projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: remove old unused commented out code
[deliverable/linux.git] / fs / btrfs / disk-io.c
CommitLineData
6cbd5570
CM		 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
e20d96d6 19#include <linux/fs.h>
d98237b3 20#include <linux/blkdev.h>
87cbda5c 21#include <linux/scatterlist.h>
22b0ebda 22#include <linux/swap.h>
0f7d52f4 23#include <linux/radix-tree.h>
35b7e476 24#include <linux/writeback.h>
d397712b 25#include <linux/buffer_head.h>
ce9adaa5 26#include <linux/workqueue.h>
a74a4b97 27#include <linux/kthread.h>
4b4e25f2 28#include <linux/freezer.h>
163e783e 29#include <linux/crc32c.h>
5a0e3ad6 30#include <linux/slab.h>
784b4e29 31#include <linux/migrate.h>
7e75bf3f 32#include <asm/unaligned.h>
4b4e25f2 33#include "compat.h"
eb60ceac
CM		 34#include "ctree.h"
35#include "disk-io.h"
e089f05c 36#include "transaction.h"
0f7d52f4 37#include "btrfs_inode.h"
0b86a832 38#include "volumes.h"
db94535d 39#include "print-tree.h"
8b712842 40#include "async-thread.h"
925baedd 41#include "locking.h"
e02119d5 42#include "tree-log.h"
fa9c0d79 43#include "free-space-cache.h"
eb60ceac 44
d1310b2e 45static struct extent_io_ops btree_extent_io_ops;
8b712842 46static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 47static void free_fs_root(struct btrfs_root *root);
acce952b 48static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
49 int read_only);
50static int btrfs_destroy_ordered_operations(struct btrfs_root *root);
51static int btrfs_destroy_ordered_extents(struct btrfs_root *root);
52static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
53 struct btrfs_root *root);
54static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t);
55static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
56static int btrfs_destroy_marked_extents(struct btrfs_root *root,
57 struct extent_io_tree *dirty_pages,
58 int mark);
59static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
60 struct extent_io_tree *pinned_extents);
61static int btrfs_cleanup_transaction(struct btrfs_root *root);
ce9adaa5 62
d352ac68
CM		 63/*
64 * end_io_wq structs are used to do processing in task context when an IO is
65 * complete. This is used during reads to verify checksums, and it is used
66 * by writes to insert metadata for new file extents after IO is complete.
67 */
ce9adaa5
CM		 68struct end_io_wq {
69 struct bio *bio;
70 bio_end_io_t *end_io;
71 void *private;
72 struct btrfs_fs_info *info;
73 int error;
22c59948 74 int metadata;
ce9adaa5 75 struct list_head list;
8b712842 76 struct btrfs_work work;
ce9adaa5 77};
0da5468f 78
d352ac68
CM		 79/*
80 * async submit bios are used to offload expensive checksumming
81 * onto the worker threads. They checksum file and metadata bios
82 * just before they are sent down the IO stack.
83 */
44b8bd7e
CM		 84struct async_submit_bio {
85 struct inode *inode;
86 struct bio *bio;
87 struct list_head list;
4a69a410
CM		 88 extent_submit_bio_hook_t *submit_bio_start;
89 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM		 90 int rw;
91 int mirror_num;
c8b97818 92 unsigned long bio_flags;
eaf25d93
CM		 93 /*
94 * bio_offset is optional, can be used if the pages in the bio
95 * can't tell us where in the file the bio should go
96 */
97 u64 bio_offset;
8b712842 98 struct btrfs_work work;
44b8bd7e
CM		 99};
100
4008c04a
CM		 101/* These are used to set the lockdep class on the extent buffer locks.
102 * The class is set by the readpage_end_io_hook after the buffer has
103 * passed csum validation but before the pages are unlocked.
104 *
105 * The lockdep class is also set by btrfs_init_new_buffer on freshly
106 * allocated blocks.
107 *
108 * The class is based on the level in the tree block, which allows lockdep
109 * to know that lower nodes nest inside the locks of higher nodes.
110 *
111 * We also add a check to make sure the highest level of the tree is
112 * the same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this
113 * code needs update as well.
114 */
115#ifdef CONFIG_DEBUG_LOCK_ALLOC
116# if BTRFS_MAX_LEVEL != 8
117# error
118# endif
119static struct lock_class_key btrfs_eb_class[BTRFS_MAX_LEVEL + 1];
120static const char *btrfs_eb_name[BTRFS_MAX_LEVEL + 1] = {
121 /* leaf */
122 "btrfs-extent-00",
123 "btrfs-extent-01",
124 "btrfs-extent-02",
125 "btrfs-extent-03",
126 "btrfs-extent-04",
127 "btrfs-extent-05",
128 "btrfs-extent-06",
129 "btrfs-extent-07",
130 /* highest possible level */
131 "btrfs-extent-08",
132};
133#endif
134
d352ac68
CM		 135/*
136 * extents on the btree inode are pretty simple, there's one extent
137 * that covers the entire device
138 */
b2950863 139static struct extent_map *btree_get_extent(struct inode *inode,
306e16ce 140 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 141 int create)
7eccb903 142{
5f39d397
CM		 143 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
144 struct extent_map *em;
145 int ret;
146
890871be 147 read_lock(&em_tree->lock);
d1310b2e 148 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 149 if (em) {
150 em->bdev =
151 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 152 read_unlock(&em_tree->lock);
5f39d397 153 goto out;
a061fc8d 154 }
890871be 155 read_unlock(&em_tree->lock);
7b13b7b1 156
172ddd60 157 em = alloc_extent_map();
5f39d397
CM		 158 if (!em) {
159 em = ERR_PTR(-ENOMEM);
160 goto out;
161 }
162 em->start = 0;
0afbaf8c 163 em->len = (u64)-1;
c8b97818 164 em->block_len = (u64)-1;
5f39d397 165 em->block_start = 0;
a061fc8d 166 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 167
890871be 168 write_lock(&em_tree->lock);
5f39d397
CM		 169 ret = add_extent_mapping(em_tree, em);
170 if (ret == -EEXIST) {
0afbaf8c
CM		 171 u64 failed_start = em->start;
172 u64 failed_len = em->len;
173
5f39d397 174 free_extent_map(em);
7b13b7b1 175 em = lookup_extent_mapping(em_tree, start, len);
0afbaf8c 176 if (em) {
7b13b7b1 177 ret = 0;
0afbaf8c
CM		 178 } else {
179 em = lookup_extent_mapping(em_tree, failed_start,
180 failed_len);
7b13b7b1 181 ret = -EIO;
0afbaf8c 182 }
5f39d397 183 } else if (ret) {
7b13b7b1
CM		 184 free_extent_map(em);
185 em = NULL;
5f39d397 186 }
890871be 187 write_unlock(&em_tree->lock);
7b13b7b1
CM		 188
189 if (ret)
190 em = ERR_PTR(ret);
5f39d397
CM		 191out:
192 return em;
7eccb903
CM		 193}
194
19c00ddc
CM		 195u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
196{
163e783e 197 return crc32c(seed, data, len);
19c00ddc
CM		 198}
199
200void btrfs_csum_final(u32 crc, char *result)
201{
7e75bf3f 202 put_unaligned_le32(~crc, result);
19c00ddc
CM		 203}
204
d352ac68
CM		 205/*
206 * compute the csum for a btree block, and either verify it or write it
207 * into the csum field of the block.
208 */
19c00ddc
CM		 209static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
210 int verify)
211{
607d432d
JB		 212 u16 csum_size =
213 btrfs_super_csum_size(&root->fs_info->super_copy);
214 char *result = NULL;
19c00ddc
CM		 215 unsigned long len;
216 unsigned long cur_len;
217 unsigned long offset = BTRFS_CSUM_SIZE;
218 char *map_token = NULL;
219 char *kaddr;
220 unsigned long map_start;
221 unsigned long map_len;
222 int err;
223 u32 crc = ~(u32)0;
607d432d 224 unsigned long inline_result;
19c00ddc
CM		 225
226 len = buf->len - offset;
d397712b 227 while (len > 0) {
19c00ddc
CM		 228 err = map_private_extent_buffer(buf, offset, 32,
229 &map_token, &kaddr,
230 &map_start, &map_len, KM_USER0);
d397712b 231 if (err)
19c00ddc 232 return 1;
19c00ddc
CM		 233 cur_len = min(len, map_len - (offset - map_start));
234 crc = btrfs_csum_data(root, kaddr + offset - map_start,
235 crc, cur_len);
236 len -= cur_len;
237 offset += cur_len;
238 unmap_extent_buffer(buf, map_token, KM_USER0);
239 }
607d432d
JB		 240 if (csum_size > sizeof(inline_result)) {
241 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
242 if (!result)
243 return 1;
244 } else {
245 result = (char *)&inline_result;
246 }
247
19c00ddc
CM		 248 btrfs_csum_final(crc, result);
249
250 if (verify) {
607d432d 251 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 252 u32 val;
253 u32 found = 0;
607d432d 254 memcpy(&found, result, csum_size);
e4204ded 255
607d432d 256 read_extent_buffer(buf, &val, 0, csum_size);
193f284d
CM		 257 if (printk_ratelimit()) {
258 printk(KERN_INFO "btrfs: %s checksum verify "
259 "failed on %llu wanted %X found %X "
260 "level %d\n",
261 root->fs_info->sb->s_id,
262 (unsigned long long)buf->start, val, found,
263 btrfs_header_level(buf));
264 }
607d432d
JB		 265 if (result != (char *)&inline_result)
266 kfree(result);
19c00ddc
CM		 267 return 1;
268 }
269 } else {
607d432d 270 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 271 }
607d432d
JB		 272 if (result != (char *)&inline_result)
273 kfree(result);
19c00ddc
CM		 274 return 0;
275}
276
d352ac68
CM		 277/*
278 * we can't consider a given block up to date unless the transid of the
279 * block matches the transid in the parent node's pointer. This is how we
280 * detect blocks that either didn't get written at all or got written
281 * in the wrong place.
282 */
1259ab75
CM		 283static int verify_parent_transid(struct extent_io_tree *io_tree,
284 struct extent_buffer *eb, u64 parent_transid)
285{
2ac55d41 286 struct extent_state *cached_state = NULL;
1259ab75
CM		 287 int ret;
288
289 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
290 return 0;
291
2ac55d41
JB		 292 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
293 0, &cached_state, GFP_NOFS);
294 if (extent_buffer_uptodate(io_tree, eb, cached_state) &&
1259ab75
CM		 295 btrfs_header_generation(eb) == parent_transid) {
296 ret = 0;
297 goto out;
298 }
193f284d
CM		 299 if (printk_ratelimit()) {
300 printk("parent transid verify failed on %llu wanted %llu "
301 "found %llu\n",
302 (unsigned long long)eb->start,
303 (unsigned long long)parent_transid,
304 (unsigned long long)btrfs_header_generation(eb));
305 }
1259ab75 306 ret = 1;
2ac55d41 307 clear_extent_buffer_uptodate(io_tree, eb, &cached_state);
33958dc6 308out:
2ac55d41
JB		 309 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
310 &cached_state, GFP_NOFS);
1259ab75 311 return ret;
1259ab75
CM		 312}
313
d352ac68
CM		 314/*
315 * helper to read a given tree block, doing retries as required when
316 * the checksums don't match and we have alternate mirrors to try.
317 */
f188591e
CM		 318static int btree_read_extent_buffer_pages(struct btrfs_root *root,
319 struct extent_buffer *eb,
ca7a79ad 320 u64 start, u64 parent_transid)
f188591e
CM		 321{
322 struct extent_io_tree *io_tree;
323 int ret;
324 int num_copies = 0;
325 int mirror_num = 0;
326
a826d6dc 327 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
f188591e
CM		 328 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
329 while (1) {
330 ret = read_extent_buffer_pages(io_tree, eb, start, 1,
331 btree_get_extent, mirror_num);
1259ab75
CM		 332 if (!ret &&
333 !verify_parent_transid(io_tree, eb, parent_transid))
f188591e 334 return ret;
d397712b 335
a826d6dc
JB		 336 /*
337 * This buffer's crc is fine, but its contents are corrupted, so
338 * there is no reason to read the other copies, they won't be
339 * any less wrong.
340 */
341 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
342 return ret;
343
f188591e
CM		 344 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
345 eb->start, eb->len);
4235298e 346 if (num_copies == 1)
f188591e 347 return ret;
4235298e 348
f188591e 349 mirror_num++;
4235298e 350 if (mirror_num > num_copies)
f188591e 351 return ret;
f188591e 352 }
f188591e
CM		 353 return -EIO;
354}
19c00ddc 355
d352ac68 356/*
d397712b
CM		 357 * checksum a dirty tree block before IO. This has extra checks to make sure
358 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 359 */
d397712b 360
b2950863 361static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
19c00ddc 362{
d1310b2e 363 struct extent_io_tree *tree;
35ebb934 364 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
19c00ddc 365 u64 found_start;
19c00ddc
CM		 366 unsigned long len;
367 struct extent_buffer *eb;
f188591e
CM		 368 int ret;
369
d1310b2e 370 tree = &BTRFS_I(page->mapping->host)->io_tree;
19c00ddc 371
eb14ab8e
CM		 372 if (page->private == EXTENT_PAGE_PRIVATE) {
373 WARN_ON(1);
19c00ddc 374 goto out;
eb14ab8e
CM		 375 }
376 if (!page->private) {
377 WARN_ON(1);
19c00ddc 378 goto out;
eb14ab8e 379 }
19c00ddc 380 len = page->private >> 2;
d397712b
CM		 381 WARN_ON(len == 0);
382
ba144192 383 eb = alloc_extent_buffer(tree, start, len, page);
91ca338d
TI		 384 if (eb == NULL) {
385 WARN_ON(1);
386 goto out;
387 }
ca7a79ad
CM		 388 ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
389 btrfs_header_generation(eb));
f188591e 390 BUG_ON(ret);
784b4e29
CM		 391 WARN_ON(!btrfs_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN));
392
19c00ddc
CM		 393 found_start = btrfs_header_bytenr(eb);
394 if (found_start != start) {
55c69072
CM		 395 WARN_ON(1);
396 goto err;
397 }
398 if (eb->first_page != page) {
55c69072
CM		 399 WARN_ON(1);
400 goto err;
401 }
402 if (!PageUptodate(page)) {
55c69072
CM		 403 WARN_ON(1);
404 goto err;
19c00ddc 405 }
19c00ddc 406 csum_tree_block(root, eb, 0);
55c69072 407err:
19c00ddc
CM		 408 free_extent_buffer(eb);
409out:
410 return 0;
411}
412
2b82032c
YZ		 413static int check_tree_block_fsid(struct btrfs_root *root,
414 struct extent_buffer *eb)
415{
416 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
417 u8 fsid[BTRFS_UUID_SIZE];
418 int ret = 1;
419
420 read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
421 BTRFS_FSID_SIZE);
422 while (fs_devices) {
423 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
424 ret = 0;
425 break;
426 }
427 fs_devices = fs_devices->seed;
428 }
429 return ret;
430}
431
a826d6dc
JB		 432#define CORRUPT(reason, eb, root, slot) \
433 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
434 "root=%llu, slot=%d\n", reason, \
435 (unsigned long long)btrfs_header_bytenr(eb), \
436 (unsigned long long)root->objectid, slot)
437
438static noinline int check_leaf(struct btrfs_root *root,
439 struct extent_buffer *leaf)
440{
441 struct btrfs_key key;
442 struct btrfs_key leaf_key;
443 u32 nritems = btrfs_header_nritems(leaf);
444 int slot;
445
446 if (nritems == 0)
447 return 0;
448
449 /* Check the 0 item */
450 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
451 BTRFS_LEAF_DATA_SIZE(root)) {
452 CORRUPT("invalid item offset size pair", leaf, root, 0);
453 return -EIO;
454 }
455
456 /*
457 * Check to make sure each items keys are in the correct order and their
458 * offsets make sense. We only have to loop through nritems-1 because
459 * we check the current slot against the next slot, which verifies the
460 * next slot's offset+size makes sense and that the current's slot
461 * offset is correct.
462 */
463 for (slot = 0; slot < nritems - 1; slot++) {
464 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
465 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
466
467 /* Make sure the keys are in the right order */
468 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
469 CORRUPT("bad key order", leaf, root, slot);
470 return -EIO;
471 }
472
473 /*
474 * Make sure the offset and ends are right, remember that the
475 * item data starts at the end of the leaf and grows towards the
476 * front.
477 */
478 if (btrfs_item_offset_nr(leaf, slot) !=
479 btrfs_item_end_nr(leaf, slot + 1)) {
480 CORRUPT("slot offset bad", leaf, root, slot);
481 return -EIO;
482 }
483
484 /*
485 * Check to make sure that we don't point outside of the leaf,
486 * just incase all the items are consistent to eachother, but
487 * all point outside of the leaf.
488 */
489 if (btrfs_item_end_nr(leaf, slot) >
490 BTRFS_LEAF_DATA_SIZE(root)) {
491 CORRUPT("slot end outside of leaf", leaf, root, slot);
492 return -EIO;
493 }
494 }
495
496 return 0;
497}
498
4008c04a
CM		 499#ifdef CONFIG_DEBUG_LOCK_ALLOC
500void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level)
501{
502 lockdep_set_class_and_name(&eb->lock,
503 &btrfs_eb_class[level],
504 btrfs_eb_name[level]);
505}
506#endif
507
b2950863 508static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
ce9adaa5
CM		 509 struct extent_state *state)
510{
511 struct extent_io_tree *tree;
512 u64 found_start;
513 int found_level;
514 unsigned long len;
515 struct extent_buffer *eb;
516 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 517 int ret = 0;
ce9adaa5
CM		 518
519 tree = &BTRFS_I(page->mapping->host)->io_tree;
520 if (page->private == EXTENT_PAGE_PRIVATE)
521 goto out;
522 if (!page->private)
523 goto out;
d397712b 524
ce9adaa5 525 len = page->private >> 2;
d397712b
CM		 526 WARN_ON(len == 0);
527
ba144192 528 eb = alloc_extent_buffer(tree, start, len, page);
91ca338d
TI		 529 if (eb == NULL) {
530 ret = -EIO;
531 goto out;
532 }
f188591e 533
ce9adaa5 534 found_start = btrfs_header_bytenr(eb);
23a07867 535 if (found_start != start) {
193f284d
CM		 536 if (printk_ratelimit()) {
537 printk(KERN_INFO "btrfs bad tree block start "
538 "%llu %llu\n",
539 (unsigned long long)found_start,
540 (unsigned long long)eb->start);
541 }
f188591e 542 ret = -EIO;
ce9adaa5
CM		 543 goto err;
544 }
545 if (eb->first_page != page) {
d397712b
CM		 546 printk(KERN_INFO "btrfs bad first page %lu %lu\n",
547 eb->first_page->index, page->index);
ce9adaa5 548 WARN_ON(1);
f188591e 549 ret = -EIO;
ce9adaa5
CM		 550 goto err;
551 }
2b82032c 552 if (check_tree_block_fsid(root, eb)) {
193f284d
CM		 553 if (printk_ratelimit()) {
554 printk(KERN_INFO "btrfs bad fsid on block %llu\n",
555 (unsigned long long)eb->start);
556 }
1259ab75
CM		 557 ret = -EIO;
558 goto err;
559 }
ce9adaa5
CM		 560 found_level = btrfs_header_level(eb);
561
4008c04a
CM		 562 btrfs_set_buffer_lockdep_class(eb, found_level);
563
ce9adaa5 564 ret = csum_tree_block(root, eb, 1);
a826d6dc 565 if (ret) {
f188591e 566 ret = -EIO;
a826d6dc
JB		 567 goto err;
568 }
569
570 /*
571 * If this is a leaf block and it is corrupt, set the corrupt bit so
572 * that we don't try and read the other copies of this block, just
573 * return -EIO.
574 */
575 if (found_level == 0 && check_leaf(root, eb)) {
576 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
577 ret = -EIO;
578 }
ce9adaa5
CM		 579
580 end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
581 end = eb->start + end - 1;
ce9adaa5
CM		 582err:
583 free_extent_buffer(eb);
584out:
f188591e 585 return ret;
ce9adaa5
CM		 586}
587
ce9adaa5 588static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM		 589{
590 struct end_io_wq *end_io_wq = bio->bi_private;
591 struct btrfs_fs_info *fs_info;
ce9adaa5 592
ce9adaa5 593 fs_info = end_io_wq->info;
ce9adaa5 594 end_io_wq->error = err;
8b712842
CM		 595 end_io_wq->work.func = end_workqueue_fn;
596 end_io_wq->work.flags = 0;
d20f7043 597
7b6d91da 598 if (bio->bi_rw & REQ_WRITE) {
0cb59c99 599 if (end_io_wq->metadata == 1)
cad321ad
CM		 600 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
601 &end_io_wq->work);
0cb59c99
JB		 602 else if (end_io_wq->metadata == 2)
603 btrfs_queue_worker(&fs_info->endio_freespace_worker,
604 &end_io_wq->work);
cad321ad
CM		 605 else
606 btrfs_queue_worker(&fs_info->endio_write_workers,
607 &end_io_wq->work);
d20f7043
CM		 608 } else {
609 if (end_io_wq->metadata)
610 btrfs_queue_worker(&fs_info->endio_meta_workers,
611 &end_io_wq->work);
612 else
613 btrfs_queue_worker(&fs_info->endio_workers,
614 &end_io_wq->work);
615 }
ce9adaa5
CM		 616}
617
0cb59c99
JB		 618/*
619 * For the metadata arg you want
620 *
621 * 0 - if data
622 * 1 - if normal metadta
623 * 2 - if writing to the free space cache area
624 */
22c59948
CM		 625int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
626 int metadata)
0b86a832 627{
ce9adaa5 628 struct end_io_wq *end_io_wq;
ce9adaa5
CM		 629 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
630 if (!end_io_wq)
631 return -ENOMEM;
632
633 end_io_wq->private = bio->bi_private;
634 end_io_wq->end_io = bio->bi_end_io;
22c59948 635 end_io_wq->info = info;
ce9adaa5
CM		 636 end_io_wq->error = 0;
637 end_io_wq->bio = bio;
22c59948 638 end_io_wq->metadata = metadata;
ce9adaa5
CM		 639
640 bio->bi_private = end_io_wq;
641 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 642 return 0;
643}
644
b64a2851 645unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 646{
4854ddd0
CM		 647 unsigned long limit = min_t(unsigned long,
648 info->workers.max_workers,
649 info->fs_devices->open_devices);
650 return 256 * limit;
651}
0986fe9e 652
4a69a410
CM		 653static void run_one_async_start(struct btrfs_work *work)
654{
4a69a410
CM		 655 struct async_submit_bio *async;
656
657 async = container_of(work, struct async_submit_bio, work);
4a69a410 658 async->submit_bio_start(async->inode, async->rw, async->bio,
eaf25d93
CM		 659 async->mirror_num, async->bio_flags,
660 async->bio_offset);
4a69a410
CM		 661}
662
663static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 664{
665 struct btrfs_fs_info *fs_info;
666 struct async_submit_bio *async;
4854ddd0 667 int limit;
8b712842
CM		 668
669 async = container_of(work, struct async_submit_bio, work);
670 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 671
b64a2851 672 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 673 limit = limit * 2 / 3;
674
8b712842 675 atomic_dec(&fs_info->nr_async_submits);
0986fe9e 676
b64a2851
CM		 677 if (atomic_read(&fs_info->nr_async_submits) < limit &&
678 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 679 wake_up(&fs_info->async_submit_wait);
680
4a69a410 681 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 682 async->mirror_num, async->bio_flags,
683 async->bio_offset);
4a69a410
CM		 684}
685
686static void run_one_async_free(struct btrfs_work *work)
687{
688 struct async_submit_bio *async;
689
690 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 691 kfree(async);
692}
693
44b8bd7e
CM		 694int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
695 int rw, struct bio *bio, int mirror_num,
c8b97818 696 unsigned long bio_flags,
eaf25d93 697 u64 bio_offset,
4a69a410
CM		 698 extent_submit_bio_hook_t *submit_bio_start,
699 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 700{
701 struct async_submit_bio *async;
702
703 async = kmalloc(sizeof(*async), GFP_NOFS);
704 if (!async)
705 return -ENOMEM;
706
707 async->inode = inode;
708 async->rw = rw;
709 async->bio = bio;
710 async->mirror_num = mirror_num;
4a69a410
CM		 711 async->submit_bio_start = submit_bio_start;
712 async->submit_bio_done = submit_bio_done;
713
714 async->work.func = run_one_async_start;
715 async->work.ordered_func = run_one_async_done;
716 async->work.ordered_free = run_one_async_free;
717
8b712842 718 async->work.flags = 0;
c8b97818 719 async->bio_flags = bio_flags;
eaf25d93 720 async->bio_offset = bio_offset;
8c8bee1d 721
cb03c743 722 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 723
7b6d91da 724 if (rw & REQ_SYNC)
d313d7a3
CM		 725 btrfs_set_work_high_prio(&async->work);
726
8b712842 727 btrfs_queue_worker(&fs_info->workers, &async->work);
9473f16c 728
d397712b 729 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 730 atomic_read(&fs_info->nr_async_submits)) {
731 wait_event(fs_info->async_submit_wait,
732 (atomic_read(&fs_info->nr_async_submits) == 0));
733 }
734
44b8bd7e
CM		 735 return 0;
736}
737
ce3ed71a
CM		 738static int btree_csum_one_bio(struct bio *bio)
739{
740 struct bio_vec *bvec = bio->bi_io_vec;
741 int bio_index = 0;
742 struct btrfs_root *root;
743
744 WARN_ON(bio->bi_vcnt <= 0);
d397712b 745 while (bio_index < bio->bi_vcnt) {
ce3ed71a
CM		 746 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
747 csum_dirty_buffer(root, bvec->bv_page);
748 bio_index++;
749 bvec++;
750 }
751 return 0;
752}
753
4a69a410
CM		 754static int __btree_submit_bio_start(struct inode *inode, int rw,
755 struct bio *bio, int mirror_num,
eaf25d93
CM		 756 unsigned long bio_flags,
757 u64 bio_offset)
22c59948 758{
8b712842
CM		 759 /*
760 * when we're called for a write, we're already in the async
5443be45 761 * submission context. Just jump into btrfs_map_bio
8b712842 762 */
4a69a410
CM		 763 btree_csum_one_bio(bio);
764 return 0;
765}
22c59948 766
4a69a410 767static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 768 int mirror_num, unsigned long bio_flags,
769 u64 bio_offset)
4a69a410 770{
8b712842 771 /*
4a69a410
CM		 772 * when we're called for a write, we're already in the async
773 * submission context. Just jump into btrfs_map_bio
8b712842 774 */
8b712842 775 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
0b86a832
CM		 776}
777
44b8bd7e 778static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 779 int mirror_num, unsigned long bio_flags,
780 u64 bio_offset)
44b8bd7e 781{
cad321ad
CM		 782 int ret;
783
784 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
785 bio, 1);
786 BUG_ON(ret);
787
7b6d91da 788 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 789 /*
790 * called for a read, do the setup so that checksum validation
791 * can happen in the async kernel threads
792 */
4a69a410 793 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
6f3577bd 794 mirror_num, 0);
44b8bd7e 795 }
d313d7a3 796
cad321ad
CM		 797 /*
798 * kthread helpers are used to submit writes so that checksumming
799 * can happen in parallel across all CPUs
800 */
44b8bd7e 801 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
c8b97818 802 inode, rw, bio, mirror_num, 0,
eaf25d93 803 bio_offset,
4a69a410
CM		 804 __btree_submit_bio_start,
805 __btree_submit_bio_done);
44b8bd7e
CM		 806}
807
3dd1462e 808#ifdef CONFIG_MIGRATION
784b4e29
CM		 809static int btree_migratepage(struct address_space *mapping,
810 struct page *newpage, struct page *page)
811{
812 /*
813 * we can't safely write a btree page from here,
814 * we haven't done the locking hook
815 */
816 if (PageDirty(page))
817 return -EAGAIN;
818 /*
819 * Buffers may be managed in a filesystem specific way.
820 * We must have no buffers or drop them.
821 */
822 if (page_has_private(page) &&
823 !try_to_release_page(page, GFP_KERNEL))
824 return -EAGAIN;
784b4e29
CM		 825 return migrate_page(mapping, newpage, page);
826}
3dd1462e 827#endif
784b4e29 828
0da5468f
CM		 829static int btree_writepage(struct page *page, struct writeback_control *wbc)
830{
d1310b2e 831 struct extent_io_tree *tree;
b9473439
CM		 832 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
833 struct extent_buffer *eb;
834 int was_dirty;
835
d1310b2e 836 tree = &BTRFS_I(page->mapping->host)->io_tree;
b9473439
CM		 837 if (!(current->flags & PF_MEMALLOC)) {
838 return extent_write_full_page(tree, page,
839 btree_get_extent, wbc);
840 }
5443be45 841
b9473439 842 redirty_page_for_writepage(wbc, page);
784b4e29 843 eb = btrfs_find_tree_block(root, page_offset(page), PAGE_CACHE_SIZE);
b9473439
CM		 844 WARN_ON(!eb);
845
846 was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
847 if (!was_dirty) {
848 spin_lock(&root->fs_info->delalloc_lock);
849 root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE;
850 spin_unlock(&root->fs_info->delalloc_lock);
5443be45 851 }
b9473439
CM		 852 free_extent_buffer(eb);
853
854 unlock_page(page);
855 return 0;
5f39d397 856}
0da5468f
CM		 857
858static int btree_writepages(struct address_space *mapping,
859 struct writeback_control *wbc)
860{
d1310b2e
CM		 861 struct extent_io_tree *tree;
862 tree = &BTRFS_I(mapping->host)->io_tree;
d8d5f3e1 863 if (wbc->sync_mode == WB_SYNC_NONE) {
b9473439 864 struct btrfs_root *root = BTRFS_I(mapping->host)->root;
793955bc 865 u64 num_dirty;
24ab9cd8 866 unsigned long thresh = 32 * 1024 * 1024;
448d640b
CM		 867
868 if (wbc->for_kupdate)
869 return 0;
870
b9473439
CM		 871 /* this is a bit racy, but that's ok */
872 num_dirty = root->fs_info->dirty_metadata_bytes;
d397712b 873 if (num_dirty < thresh)
793955bc 874 return 0;
793955bc 875 }
0da5468f
CM		 876 return extent_writepages(tree, mapping, btree_get_extent, wbc);
877}
878
b2950863 879static int btree_readpage(struct file *file, struct page *page)
5f39d397 880{
d1310b2e
CM		 881 struct extent_io_tree *tree;
882 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 883 return extent_read_full_page(tree, page, btree_get_extent);
884}
22b0ebda 885
70dec807 886static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 887{
d1310b2e
CM		 888 struct extent_io_tree *tree;
889 struct extent_map_tree *map;
5f39d397 890 int ret;
d98237b3 891
98509cfc 892 if (PageWriteback(page) || PageDirty(page))
d397712b 893 return 0;
98509cfc 894
d1310b2e
CM		 895 tree = &BTRFS_I(page->mapping->host)->io_tree;
896 map = &BTRFS_I(page->mapping->host)->extent_tree;
6af118ce 897
7b13b7b1 898 ret = try_release_extent_state(map, tree, page, gfp_flags);
d397712b 899 if (!ret)
6af118ce 900 return 0;
6af118ce
CM		 901
902 ret = try_release_extent_buffer(tree, page);
5f39d397
CM		 903 if (ret == 1) {
904 ClearPagePrivate(page);
905 set_page_private(page, 0);
906 page_cache_release(page);
907 }
6af118ce 908
d98237b3
CM		 909 return ret;
910}
911
5f39d397 912static void btree_invalidatepage(struct page *page, unsigned long offset)
d98237b3 913{
d1310b2e
CM		 914 struct extent_io_tree *tree;
915 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 916 extent_invalidatepage(tree, page, offset);
917 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 918 if (PagePrivate(page)) {
d397712b
CM		 919 printk(KERN_WARNING "btrfs warning page private not zero "
920 "on page %llu\n", (unsigned long long)page_offset(page));
9ad6b7bc
CM		 921 ClearPagePrivate(page);
922 set_page_private(page, 0);
923 page_cache_release(page);
924 }
d98237b3
CM		 925}
926
7f09410b 927static const struct address_space_operations btree_aops = {
d98237b3
CM		 928 .readpage = btree_readpage,
929 .writepage = btree_writepage,
0da5468f 930 .writepages = btree_writepages,
5f39d397
CM		 931 .releasepage = btree_releasepage,
932 .invalidatepage = btree_invalidatepage,
5a92bc88 933#ifdef CONFIG_MIGRATION
784b4e29 934 .migratepage = btree_migratepage,
5a92bc88 935#endif
d98237b3
CM		 936};
937
ca7a79ad
CM		 938int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
939 u64 parent_transid)
090d1875 940{
5f39d397
CM		 941 struct extent_buffer *buf = NULL;
942 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 943 int ret = 0;
090d1875 944
db94535d 945 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 946 if (!buf)
090d1875 947 return 0;
d1310b2e 948 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
f188591e 949 buf, 0, 0, btree_get_extent, 0);
5f39d397 950 free_extent_buffer(buf);
de428b63 951 return ret;
090d1875
CM		 952}
953
0999df54
CM		 954struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
955 u64 bytenr, u32 blocksize)
956{
957 struct inode *btree_inode = root->fs_info->btree_inode;
958 struct extent_buffer *eb;
959 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
f09d1f60 960 bytenr, blocksize);
0999df54
CM		 961 return eb;
962}
963
964struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
965 u64 bytenr, u32 blocksize)
966{
967 struct inode *btree_inode = root->fs_info->btree_inode;
968 struct extent_buffer *eb;
969
970 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
ba144192 971 bytenr, blocksize, NULL);
0999df54
CM		 972 return eb;
973}
974
975
e02119d5
CM		 976int btrfs_write_tree_block(struct extent_buffer *buf)
977{
8aa38c31
CH		 978 return filemap_fdatawrite_range(buf->first_page->mapping, buf->start,
979 buf->start + buf->len - 1);
e02119d5
CM		 980}
981
982int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
983{
8aa38c31
CH		 984 return filemap_fdatawait_range(buf->first_page->mapping,
985 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 986}
987
0999df54 988struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 989 u32 blocksize, u64 parent_transid)
0999df54
CM		 990{
991 struct extent_buffer *buf = NULL;
0999df54
CM		 992 int ret;
993
0999df54
CM		 994 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
995 if (!buf)
996 return NULL;
0999df54 997
ca7a79ad 998 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
ce9adaa5 999
d397712b 1000 if (ret == 0)
b4ce94de 1001 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
5f39d397 1002 return buf;
ce9adaa5 1003
eb60ceac
CM		 1004}
1005
e089f05c 1006int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
5f39d397 1007 struct extent_buffer *buf)
ed2ff2cb 1008{
5f39d397 1009 struct inode *btree_inode = root->fs_info->btree_inode;
55c69072 1010 if (btrfs_header_generation(buf) ==
925baedd 1011 root->fs_info->running_transaction->transid) {
b9447ef8 1012 btrfs_assert_tree_locked(buf);
b4ce94de 1013
b9473439
CM		 1014 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
1015 spin_lock(&root->fs_info->delalloc_lock);
1016 if (root->fs_info->dirty_metadata_bytes >= buf->len)
1017 root->fs_info->dirty_metadata_bytes -= buf->len;
1018 else
1019 WARN_ON(1);
1020 spin_unlock(&root->fs_info->delalloc_lock);
1021 }
b4ce94de 1022
b9473439
CM		 1023 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1024 btrfs_set_lock_blocking(buf);
d1310b2e 1025 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
55c69072 1026 buf);
925baedd 1027 }
5f39d397
CM		 1028 return 0;
1029}
1030
db94535d 1031static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
87ee04eb 1032 u32 stripesize, struct btrfs_root *root,
9f5fae2f 1033 struct btrfs_fs_info *fs_info,
e20d96d6 1034 u64 objectid)
d97e63b6 1035{
cfaa7295 1036 root->node = NULL;
a28ec197 1037 root->commit_root = NULL;
db94535d
CM		 1038 root->sectorsize = sectorsize;
1039 root->nodesize = nodesize;
1040 root->leafsize = leafsize;
87ee04eb 1041 root->stripesize = stripesize;
123abc88 1042 root->ref_cows = 0;
0b86a832 1043 root->track_dirty = 0;
c71bf099 1044 root->in_radix = 0;
d68fc57b
YZ		 1045 root->orphan_item_inserted = 0;
1046 root->orphan_cleanup_state = 0;
0b86a832 1047
9f5fae2f 1048 root->fs_info = fs_info;
0f7d52f4
CM		 1049 root->objectid = objectid;
1050 root->last_trans = 0;
13a8a7c8 1051 root->highest_objectid = 0;
58176a96 1052 root->name = NULL;
4313b399 1053 root->in_sysfs = 0;
6bef4d31 1054 root->inode_tree = RB_ROOT;
f0486c68 1055 root->block_rsv = NULL;
d68fc57b 1056 root->orphan_block_rsv = NULL;
0b86a832
CM		 1057
1058 INIT_LIST_HEAD(&root->dirty_list);
7b128766 1059 INIT_LIST_HEAD(&root->orphan_list);
5d4f98a2 1060 INIT_LIST_HEAD(&root->root_list);
925baedd 1061 spin_lock_init(&root->node_lock);
d68fc57b 1062 spin_lock_init(&root->orphan_lock);
5d4f98a2 1063 spin_lock_init(&root->inode_lock);
f0486c68 1064 spin_lock_init(&root->accounting_lock);
a2135011 1065 mutex_init(&root->objectid_mutex);
e02119d5 1066 mutex_init(&root->log_mutex);
7237f183
YZ		 1067 init_waitqueue_head(&root->log_writer_wait);
1068 init_waitqueue_head(&root->log_commit_wait[0]);
1069 init_waitqueue_head(&root->log_commit_wait[1]);
1070 atomic_set(&root->log_commit[0], 0);
1071 atomic_set(&root->log_commit[1], 0);
1072 atomic_set(&root->log_writers, 0);
1073 root->log_batch = 0;
1074 root->log_transid = 0;
257c62e1 1075 root->last_log_commit = 0;
d0c803c4 1076 extent_io_tree_init(&root->dirty_log_pages,
f993c883 1077 fs_info->btree_inode->i_mapping);
017e5369 1078
3768f368
CM		 1079 memset(&root->root_key, 0, sizeof(root->root_key));
1080 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1081 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1082 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
3f157a2f 1083 root->defrag_trans_start = fs_info->generation;
58176a96 1084 init_completion(&root->kobj_unregister);
6702ed49 1085 root->defrag_running = 0;
4d775673 1086 root->root_key.objectid = objectid;
3394e160
CM		 1087 root->anon_super.s_root = NULL;
1088 root->anon_super.s_dev = 0;
1089 INIT_LIST_HEAD(&root->anon_super.s_list);
1090 INIT_LIST_HEAD(&root->anon_super.s_instances);
1091 init_rwsem(&root->anon_super.s_umount);
1092
3768f368
CM		 1093 return 0;
1094}
1095
db94535d 1096static int find_and_setup_root(struct btrfs_root *tree_root,
9f5fae2f
CM		 1097 struct btrfs_fs_info *fs_info,
1098 u64 objectid,
e20d96d6 1099 struct btrfs_root *root)
3768f368
CM		 1100{
1101 int ret;
db94535d 1102 u32 blocksize;
84234f3a 1103 u64 generation;
3768f368 1104
db94535d 1105 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 1106 tree_root->sectorsize, tree_root->stripesize,
1107 root, fs_info, objectid);
3768f368
CM		 1108 ret = btrfs_find_last_root(tree_root, objectid,
1109 &root->root_item, &root->root_key);
4df27c4d
YZ		 1110 if (ret > 0)
1111 return -ENOENT;
3768f368
CM		 1112 BUG_ON(ret);
1113
84234f3a 1114 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1115 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1116 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1117 blocksize, generation);
68433b73
CM		 1118 if (!root->node || !btrfs_buffer_uptodate(root->node, generation)) {
1119 free_extent_buffer(root->node);
1120 return -EIO;
1121 }
4df27c4d 1122 root->commit_root = btrfs_root_node(root);
d97e63b6
CM		 1123 return 0;
1124}
1125
7237f183
YZ		 1126static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1127 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1128{
1129 struct btrfs_root *root;
1130 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1131 struct extent_buffer *leaf;
e02119d5
CM		 1132
1133 root = kzalloc(sizeof(*root), GFP_NOFS);
1134 if (!root)
7237f183 1135 return ERR_PTR(-ENOMEM);
e02119d5
CM		 1136
1137 __setup_root(tree_root->nodesize, tree_root->leafsize,
1138 tree_root->sectorsize, tree_root->stripesize,
1139 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1140
1141 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1142 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1143 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
7237f183
YZ		 1144 /*
1145 * log trees do not get reference counted because they go away
1146 * before a real commit is actually done. They do store pointers
1147 * to file data extents, and those reference counts still get
1148 * updated (along with back refs to the log tree).
1149 */
e02119d5
CM		 1150 root->ref_cows = 0;
1151
5d4f98a2
YZ		 1152 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
1153 BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0);
7237f183
YZ		 1154 if (IS_ERR(leaf)) {
1155 kfree(root);
1156 return ERR_CAST(leaf);
1157 }
e02119d5 1158
5d4f98a2
YZ		 1159 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1160 btrfs_set_header_bytenr(leaf, leaf->start);
1161 btrfs_set_header_generation(leaf, trans->transid);
1162 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1163 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1164 root->node = leaf;
e02119d5
CM		 1165
1166 write_extent_buffer(root->node, root->fs_info->fsid,
1167 (unsigned long)btrfs_header_fsid(root->node),
1168 BTRFS_FSID_SIZE);
1169 btrfs_mark_buffer_dirty(root->node);
1170 btrfs_tree_unlock(root->node);
7237f183
YZ		 1171 return root;
1172}
1173
1174int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1175 struct btrfs_fs_info *fs_info)
1176{
1177 struct btrfs_root *log_root;
1178
1179 log_root = alloc_log_tree(trans, fs_info);
1180 if (IS_ERR(log_root))
1181 return PTR_ERR(log_root);
1182 WARN_ON(fs_info->log_root_tree);
1183 fs_info->log_root_tree = log_root;
1184 return 0;
1185}
1186
1187int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1188 struct btrfs_root *root)
1189{
1190 struct btrfs_root *log_root;
1191 struct btrfs_inode_item *inode_item;
1192
1193 log_root = alloc_log_tree(trans, root->fs_info);
1194 if (IS_ERR(log_root))
1195 return PTR_ERR(log_root);
1196
1197 log_root->last_trans = trans->transid;
1198 log_root->root_key.offset = root->root_key.objectid;
1199
1200 inode_item = &log_root->root_item.inode;
1201 inode_item->generation = cpu_to_le64(1);
1202 inode_item->size = cpu_to_le64(3);
1203 inode_item->nlink = cpu_to_le32(1);
1204 inode_item->nbytes = cpu_to_le64(root->leafsize);
1205 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1206
5d4f98a2 1207 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1208
1209 WARN_ON(root->log_root);
1210 root->log_root = log_root;
1211 root->log_transid = 0;
257c62e1 1212 root->last_log_commit = 0;
e02119d5
CM		 1213 return 0;
1214}
1215
1216struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
1217 struct btrfs_key *location)
1218{
1219 struct btrfs_root *root;
1220 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1221 struct btrfs_path *path;
5f39d397 1222 struct extent_buffer *l;
84234f3a 1223 u64 generation;
db94535d 1224 u32 blocksize;
0f7d52f4
CM		 1225 int ret = 0;
1226
5eda7b5e 1227 root = kzalloc(sizeof(*root), GFP_NOFS);
0cf6c620 1228 if (!root)
0f7d52f4 1229 return ERR_PTR(-ENOMEM);
0f7d52f4 1230 if (location->offset == (u64)-1) {
db94535d 1231 ret = find_and_setup_root(tree_root, fs_info,
0f7d52f4
CM		 1232 location->objectid, root);
1233 if (ret) {
0f7d52f4
CM		 1234 kfree(root);
1235 return ERR_PTR(ret);
1236 }
13a8a7c8 1237 goto out;
0f7d52f4
CM		 1238 }
1239
db94535d 1240 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 1241 tree_root->sectorsize, tree_root->stripesize,
1242 root, fs_info, location->objectid);
0f7d52f4
CM		 1243
1244 path = btrfs_alloc_path();
db5b493a
TI		 1245 if (!path) {
1246 kfree(root);
1247 return ERR_PTR(-ENOMEM);
1248 }
0f7d52f4 1249 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
13a8a7c8
YZ		 1250 if (ret == 0) {
1251 l = path->nodes[0];
1252 read_extent_buffer(l, &root->root_item,
1253 btrfs_item_ptr_offset(l, path->slots[0]),
1254 sizeof(root->root_item));
1255 memcpy(&root->root_key, location, sizeof(*location));
0f7d52f4 1256 }
0f7d52f4
CM		 1257 btrfs_free_path(path);
1258 if (ret) {
5e540f77 1259 kfree(root);
13a8a7c8
YZ		 1260 if (ret > 0)
1261 ret = -ENOENT;
0f7d52f4
CM		 1262 return ERR_PTR(ret);
1263 }
13a8a7c8 1264
84234f3a 1265 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1266 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1267 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1268 blocksize, generation);
5d4f98a2 1269 root->commit_root = btrfs_root_node(root);
0f7d52f4 1270 BUG_ON(!root->node);
13a8a7c8 1271out:
08fe4db1 1272 if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
e02119d5 1273 root->ref_cows = 1;
08fe4db1
LZ		 1274 btrfs_check_and_init_root_item(&root->root_item);
1275 }
13a8a7c8 1276
5eda7b5e
CM		 1277 return root;
1278}
1279
edbd8d4e
CM		 1280struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
1281 struct btrfs_key *location)
5eda7b5e
CM		 1282{
1283 struct btrfs_root *root;
1284 int ret;
1285
edbd8d4e
CM		 1286 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1287 return fs_info->tree_root;
1288 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1289 return fs_info->extent_root;
8f18cf13
CM		 1290 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1291 return fs_info->chunk_root;
1292 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1293 return fs_info->dev_root;
0403e47e
YZ		 1294 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1295 return fs_info->csum_root;
4df27c4d
YZ		 1296again:
1297 spin_lock(&fs_info->fs_roots_radix_lock);
5eda7b5e
CM		 1298 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1299 (unsigned long)location->objectid);
4df27c4d 1300 spin_unlock(&fs_info->fs_roots_radix_lock);
5eda7b5e
CM		 1301 if (root)
1302 return root;
1303
e02119d5 1304 root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
5eda7b5e
CM		 1305 if (IS_ERR(root))
1306 return root;
3394e160
CM		 1307
1308 set_anon_super(&root->anon_super, NULL);
1309
d68fc57b
YZ		 1310 if (btrfs_root_refs(&root->root_item) == 0) {
1311 ret = -ENOENT;
1312 goto fail;
1313 }
1314
1315 ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
1316 if (ret < 0)
1317 goto fail;
1318 if (ret == 0)
1319 root->orphan_item_inserted = 1;
1320
4df27c4d
YZ		 1321 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1322 if (ret)
1323 goto fail;
1324
1325 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 1326 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1327 (unsigned long)root->root_key.objectid,
0f7d52f4 1328 root);
d68fc57b 1329 if (ret == 0)
4df27c4d 1330 root->in_radix = 1;
d68fc57b 1331
4df27c4d
YZ		 1332 spin_unlock(&fs_info->fs_roots_radix_lock);
1333 radix_tree_preload_end();
0f7d52f4 1334 if (ret) {
4df27c4d
YZ		 1335 if (ret == -EEXIST) {
1336 free_fs_root(root);
1337 goto again;
1338 }
1339 goto fail;
0f7d52f4 1340 }
4df27c4d
YZ		 1341
1342 ret = btrfs_find_dead_roots(fs_info->tree_root,
1343 root->root_key.objectid);
1344 WARN_ON(ret);
edbd8d4e 1345 return root;
4df27c4d
YZ		 1346fail:
1347 free_fs_root(root);
1348 return ERR_PTR(ret);
edbd8d4e
CM		 1349}
1350
04160088
CM		 1351static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1352{
1353 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1354 int ret = 0;
04160088
CM		 1355 struct btrfs_device *device;
1356 struct backing_dev_info *bdi;
b7967db7 1357
c6e30871 1358 list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1359 if (!device->bdev)
1360 continue;
04160088
CM		 1361 bdi = blk_get_backing_dev_info(device->bdev);
1362 if (bdi && bdi_congested(bdi, bdi_bits)) {
1363 ret = 1;
1364 break;
1365 }
1366 }
1367 return ret;
1368}
1369
ad081f14
JA		 1370/*
1371 * If this fails, caller must call bdi_destroy() to get rid of the
1372 * bdi again.
1373 */
04160088
CM		 1374static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1375{
ad081f14
JA		 1376 int err;
1377
1378 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1379 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1380 if (err)
1381 return err;
1382
4575c9cc 1383 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1384 bdi->congested_fn = btrfs_congested_fn;
1385 bdi->congested_data = info;
1386 return 0;
1387}
1388
ce9adaa5
CM		 1389static int bio_ready_for_csum(struct bio *bio)
1390{
1391 u64 length = 0;
1392 u64 buf_len = 0;
1393 u64 start = 0;
1394 struct page *page;
1395 struct extent_io_tree *io_tree = NULL;
ce9adaa5
CM		 1396 struct bio_vec *bvec;
1397 int i;
1398 int ret;
1399
1400 bio_for_each_segment(bvec, bio, i) {
1401 page = bvec->bv_page;
1402 if (page->private == EXTENT_PAGE_PRIVATE) {
1403 length += bvec->bv_len;
1404 continue;
1405 }
1406 if (!page->private) {
1407 length += bvec->bv_len;
1408 continue;
1409 }
1410 length = bvec->bv_len;
1411 buf_len = page->private >> 2;
1412 start = page_offset(page) + bvec->bv_offset;
1413 io_tree = &BTRFS_I(page->mapping->host)->io_tree;
ce9adaa5
CM		 1414 }
1415 /* are we fully contained in this bio? */
1416 if (buf_len <= length)
1417 return 1;
1418
1419 ret = extent_range_uptodate(io_tree, start + length,
1420 start + buf_len - 1);
ce9adaa5
CM		 1421 return ret;
1422}
1423
8b712842
CM		 1424/*
1425 * called by the kthread helper functions to finally call the bio end_io
1426 * functions. This is where read checksum verification actually happens
1427 */
1428static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1429{
ce9adaa5 1430 struct bio *bio;
8b712842
CM		 1431 struct end_io_wq *end_io_wq;
1432 struct btrfs_fs_info *fs_info;
ce9adaa5 1433 int error;
ce9adaa5 1434
8b712842
CM		 1435 end_io_wq = container_of(work, struct end_io_wq, work);
1436 bio = end_io_wq->bio;
1437 fs_info = end_io_wq->info;
ce9adaa5 1438
cad321ad 1439 /* metadata bio reads are special because the whole tree block must
8b712842
CM		 1440 * be checksummed at once. This makes sure the entire block is in
1441 * ram and up to date before trying to verify things. For
1442 * blocksize <= pagesize, it is basically a noop
1443 */
7b6d91da 1444 if (!(bio->bi_rw & REQ_WRITE) && end_io_wq->metadata &&
cad321ad 1445 !bio_ready_for_csum(bio)) {
d20f7043 1446 btrfs_queue_worker(&fs_info->endio_meta_workers,
8b712842
CM		 1447 &end_io_wq->work);
1448 return;
1449 }
1450 error = end_io_wq->error;
1451 bio->bi_private = end_io_wq->private;
1452 bio->bi_end_io = end_io_wq->end_io;
1453 kfree(end_io_wq);
8b712842 1454 bio_endio(bio, error);
44b8bd7e
CM		 1455}
1456
a74a4b97
CM		 1457static int cleaner_kthread(void *arg)
1458{
1459 struct btrfs_root *root = arg;
1460
1461 do {
a74a4b97 1462 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
76dda93c
YZ		 1463
1464 if (!(root->fs_info->sb->s_flags & MS_RDONLY) &&
1465 mutex_trylock(&root->fs_info->cleaner_mutex)) {
24bbcf04 1466 btrfs_run_delayed_iputs(root);
76dda93c
YZ		 1467 btrfs_clean_old_snapshots(root);
1468 mutex_unlock(&root->fs_info->cleaner_mutex);
1469 }
a74a4b97
CM		 1470
1471 if (freezing(current)) {
1472 refrigerator();
1473 } else {
a74a4b97 1474 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1475 if (!kthread_should_stop())
1476 schedule();
a74a4b97
CM		 1477 __set_current_state(TASK_RUNNING);
1478 }
1479 } while (!kthread_should_stop());
1480 return 0;
1481}
1482
1483static int transaction_kthread(void *arg)
1484{
1485 struct btrfs_root *root = arg;
1486 struct btrfs_trans_handle *trans;
1487 struct btrfs_transaction *cur;
8929ecfa 1488 u64 transid;
a74a4b97
CM		 1489 unsigned long now;
1490 unsigned long delay;
1491 int ret;
1492
1493 do {
a74a4b97
CM		 1494 delay = HZ * 30;
1495 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1496 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1497
8929ecfa 1498 spin_lock(&root->fs_info->new_trans_lock);
a74a4b97
CM		 1499 cur = root->fs_info->running_transaction;
1500 if (!cur) {
8929ecfa 1501 spin_unlock(&root->fs_info->new_trans_lock);
a74a4b97
CM		 1502 goto sleep;
1503 }
31153d81 1504
a74a4b97 1505 now = get_seconds();
8929ecfa
YZ		 1506 if (!cur->blocked &&
1507 (now < cur->start_time || now - cur->start_time < 30)) {
1508 spin_unlock(&root->fs_info->new_trans_lock);
a74a4b97
CM		 1509 delay = HZ * 5;
1510 goto sleep;
1511 }
8929ecfa
YZ		 1512 transid = cur->transid;
1513 spin_unlock(&root->fs_info->new_trans_lock);
56bec294 1514
8929ecfa 1515 trans = btrfs_join_transaction(root, 1);
3612b495 1516 BUG_ON(IS_ERR(trans));
8929ecfa
YZ		 1517 if (transid == trans->transid) {
1518 ret = btrfs_commit_transaction(trans, root);
1519 BUG_ON(ret);
1520 } else {
1521 btrfs_end_transaction(trans, root);
1522 }
a74a4b97
CM		 1523sleep:
1524 wake_up_process(root->fs_info->cleaner_kthread);
1525 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1526
1527 if (freezing(current)) {
1528 refrigerator();
1529 } else {
a74a4b97 1530 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1531 if (!kthread_should_stop() &&
1532 !btrfs_transaction_blocked(root->fs_info))
1533 schedule_timeout(delay);
a74a4b97
CM		 1534 __set_current_state(TASK_RUNNING);
1535 }
1536 } while (!kthread_should_stop());
1537 return 0;
1538}
1539
8a4b83cc 1540struct btrfs_root *open_ctree(struct super_block *sb,
dfe25020
CM		 1541 struct btrfs_fs_devices *fs_devices,
1542 char *options)
2e635a27 1543{
db94535d
CM		 1544 u32 sectorsize;
1545 u32 nodesize;
1546 u32 leafsize;
1547 u32 blocksize;
87ee04eb 1548 u32 stripesize;
84234f3a 1549 u64 generation;
f2b636e8 1550 u64 features;
3de4586c 1551 struct btrfs_key location;
a061fc8d 1552 struct buffer_head *bh;
e02119d5 1553 struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
e20d96d6 1554 GFP_NOFS);
d20f7043
CM		 1555 struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root),
1556 GFP_NOFS);
450ba0ea 1557 struct btrfs_root *tree_root = btrfs_sb(sb);
4891aca2 1558 struct btrfs_fs_info *fs_info = NULL;
e02119d5 1559 struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
0b86a832 1560 GFP_NOFS);
e02119d5 1561 struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
0b86a832 1562 GFP_NOFS);
e02119d5
CM		 1563 struct btrfs_root *log_tree_root;
1564
eb60ceac 1565 int ret;
e58ca020 1566 int err = -EINVAL;
4543df7e 1567
2c90e5d6 1568 struct btrfs_super_block *disk_super;
8790d502 1569
4891aca2 1570 if (!extent_root || !tree_root || !tree_root->fs_info ||
d20f7043 1571 !chunk_root || !dev_root || !csum_root) {
39279cc3
CM		 1572 err = -ENOMEM;
1573 goto fail;
1574 }
4891aca2 1575 fs_info = tree_root->fs_info;
76dda93c
YZ		 1576
1577 ret = init_srcu_struct(&fs_info->subvol_srcu);
1578 if (ret) {
1579 err = ret;
1580 goto fail;
1581 }
1582
1583 ret = setup_bdi(fs_info, &fs_info->bdi);
1584 if (ret) {
1585 err = ret;
1586 goto fail_srcu;
1587 }
1588
1589 fs_info->btree_inode = new_inode(sb);
1590 if (!fs_info->btree_inode) {
1591 err = -ENOMEM;
1592 goto fail_bdi;
1593 }
1594
1561deda
MX		 1595 fs_info->btree_inode->i_mapping->flags &= ~__GFP_FS;
1596
76dda93c 1597 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
8fd17795 1598 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 1599 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 1600 INIT_LIST_HEAD(&fs_info->delayed_iputs);
19c00ddc 1601 INIT_LIST_HEAD(&fs_info->hashers);
ea8c2819 1602 INIT_LIST_HEAD(&fs_info->delalloc_inodes);
5a3f23d5 1603 INIT_LIST_HEAD(&fs_info->ordered_operations);
11833d66 1604 INIT_LIST_HEAD(&fs_info->caching_block_groups);
1832a6d5 1605 spin_lock_init(&fs_info->delalloc_lock);
cee36a03 1606 spin_lock_init(&fs_info->new_trans_lock);
31153d81 1607 spin_lock_init(&fs_info->ref_cache_lock);
76dda93c 1608 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 1609 spin_lock_init(&fs_info->delayed_iput_lock);
19c00ddc 1610
58176a96 1611 init_completion(&fs_info->kobj_unregister);
9f5fae2f
CM		 1612 fs_info->tree_root = tree_root;
1613 fs_info->extent_root = extent_root;
d20f7043 1614 fs_info->csum_root = csum_root;
0b86a832
CM		 1615 fs_info->chunk_root = chunk_root;
1616 fs_info->dev_root = dev_root;
8a4b83cc 1617 fs_info->fs_devices = fs_devices;
0b86a832 1618 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 1619 INIT_LIST_HEAD(&fs_info->space_info);
0b86a832 1620 btrfs_mapping_init(&fs_info->mapping_tree);
f0486c68
YZ		 1621 btrfs_init_block_rsv(&fs_info->global_block_rsv);
1622 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv);
1623 btrfs_init_block_rsv(&fs_info->trans_block_rsv);
1624 btrfs_init_block_rsv(&fs_info->chunk_block_rsv);
1625 btrfs_init_block_rsv(&fs_info->empty_block_rsv);
1626 INIT_LIST_HEAD(&fs_info->durable_block_rsv_list);
1627 mutex_init(&fs_info->durable_block_rsv_mutex);
cb03c743 1628 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 1629 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 1630 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 1631 atomic_set(&fs_info->nr_async_bios, 0);
e20d96d6 1632 fs_info->sb = sb;
6f568d35 1633 fs_info->max_inline = 8192 * 1024;
9ed74f2d 1634 fs_info->metadata_ratio = 0;
c8b97818 1635
b34b086c
CM		 1636 fs_info->thread_pool_size = min_t(unsigned long,
1637 num_online_cpus() + 2, 8);
0afbaf8c 1638
3eaa2885
CM		 1639 INIT_LIST_HEAD(&fs_info->ordered_extents);
1640 spin_lock_init(&fs_info->ordered_extent_lock);
1641
a061fc8d
CM		 1642 sb->s_blocksize = 4096;
1643 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 1644 sb->s_bdi = &fs_info->bdi;
a061fc8d 1645
76dda93c
YZ		 1646 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
1647 fs_info->btree_inode->i_nlink = 1;
0afbaf8c
CM		 1648 /*
1649 * we set the i_size on the btree inode to the max possible int.
1650 * the real end of the address space is determined by all of
1651 * the devices in the system
1652 */
1653 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 1654 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 1655 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
1656
5d4f98a2 1657 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 1658 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 1659 fs_info->btree_inode->i_mapping);
a8067e02 1660 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM		 1661
1662 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 1663
76dda93c
YZ		 1664 BTRFS_I(fs_info->btree_inode)->root = tree_root;
1665 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
1666 sizeof(struct btrfs_key));
1667 BTRFS_I(fs_info->btree_inode)->dummy_inode = 1;
c65ddb52 1668 insert_inode_hash(fs_info->btree_inode);
76dda93c 1669
0f9dd46c 1670 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 1671 fs_info->block_group_cache_tree = RB_ROOT;
0f9dd46c 1672
11833d66 1673 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 1674 fs_info->btree_inode->i_mapping);
11833d66 1675 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 1676 fs_info->btree_inode->i_mapping);
11833d66 1677 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 1678 fs_info->do_barriers = 1;
e18e4809 1679
39279cc3 1680
79154b1b 1681 mutex_init(&fs_info->trans_mutex);
5a3f23d5 1682 mutex_init(&fs_info->ordered_operations_mutex);
e02119d5 1683 mutex_init(&fs_info->tree_log_mutex);
925baedd 1684 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 1685 mutex_init(&fs_info->transaction_kthread_mutex);
1686 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 1687 mutex_init(&fs_info->volume_mutex);
276e680d 1688 init_rwsem(&fs_info->extent_commit_sem);
c71bf099 1689 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 1690 init_rwsem(&fs_info->subvol_sem);
fa9c0d79
CM		 1691
1692 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
1693 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
1694
e6dcd2dc 1695 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 1696 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 1697 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 1698 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 1699
0b86a832 1700 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 1701 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 1702
a512bbf8 1703 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY		 1704 if (!bh) {
1705 err = -EINVAL;
39279cc3 1706 goto fail_iput;
20b45077 1707 }
39279cc3 1708
a061fc8d 1709 memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy));
2d69a0f8
YZ		 1710 memcpy(&fs_info->super_for_commit, &fs_info->super_copy,
1711 sizeof(fs_info->super_for_commit));
a061fc8d 1712 brelse(bh);
5f39d397 1713
a061fc8d 1714 memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE);
0b86a832 1715
5f39d397 1716 disk_super = &fs_info->super_copy;
0f7d52f4 1717 if (!btrfs_super_root(disk_super))
c6e2bac1 1718 goto fail_iput;
0f7d52f4 1719
acce952b 1720 /* check FS state, whether FS is broken. */
1721 fs_info->fs_state |= btrfs_super_flags(disk_super);
1722
1723 btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
1724
75e7cb7f
LB		 1725 /*
1726 * In the long term, we'll store the compression type in the super
1727 * block, and it'll be used for per file compression control.
1728 */
1729 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
1730
2b82032c
YZ		 1731 ret = btrfs_parse_options(tree_root, options);
1732 if (ret) {
1733 err = ret;
c6e2bac1 1734 goto fail_iput;
2b82032c 1735 }
dfe25020 1736
f2b636e8
JB		 1737 features = btrfs_super_incompat_flags(disk_super) &
1738 ~BTRFS_FEATURE_INCOMPAT_SUPP;
1739 if (features) {
1740 printk(KERN_ERR "BTRFS: couldn't mount because of "
1741 "unsupported optional features (%Lx).\n",
21380931 1742 (unsigned long long)features);
f2b636e8 1743 err = -EINVAL;
c6e2bac1 1744 goto fail_iput;
f2b636e8
JB		 1745 }
1746
5d4f98a2 1747 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae
LZ		 1748 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
1749 if (tree_root->fs_info->compress_type & BTRFS_COMPRESS_LZO)
1750 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1751 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 1752
f2b636e8
JB		 1753 features = btrfs_super_compat_ro_flags(disk_super) &
1754 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
1755 if (!(sb->s_flags & MS_RDONLY) && features) {
1756 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
1757 "unsupported option features (%Lx).\n",
21380931 1758 (unsigned long long)features);
f2b636e8 1759 err = -EINVAL;
c6e2bac1 1760 goto fail_iput;
f2b636e8 1761 }
61d92c32
CM		 1762
1763 btrfs_init_workers(&fs_info->generic_worker,
1764 "genwork", 1, NULL);
1765
5443be45 1766 btrfs_init_workers(&fs_info->workers, "worker",
61d92c32
CM		 1767 fs_info->thread_pool_size,
1768 &fs_info->generic_worker);
c8b97818 1769
771ed689 1770 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
61d92c32
CM		 1771 fs_info->thread_pool_size,
1772 &fs_info->generic_worker);
771ed689 1773
5443be45 1774 btrfs_init_workers(&fs_info->submit_workers, "submit",
b720d209 1775 min_t(u64, fs_devices->num_devices,
61d92c32
CM		 1776 fs_info->thread_pool_size),
1777 &fs_info->generic_worker);
61b49440
CM		 1778
1779 /* a higher idle thresh on the submit workers makes it much more
1780 * likely that bios will be send down in a sane order to the
1781 * devices
1782 */
1783 fs_info->submit_workers.idle_thresh = 64;
53863232 1784
771ed689 1785 fs_info->workers.idle_thresh = 16;
4a69a410 1786 fs_info->workers.ordered = 1;
61b49440 1787
771ed689
CM		 1788 fs_info->delalloc_workers.idle_thresh = 2;
1789 fs_info->delalloc_workers.ordered = 1;
1790
61d92c32
CM		 1791 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
1792 &fs_info->generic_worker);
5443be45 1793 btrfs_init_workers(&fs_info->endio_workers, "endio",
61d92c32
CM		 1794 fs_info->thread_pool_size,
1795 &fs_info->generic_worker);
d20f7043 1796 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
61d92c32
CM		 1797 fs_info->thread_pool_size,
1798 &fs_info->generic_worker);
cad321ad 1799 btrfs_init_workers(&fs_info->endio_meta_write_workers,
61d92c32
CM		 1800 "endio-meta-write", fs_info->thread_pool_size,
1801 &fs_info->generic_worker);
5443be45 1802 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
61d92c32
CM		 1803 fs_info->thread_pool_size,
1804 &fs_info->generic_worker);
0cb59c99
JB		 1805 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
1806 1, &fs_info->generic_worker);
61b49440
CM		 1807
1808 /*
1809 * endios are largely parallel and should have a very
1810 * low idle thresh
1811 */
1812 fs_info->endio_workers.idle_thresh = 4;
b51912c9
CM		 1813 fs_info->endio_meta_workers.idle_thresh = 4;
1814
9042846b
CM		 1815 fs_info->endio_write_workers.idle_thresh = 2;
1816 fs_info->endio_meta_write_workers.idle_thresh = 2;
1817
4543df7e 1818 btrfs_start_workers(&fs_info->workers, 1);
61d92c32 1819 btrfs_start_workers(&fs_info->generic_worker, 1);
1cc127b5 1820 btrfs_start_workers(&fs_info->submit_workers, 1);
771ed689 1821 btrfs_start_workers(&fs_info->delalloc_workers, 1);
247e743c 1822 btrfs_start_workers(&fs_info->fixup_workers, 1);
9042846b
CM		 1823 btrfs_start_workers(&fs_info->endio_workers, 1);
1824 btrfs_start_workers(&fs_info->endio_meta_workers, 1);
1825 btrfs_start_workers(&fs_info->endio_meta_write_workers, 1);
1826 btrfs_start_workers(&fs_info->endio_write_workers, 1);
0cb59c99 1827 btrfs_start_workers(&fs_info->endio_freespace_worker, 1);
4543df7e 1828
4575c9cc 1829 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 1830 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
1831 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 1832
db94535d
CM		 1833 nodesize = btrfs_super_nodesize(disk_super);
1834 leafsize = btrfs_super_leafsize(disk_super);
1835 sectorsize = btrfs_super_sectorsize(disk_super);
87ee04eb 1836 stripesize = btrfs_super_stripesize(disk_super);
db94535d
CM		 1837 tree_root->nodesize = nodesize;
1838 tree_root->leafsize = leafsize;
1839 tree_root->sectorsize = sectorsize;
87ee04eb 1840 tree_root->stripesize = stripesize;
a061fc8d
CM		 1841
1842 sb->s_blocksize = sectorsize;
1843 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 1844
39279cc3
CM		 1845 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
1846 sizeof(disk_super->magic))) {
d397712b 1847 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 1848 goto fail_sb_buffer;
1849 }
19c00ddc 1850
925baedd 1851 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 1852 ret = btrfs_read_sys_array(tree_root);
925baedd 1853 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 1854 if (ret) {
d397712b
CM		 1855 printk(KERN_WARNING "btrfs: failed to read the system "
1856 "array on %s\n", sb->s_id);
5d4f98a2 1857 goto fail_sb_buffer;
84eed90f 1858 }
0b86a832
CM		 1859
1860 blocksize = btrfs_level_size(tree_root,
1861 btrfs_super_chunk_root_level(disk_super));
84234f3a 1862 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 1863
1864 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1865 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1866
1867 chunk_root->node = read_tree_block(chunk_root,
1868 btrfs_super_chunk_root(disk_super),
84234f3a 1869 blocksize, generation);
0b86a832 1870 BUG_ON(!chunk_root->node);
83121942
DW		 1871 if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
1872 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
1873 sb->s_id);
1874 goto fail_chunk_root;
1875 }
5d4f98a2
YZ		 1876 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
1877 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 1878
e17cade2 1879 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
d397712b
CM		 1880 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
1881 BTRFS_UUID_SIZE);
e17cade2 1882
925baedd 1883 mutex_lock(&fs_info->chunk_mutex);
0b86a832 1884 ret = btrfs_read_chunk_tree(chunk_root);
925baedd 1885 mutex_unlock(&fs_info->chunk_mutex);
2b82032c 1886 if (ret) {
d397712b
CM		 1887 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
1888 sb->s_id);
2b82032c
YZ		 1889 goto fail_chunk_root;
1890 }
0b86a832 1891
dfe25020
CM		 1892 btrfs_close_extra_devices(fs_devices);
1893
db94535d
CM		 1894 blocksize = btrfs_level_size(tree_root,
1895 btrfs_super_root_level(disk_super));
84234f3a 1896 generation = btrfs_super_generation(disk_super);
0b86a832 1897
e20d96d6 1898 tree_root->node = read_tree_block(tree_root,
db94535d 1899 btrfs_super_root(disk_super),
84234f3a 1900 blocksize, generation);
39279cc3 1901 if (!tree_root->node)
2b82032c 1902 goto fail_chunk_root;
83121942
DW		 1903 if (!test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
1904 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
1905 sb->s_id);
1906 goto fail_tree_root;
1907 }
5d4f98a2
YZ		 1908 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
1909 tree_root->commit_root = btrfs_root_node(tree_root);
db94535d
CM		 1910
1911 ret = find_and_setup_root(tree_root, fs_info,
e20d96d6 1912 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
0b86a832 1913 if (ret)
39279cc3 1914 goto fail_tree_root;
0b86a832
CM		 1915 extent_root->track_dirty = 1;
1916
1917 ret = find_and_setup_root(tree_root, fs_info,
1918 BTRFS_DEV_TREE_OBJECTID, dev_root);
0b86a832
CM		 1919 if (ret)
1920 goto fail_extent_root;
5d4f98a2 1921 dev_root->track_dirty = 1;
3768f368 1922
d20f7043
CM		 1923 ret = find_and_setup_root(tree_root, fs_info,
1924 BTRFS_CSUM_TREE_OBJECTID, csum_root);
1925 if (ret)
5d4f98a2 1926 goto fail_dev_root;
d20f7043
CM		 1927
1928 csum_root->track_dirty = 1;
1929
8929ecfa
YZ		 1930 fs_info->generation = generation;
1931 fs_info->last_trans_committed = generation;
1932 fs_info->data_alloc_profile = (u64)-1;
1933 fs_info->metadata_alloc_profile = (u64)-1;
1934 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
1935
c59021f8 1936 ret = btrfs_init_space_info(fs_info);
1937 if (ret) {
1938 printk(KERN_ERR "Failed to initial space info: %d\n", ret);
1939 goto fail_block_groups;
1940 }
1941
1b1d1f66
JB		 1942 ret = btrfs_read_block_groups(extent_root);
1943 if (ret) {
1944 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
1945 goto fail_block_groups;
1946 }
9078a3e1 1947
a74a4b97
CM		 1948 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
1949 "btrfs-cleaner");
57506d50 1950 if (IS_ERR(fs_info->cleaner_kthread))
1b1d1f66 1951 goto fail_block_groups;
a74a4b97
CM		 1952
1953 fs_info->transaction_kthread = kthread_run(transaction_kthread,
1954 tree_root,
1955 "btrfs-transaction");
57506d50 1956 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 1957 goto fail_cleaner;
a74a4b97 1958
c289811c
CM		 1959 if (!btrfs_test_opt(tree_root, SSD) &&
1960 !btrfs_test_opt(tree_root, NOSSD) &&
1961 !fs_info->fs_devices->rotating) {
1962 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
1963 "mode\n");
1964 btrfs_set_opt(fs_info->mount_opt, SSD);
1965 }
1966
acce952b 1967 /* do not make disk changes in broken FS */
1968 if (btrfs_super_log_root(disk_super) != 0 &&
1969 !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) {
e02119d5
CM		 1970 u64 bytenr = btrfs_super_log_root(disk_super);
1971
7c2ca468 1972 if (fs_devices->rw_devices == 0) {
d397712b
CM		 1973 printk(KERN_WARNING "Btrfs log replay required "
1974 "on RO media\n");
7c2ca468
CM		 1975 err = -EIO;
1976 goto fail_trans_kthread;
1977 }
e02119d5
CM		 1978 blocksize =
1979 btrfs_level_size(tree_root,
1980 btrfs_super_log_root_level(disk_super));
d18a2c44 1981
676e4c86
DC		 1982 log_tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
1983 if (!log_tree_root) {
1984 err = -ENOMEM;
1985 goto fail_trans_kthread;
1986 }
e02119d5
CM		 1987
1988 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1989 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1990
1991 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 1992 blocksize,
1993 generation + 1);
e02119d5
CM		 1994 ret = btrfs_recover_log_trees(log_tree_root);
1995 BUG_ON(ret);
e556ce2c
YZ		 1996
1997 if (sb->s_flags & MS_RDONLY) {
1998 ret = btrfs_commit_super(tree_root);
1999 BUG_ON(ret);
2000 }
e02119d5 2001 }
1a40e23b 2002
76dda93c
YZ		 2003 ret = btrfs_find_orphan_roots(tree_root);
2004 BUG_ON(ret);
2005
7c2ca468 2006 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b
YZ		 2007 ret = btrfs_cleanup_fs_roots(fs_info);
2008 BUG_ON(ret);
2009
5d4f98a2 2010 ret = btrfs_recover_relocation(tree_root);
d7ce5843
MX		 2011 if (ret < 0) {
2012 printk(KERN_WARNING
2013 "btrfs: failed to recover relocation\n");
2014 err = -EINVAL;
2015 goto fail_trans_kthread;
2016 }
7c2ca468 2017 }
1a40e23b 2018
3de4586c
CM		 2019 location.objectid = BTRFS_FS_TREE_OBJECTID;
2020 location.type = BTRFS_ROOT_ITEM_KEY;
2021 location.offset = (u64)-1;
2022
3de4586c
CM		 2023 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
2024 if (!fs_info->fs_root)
7c2ca468 2025 goto fail_trans_kthread;
3140c9a3
DC		 2026 if (IS_ERR(fs_info->fs_root)) {
2027 err = PTR_ERR(fs_info->fs_root);
2028 goto fail_trans_kthread;
2029 }
c289811c 2030
e3acc2a6
JB		 2031 if (!(sb->s_flags & MS_RDONLY)) {
2032 down_read(&fs_info->cleanup_work_sem);
66b4ffd1
JB		 2033 err = btrfs_orphan_cleanup(fs_info->fs_root);
2034 if (!err)
2035 err = btrfs_orphan_cleanup(fs_info->tree_root);
e3acc2a6 2036 up_read(&fs_info->cleanup_work_sem);
66b4ffd1
JB		 2037 if (err) {
2038 close_ctree(tree_root);
2039 return ERR_PTR(err);
2040 }
e3acc2a6
JB		 2041 }
2042
0f7d52f4 2043 return tree_root;
39279cc3 2044
7c2ca468
CM		 2045fail_trans_kthread:
2046 kthread_stop(fs_info->transaction_kthread);
3f157a2f 2047fail_cleaner:
a74a4b97 2048 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2049
2050 /*
2051 * make sure we're done with the btree inode before we stop our
2052 * kthreads
2053 */
2054 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
2055 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
2056
1b1d1f66
JB		 2057fail_block_groups:
2058 btrfs_free_block_groups(fs_info);
d20f7043 2059 free_extent_buffer(csum_root->node);
5d4f98a2
YZ		 2060 free_extent_buffer(csum_root->commit_root);
2061fail_dev_root:
2062 free_extent_buffer(dev_root->node);
2063 free_extent_buffer(dev_root->commit_root);
0b86a832
CM		 2064fail_extent_root:
2065 free_extent_buffer(extent_root->node);
5d4f98a2 2066 free_extent_buffer(extent_root->commit_root);
39279cc3 2067fail_tree_root:
5f39d397 2068 free_extent_buffer(tree_root->node);
5d4f98a2 2069 free_extent_buffer(tree_root->commit_root);
2b82032c
YZ		 2070fail_chunk_root:
2071 free_extent_buffer(chunk_root->node);
5d4f98a2 2072 free_extent_buffer(chunk_root->commit_root);
39279cc3 2073fail_sb_buffer:
61d92c32 2074 btrfs_stop_workers(&fs_info->generic_worker);
247e743c 2075 btrfs_stop_workers(&fs_info->fixup_workers);
771ed689 2076 btrfs_stop_workers(&fs_info->delalloc_workers);
8b712842
CM		 2077 btrfs_stop_workers(&fs_info->workers);
2078 btrfs_stop_workers(&fs_info->endio_workers);
d20f7043 2079 btrfs_stop_workers(&fs_info->endio_meta_workers);
cad321ad 2080 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
e6dcd2dc 2081 btrfs_stop_workers(&fs_info->endio_write_workers);
0cb59c99 2082 btrfs_stop_workers(&fs_info->endio_freespace_worker);
1cc127b5 2083 btrfs_stop_workers(&fs_info->submit_workers);
4543df7e 2084fail_iput:
7c2ca468 2085 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
4543df7e 2086 iput(fs_info->btree_inode);
7e662854 2087
dfe25020 2088 btrfs_close_devices(fs_info->fs_devices);
84eed90f 2089 btrfs_mapping_tree_free(&fs_info->mapping_tree);
ad081f14 2090fail_bdi:
7e662854 2091 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 2092fail_srcu:
2093 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 2094fail:
39279cc3
CM		 2095 kfree(extent_root);
2096 kfree(tree_root);
2097 kfree(fs_info);
83afeac4
JM		 2098 kfree(chunk_root);
2099 kfree(dev_root);
d20f7043 2100 kfree(csum_root);
39279cc3 2101 return ERR_PTR(err);
eb60ceac
CM		 2102}
2103
f2984462
CM		 2104static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2105{
2106 char b[BDEVNAME_SIZE];
2107
2108 if (uptodate) {
2109 set_buffer_uptodate(bh);
2110 } else {
c3b9a62c 2111 if (printk_ratelimit()) {
f2984462
CM		 2112 printk(KERN_WARNING "lost page write due to "
2113 "I/O error on %s\n",
2114 bdevname(bh->b_bdev, b));
2115 }
1259ab75
CM		 2116 /* note, we dont' set_buffer_write_io_error because we have
2117 * our own ways of dealing with the IO errors
2118 */
f2984462
CM		 2119 clear_buffer_uptodate(bh);
2120 }
2121 unlock_buffer(bh);
2122 put_bh(bh);
2123}
2124
a512bbf8
YZ		 2125struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
2126{
2127 struct buffer_head *bh;
2128 struct buffer_head *latest = NULL;
2129 struct btrfs_super_block *super;
2130 int i;
2131 u64 transid = 0;
2132 u64 bytenr;
2133
2134 /* we would like to check all the supers, but that would make
2135 * a btrfs mount succeed after a mkfs from a different FS.
2136 * So, we need to add a special mount option to scan for
2137 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2138 */
2139 for (i = 0; i < 1; i++) {
2140 bytenr = btrfs_sb_offset(i);
2141 if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
2142 break;
2143 bh = __bread(bdev, bytenr / 4096, 4096);
2144 if (!bh)
2145 continue;
2146
2147 super = (struct btrfs_super_block *)bh->b_data;
2148 if (btrfs_super_bytenr(super) != bytenr ||
2149 strncmp((char *)(&super->magic), BTRFS_MAGIC,
2150 sizeof(super->magic))) {
2151 brelse(bh);
2152 continue;
2153 }
2154
2155 if (!latest || btrfs_super_generation(super) > transid) {
2156 brelse(latest);
2157 latest = bh;
2158 transid = btrfs_super_generation(super);
2159 } else {
2160 brelse(bh);
2161 }
2162 }
2163 return latest;
2164}
2165
4eedeb75
HH		 2166/*
2167 * this should be called twice, once with wait == 0 and
2168 * once with wait == 1. When wait == 0 is done, all the buffer heads
2169 * we write are pinned.
2170 *
2171 * They are released when wait == 1 is done.
2172 * max_mirrors must be the same for both runs, and it indicates how
2173 * many supers on this one device should be written.
2174 *
2175 * max_mirrors == 0 means to write them all.
2176 */
a512bbf8
YZ		 2177static int write_dev_supers(struct btrfs_device *device,
2178 struct btrfs_super_block *sb,
2179 int do_barriers, int wait, int max_mirrors)
2180{
2181 struct buffer_head *bh;
2182 int i;
2183 int ret;
2184 int errors = 0;
2185 u32 crc;
2186 u64 bytenr;
2187 int last_barrier = 0;
2188
2189 if (max_mirrors == 0)
2190 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
2191
2192 /* make sure only the last submit_bh does a barrier */
2193 if (do_barriers) {
2194 for (i = 0; i < max_mirrors; i++) {
2195 bytenr = btrfs_sb_offset(i);
2196 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
2197 device->total_bytes)
2198 break;
2199 last_barrier = i;
2200 }
2201 }
2202
2203 for (i = 0; i < max_mirrors; i++) {
2204 bytenr = btrfs_sb_offset(i);
2205 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
2206 break;
2207
2208 if (wait) {
2209 bh = __find_get_block(device->bdev, bytenr / 4096,
2210 BTRFS_SUPER_INFO_SIZE);
2211 BUG_ON(!bh);
a512bbf8 2212 wait_on_buffer(bh);
4eedeb75
HH		 2213 if (!buffer_uptodate(bh))
2214 errors++;
2215
2216 /* drop our reference */
2217 brelse(bh);
2218
2219 /* drop the reference from the wait == 0 run */
2220 brelse(bh);
2221 continue;
a512bbf8
YZ		 2222 } else {
2223 btrfs_set_super_bytenr(sb, bytenr);
2224
2225 crc = ~(u32)0;
2226 crc = btrfs_csum_data(NULL, (char *)sb +
2227 BTRFS_CSUM_SIZE, crc,
2228 BTRFS_SUPER_INFO_SIZE -
2229 BTRFS_CSUM_SIZE);
2230 btrfs_csum_final(crc, sb->csum);
2231
4eedeb75
HH		 2232 /*
2233 * one reference for us, and we leave it for the
2234 * caller
2235 */
a512bbf8
YZ		 2236 bh = __getblk(device->bdev, bytenr / 4096,
2237 BTRFS_SUPER_INFO_SIZE);
2238 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
2239
4eedeb75 2240 /* one reference for submit_bh */
a512bbf8 2241 get_bh(bh);
4eedeb75
HH		 2242
2243 set_buffer_uptodate(bh);
a512bbf8
YZ		 2244 lock_buffer(bh);
2245 bh->b_end_io = btrfs_end_buffer_write_sync;
2246 }
2247
c3b9a62c
CH		 2248 if (i == last_barrier && do_barriers)
2249 ret = submit_bh(WRITE_FLUSH_FUA, bh);
2250 else
ffbd517d 2251 ret = submit_bh(WRITE_SYNC, bh);
a512bbf8 2252
4eedeb75 2253 if (ret)
a512bbf8 2254 errors++;
a512bbf8
YZ		 2255 }
2256 return errors < i ? 0 : -1;
2257}
2258
2259int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 2260{
e5e9a520 2261 struct list_head *head;
f2984462 2262 struct btrfs_device *dev;
a061fc8d 2263 struct btrfs_super_block *sb;
f2984462 2264 struct btrfs_dev_item *dev_item;
f2984462
CM		 2265 int ret;
2266 int do_barriers;
a236aed1
CM		 2267 int max_errors;
2268 int total_errors = 0;
a061fc8d 2269 u64 flags;
f2984462 2270
a236aed1 2271 max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
f2984462
CM		 2272 do_barriers = !btrfs_test_opt(root, NOBARRIER);
2273
a061fc8d
CM		 2274 sb = &root->fs_info->super_for_commit;
2275 dev_item = &sb->dev_item;
e5e9a520
CM		 2276
2277 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
2278 head = &root->fs_info->fs_devices->devices;
c6e30871 2279 list_for_each_entry(dev, head, dev_list) {
dfe25020
CM		 2280 if (!dev->bdev) {
2281 total_errors++;
2282 continue;
2283 }
2b82032c 2284 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 2285 continue;
2286
2b82032c 2287 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 2288 btrfs_set_stack_device_type(dev_item, dev->type);
2289 btrfs_set_stack_device_id(dev_item, dev->devid);
2290 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
2291 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
2292 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
2293 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
2294 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
2295 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 2296 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 2297
a061fc8d
CM		 2298 flags = btrfs_super_flags(sb);
2299 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
2300
a512bbf8 2301 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 2302 if (ret)
2303 total_errors++;
f2984462 2304 }
a236aed1 2305 if (total_errors > max_errors) {
d397712b
CM		 2306 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2307 total_errors);
a236aed1
CM		 2308 BUG();
2309 }
f2984462 2310
a512bbf8 2311 total_errors = 0;
c6e30871 2312 list_for_each_entry(dev, head, dev_list) {
dfe25020
CM		 2313 if (!dev->bdev)
2314 continue;
2b82032c 2315 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 2316 continue;
2317
a512bbf8
YZ		 2318 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
2319 if (ret)
2320 total_errors++;
f2984462 2321 }
e5e9a520 2322 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 2323 if (total_errors > max_errors) {
d397712b
CM		 2324 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2325 total_errors);
a236aed1
CM		 2326 BUG();
2327 }
f2984462
CM		 2328 return 0;
2329}
2330
a512bbf8
YZ		 2331int write_ctree_super(struct btrfs_trans_handle *trans,
2332 struct btrfs_root *root, int max_mirrors)
eb60ceac 2333{
e66f709b 2334 int ret;
5f39d397 2335
a512bbf8 2336 ret = write_all_supers(root, max_mirrors);
5f39d397 2337 return ret;
cfaa7295
CM		 2338}
2339
5eda7b5e 2340int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
2619ba1f 2341{
4df27c4d 2342 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 2343 radix_tree_delete(&fs_info->fs_roots_radix,
2344 (unsigned long)root->root_key.objectid);
4df27c4d 2345 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 2346
2347 if (btrfs_root_refs(&root->root_item) == 0)
2348 synchronize_srcu(&fs_info->subvol_srcu);
2349
4df27c4d
YZ		 2350 free_fs_root(root);
2351 return 0;
2352}
2353
2354static void free_fs_root(struct btrfs_root *root)
2355{
2356 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
3394e160
CM		 2357 if (root->anon_super.s_dev) {
2358 down_write(&root->anon_super.s_umount);
2359 kill_anon_super(&root->anon_super);
2360 }
4df27c4d
YZ		 2361 free_extent_buffer(root->node);
2362 free_extent_buffer(root->commit_root);
d397712b 2363 kfree(root->name);
2619ba1f 2364 kfree(root);
2619ba1f
CM		 2365}
2366
35b7e476 2367static int del_fs_roots(struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 2368{
2369 int ret;
2370 struct btrfs_root *gang[8];
2371 int i;
2372
76dda93c
YZ		 2373 while (!list_empty(&fs_info->dead_roots)) {
2374 gang[0] = list_entry(fs_info->dead_roots.next,
2375 struct btrfs_root, root_list);
2376 list_del(&gang[0]->root_list);
2377
2378 if (gang[0]->in_radix) {
2379 btrfs_free_fs_root(fs_info, gang[0]);
2380 } else {
2381 free_extent_buffer(gang[0]->node);
2382 free_extent_buffer(gang[0]->commit_root);
2383 kfree(gang[0]);
2384 }
2385 }
2386
d397712b 2387 while (1) {
0f7d52f4
CM		 2388 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2389 (void **)gang, 0,
2390 ARRAY_SIZE(gang));
2391 if (!ret)
2392 break;
2619ba1f 2393 for (i = 0; i < ret; i++)
5eda7b5e 2394 btrfs_free_fs_root(fs_info, gang[i]);
0f7d52f4
CM		 2395 }
2396 return 0;
2397}
b4100d64 2398
c146afad 2399int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 2400{
c146afad
YZ		 2401 u64 root_objectid = 0;
2402 struct btrfs_root *gang[8];
2403 int i;
3768f368 2404 int ret;
e089f05c 2405
c146afad
YZ		 2406 while (1) {
2407 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2408 (void **)gang, root_objectid,
2409 ARRAY_SIZE(gang));
2410 if (!ret)
2411 break;
5d4f98a2
YZ		 2412
2413 root_objectid = gang[ret - 1]->root_key.objectid + 1;
c146afad 2414 for (i = 0; i < ret; i++) {
66b4ffd1
JB		 2415 int err;
2416
c146afad 2417 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 2418 err = btrfs_orphan_cleanup(gang[i]);
2419 if (err)
2420 return err;
c146afad
YZ		 2421 }
2422 root_objectid++;
2423 }
2424 return 0;
2425}
a2135011 2426
c146afad
YZ		 2427int btrfs_commit_super(struct btrfs_root *root)
2428{
2429 struct btrfs_trans_handle *trans;
2430 int ret;
a74a4b97 2431
c146afad 2432 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 2433 btrfs_run_delayed_iputs(root);
a74a4b97 2434 btrfs_clean_old_snapshots(root);
c146afad 2435 mutex_unlock(&root->fs_info->cleaner_mutex);
c71bf099
YZ		 2436
2437 /* wait until ongoing cleanup work done */
2438 down_write(&root->fs_info->cleanup_work_sem);
2439 up_write(&root->fs_info->cleanup_work_sem);
2440
a22285a6 2441 trans = btrfs_join_transaction(root, 1);
3612b495
TI		 2442 if (IS_ERR(trans))
2443 return PTR_ERR(trans);
54aa1f4d 2444 ret = btrfs_commit_transaction(trans, root);
c146afad
YZ		 2445 BUG_ON(ret);
2446 /* run commit again to drop the original snapshot */
a22285a6 2447 trans = btrfs_join_transaction(root, 1);
3612b495
TI		 2448 if (IS_ERR(trans))
2449 return PTR_ERR(trans);
79154b1b
CM		 2450 btrfs_commit_transaction(trans, root);
2451 ret = btrfs_write_and_wait_transaction(NULL, root);
3768f368 2452 BUG_ON(ret);
d6bfde87 2453
a512bbf8 2454 ret = write_ctree_super(NULL, root, 0);
c146afad
YZ		 2455 return ret;
2456}
2457
2458int close_ctree(struct btrfs_root *root)
2459{
2460 struct btrfs_fs_info *fs_info = root->fs_info;
2461 int ret;
2462
2463 fs_info->closing = 1;
2464 smp_mb();
2465
0af3d00b 2466 btrfs_put_block_group_cache(fs_info);
acce952b 2467
2468 /*
2469 * Here come 2 situations when btrfs is broken to flip readonly:
2470 *
2471 * 1. when btrfs flips readonly somewhere else before
2472 * btrfs_commit_super, sb->s_flags has MS_RDONLY flag,
2473 * and btrfs will skip to write sb directly to keep
2474 * ERROR state on disk.
2475 *
2476 * 2. when btrfs flips readonly just in btrfs_commit_super,
ae0e47f0 2477 * and in such case, btrfs cannot write sb via btrfs_commit_super,
acce952b 2478 * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag,
2479 * btrfs will cleanup all FS resources first and write sb then.
2480 */
c146afad 2481 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 2482 ret = btrfs_commit_super(root);
2483 if (ret)
2484 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
2485 }
2486
2487 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
2488 ret = btrfs_error_commit_super(root);
d397712b
CM		 2489 if (ret)
2490 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
c146afad 2491 }
0f7d52f4 2492
8929ecfa
YZ		 2493 kthread_stop(root->fs_info->transaction_kthread);
2494 kthread_stop(root->fs_info->cleaner_kthread);
2495
f25784b3
YZ		 2496 fs_info->closing = 2;
2497 smp_mb();
2498
b0c68f8b 2499 if (fs_info->delalloc_bytes) {
d397712b 2500 printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
21380931 2501 (unsigned long long)fs_info->delalloc_bytes);
b0c68f8b 2502 }
31153d81 2503 if (fs_info->total_ref_cache_size) {
d397712b
CM		 2504 printk(KERN_INFO "btrfs: at umount reference cache size %llu\n",
2505 (unsigned long long)fs_info->total_ref_cache_size);
31153d81 2506 }
bcc63abb 2507
5d4f98a2
YZ		 2508 free_extent_buffer(fs_info->extent_root->node);
2509 free_extent_buffer(fs_info->extent_root->commit_root);
2510 free_extent_buffer(fs_info->tree_root->node);
2511 free_extent_buffer(fs_info->tree_root->commit_root);
2512 free_extent_buffer(root->fs_info->chunk_root->node);
2513 free_extent_buffer(root->fs_info->chunk_root->commit_root);
2514 free_extent_buffer(root->fs_info->dev_root->node);
2515 free_extent_buffer(root->fs_info->dev_root->commit_root);
2516 free_extent_buffer(root->fs_info->csum_root->node);
2517 free_extent_buffer(root->fs_info->csum_root->commit_root);
d20f7043 2518
9078a3e1 2519 btrfs_free_block_groups(root->fs_info);
d10c5f31 2520
c146afad 2521 del_fs_roots(fs_info);
d10c5f31 2522
c146afad 2523 iput(fs_info->btree_inode);
9ad6b7bc 2524
61d92c32 2525 btrfs_stop_workers(&fs_info->generic_worker);
247e743c 2526 btrfs_stop_workers(&fs_info->fixup_workers);
771ed689 2527 btrfs_stop_workers(&fs_info->delalloc_workers);
8b712842
CM		 2528 btrfs_stop_workers(&fs_info->workers);
2529 btrfs_stop_workers(&fs_info->endio_workers);
d20f7043 2530 btrfs_stop_workers(&fs_info->endio_meta_workers);
cad321ad 2531 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
e6dcd2dc 2532 btrfs_stop_workers(&fs_info->endio_write_workers);
0cb59c99 2533 btrfs_stop_workers(&fs_info->endio_freespace_worker);
1cc127b5 2534 btrfs_stop_workers(&fs_info->submit_workers);
d6bfde87 2535
dfe25020 2536 btrfs_close_devices(fs_info->fs_devices);
0b86a832 2537 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 2538
04160088 2539 bdi_destroy(&fs_info->bdi);
76dda93c 2540 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 2541
0f7d52f4 2542 kfree(fs_info->extent_root);
0f7d52f4 2543 kfree(fs_info->tree_root);
0b86a832
CM		 2544 kfree(fs_info->chunk_root);
2545 kfree(fs_info->dev_root);
d20f7043 2546 kfree(fs_info->csum_root);
83a4d548
LZ		 2547 kfree(fs_info);
2548
eb60ceac
CM		 2549 return 0;
2550}
2551
1259ab75 2552int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
5f39d397 2553{
1259ab75 2554 int ret;
810191ff 2555 struct inode *btree_inode = buf->first_page->mapping->host;
1259ab75 2556
2ac55d41
JB		 2557 ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf,
2558 NULL);
1259ab75
CM		 2559 if (!ret)
2560 return ret;
2561
2562 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
2563 parent_transid);
2564 return !ret;
5f39d397
CM		 2565}
2566
2567int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 2568{
810191ff 2569 struct inode *btree_inode = buf->first_page->mapping->host;
d1310b2e 2570 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
5f39d397
CM		 2571 buf);
2572}
6702ed49 2573
5f39d397
CM		 2574void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
2575{
810191ff 2576 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
5f39d397
CM		 2577 u64 transid = btrfs_header_generation(buf);
2578 struct inode *btree_inode = root->fs_info->btree_inode;
b9473439 2579 int was_dirty;
b4ce94de 2580
b9447ef8 2581 btrfs_assert_tree_locked(buf);
ccd467d6 2582 if (transid != root->fs_info->generation) {
d397712b
CM		 2583 printk(KERN_CRIT "btrfs transid mismatch buffer %llu, "
2584 "found %llu running %llu\n",
db94535d 2585 (unsigned long long)buf->start,
d397712b
CM		 2586 (unsigned long long)transid,
2587 (unsigned long long)root->fs_info->generation);
ccd467d6
CM		 2588 WARN_ON(1);
2589 }
b9473439
CM		 2590 was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
2591 buf);
2592 if (!was_dirty) {
2593 spin_lock(&root->fs_info->delalloc_lock);
2594 root->fs_info->dirty_metadata_bytes += buf->len;
2595 spin_unlock(&root->fs_info->delalloc_lock);
2596 }
eb60ceac
CM		 2597}
2598
d3c2fdcf 2599void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
35b7e476 2600{
188de649
CM		 2601 /*
2602 * looks as though older kernels can get into trouble with
2603 * this code, they end up stuck in balance_dirty_pages forever
2604 */
d6bfde87 2605 u64 num_dirty;
771ed689 2606 unsigned long thresh = 32 * 1024 * 1024;
d6bfde87 2607
6933c02e 2608 if (current->flags & PF_MEMALLOC)
d6bfde87
CM		 2609 return;
2610
585ad2c3
CM		 2611 num_dirty = root->fs_info->dirty_metadata_bytes;
2612
d6bfde87
CM		 2613 if (num_dirty > thresh) {
2614 balance_dirty_pages_ratelimited_nr(
d7fc640e 2615 root->fs_info->btree_inode->i_mapping, 1);
d6bfde87 2616 }
188de649 2617 return;
35b7e476 2618}
6b80053d 2619
ca7a79ad 2620int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 2621{
810191ff 2622 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
ce9adaa5 2623 int ret;
ca7a79ad 2624 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
d397712b 2625 if (ret == 0)
b4ce94de 2626 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
ce9adaa5 2627 return ret;
6b80053d 2628}
0da5468f 2629
4bef0848
CM		 2630int btree_lock_page_hook(struct page *page)
2631{
2632 struct inode *inode = page->mapping->host;
b9473439 2633 struct btrfs_root *root = BTRFS_I(inode)->root;
4bef0848
CM		 2634 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2635 struct extent_buffer *eb;
2636 unsigned long len;
2637 u64 bytenr = page_offset(page);
2638
2639 if (page->private == EXTENT_PAGE_PRIVATE)
2640 goto out;
2641
2642 len = page->private >> 2;
f09d1f60 2643 eb = find_extent_buffer(io_tree, bytenr, len);
4bef0848
CM		 2644 if (!eb)
2645 goto out;
2646
2647 btrfs_tree_lock(eb);
4bef0848 2648 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
b9473439
CM		 2649
2650 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
2651 spin_lock(&root->fs_info->delalloc_lock);
2652 if (root->fs_info->dirty_metadata_bytes >= eb->len)
2653 root->fs_info->dirty_metadata_bytes -= eb->len;
2654 else
2655 WARN_ON(1);
2656 spin_unlock(&root->fs_info->delalloc_lock);
2657 }
2658
4bef0848
CM		 2659 btrfs_tree_unlock(eb);
2660 free_extent_buffer(eb);
2661out:
2662 lock_page(page);
2663 return 0;
2664}
2665
acce952b 2666static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
2667 int read_only)
2668{
2669 if (read_only)
2670 return;
2671
2672 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)
2673 printk(KERN_WARNING "warning: mount fs with errors, "
2674 "running btrfsck is recommended\n");
2675}
2676
2677int btrfs_error_commit_super(struct btrfs_root *root)
2678{
2679 int ret;
2680
2681 mutex_lock(&root->fs_info->cleaner_mutex);
2682 btrfs_run_delayed_iputs(root);
2683 mutex_unlock(&root->fs_info->cleaner_mutex);
2684
2685 down_write(&root->fs_info->cleanup_work_sem);
2686 up_write(&root->fs_info->cleanup_work_sem);
2687
2688 /* cleanup FS via transaction */
2689 btrfs_cleanup_transaction(root);
2690
2691 ret = write_ctree_super(NULL, root, 0);
2692
2693 return ret;
2694}
2695
2696static int btrfs_destroy_ordered_operations(struct btrfs_root *root)
2697{
2698 struct btrfs_inode *btrfs_inode;
2699 struct list_head splice;
2700
2701 INIT_LIST_HEAD(&splice);
2702
2703 mutex_lock(&root->fs_info->ordered_operations_mutex);
2704 spin_lock(&root->fs_info->ordered_extent_lock);
2705
2706 list_splice_init(&root->fs_info->ordered_operations, &splice);
2707 while (!list_empty(&splice)) {
2708 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
2709 ordered_operations);
2710
2711 list_del_init(&btrfs_inode->ordered_operations);
2712
2713 btrfs_invalidate_inodes(btrfs_inode->root);
2714 }
2715
2716 spin_unlock(&root->fs_info->ordered_extent_lock);
2717 mutex_unlock(&root->fs_info->ordered_operations_mutex);
2718
2719 return 0;
2720}
2721
2722static int btrfs_destroy_ordered_extents(struct btrfs_root *root)
2723{
2724 struct list_head splice;
2725 struct btrfs_ordered_extent *ordered;
2726 struct inode *inode;
2727
2728 INIT_LIST_HEAD(&splice);
2729
2730 spin_lock(&root->fs_info->ordered_extent_lock);
2731
2732 list_splice_init(&root->fs_info->ordered_extents, &splice);
2733 while (!list_empty(&splice)) {
2734 ordered = list_entry(splice.next, struct btrfs_ordered_extent,
2735 root_extent_list);
2736
2737 list_del_init(&ordered->root_extent_list);
2738 atomic_inc(&ordered->refs);
2739
2740 /* the inode may be getting freed (in sys_unlink path). */
2741 inode = igrab(ordered->inode);
2742
2743 spin_unlock(&root->fs_info->ordered_extent_lock);
2744 if (inode)
2745 iput(inode);
2746
2747 atomic_set(&ordered->refs, 1);
2748 btrfs_put_ordered_extent(ordered);
2749
2750 spin_lock(&root->fs_info->ordered_extent_lock);
2751 }
2752
2753 spin_unlock(&root->fs_info->ordered_extent_lock);
2754
2755 return 0;
2756}
2757
2758static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
2759 struct btrfs_root *root)
2760{
2761 struct rb_node *node;
2762 struct btrfs_delayed_ref_root *delayed_refs;
2763 struct btrfs_delayed_ref_node *ref;
2764 int ret = 0;
2765
2766 delayed_refs = &trans->delayed_refs;
2767
2768 spin_lock(&delayed_refs->lock);
2769 if (delayed_refs->num_entries == 0) {
cfece4db 2770 spin_unlock(&delayed_refs->lock);
acce952b 2771 printk(KERN_INFO "delayed_refs has NO entry\n");
2772 return ret;
2773 }
2774
2775 node = rb_first(&delayed_refs->root);
2776 while (node) {
2777 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2778 node = rb_next(node);
2779
2780 ref->in_tree = 0;
2781 rb_erase(&ref->rb_node, &delayed_refs->root);
2782 delayed_refs->num_entries--;
2783
2784 atomic_set(&ref->refs, 1);
2785 if (btrfs_delayed_ref_is_head(ref)) {
2786 struct btrfs_delayed_ref_head *head;
2787
2788 head = btrfs_delayed_node_to_head(ref);
2789 mutex_lock(&head->mutex);
2790 kfree(head->extent_op);
2791 delayed_refs->num_heads--;
2792 if (list_empty(&head->cluster))
2793 delayed_refs->num_heads_ready--;
2794 list_del_init(&head->cluster);
2795 mutex_unlock(&head->mutex);
2796 }
2797
2798 spin_unlock(&delayed_refs->lock);
2799 btrfs_put_delayed_ref(ref);
2800
2801 cond_resched();
2802 spin_lock(&delayed_refs->lock);
2803 }
2804
2805 spin_unlock(&delayed_refs->lock);
2806
2807 return ret;
2808}
2809
2810static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t)
2811{
2812 struct btrfs_pending_snapshot *snapshot;
2813 struct list_head splice;
2814
2815 INIT_LIST_HEAD(&splice);
2816
2817 list_splice_init(&t->pending_snapshots, &splice);
2818
2819 while (!list_empty(&splice)) {
2820 snapshot = list_entry(splice.next,
2821 struct btrfs_pending_snapshot,
2822 list);
2823
2824 list_del_init(&snapshot->list);
2825
2826 kfree(snapshot);
2827 }
2828
2829 return 0;
2830}
2831
2832static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
2833{
2834 struct btrfs_inode *btrfs_inode;
2835 struct list_head splice;
2836
2837 INIT_LIST_HEAD(&splice);
2838
2839 list_splice_init(&root->fs_info->delalloc_inodes, &splice);
2840
2841 spin_lock(&root->fs_info->delalloc_lock);
2842
2843 while (!list_empty(&splice)) {
2844 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
2845 delalloc_inodes);
2846
2847 list_del_init(&btrfs_inode->delalloc_inodes);
2848
2849 btrfs_invalidate_inodes(btrfs_inode->root);
2850 }
2851
2852 spin_unlock(&root->fs_info->delalloc_lock);
2853
2854 return 0;
2855}
2856
2857static int btrfs_destroy_marked_extents(struct btrfs_root *root,
2858 struct extent_io_tree *dirty_pages,
2859 int mark)
2860{
2861 int ret;
2862 struct page *page;
2863 struct inode *btree_inode = root->fs_info->btree_inode;
2864 struct extent_buffer *eb;
2865 u64 start = 0;
2866 u64 end;
2867 u64 offset;
2868 unsigned long index;
2869
2870 while (1) {
2871 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
2872 mark);
2873 if (ret)
2874 break;
2875
2876 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
2877 while (start <= end) {
2878 index = start >> PAGE_CACHE_SHIFT;
2879 start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
2880 page = find_get_page(btree_inode->i_mapping, index);
2881 if (!page)
2882 continue;
2883 offset = page_offset(page);
2884
2885 spin_lock(&dirty_pages->buffer_lock);
2886 eb = radix_tree_lookup(
2887 &(&BTRFS_I(page->mapping->host)->io_tree)->buffer,
2888 offset >> PAGE_CACHE_SHIFT);
2889 spin_unlock(&dirty_pages->buffer_lock);
2890 if (eb) {
2891 ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY,
2892 &eb->bflags);
2893 atomic_set(&eb->refs, 1);
2894 }
2895 if (PageWriteback(page))
2896 end_page_writeback(page);
2897
2898 lock_page(page);
2899 if (PageDirty(page)) {
2900 clear_page_dirty_for_io(page);
2901 spin_lock_irq(&page->mapping->tree_lock);
2902 radix_tree_tag_clear(&page->mapping->page_tree,
2903 page_index(page),
2904 PAGECACHE_TAG_DIRTY);
2905 spin_unlock_irq(&page->mapping->tree_lock);
2906 }
2907
2908 page->mapping->a_ops->invalidatepage(page, 0);
2909 unlock_page(page);
2910 }
2911 }
2912
2913 return ret;
2914}
2915
2916static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
2917 struct extent_io_tree *pinned_extents)
2918{
2919 struct extent_io_tree *unpin;
2920 u64 start;
2921 u64 end;
2922 int ret;
2923
2924 unpin = pinned_extents;
2925 while (1) {
2926 ret = find_first_extent_bit(unpin, 0, &start, &end,
2927 EXTENT_DIRTY);
2928 if (ret)
2929 break;
2930
2931 /* opt_discard */
5378e607
LD		 2932 if (btrfs_test_opt(root, DISCARD))
2933 ret = btrfs_error_discard_extent(root, start,
2934 end + 1 - start,
2935 NULL);
acce952b 2936
2937 clear_extent_dirty(unpin, start, end, GFP_NOFS);
2938 btrfs_error_unpin_extent_range(root, start, end);
2939 cond_resched();
2940 }
2941
2942 return 0;
2943}
2944
2945static int btrfs_cleanup_transaction(struct btrfs_root *root)
2946{
2947 struct btrfs_transaction *t;
2948 LIST_HEAD(list);
2949
2950 WARN_ON(1);
2951
2952 mutex_lock(&root->fs_info->trans_mutex);
2953 mutex_lock(&root->fs_info->transaction_kthread_mutex);
2954
2955 list_splice_init(&root->fs_info->trans_list, &list);
2956 while (!list_empty(&list)) {
2957 t = list_entry(list.next, struct btrfs_transaction, list);
2958 if (!t)
2959 break;
2960
2961 btrfs_destroy_ordered_operations(root);
2962
2963 btrfs_destroy_ordered_extents(root);
2964
2965 btrfs_destroy_delayed_refs(t, root);
2966
2967 btrfs_block_rsv_release(root,
2968 &root->fs_info->trans_block_rsv,
2969 t->dirty_pages.dirty_bytes);
2970
2971 /* FIXME: cleanup wait for commit */
2972 t->in_commit = 1;
2973 t->blocked = 1;
2974 if (waitqueue_active(&root->fs_info->transaction_blocked_wait))
2975 wake_up(&root->fs_info->transaction_blocked_wait);
2976
2977 t->blocked = 0;
2978 if (waitqueue_active(&root->fs_info->transaction_wait))
2979 wake_up(&root->fs_info->transaction_wait);
2980 mutex_unlock(&root->fs_info->trans_mutex);
2981
2982 mutex_lock(&root->fs_info->trans_mutex);
2983 t->commit_done = 1;
2984 if (waitqueue_active(&t->commit_wait))
2985 wake_up(&t->commit_wait);
2986 mutex_unlock(&root->fs_info->trans_mutex);
2987
2988 mutex_lock(&root->fs_info->trans_mutex);
2989
2990 btrfs_destroy_pending_snapshots(t);
2991
2992 btrfs_destroy_delalloc_inodes(root);
2993
2994 spin_lock(&root->fs_info->new_trans_lock);
2995 root->fs_info->running_transaction = NULL;
2996 spin_unlock(&root->fs_info->new_trans_lock);
2997
2998 btrfs_destroy_marked_extents(root, &t->dirty_pages,
2999 EXTENT_DIRTY);
3000
3001 btrfs_destroy_pinned_extent(root,
3002 root->fs_info->pinned_extents);
3003
13c5a93e 3004 atomic_set(&t->use_count, 0);
acce952b 3005 list_del_init(&t->list);
3006 memset(t, 0, sizeof(*t));
3007 kmem_cache_free(btrfs_transaction_cachep, t);
3008 }
3009
3010 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
3011 mutex_unlock(&root->fs_info->trans_mutex);
3012
3013 return 0;
3014}
3015
d1310b2e 3016static struct extent_io_ops btree_extent_io_ops = {
4bef0848 3017 .write_cache_pages_lock_hook = btree_lock_page_hook,
ce9adaa5 3018 .readpage_end_io_hook = btree_readpage_end_io_hook,
0b86a832 3019 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 3020 /* note we're sharing with inode.c for the merge bio hook */
3021 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 3022};
Linux kernel - Deliverables
This page took 0.352993 seconds and 5 git commands to generate.