projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: drop unnecessary device lock
[linux-2.6-block.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
CH		 139static struct extent_map *btree_get_extent(struct inode *inode,
140 struct page *page, size_t page_offset, u64 start, u64 len,
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
5f39d397
CM		 157 em = alloc_extent_map(GFP_NOFS);
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
19c00ddc 383 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
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
ce9adaa5 528 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
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
4854ddd0
CM		 653int btrfs_congested_async(struct btrfs_fs_info *info, int iodone)
654{
b64a2851
CM		 655 return atomic_read(&info->nr_async_bios) >
656 btrfs_async_submit_limit(info);
0986fe9e
CM		 657}
658
4a69a410
CM		 659static void run_one_async_start(struct btrfs_work *work)
660{
4a69a410
CM		 661 struct async_submit_bio *async;
662
663 async = container_of(work, struct async_submit_bio, work);
4a69a410 664 async->submit_bio_start(async->inode, async->rw, async->bio,
eaf25d93
CM		 665 async->mirror_num, async->bio_flags,
666 async->bio_offset);
4a69a410
CM		 667}
668
669static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 670{
671 struct btrfs_fs_info *fs_info;
672 struct async_submit_bio *async;
4854ddd0 673 int limit;
8b712842
CM		 674
675 async = container_of(work, struct async_submit_bio, work);
676 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 677
b64a2851 678 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 679 limit = limit * 2 / 3;
680
8b712842 681 atomic_dec(&fs_info->nr_async_submits);
0986fe9e 682
b64a2851
CM		 683 if (atomic_read(&fs_info->nr_async_submits) < limit &&
684 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 685 wake_up(&fs_info->async_submit_wait);
686
4a69a410 687 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 688 async->mirror_num, async->bio_flags,
689 async->bio_offset);
4a69a410
CM		 690}
691
692static void run_one_async_free(struct btrfs_work *work)
693{
694 struct async_submit_bio *async;
695
696 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 697 kfree(async);
698}
699
44b8bd7e
CM		 700int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
701 int rw, struct bio *bio, int mirror_num,
c8b97818 702 unsigned long bio_flags,
eaf25d93 703 u64 bio_offset,
4a69a410
CM		 704 extent_submit_bio_hook_t *submit_bio_start,
705 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 706{
707 struct async_submit_bio *async;
708
709 async = kmalloc(sizeof(*async), GFP_NOFS);
710 if (!async)
711 return -ENOMEM;
712
713 async->inode = inode;
714 async->rw = rw;
715 async->bio = bio;
716 async->mirror_num = mirror_num;
4a69a410
CM		 717 async->submit_bio_start = submit_bio_start;
718 async->submit_bio_done = submit_bio_done;
719
720 async->work.func = run_one_async_start;
721 async->work.ordered_func = run_one_async_done;
722 async->work.ordered_free = run_one_async_free;
723
8b712842 724 async->work.flags = 0;
c8b97818 725 async->bio_flags = bio_flags;
eaf25d93 726 async->bio_offset = bio_offset;
8c8bee1d 727
cb03c743 728 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 729
7b6d91da 730 if (rw & REQ_SYNC)
d313d7a3
CM		 731 btrfs_set_work_high_prio(&async->work);
732
8b712842 733 btrfs_queue_worker(&fs_info->workers, &async->work);
9473f16c 734
d397712b 735 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 736 atomic_read(&fs_info->nr_async_submits)) {
737 wait_event(fs_info->async_submit_wait,
738 (atomic_read(&fs_info->nr_async_submits) == 0));
739 }
740
44b8bd7e
CM		 741 return 0;
742}
743
ce3ed71a
CM		 744static int btree_csum_one_bio(struct bio *bio)
745{
746 struct bio_vec *bvec = bio->bi_io_vec;
747 int bio_index = 0;
748 struct btrfs_root *root;
749
750 WARN_ON(bio->bi_vcnt <= 0);
d397712b 751 while (bio_index < bio->bi_vcnt) {
ce3ed71a
CM		 752 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
753 csum_dirty_buffer(root, bvec->bv_page);
754 bio_index++;
755 bvec++;
756 }
757 return 0;
758}
759
4a69a410
CM		 760static int __btree_submit_bio_start(struct inode *inode, int rw,
761 struct bio *bio, int mirror_num,
eaf25d93
CM		 762 unsigned long bio_flags,
763 u64 bio_offset)
22c59948 764{
8b712842
CM		 765 /*
766 * when we're called for a write, we're already in the async
5443be45 767 * submission context. Just jump into btrfs_map_bio
8b712842 768 */
4a69a410
CM		 769 btree_csum_one_bio(bio);
770 return 0;
771}
22c59948 772
4a69a410 773static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 774 int mirror_num, unsigned long bio_flags,
775 u64 bio_offset)
4a69a410 776{
8b712842 777 /*
4a69a410
CM		 778 * when we're called for a write, we're already in the async
779 * submission context. Just jump into btrfs_map_bio
8b712842 780 */
8b712842 781 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
0b86a832
CM		 782}
783
44b8bd7e 784static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 785 int mirror_num, unsigned long bio_flags,
786 u64 bio_offset)
44b8bd7e 787{
cad321ad
CM		 788 int ret;
789
790 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
791 bio, 1);
792 BUG_ON(ret);
793
7b6d91da 794 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 795 /*
796 * called for a read, do the setup so that checksum validation
797 * can happen in the async kernel threads
798 */
4a69a410 799 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
6f3577bd 800 mirror_num, 0);
44b8bd7e 801 }
d313d7a3 802
cad321ad
CM		 803 /*
804 * kthread helpers are used to submit writes so that checksumming
805 * can happen in parallel across all CPUs
806 */
44b8bd7e 807 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
c8b97818 808 inode, rw, bio, mirror_num, 0,
eaf25d93 809 bio_offset,
4a69a410
CM		 810 __btree_submit_bio_start,
811 __btree_submit_bio_done);
44b8bd7e
CM		 812}
813
3dd1462e 814#ifdef CONFIG_MIGRATION
784b4e29
CM		 815static int btree_migratepage(struct address_space *mapping,
816 struct page *newpage, struct page *page)
817{
818 /*
819 * we can't safely write a btree page from here,
820 * we haven't done the locking hook
821 */
822 if (PageDirty(page))
823 return -EAGAIN;
824 /*
825 * Buffers may be managed in a filesystem specific way.
826 * We must have no buffers or drop them.
827 */
828 if (page_has_private(page) &&
829 !try_to_release_page(page, GFP_KERNEL))
830 return -EAGAIN;
784b4e29
CM		 831 return migrate_page(mapping, newpage, page);
832}
3dd1462e 833#endif
784b4e29 834
0da5468f
CM		 835static int btree_writepage(struct page *page, struct writeback_control *wbc)
836{
d1310b2e 837 struct extent_io_tree *tree;
b9473439
CM		 838 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
839 struct extent_buffer *eb;
840 int was_dirty;
841
d1310b2e 842 tree = &BTRFS_I(page->mapping->host)->io_tree;
b9473439
CM		 843 if (!(current->flags & PF_MEMALLOC)) {
844 return extent_write_full_page(tree, page,
845 btree_get_extent, wbc);
846 }
5443be45 847
b9473439 848 redirty_page_for_writepage(wbc, page);
784b4e29 849 eb = btrfs_find_tree_block(root, page_offset(page), PAGE_CACHE_SIZE);
b9473439
CM		 850 WARN_ON(!eb);
851
852 was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
853 if (!was_dirty) {
854 spin_lock(&root->fs_info->delalloc_lock);
855 root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE;
856 spin_unlock(&root->fs_info->delalloc_lock);
5443be45 857 }
b9473439
CM		 858 free_extent_buffer(eb);
859
860 unlock_page(page);
861 return 0;
5f39d397 862}
0da5468f
CM		 863
864static int btree_writepages(struct address_space *mapping,
865 struct writeback_control *wbc)
866{
d1310b2e
CM		 867 struct extent_io_tree *tree;
868 tree = &BTRFS_I(mapping->host)->io_tree;
d8d5f3e1 869 if (wbc->sync_mode == WB_SYNC_NONE) {
b9473439 870 struct btrfs_root *root = BTRFS_I(mapping->host)->root;
793955bc 871 u64 num_dirty;
24ab9cd8 872 unsigned long thresh = 32 * 1024 * 1024;
448d640b
CM		 873
874 if (wbc->for_kupdate)
875 return 0;
876
b9473439
CM		 877 /* this is a bit racy, but that's ok */
878 num_dirty = root->fs_info->dirty_metadata_bytes;
d397712b 879 if (num_dirty < thresh)
793955bc 880 return 0;
793955bc 881 }
0da5468f
CM		 882 return extent_writepages(tree, mapping, btree_get_extent, wbc);
883}
884
b2950863 885static int btree_readpage(struct file *file, struct page *page)
5f39d397 886{
d1310b2e
CM		 887 struct extent_io_tree *tree;
888 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 889 return extent_read_full_page(tree, page, btree_get_extent);
890}
22b0ebda 891
70dec807 892static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 893{
d1310b2e
CM		 894 struct extent_io_tree *tree;
895 struct extent_map_tree *map;
5f39d397 896 int ret;
d98237b3 897
98509cfc 898 if (PageWriteback(page) || PageDirty(page))
d397712b 899 return 0;
98509cfc 900
d1310b2e
CM		 901 tree = &BTRFS_I(page->mapping->host)->io_tree;
902 map = &BTRFS_I(page->mapping->host)->extent_tree;
6af118ce 903
7b13b7b1 904 ret = try_release_extent_state(map, tree, page, gfp_flags);
d397712b 905 if (!ret)
6af118ce 906 return 0;
6af118ce
CM		 907
908 ret = try_release_extent_buffer(tree, page);
5f39d397
CM		 909 if (ret == 1) {
910 ClearPagePrivate(page);
911 set_page_private(page, 0);
912 page_cache_release(page);
913 }
6af118ce 914
d98237b3
CM		 915 return ret;
916}
917
5f39d397 918static void btree_invalidatepage(struct page *page, unsigned long offset)
d98237b3 919{
d1310b2e
CM		 920 struct extent_io_tree *tree;
921 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 922 extent_invalidatepage(tree, page, offset);
923 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 924 if (PagePrivate(page)) {
d397712b
CM		 925 printk(KERN_WARNING "btrfs warning page private not zero "
926 "on page %llu\n", (unsigned long long)page_offset(page));
9ad6b7bc
CM		 927 ClearPagePrivate(page);
928 set_page_private(page, 0);
929 page_cache_release(page);
930 }
d98237b3
CM		 931}
932
7f09410b 933static const struct address_space_operations btree_aops = {
d98237b3
CM		 934 .readpage = btree_readpage,
935 .writepage = btree_writepage,
0da5468f 936 .writepages = btree_writepages,
5f39d397
CM		 937 .releasepage = btree_releasepage,
938 .invalidatepage = btree_invalidatepage,
5a92bc88 939#ifdef CONFIG_MIGRATION
784b4e29 940 .migratepage = btree_migratepage,
5a92bc88 941#endif
d98237b3
CM		 942};
943
ca7a79ad
CM		 944int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
945 u64 parent_transid)
090d1875 946{
5f39d397
CM		 947 struct extent_buffer *buf = NULL;
948 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 949 int ret = 0;
090d1875 950
db94535d 951 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 952 if (!buf)
090d1875 953 return 0;
d1310b2e 954 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
f188591e 955 buf, 0, 0, btree_get_extent, 0);
5f39d397 956 free_extent_buffer(buf);
de428b63 957 return ret;
090d1875
CM		 958}
959
0999df54
CM		 960struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
961 u64 bytenr, u32 blocksize)
962{
963 struct inode *btree_inode = root->fs_info->btree_inode;
964 struct extent_buffer *eb;
965 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
966 bytenr, blocksize, GFP_NOFS);
967 return eb;
968}
969
970struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
971 u64 bytenr, u32 blocksize)
972{
973 struct inode *btree_inode = root->fs_info->btree_inode;
974 struct extent_buffer *eb;
975
976 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
977 bytenr, blocksize, NULL, GFP_NOFS);
978 return eb;
979}
980
981
e02119d5
CM		 982int btrfs_write_tree_block(struct extent_buffer *buf)
983{
8aa38c31
CH		 984 return filemap_fdatawrite_range(buf->first_page->mapping, buf->start,
985 buf->start + buf->len - 1);
e02119d5
CM		 986}
987
988int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
989{
8aa38c31
CH		 990 return filemap_fdatawait_range(buf->first_page->mapping,
991 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 992}
993
0999df54 994struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 995 u32 blocksize, u64 parent_transid)
0999df54
CM		 996{
997 struct extent_buffer *buf = NULL;
0999df54
CM		 998 int ret;
999
0999df54
CM		 1000 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1001 if (!buf)
1002 return NULL;
0999df54 1003
ca7a79ad 1004 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
ce9adaa5 1005
d397712b 1006 if (ret == 0)
b4ce94de 1007 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
5f39d397 1008 return buf;
ce9adaa5 1009
eb60ceac
CM		 1010}
1011
e089f05c 1012int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
5f39d397 1013 struct extent_buffer *buf)
ed2ff2cb 1014{
5f39d397 1015 struct inode *btree_inode = root->fs_info->btree_inode;
55c69072 1016 if (btrfs_header_generation(buf) ==
925baedd 1017 root->fs_info->running_transaction->transid) {
b9447ef8 1018 btrfs_assert_tree_locked(buf);
b4ce94de 1019
b9473439
CM		 1020 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
1021 spin_lock(&root->fs_info->delalloc_lock);
1022 if (root->fs_info->dirty_metadata_bytes >= buf->len)
1023 root->fs_info->dirty_metadata_bytes -= buf->len;
1024 else
1025 WARN_ON(1);
1026 spin_unlock(&root->fs_info->delalloc_lock);
1027 }
b4ce94de 1028
b9473439
CM		 1029 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1030 btrfs_set_lock_blocking(buf);
d1310b2e 1031 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
55c69072 1032 buf);
925baedd 1033 }
5f39d397
CM		 1034 return 0;
1035}
1036
db94535d 1037static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
87ee04eb 1038 u32 stripesize, struct btrfs_root *root,
9f5fae2f 1039 struct btrfs_fs_info *fs_info,
e20d96d6 1040 u64 objectid)
d97e63b6 1041{
cfaa7295 1042 root->node = NULL;
a28ec197 1043 root->commit_root = NULL;
db94535d
CM		 1044 root->sectorsize = sectorsize;
1045 root->nodesize = nodesize;
1046 root->leafsize = leafsize;
87ee04eb 1047 root->stripesize = stripesize;
123abc88 1048 root->ref_cows = 0;
0b86a832 1049 root->track_dirty = 0;
c71bf099 1050 root->in_radix = 0;
d68fc57b
YZ		 1051 root->orphan_item_inserted = 0;
1052 root->orphan_cleanup_state = 0;
0b86a832 1053
9f5fae2f 1054 root->fs_info = fs_info;
0f7d52f4
CM		 1055 root->objectid = objectid;
1056 root->last_trans = 0;
13a8a7c8 1057 root->highest_objectid = 0;
58176a96 1058 root->name = NULL;
4313b399 1059 root->in_sysfs = 0;
6bef4d31 1060 root->inode_tree = RB_ROOT;
f0486c68 1061 root->block_rsv = NULL;
d68fc57b 1062 root->orphan_block_rsv = NULL;
0b86a832
CM		 1063
1064 INIT_LIST_HEAD(&root->dirty_list);
7b128766 1065 INIT_LIST_HEAD(&root->orphan_list);
5d4f98a2 1066 INIT_LIST_HEAD(&root->root_list);
d68fc57b 1067 spin_lock_init(&root->orphan_lock);
5d4f98a2 1068 spin_lock_init(&root->inode_lock);
f0486c68 1069 spin_lock_init(&root->accounting_lock);
a2135011 1070 mutex_init(&root->objectid_mutex);
e02119d5 1071 mutex_init(&root->log_mutex);
7237f183
YZ		 1072 init_waitqueue_head(&root->log_writer_wait);
1073 init_waitqueue_head(&root->log_commit_wait[0]);
1074 init_waitqueue_head(&root->log_commit_wait[1]);
1075 atomic_set(&root->log_commit[0], 0);
1076 atomic_set(&root->log_commit[1], 0);
1077 atomic_set(&root->log_writers, 0);
1078 root->log_batch = 0;
1079 root->log_transid = 0;
257c62e1 1080 root->last_log_commit = 0;
d0c803c4
CM		 1081 extent_io_tree_init(&root->dirty_log_pages,
1082 fs_info->btree_inode->i_mapping, GFP_NOFS);
017e5369 1083
3768f368
CM		 1084 memset(&root->root_key, 0, sizeof(root->root_key));
1085 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1086 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1087 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
3f157a2f 1088 root->defrag_trans_start = fs_info->generation;
58176a96 1089 init_completion(&root->kobj_unregister);
6702ed49 1090 root->defrag_running = 0;
4d775673 1091 root->root_key.objectid = objectid;
3394e160
CM		 1092 root->anon_super.s_root = NULL;
1093 root->anon_super.s_dev = 0;
1094 INIT_LIST_HEAD(&root->anon_super.s_list);
1095 INIT_LIST_HEAD(&root->anon_super.s_instances);
1096 init_rwsem(&root->anon_super.s_umount);
1097
3768f368
CM		 1098 return 0;
1099}
1100
db94535d 1101static int find_and_setup_root(struct btrfs_root *tree_root,
9f5fae2f
CM		 1102 struct btrfs_fs_info *fs_info,
1103 u64 objectid,
e20d96d6 1104 struct btrfs_root *root)
3768f368
CM		 1105{
1106 int ret;
db94535d 1107 u32 blocksize;
84234f3a 1108 u64 generation;
3768f368 1109
db94535d 1110 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 1111 tree_root->sectorsize, tree_root->stripesize,
1112 root, fs_info, objectid);
3768f368
CM		 1113 ret = btrfs_find_last_root(tree_root, objectid,
1114 &root->root_item, &root->root_key);
4df27c4d
YZ		 1115 if (ret > 0)
1116 return -ENOENT;
3768f368
CM		 1117 BUG_ON(ret);
1118
84234f3a 1119 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1120 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1121 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1122 blocksize, generation);
68433b73
CM		 1123 if (!root->node || !btrfs_buffer_uptodate(root->node, generation)) {
1124 free_extent_buffer(root->node);
1125 return -EIO;
1126 }
4df27c4d 1127 root->commit_root = btrfs_root_node(root);
d97e63b6
CM		 1128 return 0;
1129}
1130
7237f183
YZ		 1131static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1132 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1133{
1134 struct btrfs_root *root;
1135 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1136 struct extent_buffer *leaf;
e02119d5
CM		 1137
1138 root = kzalloc(sizeof(*root), GFP_NOFS);
1139 if (!root)
7237f183 1140 return ERR_PTR(-ENOMEM);
e02119d5
CM		 1141
1142 __setup_root(tree_root->nodesize, tree_root->leafsize,
1143 tree_root->sectorsize, tree_root->stripesize,
1144 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1145
1146 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1147 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1148 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
7237f183
YZ		 1149 /*
1150 * log trees do not get reference counted because they go away
1151 * before a real commit is actually done. They do store pointers
1152 * to file data extents, and those reference counts still get
1153 * updated (along with back refs to the log tree).
1154 */
e02119d5
CM		 1155 root->ref_cows = 0;
1156
5d4f98a2
YZ		 1157 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
1158 BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0);
7237f183
YZ		 1159 if (IS_ERR(leaf)) {
1160 kfree(root);
1161 return ERR_CAST(leaf);
1162 }
e02119d5 1163
5d4f98a2
YZ		 1164 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1165 btrfs_set_header_bytenr(leaf, leaf->start);
1166 btrfs_set_header_generation(leaf, trans->transid);
1167 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1168 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1169 root->node = leaf;
e02119d5
CM		 1170
1171 write_extent_buffer(root->node, root->fs_info->fsid,
1172 (unsigned long)btrfs_header_fsid(root->node),
1173 BTRFS_FSID_SIZE);
1174 btrfs_mark_buffer_dirty(root->node);
1175 btrfs_tree_unlock(root->node);
7237f183
YZ		 1176 return root;
1177}
1178
1179int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1180 struct btrfs_fs_info *fs_info)
1181{
1182 struct btrfs_root *log_root;
1183
1184 log_root = alloc_log_tree(trans, fs_info);
1185 if (IS_ERR(log_root))
1186 return PTR_ERR(log_root);
1187 WARN_ON(fs_info->log_root_tree);
1188 fs_info->log_root_tree = log_root;
1189 return 0;
1190}
1191
1192int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1193 struct btrfs_root *root)
1194{
1195 struct btrfs_root *log_root;
1196 struct btrfs_inode_item *inode_item;
1197
1198 log_root = alloc_log_tree(trans, root->fs_info);
1199 if (IS_ERR(log_root))
1200 return PTR_ERR(log_root);
1201
1202 log_root->last_trans = trans->transid;
1203 log_root->root_key.offset = root->root_key.objectid;
1204
1205 inode_item = &log_root->root_item.inode;
1206 inode_item->generation = cpu_to_le64(1);
1207 inode_item->size = cpu_to_le64(3);
1208 inode_item->nlink = cpu_to_le32(1);
1209 inode_item->nbytes = cpu_to_le64(root->leafsize);
1210 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1211
5d4f98a2 1212 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1213
1214 WARN_ON(root->log_root);
1215 root->log_root = log_root;
1216 root->log_transid = 0;
257c62e1 1217 root->last_log_commit = 0;
e02119d5
CM		 1218 return 0;
1219}
1220
1221struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
1222 struct btrfs_key *location)
1223{
1224 struct btrfs_root *root;
1225 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1226 struct btrfs_path *path;
5f39d397 1227 struct extent_buffer *l;
84234f3a 1228 u64 generation;
db94535d 1229 u32 blocksize;
0f7d52f4
CM		 1230 int ret = 0;
1231
5eda7b5e 1232 root = kzalloc(sizeof(*root), GFP_NOFS);
0cf6c620 1233 if (!root)
0f7d52f4 1234 return ERR_PTR(-ENOMEM);
0f7d52f4 1235 if (location->offset == (u64)-1) {
db94535d 1236 ret = find_and_setup_root(tree_root, fs_info,
0f7d52f4
CM		 1237 location->objectid, root);
1238 if (ret) {
0f7d52f4
CM		 1239 kfree(root);
1240 return ERR_PTR(ret);
1241 }
13a8a7c8 1242 goto out;
0f7d52f4
CM		 1243 }
1244
db94535d 1245 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 1246 tree_root->sectorsize, tree_root->stripesize,
1247 root, fs_info, location->objectid);
0f7d52f4
CM		 1248
1249 path = btrfs_alloc_path();
db5b493a
TI		 1250 if (!path) {
1251 kfree(root);
1252 return ERR_PTR(-ENOMEM);
1253 }
0f7d52f4 1254 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
13a8a7c8
YZ		 1255 if (ret == 0) {
1256 l = path->nodes[0];
1257 read_extent_buffer(l, &root->root_item,
1258 btrfs_item_ptr_offset(l, path->slots[0]),
1259 sizeof(root->root_item));
1260 memcpy(&root->root_key, location, sizeof(*location));
0f7d52f4 1261 }
0f7d52f4
CM		 1262 btrfs_free_path(path);
1263 if (ret) {
5e540f77 1264 kfree(root);
13a8a7c8
YZ		 1265 if (ret > 0)
1266 ret = -ENOENT;
0f7d52f4
CM		 1267 return ERR_PTR(ret);
1268 }
13a8a7c8 1269
84234f3a 1270 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1271 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1272 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1273 blocksize, generation);
5d4f98a2 1274 root->commit_root = btrfs_root_node(root);
0f7d52f4 1275 BUG_ON(!root->node);
13a8a7c8 1276out:
08fe4db1 1277 if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
e02119d5 1278 root->ref_cows = 1;
08fe4db1
LZ		 1279 btrfs_check_and_init_root_item(&root->root_item);
1280 }
13a8a7c8 1281
5eda7b5e
CM		 1282 return root;
1283}
1284
dc17ff8f
CM		 1285struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1286 u64 root_objectid)
1287{
1288 struct btrfs_root *root;
1289
1290 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
1291 return fs_info->tree_root;
1292 if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
1293 return fs_info->extent_root;
1294
1295 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1296 (unsigned long)root_objectid);
1297 return root;
1298}
1299
edbd8d4e
CM		 1300struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
1301 struct btrfs_key *location)
5eda7b5e
CM		 1302{
1303 struct btrfs_root *root;
1304 int ret;
1305
edbd8d4e
CM		 1306 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1307 return fs_info->tree_root;
1308 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1309 return fs_info->extent_root;
8f18cf13
CM		 1310 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1311 return fs_info->chunk_root;
1312 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1313 return fs_info->dev_root;
0403e47e
YZ		 1314 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1315 return fs_info->csum_root;
4df27c4d
YZ		 1316again:
1317 spin_lock(&fs_info->fs_roots_radix_lock);
5eda7b5e
CM		 1318 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1319 (unsigned long)location->objectid);
4df27c4d 1320 spin_unlock(&fs_info->fs_roots_radix_lock);
5eda7b5e
CM		 1321 if (root)
1322 return root;
1323
e02119d5 1324 root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
5eda7b5e
CM		 1325 if (IS_ERR(root))
1326 return root;
3394e160
CM		 1327
1328 set_anon_super(&root->anon_super, NULL);
1329
d68fc57b
YZ		 1330 if (btrfs_root_refs(&root->root_item) == 0) {
1331 ret = -ENOENT;
1332 goto fail;
1333 }
1334
1335 ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
1336 if (ret < 0)
1337 goto fail;
1338 if (ret == 0)
1339 root->orphan_item_inserted = 1;
1340
4df27c4d
YZ		 1341 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1342 if (ret)
1343 goto fail;
1344
1345 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 1346 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1347 (unsigned long)root->root_key.objectid,
0f7d52f4 1348 root);
d68fc57b 1349 if (ret == 0)
4df27c4d 1350 root->in_radix = 1;
d68fc57b 1351
4df27c4d
YZ		 1352 spin_unlock(&fs_info->fs_roots_radix_lock);
1353 radix_tree_preload_end();
0f7d52f4 1354 if (ret) {
4df27c4d
YZ		 1355 if (ret == -EEXIST) {
1356 free_fs_root(root);
1357 goto again;
1358 }
1359 goto fail;
0f7d52f4 1360 }
4df27c4d
YZ		 1361
1362 ret = btrfs_find_dead_roots(fs_info->tree_root,
1363 root->root_key.objectid);
1364 WARN_ON(ret);
edbd8d4e 1365 return root;
4df27c4d
YZ		 1366fail:
1367 free_fs_root(root);
1368 return ERR_PTR(ret);
edbd8d4e
CM		 1369}
1370
1371struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
1372 struct btrfs_key *location,
1373 const char *name, int namelen)
1374{
4df27c4d
YZ		 1375 return btrfs_read_fs_root_no_name(fs_info, location);
1376#if 0
edbd8d4e
CM		 1377 struct btrfs_root *root;
1378 int ret;
1379
1380 root = btrfs_read_fs_root_no_name(fs_info, location);
1381 if (!root)
1382 return NULL;
58176a96 1383
4313b399
CM		 1384 if (root->in_sysfs)
1385 return root;
1386
58176a96
JB		 1387 ret = btrfs_set_root_name(root, name, namelen);
1388 if (ret) {
5f39d397 1389 free_extent_buffer(root->node);
58176a96
JB		 1390 kfree(root);
1391 return ERR_PTR(ret);
1392 }
4df27c4d 1393
58176a96
JB		 1394 ret = btrfs_sysfs_add_root(root);
1395 if (ret) {
5f39d397 1396 free_extent_buffer(root->node);
58176a96
JB		 1397 kfree(root->name);
1398 kfree(root);
1399 return ERR_PTR(ret);
1400 }
4313b399 1401 root->in_sysfs = 1;
0f7d52f4 1402 return root;
4df27c4d 1403#endif
0f7d52f4 1404}
04160088
CM		 1405
1406static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1407{
1408 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1409 int ret = 0;
04160088
CM		 1410 struct btrfs_device *device;
1411 struct backing_dev_info *bdi;
b7967db7 1412
c9513edb 1413 mutex_lock(&info->fs_devices->device_list_mutex);
c6e30871 1414 list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1415 if (!device->bdev)
1416 continue;
04160088
CM		 1417 bdi = blk_get_backing_dev_info(device->bdev);
1418 if (bdi && bdi_congested(bdi, bdi_bits)) {
1419 ret = 1;
1420 break;
1421 }
1422 }
c9513edb 1423 mutex_unlock(&info->fs_devices->device_list_mutex);
04160088
CM		 1424 return ret;
1425}
1426
ad081f14
JA		 1427/*
1428 * If this fails, caller must call bdi_destroy() to get rid of the
1429 * bdi again.
1430 */
04160088
CM		 1431static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1432{
ad081f14
JA		 1433 int err;
1434
1435 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1436 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1437 if (err)
1438 return err;
1439
4575c9cc 1440 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1441 bdi->congested_fn = btrfs_congested_fn;
1442 bdi->congested_data = info;
1443 return 0;
1444}
1445
ce9adaa5
CM		 1446static int bio_ready_for_csum(struct bio *bio)
1447{
1448 u64 length = 0;
1449 u64 buf_len = 0;
1450 u64 start = 0;
1451 struct page *page;
1452 struct extent_io_tree *io_tree = NULL;
ce9adaa5
CM		 1453 struct bio_vec *bvec;
1454 int i;
1455 int ret;
1456
1457 bio_for_each_segment(bvec, bio, i) {
1458 page = bvec->bv_page;
1459 if (page->private == EXTENT_PAGE_PRIVATE) {
1460 length += bvec->bv_len;
1461 continue;
1462 }
1463 if (!page->private) {
1464 length += bvec->bv_len;
1465 continue;
1466 }
1467 length = bvec->bv_len;
1468 buf_len = page->private >> 2;
1469 start = page_offset(page) + bvec->bv_offset;
1470 io_tree = &BTRFS_I(page->mapping->host)->io_tree;
ce9adaa5
CM		 1471 }
1472 /* are we fully contained in this bio? */
1473 if (buf_len <= length)
1474 return 1;
1475
1476 ret = extent_range_uptodate(io_tree, start + length,
1477 start + buf_len - 1);
ce9adaa5
CM		 1478 return ret;
1479}
1480
8b712842
CM		 1481/*
1482 * called by the kthread helper functions to finally call the bio end_io
1483 * functions. This is where read checksum verification actually happens
1484 */
1485static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1486{
ce9adaa5 1487 struct bio *bio;
8b712842
CM		 1488 struct end_io_wq *end_io_wq;
1489 struct btrfs_fs_info *fs_info;
ce9adaa5 1490 int error;
ce9adaa5 1491
8b712842
CM		 1492 end_io_wq = container_of(work, struct end_io_wq, work);
1493 bio = end_io_wq->bio;
1494 fs_info = end_io_wq->info;
ce9adaa5 1495
cad321ad 1496 /* metadata bio reads are special because the whole tree block must
8b712842
CM		 1497 * be checksummed at once. This makes sure the entire block is in
1498 * ram and up to date before trying to verify things. For
1499 * blocksize <= pagesize, it is basically a noop
1500 */
7b6d91da 1501 if (!(bio->bi_rw & REQ_WRITE) && end_io_wq->metadata &&
cad321ad 1502 !bio_ready_for_csum(bio)) {
d20f7043 1503 btrfs_queue_worker(&fs_info->endio_meta_workers,
8b712842
CM		 1504 &end_io_wq->work);
1505 return;
1506 }
1507 error = end_io_wq->error;
1508 bio->bi_private = end_io_wq->private;
1509 bio->bi_end_io = end_io_wq->end_io;
1510 kfree(end_io_wq);
8b712842 1511 bio_endio(bio, error);
44b8bd7e
CM		 1512}
1513
a74a4b97
CM		 1514static int cleaner_kthread(void *arg)
1515{
1516 struct btrfs_root *root = arg;
1517
1518 do {
a74a4b97 1519 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
76dda93c
YZ		 1520
1521 if (!(root->fs_info->sb->s_flags & MS_RDONLY) &&
1522 mutex_trylock(&root->fs_info->cleaner_mutex)) {
24bbcf04 1523 btrfs_run_delayed_iputs(root);
76dda93c
YZ		 1524 btrfs_clean_old_snapshots(root);
1525 mutex_unlock(&root->fs_info->cleaner_mutex);
1526 }
a74a4b97
CM		 1527
1528 if (freezing(current)) {
1529 refrigerator();
1530 } else {
a74a4b97 1531 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1532 if (!kthread_should_stop())
1533 schedule();
a74a4b97
CM		 1534 __set_current_state(TASK_RUNNING);
1535 }
1536 } while (!kthread_should_stop());
1537 return 0;
1538}
1539
1540static int transaction_kthread(void *arg)
1541{
1542 struct btrfs_root *root = arg;
1543 struct btrfs_trans_handle *trans;
1544 struct btrfs_transaction *cur;
8929ecfa 1545 u64 transid;
a74a4b97
CM		 1546 unsigned long now;
1547 unsigned long delay;
1548 int ret;
1549
1550 do {
a74a4b97
CM		 1551 delay = HZ * 30;
1552 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1553 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1554
8929ecfa 1555 spin_lock(&root->fs_info->new_trans_lock);
a74a4b97
CM		 1556 cur = root->fs_info->running_transaction;
1557 if (!cur) {
8929ecfa 1558 spin_unlock(&root->fs_info->new_trans_lock);
a74a4b97
CM		 1559 goto sleep;
1560 }
31153d81 1561
a74a4b97 1562 now = get_seconds();
8929ecfa
YZ		 1563 if (!cur->blocked &&
1564 (now < cur->start_time || now - cur->start_time < 30)) {
1565 spin_unlock(&root->fs_info->new_trans_lock);
a74a4b97
CM		 1566 delay = HZ * 5;
1567 goto sleep;
1568 }
8929ecfa
YZ		 1569 transid = cur->transid;
1570 spin_unlock(&root->fs_info->new_trans_lock);
56bec294 1571
8929ecfa 1572 trans = btrfs_join_transaction(root, 1);
3612b495 1573 BUG_ON(IS_ERR(trans));
8929ecfa
YZ		 1574 if (transid == trans->transid) {
1575 ret = btrfs_commit_transaction(trans, root);
1576 BUG_ON(ret);
1577 } else {
1578 btrfs_end_transaction(trans, root);
1579 }
a74a4b97
CM		 1580sleep:
1581 wake_up_process(root->fs_info->cleaner_kthread);
1582 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1583
1584 if (freezing(current)) {
1585 refrigerator();
1586 } else {
a74a4b97 1587 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1588 if (!kthread_should_stop() &&
1589 !btrfs_transaction_blocked(root->fs_info))
1590 schedule_timeout(delay);
a74a4b97
CM		 1591 __set_current_state(TASK_RUNNING);
1592 }
1593 } while (!kthread_should_stop());
1594 return 0;
1595}
1596
8a4b83cc 1597struct btrfs_root *open_ctree(struct super_block *sb,
dfe25020
CM		 1598 struct btrfs_fs_devices *fs_devices,
1599 char *options)
2e635a27 1600{
db94535d
CM		 1601 u32 sectorsize;
1602 u32 nodesize;
1603 u32 leafsize;
1604 u32 blocksize;
87ee04eb 1605 u32 stripesize;
84234f3a 1606 u64 generation;
f2b636e8 1607 u64 features;
3de4586c 1608 struct btrfs_key location;
a061fc8d 1609 struct buffer_head *bh;
e02119d5 1610 struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
e20d96d6 1611 GFP_NOFS);
d20f7043
CM		 1612 struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root),
1613 GFP_NOFS);
450ba0ea
JB		 1614 struct btrfs_root *tree_root = btrfs_sb(sb);
1615 struct btrfs_fs_info *fs_info = tree_root->fs_info;
e02119d5 1616 struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
0b86a832 1617 GFP_NOFS);
e02119d5 1618 struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
0b86a832 1619 GFP_NOFS);
e02119d5
CM		 1620 struct btrfs_root *log_tree_root;
1621
eb60ceac 1622 int ret;
e58ca020 1623 int err = -EINVAL;
4543df7e 1624
2c90e5d6 1625 struct btrfs_super_block *disk_super;
8790d502 1626
0463bb4e 1627 if (!extent_root || !tree_root || !fs_info ||
d20f7043 1628 !chunk_root || !dev_root || !csum_root) {
39279cc3
CM		 1629 err = -ENOMEM;
1630 goto fail;
1631 }
76dda93c
YZ		 1632
1633 ret = init_srcu_struct(&fs_info->subvol_srcu);
1634 if (ret) {
1635 err = ret;
1636 goto fail;
1637 }
1638
1639 ret = setup_bdi(fs_info, &fs_info->bdi);
1640 if (ret) {
1641 err = ret;
1642 goto fail_srcu;
1643 }
1644
1645 fs_info->btree_inode = new_inode(sb);
1646 if (!fs_info->btree_inode) {
1647 err = -ENOMEM;
1648 goto fail_bdi;
1649 }
1650
1561deda
MX		 1651 fs_info->btree_inode->i_mapping->flags &= ~__GFP_FS;
1652
76dda93c 1653 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
8fd17795 1654 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 1655 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 1656 INIT_LIST_HEAD(&fs_info->delayed_iputs);
19c00ddc 1657 INIT_LIST_HEAD(&fs_info->hashers);
ea8c2819 1658 INIT_LIST_HEAD(&fs_info->delalloc_inodes);
5a3f23d5 1659 INIT_LIST_HEAD(&fs_info->ordered_operations);
11833d66 1660 INIT_LIST_HEAD(&fs_info->caching_block_groups);
1832a6d5 1661 spin_lock_init(&fs_info->delalloc_lock);
cee36a03 1662 spin_lock_init(&fs_info->new_trans_lock);
31153d81 1663 spin_lock_init(&fs_info->ref_cache_lock);
76dda93c 1664 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 1665 spin_lock_init(&fs_info->delayed_iput_lock);
19c00ddc 1666
58176a96 1667 init_completion(&fs_info->kobj_unregister);
9f5fae2f
CM		 1668 fs_info->tree_root = tree_root;
1669 fs_info->extent_root = extent_root;
d20f7043 1670 fs_info->csum_root = csum_root;
0b86a832
CM		 1671 fs_info->chunk_root = chunk_root;
1672 fs_info->dev_root = dev_root;
8a4b83cc 1673 fs_info->fs_devices = fs_devices;
0b86a832 1674 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 1675 INIT_LIST_HEAD(&fs_info->space_info);
0b86a832 1676 btrfs_mapping_init(&fs_info->mapping_tree);
f0486c68
YZ		 1677 btrfs_init_block_rsv(&fs_info->global_block_rsv);
1678 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv);
1679 btrfs_init_block_rsv(&fs_info->trans_block_rsv);
1680 btrfs_init_block_rsv(&fs_info->chunk_block_rsv);
1681 btrfs_init_block_rsv(&fs_info->empty_block_rsv);
1682 INIT_LIST_HEAD(&fs_info->durable_block_rsv_list);
1683 mutex_init(&fs_info->durable_block_rsv_mutex);
cb03c743 1684 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 1685 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 1686 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 1687 atomic_set(&fs_info->nr_async_bios, 0);
e20d96d6 1688 fs_info->sb = sb;
6f568d35 1689 fs_info->max_inline = 8192 * 1024;
9ed74f2d 1690 fs_info->metadata_ratio = 0;
c8b97818 1691
b34b086c
CM		 1692 fs_info->thread_pool_size = min_t(unsigned long,
1693 num_online_cpus() + 2, 8);
0afbaf8c 1694
3eaa2885
CM		 1695 INIT_LIST_HEAD(&fs_info->ordered_extents);
1696 spin_lock_init(&fs_info->ordered_extent_lock);
1697
a061fc8d
CM		 1698 sb->s_blocksize = 4096;
1699 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 1700 sb->s_bdi = &fs_info->bdi;
a061fc8d 1701
76dda93c
YZ		 1702 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
1703 fs_info->btree_inode->i_nlink = 1;
0afbaf8c
CM		 1704 /*
1705 * we set the i_size on the btree inode to the max possible int.
1706 * the real end of the address space is determined by all of
1707 * the devices in the system
1708 */
1709 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 1710 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 1711 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
1712
5d4f98a2 1713 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 1714 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
5f39d397
CM		 1715 fs_info->btree_inode->i_mapping,
1716 GFP_NOFS);
d1310b2e
CM		 1717 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
1718 GFP_NOFS);
1719
1720 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 1721
76dda93c
YZ		 1722 BTRFS_I(fs_info->btree_inode)->root = tree_root;
1723 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
1724 sizeof(struct btrfs_key));
1725 BTRFS_I(fs_info->btree_inode)->dummy_inode = 1;
c65ddb52 1726 insert_inode_hash(fs_info->btree_inode);
76dda93c 1727
0f9dd46c 1728 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 1729 fs_info->block_group_cache_tree = RB_ROOT;
0f9dd46c 1730
11833d66 1731 extent_io_tree_init(&fs_info->freed_extents[0],
1a5bc167 1732 fs_info->btree_inode->i_mapping, GFP_NOFS);
11833d66
YZ		 1733 extent_io_tree_init(&fs_info->freed_extents[1],
1734 fs_info->btree_inode->i_mapping, GFP_NOFS);
1735 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 1736 fs_info->do_barriers = 1;
e18e4809 1737
39279cc3 1738
79154b1b 1739 mutex_init(&fs_info->trans_mutex);
5a3f23d5 1740 mutex_init(&fs_info->ordered_operations_mutex);
e02119d5 1741 mutex_init(&fs_info->tree_log_mutex);
925baedd 1742 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 1743 mutex_init(&fs_info->transaction_kthread_mutex);
1744 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 1745 mutex_init(&fs_info->volume_mutex);
276e680d 1746 init_rwsem(&fs_info->extent_commit_sem);
c71bf099 1747 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 1748 init_rwsem(&fs_info->subvol_sem);
fa9c0d79
CM		 1749
1750 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
1751 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
1752
e6dcd2dc 1753 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 1754 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 1755 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 1756 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 1757
0b86a832 1758 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 1759 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 1760
a512bbf8 1761 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY		 1762 if (!bh) {
1763 err = -EINVAL;
39279cc3 1764 goto fail_iput;
20b45077 1765 }
39279cc3 1766
a061fc8d 1767 memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy));
2d69a0f8
YZ		 1768 memcpy(&fs_info->super_for_commit, &fs_info->super_copy,
1769 sizeof(fs_info->super_for_commit));
a061fc8d 1770 brelse(bh);
5f39d397 1771
a061fc8d 1772 memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE);
0b86a832 1773
5f39d397 1774 disk_super = &fs_info->super_copy;
0f7d52f4 1775 if (!btrfs_super_root(disk_super))
c6e2bac1 1776 goto fail_iput;
0f7d52f4 1777
acce952b 1778 /* check FS state, whether FS is broken. */
1779 fs_info->fs_state |= btrfs_super_flags(disk_super);
1780
1781 btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
1782
75e7cb7f
LB		 1783 /*
1784 * In the long term, we'll store the compression type in the super
1785 * block, and it'll be used for per file compression control.
1786 */
1787 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
1788
2b82032c
YZ		 1789 ret = btrfs_parse_options(tree_root, options);
1790 if (ret) {
1791 err = ret;
c6e2bac1 1792 goto fail_iput;
2b82032c 1793 }
dfe25020 1794
f2b636e8
JB		 1795 features = btrfs_super_incompat_flags(disk_super) &
1796 ~BTRFS_FEATURE_INCOMPAT_SUPP;
1797 if (features) {
1798 printk(KERN_ERR "BTRFS: couldn't mount because of "
1799 "unsupported optional features (%Lx).\n",
21380931 1800 (unsigned long long)features);
f2b636e8 1801 err = -EINVAL;
c6e2bac1 1802 goto fail_iput;
f2b636e8
JB		 1803 }
1804
5d4f98a2 1805 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae
LZ		 1806 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
1807 if (tree_root->fs_info->compress_type & BTRFS_COMPRESS_LZO)
1808 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1809 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 1810
f2b636e8
JB		 1811 features = btrfs_super_compat_ro_flags(disk_super) &
1812 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
1813 if (!(sb->s_flags & MS_RDONLY) && features) {
1814 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
1815 "unsupported option features (%Lx).\n",
21380931 1816 (unsigned long long)features);
f2b636e8 1817 err = -EINVAL;
c6e2bac1 1818 goto fail_iput;
f2b636e8 1819 }
61d92c32
CM		 1820
1821 btrfs_init_workers(&fs_info->generic_worker,
1822 "genwork", 1, NULL);
1823
5443be45 1824 btrfs_init_workers(&fs_info->workers, "worker",
61d92c32
CM		 1825 fs_info->thread_pool_size,
1826 &fs_info->generic_worker);
c8b97818 1827
771ed689 1828 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
61d92c32
CM		 1829 fs_info->thread_pool_size,
1830 &fs_info->generic_worker);
771ed689 1831
5443be45 1832 btrfs_init_workers(&fs_info->submit_workers, "submit",
b720d209 1833 min_t(u64, fs_devices->num_devices,
61d92c32
CM		 1834 fs_info->thread_pool_size),
1835 &fs_info->generic_worker);
61b49440
CM		 1836
1837 /* a higher idle thresh on the submit workers makes it much more
1838 * likely that bios will be send down in a sane order to the
1839 * devices
1840 */
1841 fs_info->submit_workers.idle_thresh = 64;
53863232 1842
771ed689 1843 fs_info->workers.idle_thresh = 16;
4a69a410 1844 fs_info->workers.ordered = 1;
61b49440 1845
771ed689
CM		 1846 fs_info->delalloc_workers.idle_thresh = 2;
1847 fs_info->delalloc_workers.ordered = 1;
1848
61d92c32
CM		 1849 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
1850 &fs_info->generic_worker);
5443be45 1851 btrfs_init_workers(&fs_info->endio_workers, "endio",
61d92c32
CM		 1852 fs_info->thread_pool_size,
1853 &fs_info->generic_worker);
d20f7043 1854 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
61d92c32
CM		 1855 fs_info->thread_pool_size,
1856 &fs_info->generic_worker);
cad321ad 1857 btrfs_init_workers(&fs_info->endio_meta_write_workers,
61d92c32
CM		 1858 "endio-meta-write", fs_info->thread_pool_size,
1859 &fs_info->generic_worker);
5443be45 1860 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
61d92c32
CM		 1861 fs_info->thread_pool_size,
1862 &fs_info->generic_worker);
0cb59c99
JB		 1863 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
1864 1, &fs_info->generic_worker);
61b49440
CM		 1865
1866 /*
1867 * endios are largely parallel and should have a very
1868 * low idle thresh
1869 */
1870 fs_info->endio_workers.idle_thresh = 4;
b51912c9
CM		 1871 fs_info->endio_meta_workers.idle_thresh = 4;
1872
9042846b
CM		 1873 fs_info->endio_write_workers.idle_thresh = 2;
1874 fs_info->endio_meta_write_workers.idle_thresh = 2;
1875
4543df7e 1876 btrfs_start_workers(&fs_info->workers, 1);
61d92c32 1877 btrfs_start_workers(&fs_info->generic_worker, 1);
1cc127b5 1878 btrfs_start_workers(&fs_info->submit_workers, 1);
771ed689 1879 btrfs_start_workers(&fs_info->delalloc_workers, 1);
247e743c 1880 btrfs_start_workers(&fs_info->fixup_workers, 1);
9042846b
CM		 1881 btrfs_start_workers(&fs_info->endio_workers, 1);
1882 btrfs_start_workers(&fs_info->endio_meta_workers, 1);
1883 btrfs_start_workers(&fs_info->endio_meta_write_workers, 1);
1884 btrfs_start_workers(&fs_info->endio_write_workers, 1);
0cb59c99 1885 btrfs_start_workers(&fs_info->endio_freespace_worker, 1);
4543df7e 1886
4575c9cc 1887 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 1888 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
1889 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 1890
db94535d
CM		 1891 nodesize = btrfs_super_nodesize(disk_super);
1892 leafsize = btrfs_super_leafsize(disk_super);
1893 sectorsize = btrfs_super_sectorsize(disk_super);
87ee04eb 1894 stripesize = btrfs_super_stripesize(disk_super);
db94535d
CM		 1895 tree_root->nodesize = nodesize;
1896 tree_root->leafsize = leafsize;
1897 tree_root->sectorsize = sectorsize;
87ee04eb 1898 tree_root->stripesize = stripesize;
a061fc8d
CM		 1899
1900 sb->s_blocksize = sectorsize;
1901 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 1902
39279cc3
CM		 1903 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
1904 sizeof(disk_super->magic))) {
d397712b 1905 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 1906 goto fail_sb_buffer;
1907 }
19c00ddc 1908
925baedd 1909 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 1910 ret = btrfs_read_sys_array(tree_root);
925baedd 1911 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 1912 if (ret) {
d397712b
CM		 1913 printk(KERN_WARNING "btrfs: failed to read the system "
1914 "array on %s\n", sb->s_id);
5d4f98a2 1915 goto fail_sb_buffer;
84eed90f 1916 }
0b86a832
CM		 1917
1918 blocksize = btrfs_level_size(tree_root,
1919 btrfs_super_chunk_root_level(disk_super));
84234f3a 1920 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 1921
1922 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1923 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1924
1925 chunk_root->node = read_tree_block(chunk_root,
1926 btrfs_super_chunk_root(disk_super),
84234f3a 1927 blocksize, generation);
0b86a832 1928 BUG_ON(!chunk_root->node);
83121942
DW		 1929 if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
1930 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
1931 sb->s_id);
1932 goto fail_chunk_root;
1933 }
5d4f98a2
YZ		 1934 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
1935 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 1936
e17cade2 1937 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
d397712b
CM		 1938 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
1939 BTRFS_UUID_SIZE);
e17cade2 1940
925baedd 1941 mutex_lock(&fs_info->chunk_mutex);
0b86a832 1942 ret = btrfs_read_chunk_tree(chunk_root);
925baedd 1943 mutex_unlock(&fs_info->chunk_mutex);
2b82032c 1944 if (ret) {
d397712b
CM		 1945 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
1946 sb->s_id);
2b82032c
YZ		 1947 goto fail_chunk_root;
1948 }
0b86a832 1949
dfe25020
CM		 1950 btrfs_close_extra_devices(fs_devices);
1951
db94535d
CM		 1952 blocksize = btrfs_level_size(tree_root,
1953 btrfs_super_root_level(disk_super));
84234f3a 1954 generation = btrfs_super_generation(disk_super);
0b86a832 1955
e20d96d6 1956 tree_root->node = read_tree_block(tree_root,
db94535d 1957 btrfs_super_root(disk_super),
84234f3a 1958 blocksize, generation);
39279cc3 1959 if (!tree_root->node)
2b82032c 1960 goto fail_chunk_root;
83121942
DW		 1961 if (!test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
1962 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
1963 sb->s_id);
1964 goto fail_tree_root;
1965 }
5d4f98a2
YZ		 1966 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
1967 tree_root->commit_root = btrfs_root_node(tree_root);
db94535d
CM		 1968
1969 ret = find_and_setup_root(tree_root, fs_info,
e20d96d6 1970 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
0b86a832 1971 if (ret)
39279cc3 1972 goto fail_tree_root;
0b86a832
CM		 1973 extent_root->track_dirty = 1;
1974
1975 ret = find_and_setup_root(tree_root, fs_info,
1976 BTRFS_DEV_TREE_OBJECTID, dev_root);
0b86a832
CM		 1977 if (ret)
1978 goto fail_extent_root;
5d4f98a2 1979 dev_root->track_dirty = 1;
3768f368 1980
d20f7043
CM		 1981 ret = find_and_setup_root(tree_root, fs_info,
1982 BTRFS_CSUM_TREE_OBJECTID, csum_root);
1983 if (ret)
5d4f98a2 1984 goto fail_dev_root;
d20f7043
CM		 1985
1986 csum_root->track_dirty = 1;
1987
8929ecfa
YZ		 1988 fs_info->generation = generation;
1989 fs_info->last_trans_committed = generation;
1990 fs_info->data_alloc_profile = (u64)-1;
1991 fs_info->metadata_alloc_profile = (u64)-1;
1992 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
1993
c59021f8 1994 ret = btrfs_init_space_info(fs_info);
1995 if (ret) {
1996 printk(KERN_ERR "Failed to initial space info: %d\n", ret);
1997 goto fail_block_groups;
1998 }
1999
1b1d1f66
JB		 2000 ret = btrfs_read_block_groups(extent_root);
2001 if (ret) {
2002 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
2003 goto fail_block_groups;
2004 }
9078a3e1 2005
a74a4b97
CM		 2006 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2007 "btrfs-cleaner");
57506d50 2008 if (IS_ERR(fs_info->cleaner_kthread))
1b1d1f66 2009 goto fail_block_groups;
a74a4b97
CM		 2010
2011 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2012 tree_root,
2013 "btrfs-transaction");
57506d50 2014 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2015 goto fail_cleaner;
a74a4b97 2016
c289811c
CM		 2017 if (!btrfs_test_opt(tree_root, SSD) &&
2018 !btrfs_test_opt(tree_root, NOSSD) &&
2019 !fs_info->fs_devices->rotating) {
2020 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
2021 "mode\n");
2022 btrfs_set_opt(fs_info->mount_opt, SSD);
2023 }
2024
acce952b 2025 /* do not make disk changes in broken FS */
2026 if (btrfs_super_log_root(disk_super) != 0 &&
2027 !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) {
e02119d5
CM		 2028 u64 bytenr = btrfs_super_log_root(disk_super);
2029
7c2ca468 2030 if (fs_devices->rw_devices == 0) {
d397712b
CM		 2031 printk(KERN_WARNING "Btrfs log replay required "
2032 "on RO media\n");
7c2ca468
CM		 2033 err = -EIO;
2034 goto fail_trans_kthread;
2035 }
e02119d5
CM		 2036 blocksize =
2037 btrfs_level_size(tree_root,
2038 btrfs_super_log_root_level(disk_super));
d18a2c44 2039
676e4c86
DC		 2040 log_tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
2041 if (!log_tree_root) {
2042 err = -ENOMEM;
2043 goto fail_trans_kthread;
2044 }
e02119d5
CM		 2045
2046 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2047 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2048
2049 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 2050 blocksize,
2051 generation + 1);
e02119d5
CM		 2052 ret = btrfs_recover_log_trees(log_tree_root);
2053 BUG_ON(ret);
e556ce2c
YZ		 2054
2055 if (sb->s_flags & MS_RDONLY) {
2056 ret = btrfs_commit_super(tree_root);
2057 BUG_ON(ret);
2058 }
e02119d5 2059 }
1a40e23b 2060
76dda93c
YZ		 2061 ret = btrfs_find_orphan_roots(tree_root);
2062 BUG_ON(ret);
2063
7c2ca468 2064 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b
YZ		 2065 ret = btrfs_cleanup_fs_roots(fs_info);
2066 BUG_ON(ret);
2067
5d4f98a2 2068 ret = btrfs_recover_relocation(tree_root);
d7ce5843
MX		 2069 if (ret < 0) {
2070 printk(KERN_WARNING
2071 "btrfs: failed to recover relocation\n");
2072 err = -EINVAL;
2073 goto fail_trans_kthread;
2074 }
7c2ca468 2075 }
1a40e23b 2076
3de4586c
CM		 2077 location.objectid = BTRFS_FS_TREE_OBJECTID;
2078 location.type = BTRFS_ROOT_ITEM_KEY;
2079 location.offset = (u64)-1;
2080
3de4586c
CM		 2081 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
2082 if (!fs_info->fs_root)
7c2ca468 2083 goto fail_trans_kthread;
3140c9a3
DC		 2084 if (IS_ERR(fs_info->fs_root)) {
2085 err = PTR_ERR(fs_info->fs_root);
2086 goto fail_trans_kthread;
2087 }
c289811c 2088
e3acc2a6
JB		 2089 if (!(sb->s_flags & MS_RDONLY)) {
2090 down_read(&fs_info->cleanup_work_sem);
66b4ffd1
JB		 2091 err = btrfs_orphan_cleanup(fs_info->fs_root);
2092 if (!err)
2093 err = btrfs_orphan_cleanup(fs_info->tree_root);
e3acc2a6 2094 up_read(&fs_info->cleanup_work_sem);
66b4ffd1
JB		 2095 if (err) {
2096 close_ctree(tree_root);
2097 return ERR_PTR(err);
2098 }
e3acc2a6
JB		 2099 }
2100
0f7d52f4 2101 return tree_root;
39279cc3 2102
7c2ca468
CM		 2103fail_trans_kthread:
2104 kthread_stop(fs_info->transaction_kthread);
3f157a2f 2105fail_cleaner:
a74a4b97 2106 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2107
2108 /*
2109 * make sure we're done with the btree inode before we stop our
2110 * kthreads
2111 */
2112 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
2113 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
2114
1b1d1f66
JB		 2115fail_block_groups:
2116 btrfs_free_block_groups(fs_info);
d20f7043 2117 free_extent_buffer(csum_root->node);
5d4f98a2
YZ		 2118 free_extent_buffer(csum_root->commit_root);
2119fail_dev_root:
2120 free_extent_buffer(dev_root->node);
2121 free_extent_buffer(dev_root->commit_root);
0b86a832
CM		 2122fail_extent_root:
2123 free_extent_buffer(extent_root->node);
5d4f98a2 2124 free_extent_buffer(extent_root->commit_root);
39279cc3 2125fail_tree_root:
5f39d397 2126 free_extent_buffer(tree_root->node);
5d4f98a2 2127 free_extent_buffer(tree_root->commit_root);
2b82032c
YZ		 2128fail_chunk_root:
2129 free_extent_buffer(chunk_root->node);
5d4f98a2 2130 free_extent_buffer(chunk_root->commit_root);
39279cc3 2131fail_sb_buffer:
61d92c32 2132 btrfs_stop_workers(&fs_info->generic_worker);
247e743c 2133 btrfs_stop_workers(&fs_info->fixup_workers);
771ed689 2134 btrfs_stop_workers(&fs_info->delalloc_workers);
8b712842
CM		 2135 btrfs_stop_workers(&fs_info->workers);
2136 btrfs_stop_workers(&fs_info->endio_workers);
d20f7043 2137 btrfs_stop_workers(&fs_info->endio_meta_workers);
cad321ad 2138 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
e6dcd2dc 2139 btrfs_stop_workers(&fs_info->endio_write_workers);
0cb59c99 2140 btrfs_stop_workers(&fs_info->endio_freespace_worker);
1cc127b5 2141 btrfs_stop_workers(&fs_info->submit_workers);
4543df7e 2142fail_iput:
7c2ca468 2143 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
4543df7e 2144 iput(fs_info->btree_inode);
7e662854 2145
dfe25020 2146 btrfs_close_devices(fs_info->fs_devices);
84eed90f 2147 btrfs_mapping_tree_free(&fs_info->mapping_tree);
ad081f14 2148fail_bdi:
7e662854 2149 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 2150fail_srcu:
2151 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 2152fail:
39279cc3
CM		 2153 kfree(extent_root);
2154 kfree(tree_root);
2155 kfree(fs_info);
83afeac4
JM		 2156 kfree(chunk_root);
2157 kfree(dev_root);
d20f7043 2158 kfree(csum_root);
39279cc3 2159 return ERR_PTR(err);
eb60ceac
CM		 2160}
2161
f2984462
CM		 2162static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2163{
2164 char b[BDEVNAME_SIZE];
2165
2166 if (uptodate) {
2167 set_buffer_uptodate(bh);
2168 } else {
c3b9a62c 2169 if (printk_ratelimit()) {
f2984462
CM		 2170 printk(KERN_WARNING "lost page write due to "
2171 "I/O error on %s\n",
2172 bdevname(bh->b_bdev, b));
2173 }
1259ab75
CM		 2174 /* note, we dont' set_buffer_write_io_error because we have
2175 * our own ways of dealing with the IO errors
2176 */
f2984462
CM		 2177 clear_buffer_uptodate(bh);
2178 }
2179 unlock_buffer(bh);
2180 put_bh(bh);
2181}
2182
a512bbf8
YZ		 2183struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
2184{
2185 struct buffer_head *bh;
2186 struct buffer_head *latest = NULL;
2187 struct btrfs_super_block *super;
2188 int i;
2189 u64 transid = 0;
2190 u64 bytenr;
2191
2192 /* we would like to check all the supers, but that would make
2193 * a btrfs mount succeed after a mkfs from a different FS.
2194 * So, we need to add a special mount option to scan for
2195 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2196 */
2197 for (i = 0; i < 1; i++) {
2198 bytenr = btrfs_sb_offset(i);
2199 if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
2200 break;
2201 bh = __bread(bdev, bytenr / 4096, 4096);
2202 if (!bh)
2203 continue;
2204
2205 super = (struct btrfs_super_block *)bh->b_data;
2206 if (btrfs_super_bytenr(super) != bytenr ||
2207 strncmp((char *)(&super->magic), BTRFS_MAGIC,
2208 sizeof(super->magic))) {
2209 brelse(bh);
2210 continue;
2211 }
2212
2213 if (!latest || btrfs_super_generation(super) > transid) {
2214 brelse(latest);
2215 latest = bh;
2216 transid = btrfs_super_generation(super);
2217 } else {
2218 brelse(bh);
2219 }
2220 }
2221 return latest;
2222}
2223
4eedeb75
HH		 2224/*
2225 * this should be called twice, once with wait == 0 and
2226 * once with wait == 1. When wait == 0 is done, all the buffer heads
2227 * we write are pinned.
2228 *
2229 * They are released when wait == 1 is done.
2230 * max_mirrors must be the same for both runs, and it indicates how
2231 * many supers on this one device should be written.
2232 *
2233 * max_mirrors == 0 means to write them all.
2234 */
a512bbf8
YZ		 2235static int write_dev_supers(struct btrfs_device *device,
2236 struct btrfs_super_block *sb,
2237 int do_barriers, int wait, int max_mirrors)
2238{
2239 struct buffer_head *bh;
2240 int i;
2241 int ret;
2242 int errors = 0;
2243 u32 crc;
2244 u64 bytenr;
2245 int last_barrier = 0;
2246
2247 if (max_mirrors == 0)
2248 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
2249
2250 /* make sure only the last submit_bh does a barrier */
2251 if (do_barriers) {
2252 for (i = 0; i < max_mirrors; i++) {
2253 bytenr = btrfs_sb_offset(i);
2254 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
2255 device->total_bytes)
2256 break;
2257 last_barrier = i;
2258 }
2259 }
2260
2261 for (i = 0; i < max_mirrors; i++) {
2262 bytenr = btrfs_sb_offset(i);
2263 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
2264 break;
2265
2266 if (wait) {
2267 bh = __find_get_block(device->bdev, bytenr / 4096,
2268 BTRFS_SUPER_INFO_SIZE);
2269 BUG_ON(!bh);
a512bbf8 2270 wait_on_buffer(bh);
4eedeb75
HH		 2271 if (!buffer_uptodate(bh))
2272 errors++;
2273
2274 /* drop our reference */
2275 brelse(bh);
2276
2277 /* drop the reference from the wait == 0 run */
2278 brelse(bh);
2279 continue;
a512bbf8
YZ		 2280 } else {
2281 btrfs_set_super_bytenr(sb, bytenr);
2282
2283 crc = ~(u32)0;
2284 crc = btrfs_csum_data(NULL, (char *)sb +
2285 BTRFS_CSUM_SIZE, crc,
2286 BTRFS_SUPER_INFO_SIZE -
2287 BTRFS_CSUM_SIZE);
2288 btrfs_csum_final(crc, sb->csum);
2289
4eedeb75
HH		 2290 /*
2291 * one reference for us, and we leave it for the
2292 * caller
2293 */
a512bbf8
YZ		 2294 bh = __getblk(device->bdev, bytenr / 4096,
2295 BTRFS_SUPER_INFO_SIZE);
2296 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
2297
4eedeb75 2298 /* one reference for submit_bh */
a512bbf8 2299 get_bh(bh);
4eedeb75
HH		 2300
2301 set_buffer_uptodate(bh);
a512bbf8
YZ		 2302 lock_buffer(bh);
2303 bh->b_end_io = btrfs_end_buffer_write_sync;
2304 }
2305
c3b9a62c
CH		 2306 if (i == last_barrier && do_barriers)
2307 ret = submit_bh(WRITE_FLUSH_FUA, bh);
2308 else
ffbd517d 2309 ret = submit_bh(WRITE_SYNC, bh);
a512bbf8 2310
4eedeb75 2311 if (ret)
a512bbf8 2312 errors++;
a512bbf8
YZ		 2313 }
2314 return errors < i ? 0 : -1;
2315}
2316
2317int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 2318{
e5e9a520 2319 struct list_head *head;
f2984462 2320 struct btrfs_device *dev;
a061fc8d 2321 struct btrfs_super_block *sb;
f2984462 2322 struct btrfs_dev_item *dev_item;
f2984462
CM		 2323 int ret;
2324 int do_barriers;
a236aed1
CM		 2325 int max_errors;
2326 int total_errors = 0;
a061fc8d 2327 u64 flags;
f2984462 2328
a236aed1 2329 max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
f2984462
CM		 2330 do_barriers = !btrfs_test_opt(root, NOBARRIER);
2331
a061fc8d
CM		 2332 sb = &root->fs_info->super_for_commit;
2333 dev_item = &sb->dev_item;
e5e9a520
CM		 2334
2335 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
2336 head = &root->fs_info->fs_devices->devices;
c6e30871 2337 list_for_each_entry(dev, head, dev_list) {
dfe25020
CM		 2338 if (!dev->bdev) {
2339 total_errors++;
2340 continue;
2341 }
2b82032c 2342 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 2343 continue;
2344
2b82032c 2345 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 2346 btrfs_set_stack_device_type(dev_item, dev->type);
2347 btrfs_set_stack_device_id(dev_item, dev->devid);
2348 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
2349 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
2350 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
2351 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
2352 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
2353 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 2354 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 2355
a061fc8d
CM		 2356 flags = btrfs_super_flags(sb);
2357 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
2358
a512bbf8 2359 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 2360 if (ret)
2361 total_errors++;
f2984462 2362 }
a236aed1 2363 if (total_errors > max_errors) {
d397712b
CM		 2364 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2365 total_errors);
a236aed1
CM		 2366 BUG();
2367 }
f2984462 2368
a512bbf8 2369 total_errors = 0;
c6e30871 2370 list_for_each_entry(dev, head, dev_list) {
dfe25020
CM		 2371 if (!dev->bdev)
2372 continue;
2b82032c 2373 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 2374 continue;
2375
a512bbf8
YZ		 2376 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
2377 if (ret)
2378 total_errors++;
f2984462 2379 }
e5e9a520 2380 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 2381 if (total_errors > max_errors) {
d397712b
CM		 2382 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2383 total_errors);
a236aed1
CM		 2384 BUG();
2385 }
f2984462
CM		 2386 return 0;
2387}
2388
a512bbf8
YZ		 2389int write_ctree_super(struct btrfs_trans_handle *trans,
2390 struct btrfs_root *root, int max_mirrors)
eb60ceac 2391{
e66f709b 2392 int ret;
5f39d397 2393
a512bbf8 2394 ret = write_all_supers(root, max_mirrors);
5f39d397 2395 return ret;
cfaa7295
CM		 2396}
2397
5eda7b5e 2398int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
2619ba1f 2399{
4df27c4d 2400 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 2401 radix_tree_delete(&fs_info->fs_roots_radix,
2402 (unsigned long)root->root_key.objectid);
4df27c4d 2403 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 2404
2405 if (btrfs_root_refs(&root->root_item) == 0)
2406 synchronize_srcu(&fs_info->subvol_srcu);
2407
4df27c4d
YZ		 2408 free_fs_root(root);
2409 return 0;
2410}
2411
2412static void free_fs_root(struct btrfs_root *root)
2413{
2414 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
3394e160
CM		 2415 if (root->anon_super.s_dev) {
2416 down_write(&root->anon_super.s_umount);
2417 kill_anon_super(&root->anon_super);
2418 }
4df27c4d
YZ		 2419 free_extent_buffer(root->node);
2420 free_extent_buffer(root->commit_root);
d397712b 2421 kfree(root->name);
2619ba1f 2422 kfree(root);
2619ba1f
CM		 2423}
2424
35b7e476 2425static int del_fs_roots(struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 2426{
2427 int ret;
2428 struct btrfs_root *gang[8];
2429 int i;
2430
76dda93c
YZ		 2431 while (!list_empty(&fs_info->dead_roots)) {
2432 gang[0] = list_entry(fs_info->dead_roots.next,
2433 struct btrfs_root, root_list);
2434 list_del(&gang[0]->root_list);
2435
2436 if (gang[0]->in_radix) {
2437 btrfs_free_fs_root(fs_info, gang[0]);
2438 } else {
2439 free_extent_buffer(gang[0]->node);
2440 free_extent_buffer(gang[0]->commit_root);
2441 kfree(gang[0]);
2442 }
2443 }
2444
d397712b 2445 while (1) {
0f7d52f4
CM		 2446 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2447 (void **)gang, 0,
2448 ARRAY_SIZE(gang));
2449 if (!ret)
2450 break;
2619ba1f 2451 for (i = 0; i < ret; i++)
5eda7b5e 2452 btrfs_free_fs_root(fs_info, gang[i]);
0f7d52f4
CM		 2453 }
2454 return 0;
2455}
b4100d64 2456
c146afad 2457int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 2458{
c146afad
YZ		 2459 u64 root_objectid = 0;
2460 struct btrfs_root *gang[8];
2461 int i;
3768f368 2462 int ret;
e089f05c 2463
c146afad
YZ		 2464 while (1) {
2465 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2466 (void **)gang, root_objectid,
2467 ARRAY_SIZE(gang));
2468 if (!ret)
2469 break;
5d4f98a2
YZ		 2470
2471 root_objectid = gang[ret - 1]->root_key.objectid + 1;
c146afad 2472 for (i = 0; i < ret; i++) {
66b4ffd1
JB		 2473 int err;
2474
c146afad 2475 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 2476 err = btrfs_orphan_cleanup(gang[i]);
2477 if (err)
2478 return err;
c146afad
YZ		 2479 }
2480 root_objectid++;
2481 }
2482 return 0;
2483}
a2135011 2484
c146afad
YZ		 2485int btrfs_commit_super(struct btrfs_root *root)
2486{
2487 struct btrfs_trans_handle *trans;
2488 int ret;
a74a4b97 2489
c146afad 2490 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 2491 btrfs_run_delayed_iputs(root);
a74a4b97 2492 btrfs_clean_old_snapshots(root);
c146afad 2493 mutex_unlock(&root->fs_info->cleaner_mutex);
c71bf099
YZ		 2494
2495 /* wait until ongoing cleanup work done */
2496 down_write(&root->fs_info->cleanup_work_sem);
2497 up_write(&root->fs_info->cleanup_work_sem);
2498
a22285a6 2499 trans = btrfs_join_transaction(root, 1);
3612b495
TI		 2500 if (IS_ERR(trans))
2501 return PTR_ERR(trans);
54aa1f4d 2502 ret = btrfs_commit_transaction(trans, root);
c146afad
YZ		 2503 BUG_ON(ret);
2504 /* run commit again to drop the original snapshot */
a22285a6 2505 trans = btrfs_join_transaction(root, 1);
3612b495
TI		 2506 if (IS_ERR(trans))
2507 return PTR_ERR(trans);
79154b1b
CM		 2508 btrfs_commit_transaction(trans, root);
2509 ret = btrfs_write_and_wait_transaction(NULL, root);
3768f368 2510 BUG_ON(ret);
d6bfde87 2511
a512bbf8 2512 ret = write_ctree_super(NULL, root, 0);
c146afad
YZ		 2513 return ret;
2514}
2515
2516int close_ctree(struct btrfs_root *root)
2517{
2518 struct btrfs_fs_info *fs_info = root->fs_info;
2519 int ret;
2520
2521 fs_info->closing = 1;
2522 smp_mb();
2523
0af3d00b 2524 btrfs_put_block_group_cache(fs_info);
acce952b 2525
2526 /*
2527 * Here come 2 situations when btrfs is broken to flip readonly:
2528 *
2529 * 1. when btrfs flips readonly somewhere else before
2530 * btrfs_commit_super, sb->s_flags has MS_RDONLY flag,
2531 * and btrfs will skip to write sb directly to keep
2532 * ERROR state on disk.
2533 *
2534 * 2. when btrfs flips readonly just in btrfs_commit_super,
ae0e47f0 2535 * and in such case, btrfs cannot write sb via btrfs_commit_super,
acce952b 2536 * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag,
2537 * btrfs will cleanup all FS resources first and write sb then.
2538 */
c146afad 2539 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 2540 ret = btrfs_commit_super(root);
2541 if (ret)
2542 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
2543 }
2544
2545 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
2546 ret = btrfs_error_commit_super(root);
d397712b
CM		 2547 if (ret)
2548 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
c146afad 2549 }
0f7d52f4 2550
8929ecfa
YZ		 2551 kthread_stop(root->fs_info->transaction_kthread);
2552 kthread_stop(root->fs_info->cleaner_kthread);
2553
f25784b3
YZ		 2554 fs_info->closing = 2;
2555 smp_mb();
2556
b0c68f8b 2557 if (fs_info->delalloc_bytes) {
d397712b 2558 printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
21380931 2559 (unsigned long long)fs_info->delalloc_bytes);
b0c68f8b 2560 }
31153d81 2561 if (fs_info->total_ref_cache_size) {
d397712b
CM		 2562 printk(KERN_INFO "btrfs: at umount reference cache size %llu\n",
2563 (unsigned long long)fs_info->total_ref_cache_size);
31153d81 2564 }
bcc63abb 2565
5d4f98a2
YZ		 2566 free_extent_buffer(fs_info->extent_root->node);
2567 free_extent_buffer(fs_info->extent_root->commit_root);
2568 free_extent_buffer(fs_info->tree_root->node);
2569 free_extent_buffer(fs_info->tree_root->commit_root);
2570 free_extent_buffer(root->fs_info->chunk_root->node);
2571 free_extent_buffer(root->fs_info->chunk_root->commit_root);
2572 free_extent_buffer(root->fs_info->dev_root->node);
2573 free_extent_buffer(root->fs_info->dev_root->commit_root);
2574 free_extent_buffer(root->fs_info->csum_root->node);
2575 free_extent_buffer(root->fs_info->csum_root->commit_root);
d20f7043 2576
9078a3e1 2577 btrfs_free_block_groups(root->fs_info);
d10c5f31 2578
c146afad 2579 del_fs_roots(fs_info);
d10c5f31 2580
c146afad 2581 iput(fs_info->btree_inode);
9ad6b7bc 2582
61d92c32 2583 btrfs_stop_workers(&fs_info->generic_worker);
247e743c 2584 btrfs_stop_workers(&fs_info->fixup_workers);
771ed689 2585 btrfs_stop_workers(&fs_info->delalloc_workers);
8b712842
CM		 2586 btrfs_stop_workers(&fs_info->workers);
2587 btrfs_stop_workers(&fs_info->endio_workers);
d20f7043 2588 btrfs_stop_workers(&fs_info->endio_meta_workers);
cad321ad 2589 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
e6dcd2dc 2590 btrfs_stop_workers(&fs_info->endio_write_workers);
0cb59c99 2591 btrfs_stop_workers(&fs_info->endio_freespace_worker);
1cc127b5 2592 btrfs_stop_workers(&fs_info->submit_workers);
d6bfde87 2593
dfe25020 2594 btrfs_close_devices(fs_info->fs_devices);
0b86a832 2595 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 2596
04160088 2597 bdi_destroy(&fs_info->bdi);
76dda93c 2598 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 2599
0f7d52f4 2600 kfree(fs_info->extent_root);
0f7d52f4 2601 kfree(fs_info->tree_root);
0b86a832
CM		 2602 kfree(fs_info->chunk_root);
2603 kfree(fs_info->dev_root);
d20f7043 2604 kfree(fs_info->csum_root);
83a4d548
LZ		 2605 kfree(fs_info);
2606
eb60ceac
CM		 2607 return 0;
2608}
2609
1259ab75 2610int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
5f39d397 2611{
1259ab75 2612 int ret;
810191ff 2613 struct inode *btree_inode = buf->first_page->mapping->host;
1259ab75 2614
2ac55d41
JB		 2615 ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf,
2616 NULL);
1259ab75
CM		 2617 if (!ret)
2618 return ret;
2619
2620 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
2621 parent_transid);
2622 return !ret;
5f39d397
CM		 2623}
2624
2625int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 2626{
810191ff 2627 struct inode *btree_inode = buf->first_page->mapping->host;
d1310b2e 2628 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
5f39d397
CM		 2629 buf);
2630}
6702ed49 2631
5f39d397
CM		 2632void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
2633{
810191ff 2634 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
5f39d397
CM		 2635 u64 transid = btrfs_header_generation(buf);
2636 struct inode *btree_inode = root->fs_info->btree_inode;
b9473439 2637 int was_dirty;
b4ce94de 2638
b9447ef8 2639 btrfs_assert_tree_locked(buf);
ccd467d6 2640 if (transid != root->fs_info->generation) {
d397712b
CM		 2641 printk(KERN_CRIT "btrfs transid mismatch buffer %llu, "
2642 "found %llu running %llu\n",
db94535d 2643 (unsigned long long)buf->start,
d397712b
CM		 2644 (unsigned long long)transid,
2645 (unsigned long long)root->fs_info->generation);
ccd467d6
CM		 2646 WARN_ON(1);
2647 }
b9473439
CM		 2648 was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
2649 buf);
2650 if (!was_dirty) {
2651 spin_lock(&root->fs_info->delalloc_lock);
2652 root->fs_info->dirty_metadata_bytes += buf->len;
2653 spin_unlock(&root->fs_info->delalloc_lock);
2654 }
eb60ceac
CM		 2655}
2656
d3c2fdcf 2657void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
35b7e476 2658{
188de649
CM		 2659 /*
2660 * looks as though older kernels can get into trouble with
2661 * this code, they end up stuck in balance_dirty_pages forever
2662 */
d6bfde87 2663 u64 num_dirty;
771ed689 2664 unsigned long thresh = 32 * 1024 * 1024;
d6bfde87 2665
6933c02e 2666 if (current->flags & PF_MEMALLOC)
d6bfde87
CM		 2667 return;
2668
585ad2c3
CM		 2669 num_dirty = root->fs_info->dirty_metadata_bytes;
2670
d6bfde87
CM		 2671 if (num_dirty > thresh) {
2672 balance_dirty_pages_ratelimited_nr(
d7fc640e 2673 root->fs_info->btree_inode->i_mapping, 1);
d6bfde87 2674 }
188de649 2675 return;
35b7e476 2676}
6b80053d 2677
ca7a79ad 2678int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 2679{
810191ff 2680 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
ce9adaa5 2681 int ret;
ca7a79ad 2682 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
d397712b 2683 if (ret == 0)
b4ce94de 2684 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
ce9adaa5 2685 return ret;
6b80053d 2686}
0da5468f 2687
4bef0848
CM		 2688int btree_lock_page_hook(struct page *page)
2689{
2690 struct inode *inode = page->mapping->host;
b9473439 2691 struct btrfs_root *root = BTRFS_I(inode)->root;
4bef0848
CM		 2692 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2693 struct extent_buffer *eb;
2694 unsigned long len;
2695 u64 bytenr = page_offset(page);
2696
2697 if (page->private == EXTENT_PAGE_PRIVATE)
2698 goto out;
2699
2700 len = page->private >> 2;
2701 eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS);
2702 if (!eb)
2703 goto out;
2704
2705 btrfs_tree_lock(eb);
4bef0848 2706 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
b9473439
CM		 2707
2708 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
2709 spin_lock(&root->fs_info->delalloc_lock);
2710 if (root->fs_info->dirty_metadata_bytes >= eb->len)
2711 root->fs_info->dirty_metadata_bytes -= eb->len;
2712 else
2713 WARN_ON(1);
2714 spin_unlock(&root->fs_info->delalloc_lock);
2715 }
2716
4bef0848
CM		 2717 btrfs_tree_unlock(eb);
2718 free_extent_buffer(eb);
2719out:
2720 lock_page(page);
2721 return 0;
2722}
2723
acce952b 2724static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
2725 int read_only)
2726{
2727 if (read_only)
2728 return;
2729
2730 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)
2731 printk(KERN_WARNING "warning: mount fs with errors, "
2732 "running btrfsck is recommended\n");
2733}
2734
2735int btrfs_error_commit_super(struct btrfs_root *root)
2736{
2737 int ret;
2738
2739 mutex_lock(&root->fs_info->cleaner_mutex);
2740 btrfs_run_delayed_iputs(root);
2741 mutex_unlock(&root->fs_info->cleaner_mutex);
2742
2743 down_write(&root->fs_info->cleanup_work_sem);
2744 up_write(&root->fs_info->cleanup_work_sem);
2745
2746 /* cleanup FS via transaction */
2747 btrfs_cleanup_transaction(root);
2748
2749 ret = write_ctree_super(NULL, root, 0);
2750
2751 return ret;
2752}
2753
2754static int btrfs_destroy_ordered_operations(struct btrfs_root *root)
2755{
2756 struct btrfs_inode *btrfs_inode;
2757 struct list_head splice;
2758
2759 INIT_LIST_HEAD(&splice);
2760
2761 mutex_lock(&root->fs_info->ordered_operations_mutex);
2762 spin_lock(&root->fs_info->ordered_extent_lock);
2763
2764 list_splice_init(&root->fs_info->ordered_operations, &splice);
2765 while (!list_empty(&splice)) {
2766 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
2767 ordered_operations);
2768
2769 list_del_init(&btrfs_inode->ordered_operations);
2770
2771 btrfs_invalidate_inodes(btrfs_inode->root);
2772 }
2773
2774 spin_unlock(&root->fs_info->ordered_extent_lock);
2775 mutex_unlock(&root->fs_info->ordered_operations_mutex);
2776
2777 return 0;
2778}
2779
2780static int btrfs_destroy_ordered_extents(struct btrfs_root *root)
2781{
2782 struct list_head splice;
2783 struct btrfs_ordered_extent *ordered;
2784 struct inode *inode;
2785
2786 INIT_LIST_HEAD(&splice);
2787
2788 spin_lock(&root->fs_info->ordered_extent_lock);
2789
2790 list_splice_init(&root->fs_info->ordered_extents, &splice);
2791 while (!list_empty(&splice)) {
2792 ordered = list_entry(splice.next, struct btrfs_ordered_extent,
2793 root_extent_list);
2794
2795 list_del_init(&ordered->root_extent_list);
2796 atomic_inc(&ordered->refs);
2797
2798 /* the inode may be getting freed (in sys_unlink path). */
2799 inode = igrab(ordered->inode);
2800
2801 spin_unlock(&root->fs_info->ordered_extent_lock);
2802 if (inode)
2803 iput(inode);
2804
2805 atomic_set(&ordered->refs, 1);
2806 btrfs_put_ordered_extent(ordered);
2807
2808 spin_lock(&root->fs_info->ordered_extent_lock);
2809 }
2810
2811 spin_unlock(&root->fs_info->ordered_extent_lock);
2812
2813 return 0;
2814}
2815
2816static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
2817 struct btrfs_root *root)
2818{
2819 struct rb_node *node;
2820 struct btrfs_delayed_ref_root *delayed_refs;
2821 struct btrfs_delayed_ref_node *ref;
2822 int ret = 0;
2823
2824 delayed_refs = &trans->delayed_refs;
2825
2826 spin_lock(&delayed_refs->lock);
2827 if (delayed_refs->num_entries == 0) {
cfece4db 2828 spin_unlock(&delayed_refs->lock);
acce952b 2829 printk(KERN_INFO "delayed_refs has NO entry\n");
2830 return ret;
2831 }
2832
2833 node = rb_first(&delayed_refs->root);
2834 while (node) {
2835 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2836 node = rb_next(node);
2837
2838 ref->in_tree = 0;
2839 rb_erase(&ref->rb_node, &delayed_refs->root);
2840 delayed_refs->num_entries--;
2841
2842 atomic_set(&ref->refs, 1);
2843 if (btrfs_delayed_ref_is_head(ref)) {
2844 struct btrfs_delayed_ref_head *head;
2845
2846 head = btrfs_delayed_node_to_head(ref);
2847 mutex_lock(&head->mutex);
2848 kfree(head->extent_op);
2849 delayed_refs->num_heads--;
2850 if (list_empty(&head->cluster))
2851 delayed_refs->num_heads_ready--;
2852 list_del_init(&head->cluster);
2853 mutex_unlock(&head->mutex);
2854 }
2855
2856 spin_unlock(&delayed_refs->lock);
2857 btrfs_put_delayed_ref(ref);
2858
2859 cond_resched();
2860 spin_lock(&delayed_refs->lock);
2861 }
2862
2863 spin_unlock(&delayed_refs->lock);
2864
2865 return ret;
2866}
2867
2868static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t)
2869{
2870 struct btrfs_pending_snapshot *snapshot;
2871 struct list_head splice;
2872
2873 INIT_LIST_HEAD(&splice);
2874
2875 list_splice_init(&t->pending_snapshots, &splice);
2876
2877 while (!list_empty(&splice)) {
2878 snapshot = list_entry(splice.next,
2879 struct btrfs_pending_snapshot,
2880 list);
2881
2882 list_del_init(&snapshot->list);
2883
2884 kfree(snapshot);
2885 }
2886
2887 return 0;
2888}
2889
2890static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
2891{
2892 struct btrfs_inode *btrfs_inode;
2893 struct list_head splice;
2894
2895 INIT_LIST_HEAD(&splice);
2896
2897 list_splice_init(&root->fs_info->delalloc_inodes, &splice);
2898
2899 spin_lock(&root->fs_info->delalloc_lock);
2900
2901 while (!list_empty(&splice)) {
2902 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
2903 delalloc_inodes);
2904
2905 list_del_init(&btrfs_inode->delalloc_inodes);
2906
2907 btrfs_invalidate_inodes(btrfs_inode->root);
2908 }
2909
2910 spin_unlock(&root->fs_info->delalloc_lock);
2911
2912 return 0;
2913}
2914
2915static int btrfs_destroy_marked_extents(struct btrfs_root *root,
2916 struct extent_io_tree *dirty_pages,
2917 int mark)
2918{
2919 int ret;
2920 struct page *page;
2921 struct inode *btree_inode = root->fs_info->btree_inode;
2922 struct extent_buffer *eb;
2923 u64 start = 0;
2924 u64 end;
2925 u64 offset;
2926 unsigned long index;
2927
2928 while (1) {
2929 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
2930 mark);
2931 if (ret)
2932 break;
2933
2934 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
2935 while (start <= end) {
2936 index = start >> PAGE_CACHE_SHIFT;
2937 start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
2938 page = find_get_page(btree_inode->i_mapping, index);
2939 if (!page)
2940 continue;
2941 offset = page_offset(page);
2942
2943 spin_lock(&dirty_pages->buffer_lock);
2944 eb = radix_tree_lookup(
2945 &(&BTRFS_I(page->mapping->host)->io_tree)->buffer,
2946 offset >> PAGE_CACHE_SHIFT);
2947 spin_unlock(&dirty_pages->buffer_lock);
2948 if (eb) {
2949 ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY,
2950 &eb->bflags);
2951 atomic_set(&eb->refs, 1);
2952 }
2953 if (PageWriteback(page))
2954 end_page_writeback(page);
2955
2956 lock_page(page);
2957 if (PageDirty(page)) {
2958 clear_page_dirty_for_io(page);
2959 spin_lock_irq(&page->mapping->tree_lock);
2960 radix_tree_tag_clear(&page->mapping->page_tree,
2961 page_index(page),
2962 PAGECACHE_TAG_DIRTY);
2963 spin_unlock_irq(&page->mapping->tree_lock);
2964 }
2965
2966 page->mapping->a_ops->invalidatepage(page, 0);
2967 unlock_page(page);
2968 }
2969 }
2970
2971 return ret;
2972}
2973
2974static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
2975 struct extent_io_tree *pinned_extents)
2976{
2977 struct extent_io_tree *unpin;
2978 u64 start;
2979 u64 end;
2980 int ret;
2981
2982 unpin = pinned_extents;
2983 while (1) {
2984 ret = find_first_extent_bit(unpin, 0, &start, &end,
2985 EXTENT_DIRTY);
2986 if (ret)
2987 break;
2988
2989 /* opt_discard */
5378e607
LD		 2990 if (btrfs_test_opt(root, DISCARD))
2991 ret = btrfs_error_discard_extent(root, start,
2992 end + 1 - start,
2993 NULL);
acce952b 2994
2995 clear_extent_dirty(unpin, start, end, GFP_NOFS);
2996 btrfs_error_unpin_extent_range(root, start, end);
2997 cond_resched();
2998 }
2999
3000 return 0;
3001}
3002
3003static int btrfs_cleanup_transaction(struct btrfs_root *root)
3004{
3005 struct btrfs_transaction *t;
3006 LIST_HEAD(list);
3007
3008 WARN_ON(1);
3009
3010 mutex_lock(&root->fs_info->trans_mutex);
3011 mutex_lock(&root->fs_info->transaction_kthread_mutex);
3012
3013 list_splice_init(&root->fs_info->trans_list, &list);
3014 while (!list_empty(&list)) {
3015 t = list_entry(list.next, struct btrfs_transaction, list);
3016 if (!t)
3017 break;
3018
3019 btrfs_destroy_ordered_operations(root);
3020
3021 btrfs_destroy_ordered_extents(root);
3022
3023 btrfs_destroy_delayed_refs(t, root);
3024
3025 btrfs_block_rsv_release(root,
3026 &root->fs_info->trans_block_rsv,
3027 t->dirty_pages.dirty_bytes);
3028
3029 /* FIXME: cleanup wait for commit */
3030 t->in_commit = 1;
3031 t->blocked = 1;
3032 if (waitqueue_active(&root->fs_info->transaction_blocked_wait))
3033 wake_up(&root->fs_info->transaction_blocked_wait);
3034
3035 t->blocked = 0;
3036 if (waitqueue_active(&root->fs_info->transaction_wait))
3037 wake_up(&root->fs_info->transaction_wait);
3038 mutex_unlock(&root->fs_info->trans_mutex);
3039
3040 mutex_lock(&root->fs_info->trans_mutex);
3041 t->commit_done = 1;
3042 if (waitqueue_active(&t->commit_wait))
3043 wake_up(&t->commit_wait);
3044 mutex_unlock(&root->fs_info->trans_mutex);
3045
3046 mutex_lock(&root->fs_info->trans_mutex);
3047
3048 btrfs_destroy_pending_snapshots(t);
3049
3050 btrfs_destroy_delalloc_inodes(root);
3051
3052 spin_lock(&root->fs_info->new_trans_lock);
3053 root->fs_info->running_transaction = NULL;
3054 spin_unlock(&root->fs_info->new_trans_lock);
3055
3056 btrfs_destroy_marked_extents(root, &t->dirty_pages,
3057 EXTENT_DIRTY);
3058
3059 btrfs_destroy_pinned_extent(root,
3060 root->fs_info->pinned_extents);
3061
13c5a93e 3062 atomic_set(&t->use_count, 0);
acce952b 3063 list_del_init(&t->list);
3064 memset(t, 0, sizeof(*t));
3065 kmem_cache_free(btrfs_transaction_cachep, t);
3066 }
3067
3068 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
3069 mutex_unlock(&root->fs_info->trans_mutex);
3070
3071 return 0;
3072}
3073
d1310b2e 3074static struct extent_io_ops btree_extent_io_ops = {
4bef0848 3075 .write_cache_pages_lock_hook = btree_lock_page_hook,
ce9adaa5 3076 .readpage_end_io_hook = btree_readpage_end_io_hook,
0b86a832 3077 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 3078 /* note we're sharing with inode.c for the merge bio hook */
3079 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 3080};
Linux 6.x block layer and io_uring tree(s)