projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: break out of orphan cleanup if we can't make progress
[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>
7a36ddec 32#include <linux/ratelimit.h>
7e75bf3f 33#include <asm/unaligned.h>
4b4e25f2 34#include "compat.h"
eb60ceac
CM		 35#include "ctree.h"
36#include "disk-io.h"
e089f05c 37#include "transaction.h"
0f7d52f4 38#include "btrfs_inode.h"
0b86a832 39#include "volumes.h"
db94535d 40#include "print-tree.h"
8b712842 41#include "async-thread.h"
925baedd 42#include "locking.h"
e02119d5 43#include "tree-log.h"
fa9c0d79 44#include "free-space-cache.h"
581bb050 45#include "inode-map.h"
eb60ceac 46
d1310b2e 47static struct extent_io_ops btree_extent_io_ops;
8b712842 48static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 49static void free_fs_root(struct btrfs_root *root);
acce952b 50static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
51 int read_only);
52static int btrfs_destroy_ordered_operations(struct btrfs_root *root);
53static int btrfs_destroy_ordered_extents(struct btrfs_root *root);
54static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
55 struct btrfs_root *root);
56static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t);
57static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
58static int btrfs_destroy_marked_extents(struct btrfs_root *root,
59 struct extent_io_tree *dirty_pages,
60 int mark);
61static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
62 struct extent_io_tree *pinned_extents);
63static int btrfs_cleanup_transaction(struct btrfs_root *root);
ce9adaa5 64
d352ac68
CM		 65/*
66 * end_io_wq structs are used to do processing in task context when an IO is
67 * complete. This is used during reads to verify checksums, and it is used
68 * by writes to insert metadata for new file extents after IO is complete.
69 */
ce9adaa5
CM		 70struct end_io_wq {
71 struct bio *bio;
72 bio_end_io_t *end_io;
73 void *private;
74 struct btrfs_fs_info *info;
75 int error;
22c59948 76 int metadata;
ce9adaa5 77 struct list_head list;
8b712842 78 struct btrfs_work work;
ce9adaa5 79};
0da5468f 80
d352ac68
CM		 81/*
82 * async submit bios are used to offload expensive checksumming
83 * onto the worker threads. They checksum file and metadata bios
84 * just before they are sent down the IO stack.
85 */
44b8bd7e
CM		 86struct async_submit_bio {
87 struct inode *inode;
88 struct bio *bio;
89 struct list_head list;
4a69a410
CM		 90 extent_submit_bio_hook_t *submit_bio_start;
91 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM		 92 int rw;
93 int mirror_num;
c8b97818 94 unsigned long bio_flags;
eaf25d93
CM		 95 /*
96 * bio_offset is optional, can be used if the pages in the bio
97 * can't tell us where in the file the bio should go
98 */
99 u64 bio_offset;
8b712842 100 struct btrfs_work work;
44b8bd7e
CM		 101};
102
85d4e461
CM		 103/*
104 * Lockdep class keys for extent_buffer->lock's in this root. For a given
105 * eb, the lockdep key is determined by the btrfs_root it belongs to and
106 * the level the eb occupies in the tree.
107 *
108 * Different roots are used for different purposes and may nest inside each
109 * other and they require separate keysets. As lockdep keys should be
110 * static, assign keysets according to the purpose of the root as indicated
111 * by btrfs_root->objectid. This ensures that all special purpose roots
112 * have separate keysets.
4008c04a 113 *
85d4e461
CM		 114 * Lock-nesting across peer nodes is always done with the immediate parent
115 * node locked thus preventing deadlock. As lockdep doesn't know this, use
116 * subclass to avoid triggering lockdep warning in such cases.
4008c04a 117 *
85d4e461
CM		 118 * The key is set by the readpage_end_io_hook after the buffer has passed
119 * csum validation but before the pages are unlocked. It is also set by
120 * btrfs_init_new_buffer on freshly allocated blocks.
4008c04a 121 *
85d4e461
CM		 122 * We also add a check to make sure the highest level of the tree is the
123 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
124 * needs update as well.
4008c04a
CM		 125 */
126#ifdef CONFIG_DEBUG_LOCK_ALLOC
127# if BTRFS_MAX_LEVEL != 8
128# error
129# endif
85d4e461
CM		 130
131static struct btrfs_lockdep_keyset {
132 u64 id; /* root objectid */
133 const char *name_stem; /* lock name stem */
134 char names[BTRFS_MAX_LEVEL + 1][20];
135 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
136} btrfs_lockdep_keysets[] = {
137 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
138 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
139 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
140 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
141 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
142 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
143 { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" },
144 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
145 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
146 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
147 { .id = 0, .name_stem = "tree" },
4008c04a 148};
85d4e461
CM		 149
150void __init btrfs_init_lockdep(void)
151{
152 int i, j;
153
154 /* initialize lockdep class names */
155 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
156 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
157
158 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
159 snprintf(ks->names[j], sizeof(ks->names[j]),
160 "btrfs-%s-%02d", ks->name_stem, j);
161 }
162}
163
164void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
165 int level)
166{
167 struct btrfs_lockdep_keyset *ks;
168
169 BUG_ON(level >= ARRAY_SIZE(ks->keys));
170
171 /* find the matching keyset, id 0 is the default entry */
172 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
173 if (ks->id == objectid)
174 break;
175
176 lockdep_set_class_and_name(&eb->lock,
177 &ks->keys[level], ks->names[level]);
178}
179
4008c04a
CM		 180#endif
181
d352ac68
CM		 182/*
183 * extents on the btree inode are pretty simple, there's one extent
184 * that covers the entire device
185 */
b2950863 186static struct extent_map *btree_get_extent(struct inode *inode,
306e16ce 187 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 188 int create)
7eccb903 189{
5f39d397
CM		 190 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
191 struct extent_map *em;
192 int ret;
193
890871be 194 read_lock(&em_tree->lock);
d1310b2e 195 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 196 if (em) {
197 em->bdev =
198 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 199 read_unlock(&em_tree->lock);
5f39d397 200 goto out;
a061fc8d 201 }
890871be 202 read_unlock(&em_tree->lock);
7b13b7b1 203
172ddd60 204 em = alloc_extent_map();
5f39d397
CM		 205 if (!em) {
206 em = ERR_PTR(-ENOMEM);
207 goto out;
208 }
209 em->start = 0;
0afbaf8c 210 em->len = (u64)-1;
c8b97818 211 em->block_len = (u64)-1;
5f39d397 212 em->block_start = 0;
a061fc8d 213 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 214
890871be 215 write_lock(&em_tree->lock);
5f39d397
CM		 216 ret = add_extent_mapping(em_tree, em);
217 if (ret == -EEXIST) {
0afbaf8c
CM		 218 u64 failed_start = em->start;
219 u64 failed_len = em->len;
220
5f39d397 221 free_extent_map(em);
7b13b7b1 222 em = lookup_extent_mapping(em_tree, start, len);
0afbaf8c 223 if (em) {
7b13b7b1 224 ret = 0;
0afbaf8c
CM		 225 } else {
226 em = lookup_extent_mapping(em_tree, failed_start,
227 failed_len);
7b13b7b1 228 ret = -EIO;
0afbaf8c 229 }
5f39d397 230 } else if (ret) {
7b13b7b1
CM		 231 free_extent_map(em);
232 em = NULL;
5f39d397 233 }
890871be 234 write_unlock(&em_tree->lock);
7b13b7b1
CM		 235
236 if (ret)
237 em = ERR_PTR(ret);
5f39d397
CM		 238out:
239 return em;
7eccb903
CM		 240}
241
19c00ddc
CM		 242u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
243{
163e783e 244 return crc32c(seed, data, len);
19c00ddc
CM		 245}
246
247void btrfs_csum_final(u32 crc, char *result)
248{
7e75bf3f 249 put_unaligned_le32(~crc, result);
19c00ddc
CM		 250}
251
d352ac68
CM		 252/*
253 * compute the csum for a btree block, and either verify it or write it
254 * into the csum field of the block.
255 */
19c00ddc
CM		 256static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
257 int verify)
258{
607d432d
JB		 259 u16 csum_size =
260 btrfs_super_csum_size(&root->fs_info->super_copy);
261 char *result = NULL;
19c00ddc
CM		 262 unsigned long len;
263 unsigned long cur_len;
264 unsigned long offset = BTRFS_CSUM_SIZE;
19c00ddc
CM		 265 char *kaddr;
266 unsigned long map_start;
267 unsigned long map_len;
268 int err;
269 u32 crc = ~(u32)0;
607d432d 270 unsigned long inline_result;
19c00ddc
CM		 271
272 len = buf->len - offset;
d397712b 273 while (len > 0) {
19c00ddc 274 err = map_private_extent_buffer(buf, offset, 32,
a6591715 275 &kaddr, &map_start, &map_len);
d397712b 276 if (err)
19c00ddc 277 return 1;
19c00ddc
CM		 278 cur_len = min(len, map_len - (offset - map_start));
279 crc = btrfs_csum_data(root, kaddr + offset - map_start,
280 crc, cur_len);
281 len -= cur_len;
282 offset += cur_len;
19c00ddc 283 }
607d432d
JB		 284 if (csum_size > sizeof(inline_result)) {
285 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
286 if (!result)
287 return 1;
288 } else {
289 result = (char *)&inline_result;
290 }
291
19c00ddc
CM		 292 btrfs_csum_final(crc, result);
293
294 if (verify) {
607d432d 295 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 296 u32 val;
297 u32 found = 0;
607d432d 298 memcpy(&found, result, csum_size);
e4204ded 299
607d432d 300 read_extent_buffer(buf, &val, 0, csum_size);
7a36ddec 301 printk_ratelimited(KERN_INFO "btrfs: %s checksum verify "
193f284d
CM		 302 "failed on %llu wanted %X found %X "
303 "level %d\n",
304 root->fs_info->sb->s_id,
305 (unsigned long long)buf->start, val, found,
306 btrfs_header_level(buf));
607d432d
JB		 307 if (result != (char *)&inline_result)
308 kfree(result);
19c00ddc
CM		 309 return 1;
310 }
311 } else {
607d432d 312 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 313 }
607d432d
JB		 314 if (result != (char *)&inline_result)
315 kfree(result);
19c00ddc
CM		 316 return 0;
317}
318
d352ac68
CM		 319/*
320 * we can't consider a given block up to date unless the transid of the
321 * block matches the transid in the parent node's pointer. This is how we
322 * detect blocks that either didn't get written at all or got written
323 * in the wrong place.
324 */
1259ab75
CM		 325static int verify_parent_transid(struct extent_io_tree *io_tree,
326 struct extent_buffer *eb, u64 parent_transid)
327{
2ac55d41 328 struct extent_state *cached_state = NULL;
1259ab75
CM		 329 int ret;
330
331 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
332 return 0;
333
2ac55d41
JB		 334 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
335 0, &cached_state, GFP_NOFS);
336 if (extent_buffer_uptodate(io_tree, eb, cached_state) &&
1259ab75
CM		 337 btrfs_header_generation(eb) == parent_transid) {
338 ret = 0;
339 goto out;
340 }
7a36ddec 341 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
193f284d
CM		 342 "found %llu\n",
343 (unsigned long long)eb->start,
344 (unsigned long long)parent_transid,
345 (unsigned long long)btrfs_header_generation(eb));
1259ab75 346 ret = 1;
2ac55d41 347 clear_extent_buffer_uptodate(io_tree, eb, &cached_state);
33958dc6 348out:
2ac55d41
JB		 349 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
350 &cached_state, GFP_NOFS);
1259ab75 351 return ret;
1259ab75
CM		 352}
353
d352ac68
CM		 354/*
355 * helper to read a given tree block, doing retries as required when
356 * the checksums don't match and we have alternate mirrors to try.
357 */
f188591e
CM		 358static int btree_read_extent_buffer_pages(struct btrfs_root *root,
359 struct extent_buffer *eb,
ca7a79ad 360 u64 start, u64 parent_transid)
f188591e
CM		 361{
362 struct extent_io_tree *io_tree;
363 int ret;
364 int num_copies = 0;
365 int mirror_num = 0;
366
a826d6dc 367 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
f188591e
CM		 368 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
369 while (1) {
370 ret = read_extent_buffer_pages(io_tree, eb, start, 1,
371 btree_get_extent, mirror_num);
1259ab75
CM		 372 if (!ret &&
373 !verify_parent_transid(io_tree, eb, parent_transid))
f188591e 374 return ret;
d397712b 375
a826d6dc
JB		 376 /*
377 * This buffer's crc is fine, but its contents are corrupted, so
378 * there is no reason to read the other copies, they won't be
379 * any less wrong.
380 */
381 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
382 return ret;
383
f188591e
CM		 384 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
385 eb->start, eb->len);
4235298e 386 if (num_copies == 1)
f188591e 387 return ret;
4235298e 388
f188591e 389 mirror_num++;
4235298e 390 if (mirror_num > num_copies)
f188591e 391 return ret;
f188591e 392 }
f188591e
CM		 393 return -EIO;
394}
19c00ddc 395
d352ac68 396/*
d397712b
CM		 397 * checksum a dirty tree block before IO. This has extra checks to make sure
398 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 399 */
d397712b 400
b2950863 401static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
19c00ddc 402{
d1310b2e 403 struct extent_io_tree *tree;
35ebb934 404 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
19c00ddc 405 u64 found_start;
19c00ddc
CM		 406 unsigned long len;
407 struct extent_buffer *eb;
f188591e
CM		 408 int ret;
409
d1310b2e 410 tree = &BTRFS_I(page->mapping->host)->io_tree;
19c00ddc 411
eb14ab8e
CM		 412 if (page->private == EXTENT_PAGE_PRIVATE) {
413 WARN_ON(1);
19c00ddc 414 goto out;
eb14ab8e
CM		 415 }
416 if (!page->private) {
417 WARN_ON(1);
19c00ddc 418 goto out;
eb14ab8e 419 }
19c00ddc 420 len = page->private >> 2;
d397712b
CM		 421 WARN_ON(len == 0);
422
ba144192 423 eb = alloc_extent_buffer(tree, start, len, page);
91ca338d
TI		 424 if (eb == NULL) {
425 WARN_ON(1);
426 goto out;
427 }
ca7a79ad
CM		 428 ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
429 btrfs_header_generation(eb));
f188591e 430 BUG_ON(ret);
784b4e29
CM		 431 WARN_ON(!btrfs_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN));
432
19c00ddc
CM		 433 found_start = btrfs_header_bytenr(eb);
434 if (found_start != start) {
55c69072
CM		 435 WARN_ON(1);
436 goto err;
437 }
438 if (eb->first_page != page) {
55c69072
CM		 439 WARN_ON(1);
440 goto err;
441 }
442 if (!PageUptodate(page)) {
55c69072
CM		 443 WARN_ON(1);
444 goto err;
19c00ddc 445 }
19c00ddc 446 csum_tree_block(root, eb, 0);
55c69072 447err:
19c00ddc
CM		 448 free_extent_buffer(eb);
449out:
450 return 0;
451}
452
2b82032c
YZ		 453static int check_tree_block_fsid(struct btrfs_root *root,
454 struct extent_buffer *eb)
455{
456 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
457 u8 fsid[BTRFS_UUID_SIZE];
458 int ret = 1;
459
460 read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
461 BTRFS_FSID_SIZE);
462 while (fs_devices) {
463 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
464 ret = 0;
465 break;
466 }
467 fs_devices = fs_devices->seed;
468 }
469 return ret;
470}
471
a826d6dc
JB		 472#define CORRUPT(reason, eb, root, slot) \
473 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
474 "root=%llu, slot=%d\n", reason, \
475 (unsigned long long)btrfs_header_bytenr(eb), \
476 (unsigned long long)root->objectid, slot)
477
478static noinline int check_leaf(struct btrfs_root *root,
479 struct extent_buffer *leaf)
480{
481 struct btrfs_key key;
482 struct btrfs_key leaf_key;
483 u32 nritems = btrfs_header_nritems(leaf);
484 int slot;
485
486 if (nritems == 0)
487 return 0;
488
489 /* Check the 0 item */
490 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
491 BTRFS_LEAF_DATA_SIZE(root)) {
492 CORRUPT("invalid item offset size pair", leaf, root, 0);
493 return -EIO;
494 }
495
496 /*
497 * Check to make sure each items keys are in the correct order and their
498 * offsets make sense. We only have to loop through nritems-1 because
499 * we check the current slot against the next slot, which verifies the
500 * next slot's offset+size makes sense and that the current's slot
501 * offset is correct.
502 */
503 for (slot = 0; slot < nritems - 1; slot++) {
504 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
505 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
506
507 /* Make sure the keys are in the right order */
508 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
509 CORRUPT("bad key order", leaf, root, slot);
510 return -EIO;
511 }
512
513 /*
514 * Make sure the offset and ends are right, remember that the
515 * item data starts at the end of the leaf and grows towards the
516 * front.
517 */
518 if (btrfs_item_offset_nr(leaf, slot) !=
519 btrfs_item_end_nr(leaf, slot + 1)) {
520 CORRUPT("slot offset bad", leaf, root, slot);
521 return -EIO;
522 }
523
524 /*
525 * Check to make sure that we don't point outside of the leaf,
526 * just incase all the items are consistent to eachother, but
527 * all point outside of the leaf.
528 */
529 if (btrfs_item_end_nr(leaf, slot) >
530 BTRFS_LEAF_DATA_SIZE(root)) {
531 CORRUPT("slot end outside of leaf", leaf, root, slot);
532 return -EIO;
533 }
534 }
535
536 return 0;
537}
538
b2950863 539static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
ce9adaa5
CM		 540 struct extent_state *state)
541{
542 struct extent_io_tree *tree;
543 u64 found_start;
544 int found_level;
545 unsigned long len;
546 struct extent_buffer *eb;
547 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 548 int ret = 0;
ce9adaa5
CM		 549
550 tree = &BTRFS_I(page->mapping->host)->io_tree;
551 if (page->private == EXTENT_PAGE_PRIVATE)
552 goto out;
553 if (!page->private)
554 goto out;
d397712b 555
ce9adaa5 556 len = page->private >> 2;
d397712b
CM		 557 WARN_ON(len == 0);
558
ba144192 559 eb = alloc_extent_buffer(tree, start, len, page);
91ca338d
TI		 560 if (eb == NULL) {
561 ret = -EIO;
562 goto out;
563 }
f188591e 564
ce9adaa5 565 found_start = btrfs_header_bytenr(eb);
23a07867 566 if (found_start != start) {
7a36ddec 567 printk_ratelimited(KERN_INFO "btrfs bad tree block start "
193f284d
CM		 568 "%llu %llu\n",
569 (unsigned long long)found_start,
570 (unsigned long long)eb->start);
f188591e 571 ret = -EIO;
ce9adaa5
CM		 572 goto err;
573 }
574 if (eb->first_page != page) {
d397712b
CM		 575 printk(KERN_INFO "btrfs bad first page %lu %lu\n",
576 eb->first_page->index, page->index);
ce9adaa5 577 WARN_ON(1);
f188591e 578 ret = -EIO;
ce9adaa5
CM		 579 goto err;
580 }
2b82032c 581 if (check_tree_block_fsid(root, eb)) {
7a36ddec 582 printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n",
193f284d 583 (unsigned long long)eb->start);
1259ab75
CM		 584 ret = -EIO;
585 goto err;
586 }
ce9adaa5
CM		 587 found_level = btrfs_header_level(eb);
588
85d4e461
CM		 589 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
590 eb, found_level);
4008c04a 591
ce9adaa5 592 ret = csum_tree_block(root, eb, 1);
a826d6dc 593 if (ret) {
f188591e 594 ret = -EIO;
a826d6dc
JB		 595 goto err;
596 }
597
598 /*
599 * If this is a leaf block and it is corrupt, set the corrupt bit so
600 * that we don't try and read the other copies of this block, just
601 * return -EIO.
602 */
603 if (found_level == 0 && check_leaf(root, eb)) {
604 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
605 ret = -EIO;
606 }
ce9adaa5
CM		 607
608 end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
609 end = eb->start + end - 1;
ce9adaa5
CM		 610err:
611 free_extent_buffer(eb);
612out:
f188591e 613 return ret;
ce9adaa5
CM		 614}
615
ce9adaa5 616static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM		 617{
618 struct end_io_wq *end_io_wq = bio->bi_private;
619 struct btrfs_fs_info *fs_info;
ce9adaa5 620
ce9adaa5 621 fs_info = end_io_wq->info;
ce9adaa5 622 end_io_wq->error = err;
8b712842
CM		 623 end_io_wq->work.func = end_workqueue_fn;
624 end_io_wq->work.flags = 0;
d20f7043 625
7b6d91da 626 if (bio->bi_rw & REQ_WRITE) {
0cb59c99 627 if (end_io_wq->metadata == 1)
cad321ad
CM		 628 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
629 &end_io_wq->work);
0cb59c99
JB		 630 else if (end_io_wq->metadata == 2)
631 btrfs_queue_worker(&fs_info->endio_freespace_worker,
632 &end_io_wq->work);
cad321ad
CM		 633 else
634 btrfs_queue_worker(&fs_info->endio_write_workers,
635 &end_io_wq->work);
d20f7043
CM		 636 } else {
637 if (end_io_wq->metadata)
638 btrfs_queue_worker(&fs_info->endio_meta_workers,
639 &end_io_wq->work);
640 else
641 btrfs_queue_worker(&fs_info->endio_workers,
642 &end_io_wq->work);
643 }
ce9adaa5
CM		 644}
645
0cb59c99
JB		 646/*
647 * For the metadata arg you want
648 *
649 * 0 - if data
650 * 1 - if normal metadta
651 * 2 - if writing to the free space cache area
652 */
22c59948
CM		 653int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
654 int metadata)
0b86a832 655{
ce9adaa5 656 struct end_io_wq *end_io_wq;
ce9adaa5
CM		 657 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
658 if (!end_io_wq)
659 return -ENOMEM;
660
661 end_io_wq->private = bio->bi_private;
662 end_io_wq->end_io = bio->bi_end_io;
22c59948 663 end_io_wq->info = info;
ce9adaa5
CM		 664 end_io_wq->error = 0;
665 end_io_wq->bio = bio;
22c59948 666 end_io_wq->metadata = metadata;
ce9adaa5
CM		 667
668 bio->bi_private = end_io_wq;
669 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 670 return 0;
671}
672
b64a2851 673unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 674{
4854ddd0
CM		 675 unsigned long limit = min_t(unsigned long,
676 info->workers.max_workers,
677 info->fs_devices->open_devices);
678 return 256 * limit;
679}
0986fe9e 680
4a69a410
CM		 681static void run_one_async_start(struct btrfs_work *work)
682{
4a69a410
CM		 683 struct async_submit_bio *async;
684
685 async = container_of(work, struct async_submit_bio, work);
4a69a410 686 async->submit_bio_start(async->inode, async->rw, async->bio,
eaf25d93
CM		 687 async->mirror_num, async->bio_flags,
688 async->bio_offset);
4a69a410
CM		 689}
690
691static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 692{
693 struct btrfs_fs_info *fs_info;
694 struct async_submit_bio *async;
4854ddd0 695 int limit;
8b712842
CM		 696
697 async = container_of(work, struct async_submit_bio, work);
698 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 699
b64a2851 700 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 701 limit = limit * 2 / 3;
702
8b712842 703 atomic_dec(&fs_info->nr_async_submits);
0986fe9e 704
b64a2851
CM		 705 if (atomic_read(&fs_info->nr_async_submits) < limit &&
706 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 707 wake_up(&fs_info->async_submit_wait);
708
4a69a410 709 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 710 async->mirror_num, async->bio_flags,
711 async->bio_offset);
4a69a410
CM		 712}
713
714static void run_one_async_free(struct btrfs_work *work)
715{
716 struct async_submit_bio *async;
717
718 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 719 kfree(async);
720}
721
44b8bd7e
CM		 722int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
723 int rw, struct bio *bio, int mirror_num,
c8b97818 724 unsigned long bio_flags,
eaf25d93 725 u64 bio_offset,
4a69a410
CM		 726 extent_submit_bio_hook_t *submit_bio_start,
727 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 728{
729 struct async_submit_bio *async;
730
731 async = kmalloc(sizeof(*async), GFP_NOFS);
732 if (!async)
733 return -ENOMEM;
734
735 async->inode = inode;
736 async->rw = rw;
737 async->bio = bio;
738 async->mirror_num = mirror_num;
4a69a410
CM		 739 async->submit_bio_start = submit_bio_start;
740 async->submit_bio_done = submit_bio_done;
741
742 async->work.func = run_one_async_start;
743 async->work.ordered_func = run_one_async_done;
744 async->work.ordered_free = run_one_async_free;
745
8b712842 746 async->work.flags = 0;
c8b97818 747 async->bio_flags = bio_flags;
eaf25d93 748 async->bio_offset = bio_offset;
8c8bee1d 749
cb03c743 750 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 751
7b6d91da 752 if (rw & REQ_SYNC)
d313d7a3
CM		 753 btrfs_set_work_high_prio(&async->work);
754
8b712842 755 btrfs_queue_worker(&fs_info->workers, &async->work);
9473f16c 756
d397712b 757 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 758 atomic_read(&fs_info->nr_async_submits)) {
759 wait_event(fs_info->async_submit_wait,
760 (atomic_read(&fs_info->nr_async_submits) == 0));
761 }
762
44b8bd7e
CM		 763 return 0;
764}
765
ce3ed71a
CM		 766static int btree_csum_one_bio(struct bio *bio)
767{
768 struct bio_vec *bvec = bio->bi_io_vec;
769 int bio_index = 0;
770 struct btrfs_root *root;
771
772 WARN_ON(bio->bi_vcnt <= 0);
d397712b 773 while (bio_index < bio->bi_vcnt) {
ce3ed71a
CM		 774 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
775 csum_dirty_buffer(root, bvec->bv_page);
776 bio_index++;
777 bvec++;
778 }
779 return 0;
780}
781
4a69a410
CM		 782static int __btree_submit_bio_start(struct inode *inode, int rw,
783 struct bio *bio, int mirror_num,
eaf25d93
CM		 784 unsigned long bio_flags,
785 u64 bio_offset)
22c59948 786{
8b712842
CM		 787 /*
788 * when we're called for a write, we're already in the async
5443be45 789 * submission context. Just jump into btrfs_map_bio
8b712842 790 */
4a69a410
CM		 791 btree_csum_one_bio(bio);
792 return 0;
793}
22c59948 794
4a69a410 795static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 796 int mirror_num, unsigned long bio_flags,
797 u64 bio_offset)
4a69a410 798{
8b712842 799 /*
4a69a410
CM		 800 * when we're called for a write, we're already in the async
801 * submission context. Just jump into btrfs_map_bio
8b712842 802 */
8b712842 803 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
0b86a832
CM		 804}
805
44b8bd7e 806static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 807 int mirror_num, unsigned long bio_flags,
808 u64 bio_offset)
44b8bd7e 809{
cad321ad
CM		 810 int ret;
811
812 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
813 bio, 1);
814 BUG_ON(ret);
815
7b6d91da 816 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 817 /*
818 * called for a read, do the setup so that checksum validation
819 * can happen in the async kernel threads
820 */
4a69a410 821 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
6f3577bd 822 mirror_num, 0);
44b8bd7e 823 }
d313d7a3 824
cad321ad
CM		 825 /*
826 * kthread helpers are used to submit writes so that checksumming
827 * can happen in parallel across all CPUs
828 */
44b8bd7e 829 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
c8b97818 830 inode, rw, bio, mirror_num, 0,
eaf25d93 831 bio_offset,
4a69a410
CM		 832 __btree_submit_bio_start,
833 __btree_submit_bio_done);
44b8bd7e
CM		 834}
835
3dd1462e 836#ifdef CONFIG_MIGRATION
784b4e29
CM		 837static int btree_migratepage(struct address_space *mapping,
838 struct page *newpage, struct page *page)
839{
840 /*
841 * we can't safely write a btree page from here,
842 * we haven't done the locking hook
843 */
844 if (PageDirty(page))
845 return -EAGAIN;
846 /*
847 * Buffers may be managed in a filesystem specific way.
848 * We must have no buffers or drop them.
849 */
850 if (page_has_private(page) &&
851 !try_to_release_page(page, GFP_KERNEL))
852 return -EAGAIN;
784b4e29
CM		 853 return migrate_page(mapping, newpage, page);
854}
3dd1462e 855#endif
784b4e29 856
0da5468f
CM		 857static int btree_writepage(struct page *page, struct writeback_control *wbc)
858{
d1310b2e 859 struct extent_io_tree *tree;
b9473439
CM		 860 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
861 struct extent_buffer *eb;
862 int was_dirty;
863
d1310b2e 864 tree = &BTRFS_I(page->mapping->host)->io_tree;
b9473439
CM		 865 if (!(current->flags & PF_MEMALLOC)) {
866 return extent_write_full_page(tree, page,
867 btree_get_extent, wbc);
868 }
5443be45 869
b9473439 870 redirty_page_for_writepage(wbc, page);
784b4e29 871 eb = btrfs_find_tree_block(root, page_offset(page), PAGE_CACHE_SIZE);
b9473439
CM		 872 WARN_ON(!eb);
873
874 was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
875 if (!was_dirty) {
876 spin_lock(&root->fs_info->delalloc_lock);
877 root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE;
878 spin_unlock(&root->fs_info->delalloc_lock);
5443be45 879 }
b9473439
CM		 880 free_extent_buffer(eb);
881
882 unlock_page(page);
883 return 0;
5f39d397 884}
0da5468f
CM		 885
886static int btree_writepages(struct address_space *mapping,
887 struct writeback_control *wbc)
888{
d1310b2e
CM		 889 struct extent_io_tree *tree;
890 tree = &BTRFS_I(mapping->host)->io_tree;
d8d5f3e1 891 if (wbc->sync_mode == WB_SYNC_NONE) {
b9473439 892 struct btrfs_root *root = BTRFS_I(mapping->host)->root;
793955bc 893 u64 num_dirty;
24ab9cd8 894 unsigned long thresh = 32 * 1024 * 1024;
448d640b
CM		 895
896 if (wbc->for_kupdate)
897 return 0;
898
b9473439
CM		 899 /* this is a bit racy, but that's ok */
900 num_dirty = root->fs_info->dirty_metadata_bytes;
d397712b 901 if (num_dirty < thresh)
793955bc 902 return 0;
793955bc 903 }
0da5468f
CM		 904 return extent_writepages(tree, mapping, btree_get_extent, wbc);
905}
906
b2950863 907static int btree_readpage(struct file *file, struct page *page)
5f39d397 908{
d1310b2e
CM		 909 struct extent_io_tree *tree;
910 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 911 return extent_read_full_page(tree, page, btree_get_extent);
912}
22b0ebda 913
70dec807 914static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 915{
d1310b2e
CM		 916 struct extent_io_tree *tree;
917 struct extent_map_tree *map;
5f39d397 918 int ret;
d98237b3 919
98509cfc 920 if (PageWriteback(page) || PageDirty(page))
d397712b 921 return 0;
98509cfc 922
d1310b2e
CM		 923 tree = &BTRFS_I(page->mapping->host)->io_tree;
924 map = &BTRFS_I(page->mapping->host)->extent_tree;
6af118ce 925
7b13b7b1 926 ret = try_release_extent_state(map, tree, page, gfp_flags);
d397712b 927 if (!ret)
6af118ce 928 return 0;
6af118ce
CM		 929
930 ret = try_release_extent_buffer(tree, page);
5f39d397
CM		 931 if (ret == 1) {
932 ClearPagePrivate(page);
933 set_page_private(page, 0);
934 page_cache_release(page);
935 }
6af118ce 936
d98237b3
CM		 937 return ret;
938}
939
5f39d397 940static void btree_invalidatepage(struct page *page, unsigned long offset)
d98237b3 941{
d1310b2e
CM		 942 struct extent_io_tree *tree;
943 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 944 extent_invalidatepage(tree, page, offset);
945 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 946 if (PagePrivate(page)) {
d397712b
CM		 947 printk(KERN_WARNING "btrfs warning page private not zero "
948 "on page %llu\n", (unsigned long long)page_offset(page));
9ad6b7bc
CM		 949 ClearPagePrivate(page);
950 set_page_private(page, 0);
951 page_cache_release(page);
952 }
d98237b3
CM		 953}
954
7f09410b 955static const struct address_space_operations btree_aops = {
d98237b3
CM		 956 .readpage = btree_readpage,
957 .writepage = btree_writepage,
0da5468f 958 .writepages = btree_writepages,
5f39d397
CM		 959 .releasepage = btree_releasepage,
960 .invalidatepage = btree_invalidatepage,
5a92bc88 961#ifdef CONFIG_MIGRATION
784b4e29 962 .migratepage = btree_migratepage,
5a92bc88 963#endif
d98237b3
CM		 964};
965
ca7a79ad
CM		 966int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
967 u64 parent_transid)
090d1875 968{
5f39d397
CM		 969 struct extent_buffer *buf = NULL;
970 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 971 int ret = 0;
090d1875 972
db94535d 973 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 974 if (!buf)
090d1875 975 return 0;
d1310b2e 976 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
f188591e 977 buf, 0, 0, btree_get_extent, 0);
5f39d397 978 free_extent_buffer(buf);
de428b63 979 return ret;
090d1875
CM		 980}
981
0999df54
CM		 982struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
983 u64 bytenr, u32 blocksize)
984{
985 struct inode *btree_inode = root->fs_info->btree_inode;
986 struct extent_buffer *eb;
987 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
f09d1f60 988 bytenr, blocksize);
0999df54
CM		 989 return eb;
990}
991
992struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
993 u64 bytenr, u32 blocksize)
994{
995 struct inode *btree_inode = root->fs_info->btree_inode;
996 struct extent_buffer *eb;
997
998 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
ba144192 999 bytenr, blocksize, NULL);
0999df54
CM		 1000 return eb;
1001}
1002
1003
e02119d5
CM		 1004int btrfs_write_tree_block(struct extent_buffer *buf)
1005{
8aa38c31
CH		 1006 return filemap_fdatawrite_range(buf->first_page->mapping, buf->start,
1007 buf->start + buf->len - 1);
e02119d5
CM		 1008}
1009
1010int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1011{
8aa38c31
CH		 1012 return filemap_fdatawait_range(buf->first_page->mapping,
1013 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1014}
1015
0999df54 1016struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 1017 u32 blocksize, u64 parent_transid)
0999df54
CM		 1018{
1019 struct extent_buffer *buf = NULL;
0999df54
CM		 1020 int ret;
1021
0999df54
CM		 1022 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1023 if (!buf)
1024 return NULL;
0999df54 1025
ca7a79ad 1026 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
ce9adaa5 1027
d397712b 1028 if (ret == 0)
b4ce94de 1029 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
5f39d397 1030 return buf;
ce9adaa5 1031
eb60ceac
CM		 1032}
1033
e089f05c 1034int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
5f39d397 1035 struct extent_buffer *buf)
ed2ff2cb 1036{
5f39d397 1037 struct inode *btree_inode = root->fs_info->btree_inode;
55c69072 1038 if (btrfs_header_generation(buf) ==
925baedd 1039 root->fs_info->running_transaction->transid) {
b9447ef8 1040 btrfs_assert_tree_locked(buf);
b4ce94de 1041
b9473439
CM		 1042 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
1043 spin_lock(&root->fs_info->delalloc_lock);
1044 if (root->fs_info->dirty_metadata_bytes >= buf->len)
1045 root->fs_info->dirty_metadata_bytes -= buf->len;
1046 else
1047 WARN_ON(1);
1048 spin_unlock(&root->fs_info->delalloc_lock);
1049 }
b4ce94de 1050
b9473439
CM		 1051 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1052 btrfs_set_lock_blocking(buf);
d1310b2e 1053 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
55c69072 1054 buf);
925baedd 1055 }
5f39d397
CM		 1056 return 0;
1057}
1058
db94535d 1059static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
87ee04eb 1060 u32 stripesize, struct btrfs_root *root,
9f5fae2f 1061 struct btrfs_fs_info *fs_info,
e20d96d6 1062 u64 objectid)
d97e63b6 1063{
cfaa7295 1064 root->node = NULL;
a28ec197 1065 root->commit_root = NULL;
db94535d
CM		 1066 root->sectorsize = sectorsize;
1067 root->nodesize = nodesize;
1068 root->leafsize = leafsize;
87ee04eb 1069 root->stripesize = stripesize;
123abc88 1070 root->ref_cows = 0;
0b86a832 1071 root->track_dirty = 0;
c71bf099 1072 root->in_radix = 0;
d68fc57b
YZ		 1073 root->orphan_item_inserted = 0;
1074 root->orphan_cleanup_state = 0;
0b86a832 1075
9f5fae2f 1076 root->fs_info = fs_info;
0f7d52f4
CM		 1077 root->objectid = objectid;
1078 root->last_trans = 0;
13a8a7c8 1079 root->highest_objectid = 0;
58176a96 1080 root->name = NULL;
6bef4d31 1081 root->inode_tree = RB_ROOT;
16cdcec7 1082 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1083 root->block_rsv = NULL;
d68fc57b 1084 root->orphan_block_rsv = NULL;
0b86a832
CM		 1085
1086 INIT_LIST_HEAD(&root->dirty_list);
7b128766 1087 INIT_LIST_HEAD(&root->orphan_list);
5d4f98a2 1088 INIT_LIST_HEAD(&root->root_list);
d68fc57b 1089 spin_lock_init(&root->orphan_lock);
5d4f98a2 1090 spin_lock_init(&root->inode_lock);
f0486c68 1091 spin_lock_init(&root->accounting_lock);
a2135011 1092 mutex_init(&root->objectid_mutex);
e02119d5 1093 mutex_init(&root->log_mutex);
7237f183
YZ		 1094 init_waitqueue_head(&root->log_writer_wait);
1095 init_waitqueue_head(&root->log_commit_wait[0]);
1096 init_waitqueue_head(&root->log_commit_wait[1]);
1097 atomic_set(&root->log_commit[0], 0);
1098 atomic_set(&root->log_commit[1], 0);
1099 atomic_set(&root->log_writers, 0);
1100 root->log_batch = 0;
1101 root->log_transid = 0;
257c62e1 1102 root->last_log_commit = 0;
d0c803c4 1103 extent_io_tree_init(&root->dirty_log_pages,
f993c883 1104 fs_info->btree_inode->i_mapping);
017e5369 1105
3768f368
CM		 1106 memset(&root->root_key, 0, sizeof(root->root_key));
1107 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1108 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1109 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
3f157a2f 1110 root->defrag_trans_start = fs_info->generation;
58176a96 1111 init_completion(&root->kobj_unregister);
6702ed49 1112 root->defrag_running = 0;
4d775673 1113 root->root_key.objectid = objectid;
0ee5dc67 1114 root->anon_dev = 0;
3768f368
CM		 1115 return 0;
1116}
1117
db94535d 1118static int find_and_setup_root(struct btrfs_root *tree_root,
9f5fae2f
CM		 1119 struct btrfs_fs_info *fs_info,
1120 u64 objectid,
e20d96d6 1121 struct btrfs_root *root)
3768f368
CM		 1122{
1123 int ret;
db94535d 1124 u32 blocksize;
84234f3a 1125 u64 generation;
3768f368 1126
db94535d 1127 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 1128 tree_root->sectorsize, tree_root->stripesize,
1129 root, fs_info, objectid);
3768f368
CM		 1130 ret = btrfs_find_last_root(tree_root, objectid,
1131 &root->root_item, &root->root_key);
4df27c4d
YZ		 1132 if (ret > 0)
1133 return -ENOENT;
3768f368
CM		 1134 BUG_ON(ret);
1135
84234f3a 1136 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1137 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1138 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1139 blocksize, generation);
68433b73
CM		 1140 if (!root->node || !btrfs_buffer_uptodate(root->node, generation)) {
1141 free_extent_buffer(root->node);
1142 return -EIO;
1143 }
4df27c4d 1144 root->commit_root = btrfs_root_node(root);
d97e63b6
CM		 1145 return 0;
1146}
1147
7237f183
YZ		 1148static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1149 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1150{
1151 struct btrfs_root *root;
1152 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1153 struct extent_buffer *leaf;
e02119d5
CM		 1154
1155 root = kzalloc(sizeof(*root), GFP_NOFS);
1156 if (!root)
7237f183 1157 return ERR_PTR(-ENOMEM);
e02119d5
CM		 1158
1159 __setup_root(tree_root->nodesize, tree_root->leafsize,
1160 tree_root->sectorsize, tree_root->stripesize,
1161 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1162
1163 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1164 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1165 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
7237f183
YZ		 1166 /*
1167 * log trees do not get reference counted because they go away
1168 * before a real commit is actually done. They do store pointers
1169 * to file data extents, and those reference counts still get
1170 * updated (along with back refs to the log tree).
1171 */
e02119d5
CM		 1172 root->ref_cows = 0;
1173
5d4f98a2
YZ		 1174 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
1175 BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0);
7237f183
YZ		 1176 if (IS_ERR(leaf)) {
1177 kfree(root);
1178 return ERR_CAST(leaf);
1179 }
e02119d5 1180
5d4f98a2
YZ		 1181 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1182 btrfs_set_header_bytenr(leaf, leaf->start);
1183 btrfs_set_header_generation(leaf, trans->transid);
1184 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1185 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1186 root->node = leaf;
e02119d5
CM		 1187
1188 write_extent_buffer(root->node, root->fs_info->fsid,
1189 (unsigned long)btrfs_header_fsid(root->node),
1190 BTRFS_FSID_SIZE);
1191 btrfs_mark_buffer_dirty(root->node);
1192 btrfs_tree_unlock(root->node);
7237f183
YZ		 1193 return root;
1194}
1195
1196int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1197 struct btrfs_fs_info *fs_info)
1198{
1199 struct btrfs_root *log_root;
1200
1201 log_root = alloc_log_tree(trans, fs_info);
1202 if (IS_ERR(log_root))
1203 return PTR_ERR(log_root);
1204 WARN_ON(fs_info->log_root_tree);
1205 fs_info->log_root_tree = log_root;
1206 return 0;
1207}
1208
1209int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1210 struct btrfs_root *root)
1211{
1212 struct btrfs_root *log_root;
1213 struct btrfs_inode_item *inode_item;
1214
1215 log_root = alloc_log_tree(trans, root->fs_info);
1216 if (IS_ERR(log_root))
1217 return PTR_ERR(log_root);
1218
1219 log_root->last_trans = trans->transid;
1220 log_root->root_key.offset = root->root_key.objectid;
1221
1222 inode_item = &log_root->root_item.inode;
1223 inode_item->generation = cpu_to_le64(1);
1224 inode_item->size = cpu_to_le64(3);
1225 inode_item->nlink = cpu_to_le32(1);
1226 inode_item->nbytes = cpu_to_le64(root->leafsize);
1227 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1228
5d4f98a2 1229 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1230
1231 WARN_ON(root->log_root);
1232 root->log_root = log_root;
1233 root->log_transid = 0;
257c62e1 1234 root->last_log_commit = 0;
e02119d5
CM		 1235 return 0;
1236}
1237
1238struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
1239 struct btrfs_key *location)
1240{
1241 struct btrfs_root *root;
1242 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1243 struct btrfs_path *path;
5f39d397 1244 struct extent_buffer *l;
84234f3a 1245 u64 generation;
db94535d 1246 u32 blocksize;
0f7d52f4
CM		 1247 int ret = 0;
1248
5eda7b5e 1249 root = kzalloc(sizeof(*root), GFP_NOFS);
0cf6c620 1250 if (!root)
0f7d52f4 1251 return ERR_PTR(-ENOMEM);
0f7d52f4 1252 if (location->offset == (u64)-1) {
db94535d 1253 ret = find_and_setup_root(tree_root, fs_info,
0f7d52f4
CM		 1254 location->objectid, root);
1255 if (ret) {
0f7d52f4
CM		 1256 kfree(root);
1257 return ERR_PTR(ret);
1258 }
13a8a7c8 1259 goto out;
0f7d52f4
CM		 1260 }
1261
db94535d 1262 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 1263 tree_root->sectorsize, tree_root->stripesize,
1264 root, fs_info, location->objectid);
0f7d52f4
CM		 1265
1266 path = btrfs_alloc_path();
db5b493a
TI		 1267 if (!path) {
1268 kfree(root);
1269 return ERR_PTR(-ENOMEM);
1270 }
0f7d52f4 1271 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
13a8a7c8
YZ		 1272 if (ret == 0) {
1273 l = path->nodes[0];
1274 read_extent_buffer(l, &root->root_item,
1275 btrfs_item_ptr_offset(l, path->slots[0]),
1276 sizeof(root->root_item));
1277 memcpy(&root->root_key, location, sizeof(*location));
0f7d52f4 1278 }
0f7d52f4
CM		 1279 btrfs_free_path(path);
1280 if (ret) {
5e540f77 1281 kfree(root);
13a8a7c8
YZ		 1282 if (ret > 0)
1283 ret = -ENOENT;
0f7d52f4
CM		 1284 return ERR_PTR(ret);
1285 }
13a8a7c8 1286
84234f3a 1287 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1288 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1289 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1290 blocksize, generation);
5d4f98a2 1291 root->commit_root = btrfs_root_node(root);
0f7d52f4 1292 BUG_ON(!root->node);
13a8a7c8 1293out:
08fe4db1 1294 if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
e02119d5 1295 root->ref_cows = 1;
08fe4db1
LZ		 1296 btrfs_check_and_init_root_item(&root->root_item);
1297 }
13a8a7c8 1298
5eda7b5e
CM		 1299 return root;
1300}
1301
edbd8d4e
CM		 1302struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
1303 struct btrfs_key *location)
5eda7b5e
CM		 1304{
1305 struct btrfs_root *root;
1306 int ret;
1307
edbd8d4e
CM		 1308 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1309 return fs_info->tree_root;
1310 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1311 return fs_info->extent_root;
8f18cf13
CM		 1312 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1313 return fs_info->chunk_root;
1314 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1315 return fs_info->dev_root;
0403e47e
YZ		 1316 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1317 return fs_info->csum_root;
4df27c4d
YZ		 1318again:
1319 spin_lock(&fs_info->fs_roots_radix_lock);
5eda7b5e
CM		 1320 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1321 (unsigned long)location->objectid);
4df27c4d 1322 spin_unlock(&fs_info->fs_roots_radix_lock);
5eda7b5e
CM		 1323 if (root)
1324 return root;
1325
e02119d5 1326 root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
5eda7b5e
CM		 1327 if (IS_ERR(root))
1328 return root;
3394e160 1329
581bb050 1330 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
581bb050
LZ		 1331 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1332 GFP_NOFS);
35a30d7c
DS		 1333 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1334 ret = -ENOMEM;
581bb050 1335 goto fail;
35a30d7c 1336 }
581bb050
LZ		 1337
1338 btrfs_init_free_ino_ctl(root);
1339 mutex_init(&root->fs_commit_mutex);
1340 spin_lock_init(&root->cache_lock);
1341 init_waitqueue_head(&root->cache_wait);
1342
0ee5dc67 1343 ret = get_anon_bdev(&root->anon_dev);
ac08aedf
CM		 1344 if (ret)
1345 goto fail;
3394e160 1346
d68fc57b
YZ		 1347 if (btrfs_root_refs(&root->root_item) == 0) {
1348 ret = -ENOENT;
1349 goto fail;
1350 }
1351
1352 ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
1353 if (ret < 0)
1354 goto fail;
1355 if (ret == 0)
1356 root->orphan_item_inserted = 1;
1357
4df27c4d
YZ		 1358 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1359 if (ret)
1360 goto fail;
1361
1362 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 1363 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1364 (unsigned long)root->root_key.objectid,
0f7d52f4 1365 root);
d68fc57b 1366 if (ret == 0)
4df27c4d 1367 root->in_radix = 1;
d68fc57b 1368
4df27c4d
YZ		 1369 spin_unlock(&fs_info->fs_roots_radix_lock);
1370 radix_tree_preload_end();
0f7d52f4 1371 if (ret) {
4df27c4d
YZ		 1372 if (ret == -EEXIST) {
1373 free_fs_root(root);
1374 goto again;
1375 }
1376 goto fail;
0f7d52f4 1377 }
4df27c4d
YZ		 1378
1379 ret = btrfs_find_dead_roots(fs_info->tree_root,
1380 root->root_key.objectid);
1381 WARN_ON(ret);
edbd8d4e 1382 return root;
4df27c4d
YZ		 1383fail:
1384 free_fs_root(root);
1385 return ERR_PTR(ret);
edbd8d4e
CM		 1386}
1387
04160088
CM		 1388static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1389{
1390 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1391 int ret = 0;
04160088
CM		 1392 struct btrfs_device *device;
1393 struct backing_dev_info *bdi;
b7967db7 1394
1f78160c
XG		 1395 rcu_read_lock();
1396 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1397 if (!device->bdev)
1398 continue;
04160088
CM		 1399 bdi = blk_get_backing_dev_info(device->bdev);
1400 if (bdi && bdi_congested(bdi, bdi_bits)) {
1401 ret = 1;
1402 break;
1403 }
1404 }
1f78160c 1405 rcu_read_unlock();
04160088
CM		 1406 return ret;
1407}
1408
ad081f14
JA		 1409/*
1410 * If this fails, caller must call bdi_destroy() to get rid of the
1411 * bdi again.
1412 */
04160088
CM		 1413static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1414{
ad081f14
JA		 1415 int err;
1416
1417 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1418 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1419 if (err)
1420 return err;
1421
4575c9cc 1422 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1423 bdi->congested_fn = btrfs_congested_fn;
1424 bdi->congested_data = info;
1425 return 0;
1426}
1427
ce9adaa5
CM		 1428static int bio_ready_for_csum(struct bio *bio)
1429{
1430 u64 length = 0;
1431 u64 buf_len = 0;
1432 u64 start = 0;
1433 struct page *page;
1434 struct extent_io_tree *io_tree = NULL;
ce9adaa5
CM		 1435 struct bio_vec *bvec;
1436 int i;
1437 int ret;
1438
1439 bio_for_each_segment(bvec, bio, i) {
1440 page = bvec->bv_page;
1441 if (page->private == EXTENT_PAGE_PRIVATE) {
1442 length += bvec->bv_len;
1443 continue;
1444 }
1445 if (!page->private) {
1446 length += bvec->bv_len;
1447 continue;
1448 }
1449 length = bvec->bv_len;
1450 buf_len = page->private >> 2;
1451 start = page_offset(page) + bvec->bv_offset;
1452 io_tree = &BTRFS_I(page->mapping->host)->io_tree;
ce9adaa5
CM		 1453 }
1454 /* are we fully contained in this bio? */
1455 if (buf_len <= length)
1456 return 1;
1457
1458 ret = extent_range_uptodate(io_tree, start + length,
1459 start + buf_len - 1);
ce9adaa5
CM		 1460 return ret;
1461}
1462
8b712842
CM		 1463/*
1464 * called by the kthread helper functions to finally call the bio end_io
1465 * functions. This is where read checksum verification actually happens
1466 */
1467static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1468{
ce9adaa5 1469 struct bio *bio;
8b712842
CM		 1470 struct end_io_wq *end_io_wq;
1471 struct btrfs_fs_info *fs_info;
ce9adaa5 1472 int error;
ce9adaa5 1473
8b712842
CM		 1474 end_io_wq = container_of(work, struct end_io_wq, work);
1475 bio = end_io_wq->bio;
1476 fs_info = end_io_wq->info;
ce9adaa5 1477
cad321ad 1478 /* metadata bio reads are special because the whole tree block must
8b712842
CM		 1479 * be checksummed at once. This makes sure the entire block is in
1480 * ram and up to date before trying to verify things. For
1481 * blocksize <= pagesize, it is basically a noop
1482 */
7b6d91da 1483 if (!(bio->bi_rw & REQ_WRITE) && end_io_wq->metadata &&
cad321ad 1484 !bio_ready_for_csum(bio)) {
d20f7043 1485 btrfs_queue_worker(&fs_info->endio_meta_workers,
8b712842
CM		 1486 &end_io_wq->work);
1487 return;
1488 }
1489 error = end_io_wq->error;
1490 bio->bi_private = end_io_wq->private;
1491 bio->bi_end_io = end_io_wq->end_io;
1492 kfree(end_io_wq);
8b712842 1493 bio_endio(bio, error);
44b8bd7e
CM		 1494}
1495
a74a4b97
CM		 1496static int cleaner_kthread(void *arg)
1497{
1498 struct btrfs_root *root = arg;
1499
1500 do {
a74a4b97 1501 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
76dda93c
YZ		 1502
1503 if (!(root->fs_info->sb->s_flags & MS_RDONLY) &&
1504 mutex_trylock(&root->fs_info->cleaner_mutex)) {
24bbcf04 1505 btrfs_run_delayed_iputs(root);
76dda93c
YZ		 1506 btrfs_clean_old_snapshots(root);
1507 mutex_unlock(&root->fs_info->cleaner_mutex);
4cb5300b 1508 btrfs_run_defrag_inodes(root->fs_info);
76dda93c 1509 }
a74a4b97
CM		 1510
1511 if (freezing(current)) {
1512 refrigerator();
1513 } else {
a74a4b97 1514 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1515 if (!kthread_should_stop())
1516 schedule();
a74a4b97
CM		 1517 __set_current_state(TASK_RUNNING);
1518 }
1519 } while (!kthread_should_stop());
1520 return 0;
1521}
1522
1523static int transaction_kthread(void *arg)
1524{
1525 struct btrfs_root *root = arg;
1526 struct btrfs_trans_handle *trans;
1527 struct btrfs_transaction *cur;
8929ecfa 1528 u64 transid;
a74a4b97
CM		 1529 unsigned long now;
1530 unsigned long delay;
1531 int ret;
1532
1533 do {
a74a4b97
CM		 1534 delay = HZ * 30;
1535 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1536 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1537
a4abeea4 1538 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1539 cur = root->fs_info->running_transaction;
1540 if (!cur) {
a4abeea4 1541 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1542 goto sleep;
1543 }
31153d81 1544
a74a4b97 1545 now = get_seconds();
8929ecfa
YZ		 1546 if (!cur->blocked &&
1547 (now < cur->start_time || now - cur->start_time < 30)) {
a4abeea4 1548 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1549 delay = HZ * 5;
1550 goto sleep;
1551 }
8929ecfa 1552 transid = cur->transid;
a4abeea4 1553 spin_unlock(&root->fs_info->trans_lock);
56bec294 1554
7a7eaa40 1555 trans = btrfs_join_transaction(root);
3612b495 1556 BUG_ON(IS_ERR(trans));
8929ecfa
YZ		 1557 if (transid == trans->transid) {
1558 ret = btrfs_commit_transaction(trans, root);
1559 BUG_ON(ret);
1560 } else {
1561 btrfs_end_transaction(trans, root);
1562 }
a74a4b97
CM		 1563sleep:
1564 wake_up_process(root->fs_info->cleaner_kthread);
1565 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1566
1567 if (freezing(current)) {
1568 refrigerator();
1569 } else {
a74a4b97 1570 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1571 if (!kthread_should_stop() &&
1572 !btrfs_transaction_blocked(root->fs_info))
1573 schedule_timeout(delay);
a74a4b97
CM		 1574 __set_current_state(TASK_RUNNING);
1575 }
1576 } while (!kthread_should_stop());
1577 return 0;
1578}
1579
8a4b83cc 1580struct btrfs_root *open_ctree(struct super_block *sb,
dfe25020
CM		 1581 struct btrfs_fs_devices *fs_devices,
1582 char *options)
2e635a27 1583{
db94535d
CM		 1584 u32 sectorsize;
1585 u32 nodesize;
1586 u32 leafsize;
1587 u32 blocksize;
87ee04eb 1588 u32 stripesize;
84234f3a 1589 u64 generation;
f2b636e8 1590 u64 features;
3de4586c 1591 struct btrfs_key location;
a061fc8d 1592 struct buffer_head *bh;
e02119d5 1593 struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
e20d96d6 1594 GFP_NOFS);
d20f7043
CM		 1595 struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root),
1596 GFP_NOFS);
450ba0ea 1597 struct btrfs_root *tree_root = btrfs_sb(sb);
4891aca2 1598 struct btrfs_fs_info *fs_info = NULL;
e02119d5 1599 struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
0b86a832 1600 GFP_NOFS);
e02119d5 1601 struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
0b86a832 1602 GFP_NOFS);
e02119d5
CM		 1603 struct btrfs_root *log_tree_root;
1604
eb60ceac 1605 int ret;
e58ca020 1606 int err = -EINVAL;
4543df7e 1607
2c90e5d6 1608 struct btrfs_super_block *disk_super;
8790d502 1609
4891aca2 1610 if (!extent_root || !tree_root || !tree_root->fs_info ||
d20f7043 1611 !chunk_root || !dev_root || !csum_root) {
39279cc3
CM		 1612 err = -ENOMEM;
1613 goto fail;
1614 }
4891aca2 1615 fs_info = tree_root->fs_info;
76dda93c
YZ		 1616
1617 ret = init_srcu_struct(&fs_info->subvol_srcu);
1618 if (ret) {
1619 err = ret;
1620 goto fail;
1621 }
1622
1623 ret = setup_bdi(fs_info, &fs_info->bdi);
1624 if (ret) {
1625 err = ret;
1626 goto fail_srcu;
1627 }
1628
1629 fs_info->btree_inode = new_inode(sb);
1630 if (!fs_info->btree_inode) {
1631 err = -ENOMEM;
1632 goto fail_bdi;
1633 }
1634
a6591715 1635 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 1636
76dda93c 1637 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
8fd17795 1638 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 1639 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 1640 INIT_LIST_HEAD(&fs_info->delayed_iputs);
19c00ddc 1641 INIT_LIST_HEAD(&fs_info->hashers);
ea8c2819 1642 INIT_LIST_HEAD(&fs_info->delalloc_inodes);
5a3f23d5 1643 INIT_LIST_HEAD(&fs_info->ordered_operations);
11833d66 1644 INIT_LIST_HEAD(&fs_info->caching_block_groups);
1832a6d5 1645 spin_lock_init(&fs_info->delalloc_lock);
a4abeea4 1646 spin_lock_init(&fs_info->trans_lock);
31153d81 1647 spin_lock_init(&fs_info->ref_cache_lock);
76dda93c 1648 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 1649 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 1650 spin_lock_init(&fs_info->defrag_inodes_lock);
7585717f 1651 mutex_init(&fs_info->reloc_mutex);
19c00ddc 1652
58176a96 1653 init_completion(&fs_info->kobj_unregister);
9f5fae2f
CM		 1654 fs_info->tree_root = tree_root;
1655 fs_info->extent_root = extent_root;
d20f7043 1656 fs_info->csum_root = csum_root;
0b86a832
CM		 1657 fs_info->chunk_root = chunk_root;
1658 fs_info->dev_root = dev_root;
8a4b83cc 1659 fs_info->fs_devices = fs_devices;
0b86a832 1660 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 1661 INIT_LIST_HEAD(&fs_info->space_info);
0b86a832 1662 btrfs_mapping_init(&fs_info->mapping_tree);
f0486c68
YZ		 1663 btrfs_init_block_rsv(&fs_info->global_block_rsv);
1664 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv);
1665 btrfs_init_block_rsv(&fs_info->trans_block_rsv);
1666 btrfs_init_block_rsv(&fs_info->chunk_block_rsv);
1667 btrfs_init_block_rsv(&fs_info->empty_block_rsv);
cb03c743 1668 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 1669 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 1670 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 1671 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 1672 atomic_set(&fs_info->defrag_running, 0);
e20d96d6 1673 fs_info->sb = sb;
6f568d35 1674 fs_info->max_inline = 8192 * 1024;
9ed74f2d 1675 fs_info->metadata_ratio = 0;
4cb5300b 1676 fs_info->defrag_inodes = RB_ROOT;
a4abeea4 1677 fs_info->trans_no_join = 0;
c8b97818 1678
b34b086c
CM		 1679 fs_info->thread_pool_size = min_t(unsigned long,
1680 num_online_cpus() + 2, 8);
0afbaf8c 1681
3eaa2885
CM		 1682 INIT_LIST_HEAD(&fs_info->ordered_extents);
1683 spin_lock_init(&fs_info->ordered_extent_lock);
16cdcec7
MX		 1684 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
1685 GFP_NOFS);
1686 if (!fs_info->delayed_root) {
1687 err = -ENOMEM;
1688 goto fail_iput;
1689 }
1690 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 1691
a2de733c
AJ		 1692 mutex_init(&fs_info->scrub_lock);
1693 atomic_set(&fs_info->scrubs_running, 0);
1694 atomic_set(&fs_info->scrub_pause_req, 0);
1695 atomic_set(&fs_info->scrubs_paused, 0);
1696 atomic_set(&fs_info->scrub_cancel_req, 0);
1697 init_waitqueue_head(&fs_info->scrub_pause_wait);
1698 init_rwsem(&fs_info->scrub_super_lock);
1699 fs_info->scrub_workers_refcnt = 0;
a2de733c 1700
a061fc8d
CM		 1701 sb->s_blocksize = 4096;
1702 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 1703 sb->s_bdi = &fs_info->bdi;
a061fc8d 1704
76dda93c
YZ		 1705 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
1706 fs_info->btree_inode->i_nlink = 1;
0afbaf8c
CM		 1707 /*
1708 * we set the i_size on the btree inode to the max possible int.
1709 * the real end of the address space is determined by all of
1710 * the devices in the system
1711 */
1712 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 1713 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 1714 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
1715
5d4f98a2 1716 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 1717 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 1718 fs_info->btree_inode->i_mapping);
a8067e02 1719 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM		 1720
1721 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 1722
76dda93c
YZ		 1723 BTRFS_I(fs_info->btree_inode)->root = tree_root;
1724 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
1725 sizeof(struct btrfs_key));
1726 BTRFS_I(fs_info->btree_inode)->dummy_inode = 1;
c65ddb52 1727 insert_inode_hash(fs_info->btree_inode);
76dda93c 1728
0f9dd46c 1729 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 1730 fs_info->block_group_cache_tree = RB_ROOT;
0f9dd46c 1731
11833d66 1732 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 1733 fs_info->btree_inode->i_mapping);
11833d66 1734 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 1735 fs_info->btree_inode->i_mapping);
11833d66 1736 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 1737 fs_info->do_barriers = 1;
e18e4809 1738
39279cc3 1739
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;
16cdcec7 1764 goto fail_alloc;
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))
16cdcec7 1776 goto fail_alloc;
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;
16cdcec7 1792 goto fail_alloc;
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;
16cdcec7 1802 goto fail_alloc;
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;
16cdcec7 1818 goto fail_alloc;
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 1836
bab39bf9
JB		 1837 btrfs_init_workers(&fs_info->caching_workers, "cache",
1838 2, &fs_info->generic_worker);
1839
61b49440
CM		 1840 /* a higher idle thresh on the submit workers makes it much more
1841 * likely that bios will be send down in a sane order to the
1842 * devices
1843 */
1844 fs_info->submit_workers.idle_thresh = 64;
53863232 1845
771ed689 1846 fs_info->workers.idle_thresh = 16;
4a69a410 1847 fs_info->workers.ordered = 1;
61b49440 1848
771ed689
CM		 1849 fs_info->delalloc_workers.idle_thresh = 2;
1850 fs_info->delalloc_workers.ordered = 1;
1851
61d92c32
CM		 1852 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
1853 &fs_info->generic_worker);
5443be45 1854 btrfs_init_workers(&fs_info->endio_workers, "endio",
61d92c32
CM		 1855 fs_info->thread_pool_size,
1856 &fs_info->generic_worker);
d20f7043 1857 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
61d92c32
CM		 1858 fs_info->thread_pool_size,
1859 &fs_info->generic_worker);
cad321ad 1860 btrfs_init_workers(&fs_info->endio_meta_write_workers,
61d92c32
CM		 1861 "endio-meta-write", fs_info->thread_pool_size,
1862 &fs_info->generic_worker);
5443be45 1863 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
61d92c32
CM		 1864 fs_info->thread_pool_size,
1865 &fs_info->generic_worker);
0cb59c99
JB		 1866 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
1867 1, &fs_info->generic_worker);
16cdcec7
MX		 1868 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
1869 fs_info->thread_pool_size,
1870 &fs_info->generic_worker);
61b49440
CM		 1871
1872 /*
1873 * endios are largely parallel and should have a very
1874 * low idle thresh
1875 */
1876 fs_info->endio_workers.idle_thresh = 4;
b51912c9
CM		 1877 fs_info->endio_meta_workers.idle_thresh = 4;
1878
9042846b
CM		 1879 fs_info->endio_write_workers.idle_thresh = 2;
1880 fs_info->endio_meta_write_workers.idle_thresh = 2;
1881
4543df7e 1882 btrfs_start_workers(&fs_info->workers, 1);
61d92c32 1883 btrfs_start_workers(&fs_info->generic_worker, 1);
1cc127b5 1884 btrfs_start_workers(&fs_info->submit_workers, 1);
771ed689 1885 btrfs_start_workers(&fs_info->delalloc_workers, 1);
247e743c 1886 btrfs_start_workers(&fs_info->fixup_workers, 1);
9042846b
CM		 1887 btrfs_start_workers(&fs_info->endio_workers, 1);
1888 btrfs_start_workers(&fs_info->endio_meta_workers, 1);
1889 btrfs_start_workers(&fs_info->endio_meta_write_workers, 1);
1890 btrfs_start_workers(&fs_info->endio_write_workers, 1);
0cb59c99 1891 btrfs_start_workers(&fs_info->endio_freespace_worker, 1);
16cdcec7 1892 btrfs_start_workers(&fs_info->delayed_workers, 1);
bab39bf9 1893 btrfs_start_workers(&fs_info->caching_workers, 1);
4543df7e 1894
4575c9cc 1895 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 1896 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
1897 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 1898
db94535d
CM		 1899 nodesize = btrfs_super_nodesize(disk_super);
1900 leafsize = btrfs_super_leafsize(disk_super);
1901 sectorsize = btrfs_super_sectorsize(disk_super);
87ee04eb 1902 stripesize = btrfs_super_stripesize(disk_super);
db94535d
CM		 1903 tree_root->nodesize = nodesize;
1904 tree_root->leafsize = leafsize;
1905 tree_root->sectorsize = sectorsize;
87ee04eb 1906 tree_root->stripesize = stripesize;
a061fc8d
CM		 1907
1908 sb->s_blocksize = sectorsize;
1909 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 1910
39279cc3
CM		 1911 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
1912 sizeof(disk_super->magic))) {
d397712b 1913 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 1914 goto fail_sb_buffer;
1915 }
19c00ddc 1916
925baedd 1917 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 1918 ret = btrfs_read_sys_array(tree_root);
925baedd 1919 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 1920 if (ret) {
d397712b
CM		 1921 printk(KERN_WARNING "btrfs: failed to read the system "
1922 "array on %s\n", sb->s_id);
5d4f98a2 1923 goto fail_sb_buffer;
84eed90f 1924 }
0b86a832
CM		 1925
1926 blocksize = btrfs_level_size(tree_root,
1927 btrfs_super_chunk_root_level(disk_super));
84234f3a 1928 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 1929
1930 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1931 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1932
1933 chunk_root->node = read_tree_block(chunk_root,
1934 btrfs_super_chunk_root(disk_super),
84234f3a 1935 blocksize, generation);
0b86a832 1936 BUG_ON(!chunk_root->node);
83121942
DW		 1937 if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
1938 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
1939 sb->s_id);
1940 goto fail_chunk_root;
1941 }
5d4f98a2
YZ		 1942 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
1943 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 1944
e17cade2 1945 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
d397712b
CM		 1946 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
1947 BTRFS_UUID_SIZE);
e17cade2 1948
925baedd 1949 mutex_lock(&fs_info->chunk_mutex);
0b86a832 1950 ret = btrfs_read_chunk_tree(chunk_root);
925baedd 1951 mutex_unlock(&fs_info->chunk_mutex);
2b82032c 1952 if (ret) {
d397712b
CM		 1953 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
1954 sb->s_id);
2b82032c
YZ		 1955 goto fail_chunk_root;
1956 }
0b86a832 1957
dfe25020
CM		 1958 btrfs_close_extra_devices(fs_devices);
1959
db94535d
CM		 1960 blocksize = btrfs_level_size(tree_root,
1961 btrfs_super_root_level(disk_super));
84234f3a 1962 generation = btrfs_super_generation(disk_super);
0b86a832 1963
e20d96d6 1964 tree_root->node = read_tree_block(tree_root,
db94535d 1965 btrfs_super_root(disk_super),
84234f3a 1966 blocksize, generation);
39279cc3 1967 if (!tree_root->node)
2b82032c 1968 goto fail_chunk_root;
83121942
DW		 1969 if (!test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
1970 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
1971 sb->s_id);
1972 goto fail_tree_root;
1973 }
5d4f98a2
YZ		 1974 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
1975 tree_root->commit_root = btrfs_root_node(tree_root);
db94535d
CM		 1976
1977 ret = find_and_setup_root(tree_root, fs_info,
e20d96d6 1978 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
0b86a832 1979 if (ret)
39279cc3 1980 goto fail_tree_root;
0b86a832
CM		 1981 extent_root->track_dirty = 1;
1982
1983 ret = find_and_setup_root(tree_root, fs_info,
1984 BTRFS_DEV_TREE_OBJECTID, dev_root);
0b86a832
CM		 1985 if (ret)
1986 goto fail_extent_root;
5d4f98a2 1987 dev_root->track_dirty = 1;
3768f368 1988
d20f7043
CM		 1989 ret = find_and_setup_root(tree_root, fs_info,
1990 BTRFS_CSUM_TREE_OBJECTID, csum_root);
1991 if (ret)
5d4f98a2 1992 goto fail_dev_root;
d20f7043
CM		 1993
1994 csum_root->track_dirty = 1;
1995
8929ecfa
YZ		 1996 fs_info->generation = generation;
1997 fs_info->last_trans_committed = generation;
1998 fs_info->data_alloc_profile = (u64)-1;
1999 fs_info->metadata_alloc_profile = (u64)-1;
2000 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
2001
c59021f8 2002 ret = btrfs_init_space_info(fs_info);
2003 if (ret) {
2004 printk(KERN_ERR "Failed to initial space info: %d\n", ret);
2005 goto fail_block_groups;
2006 }
2007
1b1d1f66
JB		 2008 ret = btrfs_read_block_groups(extent_root);
2009 if (ret) {
2010 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
2011 goto fail_block_groups;
2012 }
9078a3e1 2013
a74a4b97
CM		 2014 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2015 "btrfs-cleaner");
57506d50 2016 if (IS_ERR(fs_info->cleaner_kthread))
1b1d1f66 2017 goto fail_block_groups;
a74a4b97
CM		 2018
2019 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2020 tree_root,
2021 "btrfs-transaction");
57506d50 2022 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2023 goto fail_cleaner;
a74a4b97 2024
c289811c
CM		 2025 if (!btrfs_test_opt(tree_root, SSD) &&
2026 !btrfs_test_opt(tree_root, NOSSD) &&
2027 !fs_info->fs_devices->rotating) {
2028 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
2029 "mode\n");
2030 btrfs_set_opt(fs_info->mount_opt, SSD);
2031 }
2032
acce952b 2033 /* do not make disk changes in broken FS */
2034 if (btrfs_super_log_root(disk_super) != 0 &&
2035 !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) {
e02119d5
CM		 2036 u64 bytenr = btrfs_super_log_root(disk_super);
2037
7c2ca468 2038 if (fs_devices->rw_devices == 0) {
d397712b
CM		 2039 printk(KERN_WARNING "Btrfs log replay required "
2040 "on RO media\n");
7c2ca468
CM		 2041 err = -EIO;
2042 goto fail_trans_kthread;
2043 }
e02119d5
CM		 2044 blocksize =
2045 btrfs_level_size(tree_root,
2046 btrfs_super_log_root_level(disk_super));
d18a2c44 2047
676e4c86
DC		 2048 log_tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
2049 if (!log_tree_root) {
2050 err = -ENOMEM;
2051 goto fail_trans_kthread;
2052 }
e02119d5
CM		 2053
2054 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2055 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2056
2057 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 2058 blocksize,
2059 generation + 1);
e02119d5
CM		 2060 ret = btrfs_recover_log_trees(log_tree_root);
2061 BUG_ON(ret);
e556ce2c
YZ		 2062
2063 if (sb->s_flags & MS_RDONLY) {
2064 ret = btrfs_commit_super(tree_root);
2065 BUG_ON(ret);
2066 }
e02119d5 2067 }
1a40e23b 2068
76dda93c
YZ		 2069 ret = btrfs_find_orphan_roots(tree_root);
2070 BUG_ON(ret);
2071
7c2ca468 2072 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b
YZ		 2073 ret = btrfs_cleanup_fs_roots(fs_info);
2074 BUG_ON(ret);
2075
5d4f98a2 2076 ret = btrfs_recover_relocation(tree_root);
d7ce5843
MX		 2077 if (ret < 0) {
2078 printk(KERN_WARNING
2079 "btrfs: failed to recover relocation\n");
2080 err = -EINVAL;
2081 goto fail_trans_kthread;
2082 }
7c2ca468 2083 }
1a40e23b 2084
3de4586c
CM		 2085 location.objectid = BTRFS_FS_TREE_OBJECTID;
2086 location.type = BTRFS_ROOT_ITEM_KEY;
2087 location.offset = (u64)-1;
2088
3de4586c
CM		 2089 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
2090 if (!fs_info->fs_root)
7c2ca468 2091 goto fail_trans_kthread;
3140c9a3
DC		 2092 if (IS_ERR(fs_info->fs_root)) {
2093 err = PTR_ERR(fs_info->fs_root);
2094 goto fail_trans_kthread;
2095 }
c289811c 2096
e3acc2a6
JB		 2097 if (!(sb->s_flags & MS_RDONLY)) {
2098 down_read(&fs_info->cleanup_work_sem);
66b4ffd1
JB		 2099 err = btrfs_orphan_cleanup(fs_info->fs_root);
2100 if (!err)
2101 err = btrfs_orphan_cleanup(fs_info->tree_root);
e3acc2a6 2102 up_read(&fs_info->cleanup_work_sem);
66b4ffd1
JB		 2103 if (err) {
2104 close_ctree(tree_root);
2105 return ERR_PTR(err);
2106 }
e3acc2a6
JB		 2107 }
2108
0f7d52f4 2109 return tree_root;
39279cc3 2110
7c2ca468
CM		 2111fail_trans_kthread:
2112 kthread_stop(fs_info->transaction_kthread);
3f157a2f 2113fail_cleaner:
a74a4b97 2114 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2115
2116 /*
2117 * make sure we're done with the btree inode before we stop our
2118 * kthreads
2119 */
2120 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
2121 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
2122
1b1d1f66
JB		 2123fail_block_groups:
2124 btrfs_free_block_groups(fs_info);
d20f7043 2125 free_extent_buffer(csum_root->node);
5d4f98a2
YZ		 2126 free_extent_buffer(csum_root->commit_root);
2127fail_dev_root:
2128 free_extent_buffer(dev_root->node);
2129 free_extent_buffer(dev_root->commit_root);
0b86a832
CM		 2130fail_extent_root:
2131 free_extent_buffer(extent_root->node);
5d4f98a2 2132 free_extent_buffer(extent_root->commit_root);
39279cc3 2133fail_tree_root:
5f39d397 2134 free_extent_buffer(tree_root->node);
5d4f98a2 2135 free_extent_buffer(tree_root->commit_root);
2b82032c
YZ		 2136fail_chunk_root:
2137 free_extent_buffer(chunk_root->node);
5d4f98a2 2138 free_extent_buffer(chunk_root->commit_root);
39279cc3 2139fail_sb_buffer:
61d92c32 2140 btrfs_stop_workers(&fs_info->generic_worker);
247e743c 2141 btrfs_stop_workers(&fs_info->fixup_workers);
771ed689 2142 btrfs_stop_workers(&fs_info->delalloc_workers);
8b712842
CM		 2143 btrfs_stop_workers(&fs_info->workers);
2144 btrfs_stop_workers(&fs_info->endio_workers);
d20f7043 2145 btrfs_stop_workers(&fs_info->endio_meta_workers);
cad321ad 2146 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
e6dcd2dc 2147 btrfs_stop_workers(&fs_info->endio_write_workers);
0cb59c99 2148 btrfs_stop_workers(&fs_info->endio_freespace_worker);
1cc127b5 2149 btrfs_stop_workers(&fs_info->submit_workers);
16cdcec7 2150 btrfs_stop_workers(&fs_info->delayed_workers);
bab39bf9 2151 btrfs_stop_workers(&fs_info->caching_workers);
16cdcec7
MX		 2152fail_alloc:
2153 kfree(fs_info->delayed_root);
4543df7e 2154fail_iput:
7c2ca468 2155 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
4543df7e 2156 iput(fs_info->btree_inode);
7e662854 2157
dfe25020 2158 btrfs_close_devices(fs_info->fs_devices);
84eed90f 2159 btrfs_mapping_tree_free(&fs_info->mapping_tree);
ad081f14 2160fail_bdi:
7e662854 2161 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 2162fail_srcu:
2163 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 2164fail:
39279cc3
CM		 2165 kfree(extent_root);
2166 kfree(tree_root);
2167 kfree(fs_info);
83afeac4
JM		 2168 kfree(chunk_root);
2169 kfree(dev_root);
d20f7043 2170 kfree(csum_root);
39279cc3 2171 return ERR_PTR(err);
eb60ceac
CM		 2172}
2173
f2984462
CM		 2174static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2175{
2176 char b[BDEVNAME_SIZE];
2177
2178 if (uptodate) {
2179 set_buffer_uptodate(bh);
2180 } else {
7a36ddec 2181 printk_ratelimited(KERN_WARNING "lost page write due to "
f2984462
CM		 2182 "I/O error on %s\n",
2183 bdevname(bh->b_bdev, b));
1259ab75
CM		 2184 /* note, we dont' set_buffer_write_io_error because we have
2185 * our own ways of dealing with the IO errors
2186 */
f2984462
CM		 2187 clear_buffer_uptodate(bh);
2188 }
2189 unlock_buffer(bh);
2190 put_bh(bh);
2191}
2192
a512bbf8
YZ		 2193struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
2194{
2195 struct buffer_head *bh;
2196 struct buffer_head *latest = NULL;
2197 struct btrfs_super_block *super;
2198 int i;
2199 u64 transid = 0;
2200 u64 bytenr;
2201
2202 /* we would like to check all the supers, but that would make
2203 * a btrfs mount succeed after a mkfs from a different FS.
2204 * So, we need to add a special mount option to scan for
2205 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2206 */
2207 for (i = 0; i < 1; i++) {
2208 bytenr = btrfs_sb_offset(i);
2209 if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
2210 break;
2211 bh = __bread(bdev, bytenr / 4096, 4096);
2212 if (!bh)
2213 continue;
2214
2215 super = (struct btrfs_super_block *)bh->b_data;
2216 if (btrfs_super_bytenr(super) != bytenr ||
2217 strncmp((char *)(&super->magic), BTRFS_MAGIC,
2218 sizeof(super->magic))) {
2219 brelse(bh);
2220 continue;
2221 }
2222
2223 if (!latest || btrfs_super_generation(super) > transid) {
2224 brelse(latest);
2225 latest = bh;
2226 transid = btrfs_super_generation(super);
2227 } else {
2228 brelse(bh);
2229 }
2230 }
2231 return latest;
2232}
2233
4eedeb75
HH		 2234/*
2235 * this should be called twice, once with wait == 0 and
2236 * once with wait == 1. When wait == 0 is done, all the buffer heads
2237 * we write are pinned.
2238 *
2239 * They are released when wait == 1 is done.
2240 * max_mirrors must be the same for both runs, and it indicates how
2241 * many supers on this one device should be written.
2242 *
2243 * max_mirrors == 0 means to write them all.
2244 */
a512bbf8
YZ		 2245static int write_dev_supers(struct btrfs_device *device,
2246 struct btrfs_super_block *sb,
2247 int do_barriers, int wait, int max_mirrors)
2248{
2249 struct buffer_head *bh;
2250 int i;
2251 int ret;
2252 int errors = 0;
2253 u32 crc;
2254 u64 bytenr;
2255 int last_barrier = 0;
2256
2257 if (max_mirrors == 0)
2258 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
2259
2260 /* make sure only the last submit_bh does a barrier */
2261 if (do_barriers) {
2262 for (i = 0; i < max_mirrors; i++) {
2263 bytenr = btrfs_sb_offset(i);
2264 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
2265 device->total_bytes)
2266 break;
2267 last_barrier = i;
2268 }
2269 }
2270
2271 for (i = 0; i < max_mirrors; i++) {
2272 bytenr = btrfs_sb_offset(i);
2273 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
2274 break;
2275
2276 if (wait) {
2277 bh = __find_get_block(device->bdev, bytenr / 4096,
2278 BTRFS_SUPER_INFO_SIZE);
2279 BUG_ON(!bh);
a512bbf8 2280 wait_on_buffer(bh);
4eedeb75
HH		 2281 if (!buffer_uptodate(bh))
2282 errors++;
2283
2284 /* drop our reference */
2285 brelse(bh);
2286
2287 /* drop the reference from the wait == 0 run */
2288 brelse(bh);
2289 continue;
a512bbf8
YZ		 2290 } else {
2291 btrfs_set_super_bytenr(sb, bytenr);
2292
2293 crc = ~(u32)0;
2294 crc = btrfs_csum_data(NULL, (char *)sb +
2295 BTRFS_CSUM_SIZE, crc,
2296 BTRFS_SUPER_INFO_SIZE -
2297 BTRFS_CSUM_SIZE);
2298 btrfs_csum_final(crc, sb->csum);
2299
4eedeb75
HH		 2300 /*
2301 * one reference for us, and we leave it for the
2302 * caller
2303 */
a512bbf8
YZ		 2304 bh = __getblk(device->bdev, bytenr / 4096,
2305 BTRFS_SUPER_INFO_SIZE);
2306 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
2307
4eedeb75 2308 /* one reference for submit_bh */
a512bbf8 2309 get_bh(bh);
4eedeb75
HH		 2310
2311 set_buffer_uptodate(bh);
a512bbf8
YZ		 2312 lock_buffer(bh);
2313 bh->b_end_io = btrfs_end_buffer_write_sync;
2314 }
2315
c3b9a62c
CH		 2316 if (i == last_barrier && do_barriers)
2317 ret = submit_bh(WRITE_FLUSH_FUA, bh);
2318 else
ffbd517d 2319 ret = submit_bh(WRITE_SYNC, bh);
a512bbf8 2320
4eedeb75 2321 if (ret)
a512bbf8 2322 errors++;
a512bbf8
YZ		 2323 }
2324 return errors < i ? 0 : -1;
2325}
2326
2327int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 2328{
e5e9a520 2329 struct list_head *head;
f2984462 2330 struct btrfs_device *dev;
a061fc8d 2331 struct btrfs_super_block *sb;
f2984462 2332 struct btrfs_dev_item *dev_item;
f2984462
CM		 2333 int ret;
2334 int do_barriers;
a236aed1
CM		 2335 int max_errors;
2336 int total_errors = 0;
a061fc8d 2337 u64 flags;
f2984462 2338
a236aed1 2339 max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
f2984462
CM		 2340 do_barriers = !btrfs_test_opt(root, NOBARRIER);
2341
a061fc8d
CM		 2342 sb = &root->fs_info->super_for_commit;
2343 dev_item = &sb->dev_item;
e5e9a520 2344
174ba509 2345 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 2346 head = &root->fs_info->fs_devices->devices;
1f78160c 2347 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 2348 if (!dev->bdev) {
2349 total_errors++;
2350 continue;
2351 }
2b82032c 2352 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 2353 continue;
2354
2b82032c 2355 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 2356 btrfs_set_stack_device_type(dev_item, dev->type);
2357 btrfs_set_stack_device_id(dev_item, dev->devid);
2358 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
2359 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
2360 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
2361 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
2362 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
2363 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 2364 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 2365
a061fc8d
CM		 2366 flags = btrfs_super_flags(sb);
2367 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
2368
a512bbf8 2369 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 2370 if (ret)
2371 total_errors++;
f2984462 2372 }
a236aed1 2373 if (total_errors > max_errors) {
d397712b
CM		 2374 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2375 total_errors);
a236aed1
CM		 2376 BUG();
2377 }
f2984462 2378
a512bbf8 2379 total_errors = 0;
1f78160c 2380 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 2381 if (!dev->bdev)
2382 continue;
2b82032c 2383 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 2384 continue;
2385
a512bbf8
YZ		 2386 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
2387 if (ret)
2388 total_errors++;
f2984462 2389 }
174ba509 2390 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 2391 if (total_errors > max_errors) {
d397712b
CM		 2392 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2393 total_errors);
a236aed1
CM		 2394 BUG();
2395 }
f2984462
CM		 2396 return 0;
2397}
2398
a512bbf8
YZ		 2399int write_ctree_super(struct btrfs_trans_handle *trans,
2400 struct btrfs_root *root, int max_mirrors)
eb60ceac 2401{
e66f709b 2402 int ret;
5f39d397 2403
a512bbf8 2404 ret = write_all_supers(root, max_mirrors);
5f39d397 2405 return ret;
cfaa7295
CM		 2406}
2407
5eda7b5e 2408int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
2619ba1f 2409{
4df27c4d 2410 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 2411 radix_tree_delete(&fs_info->fs_roots_radix,
2412 (unsigned long)root->root_key.objectid);
4df27c4d 2413 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 2414
2415 if (btrfs_root_refs(&root->root_item) == 0)
2416 synchronize_srcu(&fs_info->subvol_srcu);
2417
581bb050
LZ		 2418 __btrfs_remove_free_space_cache(root->free_ino_pinned);
2419 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d
YZ		 2420 free_fs_root(root);
2421 return 0;
2422}
2423
2424static void free_fs_root(struct btrfs_root *root)
2425{
82d5902d 2426 iput(root->cache_inode);
4df27c4d 2427 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
0ee5dc67
AV		 2428 if (root->anon_dev)
2429 free_anon_bdev(root->anon_dev);
4df27c4d
YZ		 2430 free_extent_buffer(root->node);
2431 free_extent_buffer(root->commit_root);
581bb050
LZ		 2432 kfree(root->free_ino_ctl);
2433 kfree(root->free_ino_pinned);
d397712b 2434 kfree(root->name);
2619ba1f 2435 kfree(root);
2619ba1f
CM		 2436}
2437
35b7e476 2438static int del_fs_roots(struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 2439{
2440 int ret;
2441 struct btrfs_root *gang[8];
2442 int i;
2443
76dda93c
YZ		 2444 while (!list_empty(&fs_info->dead_roots)) {
2445 gang[0] = list_entry(fs_info->dead_roots.next,
2446 struct btrfs_root, root_list);
2447 list_del(&gang[0]->root_list);
2448
2449 if (gang[0]->in_radix) {
2450 btrfs_free_fs_root(fs_info, gang[0]);
2451 } else {
2452 free_extent_buffer(gang[0]->node);
2453 free_extent_buffer(gang[0]->commit_root);
2454 kfree(gang[0]);
2455 }
2456 }
2457
d397712b 2458 while (1) {
0f7d52f4
CM		 2459 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2460 (void **)gang, 0,
2461 ARRAY_SIZE(gang));
2462 if (!ret)
2463 break;
2619ba1f 2464 for (i = 0; i < ret; i++)
5eda7b5e 2465 btrfs_free_fs_root(fs_info, gang[i]);
0f7d52f4
CM		 2466 }
2467 return 0;
2468}
b4100d64 2469
c146afad 2470int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 2471{
c146afad
YZ		 2472 u64 root_objectid = 0;
2473 struct btrfs_root *gang[8];
2474 int i;
3768f368 2475 int ret;
e089f05c 2476
c146afad
YZ		 2477 while (1) {
2478 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2479 (void **)gang, root_objectid,
2480 ARRAY_SIZE(gang));
2481 if (!ret)
2482 break;
5d4f98a2
YZ		 2483
2484 root_objectid = gang[ret - 1]->root_key.objectid + 1;
c146afad 2485 for (i = 0; i < ret; i++) {
66b4ffd1
JB		 2486 int err;
2487
c146afad 2488 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 2489 err = btrfs_orphan_cleanup(gang[i]);
2490 if (err)
2491 return err;
c146afad
YZ		 2492 }
2493 root_objectid++;
2494 }
2495 return 0;
2496}
a2135011 2497
c146afad
YZ		 2498int btrfs_commit_super(struct btrfs_root *root)
2499{
2500 struct btrfs_trans_handle *trans;
2501 int ret;
a74a4b97 2502
c146afad 2503 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 2504 btrfs_run_delayed_iputs(root);
a74a4b97 2505 btrfs_clean_old_snapshots(root);
c146afad 2506 mutex_unlock(&root->fs_info->cleaner_mutex);
c71bf099
YZ		 2507
2508 /* wait until ongoing cleanup work done */
2509 down_write(&root->fs_info->cleanup_work_sem);
2510 up_write(&root->fs_info->cleanup_work_sem);
2511
7a7eaa40 2512 trans = btrfs_join_transaction(root);
3612b495
TI		 2513 if (IS_ERR(trans))
2514 return PTR_ERR(trans);
54aa1f4d 2515 ret = btrfs_commit_transaction(trans, root);
c146afad
YZ		 2516 BUG_ON(ret);
2517 /* run commit again to drop the original snapshot */
7a7eaa40 2518 trans = btrfs_join_transaction(root);
3612b495
TI		 2519 if (IS_ERR(trans))
2520 return PTR_ERR(trans);
79154b1b
CM		 2521 btrfs_commit_transaction(trans, root);
2522 ret = btrfs_write_and_wait_transaction(NULL, root);
3768f368 2523 BUG_ON(ret);
d6bfde87 2524
a512bbf8 2525 ret = write_ctree_super(NULL, root, 0);
c146afad
YZ		 2526 return ret;
2527}
2528
2529int close_ctree(struct btrfs_root *root)
2530{
2531 struct btrfs_fs_info *fs_info = root->fs_info;
2532 int ret;
2533
2534 fs_info->closing = 1;
2535 smp_mb();
2536
a2de733c 2537 btrfs_scrub_cancel(root);
4cb5300b
CM		 2538
2539 /* wait for any defraggers to finish */
2540 wait_event(fs_info->transaction_wait,
2541 (atomic_read(&fs_info->defrag_running) == 0));
2542
2543 /* clear out the rbtree of defraggable inodes */
2544 btrfs_run_defrag_inodes(root->fs_info);
2545
acce952b 2546 /*
2547 * Here come 2 situations when btrfs is broken to flip readonly:
2548 *
2549 * 1. when btrfs flips readonly somewhere else before
2550 * btrfs_commit_super, sb->s_flags has MS_RDONLY flag,
2551 * and btrfs will skip to write sb directly to keep
2552 * ERROR state on disk.
2553 *
2554 * 2. when btrfs flips readonly just in btrfs_commit_super,
ae0e47f0 2555 * and in such case, btrfs cannot write sb via btrfs_commit_super,
acce952b 2556 * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag,
2557 * btrfs will cleanup all FS resources first and write sb then.
2558 */
c146afad 2559 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 2560 ret = btrfs_commit_super(root);
2561 if (ret)
2562 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
2563 }
2564
2565 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
2566 ret = btrfs_error_commit_super(root);
d397712b
CM		 2567 if (ret)
2568 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
c146afad 2569 }
0f7d52f4 2570
300e4f8a
JB		 2571 btrfs_put_block_group_cache(fs_info);
2572
8929ecfa
YZ		 2573 kthread_stop(root->fs_info->transaction_kthread);
2574 kthread_stop(root->fs_info->cleaner_kthread);
2575
f25784b3
YZ		 2576 fs_info->closing = 2;
2577 smp_mb();
2578
b0c68f8b 2579 if (fs_info->delalloc_bytes) {
d397712b 2580 printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
21380931 2581 (unsigned long long)fs_info->delalloc_bytes);
b0c68f8b 2582 }
31153d81 2583 if (fs_info->total_ref_cache_size) {
d397712b
CM		 2584 printk(KERN_INFO "btrfs: at umount reference cache size %llu\n",
2585 (unsigned long long)fs_info->total_ref_cache_size);
31153d81 2586 }
bcc63abb 2587
5d4f98a2
YZ		 2588 free_extent_buffer(fs_info->extent_root->node);
2589 free_extent_buffer(fs_info->extent_root->commit_root);
2590 free_extent_buffer(fs_info->tree_root->node);
2591 free_extent_buffer(fs_info->tree_root->commit_root);
2592 free_extent_buffer(root->fs_info->chunk_root->node);
2593 free_extent_buffer(root->fs_info->chunk_root->commit_root);
2594 free_extent_buffer(root->fs_info->dev_root->node);
2595 free_extent_buffer(root->fs_info->dev_root->commit_root);
2596 free_extent_buffer(root->fs_info->csum_root->node);
2597 free_extent_buffer(root->fs_info->csum_root->commit_root);
d20f7043 2598
9078a3e1 2599 btrfs_free_block_groups(root->fs_info);
d10c5f31 2600
c146afad 2601 del_fs_roots(fs_info);
d10c5f31 2602
c146afad 2603 iput(fs_info->btree_inode);
16cdcec7 2604 kfree(fs_info->delayed_root);
9ad6b7bc 2605
61d92c32 2606 btrfs_stop_workers(&fs_info->generic_worker);
247e743c 2607 btrfs_stop_workers(&fs_info->fixup_workers);
771ed689 2608 btrfs_stop_workers(&fs_info->delalloc_workers);
8b712842
CM		 2609 btrfs_stop_workers(&fs_info->workers);
2610 btrfs_stop_workers(&fs_info->endio_workers);
d20f7043 2611 btrfs_stop_workers(&fs_info->endio_meta_workers);
cad321ad 2612 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
e6dcd2dc 2613 btrfs_stop_workers(&fs_info->endio_write_workers);
0cb59c99 2614 btrfs_stop_workers(&fs_info->endio_freespace_worker);
1cc127b5 2615 btrfs_stop_workers(&fs_info->submit_workers);
16cdcec7 2616 btrfs_stop_workers(&fs_info->delayed_workers);
bab39bf9 2617 btrfs_stop_workers(&fs_info->caching_workers);
d6bfde87 2618
dfe25020 2619 btrfs_close_devices(fs_info->fs_devices);
0b86a832 2620 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 2621
04160088 2622 bdi_destroy(&fs_info->bdi);
76dda93c 2623 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 2624
0f7d52f4 2625 kfree(fs_info->extent_root);
0f7d52f4 2626 kfree(fs_info->tree_root);
0b86a832
CM		 2627 kfree(fs_info->chunk_root);
2628 kfree(fs_info->dev_root);
d20f7043 2629 kfree(fs_info->csum_root);
83a4d548
LZ		 2630 kfree(fs_info);
2631
eb60ceac
CM		 2632 return 0;
2633}
2634
1259ab75 2635int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
5f39d397 2636{
1259ab75 2637 int ret;
810191ff 2638 struct inode *btree_inode = buf->first_page->mapping->host;
1259ab75 2639
2ac55d41
JB		 2640 ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf,
2641 NULL);
1259ab75
CM		 2642 if (!ret)
2643 return ret;
2644
2645 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
2646 parent_transid);
2647 return !ret;
5f39d397
CM		 2648}
2649
2650int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 2651{
810191ff 2652 struct inode *btree_inode = buf->first_page->mapping->host;
d1310b2e 2653 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
5f39d397
CM		 2654 buf);
2655}
6702ed49 2656
5f39d397
CM		 2657void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
2658{
810191ff 2659 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
5f39d397
CM		 2660 u64 transid = btrfs_header_generation(buf);
2661 struct inode *btree_inode = root->fs_info->btree_inode;
b9473439 2662 int was_dirty;
b4ce94de 2663
b9447ef8 2664 btrfs_assert_tree_locked(buf);
ccd467d6 2665 if (transid != root->fs_info->generation) {
d397712b
CM		 2666 printk(KERN_CRIT "btrfs transid mismatch buffer %llu, "
2667 "found %llu running %llu\n",
db94535d 2668 (unsigned long long)buf->start,
d397712b
CM		 2669 (unsigned long long)transid,
2670 (unsigned long long)root->fs_info->generation);
ccd467d6
CM		 2671 WARN_ON(1);
2672 }
b9473439
CM		 2673 was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
2674 buf);
2675 if (!was_dirty) {
2676 spin_lock(&root->fs_info->delalloc_lock);
2677 root->fs_info->dirty_metadata_bytes += buf->len;
2678 spin_unlock(&root->fs_info->delalloc_lock);
2679 }
eb60ceac
CM		 2680}
2681
d3c2fdcf 2682void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
16cdcec7
MX		 2683{
2684 /*
2685 * looks as though older kernels can get into trouble with
2686 * this code, they end up stuck in balance_dirty_pages forever
2687 */
2688 u64 num_dirty;
2689 unsigned long thresh = 32 * 1024 * 1024;
2690
2691 if (current->flags & PF_MEMALLOC)
2692 return;
2693
2694 btrfs_balance_delayed_items(root);
2695
2696 num_dirty = root->fs_info->dirty_metadata_bytes;
2697
2698 if (num_dirty > thresh) {
2699 balance_dirty_pages_ratelimited_nr(
2700 root->fs_info->btree_inode->i_mapping, 1);
2701 }
2702 return;
2703}
2704
2705void __btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
35b7e476 2706{
188de649
CM		 2707 /*
2708 * looks as though older kernels can get into trouble with
2709 * this code, they end up stuck in balance_dirty_pages forever
2710 */
d6bfde87 2711 u64 num_dirty;
771ed689 2712 unsigned long thresh = 32 * 1024 * 1024;
d6bfde87 2713
6933c02e 2714 if (current->flags & PF_MEMALLOC)
d6bfde87
CM		 2715 return;
2716
585ad2c3
CM		 2717 num_dirty = root->fs_info->dirty_metadata_bytes;
2718
d6bfde87
CM		 2719 if (num_dirty > thresh) {
2720 balance_dirty_pages_ratelimited_nr(
d7fc640e 2721 root->fs_info->btree_inode->i_mapping, 1);
d6bfde87 2722 }
188de649 2723 return;
35b7e476 2724}
6b80053d 2725
ca7a79ad 2726int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 2727{
810191ff 2728 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
ce9adaa5 2729 int ret;
ca7a79ad 2730 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
d397712b 2731 if (ret == 0)
b4ce94de 2732 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
ce9adaa5 2733 return ret;
6b80053d 2734}
0da5468f 2735
4bef0848
CM		 2736int btree_lock_page_hook(struct page *page)
2737{
2738 struct inode *inode = page->mapping->host;
b9473439 2739 struct btrfs_root *root = BTRFS_I(inode)->root;
4bef0848
CM		 2740 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2741 struct extent_buffer *eb;
2742 unsigned long len;
2743 u64 bytenr = page_offset(page);
2744
2745 if (page->private == EXTENT_PAGE_PRIVATE)
2746 goto out;
2747
2748 len = page->private >> 2;
f09d1f60 2749 eb = find_extent_buffer(io_tree, bytenr, len);
4bef0848
CM		 2750 if (!eb)
2751 goto out;
2752
2753 btrfs_tree_lock(eb);
4bef0848 2754 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
b9473439
CM		 2755
2756 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
2757 spin_lock(&root->fs_info->delalloc_lock);
2758 if (root->fs_info->dirty_metadata_bytes >= eb->len)
2759 root->fs_info->dirty_metadata_bytes -= eb->len;
2760 else
2761 WARN_ON(1);
2762 spin_unlock(&root->fs_info->delalloc_lock);
2763 }
2764
4bef0848
CM		 2765 btrfs_tree_unlock(eb);
2766 free_extent_buffer(eb);
2767out:
2768 lock_page(page);
2769 return 0;
2770}
2771
acce952b 2772static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
2773 int read_only)
2774{
2775 if (read_only)
2776 return;
2777
2778 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)
2779 printk(KERN_WARNING "warning: mount fs with errors, "
2780 "running btrfsck is recommended\n");
2781}
2782
2783int btrfs_error_commit_super(struct btrfs_root *root)
2784{
2785 int ret;
2786
2787 mutex_lock(&root->fs_info->cleaner_mutex);
2788 btrfs_run_delayed_iputs(root);
2789 mutex_unlock(&root->fs_info->cleaner_mutex);
2790
2791 down_write(&root->fs_info->cleanup_work_sem);
2792 up_write(&root->fs_info->cleanup_work_sem);
2793
2794 /* cleanup FS via transaction */
2795 btrfs_cleanup_transaction(root);
2796
2797 ret = write_ctree_super(NULL, root, 0);
2798
2799 return ret;
2800}
2801
2802static int btrfs_destroy_ordered_operations(struct btrfs_root *root)
2803{
2804 struct btrfs_inode *btrfs_inode;
2805 struct list_head splice;
2806
2807 INIT_LIST_HEAD(&splice);
2808
2809 mutex_lock(&root->fs_info->ordered_operations_mutex);
2810 spin_lock(&root->fs_info->ordered_extent_lock);
2811
2812 list_splice_init(&root->fs_info->ordered_operations, &splice);
2813 while (!list_empty(&splice)) {
2814 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
2815 ordered_operations);
2816
2817 list_del_init(&btrfs_inode->ordered_operations);
2818
2819 btrfs_invalidate_inodes(btrfs_inode->root);
2820 }
2821
2822 spin_unlock(&root->fs_info->ordered_extent_lock);
2823 mutex_unlock(&root->fs_info->ordered_operations_mutex);
2824
2825 return 0;
2826}
2827
2828static int btrfs_destroy_ordered_extents(struct btrfs_root *root)
2829{
2830 struct list_head splice;
2831 struct btrfs_ordered_extent *ordered;
2832 struct inode *inode;
2833
2834 INIT_LIST_HEAD(&splice);
2835
2836 spin_lock(&root->fs_info->ordered_extent_lock);
2837
2838 list_splice_init(&root->fs_info->ordered_extents, &splice);
2839 while (!list_empty(&splice)) {
2840 ordered = list_entry(splice.next, struct btrfs_ordered_extent,
2841 root_extent_list);
2842
2843 list_del_init(&ordered->root_extent_list);
2844 atomic_inc(&ordered->refs);
2845
2846 /* the inode may be getting freed (in sys_unlink path). */
2847 inode = igrab(ordered->inode);
2848
2849 spin_unlock(&root->fs_info->ordered_extent_lock);
2850 if (inode)
2851 iput(inode);
2852
2853 atomic_set(&ordered->refs, 1);
2854 btrfs_put_ordered_extent(ordered);
2855
2856 spin_lock(&root->fs_info->ordered_extent_lock);
2857 }
2858
2859 spin_unlock(&root->fs_info->ordered_extent_lock);
2860
2861 return 0;
2862}
2863
2864static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
2865 struct btrfs_root *root)
2866{
2867 struct rb_node *node;
2868 struct btrfs_delayed_ref_root *delayed_refs;
2869 struct btrfs_delayed_ref_node *ref;
2870 int ret = 0;
2871
2872 delayed_refs = &trans->delayed_refs;
2873
2874 spin_lock(&delayed_refs->lock);
2875 if (delayed_refs->num_entries == 0) {
cfece4db 2876 spin_unlock(&delayed_refs->lock);
acce952b 2877 printk(KERN_INFO "delayed_refs has NO entry\n");
2878 return ret;
2879 }
2880
2881 node = rb_first(&delayed_refs->root);
2882 while (node) {
2883 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2884 node = rb_next(node);
2885
2886 ref->in_tree = 0;
2887 rb_erase(&ref->rb_node, &delayed_refs->root);
2888 delayed_refs->num_entries--;
2889
2890 atomic_set(&ref->refs, 1);
2891 if (btrfs_delayed_ref_is_head(ref)) {
2892 struct btrfs_delayed_ref_head *head;
2893
2894 head = btrfs_delayed_node_to_head(ref);
2895 mutex_lock(&head->mutex);
2896 kfree(head->extent_op);
2897 delayed_refs->num_heads--;
2898 if (list_empty(&head->cluster))
2899 delayed_refs->num_heads_ready--;
2900 list_del_init(&head->cluster);
2901 mutex_unlock(&head->mutex);
2902 }
2903
2904 spin_unlock(&delayed_refs->lock);
2905 btrfs_put_delayed_ref(ref);
2906
2907 cond_resched();
2908 spin_lock(&delayed_refs->lock);
2909 }
2910
2911 spin_unlock(&delayed_refs->lock);
2912
2913 return ret;
2914}
2915
2916static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t)
2917{
2918 struct btrfs_pending_snapshot *snapshot;
2919 struct list_head splice;
2920
2921 INIT_LIST_HEAD(&splice);
2922
2923 list_splice_init(&t->pending_snapshots, &splice);
2924
2925 while (!list_empty(&splice)) {
2926 snapshot = list_entry(splice.next,
2927 struct btrfs_pending_snapshot,
2928 list);
2929
2930 list_del_init(&snapshot->list);
2931
2932 kfree(snapshot);
2933 }
2934
2935 return 0;
2936}
2937
2938static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
2939{
2940 struct btrfs_inode *btrfs_inode;
2941 struct list_head splice;
2942
2943 INIT_LIST_HEAD(&splice);
2944
acce952b 2945 spin_lock(&root->fs_info->delalloc_lock);
5be76758 2946 list_splice_init(&root->fs_info->delalloc_inodes, &splice);
acce952b 2947
2948 while (!list_empty(&splice)) {
2949 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
2950 delalloc_inodes);
2951
2952 list_del_init(&btrfs_inode->delalloc_inodes);
2953
2954 btrfs_invalidate_inodes(btrfs_inode->root);
2955 }
2956
2957 spin_unlock(&root->fs_info->delalloc_lock);
2958
2959 return 0;
2960}
2961
2962static int btrfs_destroy_marked_extents(struct btrfs_root *root,
2963 struct extent_io_tree *dirty_pages,
2964 int mark)
2965{
2966 int ret;
2967 struct page *page;
2968 struct inode *btree_inode = root->fs_info->btree_inode;
2969 struct extent_buffer *eb;
2970 u64 start = 0;
2971 u64 end;
2972 u64 offset;
2973 unsigned long index;
2974
2975 while (1) {
2976 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
2977 mark);
2978 if (ret)
2979 break;
2980
2981 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
2982 while (start <= end) {
2983 index = start >> PAGE_CACHE_SHIFT;
2984 start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
2985 page = find_get_page(btree_inode->i_mapping, index);
2986 if (!page)
2987 continue;
2988 offset = page_offset(page);
2989
2990 spin_lock(&dirty_pages->buffer_lock);
2991 eb = radix_tree_lookup(
2992 &(&BTRFS_I(page->mapping->host)->io_tree)->buffer,
2993 offset >> PAGE_CACHE_SHIFT);
2994 spin_unlock(&dirty_pages->buffer_lock);
2995 if (eb) {
2996 ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY,
2997 &eb->bflags);
2998 atomic_set(&eb->refs, 1);
2999 }
3000 if (PageWriteback(page))
3001 end_page_writeback(page);
3002
3003 lock_page(page);
3004 if (PageDirty(page)) {
3005 clear_page_dirty_for_io(page);
3006 spin_lock_irq(&page->mapping->tree_lock);
3007 radix_tree_tag_clear(&page->mapping->page_tree,
3008 page_index(page),
3009 PAGECACHE_TAG_DIRTY);
3010 spin_unlock_irq(&page->mapping->tree_lock);
3011 }
3012
3013 page->mapping->a_ops->invalidatepage(page, 0);
3014 unlock_page(page);
3015 }
3016 }
3017
3018 return ret;
3019}
3020
3021static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3022 struct extent_io_tree *pinned_extents)
3023{
3024 struct extent_io_tree *unpin;
3025 u64 start;
3026 u64 end;
3027 int ret;
3028
3029 unpin = pinned_extents;
3030 while (1) {
3031 ret = find_first_extent_bit(unpin, 0, &start, &end,
3032 EXTENT_DIRTY);
3033 if (ret)
3034 break;
3035
3036 /* opt_discard */
5378e607
LD		 3037 if (btrfs_test_opt(root, DISCARD))
3038 ret = btrfs_error_discard_extent(root, start,
3039 end + 1 - start,
3040 NULL);
acce952b 3041
3042 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3043 btrfs_error_unpin_extent_range(root, start, end);
3044 cond_resched();
3045 }
3046
3047 return 0;
3048}
3049
3050static int btrfs_cleanup_transaction(struct btrfs_root *root)
3051{
3052 struct btrfs_transaction *t;
3053 LIST_HEAD(list);
3054
3055 WARN_ON(1);
3056
acce952b 3057 mutex_lock(&root->fs_info->transaction_kthread_mutex);
3058
a4abeea4 3059 spin_lock(&root->fs_info->trans_lock);
acce952b 3060 list_splice_init(&root->fs_info->trans_list, &list);
a4abeea4
JB		 3061 root->fs_info->trans_no_join = 1;
3062 spin_unlock(&root->fs_info->trans_lock);
3063
acce952b 3064 while (!list_empty(&list)) {
3065 t = list_entry(list.next, struct btrfs_transaction, list);
3066 if (!t)
3067 break;
3068
3069 btrfs_destroy_ordered_operations(root);
3070
3071 btrfs_destroy_ordered_extents(root);
3072
3073 btrfs_destroy_delayed_refs(t, root);
3074
3075 btrfs_block_rsv_release(root,
3076 &root->fs_info->trans_block_rsv,
3077 t->dirty_pages.dirty_bytes);
3078
3079 /* FIXME: cleanup wait for commit */
3080 t->in_commit = 1;
3081 t->blocked = 1;
3082 if (waitqueue_active(&root->fs_info->transaction_blocked_wait))
3083 wake_up(&root->fs_info->transaction_blocked_wait);
3084
3085 t->blocked = 0;
3086 if (waitqueue_active(&root->fs_info->transaction_wait))
3087 wake_up(&root->fs_info->transaction_wait);
acce952b 3088
acce952b 3089 t->commit_done = 1;
3090 if (waitqueue_active(&t->commit_wait))
3091 wake_up(&t->commit_wait);
acce952b 3092
3093 btrfs_destroy_pending_snapshots(t);
3094
3095 btrfs_destroy_delalloc_inodes(root);
3096
a4abeea4 3097 spin_lock(&root->fs_info->trans_lock);
acce952b 3098 root->fs_info->running_transaction = NULL;
a4abeea4 3099 spin_unlock(&root->fs_info->trans_lock);
acce952b 3100
3101 btrfs_destroy_marked_extents(root, &t->dirty_pages,
3102 EXTENT_DIRTY);
3103
3104 btrfs_destroy_pinned_extent(root,
3105 root->fs_info->pinned_extents);
3106
13c5a93e 3107 atomic_set(&t->use_count, 0);
acce952b 3108 list_del_init(&t->list);
3109 memset(t, 0, sizeof(*t));
3110 kmem_cache_free(btrfs_transaction_cachep, t);
3111 }
3112
a4abeea4
JB		 3113 spin_lock(&root->fs_info->trans_lock);
3114 root->fs_info->trans_no_join = 0;
3115 spin_unlock(&root->fs_info->trans_lock);
acce952b 3116 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 3117
3118 return 0;
3119}
3120
d1310b2e 3121static struct extent_io_ops btree_extent_io_ops = {
4bef0848 3122 .write_cache_pages_lock_hook = btree_lock_page_hook,
ce9adaa5 3123 .readpage_end_io_hook = btree_readpage_end_io_hook,
0b86a832 3124 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 3125 /* note we're sharing with inode.c for the merge bio hook */
3126 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 3127};
Linux 6.x block layer and io_uring tree(s)