projects / linux-2.6-block.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
Match the extent tree code to btrfs-progs for multi-device merging
[linux-2.6-block.git] / fs / btrfs / disk-io.c
... / ...
CommitLineData
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
34static 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
46static struct extent_io_ops btree_extent_io_ops;
47
48struct 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)->io_tree,
54 bytenr, blocksize, GFP_NOFS);
55 return eb;
56}
57
58struct 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)->io_tree,
65 bytenr, blocksize, NULL, GFP_NOFS);
66 return eb;
67}
68
69struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
70 size_t page_offset, u64 start, u64 len,
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
77again:
78 spin_lock(&em_tree->lock);
79 em = lookup_extent_mapping(em_tree, start, len);
80 spin_unlock(&em_tree->lock);
81 if (em) {
82 goto out;
83 }
84 em = alloc_extent_map(GFP_NOFS);
85 if (!em) {
86 em = ERR_PTR(-ENOMEM);
87 goto out;
88 }
89 em->start = 0;
90 em->len = i_size_read(inode);
91 em->block_start = 0;
92 em->bdev = inode->i_sb->s_bdev;
93
94 spin_lock(&em_tree->lock);
95 ret = add_extent_mapping(em_tree, em);
96 spin_unlock(&em_tree->lock);
97
98 if (ret == -EEXIST) {
99 free_extent_map(em);
100 em = NULL;
101 goto again;
102 } else if (ret) {
103 em = ERR_PTR(ret);
104 }
105out:
106 return em;
107}
108
109u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
110{
111 return crc32c(seed, data, len);
112}
113
114void btrfs_csum_final(u32 crc, char *result)
115{
116 *(__le32 *)result = ~cpu_to_le32(crc);
117}
118
119static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
120 int verify)
121{
122 char result[BTRFS_CRC32_SIZE];
123 unsigned long len;
124 unsigned long cur_len;
125 unsigned long offset = BTRFS_CSUM_SIZE;
126 char *map_token = NULL;
127 char *kaddr;
128 unsigned long map_start;
129 unsigned long map_len;
130 int err;
131 u32 crc = ~(u32)0;
132
133 len = buf->len - offset;
134 while(len > 0) {
135 err = map_private_extent_buffer(buf, offset, 32,
136 &map_token, &kaddr,
137 &map_start, &map_len, KM_USER0);
138 if (err) {
139 printk("failed to map extent buffer! %lu\n",
140 offset);
141 return 1;
142 }
143 cur_len = min(len, map_len - (offset - map_start));
144 crc = btrfs_csum_data(root, kaddr + offset - map_start,
145 crc, cur_len);
146 len -= cur_len;
147 offset += cur_len;
148 unmap_extent_buffer(buf, map_token, KM_USER0);
149 }
150 btrfs_csum_final(crc, result);
151
152 if (verify) {
153 int from_this_trans = 0;
154
155 if (root->fs_info->running_transaction &&
156 btrfs_header_generation(buf) ==
157 root->fs_info->running_transaction->transid)
158 from_this_trans = 1;
159
160 /* FIXME, this is not good */
161 if (from_this_trans == 0 &&
162 memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) {
163 u32 val;
164 u32 found = 0;
165 memcpy(&found, result, BTRFS_CRC32_SIZE);
166
167 read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE);
168 printk("btrfs: %s checksum verify failed on %llu "
169 "wanted %X found %X from_this_trans %d\n",
170 root->fs_info->sb->s_id,
171 buf->start, val, found, from_this_trans);
172 return 1;
173 }
174 } else {
175 write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE);
176 }
177 return 0;
178}
179
180
181int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
182{
183 struct extent_io_tree *tree;
184 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
185 u64 found_start;
186 int found_level;
187 unsigned long len;
188 struct extent_buffer *eb;
189 tree = &BTRFS_I(page->mapping->host)->io_tree;
190
191 if (page->private == EXTENT_PAGE_PRIVATE)
192 goto out;
193 if (!page->private)
194 goto out;
195 len = page->private >> 2;
196 if (len == 0) {
197 WARN_ON(1);
198 }
199 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
200 read_extent_buffer_pages(tree, eb, start + PAGE_CACHE_SIZE, 1,
201 btree_get_extent);
202 btrfs_clear_buffer_defrag(eb);
203 found_start = btrfs_header_bytenr(eb);
204 if (found_start != start) {
205 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
206 start, found_start, len);
207 WARN_ON(1);
208 goto err;
209 }
210 if (eb->first_page != page) {
211 printk("bad first page %lu %lu\n", eb->first_page->index,
212 page->index);
213 WARN_ON(1);
214 goto err;
215 }
216 if (!PageUptodate(page)) {
217 printk("csum not up to date page %lu\n", page->index);
218 WARN_ON(1);
219 goto err;
220 }
221 found_level = btrfs_header_level(eb);
222 csum_tree_block(root, eb, 0);
223err:
224 free_extent_buffer(eb);
225out:
226 return 0;
227}
228
229static int btree_writepage_io_hook(struct page *page, u64 start, u64 end)
230{
231 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
232
233 csum_dirty_buffer(root, page);
234 return 0;
235}
236
237static int btree_writepage(struct page *page, struct writeback_control *wbc)
238{
239 struct extent_io_tree *tree;
240 tree = &BTRFS_I(page->mapping->host)->io_tree;
241 return extent_write_full_page(tree, page, btree_get_extent, wbc);
242}
243
244static int btree_writepages(struct address_space *mapping,
245 struct writeback_control *wbc)
246{
247 struct extent_io_tree *tree;
248 tree = &BTRFS_I(mapping->host)->io_tree;
249 if (wbc->sync_mode == WB_SYNC_NONE) {
250 u64 num_dirty;
251 u64 start = 0;
252 unsigned long thresh = 96 * 1024 * 1024;
253
254 if (wbc->for_kupdate)
255 return 0;
256
257 if (current_is_pdflush()) {
258 thresh = 96 * 1024 * 1024;
259 } else {
260 thresh = 8 * 1024 * 1024;
261 }
262 num_dirty = count_range_bits(tree, &start, (u64)-1,
263 thresh, EXTENT_DIRTY);
264 if (num_dirty < thresh) {
265 return 0;
266 }
267 }
268 return extent_writepages(tree, mapping, btree_get_extent, wbc);
269}
270
271int btree_readpage(struct file *file, struct page *page)
272{
273 struct extent_io_tree *tree;
274 tree = &BTRFS_I(page->mapping->host)->io_tree;
275 return extent_read_full_page(tree, page, btree_get_extent);
276}
277
278static int btree_releasepage(struct page *page, gfp_t gfp_flags)
279{
280 struct extent_io_tree *tree;
281 struct extent_map_tree *map;
282 int ret;
283
284 tree = &BTRFS_I(page->mapping->host)->io_tree;
285 map = &BTRFS_I(page->mapping->host)->extent_tree;
286 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
287 if (ret == 1) {
288 ClearPagePrivate(page);
289 set_page_private(page, 0);
290 page_cache_release(page);
291 }
292 return ret;
293}
294
295static void btree_invalidatepage(struct page *page, unsigned long offset)
296{
297 struct extent_io_tree *tree;
298 tree = &BTRFS_I(page->mapping->host)->io_tree;
299 extent_invalidatepage(tree, page, offset);
300 btree_releasepage(page, GFP_NOFS);
301}
302
303#if 0
304static int btree_writepage(struct page *page, struct writeback_control *wbc)
305{
306 struct buffer_head *bh;
307 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
308 struct buffer_head *head;
309 if (!page_has_buffers(page)) {
310 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
311 (1 << BH_Dirty)|(1 << BH_Uptodate));
312 }
313 head = page_buffers(page);
314 bh = head;
315 do {
316 if (buffer_dirty(bh))
317 csum_tree_block(root, bh, 0);
318 bh = bh->b_this_page;
319 } while (bh != head);
320 return block_write_full_page(page, btree_get_block, wbc);
321}
322#endif
323
324static struct address_space_operations btree_aops = {
325 .readpage = btree_readpage,
326 .writepage = btree_writepage,
327 .writepages = btree_writepages,
328 .releasepage = btree_releasepage,
329 .invalidatepage = btree_invalidatepage,
330 .sync_page = block_sync_page,
331};
332
333int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
334{
335 struct extent_buffer *buf = NULL;
336 struct inode *btree_inode = root->fs_info->btree_inode;
337 int ret = 0;
338
339 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
340 if (!buf)
341 return 0;
342 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
343 buf, 0, 0, btree_get_extent);
344 free_extent_buffer(buf);
345 return ret;
346}
347
348struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
349 u32 blocksize)
350{
351 struct extent_buffer *buf = NULL;
352 struct inode *btree_inode = root->fs_info->btree_inode;
353 struct extent_io_tree *io_tree;
354 u64 end;
355 int ret;
356
357 io_tree = &BTRFS_I(btree_inode)->io_tree;
358
359 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
360 if (!buf)
361 return NULL;
362 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, buf, 0, 1,
363 btree_get_extent);
364
365 if (buf->flags & EXTENT_CSUM)
366 return buf;
367
368 end = buf->start + PAGE_CACHE_SIZE - 1;
369 if (test_range_bit(io_tree, buf->start, end, EXTENT_CSUM, 1)) {
370 buf->flags |= EXTENT_CSUM;
371 return buf;
372 }
373
374 lock_extent(io_tree, buf->start, end, GFP_NOFS);
375
376 if (test_range_bit(io_tree, buf->start, end, EXTENT_CSUM, 1)) {
377 buf->flags |= EXTENT_CSUM;
378 goto out_unlock;
379 }
380
381 ret = csum_tree_block(root, buf, 1);
382 set_extent_bits(io_tree, buf->start, end, EXTENT_CSUM, GFP_NOFS);
383 buf->flags |= EXTENT_CSUM;
384
385out_unlock:
386 unlock_extent(io_tree, buf->start, end, GFP_NOFS);
387 return buf;
388}
389
390int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
391 struct extent_buffer *buf)
392{
393 struct inode *btree_inode = root->fs_info->btree_inode;
394 if (btrfs_header_generation(buf) ==
395 root->fs_info->running_transaction->transid)
396 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
397 buf);
398 return 0;
399}
400
401int wait_on_tree_block_writeback(struct btrfs_root *root,
402 struct extent_buffer *buf)
403{
404 struct inode *btree_inode = root->fs_info->btree_inode;
405 wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree,
406 buf);
407 return 0;
408}
409
410static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
411 u32 stripesize, struct btrfs_root *root,
412 struct btrfs_fs_info *fs_info,
413 u64 objectid)
414{
415 root->node = NULL;
416 root->inode = NULL;
417 root->commit_root = NULL;
418 root->sectorsize = sectorsize;
419 root->nodesize = nodesize;
420 root->leafsize = leafsize;
421 root->stripesize = stripesize;
422 root->ref_cows = 0;
423 root->fs_info = fs_info;
424 root->objectid = objectid;
425 root->last_trans = 0;
426 root->highest_inode = 0;
427 root->last_inode_alloc = 0;
428 root->name = NULL;
429 root->in_sysfs = 0;
430 memset(&root->root_key, 0, sizeof(root->root_key));
431 memset(&root->root_item, 0, sizeof(root->root_item));
432 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
433 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
434 init_completion(&root->kobj_unregister);
435 root->defrag_running = 0;
436 root->defrag_level = 0;
437 root->root_key.objectid = objectid;
438 return 0;
439}
440
441static int find_and_setup_root(struct btrfs_root *tree_root,
442 struct btrfs_fs_info *fs_info,
443 u64 objectid,
444 struct btrfs_root *root)
445{
446 int ret;
447 u32 blocksize;
448
449 __setup_root(tree_root->nodesize, tree_root->leafsize,
450 tree_root->sectorsize, tree_root->stripesize,
451 root, fs_info, objectid);
452 ret = btrfs_find_last_root(tree_root, objectid,
453 &root->root_item, &root->root_key);
454 BUG_ON(ret);
455
456 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
457 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
458 blocksize);
459 BUG_ON(!root->node);
460 return 0;
461}
462
463struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
464 struct btrfs_key *location)
465{
466 struct btrfs_root *root;
467 struct btrfs_root *tree_root = fs_info->tree_root;
468 struct btrfs_path *path;
469 struct extent_buffer *l;
470 u64 highest_inode;
471 u32 blocksize;
472 int ret = 0;
473
474 root = kzalloc(sizeof(*root), GFP_NOFS);
475 if (!root)
476 return ERR_PTR(-ENOMEM);
477 if (location->offset == (u64)-1) {
478 ret = find_and_setup_root(tree_root, fs_info,
479 location->objectid, root);
480 if (ret) {
481 kfree(root);
482 return ERR_PTR(ret);
483 }
484 goto insert;
485 }
486
487 __setup_root(tree_root->nodesize, tree_root->leafsize,
488 tree_root->sectorsize, tree_root->stripesize,
489 root, fs_info, location->objectid);
490
491 path = btrfs_alloc_path();
492 BUG_ON(!path);
493 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
494 if (ret != 0) {
495 if (ret > 0)
496 ret = -ENOENT;
497 goto out;
498 }
499 l = path->nodes[0];
500 read_extent_buffer(l, &root->root_item,
501 btrfs_item_ptr_offset(l, path->slots[0]),
502 sizeof(root->root_item));
503 memcpy(&root->root_key, location, sizeof(*location));
504 ret = 0;
505out:
506 btrfs_release_path(root, path);
507 btrfs_free_path(path);
508 if (ret) {
509 kfree(root);
510 return ERR_PTR(ret);
511 }
512 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
513 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
514 blocksize);
515 BUG_ON(!root->node);
516insert:
517 root->ref_cows = 1;
518 ret = btrfs_find_highest_inode(root, &highest_inode);
519 if (ret == 0) {
520 root->highest_inode = highest_inode;
521 root->last_inode_alloc = highest_inode;
522 }
523 return root;
524}
525
526struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
527 u64 root_objectid)
528{
529 struct btrfs_root *root;
530
531 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
532 return fs_info->tree_root;
533 if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
534 return fs_info->extent_root;
535
536 root = radix_tree_lookup(&fs_info->fs_roots_radix,
537 (unsigned long)root_objectid);
538 return root;
539}
540
541struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
542 struct btrfs_key *location)
543{
544 struct btrfs_root *root;
545 int ret;
546
547 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
548 return fs_info->tree_root;
549 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
550 return fs_info->extent_root;
551
552 root = radix_tree_lookup(&fs_info->fs_roots_radix,
553 (unsigned long)location->objectid);
554 if (root)
555 return root;
556
557 root = btrfs_read_fs_root_no_radix(fs_info, location);
558 if (IS_ERR(root))
559 return root;
560 ret = radix_tree_insert(&fs_info->fs_roots_radix,
561 (unsigned long)root->root_key.objectid,
562 root);
563 if (ret) {
564 free_extent_buffer(root->node);
565 kfree(root);
566 return ERR_PTR(ret);
567 }
568 ret = btrfs_find_dead_roots(fs_info->tree_root,
569 root->root_key.objectid, root);
570 BUG_ON(ret);
571
572 return root;
573}
574
575struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
576 struct btrfs_key *location,
577 const char *name, int namelen)
578{
579 struct btrfs_root *root;
580 int ret;
581
582 root = btrfs_read_fs_root_no_name(fs_info, location);
583 if (!root)
584 return NULL;
585
586 if (root->in_sysfs)
587 return root;
588
589 ret = btrfs_set_root_name(root, name, namelen);
590 if (ret) {
591 free_extent_buffer(root->node);
592 kfree(root);
593 return ERR_PTR(ret);
594 }
595
596 ret = btrfs_sysfs_add_root(root);
597 if (ret) {
598 free_extent_buffer(root->node);
599 kfree(root->name);
600 kfree(root);
601 return ERR_PTR(ret);
602 }
603 root->in_sysfs = 1;
604 return root;
605}
606#if 0
607static int add_hasher(struct btrfs_fs_info *info, char *type) {
608 struct btrfs_hasher *hasher;
609
610 hasher = kmalloc(sizeof(*hasher), GFP_NOFS);
611 if (!hasher)
612 return -ENOMEM;
613 hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC);
614 if (!hasher->hash_tfm) {
615 kfree(hasher);
616 return -EINVAL;
617 }
618 spin_lock(&info->hash_lock);
619 list_add(&hasher->list, &info->hashers);
620 spin_unlock(&info->hash_lock);
621 return 0;
622}
623#endif
624struct btrfs_root *open_ctree(struct super_block *sb)
625{
626 u32 sectorsize;
627 u32 nodesize;
628 u32 leafsize;
629 u32 blocksize;
630 u32 stripesize;
631 struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
632 GFP_NOFS);
633 struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
634 GFP_NOFS);
635 struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
636 GFP_NOFS);
637 int ret;
638 int err = -EIO;
639 struct btrfs_super_block *disk_super;
640
641 if (!extent_root || !tree_root || !fs_info) {
642 err = -ENOMEM;
643 goto fail;
644 }
645 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
646 INIT_LIST_HEAD(&fs_info->trans_list);
647 INIT_LIST_HEAD(&fs_info->dead_roots);
648 INIT_LIST_HEAD(&fs_info->hashers);
649 spin_lock_init(&fs_info->hash_lock);
650 spin_lock_init(&fs_info->delalloc_lock);
651 spin_lock_init(&fs_info->new_trans_lock);
652
653 memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj));
654 init_completion(&fs_info->kobj_unregister);
655 sb_set_blocksize(sb, 4096);
656 fs_info->running_transaction = NULL;
657 fs_info->last_trans_committed = 0;
658 fs_info->tree_root = tree_root;
659 fs_info->extent_root = extent_root;
660 fs_info->sb = sb;
661 fs_info->throttles = 0;
662 fs_info->mount_opt = 0;
663 fs_info->max_extent = (u64)-1;
664 fs_info->max_inline = 8192 * 1024;
665 fs_info->delalloc_bytes = 0;
666 fs_info->btree_inode = new_inode(sb);
667 fs_info->btree_inode->i_ino = 1;
668 fs_info->btree_inode->i_nlink = 1;
669 fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
670 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
671 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
672 fs_info->btree_inode->i_mapping,
673 GFP_NOFS);
674 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
675 GFP_NOFS);
676
677 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
678
679 extent_io_tree_init(&fs_info->free_space_cache,
680 fs_info->btree_inode->i_mapping, GFP_NOFS);
681 extent_io_tree_init(&fs_info->block_group_cache,
682 fs_info->btree_inode->i_mapping, GFP_NOFS);
683 extent_io_tree_init(&fs_info->pinned_extents,
684 fs_info->btree_inode->i_mapping, GFP_NOFS);
685 extent_io_tree_init(&fs_info->pending_del,
686 fs_info->btree_inode->i_mapping, GFP_NOFS);
687 extent_io_tree_init(&fs_info->extent_ins,
688 fs_info->btree_inode->i_mapping, GFP_NOFS);
689 fs_info->do_barriers = 1;
690 fs_info->closing = 0;
691 fs_info->total_pinned = 0;
692 fs_info->last_alloc = 0;
693 fs_info->last_data_alloc = 0;
694
695#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
696 INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info);
697#else
698 INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
699#endif
700 BTRFS_I(fs_info->btree_inode)->root = tree_root;
701 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
702 sizeof(struct btrfs_key));
703 insert_inode_hash(fs_info->btree_inode);
704 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
705
706 mutex_init(&fs_info->trans_mutex);
707 mutex_init(&fs_info->fs_mutex);
708
709#if 0
710 ret = add_hasher(fs_info, "crc32c");
711 if (ret) {
712 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
713 err = -ENOMEM;
714 goto fail_iput;
715 }
716#endif
717 __setup_root(512, 512, 512, 512, tree_root,
718 fs_info, BTRFS_ROOT_TREE_OBJECTID);
719
720 fs_info->sb_buffer = read_tree_block(tree_root,
721 BTRFS_SUPER_INFO_OFFSET,
722 512);
723
724 if (!fs_info->sb_buffer)
725 goto fail_iput;
726
727 read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0,
728 sizeof(fs_info->super_copy));
729
730 read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
731 (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
732 BTRFS_FSID_SIZE);
733 disk_super = &fs_info->super_copy;
734 if (!btrfs_super_root(disk_super))
735 goto fail_sb_buffer;
736
737 nodesize = btrfs_super_nodesize(disk_super);
738 leafsize = btrfs_super_leafsize(disk_super);
739 sectorsize = btrfs_super_sectorsize(disk_super);
740 stripesize = btrfs_super_stripesize(disk_super);
741 tree_root->nodesize = nodesize;
742 tree_root->leafsize = leafsize;
743 tree_root->sectorsize = sectorsize;
744 tree_root->stripesize = stripesize;
745 sb_set_blocksize(sb, sectorsize);
746
747 i_size_write(fs_info->btree_inode,
748 btrfs_super_total_bytes(disk_super));
749
750 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
751 sizeof(disk_super->magic))) {
752 printk("btrfs: valid FS not found on %s\n", sb->s_id);
753 goto fail_sb_buffer;
754 }
755
756 blocksize = btrfs_level_size(tree_root,
757 btrfs_super_root_level(disk_super));
758
759 tree_root->node = read_tree_block(tree_root,
760 btrfs_super_root(disk_super),
761 blocksize);
762 if (!tree_root->node)
763 goto fail_sb_buffer;
764
765 mutex_lock(&fs_info->fs_mutex);
766
767 ret = find_and_setup_root(tree_root, fs_info,
768 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
769 if (ret) {
770 mutex_unlock(&fs_info->fs_mutex);
771 goto fail_tree_root;
772 }
773
774 btrfs_read_block_groups(extent_root);
775
776 fs_info->generation = btrfs_super_generation(disk_super) + 1;
777 mutex_unlock(&fs_info->fs_mutex);
778 return tree_root;
779
780fail_tree_root:
781 free_extent_buffer(tree_root->node);
782fail_sb_buffer:
783 free_extent_buffer(fs_info->sb_buffer);
784fail_iput:
785 iput(fs_info->btree_inode);
786fail:
787 kfree(extent_root);
788 kfree(tree_root);
789 kfree(fs_info);
790 return ERR_PTR(err);
791}
792
793int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
794 *root)
795{
796 int ret;
797 struct extent_buffer *super = root->fs_info->sb_buffer;
798 struct inode *btree_inode = root->fs_info->btree_inode;
799 struct super_block *sb = root->fs_info->sb;
800
801 if (!btrfs_test_opt(root, NOBARRIER))
802 blkdev_issue_flush(sb->s_bdev, NULL);
803 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, super);
804 ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping,
805 super->start, super->len);
806 if (!btrfs_test_opt(root, NOBARRIER))
807 blkdev_issue_flush(sb->s_bdev, NULL);
808 return ret;
809}
810
811int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
812{
813 radix_tree_delete(&fs_info->fs_roots_radix,
814 (unsigned long)root->root_key.objectid);
815 if (root->in_sysfs)
816 btrfs_sysfs_del_root(root);
817 if (root->inode)
818 iput(root->inode);
819 if (root->node)
820 free_extent_buffer(root->node);
821 if (root->commit_root)
822 free_extent_buffer(root->commit_root);
823 if (root->name)
824 kfree(root->name);
825 kfree(root);
826 return 0;
827}
828
829static int del_fs_roots(struct btrfs_fs_info *fs_info)
830{
831 int ret;
832 struct btrfs_root *gang[8];
833 int i;
834
835 while(1) {
836 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
837 (void **)gang, 0,
838 ARRAY_SIZE(gang));
839 if (!ret)
840 break;
841 for (i = 0; i < ret; i++)
842 btrfs_free_fs_root(fs_info, gang[i]);
843 }
844 return 0;
845}
846
847int close_ctree(struct btrfs_root *root)
848{
849 int ret;
850 struct btrfs_trans_handle *trans;
851 struct btrfs_fs_info *fs_info = root->fs_info;
852
853 fs_info->closing = 1;
854 btrfs_transaction_flush_work(root);
855 mutex_lock(&fs_info->fs_mutex);
856 btrfs_defrag_dirty_roots(root->fs_info);
857 trans = btrfs_start_transaction(root, 1);
858 ret = btrfs_commit_transaction(trans, root);
859 /* run commit again to drop the original snapshot */
860 trans = btrfs_start_transaction(root, 1);
861 btrfs_commit_transaction(trans, root);
862 ret = btrfs_write_and_wait_transaction(NULL, root);
863 BUG_ON(ret);
864 write_ctree_super(NULL, root);
865 mutex_unlock(&fs_info->fs_mutex);
866
867 if (fs_info->delalloc_bytes) {
868 printk("btrfs: at unmount delalloc count %Lu\n",
869 fs_info->delalloc_bytes);
870 }
871 if (fs_info->extent_root->node)
872 free_extent_buffer(fs_info->extent_root->node);
873
874 if (fs_info->tree_root->node)
875 free_extent_buffer(fs_info->tree_root->node);
876
877 free_extent_buffer(fs_info->sb_buffer);
878
879 btrfs_free_block_groups(root->fs_info);
880 del_fs_roots(fs_info);
881
882 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
883
884 extent_io_tree_empty_lru(&fs_info->free_space_cache);
885 extent_io_tree_empty_lru(&fs_info->block_group_cache);
886 extent_io_tree_empty_lru(&fs_info->pinned_extents);
887 extent_io_tree_empty_lru(&fs_info->pending_del);
888 extent_io_tree_empty_lru(&fs_info->extent_ins);
889 extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
890
891 truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
892
893 iput(fs_info->btree_inode);
894#if 0
895 while(!list_empty(&fs_info->hashers)) {
896 struct btrfs_hasher *hasher;
897 hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher,
898 hashers);
899 list_del(&hasher->hashers);
900 crypto_free_hash(&fs_info->hash_tfm);
901 kfree(hasher);
902 }
903#endif
904 kfree(fs_info->extent_root);
905 kfree(fs_info->tree_root);
906 return 0;
907}
908
909int btrfs_buffer_uptodate(struct extent_buffer *buf)
910{
911 struct inode *btree_inode = buf->first_page->mapping->host;
912 return extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf);
913}
914
915int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
916{
917 struct inode *btree_inode = buf->first_page->mapping->host;
918 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
919 buf);
920}
921
922void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
923{
924 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
925 u64 transid = btrfs_header_generation(buf);
926 struct inode *btree_inode = root->fs_info->btree_inode;
927
928 if (transid != root->fs_info->generation) {
929 printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
930 (unsigned long long)buf->start,
931 transid, root->fs_info->generation);
932 WARN_ON(1);
933 }
934 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf);
935}
936
937void btrfs_throttle(struct btrfs_root *root)
938{
939 struct backing_dev_info *bdi;
940
941 bdi = root->fs_info->sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
942 if (root->fs_info->throttles && bdi_write_congested(bdi)) {
943#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18)
944 congestion_wait(WRITE, HZ/20);
945#else
946 blk_congestion_wait(WRITE, HZ/20);
947#endif
948 }
949}
950
951void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
952{
953 balance_dirty_pages_ratelimited_nr(
954 root->fs_info->btree_inode->i_mapping, 1);
955}
956
957void btrfs_set_buffer_defrag(struct extent_buffer *buf)
958{
959 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
960 struct inode *btree_inode = root->fs_info->btree_inode;
961 set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
962 buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS);
963}
964
965void btrfs_set_buffer_defrag_done(struct extent_buffer *buf)
966{
967 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
968 struct inode *btree_inode = root->fs_info->btree_inode;
969 set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
970 buf->start + buf->len - 1, EXTENT_DEFRAG_DONE,
971 GFP_NOFS);
972}
973
974int btrfs_buffer_defrag(struct extent_buffer *buf)
975{
976 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
977 struct inode *btree_inode = root->fs_info->btree_inode;
978 return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
979 buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0);
980}
981
982int btrfs_buffer_defrag_done(struct extent_buffer *buf)
983{
984 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
985 struct inode *btree_inode = root->fs_info->btree_inode;
986 return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
987 buf->start, buf->start + buf->len - 1,
988 EXTENT_DEFRAG_DONE, 0);
989}
990
991int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf)
992{
993 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
994 struct inode *btree_inode = root->fs_info->btree_inode;
995 return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
996 buf->start, buf->start + buf->len - 1,
997 EXTENT_DEFRAG_DONE, GFP_NOFS);
998}
999
1000int btrfs_clear_buffer_defrag(struct extent_buffer *buf)
1001{
1002 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1003 struct inode *btree_inode = root->fs_info->btree_inode;
1004 return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
1005 buf->start, buf->start + buf->len - 1,
1006 EXTENT_DEFRAG, GFP_NOFS);
1007}
1008
1009int btrfs_read_buffer(struct extent_buffer *buf)
1010{
1011 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1012 struct inode *btree_inode = root->fs_info->btree_inode;
1013 return read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
1014 buf, 0, 1, btree_get_extent);
1015}
1016
1017static struct extent_io_ops btree_extent_io_ops = {
1018 .writepage_io_hook = btree_writepage_io_hook,
1019};
Linux 6.x block layer and io_uring tree(s)