projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Btrfs: Add chunk uuids and update multi-device back references
[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/scatterlist.h>
22 #include <linux/swap.h>
23 #include <linux/radix-tree.h>
24 #include <linux/writeback.h>
25 #include <linux/buffer_head.h> // for block_sync_page
26 #include <linux/workqueue.h>
27 #include "crc32c.h"
28 #include "ctree.h"
29 #include "disk-io.h"
30 #include "transaction.h"
31 #include "btrfs_inode.h"
32 #include "volumes.h"
33 #include "print-tree.h"
34
35 #if 0
36 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
37 {
38 if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) {
39 printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n",
40 (unsigned long long)extent_buffer_blocknr(buf),
41 (unsigned long long)btrfs_header_blocknr(buf));
42 return 1;
43 }
44 return 0;
45 }
46 #endif
47
48 static struct extent_io_ops btree_extent_io_ops;
49 static struct workqueue_struct *end_io_workqueue;
50
51 struct end_io_wq {
52 struct bio *bio;
53 bio_end_io_t *end_io;
54 void *private;
55 struct btrfs_fs_info *info;
56 int error;
57 int metadata;
58 struct list_head list;
59 };
60
61 struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
62 size_t page_offset, u64 start, u64 len,
63 int create)
64 {
65 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
66 struct extent_map *em;
67 int ret;
68
69 again:
70 spin_lock(&em_tree->lock);
71 em = lookup_extent_mapping(em_tree, start, len);
72 spin_unlock(&em_tree->lock);
73 if (em) {
74 goto out;
75 }
76 em = alloc_extent_map(GFP_NOFS);
77 if (!em) {
78 em = ERR_PTR(-ENOMEM);
79 goto out;
80 }
81 em->start = 0;
82 em->len = i_size_read(inode);
83 em->block_start = 0;
84 em->bdev = inode->i_sb->s_bdev;
85
86 spin_lock(&em_tree->lock);
87 ret = add_extent_mapping(em_tree, em);
88 spin_unlock(&em_tree->lock);
89
90 if (ret == -EEXIST) {
91 free_extent_map(em);
92 em = NULL;
93 goto again;
94 } else if (ret) {
95 em = ERR_PTR(ret);
96 }
97 out:
98 return em;
99 }
100
101 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
102 {
103 return btrfs_crc32c(seed, data, len);
104 }
105
106 void btrfs_csum_final(u32 crc, char *result)
107 {
108 *(__le32 *)result = ~cpu_to_le32(crc);
109 }
110
111 static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
112 int verify)
113 {
114 char result[BTRFS_CRC32_SIZE];
115 unsigned long len;
116 unsigned long cur_len;
117 unsigned long offset = BTRFS_CSUM_SIZE;
118 char *map_token = NULL;
119 char *kaddr;
120 unsigned long map_start;
121 unsigned long map_len;
122 int err;
123 u32 crc = ~(u32)0;
124
125 len = buf->len - offset;
126 while(len > 0) {
127 err = map_private_extent_buffer(buf, offset, 32,
128 &map_token, &kaddr,
129 &map_start, &map_len, KM_USER0);
130 if (err) {
131 printk("failed to map extent buffer! %lu\n",
132 offset);
133 return 1;
134 }
135 cur_len = min(len, map_len - (offset - map_start));
136 crc = btrfs_csum_data(root, kaddr + offset - map_start,
137 crc, cur_len);
138 len -= cur_len;
139 offset += cur_len;
140 unmap_extent_buffer(buf, map_token, KM_USER0);
141 }
142 btrfs_csum_final(crc, result);
143
144 if (verify) {
145 int from_this_trans = 0;
146
147 if (root->fs_info->running_transaction &&
148 btrfs_header_generation(buf) ==
149 root->fs_info->running_transaction->transid)
150 from_this_trans = 1;
151
152 /* FIXME, this is not good */
153 if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) {
154 u32 val;
155 u32 found = 0;
156 memcpy(&found, result, BTRFS_CRC32_SIZE);
157
158 read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE);
159 printk("btrfs: %s checksum verify failed on %llu "
160 "wanted %X found %X from_this_trans %d "
161 "level %d\n",
162 root->fs_info->sb->s_id,
163 buf->start, val, found, from_this_trans,
164 btrfs_header_level(buf));
165 return 1;
166 }
167 } else {
168 write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE);
169 }
170 return 0;
171 }
172
173 static int btree_read_extent_buffer_pages(struct btrfs_root *root,
174 struct extent_buffer *eb,
175 u64 start)
176 {
177 struct extent_io_tree *io_tree;
178 int ret;
179 int num_copies = 0;
180 int mirror_num = 0;
181
182 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
183 while (1) {
184 ret = read_extent_buffer_pages(io_tree, eb, start, 1,
185 btree_get_extent, mirror_num);
186 if (!ret) {
187 if (mirror_num)
188 printk("good read %Lu mirror %d total %d\n", eb->start, mirror_num, num_copies);
189 return ret;
190 }
191 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
192 eb->start, eb->len);
193 printk("failed to read %Lu mirror %d total %d\n", eb->start, mirror_num, num_copies);
194 if (num_copies == 1) {
195 printk("reading %Lu failed only one copy\n", eb->start);
196 return ret;
197 }
198 mirror_num++;
199 if (mirror_num > num_copies) {
200 printk("bailing at mirror %d of %d\n", mirror_num, num_copies);
201 return ret;
202 }
203 }
204 printk("read extent buffer page last\n");
205 return -EIO;
206 }
207
208 int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
209 {
210 struct extent_io_tree *tree;
211 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
212 u64 found_start;
213 int found_level;
214 unsigned long len;
215 struct extent_buffer *eb;
216 int ret;
217
218 tree = &BTRFS_I(page->mapping->host)->io_tree;
219
220 if (page->private == EXTENT_PAGE_PRIVATE)
221 goto out;
222 if (!page->private)
223 goto out;
224 len = page->private >> 2;
225 if (len == 0) {
226 WARN_ON(1);
227 }
228 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
229 ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE);
230 BUG_ON(ret);
231 btrfs_clear_buffer_defrag(eb);
232 found_start = btrfs_header_bytenr(eb);
233 if (found_start != start) {
234 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
235 start, found_start, len);
236 WARN_ON(1);
237 goto err;
238 }
239 if (eb->first_page != page) {
240 printk("bad first page %lu %lu\n", eb->first_page->index,
241 page->index);
242 WARN_ON(1);
243 goto err;
244 }
245 if (!PageUptodate(page)) {
246 printk("csum not up to date page %lu\n", page->index);
247 WARN_ON(1);
248 goto err;
249 }
250 found_level = btrfs_header_level(eb);
251 spin_lock(&root->fs_info->hash_lock);
252 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
253 spin_unlock(&root->fs_info->hash_lock);
254 csum_tree_block(root, eb, 0);
255 err:
256 free_extent_buffer(eb);
257 out:
258 return 0;
259 }
260
261 static int btree_writepage_io_hook(struct page *page, u64 start, u64 end)
262 {
263 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
264
265 csum_dirty_buffer(root, page);
266 return 0;
267 }
268
269 int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
270 struct extent_state *state)
271 {
272 struct extent_io_tree *tree;
273 u64 found_start;
274 int found_level;
275 unsigned long len;
276 struct extent_buffer *eb;
277 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
278 int ret = 0;
279
280 tree = &BTRFS_I(page->mapping->host)->io_tree;
281 if (page->private == EXTENT_PAGE_PRIVATE)
282 goto out;
283 if (!page->private)
284 goto out;
285 len = page->private >> 2;
286 if (len == 0) {
287 WARN_ON(1);
288 }
289 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
290
291 btrfs_clear_buffer_defrag(eb);
292 found_start = btrfs_header_bytenr(eb);
293 if (found_start != start) {
294 printk("bad start on %Lu found %Lu\n", eb->start, found_start);
295 ret = -EIO;
296 goto err;
297 }
298 if (eb->first_page != page) {
299 printk("bad first page %lu %lu\n", eb->first_page->index,
300 page->index);
301 WARN_ON(1);
302 ret = -EIO;
303 goto err;
304 }
305 found_level = btrfs_header_level(eb);
306
307 ret = csum_tree_block(root, eb, 1);
308 if (ret)
309 ret = -EIO;
310
311 end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
312 end = eb->start + end - 1;
313 release_extent_buffer_tail_pages(eb);
314 err:
315 free_extent_buffer(eb);
316 out:
317 return ret;
318 }
319
320 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
321 static void end_workqueue_bio(struct bio *bio, int err)
322 #else
323 static int end_workqueue_bio(struct bio *bio,
324 unsigned int bytes_done, int err)
325 #endif
326 {
327 struct end_io_wq *end_io_wq = bio->bi_private;
328 struct btrfs_fs_info *fs_info;
329 unsigned long flags;
330
331 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
332 if (bio->bi_size)
333 return 1;
334 #endif
335
336 fs_info = end_io_wq->info;
337 spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
338 end_io_wq->error = err;
339 list_add_tail(&end_io_wq->list, &fs_info->end_io_work_list);
340 spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags);
341 queue_work(end_io_workqueue, &fs_info->end_io_work);
342
343 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
344 return 0;
345 #endif
346 }
347
348 int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
349 int metadata)
350 {
351 struct end_io_wq *end_io_wq;
352 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
353 if (!end_io_wq)
354 return -ENOMEM;
355
356 end_io_wq->private = bio->bi_private;
357 end_io_wq->end_io = bio->bi_end_io;
358 end_io_wq->info = info;
359 end_io_wq->error = 0;
360 end_io_wq->bio = bio;
361 end_io_wq->metadata = metadata;
362
363 bio->bi_private = end_io_wq;
364 bio->bi_end_io = end_workqueue_bio;
365 return 0;
366 }
367
368 static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
369 int mirror_num)
370 {
371 struct btrfs_root *root = BTRFS_I(inode)->root;
372 u64 offset;
373 int ret;
374
375 offset = bio->bi_sector << 9;
376
377 if (rw & (1 << BIO_RW)) {
378 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num);
379 }
380
381 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1);
382 BUG_ON(ret);
383
384 if (offset == BTRFS_SUPER_INFO_OFFSET) {
385 bio->bi_bdev = root->fs_info->fs_devices->latest_bdev;
386 submit_bio(rw, bio);
387 return 0;
388 }
389 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num);
390 }
391
392 static int btree_writepage(struct page *page, struct writeback_control *wbc)
393 {
394 struct extent_io_tree *tree;
395 tree = &BTRFS_I(page->mapping->host)->io_tree;
396 return extent_write_full_page(tree, page, btree_get_extent, wbc);
397 }
398
399 static int btree_writepages(struct address_space *mapping,
400 struct writeback_control *wbc)
401 {
402 struct extent_io_tree *tree;
403 tree = &BTRFS_I(mapping->host)->io_tree;
404 if (wbc->sync_mode == WB_SYNC_NONE) {
405 u64 num_dirty;
406 u64 start = 0;
407 unsigned long thresh = 96 * 1024 * 1024;
408
409 if (wbc->for_kupdate)
410 return 0;
411
412 if (current_is_pdflush()) {
413 thresh = 96 * 1024 * 1024;
414 } else {
415 thresh = 8 * 1024 * 1024;
416 }
417 num_dirty = count_range_bits(tree, &start, (u64)-1,
418 thresh, EXTENT_DIRTY);
419 if (num_dirty < thresh) {
420 return 0;
421 }
422 }
423 return extent_writepages(tree, mapping, btree_get_extent, wbc);
424 }
425
426 int btree_readpage(struct file *file, struct page *page)
427 {
428 struct extent_io_tree *tree;
429 tree = &BTRFS_I(page->mapping->host)->io_tree;
430 return extent_read_full_page(tree, page, btree_get_extent);
431 }
432
433 static int btree_releasepage(struct page *page, gfp_t gfp_flags)
434 {
435 struct extent_io_tree *tree;
436 struct extent_map_tree *map;
437 int ret;
438
439 if (page_count(page) > 3) {
440 /* once for page->private, once for the caller, once
441 * once for the page cache
442 */
443 return 0;
444 }
445 tree = &BTRFS_I(page->mapping->host)->io_tree;
446 map = &BTRFS_I(page->mapping->host)->extent_tree;
447 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
448 if (ret == 1) {
449 invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE);
450 ClearPagePrivate(page);
451 set_page_private(page, 0);
452 page_cache_release(page);
453 }
454 return ret;
455 }
456
457 static void btree_invalidatepage(struct page *page, unsigned long offset)
458 {
459 struct extent_io_tree *tree;
460 tree = &BTRFS_I(page->mapping->host)->io_tree;
461 extent_invalidatepage(tree, page, offset);
462 btree_releasepage(page, GFP_NOFS);
463 }
464
465 #if 0
466 static int btree_writepage(struct page *page, struct writeback_control *wbc)
467 {
468 struct buffer_head *bh;
469 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
470 struct buffer_head *head;
471 if (!page_has_buffers(page)) {
472 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
473 (1 << BH_Dirty)|(1 << BH_Uptodate));
474 }
475 head = page_buffers(page);
476 bh = head;
477 do {
478 if (buffer_dirty(bh))
479 csum_tree_block(root, bh, 0);
480 bh = bh->b_this_page;
481 } while (bh != head);
482 return block_write_full_page(page, btree_get_block, wbc);
483 }
484 #endif
485
486 static struct address_space_operations btree_aops = {
487 .readpage = btree_readpage,
488 .writepage = btree_writepage,
489 .writepages = btree_writepages,
490 .releasepage = btree_releasepage,
491 .invalidatepage = btree_invalidatepage,
492 .sync_page = block_sync_page,
493 };
494
495 int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
496 {
497 struct extent_buffer *buf = NULL;
498 struct inode *btree_inode = root->fs_info->btree_inode;
499 int ret = 0;
500
501 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
502 if (!buf)
503 return 0;
504 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
505 buf, 0, 0, btree_get_extent, 0);
506 free_extent_buffer(buf);
507 return ret;
508 }
509
510 static int close_all_devices(struct btrfs_fs_info *fs_info)
511 {
512 struct list_head *list;
513 struct list_head *next;
514 struct btrfs_device *device;
515
516 list = &fs_info->fs_devices->devices;
517 list_for_each(next, list) {
518 device = list_entry(next, struct btrfs_device, dev_list);
519 if (device->bdev && device->bdev != fs_info->sb->s_bdev)
520 close_bdev_excl(device->bdev);
521 device->bdev = NULL;
522 }
523 return 0;
524 }
525
526 int btrfs_verify_block_csum(struct btrfs_root *root,
527 struct extent_buffer *buf)
528 {
529 return btrfs_buffer_uptodate(buf);
530 }
531
532 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
533 u64 bytenr, u32 blocksize)
534 {
535 struct inode *btree_inode = root->fs_info->btree_inode;
536 struct extent_buffer *eb;
537 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
538 bytenr, blocksize, GFP_NOFS);
539 return eb;
540 }
541
542 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
543 u64 bytenr, u32 blocksize)
544 {
545 struct inode *btree_inode = root->fs_info->btree_inode;
546 struct extent_buffer *eb;
547
548 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
549 bytenr, blocksize, NULL, GFP_NOFS);
550 return eb;
551 }
552
553
554 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
555 u32 blocksize)
556 {
557 struct extent_buffer *buf = NULL;
558 struct inode *btree_inode = root->fs_info->btree_inode;
559 struct extent_io_tree *io_tree;
560 int ret;
561
562 io_tree = &BTRFS_I(btree_inode)->io_tree;
563
564 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
565 if (!buf)
566 return NULL;
567
568 ret = btree_read_extent_buffer_pages(root, buf, 0);
569
570 if (ret == 0) {
571 buf->flags |= EXTENT_UPTODATE;
572 }
573 return buf;
574
575 }
576
577 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
578 struct extent_buffer *buf)
579 {
580 struct inode *btree_inode = root->fs_info->btree_inode;
581 if (btrfs_header_generation(buf) ==
582 root->fs_info->running_transaction->transid)
583 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
584 buf);
585 return 0;
586 }
587
588 int wait_on_tree_block_writeback(struct btrfs_root *root,
589 struct extent_buffer *buf)
590 {
591 struct inode *btree_inode = root->fs_info->btree_inode;
592 wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree,
593 buf);
594 return 0;
595 }
596
597 static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
598 u32 stripesize, struct btrfs_root *root,
599 struct btrfs_fs_info *fs_info,
600 u64 objectid)
601 {
602 root->node = NULL;
603 root->inode = NULL;
604 root->commit_root = NULL;
605 root->sectorsize = sectorsize;
606 root->nodesize = nodesize;
607 root->leafsize = leafsize;
608 root->stripesize = stripesize;
609 root->ref_cows = 0;
610 root->track_dirty = 0;
611
612 root->fs_info = fs_info;
613 root->objectid = objectid;
614 root->last_trans = 0;
615 root->highest_inode = 0;
616 root->last_inode_alloc = 0;
617 root->name = NULL;
618 root->in_sysfs = 0;
619
620 INIT_LIST_HEAD(&root->dirty_list);
621 memset(&root->root_key, 0, sizeof(root->root_key));
622 memset(&root->root_item, 0, sizeof(root->root_item));
623 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
624 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
625 init_completion(&root->kobj_unregister);
626 root->defrag_running = 0;
627 root->defrag_level = 0;
628 root->root_key.objectid = objectid;
629 return 0;
630 }
631
632 static int find_and_setup_root(struct btrfs_root *tree_root,
633 struct btrfs_fs_info *fs_info,
634 u64 objectid,
635 struct btrfs_root *root)
636 {
637 int ret;
638 u32 blocksize;
639
640 __setup_root(tree_root->nodesize, tree_root->leafsize,
641 tree_root->sectorsize, tree_root->stripesize,
642 root, fs_info, objectid);
643 ret = btrfs_find_last_root(tree_root, objectid,
644 &root->root_item, &root->root_key);
645 BUG_ON(ret);
646
647 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
648 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
649 blocksize);
650 BUG_ON(!root->node);
651 return 0;
652 }
653
654 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
655 struct btrfs_key *location)
656 {
657 struct btrfs_root *root;
658 struct btrfs_root *tree_root = fs_info->tree_root;
659 struct btrfs_path *path;
660 struct extent_buffer *l;
661 u64 highest_inode;
662 u32 blocksize;
663 int ret = 0;
664
665 root = kzalloc(sizeof(*root), GFP_NOFS);
666 if (!root)
667 return ERR_PTR(-ENOMEM);
668 if (location->offset == (u64)-1) {
669 ret = find_and_setup_root(tree_root, fs_info,
670 location->objectid, root);
671 if (ret) {
672 kfree(root);
673 return ERR_PTR(ret);
674 }
675 goto insert;
676 }
677
678 __setup_root(tree_root->nodesize, tree_root->leafsize,
679 tree_root->sectorsize, tree_root->stripesize,
680 root, fs_info, location->objectid);
681
682 path = btrfs_alloc_path();
683 BUG_ON(!path);
684 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
685 if (ret != 0) {
686 if (ret > 0)
687 ret = -ENOENT;
688 goto out;
689 }
690 l = path->nodes[0];
691 read_extent_buffer(l, &root->root_item,
692 btrfs_item_ptr_offset(l, path->slots[0]),
693 sizeof(root->root_item));
694 memcpy(&root->root_key, location, sizeof(*location));
695 ret = 0;
696 out:
697 btrfs_release_path(root, path);
698 btrfs_free_path(path);
699 if (ret) {
700 kfree(root);
701 return ERR_PTR(ret);
702 }
703 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
704 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
705 blocksize);
706 BUG_ON(!root->node);
707 insert:
708 root->ref_cows = 1;
709 ret = btrfs_find_highest_inode(root, &highest_inode);
710 if (ret == 0) {
711 root->highest_inode = highest_inode;
712 root->last_inode_alloc = highest_inode;
713 }
714 return root;
715 }
716
717 struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
718 u64 root_objectid)
719 {
720 struct btrfs_root *root;
721
722 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
723 return fs_info->tree_root;
724 if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
725 return fs_info->extent_root;
726
727 root = radix_tree_lookup(&fs_info->fs_roots_radix,
728 (unsigned long)root_objectid);
729 return root;
730 }
731
732 struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
733 struct btrfs_key *location)
734 {
735 struct btrfs_root *root;
736 int ret;
737
738 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
739 return fs_info->tree_root;
740 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
741 return fs_info->extent_root;
742
743 root = radix_tree_lookup(&fs_info->fs_roots_radix,
744 (unsigned long)location->objectid);
745 if (root)
746 return root;
747
748 root = btrfs_read_fs_root_no_radix(fs_info, location);
749 if (IS_ERR(root))
750 return root;
751 ret = radix_tree_insert(&fs_info->fs_roots_radix,
752 (unsigned long)root->root_key.objectid,
753 root);
754 if (ret) {
755 free_extent_buffer(root->node);
756 kfree(root);
757 return ERR_PTR(ret);
758 }
759 ret = btrfs_find_dead_roots(fs_info->tree_root,
760 root->root_key.objectid, root);
761 BUG_ON(ret);
762
763 return root;
764 }
765
766 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
767 struct btrfs_key *location,
768 const char *name, int namelen)
769 {
770 struct btrfs_root *root;
771 int ret;
772
773 root = btrfs_read_fs_root_no_name(fs_info, location);
774 if (!root)
775 return NULL;
776
777 if (root->in_sysfs)
778 return root;
779
780 ret = btrfs_set_root_name(root, name, namelen);
781 if (ret) {
782 free_extent_buffer(root->node);
783 kfree(root);
784 return ERR_PTR(ret);
785 }
786
787 ret = btrfs_sysfs_add_root(root);
788 if (ret) {
789 free_extent_buffer(root->node);
790 kfree(root->name);
791 kfree(root);
792 return ERR_PTR(ret);
793 }
794 root->in_sysfs = 1;
795 return root;
796 }
797 #if 0
798 static int add_hasher(struct btrfs_fs_info *info, char *type) {
799 struct btrfs_hasher *hasher;
800
801 hasher = kmalloc(sizeof(*hasher), GFP_NOFS);
802 if (!hasher)
803 return -ENOMEM;
804 hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC);
805 if (!hasher->hash_tfm) {
806 kfree(hasher);
807 return -EINVAL;
808 }
809 spin_lock(&info->hash_lock);
810 list_add(&hasher->list, &info->hashers);
811 spin_unlock(&info->hash_lock);
812 return 0;
813 }
814 #endif
815
816 static int btrfs_congested_fn(void *congested_data, int bdi_bits)
817 {
818 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
819 int ret = 0;
820 struct list_head *cur;
821 struct btrfs_device *device;
822 struct backing_dev_info *bdi;
823
824 list_for_each(cur, &info->fs_devices->devices) {
825 device = list_entry(cur, struct btrfs_device, dev_list);
826 bdi = blk_get_backing_dev_info(device->bdev);
827 if (bdi && bdi_congested(bdi, bdi_bits)) {
828 ret = 1;
829 break;
830 }
831 }
832 return ret;
833 }
834
835 void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
836 {
837 struct list_head *cur;
838 struct btrfs_device *device;
839 struct btrfs_fs_info *info;
840
841 info = (struct btrfs_fs_info *)bdi->unplug_io_data;
842 list_for_each(cur, &info->fs_devices->devices) {
843 device = list_entry(cur, struct btrfs_device, dev_list);
844 bdi = blk_get_backing_dev_info(device->bdev);
845 if (bdi->unplug_io_fn) {
846 bdi->unplug_io_fn(bdi, page);
847 }
848 }
849 }
850
851 static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
852 {
853 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
854 bdi_init(bdi);
855 #endif
856 bdi->ra_pages = default_backing_dev_info.ra_pages * 4;
857 bdi->state = 0;
858 bdi->capabilities = default_backing_dev_info.capabilities;
859 bdi->unplug_io_fn = btrfs_unplug_io_fn;
860 bdi->unplug_io_data = info;
861 bdi->congested_fn = btrfs_congested_fn;
862 bdi->congested_data = info;
863 return 0;
864 }
865
866 static int bio_ready_for_csum(struct bio *bio)
867 {
868 u64 length = 0;
869 u64 buf_len = 0;
870 u64 start = 0;
871 struct page *page;
872 struct extent_io_tree *io_tree = NULL;
873 struct btrfs_fs_info *info = NULL;
874 struct bio_vec *bvec;
875 int i;
876 int ret;
877
878 bio_for_each_segment(bvec, bio, i) {
879 page = bvec->bv_page;
880 if (page->private == EXTENT_PAGE_PRIVATE) {
881 length += bvec->bv_len;
882 continue;
883 }
884 if (!page->private) {
885 length += bvec->bv_len;
886 continue;
887 }
888 length = bvec->bv_len;
889 buf_len = page->private >> 2;
890 start = page_offset(page) + bvec->bv_offset;
891 io_tree = &BTRFS_I(page->mapping->host)->io_tree;
892 info = BTRFS_I(page->mapping->host)->root->fs_info;
893 }
894 /* are we fully contained in this bio? */
895 if (buf_len <= length)
896 return 1;
897
898 ret = extent_range_uptodate(io_tree, start + length,
899 start + buf_len - 1);
900 if (ret == 1)
901 return ret;
902 return ret;
903 }
904
905 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
906 void btrfs_end_io_csum(void *p)
907 #else
908 void btrfs_end_io_csum(struct work_struct *work)
909 #endif
910 {
911 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
912 struct btrfs_fs_info *fs_info = p;
913 #else
914 struct btrfs_fs_info *fs_info = container_of(work,
915 struct btrfs_fs_info,
916 end_io_work);
917 #endif
918 unsigned long flags;
919 struct end_io_wq *end_io_wq;
920 struct bio *bio;
921 struct list_head *next;
922 int error;
923 int was_empty;
924
925 while(1) {
926 spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
927 if (list_empty(&fs_info->end_io_work_list)) {
928 spin_unlock_irqrestore(&fs_info->end_io_work_lock,
929 flags);
930 return;
931 }
932 next = fs_info->end_io_work_list.next;
933 list_del(next);
934 spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags);
935
936 end_io_wq = list_entry(next, struct end_io_wq, list);
937
938 bio = end_io_wq->bio;
939 if (end_io_wq->metadata && !bio_ready_for_csum(bio)) {
940 spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
941 was_empty = list_empty(&fs_info->end_io_work_list);
942 list_add_tail(&end_io_wq->list,
943 &fs_info->end_io_work_list);
944 spin_unlock_irqrestore(&fs_info->end_io_work_lock,
945 flags);
946 if (was_empty)
947 return;
948 continue;
949 }
950 error = end_io_wq->error;
951 bio->bi_private = end_io_wq->private;
952 bio->bi_end_io = end_io_wq->end_io;
953 kfree(end_io_wq);
954 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
955 bio_endio(bio, bio->bi_size, error);
956 #else
957 bio_endio(bio, error);
958 #endif
959 }
960 }
961
962 struct btrfs_root *open_ctree(struct super_block *sb,
963 struct btrfs_fs_devices *fs_devices)
964 {
965 u32 sectorsize;
966 u32 nodesize;
967 u32 leafsize;
968 u32 blocksize;
969 u32 stripesize;
970 struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
971 GFP_NOFS);
972 struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
973 GFP_NOFS);
974 struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
975 GFP_NOFS);
976 struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root),
977 GFP_NOFS);
978 struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root),
979 GFP_NOFS);
980 int ret;
981 int err = -EINVAL;
982 struct btrfs_super_block *disk_super;
983
984 if (!extent_root || !tree_root || !fs_info) {
985 err = -ENOMEM;
986 goto fail;
987 }
988 end_io_workqueue = create_workqueue("btrfs-end-io");
989 BUG_ON(!end_io_workqueue);
990
991 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
992 INIT_LIST_HEAD(&fs_info->trans_list);
993 INIT_LIST_HEAD(&fs_info->dead_roots);
994 INIT_LIST_HEAD(&fs_info->hashers);
995 INIT_LIST_HEAD(&fs_info->end_io_work_list);
996 spin_lock_init(&fs_info->hash_lock);
997 spin_lock_init(&fs_info->end_io_work_lock);
998 spin_lock_init(&fs_info->delalloc_lock);
999 spin_lock_init(&fs_info->new_trans_lock);
1000
1001 init_completion(&fs_info->kobj_unregister);
1002 sb_set_blocksize(sb, BTRFS_SUPER_INFO_SIZE);
1003 fs_info->tree_root = tree_root;
1004 fs_info->extent_root = extent_root;
1005 fs_info->chunk_root = chunk_root;
1006 fs_info->dev_root = dev_root;
1007 fs_info->fs_devices = fs_devices;
1008 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
1009 INIT_LIST_HEAD(&fs_info->space_info);
1010 btrfs_mapping_init(&fs_info->mapping_tree);
1011 fs_info->sb = sb;
1012 fs_info->max_extent = (u64)-1;
1013 fs_info->max_inline = 8192 * 1024;
1014 setup_bdi(fs_info, &fs_info->bdi);
1015 fs_info->btree_inode = new_inode(sb);
1016 fs_info->btree_inode->i_ino = 1;
1017 fs_info->btree_inode->i_nlink = 1;
1018 fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
1019 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
1020 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
1021
1022 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
1023 fs_info->btree_inode->i_mapping,
1024 GFP_NOFS);
1025 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
1026 GFP_NOFS);
1027
1028 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
1029
1030 extent_io_tree_init(&fs_info->free_space_cache,
1031 fs_info->btree_inode->i_mapping, GFP_NOFS);
1032 extent_io_tree_init(&fs_info->block_group_cache,
1033 fs_info->btree_inode->i_mapping, GFP_NOFS);
1034 extent_io_tree_init(&fs_info->pinned_extents,
1035 fs_info->btree_inode->i_mapping, GFP_NOFS);
1036 extent_io_tree_init(&fs_info->pending_del,
1037 fs_info->btree_inode->i_mapping, GFP_NOFS);
1038 extent_io_tree_init(&fs_info->extent_ins,
1039 fs_info->btree_inode->i_mapping, GFP_NOFS);
1040 fs_info->do_barriers = 1;
1041
1042 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1043 INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum, fs_info);
1044 INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info);
1045 #else
1046 INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum);
1047 INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
1048 #endif
1049 BTRFS_I(fs_info->btree_inode)->root = tree_root;
1050 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
1051 sizeof(struct btrfs_key));
1052 insert_inode_hash(fs_info->btree_inode);
1053 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1054
1055 mutex_init(&fs_info->trans_mutex);
1056 mutex_init(&fs_info->fs_mutex);
1057
1058 #if 0
1059 ret = add_hasher(fs_info, "crc32c");
1060 if (ret) {
1061 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
1062 err = -ENOMEM;
1063 goto fail_iput;
1064 }
1065 #endif
1066 __setup_root(4096, 4096, 4096, 4096, tree_root,
1067 fs_info, BTRFS_ROOT_TREE_OBJECTID);
1068
1069 fs_info->sb_buffer = read_tree_block(tree_root,
1070 BTRFS_SUPER_INFO_OFFSET,
1071 4096);
1072
1073 if (!fs_info->sb_buffer)
1074 goto fail_iput;
1075
1076 read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0,
1077 sizeof(fs_info->super_copy));
1078
1079 read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
1080 (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
1081 BTRFS_FSID_SIZE);
1082
1083 disk_super = &fs_info->super_copy;
1084 if (!btrfs_super_root(disk_super))
1085 goto fail_sb_buffer;
1086
1087 if (btrfs_super_num_devices(disk_super) != fs_devices->num_devices) {
1088 printk("Btrfs: wanted %llu devices, but found %llu\n",
1089 (unsigned long long)btrfs_super_num_devices(disk_super),
1090 (unsigned long long)fs_devices->num_devices);
1091 goto fail_sb_buffer;
1092 }
1093 nodesize = btrfs_super_nodesize(disk_super);
1094 leafsize = btrfs_super_leafsize(disk_super);
1095 sectorsize = btrfs_super_sectorsize(disk_super);
1096 stripesize = btrfs_super_stripesize(disk_super);
1097 tree_root->nodesize = nodesize;
1098 tree_root->leafsize = leafsize;
1099 tree_root->sectorsize = sectorsize;
1100 tree_root->stripesize = stripesize;
1101 sb_set_blocksize(sb, sectorsize);
1102
1103 i_size_write(fs_info->btree_inode,
1104 btrfs_super_total_bytes(disk_super));
1105
1106 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
1107 sizeof(disk_super->magic))) {
1108 printk("btrfs: valid FS not found on %s\n", sb->s_id);
1109 goto fail_sb_buffer;
1110 }
1111
1112 mutex_lock(&fs_info->fs_mutex);
1113
1114 ret = btrfs_read_sys_array(tree_root);
1115 BUG_ON(ret);
1116
1117 blocksize = btrfs_level_size(tree_root,
1118 btrfs_super_chunk_root_level(disk_super));
1119
1120 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1121 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1122
1123 chunk_root->node = read_tree_block(chunk_root,
1124 btrfs_super_chunk_root(disk_super),
1125 blocksize);
1126 BUG_ON(!chunk_root->node);
1127
1128 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
1129 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
1130 BTRFS_UUID_SIZE);
1131
1132 ret = btrfs_read_chunk_tree(chunk_root);
1133 BUG_ON(ret);
1134
1135 blocksize = btrfs_level_size(tree_root,
1136 btrfs_super_root_level(disk_super));
1137
1138
1139 tree_root->node = read_tree_block(tree_root,
1140 btrfs_super_root(disk_super),
1141 blocksize);
1142 if (!tree_root->node)
1143 goto fail_sb_buffer;
1144
1145
1146 ret = find_and_setup_root(tree_root, fs_info,
1147 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
1148 if (ret)
1149 goto fail_tree_root;
1150 extent_root->track_dirty = 1;
1151
1152 ret = find_and_setup_root(tree_root, fs_info,
1153 BTRFS_DEV_TREE_OBJECTID, dev_root);
1154 dev_root->track_dirty = 1;
1155
1156 if (ret)
1157 goto fail_extent_root;
1158
1159 btrfs_read_block_groups(extent_root);
1160
1161 fs_info->generation = btrfs_super_generation(disk_super) + 1;
1162 fs_info->data_alloc_profile = (u64)-1;
1163 fs_info->metadata_alloc_profile = (u64)-1;
1164 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
1165
1166 mutex_unlock(&fs_info->fs_mutex);
1167 return tree_root;
1168
1169 fail_extent_root:
1170 free_extent_buffer(extent_root->node);
1171 fail_tree_root:
1172 mutex_unlock(&fs_info->fs_mutex);
1173 free_extent_buffer(tree_root->node);
1174 fail_sb_buffer:
1175 free_extent_buffer(fs_info->sb_buffer);
1176 extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
1177 fail_iput:
1178 iput(fs_info->btree_inode);
1179 fail:
1180 close_all_devices(fs_info);
1181 kfree(extent_root);
1182 kfree(tree_root);
1183 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
1184 bdi_destroy(&fs_info->bdi);
1185 #endif
1186 kfree(fs_info);
1187 return ERR_PTR(err);
1188 }
1189
1190 static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
1191 {
1192 char b[BDEVNAME_SIZE];
1193
1194 if (uptodate) {
1195 set_buffer_uptodate(bh);
1196 } else {
1197 if (!buffer_eopnotsupp(bh) && printk_ratelimit()) {
1198 printk(KERN_WARNING "lost page write due to "
1199 "I/O error on %s\n",
1200 bdevname(bh->b_bdev, b));
1201 }
1202 set_buffer_write_io_error(bh);
1203 clear_buffer_uptodate(bh);
1204 }
1205 unlock_buffer(bh);
1206 put_bh(bh);
1207 }
1208
1209 int write_all_supers(struct btrfs_root *root)
1210 {
1211 struct list_head *cur;
1212 struct list_head *head = &root->fs_info->fs_devices->devices;
1213 struct btrfs_device *dev;
1214 struct extent_buffer *sb;
1215 struct btrfs_dev_item *dev_item;
1216 struct buffer_head *bh;
1217 int ret;
1218 int do_barriers;
1219
1220 do_barriers = !btrfs_test_opt(root, NOBARRIER);
1221
1222 sb = root->fs_info->sb_buffer;
1223 dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block,
1224 dev_item);
1225 list_for_each(cur, head) {
1226 dev = list_entry(cur, struct btrfs_device, dev_list);
1227 btrfs_set_device_type(sb, dev_item, dev->type);
1228 btrfs_set_device_id(sb, dev_item, dev->devid);
1229 btrfs_set_device_total_bytes(sb, dev_item, dev->total_bytes);
1230 btrfs_set_device_bytes_used(sb, dev_item, dev->bytes_used);
1231 btrfs_set_device_io_align(sb, dev_item, dev->io_align);
1232 btrfs_set_device_io_width(sb, dev_item, dev->io_width);
1233 btrfs_set_device_sector_size(sb, dev_item, dev->sector_size);
1234 write_extent_buffer(sb, dev->uuid,
1235 (unsigned long)btrfs_device_uuid(dev_item),
1236 BTRFS_UUID_SIZE);
1237
1238 btrfs_set_header_flag(sb, BTRFS_HEADER_FLAG_WRITTEN);
1239 csum_tree_block(root, sb, 0);
1240
1241 bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET /
1242 root->fs_info->sb->s_blocksize,
1243 BTRFS_SUPER_INFO_SIZE);
1244
1245 read_extent_buffer(sb, bh->b_data, 0, BTRFS_SUPER_INFO_SIZE);
1246 dev->pending_io = bh;
1247
1248 get_bh(bh);
1249 set_buffer_uptodate(bh);
1250 lock_buffer(bh);
1251 bh->b_end_io = btrfs_end_buffer_write_sync;
1252
1253 if (do_barriers && dev->barriers) {
1254 ret = submit_bh(WRITE_BARRIER, bh);
1255 if (ret == -EOPNOTSUPP) {
1256 printk("btrfs: disabling barriers on dev %s\n",
1257 dev->name);
1258 set_buffer_uptodate(bh);
1259 dev->barriers = 0;
1260 get_bh(bh);
1261 lock_buffer(bh);
1262 ret = submit_bh(WRITE, bh);
1263 }
1264 } else {
1265 ret = submit_bh(WRITE, bh);
1266 }
1267 BUG_ON(ret);
1268 }
1269
1270 list_for_each(cur, head) {
1271 dev = list_entry(cur, struct btrfs_device, dev_list);
1272 BUG_ON(!dev->pending_io);
1273 bh = dev->pending_io;
1274 wait_on_buffer(bh);
1275 if (!buffer_uptodate(dev->pending_io)) {
1276 if (do_barriers && dev->barriers) {
1277 printk("btrfs: disabling barriers on dev %s\n",
1278 dev->name);
1279 set_buffer_uptodate(bh);
1280 get_bh(bh);
1281 lock_buffer(bh);
1282 dev->barriers = 0;
1283 ret = submit_bh(WRITE, bh);
1284 BUG_ON(ret);
1285 wait_on_buffer(bh);
1286 BUG_ON(!buffer_uptodate(bh));
1287 } else {
1288 BUG();
1289 }
1290
1291 }
1292 dev->pending_io = NULL;
1293 brelse(bh);
1294 }
1295 return 0;
1296 }
1297
1298 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
1299 *root)
1300 {
1301 int ret;
1302
1303 ret = write_all_supers(root);
1304 #if 0
1305 if (!btrfs_test_opt(root, NOBARRIER))
1306 blkdev_issue_flush(sb->s_bdev, NULL);
1307 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, super);
1308 ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping,
1309 super->start, super->len);
1310 if (!btrfs_test_opt(root, NOBARRIER))
1311 blkdev_issue_flush(sb->s_bdev, NULL);
1312 #endif
1313 return ret;
1314 }
1315
1316 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
1317 {
1318 radix_tree_delete(&fs_info->fs_roots_radix,
1319 (unsigned long)root->root_key.objectid);
1320 if (root->in_sysfs)
1321 btrfs_sysfs_del_root(root);
1322 if (root->inode)
1323 iput(root->inode);
1324 if (root->node)
1325 free_extent_buffer(root->node);
1326 if (root->commit_root)
1327 free_extent_buffer(root->commit_root);
1328 if (root->name)
1329 kfree(root->name);
1330 kfree(root);
1331 return 0;
1332 }
1333
1334 static int del_fs_roots(struct btrfs_fs_info *fs_info)
1335 {
1336 int ret;
1337 struct btrfs_root *gang[8];
1338 int i;
1339
1340 while(1) {
1341 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
1342 (void **)gang, 0,
1343 ARRAY_SIZE(gang));
1344 if (!ret)
1345 break;
1346 for (i = 0; i < ret; i++)
1347 btrfs_free_fs_root(fs_info, gang[i]);
1348 }
1349 return 0;
1350 }
1351
1352 int close_ctree(struct btrfs_root *root)
1353 {
1354 int ret;
1355 struct btrfs_trans_handle *trans;
1356 struct btrfs_fs_info *fs_info = root->fs_info;
1357
1358 fs_info->closing = 1;
1359 btrfs_transaction_flush_work(root);
1360 mutex_lock(&fs_info->fs_mutex);
1361 btrfs_defrag_dirty_roots(root->fs_info);
1362 trans = btrfs_start_transaction(root, 1);
1363 ret = btrfs_commit_transaction(trans, root);
1364 /* run commit again to drop the original snapshot */
1365 trans = btrfs_start_transaction(root, 1);
1366 btrfs_commit_transaction(trans, root);
1367 ret = btrfs_write_and_wait_transaction(NULL, root);
1368 BUG_ON(ret);
1369 write_ctree_super(NULL, root);
1370 mutex_unlock(&fs_info->fs_mutex);
1371
1372 if (fs_info->delalloc_bytes) {
1373 printk("btrfs: at unmount delalloc count %Lu\n",
1374 fs_info->delalloc_bytes);
1375 }
1376 if (fs_info->extent_root->node)
1377 free_extent_buffer(fs_info->extent_root->node);
1378
1379 if (fs_info->tree_root->node)
1380 free_extent_buffer(fs_info->tree_root->node);
1381
1382 if (root->fs_info->chunk_root->node);
1383 free_extent_buffer(root->fs_info->chunk_root->node);
1384
1385 if (root->fs_info->dev_root->node);
1386 free_extent_buffer(root->fs_info->dev_root->node);
1387
1388 free_extent_buffer(fs_info->sb_buffer);
1389
1390 btrfs_free_block_groups(root->fs_info);
1391 del_fs_roots(fs_info);
1392
1393 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
1394
1395 extent_io_tree_empty_lru(&fs_info->free_space_cache);
1396 extent_io_tree_empty_lru(&fs_info->block_group_cache);
1397 extent_io_tree_empty_lru(&fs_info->pinned_extents);
1398 extent_io_tree_empty_lru(&fs_info->pending_del);
1399 extent_io_tree_empty_lru(&fs_info->extent_ins);
1400 extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
1401
1402 truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
1403 flush_workqueue(end_io_workqueue);
1404 destroy_workqueue(end_io_workqueue);
1405
1406 iput(fs_info->btree_inode);
1407 #if 0
1408 while(!list_empty(&fs_info->hashers)) {
1409 struct btrfs_hasher *hasher;
1410 hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher,
1411 hashers);
1412 list_del(&hasher->hashers);
1413 crypto_free_hash(&fs_info->hash_tfm);
1414 kfree(hasher);
1415 }
1416 #endif
1417 close_all_devices(fs_info);
1418 btrfs_mapping_tree_free(&fs_info->mapping_tree);
1419
1420 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
1421 bdi_destroy(&fs_info->bdi);
1422 #endif
1423
1424 kfree(fs_info->extent_root);
1425 kfree(fs_info->tree_root);
1426 kfree(fs_info->chunk_root);
1427 kfree(fs_info->dev_root);
1428 return 0;
1429 }
1430
1431 int btrfs_buffer_uptodate(struct extent_buffer *buf)
1432 {
1433 struct inode *btree_inode = buf->first_page->mapping->host;
1434 return extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf);
1435 }
1436
1437 int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
1438 {
1439 struct inode *btree_inode = buf->first_page->mapping->host;
1440 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
1441 buf);
1442 }
1443
1444 void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
1445 {
1446 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1447 u64 transid = btrfs_header_generation(buf);
1448 struct inode *btree_inode = root->fs_info->btree_inode;
1449
1450 if (transid != root->fs_info->generation) {
1451 printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
1452 (unsigned long long)buf->start,
1453 transid, root->fs_info->generation);
1454 WARN_ON(1);
1455 }
1456 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf);
1457 }
1458
1459 void btrfs_throttle(struct btrfs_root *root)
1460 {
1461 struct backing_dev_info *bdi;
1462
1463 bdi = root->fs_info->sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
1464 if (root->fs_info->throttles && bdi_write_congested(bdi)) {
1465 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18)
1466 congestion_wait(WRITE, HZ/20);
1467 #else
1468 blk_congestion_wait(WRITE, HZ/20);
1469 #endif
1470 }
1471 }
1472
1473 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
1474 {
1475 balance_dirty_pages_ratelimited_nr(
1476 root->fs_info->btree_inode->i_mapping, 1);
1477 }
1478
1479 void btrfs_set_buffer_defrag(struct extent_buffer *buf)
1480 {
1481 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1482 struct inode *btree_inode = root->fs_info->btree_inode;
1483 set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
1484 buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS);
1485 }
1486
1487 void btrfs_set_buffer_defrag_done(struct extent_buffer *buf)
1488 {
1489 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1490 struct inode *btree_inode = root->fs_info->btree_inode;
1491 set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
1492 buf->start + buf->len - 1, EXTENT_DEFRAG_DONE,
1493 GFP_NOFS);
1494 }
1495
1496 int btrfs_buffer_defrag(struct extent_buffer *buf)
1497 {
1498 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1499 struct inode *btree_inode = root->fs_info->btree_inode;
1500 return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
1501 buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0);
1502 }
1503
1504 int btrfs_buffer_defrag_done(struct extent_buffer *buf)
1505 {
1506 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1507 struct inode *btree_inode = root->fs_info->btree_inode;
1508 return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
1509 buf->start, buf->start + buf->len - 1,
1510 EXTENT_DEFRAG_DONE, 0);
1511 }
1512
1513 int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf)
1514 {
1515 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1516 struct inode *btree_inode = root->fs_info->btree_inode;
1517 return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
1518 buf->start, buf->start + buf->len - 1,
1519 EXTENT_DEFRAG_DONE, GFP_NOFS);
1520 }
1521
1522 int btrfs_clear_buffer_defrag(struct extent_buffer *buf)
1523 {
1524 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1525 struct inode *btree_inode = root->fs_info->btree_inode;
1526 return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
1527 buf->start, buf->start + buf->len - 1,
1528 EXTENT_DEFRAG, GFP_NOFS);
1529 }
1530
1531 int btrfs_read_buffer(struct extent_buffer *buf)
1532 {
1533 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1534 int ret;
1535 ret = btree_read_extent_buffer_pages(root, buf, 0);
1536 if (ret == 0) {
1537 buf->flags |= EXTENT_UPTODATE;
1538 }
1539 return ret;
1540 }
1541
1542 static struct extent_io_ops btree_extent_io_ops = {
1543 .writepage_io_hook = btree_writepage_io_hook,
1544 .readpage_end_io_hook = btree_readpage_end_io_hook,
1545 .submit_bio_hook = btree_submit_bio_hook,
1546 /* note we're sharing with inode.c for the merge bio hook */
1547 .merge_bio_hook = btrfs_merge_bio_hook,
1548 };
Linux kernel - Deliverables
This page took 0.060567 seconds and 6 git commands to generate.