree_inode;
344 	clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
345 	return 0;
346 }
347 
348 int wait_on_tree_block_writeback(struct btrfs_root *root,
349 				 struct extent_buffer *buf)
350 {
351 	struct inode *btree_inode = root->fs_info->btree_inode;
352 	wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->extent_tree,
353 					buf);
354 	return 0;
355 }
356 
357 static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
358 			u32 stripesize, struct btrfs_root *root,
359 			struct btrfs_fs_info *fs_info,
360 			u64 objectid)
361 {
362 	root->node = NULL;
363 	root->inode = NULL;
364 	root->commit_root = NULL;
365 	root->sectorsize = sectorsize;
366 	root->nodesize = nodesize;
367 	root->leafsize = leafsize;
368 	root->stripesize = stripesize;
369 	root->ref_cows = 0;
370 	root->fs_info = fs_info;
371 	root->objectid = objectid;
372 	root->last_trans = 0;
373 	root->highest_inode = 0;
374 	root->last_inode_alloc = 0;
375 	root->name = NULL;
376 	memset(&root->root_key, 0, sizeof(root->root_key));
377 	memset(&root->root_item, 0, sizeof(root->root_item));
378 	memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
379 	memset(&root->root_kobj, 0, sizeof(root->root_kobj));
380 	init_completion(&root->kobj_unregister);
381 	init_rwsem(&root->snap_sem);
382 	root->defrag_running = 0;
383 	root->defrag_level = 0;
384 	root->root_key.objectid = objectid;
385 	return 0;
386 }
387 
388 static int find_and_setup_root(struct btrfs_root *tree_root,
389 			       struct btrfs_fs_info *fs_info,
390 			       u64 objectid,
391 			       struct btrfs_root *root)
392 {
393 	int ret;
394 	u32 blocksize;
395 
396 	__setup_root(tree_root->nodesize, tree_root->leafsize,
397 		     tree_root->sectorsize, tree_root->stripesize,
398 		     root, fs_info, objectid);
399 	ret = btrfs_find_last_root(tree_root, objectid,
400 				   &root->root_item, &root->root_key);
401 	BUG_ON(ret);
402 
403 	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
404 	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
405 				     blocksize);
406 	BUG_ON(!root->node);
407 	return 0;
408 }
409 
410 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
411 					       struct btrfs_key *location)
412 {
413 	struct btrfs_root *root;
414 	struct btrfs_root *tree_root = fs_info->tree_root;
415 	struct btrfs_path *path;
416 	struct extent_buffer *l;
417 	u64 highest_inode;
418 	u32 blocksize;
419 	int ret = 0;
420 
421 	root = kzalloc(sizeof(*root), GFP_NOFS);
422 	if (!root)
423 		return ERR_PTR(-ENOMEM);
424 	if (location->offset == (u64)-1) {
425 		ret = find_and_setup_root(tree_root, fs_info,
426 					  location->objectid, root);
427 		if (ret) {
428 			kfree(root);
429 			return ERR_PTR(ret);
430 		}
431 		goto insert;
432 	}
433 
434 	__setup_root(tree_root->nodesize, tree_root->leafsize,
435 		     tree_root->sectorsize, tree_root->stripesize,
436 		     root, fs_info, location->objectid);
437 
438 	path = btrfs_alloc_path();
439 	BUG_ON(!path);
440 	ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
441 	if (ret != 0) {
442 		if (ret > 0)
443 			ret = -ENOENT;
444 		goto out;
445 	}
446 	l = path->nodes[0];
447 	read_extent_buffer(l, &root->root_item,
448 	       btrfs_item_ptr_offset(l, path->slots[0]),
449 	       sizeof(root->root_item));
450 	memcpy(&root->root_key, location, sizeof(*location));
451 	ret = 0;
452 out:
453 	btrfs_release_path(root, path);
454 	btrfs_free_path(path);
455 	if (ret) {
456 		kfree(root);
457 		return ERR_PTR(ret);
458 	}
459 	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
460 	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
461 				     blocksize);
462 	BUG_ON(!root->node);
463 insert:
464 	root->ref_cows = 1;
465 	ret = btrfs_find_highest_inode(root, &highest_inode);
466 	if (ret == 0) {
467 		root->highest_inode = highest_inode;
468 		root->last_inode_alloc = highest_inode;
469 	}
470 	return root;
471 }
472 
473 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
474 				      struct btrfs_key *location,
475 				      const char *name, int namelen)
476 {
477 	struct btrfs_root *root;
478 	int ret;
479 
480 	root = radix_tree_lookup(&fs_info->fs_roots_radix,
481 				 (unsigned long)location->objectid);
482 	if (root)
483 		return root;
484 
485 	root = btrfs_read_fs_root_no_radix(fs_info, location);
486 	if (IS_ERR(root))
487 		return root;
488 	ret = radix_tree_insert(&fs_info->fs_roots_radix,
489 				(unsigned long)root->root_key.objectid,
490 				root);
491 	if (ret) {
492 		free_extent_buffer(root->node);
493 		kfree(root);
494 		return ERR_PTR(ret);
495 	}
496 
497 	ret = btrfs_set_root_name(root, name, namelen);
498 	if (ret) {
499 		free_extent_buffer(root->node);
500 		kfree(root);
501 		return ERR_PTR(ret);
502 	}
503 
504 	ret = btrfs_sysfs_add_root(root);
505 	if (ret) {
506 		free_extent_buffer(root->node);
507 		kfree(root->name);
508 		kfree(root);
509 		return ERR_PTR(ret);
510 	}
511 
512 	ret = btrfs_find_dead_roots(fs_info->tree_root,
513 				    root->root_key.objectid, root);
514 	BUG_ON(ret);
515 
516 	return root;
517 }
518 #if 0
519 static int add_hasher(struct btrfs_fs_info *info, char *type) {
520 	struct btrfs_hasher *hasher;
521 
522 	hasher = kmalloc(sizeof(*hasher), GFP_NOFS);
523 	if (!hasher)
524 		return -ENOMEM;
525 	hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC);
526 	if (!hasher->hash_tfm) {
527 		kfree(hasher);
528 		return -EINVAL;
529 	}
530 	spin_lock(&info->hash_lock);
531 	list_add(&hasher->list, &info->hashers);
532 	spin_unlock(&info->hash_lock);
533 	return 0;
534 }
535 #endif
536 struct btrfs_root *open_ctree(struct super_block *sb)
537 {
538 	u32 sectorsize;
539 	u32 nodesize;
540 	u32 leafsize;
541 	u32 blocksize;
542 	u32 stripesize;
543 	struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
544 						 GFP_NOFS);
545 	struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
546 					       GFP_NOFS);
547 	struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
548 						GFP_NOFS);
549 	int ret;
550 	int err = -EIO;
551 	struct btrfs_super_block *disk_super;
552 
553 	if (!extent_root || !tree_root || !fs_info) {
554 		err = -ENOMEM;
555 		goto fail;
556 	}
557 	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
558 	INIT_LIST_HEAD(&fs_info->trans_list);
559 	INIT_LIST_HEAD(&fs_info->dead_roots);
560 	INIT_LIST_HEAD(&fs_info->hashers);
561 	spin_lock_init(&fs_info->hash_lock);
562 
563 	memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj));
564 	init_completion(&fs_info->kobj_unregister);
565 	sb_set_blocksize(sb, 4096);
566 	fs_info->running_transaction = NULL;
567 	fs_info->last_trans_committed = 0;
568 	fs_info->tree_root = tree_root;
569 	fs_info->extent_root = extent_root;
570 	fs_info->sb = sb;
571 	fs_info->mount_opt = 0;
572 	fs_info->max_extent = (u64)-1;
573 	fs_info->btree_inode = new_inode(sb);
574 	fs_info->btree_inode->i_ino = 1;
575 	fs_info->btree_inode->i_nlink = 1;
576 	fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
577 	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
578 	extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
579 			     fs_info->btree_inode->i_mapping,
580 			     GFP_NOFS);
581 	BTRFS_I(fs_info->btree_inode)->extent_tree.ops = &btree_extent_map_ops;
582 
583 	extent_map_tree_init(&fs_info->free_space_cache,
584 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
585 	extent_map_tree_init(&fs_info->block_group_cache,
586 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
587 	extent_map_tree_init(&fs_info->pinned_extents,
588 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
589 	extent_map_tree_init(&fs_info->pending_del,
590 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
591 	extent_map_tree_init(&fs_info->extent_ins,
592 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
593 	fs_info->do_barriers = 1;
594 	fs_info->closing = 0;
595 	fs_info->total_pinned = 0;
596 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
597 	INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info);
598 #else
599 	INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
600 #endif
601 	BTRFS_I(fs_info->btree_inode)->root = tree_root;
602 	memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
603 	       sizeof(struct btrfs_key));
604 	insert_inode_hash(fs_info->btree_inode);
605 	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
606 
607 	mutex_init(&fs_info->trans_mutex);
608 	mutex_init(&fs_info->fs_mutex);
609 
610 #if 0
611 	ret = add_hasher(fs_info, "crc32c");
612 	if (ret) {
613 		printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
614 		err = -ENOMEM;
615 		goto fail_iput;
616 	}
617 #endif
618 	__setup_root(512, 512, 512, 512, tree_root,
619 		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
620 
621 	fs_info->sb_buffer = read_tree_block(tree_root,
622 					     BTRFS_SUPER_INFO_OFFSET,
623 					     512);
624 
625 	if (!fs_info->sb_buffer)
626 		goto fail_iput;
627 
628 	read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0,
629 			   sizeof(fs_info->super_copy));
630 
631 	read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
632 			   (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
633 			   BTRFS_FSID_SIZE);
634 	disk_super = &fs_info->super_copy;
635 	if (!btrfs_super_root(disk_super))
636 		goto fail_sb_buffer;
637 
638 	nodesize = btrfs_super_nodesize(disk_super);
639 	leafsize = btrfs_super_leafsize(disk_super);
640 	sectorsize = btrfs_super_sectorsize(disk_super);
641 	stripesize = btrfs_super_stripesize(disk_super);
642 	tree_root->nodesize = nodesize;
643 	tree_root->leafsize = leafsize;
644 	tree_root->sectorsize = sectorsize;
645 	tree_root->stripesize = stripesize;
646 	sb_set_blocksize(sb, sectorsize);
647 
648 	i_size_write(fs_info->btree_inode,
649 		     btrfs_super_total_bytes(disk_super));
650 
651 	if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
652 		    sizeof(disk_super->magic))) {
653 		printk("btrfs: valid FS not found on %s\n", sb->s_id);
654 		goto fail_sb_buffer;
655 	}
656 
657 	blocksize = btrfs_level_size(tree_root,
658 				     btrfs_super_root_level(disk_super));
659 
660 	tree_root->node = read_tree_block(tree_root,
661 					  btrfs_super_root(disk_super),
662 					  blocksize);
663 	if (!tree_root->node)
664 		goto fail_sb_buffer;
665 
666 	mutex_lock(&fs_info->fs_mutex);
667 
668 	ret = find_and_setup_root(tree_root, fs_info,
669 				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
670 	if (ret) {
671 		mutex_unlock(&fs_info->fs_mutex);
672 		goto fail_tree_root;
673 	}
674 
675 	btrfs_read_block_groups(extent_root);
676 
677 	fs_info->generation = btrfs_super_generation(disk_super) + 1;
678 	mutex_unlock(&fs_info->fs_mutex);
679 	return tree_root;
680 
681 fail_tree_root:
682 	free_extent_buffer(tree_root->node);
683 fail_sb_buffer:
684 	free_extent_buffer(fs_info->sb_buffer);
685 fail_iput:
686 	iput(fs_info->btree_inode);
687 fail:
688 	kfree(extent_root);
689 	kfree(tree_root);
690 	kfree(fs_info);
691 	return ERR_PTR(err);
692 }
693 
694 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
695 		      *root)
696 {
697 	int ret;
698 	struct extent_buffer *super = root->fs_info->sb_buffer;
699 	struct inode *btree_inode = root->fs_info->btree_inode;
700 
701 	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, super);
702 	ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping,
703 				     super->start, super->len);
704 	return ret;
705 }
706 
707 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
708 {
709 	radix_tree_delete(&fs_info->fs_roots_radix,
710 			  (unsigned long)root->root_key.objectid);
711 	btrfs_sysfs_del_root(root);
712 	if (root->inode)
713 		iput(root->inode);
714 	if (root->node)
715 		free_extent_buffer(root->node);
716 	if (root->commit_root)
717 		free_extent_buffer(root->commit_root);
718 	if (root->name)
719 		kfree(root->name);
720 	kfree(root);
721 	return 0;
722 }
723 
724 static int del_fs_roots(struct btrfs_fs_info *fs_info)
725 {
726 	int ret;
727 	struct btrfs_root *gang[8];
728 	int i;
729 
730 	while(1) {
731 		ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
732 					     (void **)gang, 0,
733 					     ARRAY_SIZE(gang));
734 		if (!ret)
735 			break;
736 		for (i = 0; i < ret; i++)
737 			btrfs_free_fs_root(fs_info, gang[i]);
738 	}
739 	return 0;
740 }
741 
742 int close_ctree(struct btrfs_root *root)
743 {
744 	int ret;
745 	struct btrfs_trans_handle *trans;
746 	struct btrfs_fs_info *fs_info = root->fs_info;
747 
748 	fs_info->closing = 1;
749 	btrfs_transaction_flush_work(root);
750 	mutex_lock(&fs_info->fs_mutex);
751 	btrfs_defrag_dirty_roots(root->fs_info);
752 	trans = btrfs_start_transaction(root, 1);
753 	ret = btrfs_commit_transaction(trans, root);
754 	/* run commit again to  drop the original snapshot */
755 	trans = btrfs_start_transaction(root, 1);
756 	btrfs_commit_transaction(trans, root);
757 	ret = btrfs_write_and_wait_transaction(NULL, root);
758 	BUG_ON(ret);
759 	write_ctree_super(NULL, root);
760 	mutex_unlock(&fs_info->fs_mutex);
761 
762 	if (fs_info->extent_root->node)
763 		free_extent_buffer(fs_info->extent_root->node);
764 
765 	if (fs_info->tree_root->node)
766 		free_extent_buffer(fs_info->tree_root->node);
767 
768 	free_extent_buffer(fs_info->sb_buffer);
769 
770 	btrfs_free_block_groups(root->fs_info);
771 	del_fs_roots(fs_info);
772 
773 	filemap_write_and_wait(fs_info->btree_inode->i_mapping);
774 
775 	extent_map_tree_empty_lru(&fs_info->free_space_cache);
776 	extent_map_tree_empty_lru(&fs_info->block_group_cache);
777 	extent_map_tree_empty_lru(&fs_info->pinned_extents);
778 	extent_map_tree_empty_lru(&fs_info->pending_del);
779 	extent_map_tree_empty_lru(&fs_info->extent_ins);
780 	extent_map_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->extent_tree);
781 
782 	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
783 
784 	iput(fs_info->btree_inode);
785 #if 0
786 	while(!list_empty(&fs_info->hashers)) {
787 		struct btrfs_hasher *hasher;
788 		hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher,
789 				    hashers);
790 		list_del(&hasher->hashers);
791 		crypto_free_hash(&fs_info->hash_tfm);
792 		kfree(hasher);
793 	}
794 #endif
795 	kfree(fs_info->extent_root);
796 	kfree(fs_info->tree_root);
797 	return 0;
798 }
799 
800 int btrfs_buffer_uptodate(struct extent_buffer *buf)
801 {
802 	struct inode *btree_inode = buf->first_page->mapping->host;
803 	return extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree, buf);
804 }
805 
806 int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
807 {
808 	struct inode *btree_inode = buf->first_page->mapping->host;
809 	return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree,
810 					  buf);
811 }
812 
813 void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
814 {
815 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
816 	u64 transid = btrfs_header_generation(buf);
817 	struct inode *btree_inode = root->fs_info->btree_inode;
818 
819 	if (transid != root->fs_info->generation) {
820 		printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
821 			(unsigned long long)buf->start,
822 			transid, root->fs_info->generation);
823 		WARN_ON(1);
824 	}
825 	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
826 }
827 
828 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
829 {
830 	balance_dirty_pages_ratelimited_nr(
831 			root->fs_info->btree_inode->i_mapping, 1);
832 }
833 
834 void btrfs_set_buffer_defrag(struct extent_buffer *buf)
835 {
836 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
837 	struct inode *btree_inode = root->fs_info->btree_inode;
838 	set_extent_bits(&BTRFS_I(btree_inode)->extent_tree, buf->start,
839 			buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS);
840 }
841 
842 void btrfs_set_buffer_defrag_done(struct extent_buffer *buf)
843 {
844 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
845 	struct inode *btree_inode = root->fs_info->btree_inode;
846 	set_extent_bits(&BTRFS_I(btree_inode)->extent_tree, buf->start,
847 			buf->start + buf->len - 1, EXTENT_DEFRAG_DONE,
848 			GFP_NOFS);
849 }
850 
851 int btrfs_buffer_defrag(struct extent_buffer *buf)
852 {
853 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
854 	struct inode *btree_inode = root->fs_info->btree_inode;
855 	return test_range_bit(&BTRFS_I(btree_inode)->extent_tree,
856 		     buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0);
857 }
858 
859 int btrfs_buffer_defrag_done(struct extent_buffer *buf)
860 {
861 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
862 	struct inode *btree_inode = root->fs_info->btree_inode;
863 	return test_range_bit(&BTRFS_I(btree_inode)->extent_tree,
864 		     buf->start, buf->start + buf->len - 1,
865 		     EXTENT_DEFRAG_DONE, 0);
866 }
867 
868 int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf)
869 {
870 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
871 	struct inode *btree_inode = root->fs_info->btree_inode;
872 	return clear_extent_bits(&BTRFS_I(btree_inode)->extent_tree,
873 		     buf->start, buf->start + buf->len - 1,
874 		     EXTENT_DEFRAG_DONE, GFP_NOFS);
875 }
876 
877 int btrfs_clear_buffer_defrag(struct extent_buffer *buf)
878 {
879 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
880 	struct inode *btree_inode = root->fs_info->btree_inode;
881 	return clear_extent_bits(&BTRFS_I(btree_inode)->extent_tree,
882 		     buf->start, buf->start + buf->len - 1,
883 		     EXTENT_DEFRAG, GFP_NOFS);
884 }
885 
886 int btrfs_read_buffer(struct extent_buffer *buf)
887 {
888 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
889 	struct inode *btree_inode = root->fs_info->btree_inode;
890 	return read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
891 					buf, 0, 1);
892 }
893 
894 static struct extent_map_ops btree_extent_map_ops = {
895 	.writepage_io_hook = btree_writepage_io_hook,
896 };
897