projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: fix error labels in init_btrfs_fs
[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>
5a0e3ad6 29#include <linux/slab.h>
784b4e29 30#include <linux/migrate.h>
7a36ddec 31#include <linux/ratelimit.h>
6463fe58 32#include <linux/uuid.h>
803b2f54 33#include <linux/semaphore.h>
7e75bf3f 34#include <asm/unaligned.h>
eb60ceac
CM		 35#include "ctree.h"
36#include "disk-io.h"
0b947aff 37#include "hash.h"
e089f05c 38#include "transaction.h"
0f7d52f4 39#include "btrfs_inode.h"
0b86a832 40#include "volumes.h"
db94535d 41#include "print-tree.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"
21adbd5c 46#include "check-integrity.h"
606686ee 47#include "rcu-string.h"
8dabb742 48#include "dev-replace.h"
53b381b3 49#include "raid56.h"
5ac1d209 50#include "sysfs.h"
fcebe456 51#include "qgroup.h"
eb60ceac 52
de0022b9
JB		 53#ifdef CONFIG_X86
54#include <asm/cpufeature.h>
55#endif
56
d1310b2e 57static struct extent_io_ops btree_extent_io_ops;
8b712842 58static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 59static void free_fs_root(struct btrfs_root *root);
fcd1f065 60static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 61 int read_only);
143bede5 62static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
acce952b 63static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
64 struct btrfs_root *root);
143bede5 65static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
acce952b 66static int btrfs_destroy_marked_extents(struct btrfs_root *root,
67 struct extent_io_tree *dirty_pages,
68 int mark);
69static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
70 struct extent_io_tree *pinned_extents);
48a3b636
ES		 71static int btrfs_cleanup_transaction(struct btrfs_root *root);
72static void btrfs_error_commit_super(struct btrfs_root *root);
ce9adaa5 73
d352ac68
CM		 74/*
75 * end_io_wq structs are used to do processing in task context when an IO is
76 * complete. This is used during reads to verify checksums, and it is used
77 * by writes to insert metadata for new file extents after IO is complete.
78 */
ce9adaa5
CM		 79struct end_io_wq {
80 struct bio *bio;
81 bio_end_io_t *end_io;
82 void *private;
83 struct btrfs_fs_info *info;
84 int error;
bfebd8b5 85 enum btrfs_wq_endio_type metadata;
ce9adaa5 86 struct list_head list;
8b712842 87 struct btrfs_work work;
ce9adaa5 88};
0da5468f 89
d352ac68
CM		 90/*
91 * async submit bios are used to offload expensive checksumming
92 * onto the worker threads. They checksum file and metadata bios
93 * just before they are sent down the IO stack.
94 */
44b8bd7e
CM		 95struct async_submit_bio {
96 struct inode *inode;
97 struct bio *bio;
98 struct list_head list;
4a69a410
CM		 99 extent_submit_bio_hook_t *submit_bio_start;
100 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM		 101 int rw;
102 int mirror_num;
c8b97818 103 unsigned long bio_flags;
eaf25d93
CM		 104 /*
105 * bio_offset is optional, can be used if the pages in the bio
106 * can't tell us where in the file the bio should go
107 */
108 u64 bio_offset;
8b712842 109 struct btrfs_work work;
79787eaa 110 int error;
44b8bd7e
CM		 111};
112
85d4e461
CM		 113/*
114 * Lockdep class keys for extent_buffer->lock's in this root. For a given
115 * eb, the lockdep key is determined by the btrfs_root it belongs to and
116 * the level the eb occupies in the tree.
117 *
118 * Different roots are used for different purposes and may nest inside each
119 * other and they require separate keysets. As lockdep keys should be
120 * static, assign keysets according to the purpose of the root as indicated
121 * by btrfs_root->objectid. This ensures that all special purpose roots
122 * have separate keysets.
4008c04a 123 *
85d4e461
CM		 124 * Lock-nesting across peer nodes is always done with the immediate parent
125 * node locked thus preventing deadlock. As lockdep doesn't know this, use
126 * subclass to avoid triggering lockdep warning in such cases.
4008c04a 127 *
85d4e461
CM		 128 * The key is set by the readpage_end_io_hook after the buffer has passed
129 * csum validation but before the pages are unlocked. It is also set by
130 * btrfs_init_new_buffer on freshly allocated blocks.
4008c04a 131 *
85d4e461
CM		 132 * We also add a check to make sure the highest level of the tree is the
133 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
134 * needs update as well.
4008c04a
CM		 135 */
136#ifdef CONFIG_DEBUG_LOCK_ALLOC
137# if BTRFS_MAX_LEVEL != 8
138# error
139# endif
85d4e461
CM		 140
141static struct btrfs_lockdep_keyset {
142 u64 id; /* root objectid */
143 const char *name_stem; /* lock name stem */
144 char names[BTRFS_MAX_LEVEL + 1][20];
145 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
146} btrfs_lockdep_keysets[] = {
147 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
148 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
149 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
150 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
151 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
152 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
60b62978 153 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
85d4e461
CM		 154 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
155 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
156 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
13fd8da9 157 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
85d4e461 158 { .id = 0, .name_stem = "tree" },
4008c04a 159};
85d4e461
CM		 160
161void __init btrfs_init_lockdep(void)
162{
163 int i, j;
164
165 /* initialize lockdep class names */
166 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
167 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
168
169 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
170 snprintf(ks->names[j], sizeof(ks->names[j]),
171 "btrfs-%s-%02d", ks->name_stem, j);
172 }
173}
174
175void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
176 int level)
177{
178 struct btrfs_lockdep_keyset *ks;
179
180 BUG_ON(level >= ARRAY_SIZE(ks->keys));
181
182 /* find the matching keyset, id 0 is the default entry */
183 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
184 if (ks->id == objectid)
185 break;
186
187 lockdep_set_class_and_name(&eb->lock,
188 &ks->keys[level], ks->names[level]);
189}
190
4008c04a
CM		 191#endif
192
d352ac68
CM		 193/*
194 * extents on the btree inode are pretty simple, there's one extent
195 * that covers the entire device
196 */
b2950863 197static struct extent_map *btree_get_extent(struct inode *inode,
306e16ce 198 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 199 int create)
7eccb903 200{
5f39d397
CM		 201 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
202 struct extent_map *em;
203 int ret;
204
890871be 205 read_lock(&em_tree->lock);
d1310b2e 206 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 207 if (em) {
208 em->bdev =
209 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 210 read_unlock(&em_tree->lock);
5f39d397 211 goto out;
a061fc8d 212 }
890871be 213 read_unlock(&em_tree->lock);
7b13b7b1 214
172ddd60 215 em = alloc_extent_map();
5f39d397
CM		 216 if (!em) {
217 em = ERR_PTR(-ENOMEM);
218 goto out;
219 }
220 em->start = 0;
0afbaf8c 221 em->len = (u64)-1;
c8b97818 222 em->block_len = (u64)-1;
5f39d397 223 em->block_start = 0;
a061fc8d 224 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 225
890871be 226 write_lock(&em_tree->lock);
09a2a8f9 227 ret = add_extent_mapping(em_tree, em, 0);
5f39d397
CM		 228 if (ret == -EEXIST) {
229 free_extent_map(em);
7b13b7b1 230 em = lookup_extent_mapping(em_tree, start, len);
b4f359ab 231 if (!em)
0433f20d 232 em = ERR_PTR(-EIO);
5f39d397 233 } else if (ret) {
7b13b7b1 234 free_extent_map(em);
0433f20d 235 em = ERR_PTR(ret);
5f39d397 236 }
890871be 237 write_unlock(&em_tree->lock);
7b13b7b1 238
5f39d397
CM		 239out:
240 return em;
7eccb903
CM		 241}
242
b0496686 243u32 btrfs_csum_data(char *data, u32 seed, size_t len)
19c00ddc 244{
0b947aff 245 return btrfs_crc32c(seed, data, len);
19c00ddc
CM		 246}
247
248void btrfs_csum_final(u32 crc, char *result)
249{
7e75bf3f 250 put_unaligned_le32(~crc, result);
19c00ddc
CM		 251}
252
d352ac68
CM		 253/*
254 * compute the csum for a btree block, and either verify it or write it
255 * into the csum field of the block.
256 */
19c00ddc
CM		 257static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
258 int verify)
259{
6c41761f 260 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
607d432d 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 278 cur_len = min(len, map_len - (offset - map_start));
b0496686 279 crc = btrfs_csum_data(kaddr + offset - map_start,
19c00ddc
CM		 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);
efe120a0
FH		 301 printk_ratelimited(KERN_INFO
302 "BTRFS: %s checksum verify failed on %llu wanted %X found %X "
303 "level %d\n",
304 root->fs_info->sb->s_id, buf->start,
305 val, found, btrfs_header_level(buf));
607d432d
JB		 306 if (result != (char *)&inline_result)
307 kfree(result);
19c00ddc
CM		 308 return 1;
309 }
310 } else {
607d432d 311 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 312 }
607d432d
JB		 313 if (result != (char *)&inline_result)
314 kfree(result);
19c00ddc
CM		 315 return 0;
316}
317
d352ac68
CM		 318/*
319 * we can't consider a given block up to date unless the transid of the
320 * block matches the transid in the parent node's pointer. This is how we
321 * detect blocks that either didn't get written at all or got written
322 * in the wrong place.
323 */
1259ab75 324static int verify_parent_transid(struct extent_io_tree *io_tree,
b9fab919
CM		 325 struct extent_buffer *eb, u64 parent_transid,
326 int atomic)
1259ab75 327{
2ac55d41 328 struct extent_state *cached_state = NULL;
1259ab75 329 int ret;
a26e8c9f
JB		 330 bool need_lock = (current->journal_info ==
331 (void *)BTRFS_SEND_TRANS_STUB);
1259ab75
CM		 332
333 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
334 return 0;
335
b9fab919
CM		 336 if (atomic)
337 return -EAGAIN;
338
a26e8c9f
JB		 339 if (need_lock) {
340 btrfs_tree_read_lock(eb);
341 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
342 }
343
2ac55d41 344 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
d0082371 345 0, &cached_state);
0b32f4bb 346 if (extent_buffer_uptodate(eb) &&
1259ab75
CM		 347 btrfs_header_generation(eb) == parent_transid) {
348 ret = 0;
349 goto out;
350 }
29549aec
WS		 351 printk_ratelimited(KERN_INFO "BTRFS (device %s): parent transid verify failed on %llu wanted %llu found %llu\n",
352 eb->fs_info->sb->s_id, eb->start,
353 parent_transid, btrfs_header_generation(eb));
1259ab75 354 ret = 1;
a26e8c9f
JB		 355
356 /*
357 * Things reading via commit roots that don't have normal protection,
358 * like send, can have a really old block in cache that may point at a
359 * block that has been free'd and re-allocated. So don't clear uptodate
360 * if we find an eb that is under IO (dirty/writeback) because we could
361 * end up reading in the stale data and then writing it back out and
362 * making everybody very sad.
363 */
364 if (!extent_buffer_under_io(eb))
365 clear_extent_buffer_uptodate(eb);
33958dc6 366out:
2ac55d41
JB		 367 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
368 &cached_state, GFP_NOFS);
472b909f
JB		 369 if (need_lock)
370 btrfs_tree_read_unlock_blocking(eb);
1259ab75 371 return ret;
1259ab75
CM		 372}
373
1104a885
DS		 374/*
375 * Return 0 if the superblock checksum type matches the checksum value of that
376 * algorithm. Pass the raw disk superblock data.
377 */
378static int btrfs_check_super_csum(char *raw_disk_sb)
379{
380 struct btrfs_super_block *disk_sb =
381 (struct btrfs_super_block *)raw_disk_sb;
382 u16 csum_type = btrfs_super_csum_type(disk_sb);
383 int ret = 0;
384
385 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
386 u32 crc = ~(u32)0;
387 const int csum_size = sizeof(crc);
388 char result[csum_size];
389
390 /*
391 * The super_block structure does not span the whole
392 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
393 * is filled with zeros and is included in the checkum.
394 */
395 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
396 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
397 btrfs_csum_final(crc, result);
398
399 if (memcmp(raw_disk_sb, result, csum_size))
400 ret = 1;
667e7d94
CM		 401
402 if (ret && btrfs_super_generation(disk_sb) < 10) {
efe120a0
FH		 403 printk(KERN_WARNING
404 "BTRFS: super block crcs don't match, older mkfs detected\n");
667e7d94
CM		 405 ret = 0;
406 }
1104a885
DS		 407 }
408
409 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
efe120a0 410 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
1104a885
DS		 411 csum_type);
412 ret = 1;
413 }
414
415 return ret;
416}
417
d352ac68
CM		 418/*
419 * helper to read a given tree block, doing retries as required when
420 * the checksums don't match and we have alternate mirrors to try.
421 */
f188591e
CM		 422static int btree_read_extent_buffer_pages(struct btrfs_root *root,
423 struct extent_buffer *eb,
ca7a79ad 424 u64 start, u64 parent_transid)
f188591e
CM		 425{
426 struct extent_io_tree *io_tree;
ea466794 427 int failed = 0;
f188591e
CM		 428 int ret;
429 int num_copies = 0;
430 int mirror_num = 0;
ea466794 431 int failed_mirror = 0;
f188591e 432
a826d6dc 433 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
f188591e
CM		 434 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
435 while (1) {
bb82ab88
AJ		 436 ret = read_extent_buffer_pages(io_tree, eb, start,
437 WAIT_COMPLETE,
f188591e 438 btree_get_extent, mirror_num);
256dd1bb
SB		 439 if (!ret) {
440 if (!verify_parent_transid(io_tree, eb,
b9fab919 441 parent_transid, 0))
256dd1bb
SB		 442 break;
443 else
444 ret = -EIO;
445 }
d397712b 446
a826d6dc
JB		 447 /*
448 * This buffer's crc is fine, but its contents are corrupted, so
449 * there is no reason to read the other copies, they won't be
450 * any less wrong.
451 */
452 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
ea466794
JB		 453 break;
454
5d964051 455 num_copies = btrfs_num_copies(root->fs_info,
f188591e 456 eb->start, eb->len);
4235298e 457 if (num_copies == 1)
ea466794 458 break;
4235298e 459
5cf1ab56
JB		 460 if (!failed_mirror) {
461 failed = 1;
462 failed_mirror = eb->read_mirror;
463 }
464
f188591e 465 mirror_num++;
ea466794
JB		 466 if (mirror_num == failed_mirror)
467 mirror_num++;
468
4235298e 469 if (mirror_num > num_copies)
ea466794 470 break;
f188591e 471 }
ea466794 472
c0901581 473 if (failed && !ret && failed_mirror)
ea466794
JB		 474 repair_eb_io_failure(root, eb, failed_mirror);
475
476 return ret;
f188591e 477}
19c00ddc 478
d352ac68 479/*
d397712b
CM		 480 * checksum a dirty tree block before IO. This has extra checks to make sure
481 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 482 */
d397712b 483
b2950863 484static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
19c00ddc 485{
4eee4fa4 486 u64 start = page_offset(page);
19c00ddc 487 u64 found_start;
19c00ddc 488 struct extent_buffer *eb;
f188591e 489
4f2de97a
JB		 490 eb = (struct extent_buffer *)page->private;
491 if (page != eb->pages[0])
492 return 0;
19c00ddc 493 found_start = btrfs_header_bytenr(eb);
fae7f21c 494 if (WARN_ON(found_start != start || !PageUptodate(page)))
4f2de97a 495 return 0;
19c00ddc 496 csum_tree_block(root, eb, 0);
19c00ddc
CM		 497 return 0;
498}
499
2b82032c
YZ		 500static int check_tree_block_fsid(struct btrfs_root *root,
501 struct extent_buffer *eb)
502{
503 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
504 u8 fsid[BTRFS_UUID_SIZE];
505 int ret = 1;
506
0a4e5586 507 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
2b82032c
YZ		 508 while (fs_devices) {
509 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
510 ret = 0;
511 break;
512 }
513 fs_devices = fs_devices->seed;
514 }
515 return ret;
516}
517
a826d6dc 518#define CORRUPT(reason, eb, root, slot) \
efe120a0
FH		 519 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
520 "root=%llu, slot=%d", reason, \
c1c9ff7c 521 btrfs_header_bytenr(eb), root->objectid, slot)
a826d6dc
JB		 522
523static noinline int check_leaf(struct btrfs_root *root,
524 struct extent_buffer *leaf)
525{
526 struct btrfs_key key;
527 struct btrfs_key leaf_key;
528 u32 nritems = btrfs_header_nritems(leaf);
529 int slot;
530
531 if (nritems == 0)
532 return 0;
533
534 /* Check the 0 item */
535 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
536 BTRFS_LEAF_DATA_SIZE(root)) {
537 CORRUPT("invalid item offset size pair", leaf, root, 0);
538 return -EIO;
539 }
540
541 /*
542 * Check to make sure each items keys are in the correct order and their
543 * offsets make sense. We only have to loop through nritems-1 because
544 * we check the current slot against the next slot, which verifies the
545 * next slot's offset+size makes sense and that the current's slot
546 * offset is correct.
547 */
548 for (slot = 0; slot < nritems - 1; slot++) {
549 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
550 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
551
552 /* Make sure the keys are in the right order */
553 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
554 CORRUPT("bad key order", leaf, root, slot);
555 return -EIO;
556 }
557
558 /*
559 * Make sure the offset and ends are right, remember that the
560 * item data starts at the end of the leaf and grows towards the
561 * front.
562 */
563 if (btrfs_item_offset_nr(leaf, slot) !=
564 btrfs_item_end_nr(leaf, slot + 1)) {
565 CORRUPT("slot offset bad", leaf, root, slot);
566 return -EIO;
567 }
568
569 /*
570 * Check to make sure that we don't point outside of the leaf,
571 * just incase all the items are consistent to eachother, but
572 * all point outside of the leaf.
573 */
574 if (btrfs_item_end_nr(leaf, slot) >
575 BTRFS_LEAF_DATA_SIZE(root)) {
576 CORRUPT("slot end outside of leaf", leaf, root, slot);
577 return -EIO;
578 }
579 }
580
581 return 0;
582}
583
facc8a22
MX		 584static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
585 u64 phy_offset, struct page *page,
586 u64 start, u64 end, int mirror)
ce9adaa5 587{
ce9adaa5
CM		 588 u64 found_start;
589 int found_level;
ce9adaa5
CM		 590 struct extent_buffer *eb;
591 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 592 int ret = 0;
727011e0 593 int reads_done;
ce9adaa5 594
ce9adaa5
CM		 595 if (!page->private)
596 goto out;
d397712b 597
4f2de97a 598 eb = (struct extent_buffer *)page->private;
d397712b 599
0b32f4bb
JB		 600 /* the pending IO might have been the only thing that kept this buffer
601 * in memory. Make sure we have a ref for all this other checks
602 */
603 extent_buffer_get(eb);
604
605 reads_done = atomic_dec_and_test(&eb->io_pages);
727011e0
CM		 606 if (!reads_done)
607 goto err;
f188591e 608
5cf1ab56 609 eb->read_mirror = mirror;
ea466794
JB		 610 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
611 ret = -EIO;
612 goto err;
613 }
614
ce9adaa5 615 found_start = btrfs_header_bytenr(eb);
727011e0 616 if (found_start != eb->start) {
29549aec 617 printk_ratelimited(KERN_INFO "BTRFS (device %s): bad tree block start "
193f284d 618 "%llu %llu\n",
29549aec 619 eb->fs_info->sb->s_id, found_start, eb->start);
f188591e 620 ret = -EIO;
ce9adaa5
CM		 621 goto err;
622 }
2b82032c 623 if (check_tree_block_fsid(root, eb)) {
29549aec
WS		 624 printk_ratelimited(KERN_INFO "BTRFS (device %s): bad fsid on block %llu\n",
625 eb->fs_info->sb->s_id, eb->start);
1259ab75
CM		 626 ret = -EIO;
627 goto err;
628 }
ce9adaa5 629 found_level = btrfs_header_level(eb);
1c24c3ce 630 if (found_level >= BTRFS_MAX_LEVEL) {
efe120a0 631 btrfs_info(root->fs_info, "bad tree block level %d",
1c24c3ce
JB		 632 (int)btrfs_header_level(eb));
633 ret = -EIO;
634 goto err;
635 }
ce9adaa5 636
85d4e461
CM		 637 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
638 eb, found_level);
4008c04a 639
ce9adaa5 640 ret = csum_tree_block(root, eb, 1);
a826d6dc 641 if (ret) {
f188591e 642 ret = -EIO;
a826d6dc
JB		 643 goto err;
644 }
645
646 /*
647 * If this is a leaf block and it is corrupt, set the corrupt bit so
648 * that we don't try and read the other copies of this block, just
649 * return -EIO.
650 */
651 if (found_level == 0 && check_leaf(root, eb)) {
652 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
653 ret = -EIO;
654 }
ce9adaa5 655
0b32f4bb
JB		 656 if (!ret)
657 set_extent_buffer_uptodate(eb);
ce9adaa5 658err:
79fb65a1
JB		 659 if (reads_done &&
660 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 661 btree_readahead_hook(root, eb, eb->start, ret);
4bb31e92 662
53b381b3
DW		 663 if (ret) {
664 /*
665 * our io error hook is going to dec the io pages
666 * again, we have to make sure it has something
667 * to decrement
668 */
669 atomic_inc(&eb->io_pages);
0b32f4bb 670 clear_extent_buffer_uptodate(eb);
53b381b3 671 }
0b32f4bb 672 free_extent_buffer(eb);
ce9adaa5 673out:
f188591e 674 return ret;
ce9adaa5
CM		 675}
676
ea466794 677static int btree_io_failed_hook(struct page *page, int failed_mirror)
4bb31e92 678{
4bb31e92
AJ		 679 struct extent_buffer *eb;
680 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
681
4f2de97a 682 eb = (struct extent_buffer *)page->private;
ea466794 683 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
5cf1ab56 684 eb->read_mirror = failed_mirror;
53b381b3 685 atomic_dec(&eb->io_pages);
ea466794 686 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 687 btree_readahead_hook(root, eb, eb->start, -EIO);
4bb31e92
AJ		 688 return -EIO; /* we fixed nothing */
689}
690
ce9adaa5 691static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM		 692{
693 struct end_io_wq *end_io_wq = bio->bi_private;
694 struct btrfs_fs_info *fs_info;
9e0af237
LB		 695 struct btrfs_workqueue *wq;
696 btrfs_work_func_t func;
ce9adaa5 697
ce9adaa5 698 fs_info = end_io_wq->info;
ce9adaa5 699 end_io_wq->error = err;
d20f7043 700
7b6d91da 701 if (bio->bi_rw & REQ_WRITE) {
9e0af237
LB		 702 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) {
703 wq = fs_info->endio_meta_write_workers;
704 func = btrfs_endio_meta_write_helper;
705 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) {
706 wq = fs_info->endio_freespace_worker;
707 func = btrfs_freespace_write_helper;
708 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
709 wq = fs_info->endio_raid56_workers;
710 func = btrfs_endio_raid56_helper;
711 } else {
712 wq = fs_info->endio_write_workers;
713 func = btrfs_endio_write_helper;
714 }
d20f7043 715 } else {
8b110e39
MX		 716 if (unlikely(end_io_wq->metadata ==
717 BTRFS_WQ_ENDIO_DIO_REPAIR)) {
718 wq = fs_info->endio_repair_workers;
719 func = btrfs_endio_repair_helper;
720 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
9e0af237
LB		 721 wq = fs_info->endio_raid56_workers;
722 func = btrfs_endio_raid56_helper;
723 } else if (end_io_wq->metadata) {
724 wq = fs_info->endio_meta_workers;
725 func = btrfs_endio_meta_helper;
726 } else {
727 wq = fs_info->endio_workers;
728 func = btrfs_endio_helper;
729 }
d20f7043 730 }
9e0af237
LB		 731
732 btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL);
733 btrfs_queue_work(wq, &end_io_wq->work);
ce9adaa5
CM		 734}
735
22c59948 736int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
bfebd8b5 737 enum btrfs_wq_endio_type metadata)
0b86a832 738{
ce9adaa5 739 struct end_io_wq *end_io_wq;
8b110e39 740
ce9adaa5
CM		 741 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
742 if (!end_io_wq)
743 return -ENOMEM;
744
745 end_io_wq->private = bio->bi_private;
746 end_io_wq->end_io = bio->bi_end_io;
22c59948 747 end_io_wq->info = info;
ce9adaa5
CM		 748 end_io_wq->error = 0;
749 end_io_wq->bio = bio;
22c59948 750 end_io_wq->metadata = metadata;
ce9adaa5
CM		 751
752 bio->bi_private = end_io_wq;
753 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 754 return 0;
755}
756
b64a2851 757unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 758{
4854ddd0 759 unsigned long limit = min_t(unsigned long,
5cdc7ad3 760 info->thread_pool_size,
4854ddd0
CM		 761 info->fs_devices->open_devices);
762 return 256 * limit;
763}
0986fe9e 764
4a69a410
CM		 765static void run_one_async_start(struct btrfs_work *work)
766{
4a69a410 767 struct async_submit_bio *async;
79787eaa 768 int ret;
4a69a410
CM		 769
770 async = container_of(work, struct async_submit_bio, work);
79787eaa
JM		 771 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
772 async->mirror_num, async->bio_flags,
773 async->bio_offset);
774 if (ret)
775 async->error = ret;
4a69a410
CM		 776}
777
778static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 779{
780 struct btrfs_fs_info *fs_info;
781 struct async_submit_bio *async;
4854ddd0 782 int limit;
8b712842
CM		 783
784 async = container_of(work, struct async_submit_bio, work);
785 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 786
b64a2851 787 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 788 limit = limit * 2 / 3;
789
66657b31 790 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
b64a2851 791 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 792 wake_up(&fs_info->async_submit_wait);
793
79787eaa
JM		 794 /* If an error occured we just want to clean up the bio and move on */
795 if (async->error) {
796 bio_endio(async->bio, async->error);
797 return;
798 }
799
4a69a410 800 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 801 async->mirror_num, async->bio_flags,
802 async->bio_offset);
4a69a410
CM		 803}
804
805static void run_one_async_free(struct btrfs_work *work)
806{
807 struct async_submit_bio *async;
808
809 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 810 kfree(async);
811}
812
44b8bd7e
CM		 813int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
814 int rw, struct bio *bio, int mirror_num,
c8b97818 815 unsigned long bio_flags,
eaf25d93 816 u64 bio_offset,
4a69a410
CM		 817 extent_submit_bio_hook_t *submit_bio_start,
818 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 819{
820 struct async_submit_bio *async;
821
822 async = kmalloc(sizeof(*async), GFP_NOFS);
823 if (!async)
824 return -ENOMEM;
825
826 async->inode = inode;
827 async->rw = rw;
828 async->bio = bio;
829 async->mirror_num = mirror_num;
4a69a410
CM		 830 async->submit_bio_start = submit_bio_start;
831 async->submit_bio_done = submit_bio_done;
832
9e0af237 833 btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start,
5cdc7ad3 834 run_one_async_done, run_one_async_free);
4a69a410 835
c8b97818 836 async->bio_flags = bio_flags;
eaf25d93 837 async->bio_offset = bio_offset;
8c8bee1d 838
79787eaa
JM		 839 async->error = 0;
840
cb03c743 841 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 842
7b6d91da 843 if (rw & REQ_SYNC)
5cdc7ad3 844 btrfs_set_work_high_priority(&async->work);
d313d7a3 845
5cdc7ad3 846 btrfs_queue_work(fs_info->workers, &async->work);
9473f16c 847
d397712b 848 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 849 atomic_read(&fs_info->nr_async_submits)) {
850 wait_event(fs_info->async_submit_wait,
851 (atomic_read(&fs_info->nr_async_submits) == 0));
852 }
853
44b8bd7e
CM		 854 return 0;
855}
856
ce3ed71a
CM		 857static int btree_csum_one_bio(struct bio *bio)
858{
2c30c71b 859 struct bio_vec *bvec;
ce3ed71a 860 struct btrfs_root *root;
2c30c71b 861 int i, ret = 0;
ce3ed71a 862
2c30c71b 863 bio_for_each_segment_all(bvec, bio, i) {
ce3ed71a 864 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
79787eaa
JM		 865 ret = csum_dirty_buffer(root, bvec->bv_page);
866 if (ret)
867 break;
ce3ed71a 868 }
2c30c71b 869
79787eaa 870 return ret;
ce3ed71a
CM		 871}
872
4a69a410
CM		 873static int __btree_submit_bio_start(struct inode *inode, int rw,
874 struct bio *bio, int mirror_num,
eaf25d93
CM		 875 unsigned long bio_flags,
876 u64 bio_offset)
22c59948 877{
8b712842
CM		 878 /*
879 * when we're called for a write, we're already in the async
5443be45 880 * submission context. Just jump into btrfs_map_bio
8b712842 881 */
79787eaa 882 return btree_csum_one_bio(bio);
4a69a410 883}
22c59948 884
4a69a410 885static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 886 int mirror_num, unsigned long bio_flags,
887 u64 bio_offset)
4a69a410 888{
61891923
SB		 889 int ret;
890
8b712842 891 /*
4a69a410
CM		 892 * when we're called for a write, we're already in the async
893 * submission context. Just jump into btrfs_map_bio
8b712842 894 */
61891923
SB		 895 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
896 if (ret)
897 bio_endio(bio, ret);
898 return ret;
0b86a832
CM		 899}
900
de0022b9
JB		 901static int check_async_write(struct inode *inode, unsigned long bio_flags)
902{
903 if (bio_flags & EXTENT_BIO_TREE_LOG)
904 return 0;
905#ifdef CONFIG_X86
906 if (cpu_has_xmm4_2)
907 return 0;
908#endif
909 return 1;
910}
911
44b8bd7e 912static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 913 int mirror_num, unsigned long bio_flags,
914 u64 bio_offset)
44b8bd7e 915{
de0022b9 916 int async = check_async_write(inode, bio_flags);
cad321ad
CM		 917 int ret;
918
7b6d91da 919 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 920 /*
921 * called for a read, do the setup so that checksum validation
922 * can happen in the async kernel threads
923 */
f3f266ab 924 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
bfebd8b5 925 bio, BTRFS_WQ_ENDIO_METADATA);
1d4284bd 926 if (ret)
61891923
SB		 927 goto out_w_error;
928 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
929 mirror_num, 0);
de0022b9
JB		 930 } else if (!async) {
931 ret = btree_csum_one_bio(bio);
932 if (ret)
61891923
SB		 933 goto out_w_error;
934 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
935 mirror_num, 0);
936 } else {
937 /*
938 * kthread helpers are used to submit writes so that
939 * checksumming can happen in parallel across all CPUs
940 */
941 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
942 inode, rw, bio, mirror_num, 0,
943 bio_offset,
944 __btree_submit_bio_start,
945 __btree_submit_bio_done);
44b8bd7e 946 }
d313d7a3 947
61891923
SB		 948 if (ret) {
949out_w_error:
950 bio_endio(bio, ret);
951 }
952 return ret;
44b8bd7e
CM		 953}
954
3dd1462e 955#ifdef CONFIG_MIGRATION
784b4e29 956static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG		 957 struct page *newpage, struct page *page,
958 enum migrate_mode mode)
784b4e29
CM		 959{
960 /*
961 * we can't safely write a btree page from here,
962 * we haven't done the locking hook
963 */
964 if (PageDirty(page))
965 return -EAGAIN;
966 /*
967 * Buffers may be managed in a filesystem specific way.
968 * We must have no buffers or drop them.
969 */
970 if (page_has_private(page) &&
971 !try_to_release_page(page, GFP_KERNEL))
972 return -EAGAIN;
a6bc32b8 973 return migrate_page(mapping, newpage, page, mode);
784b4e29 974}
3dd1462e 975#endif
784b4e29 976
0da5468f
CM		 977
978static int btree_writepages(struct address_space *mapping,
979 struct writeback_control *wbc)
980{
e2d84521
MX		 981 struct btrfs_fs_info *fs_info;
982 int ret;
983
d8d5f3e1 984 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM		 985
986 if (wbc->for_kupdate)
987 return 0;
988
e2d84521 989 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 990 /* this is a bit racy, but that's ok */
e2d84521
MX		 991 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
992 BTRFS_DIRTY_METADATA_THRESH);
993 if (ret < 0)
793955bc 994 return 0;
793955bc 995 }
0b32f4bb 996 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM		 997}
998
b2950863 999static int btree_readpage(struct file *file, struct page *page)
5f39d397 1000{
d1310b2e
CM		 1001 struct extent_io_tree *tree;
1002 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 1003 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 1004}
22b0ebda 1005
70dec807 1006static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 1007{
98509cfc 1008 if (PageWriteback(page) || PageDirty(page))
d397712b 1009 return 0;
0c4e538b 1010
f7a52a40 1011 return try_release_extent_buffer(page);
d98237b3
CM		 1012}
1013
d47992f8
LC		 1014static void btree_invalidatepage(struct page *page, unsigned int offset,
1015 unsigned int length)
d98237b3 1016{
d1310b2e
CM		 1017 struct extent_io_tree *tree;
1018 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 1019 extent_invalidatepage(tree, page, offset);
1020 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 1021 if (PagePrivate(page)) {
efe120a0
FH		 1022 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1023 "page private not zero on page %llu",
1024 (unsigned long long)page_offset(page));
9ad6b7bc
CM		 1025 ClearPagePrivate(page);
1026 set_page_private(page, 0);
1027 page_cache_release(page);
1028 }
d98237b3
CM		 1029}
1030
0b32f4bb
JB		 1031static int btree_set_page_dirty(struct page *page)
1032{
bb146eb2 1033#ifdef DEBUG
0b32f4bb
JB		 1034 struct extent_buffer *eb;
1035
1036 BUG_ON(!PagePrivate(page));
1037 eb = (struct extent_buffer *)page->private;
1038 BUG_ON(!eb);
1039 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1040 BUG_ON(!atomic_read(&eb->refs));
1041 btrfs_assert_tree_locked(eb);
bb146eb2 1042#endif
0b32f4bb
JB		 1043 return __set_page_dirty_nobuffers(page);
1044}
1045
7f09410b 1046static const struct address_space_operations btree_aops = {
d98237b3 1047 .readpage = btree_readpage,
0da5468f 1048 .writepages = btree_writepages,
5f39d397
CM		 1049 .releasepage = btree_releasepage,
1050 .invalidatepage = btree_invalidatepage,
5a92bc88 1051#ifdef CONFIG_MIGRATION
784b4e29 1052 .migratepage = btree_migratepage,
5a92bc88 1053#endif
0b32f4bb 1054 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM		 1055};
1056
ca7a79ad
CM		 1057int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1058 u64 parent_transid)
090d1875 1059{
5f39d397
CM		 1060 struct extent_buffer *buf = NULL;
1061 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 1062 int ret = 0;
090d1875 1063
db94535d 1064 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 1065 if (!buf)
090d1875 1066 return 0;
d1310b2e 1067 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
bb82ab88 1068 buf, 0, WAIT_NONE, btree_get_extent, 0);
5f39d397 1069 free_extent_buffer(buf);
de428b63 1070 return ret;
090d1875
CM		 1071}
1072
ab0fff03
AJ		 1073int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1074 int mirror_num, struct extent_buffer **eb)
1075{
1076 struct extent_buffer *buf = NULL;
1077 struct inode *btree_inode = root->fs_info->btree_inode;
1078 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1079 int ret;
1080
1081 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1082 if (!buf)
1083 return 0;
1084
1085 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1086
1087 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1088 btree_get_extent, mirror_num);
1089 if (ret) {
1090 free_extent_buffer(buf);
1091 return ret;
1092 }
1093
1094 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1095 free_extent_buffer(buf);
1096 return -EIO;
0b32f4bb 1097 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ		 1098 *eb = buf;
1099 } else {
1100 free_extent_buffer(buf);
1101 }
1102 return 0;
1103}
1104
0999df54
CM		 1105struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1106 u64 bytenr, u32 blocksize)
1107{
f28491e0 1108 return find_extent_buffer(root->fs_info, bytenr);
0999df54
CM		 1109}
1110
1111struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1112 u64 bytenr, u32 blocksize)
1113{
faa2dbf0
JB		 1114#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1115 if (unlikely(test_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state)))
1116 return alloc_test_extent_buffer(root->fs_info, bytenr,
1117 blocksize);
1118#endif
f28491e0 1119 return alloc_extent_buffer(root->fs_info, bytenr, blocksize);
0999df54
CM		 1120}
1121
1122
e02119d5
CM		 1123int btrfs_write_tree_block(struct extent_buffer *buf)
1124{
727011e0 1125 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1126 buf->start + buf->len - 1);
e02119d5
CM		 1127}
1128
1129int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1130{
727011e0 1131 return filemap_fdatawait_range(buf->pages[0]->mapping,
8aa38c31 1132 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1133}
1134
0999df54 1135struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 1136 u32 blocksize, u64 parent_transid)
0999df54
CM		 1137{
1138 struct extent_buffer *buf = NULL;
0999df54
CM		 1139 int ret;
1140
0999df54
CM		 1141 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1142 if (!buf)
1143 return NULL;
0999df54 1144
ca7a79ad 1145 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
0f0fe8f7
FDBM		 1146 if (ret) {
1147 free_extent_buffer(buf);
1148 return NULL;
1149 }
5f39d397 1150 return buf;
ce9adaa5 1151
eb60ceac
CM		 1152}
1153
d5c13f92
JM		 1154void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1155 struct extent_buffer *buf)
ed2ff2cb 1156{
e2d84521
MX		 1157 struct btrfs_fs_info *fs_info = root->fs_info;
1158
55c69072 1159 if (btrfs_header_generation(buf) ==
e2d84521 1160 fs_info->running_transaction->transid) {
b9447ef8 1161 btrfs_assert_tree_locked(buf);
b4ce94de 1162
b9473439 1163 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
e2d84521
MX		 1164 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1165 -buf->len,
1166 fs_info->dirty_metadata_batch);
ed7b63eb
JB		 1167 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1168 btrfs_set_lock_blocking(buf);
1169 clear_extent_buffer_dirty(buf);
1170 }
925baedd 1171 }
5f39d397
CM		 1172}
1173
8257b2dc
MX		 1174static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1175{
1176 struct btrfs_subvolume_writers *writers;
1177 int ret;
1178
1179 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1180 if (!writers)
1181 return ERR_PTR(-ENOMEM);
1182
1183 ret = percpu_counter_init(&writers->counter, 0);
1184 if (ret < 0) {
1185 kfree(writers);
1186 return ERR_PTR(ret);
1187 }
1188
1189 init_waitqueue_head(&writers->wait);
1190 return writers;
1191}
1192
1193static void
1194btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1195{
1196 percpu_counter_destroy(&writers->counter);
1197 kfree(writers);
1198}
1199
707e8a07
DS		 1200static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize,
1201 struct btrfs_root *root, struct btrfs_fs_info *fs_info,
143bede5 1202 u64 objectid)
d97e63b6 1203{
cfaa7295 1204 root->node = NULL;
a28ec197 1205 root->commit_root = NULL;
db94535d
CM		 1206 root->sectorsize = sectorsize;
1207 root->nodesize = nodesize;
87ee04eb 1208 root->stripesize = stripesize;
27cdeb70 1209 root->state = 0;
d68fc57b 1210 root->orphan_cleanup_state = 0;
0b86a832 1211
0f7d52f4
CM		 1212 root->objectid = objectid;
1213 root->last_trans = 0;
13a8a7c8 1214 root->highest_objectid = 0;
eb73c1b7 1215 root->nr_delalloc_inodes = 0;
199c2a9c 1216 root->nr_ordered_extents = 0;
58176a96 1217 root->name = NULL;
6bef4d31 1218 root->inode_tree = RB_ROOT;
16cdcec7 1219 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1220 root->block_rsv = NULL;
d68fc57b 1221 root->orphan_block_rsv = NULL;
0b86a832
CM		 1222
1223 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1224 INIT_LIST_HEAD(&root->root_list);
eb73c1b7
MX		 1225 INIT_LIST_HEAD(&root->delalloc_inodes);
1226 INIT_LIST_HEAD(&root->delalloc_root);
199c2a9c
MX		 1227 INIT_LIST_HEAD(&root->ordered_extents);
1228 INIT_LIST_HEAD(&root->ordered_root);
2ab28f32
JB		 1229 INIT_LIST_HEAD(&root->logged_list[0]);
1230 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1231 spin_lock_init(&root->orphan_lock);
5d4f98a2 1232 spin_lock_init(&root->inode_lock);
eb73c1b7 1233 spin_lock_init(&root->delalloc_lock);
199c2a9c 1234 spin_lock_init(&root->ordered_extent_lock);
f0486c68 1235 spin_lock_init(&root->accounting_lock);
2ab28f32
JB		 1236 spin_lock_init(&root->log_extents_lock[0]);
1237 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1238 mutex_init(&root->objectid_mutex);
e02119d5 1239 mutex_init(&root->log_mutex);
31f3d255 1240 mutex_init(&root->ordered_extent_mutex);
573bfb72 1241 mutex_init(&root->delalloc_mutex);
7237f183
YZ		 1242 init_waitqueue_head(&root->log_writer_wait);
1243 init_waitqueue_head(&root->log_commit_wait[0]);
1244 init_waitqueue_head(&root->log_commit_wait[1]);
8b050d35
MX		 1245 INIT_LIST_HEAD(&root->log_ctxs[0]);
1246 INIT_LIST_HEAD(&root->log_ctxs[1]);
7237f183
YZ		 1247 atomic_set(&root->log_commit[0], 0);
1248 atomic_set(&root->log_commit[1], 0);
1249 atomic_set(&root->log_writers, 0);
2ecb7923 1250 atomic_set(&root->log_batch, 0);
8a35d95f 1251 atomic_set(&root->orphan_inodes, 0);
b0feb9d9 1252 atomic_set(&root->refs, 1);
8257b2dc 1253 atomic_set(&root->will_be_snapshoted, 0);
7237f183 1254 root->log_transid = 0;
d1433deb 1255 root->log_transid_committed = -1;
257c62e1 1256 root->last_log_commit = 0;
06ea65a3
JB		 1257 if (fs_info)
1258 extent_io_tree_init(&root->dirty_log_pages,
1259 fs_info->btree_inode->i_mapping);
017e5369 1260
3768f368
CM		 1261 memset(&root->root_key, 0, sizeof(root->root_key));
1262 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1263 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1264 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
06ea65a3
JB		 1265 if (fs_info)
1266 root->defrag_trans_start = fs_info->generation;
1267 else
1268 root->defrag_trans_start = 0;
58176a96 1269 init_completion(&root->kobj_unregister);
4d775673 1270 root->root_key.objectid = objectid;
0ee5dc67 1271 root->anon_dev = 0;
8ea05e3a 1272
5f3ab90a 1273 spin_lock_init(&root->root_item_lock);
3768f368
CM		 1274}
1275
f84a8bd6 1276static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
6f07e42e
AV		 1277{
1278 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1279 if (root)
1280 root->fs_info = fs_info;
1281 return root;
1282}
1283
06ea65a3
JB		 1284#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1285/* Should only be used by the testing infrastructure */
1286struct btrfs_root *btrfs_alloc_dummy_root(void)
1287{
1288 struct btrfs_root *root;
1289
1290 root = btrfs_alloc_root(NULL);
1291 if (!root)
1292 return ERR_PTR(-ENOMEM);
707e8a07 1293 __setup_root(4096, 4096, 4096, root, NULL, 1);
27cdeb70 1294 set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state);
faa2dbf0 1295 root->alloc_bytenr = 0;
06ea65a3
JB		 1296
1297 return root;
1298}
1299#endif
1300
20897f5c
AJ		 1301struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1302 struct btrfs_fs_info *fs_info,
1303 u64 objectid)
1304{
1305 struct extent_buffer *leaf;
1306 struct btrfs_root *tree_root = fs_info->tree_root;
1307 struct btrfs_root *root;
1308 struct btrfs_key key;
1309 int ret = 0;
6463fe58 1310 uuid_le uuid;
20897f5c
AJ		 1311
1312 root = btrfs_alloc_root(fs_info);
1313 if (!root)
1314 return ERR_PTR(-ENOMEM);
1315
707e8a07
DS		 1316 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1317 tree_root->stripesize, root, fs_info, objectid);
20897f5c
AJ		 1318 root->root_key.objectid = objectid;
1319 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1320 root->root_key.offset = 0;
1321
707e8a07 1322 leaf = btrfs_alloc_free_block(trans, root, root->nodesize,
20897f5c
AJ		 1323 0, objectid, NULL, 0, 0, 0);
1324 if (IS_ERR(leaf)) {
1325 ret = PTR_ERR(leaf);
1dd05682 1326 leaf = NULL;
20897f5c
AJ		 1327 goto fail;
1328 }
1329
20897f5c
AJ		 1330 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1331 btrfs_set_header_bytenr(leaf, leaf->start);
1332 btrfs_set_header_generation(leaf, trans->transid);
1333 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1334 btrfs_set_header_owner(leaf, objectid);
1335 root->node = leaf;
1336
0a4e5586 1337 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
20897f5c
AJ		 1338 BTRFS_FSID_SIZE);
1339 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
b308bc2f 1340 btrfs_header_chunk_tree_uuid(leaf),
20897f5c
AJ		 1341 BTRFS_UUID_SIZE);
1342 btrfs_mark_buffer_dirty(leaf);
1343
1344 root->commit_root = btrfs_root_node(root);
27cdeb70 1345 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
20897f5c
AJ		 1346
1347 root->root_item.flags = 0;
1348 root->root_item.byte_limit = 0;
1349 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1350 btrfs_set_root_generation(&root->root_item, trans->transid);
1351 btrfs_set_root_level(&root->root_item, 0);
1352 btrfs_set_root_refs(&root->root_item, 1);
1353 btrfs_set_root_used(&root->root_item, leaf->len);
1354 btrfs_set_root_last_snapshot(&root->root_item, 0);
1355 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB		 1356 uuid_le_gen(&uuid);
1357 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ		 1358 root->root_item.drop_level = 0;
1359
1360 key.objectid = objectid;
1361 key.type = BTRFS_ROOT_ITEM_KEY;
1362 key.offset = 0;
1363 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1364 if (ret)
1365 goto fail;
1366
1367 btrfs_tree_unlock(leaf);
1368
1dd05682
TI		 1369 return root;
1370
20897f5c 1371fail:
1dd05682
TI		 1372 if (leaf) {
1373 btrfs_tree_unlock(leaf);
59885b39 1374 free_extent_buffer(root->commit_root);
1dd05682
TI		 1375 free_extent_buffer(leaf);
1376 }
1377 kfree(root);
20897f5c 1378
1dd05682 1379 return ERR_PTR(ret);
20897f5c
AJ		 1380}
1381
7237f183
YZ		 1382static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1383 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1384{
1385 struct btrfs_root *root;
1386 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1387 struct extent_buffer *leaf;
e02119d5 1388
6f07e42e 1389 root = btrfs_alloc_root(fs_info);
e02119d5 1390 if (!root)
7237f183 1391 return ERR_PTR(-ENOMEM);
e02119d5 1392
707e8a07
DS		 1393 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1394 tree_root->stripesize, root, fs_info,
1395 BTRFS_TREE_LOG_OBJECTID);
e02119d5
CM		 1396
1397 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1398 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1399 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
27cdeb70 1400
7237f183 1401 /*
27cdeb70
MX		 1402 * DON'T set REF_COWS for log trees
1403 *
7237f183
YZ		 1404 * log trees do not get reference counted because they go away
1405 * before a real commit is actually done. They do store pointers
1406 * to file data extents, and those reference counts still get
1407 * updated (along with back refs to the log tree).
1408 */
e02119d5 1409
707e8a07 1410 leaf = btrfs_alloc_free_block(trans, root, root->nodesize, 0,
66d7e7f0 1411 BTRFS_TREE_LOG_OBJECTID, NULL,
5581a51a 1412 0, 0, 0);
7237f183
YZ		 1413 if (IS_ERR(leaf)) {
1414 kfree(root);
1415 return ERR_CAST(leaf);
1416 }
e02119d5 1417
5d4f98a2
YZ		 1418 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1419 btrfs_set_header_bytenr(leaf, leaf->start);
1420 btrfs_set_header_generation(leaf, trans->transid);
1421 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1422 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1423 root->node = leaf;
e02119d5
CM		 1424
1425 write_extent_buffer(root->node, root->fs_info->fsid,
0a4e5586 1426 btrfs_header_fsid(), BTRFS_FSID_SIZE);
e02119d5
CM		 1427 btrfs_mark_buffer_dirty(root->node);
1428 btrfs_tree_unlock(root->node);
7237f183
YZ		 1429 return root;
1430}
1431
1432int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1433 struct btrfs_fs_info *fs_info)
1434{
1435 struct btrfs_root *log_root;
1436
1437 log_root = alloc_log_tree(trans, fs_info);
1438 if (IS_ERR(log_root))
1439 return PTR_ERR(log_root);
1440 WARN_ON(fs_info->log_root_tree);
1441 fs_info->log_root_tree = log_root;
1442 return 0;
1443}
1444
1445int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1446 struct btrfs_root *root)
1447{
1448 struct btrfs_root *log_root;
1449 struct btrfs_inode_item *inode_item;
1450
1451 log_root = alloc_log_tree(trans, root->fs_info);
1452 if (IS_ERR(log_root))
1453 return PTR_ERR(log_root);
1454
1455 log_root->last_trans = trans->transid;
1456 log_root->root_key.offset = root->root_key.objectid;
1457
1458 inode_item = &log_root->root_item.inode;
3cae210f
QW		 1459 btrfs_set_stack_inode_generation(inode_item, 1);
1460 btrfs_set_stack_inode_size(inode_item, 3);
1461 btrfs_set_stack_inode_nlink(inode_item, 1);
707e8a07 1462 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
3cae210f 1463 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
7237f183 1464
5d4f98a2 1465 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1466
1467 WARN_ON(root->log_root);
1468 root->log_root = log_root;
1469 root->log_transid = 0;
d1433deb 1470 root->log_transid_committed = -1;
257c62e1 1471 root->last_log_commit = 0;
e02119d5
CM		 1472 return 0;
1473}
1474
35a3621b
SB		 1475static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1476 struct btrfs_key *key)
e02119d5
CM		 1477{
1478 struct btrfs_root *root;
1479 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1480 struct btrfs_path *path;
84234f3a 1481 u64 generation;
db94535d 1482 u32 blocksize;
cb517eab 1483 int ret;
0f7d52f4 1484
cb517eab
MX		 1485 path = btrfs_alloc_path();
1486 if (!path)
0f7d52f4 1487 return ERR_PTR(-ENOMEM);
cb517eab
MX		 1488
1489 root = btrfs_alloc_root(fs_info);
1490 if (!root) {
1491 ret = -ENOMEM;
1492 goto alloc_fail;
0f7d52f4
CM		 1493 }
1494
707e8a07
DS		 1495 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1496 tree_root->stripesize, root, fs_info, key->objectid);
0f7d52f4 1497
cb517eab
MX		 1498 ret = btrfs_find_root(tree_root, key, path,
1499 &root->root_item, &root->root_key);
0f7d52f4 1500 if (ret) {
13a8a7c8
YZ		 1501 if (ret > 0)
1502 ret = -ENOENT;
cb517eab 1503 goto find_fail;
0f7d52f4 1504 }
13a8a7c8 1505
84234f3a 1506 generation = btrfs_root_generation(&root->root_item);
707e8a07 1507 blocksize = root->nodesize;
db94535d 1508 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1509 blocksize, generation);
cb517eab
MX		 1510 if (!root->node) {
1511 ret = -ENOMEM;
1512 goto find_fail;
1513 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1514 ret = -EIO;
1515 goto read_fail;
416bc658 1516 }
5d4f98a2 1517 root->commit_root = btrfs_root_node(root);
13a8a7c8 1518out:
cb517eab
MX		 1519 btrfs_free_path(path);
1520 return root;
1521
1522read_fail:
1523 free_extent_buffer(root->node);
1524find_fail:
1525 kfree(root);
1526alloc_fail:
1527 root = ERR_PTR(ret);
1528 goto out;
1529}
1530
1531struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1532 struct btrfs_key *location)
1533{
1534 struct btrfs_root *root;
1535
1536 root = btrfs_read_tree_root(tree_root, location);
1537 if (IS_ERR(root))
1538 return root;
1539
1540 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
27cdeb70 1541 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
08fe4db1
LZ		 1542 btrfs_check_and_init_root_item(&root->root_item);
1543 }
13a8a7c8 1544
5eda7b5e
CM		 1545 return root;
1546}
1547
cb517eab
MX		 1548int btrfs_init_fs_root(struct btrfs_root *root)
1549{
1550 int ret;
8257b2dc 1551 struct btrfs_subvolume_writers *writers;
cb517eab
MX		 1552
1553 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1554 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1555 GFP_NOFS);
1556 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1557 ret = -ENOMEM;
1558 goto fail;
1559 }
1560
8257b2dc
MX		 1561 writers = btrfs_alloc_subvolume_writers();
1562 if (IS_ERR(writers)) {
1563 ret = PTR_ERR(writers);
1564 goto fail;
1565 }
1566 root->subv_writers = writers;
1567
cb517eab 1568 btrfs_init_free_ino_ctl(root);
57cdc8db
DS		 1569 spin_lock_init(&root->ino_cache_lock);
1570 init_waitqueue_head(&root->ino_cache_wait);
cb517eab
MX		 1571
1572 ret = get_anon_bdev(&root->anon_dev);
1573 if (ret)
8257b2dc 1574 goto free_writers;
cb517eab 1575 return 0;
8257b2dc
MX		 1576
1577free_writers:
1578 btrfs_free_subvolume_writers(root->subv_writers);
cb517eab
MX		 1579fail:
1580 kfree(root->free_ino_ctl);
1581 kfree(root->free_ino_pinned);
1582 return ret;
1583}
1584
171170c1
ST		 1585static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1586 u64 root_id)
cb517eab
MX		 1587{
1588 struct btrfs_root *root;
1589
1590 spin_lock(&fs_info->fs_roots_radix_lock);
1591 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1592 (unsigned long)root_id);
1593 spin_unlock(&fs_info->fs_roots_radix_lock);
1594 return root;
1595}
1596
1597int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1598 struct btrfs_root *root)
1599{
1600 int ret;
1601
1602 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1603 if (ret)
1604 return ret;
1605
1606 spin_lock(&fs_info->fs_roots_radix_lock);
1607 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1608 (unsigned long)root->root_key.objectid,
1609 root);
1610 if (ret == 0)
27cdeb70 1611 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
cb517eab
MX		 1612 spin_unlock(&fs_info->fs_roots_radix_lock);
1613 radix_tree_preload_end();
1614
1615 return ret;
1616}
1617
c00869f1
MX		 1618struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1619 struct btrfs_key *location,
1620 bool check_ref)
5eda7b5e
CM		 1621{
1622 struct btrfs_root *root;
1623 int ret;
1624
edbd8d4e
CM		 1625 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1626 return fs_info->tree_root;
1627 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1628 return fs_info->extent_root;
8f18cf13
CM		 1629 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1630 return fs_info->chunk_root;
1631 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1632 return fs_info->dev_root;
0403e47e
YZ		 1633 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1634 return fs_info->csum_root;
bcef60f2
AJ		 1635 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1636 return fs_info->quota_root ? fs_info->quota_root :
1637 ERR_PTR(-ENOENT);
f7a81ea4
SB		 1638 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1639 return fs_info->uuid_root ? fs_info->uuid_root :
1640 ERR_PTR(-ENOENT);
4df27c4d 1641again:
cb517eab 1642 root = btrfs_lookup_fs_root(fs_info, location->objectid);
48475471 1643 if (root) {
c00869f1 1644 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
48475471 1645 return ERR_PTR(-ENOENT);
5eda7b5e 1646 return root;
48475471 1647 }
5eda7b5e 1648
cb517eab 1649 root = btrfs_read_fs_root(fs_info->tree_root, location);
5eda7b5e
CM		 1650 if (IS_ERR(root))
1651 return root;
3394e160 1652
c00869f1 1653 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
cb517eab 1654 ret = -ENOENT;
581bb050 1655 goto fail;
35a30d7c 1656 }
581bb050 1657
cb517eab 1658 ret = btrfs_init_fs_root(root);
ac08aedf
CM		 1659 if (ret)
1660 goto fail;
3394e160 1661
3f870c28
KN		 1662 ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID,
1663 location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL);
d68fc57b
YZ		 1664 if (ret < 0)
1665 goto fail;
1666 if (ret == 0)
27cdeb70 1667 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
d68fc57b 1668
cb517eab 1669 ret = btrfs_insert_fs_root(fs_info, root);
0f7d52f4 1670 if (ret) {
4df27c4d
YZ		 1671 if (ret == -EEXIST) {
1672 free_fs_root(root);
1673 goto again;
1674 }
1675 goto fail;
0f7d52f4 1676 }
edbd8d4e 1677 return root;
4df27c4d
YZ		 1678fail:
1679 free_fs_root(root);
1680 return ERR_PTR(ret);
edbd8d4e
CM		 1681}
1682
04160088
CM		 1683static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1684{
1685 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1686 int ret = 0;
04160088
CM		 1687 struct btrfs_device *device;
1688 struct backing_dev_info *bdi;
b7967db7 1689
1f78160c
XG		 1690 rcu_read_lock();
1691 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1692 if (!device->bdev)
1693 continue;
04160088
CM		 1694 bdi = blk_get_backing_dev_info(device->bdev);
1695 if (bdi && bdi_congested(bdi, bdi_bits)) {
1696 ret = 1;
1697 break;
1698 }
1699 }
1f78160c 1700 rcu_read_unlock();
04160088
CM		 1701 return ret;
1702}
1703
04160088
CM		 1704static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1705{
ad081f14
JA		 1706 int err;
1707
1708 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1709 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1710 if (err)
1711 return err;
1712
4575c9cc 1713 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1714 bdi->congested_fn = btrfs_congested_fn;
1715 bdi->congested_data = info;
1716 return 0;
1717}
1718
8b712842
CM		 1719/*
1720 * called by the kthread helper functions to finally call the bio end_io
1721 * functions. This is where read checksum verification actually happens
1722 */
1723static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1724{
ce9adaa5 1725 struct bio *bio;
8b712842 1726 struct end_io_wq *end_io_wq;
ce9adaa5 1727 int error;
ce9adaa5 1728
8b712842
CM		 1729 end_io_wq = container_of(work, struct end_io_wq, work);
1730 bio = end_io_wq->bio;
ce9adaa5 1731
8b712842
CM		 1732 error = end_io_wq->error;
1733 bio->bi_private = end_io_wq->private;
1734 bio->bi_end_io = end_io_wq->end_io;
1735 kfree(end_io_wq);
bc1e79ac 1736 bio_endio_nodec(bio, error);
44b8bd7e
CM		 1737}
1738
a74a4b97
CM		 1739static int cleaner_kthread(void *arg)
1740{
1741 struct btrfs_root *root = arg;
d0278245 1742 int again;
a74a4b97
CM		 1743
1744 do {
d0278245 1745 again = 0;
a74a4b97 1746
d0278245 1747 /* Make the cleaner go to sleep early. */
babbf170 1748 if (btrfs_need_cleaner_sleep(root))
d0278245
MX		 1749 goto sleep;
1750
1751 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1752 goto sleep;
1753
dc7f370c
MX		 1754 /*
1755 * Avoid the problem that we change the status of the fs
1756 * during the above check and trylock.
1757 */
babbf170 1758 if (btrfs_need_cleaner_sleep(root)) {
dc7f370c
MX		 1759 mutex_unlock(&root->fs_info->cleaner_mutex);
1760 goto sleep;
76dda93c 1761 }
a74a4b97 1762
d0278245 1763 btrfs_run_delayed_iputs(root);
47ab2a6c 1764 btrfs_delete_unused_bgs(root->fs_info);
d0278245
MX		 1765 again = btrfs_clean_one_deleted_snapshot(root);
1766 mutex_unlock(&root->fs_info->cleaner_mutex);
1767
1768 /*
05323cd1
MX		 1769 * The defragger has dealt with the R/O remount and umount,
1770 * needn't do anything special here.
d0278245
MX		 1771 */
1772 btrfs_run_defrag_inodes(root->fs_info);
1773sleep:
9d1a2a3a 1774 if (!try_to_freeze() && !again) {
a74a4b97 1775 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1776 if (!kthread_should_stop())
1777 schedule();
a74a4b97
CM		 1778 __set_current_state(TASK_RUNNING);
1779 }
1780 } while (!kthread_should_stop());
1781 return 0;
1782}
1783
1784static int transaction_kthread(void *arg)
1785{
1786 struct btrfs_root *root = arg;
1787 struct btrfs_trans_handle *trans;
1788 struct btrfs_transaction *cur;
8929ecfa 1789 u64 transid;
a74a4b97
CM		 1790 unsigned long now;
1791 unsigned long delay;
914b2007 1792 bool cannot_commit;
a74a4b97
CM		 1793
1794 do {
914b2007 1795 cannot_commit = false;
8b87dc17 1796 delay = HZ * root->fs_info->commit_interval;
a74a4b97
CM		 1797 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1798
a4abeea4 1799 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1800 cur = root->fs_info->running_transaction;
1801 if (!cur) {
a4abeea4 1802 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1803 goto sleep;
1804 }
31153d81 1805
a74a4b97 1806 now = get_seconds();
4a9d8bde 1807 if (cur->state < TRANS_STATE_BLOCKED &&
8b87dc17
DS		 1808 (now < cur->start_time ||
1809 now - cur->start_time < root->fs_info->commit_interval)) {
a4abeea4 1810 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1811 delay = HZ * 5;
1812 goto sleep;
1813 }
8929ecfa 1814 transid = cur->transid;
a4abeea4 1815 spin_unlock(&root->fs_info->trans_lock);
56bec294 1816
79787eaa 1817 /* If the file system is aborted, this will always fail. */
354aa0fb 1818 trans = btrfs_attach_transaction(root);
914b2007 1819 if (IS_ERR(trans)) {
354aa0fb
MX		 1820 if (PTR_ERR(trans) != -ENOENT)
1821 cannot_commit = true;
79787eaa 1822 goto sleep;
914b2007 1823 }
8929ecfa 1824 if (transid == trans->transid) {
79787eaa 1825 btrfs_commit_transaction(trans, root);
8929ecfa
YZ		 1826 } else {
1827 btrfs_end_transaction(trans, root);
1828 }
a74a4b97
CM		 1829sleep:
1830 wake_up_process(root->fs_info->cleaner_kthread);
1831 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1832
4e121c06
JB		 1833 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1834 &root->fs_info->fs_state)))
1835 btrfs_cleanup_transaction(root);
a0acae0e 1836 if (!try_to_freeze()) {
a74a4b97 1837 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa 1838 if (!kthread_should_stop() &&
914b2007
JK		 1839 (!btrfs_transaction_blocked(root->fs_info) ||
1840 cannot_commit))
8929ecfa 1841 schedule_timeout(delay);
a74a4b97
CM		 1842 __set_current_state(TASK_RUNNING);
1843 }
1844 } while (!kthread_should_stop());
1845 return 0;
1846}
1847
af31f5e5
CM		 1848/*
1849 * this will find the highest generation in the array of
1850 * root backups. The index of the highest array is returned,
1851 * or -1 if we can't find anything.
1852 *
1853 * We check to make sure the array is valid by comparing the
1854 * generation of the latest root in the array with the generation
1855 * in the super block. If they don't match we pitch it.
1856 */
1857static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1858{
1859 u64 cur;
1860 int newest_index = -1;
1861 struct btrfs_root_backup *root_backup;
1862 int i;
1863
1864 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1865 root_backup = info->super_copy->super_roots + i;
1866 cur = btrfs_backup_tree_root_gen(root_backup);
1867 if (cur == newest_gen)
1868 newest_index = i;
1869 }
1870
1871 /* check to see if we actually wrapped around */
1872 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1873 root_backup = info->super_copy->super_roots;
1874 cur = btrfs_backup_tree_root_gen(root_backup);
1875 if (cur == newest_gen)
1876 newest_index = 0;
1877 }
1878 return newest_index;
1879}
1880
1881
1882/*
1883 * find the oldest backup so we know where to store new entries
1884 * in the backup array. This will set the backup_root_index
1885 * field in the fs_info struct
1886 */
1887static void find_oldest_super_backup(struct btrfs_fs_info *info,
1888 u64 newest_gen)
1889{
1890 int newest_index = -1;
1891
1892 newest_index = find_newest_super_backup(info, newest_gen);
1893 /* if there was garbage in there, just move along */
1894 if (newest_index == -1) {
1895 info->backup_root_index = 0;
1896 } else {
1897 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1898 }
1899}
1900
1901/*
1902 * copy all the root pointers into the super backup array.
1903 * this will bump the backup pointer by one when it is
1904 * done
1905 */
1906static void backup_super_roots(struct btrfs_fs_info *info)
1907{
1908 int next_backup;
1909 struct btrfs_root_backup *root_backup;
1910 int last_backup;
1911
1912 next_backup = info->backup_root_index;
1913 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1914 BTRFS_NUM_BACKUP_ROOTS;
1915
1916 /*
1917 * just overwrite the last backup if we're at the same generation
1918 * this happens only at umount
1919 */
1920 root_backup = info->super_for_commit->super_roots + last_backup;
1921 if (btrfs_backup_tree_root_gen(root_backup) ==
1922 btrfs_header_generation(info->tree_root->node))
1923 next_backup = last_backup;
1924
1925 root_backup = info->super_for_commit->super_roots + next_backup;
1926
1927 /*
1928 * make sure all of our padding and empty slots get zero filled
1929 * regardless of which ones we use today
1930 */
1931 memset(root_backup, 0, sizeof(*root_backup));
1932
1933 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1934
1935 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1936 btrfs_set_backup_tree_root_gen(root_backup,
1937 btrfs_header_generation(info->tree_root->node));
1938
1939 btrfs_set_backup_tree_root_level(root_backup,
1940 btrfs_header_level(info->tree_root->node));
1941
1942 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1943 btrfs_set_backup_chunk_root_gen(root_backup,
1944 btrfs_header_generation(info->chunk_root->node));
1945 btrfs_set_backup_chunk_root_level(root_backup,
1946 btrfs_header_level(info->chunk_root->node));
1947
1948 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1949 btrfs_set_backup_extent_root_gen(root_backup,
1950 btrfs_header_generation(info->extent_root->node));
1951 btrfs_set_backup_extent_root_level(root_backup,
1952 btrfs_header_level(info->extent_root->node));
1953
7c7e82a7
CM		 1954 /*
1955 * we might commit during log recovery, which happens before we set
1956 * the fs_root. Make sure it is valid before we fill it in.
1957 */
1958 if (info->fs_root && info->fs_root->node) {
1959 btrfs_set_backup_fs_root(root_backup,
1960 info->fs_root->node->start);
1961 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 1962 btrfs_header_generation(info->fs_root->node));
7c7e82a7 1963 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 1964 btrfs_header_level(info->fs_root->node));
7c7e82a7 1965 }
af31f5e5
CM		 1966
1967 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1968 btrfs_set_backup_dev_root_gen(root_backup,
1969 btrfs_header_generation(info->dev_root->node));
1970 btrfs_set_backup_dev_root_level(root_backup,
1971 btrfs_header_level(info->dev_root->node));
1972
1973 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1974 btrfs_set_backup_csum_root_gen(root_backup,
1975 btrfs_header_generation(info->csum_root->node));
1976 btrfs_set_backup_csum_root_level(root_backup,
1977 btrfs_header_level(info->csum_root->node));
1978
1979 btrfs_set_backup_total_bytes(root_backup,
1980 btrfs_super_total_bytes(info->super_copy));
1981 btrfs_set_backup_bytes_used(root_backup,
1982 btrfs_super_bytes_used(info->super_copy));
1983 btrfs_set_backup_num_devices(root_backup,
1984 btrfs_super_num_devices(info->super_copy));
1985
1986 /*
1987 * if we don't copy this out to the super_copy, it won't get remembered
1988 * for the next commit
1989 */
1990 memcpy(&info->super_copy->super_roots,
1991 &info->super_for_commit->super_roots,
1992 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1993}
1994
1995/*
1996 * this copies info out of the root backup array and back into
1997 * the in-memory super block. It is meant to help iterate through
1998 * the array, so you send it the number of backups you've already
1999 * tried and the last backup index you used.
2000 *
2001 * this returns -1 when it has tried all the backups
2002 */
2003static noinline int next_root_backup(struct btrfs_fs_info *info,
2004 struct btrfs_super_block *super,
2005 int *num_backups_tried, int *backup_index)
2006{
2007 struct btrfs_root_backup *root_backup;
2008 int newest = *backup_index;
2009
2010 if (*num_backups_tried == 0) {
2011 u64 gen = btrfs_super_generation(super);
2012
2013 newest = find_newest_super_backup(info, gen);
2014 if (newest == -1)
2015 return -1;
2016
2017 *backup_index = newest;
2018 *num_backups_tried = 1;
2019 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2020 /* we've tried all the backups, all done */
2021 return -1;
2022 } else {
2023 /* jump to the next oldest backup */
2024 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2025 BTRFS_NUM_BACKUP_ROOTS;
2026 *backup_index = newest;
2027 *num_backups_tried += 1;
2028 }
2029 root_backup = super->super_roots + newest;
2030
2031 btrfs_set_super_generation(super,
2032 btrfs_backup_tree_root_gen(root_backup));
2033 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2034 btrfs_set_super_root_level(super,
2035 btrfs_backup_tree_root_level(root_backup));
2036 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2037
2038 /*
2039 * fixme: the total bytes and num_devices need to match or we should
2040 * need a fsck
2041 */
2042 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2043 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2044 return 0;
2045}
2046
7abadb64
LB		 2047/* helper to cleanup workers */
2048static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2049{
dc6e3209 2050 btrfs_destroy_workqueue(fs_info->fixup_workers);
afe3d242 2051 btrfs_destroy_workqueue(fs_info->delalloc_workers);
5cdc7ad3 2052 btrfs_destroy_workqueue(fs_info->workers);
fccb5d86
QW		 2053 btrfs_destroy_workqueue(fs_info->endio_workers);
2054 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2055 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
8b110e39 2056 btrfs_destroy_workqueue(fs_info->endio_repair_workers);
d05a33ac 2057 btrfs_destroy_workqueue(fs_info->rmw_workers);
fccb5d86
QW		 2058 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2059 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2060 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
a8c93d4e 2061 btrfs_destroy_workqueue(fs_info->submit_workers);
5b3bc44e 2062 btrfs_destroy_workqueue(fs_info->delayed_workers);
e66f0bb1 2063 btrfs_destroy_workqueue(fs_info->caching_workers);
736cfa15 2064 btrfs_destroy_workqueue(fs_info->readahead_workers);
a44903ab 2065 btrfs_destroy_workqueue(fs_info->flush_workers);
fc97fab0 2066 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
a79b7d4b 2067 btrfs_destroy_workqueue(fs_info->extent_workers);
7abadb64
LB		 2068}
2069
2e9f5954
R		 2070static void free_root_extent_buffers(struct btrfs_root *root)
2071{
2072 if (root) {
2073 free_extent_buffer(root->node);
2074 free_extent_buffer(root->commit_root);
2075 root->node = NULL;
2076 root->commit_root = NULL;
2077 }
2078}
2079
af31f5e5
CM		 2080/* helper to cleanup tree roots */
2081static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2082{
2e9f5954 2083 free_root_extent_buffers(info->tree_root);
655b09fe 2084
2e9f5954
R		 2085 free_root_extent_buffers(info->dev_root);
2086 free_root_extent_buffers(info->extent_root);
2087 free_root_extent_buffers(info->csum_root);
2088 free_root_extent_buffers(info->quota_root);
2089 free_root_extent_buffers(info->uuid_root);
2090 if (chunk_root)
2091 free_root_extent_buffers(info->chunk_root);
af31f5e5
CM		 2092}
2093
faa2dbf0 2094void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
171f6537
JB		 2095{
2096 int ret;
2097 struct btrfs_root *gang[8];
2098 int i;
2099
2100 while (!list_empty(&fs_info->dead_roots)) {
2101 gang[0] = list_entry(fs_info->dead_roots.next,
2102 struct btrfs_root, root_list);
2103 list_del(&gang[0]->root_list);
2104
27cdeb70 2105 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
cb517eab 2106 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
171f6537
JB		 2107 } else {
2108 free_extent_buffer(gang[0]->node);
2109 free_extent_buffer(gang[0]->commit_root);
b0feb9d9 2110 btrfs_put_fs_root(gang[0]);
171f6537
JB		 2111 }
2112 }
2113
2114 while (1) {
2115 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2116 (void **)gang, 0,
2117 ARRAY_SIZE(gang));
2118 if (!ret)
2119 break;
2120 for (i = 0; i < ret; i++)
cb517eab 2121 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
171f6537 2122 }
1a4319cc
LB		 2123
2124 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2125 btrfs_free_log_root_tree(NULL, fs_info);
2126 btrfs_destroy_pinned_extent(fs_info->tree_root,
2127 fs_info->pinned_extents);
2128 }
171f6537 2129}
af31f5e5 2130
ad2b2c80
AV		 2131int open_ctree(struct super_block *sb,
2132 struct btrfs_fs_devices *fs_devices,
2133 char *options)
2e635a27 2134{
db94535d
CM		 2135 u32 sectorsize;
2136 u32 nodesize;
db94535d 2137 u32 blocksize;
87ee04eb 2138 u32 stripesize;
84234f3a 2139 u64 generation;
f2b636e8 2140 u64 features;
3de4586c 2141 struct btrfs_key location;
a061fc8d 2142 struct buffer_head *bh;
4d34b278 2143 struct btrfs_super_block *disk_super;
815745cf 2144 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2145 struct btrfs_root *tree_root;
4d34b278
ID		 2146 struct btrfs_root *extent_root;
2147 struct btrfs_root *csum_root;
2148 struct btrfs_root *chunk_root;
2149 struct btrfs_root *dev_root;
bcef60f2 2150 struct btrfs_root *quota_root;
f7a81ea4 2151 struct btrfs_root *uuid_root;
e02119d5 2152 struct btrfs_root *log_tree_root;
eb60ceac 2153 int ret;
e58ca020 2154 int err = -EINVAL;
af31f5e5
CM		 2155 int num_backups_tried = 0;
2156 int backup_index = 0;
5cdc7ad3
QW		 2157 int max_active;
2158 int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
70f80175
SB		 2159 bool create_uuid_tree;
2160 bool check_uuid_tree;
4543df7e 2161
f84a8bd6 2162 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
6f07e42e 2163 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
cb517eab 2164 if (!tree_root || !chunk_root) {
39279cc3
CM		 2165 err = -ENOMEM;
2166 goto fail;
2167 }
76dda93c
YZ		 2168
2169 ret = init_srcu_struct(&fs_info->subvol_srcu);
2170 if (ret) {
2171 err = ret;
2172 goto fail;
2173 }
2174
2175 ret = setup_bdi(fs_info, &fs_info->bdi);
2176 if (ret) {
2177 err = ret;
2178 goto fail_srcu;
2179 }
2180
e2d84521
MX		 2181 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2182 if (ret) {
2183 err = ret;
2184 goto fail_bdi;
2185 }
2186 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2187 (1 + ilog2(nr_cpu_ids));
2188
963d678b
MX		 2189 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2190 if (ret) {
2191 err = ret;
2192 goto fail_dirty_metadata_bytes;
2193 }
2194
c404e0dc
MX		 2195 ret = percpu_counter_init(&fs_info->bio_counter, 0);
2196 if (ret) {
2197 err = ret;
2198 goto fail_delalloc_bytes;
2199 }
2200
76dda93c
YZ		 2201 fs_info->btree_inode = new_inode(sb);
2202 if (!fs_info->btree_inode) {
2203 err = -ENOMEM;
c404e0dc 2204 goto fail_bio_counter;
76dda93c
YZ		 2205 }
2206
a6591715 2207 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2208
76dda93c 2209 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
f28491e0 2210 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
8fd17795 2211 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2212 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2213 INIT_LIST_HEAD(&fs_info->delayed_iputs);
eb73c1b7 2214 INIT_LIST_HEAD(&fs_info->delalloc_roots);
11833d66 2215 INIT_LIST_HEAD(&fs_info->caching_block_groups);
eb73c1b7 2216 spin_lock_init(&fs_info->delalloc_root_lock);
a4abeea4 2217 spin_lock_init(&fs_info->trans_lock);
76dda93c 2218 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2219 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2220 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 2221 spin_lock_init(&fs_info->free_chunk_lock);
f29021b2 2222 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2223 spin_lock_init(&fs_info->super_lock);
fcebe456 2224 spin_lock_init(&fs_info->qgroup_op_lock);
f28491e0 2225 spin_lock_init(&fs_info->buffer_lock);
47ab2a6c 2226 spin_lock_init(&fs_info->unused_bgs_lock);
f29021b2 2227 rwlock_init(&fs_info->tree_mod_log_lock);
7585717f 2228 mutex_init(&fs_info->reloc_mutex);
573bfb72 2229 mutex_init(&fs_info->delalloc_root_mutex);
de98ced9 2230 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2231
58176a96 2232 init_completion(&fs_info->kobj_unregister);
0b86a832 2233 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2234 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2235 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
47ab2a6c 2236 INIT_LIST_HEAD(&fs_info->unused_bgs);
0b86a832 2237 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX		 2238 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2239 BTRFS_BLOCK_RSV_GLOBAL);
2240 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2241 BTRFS_BLOCK_RSV_DELALLOC);
2242 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2243 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2244 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2245 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2246 BTRFS_BLOCK_RSV_DELOPS);
cb03c743 2247 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 2248 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 2249 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 2250 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 2251 atomic_set(&fs_info->defrag_running, 0);
fcebe456 2252 atomic_set(&fs_info->qgroup_op_seq, 0);
fc36ed7e 2253 atomic64_set(&fs_info->tree_mod_seq, 0);
e20d96d6 2254 fs_info->sb = sb;
95ac567a 2255 fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE;
9ed74f2d 2256 fs_info->metadata_ratio = 0;
4cb5300b 2257 fs_info->defrag_inodes = RB_ROOT;
2bf64758 2258 fs_info->free_chunk_space = 0;
f29021b2 2259 fs_info->tree_mod_log = RB_ROOT;
8b87dc17 2260 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
0a2b2a84 2261 fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
90519d66
AJ		 2262 /* readahead state */
2263 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2264 spin_lock_init(&fs_info->reada_lock);
c8b97818 2265
b34b086c
CM		 2266 fs_info->thread_pool_size = min_t(unsigned long,
2267 num_online_cpus() + 2, 8);
0afbaf8c 2268
199c2a9c
MX		 2269 INIT_LIST_HEAD(&fs_info->ordered_roots);
2270 spin_lock_init(&fs_info->ordered_root_lock);
16cdcec7
MX		 2271 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2272 GFP_NOFS);
2273 if (!fs_info->delayed_root) {
2274 err = -ENOMEM;
2275 goto fail_iput;
2276 }
2277 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2278
a2de733c
AJ		 2279 mutex_init(&fs_info->scrub_lock);
2280 atomic_set(&fs_info->scrubs_running, 0);
2281 atomic_set(&fs_info->scrub_pause_req, 0);
2282 atomic_set(&fs_info->scrubs_paused, 0);
2283 atomic_set(&fs_info->scrub_cancel_req, 0);
c404e0dc 2284 init_waitqueue_head(&fs_info->replace_wait);
a2de733c 2285 init_waitqueue_head(&fs_info->scrub_pause_wait);
a2de733c 2286 fs_info->scrub_workers_refcnt = 0;
21adbd5c
SB		 2287#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2288 fs_info->check_integrity_print_mask = 0;
2289#endif
a2de733c 2290
c9e9f97b
ID		 2291 spin_lock_init(&fs_info->balance_lock);
2292 mutex_init(&fs_info->balance_mutex);
837d5b6e
ID		 2293 atomic_set(&fs_info->balance_running, 0);
2294 atomic_set(&fs_info->balance_pause_req, 0);
a7e99c69 2295 atomic_set(&fs_info->balance_cancel_req, 0);
c9e9f97b 2296 fs_info->balance_ctl = NULL;
837d5b6e 2297 init_waitqueue_head(&fs_info->balance_wait_q);
21c7e756 2298 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
a2de733c 2299
a061fc8d
CM		 2300 sb->s_blocksize = 4096;
2301 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 2302 sb->s_bdi = &fs_info->bdi;
a061fc8d 2303
76dda93c 2304 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
bfe86848 2305 set_nlink(fs_info->btree_inode, 1);
0afbaf8c
CM		 2306 /*
2307 * we set the i_size on the btree inode to the max possible int.
2308 * the real end of the address space is determined by all of
2309 * the devices in the system
2310 */
2311 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 2312 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 2313 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2314
5d4f98a2 2315 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 2316 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 2317 fs_info->btree_inode->i_mapping);
0b32f4bb 2318 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
a8067e02 2319 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM		 2320
2321 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 2322
76dda93c
YZ		 2323 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2324 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2325 sizeof(struct btrfs_key));
72ac3c0d
JB		 2326 set_bit(BTRFS_INODE_DUMMY,
2327 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
778ba82b 2328 btrfs_insert_inode_hash(fs_info->btree_inode);
76dda93c 2329
0f9dd46c 2330 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2331 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2332 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2333
11833d66 2334 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 2335 fs_info->btree_inode->i_mapping);
11833d66 2336 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 2337 fs_info->btree_inode->i_mapping);
11833d66 2338 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 2339 fs_info->do_barriers = 1;
e18e4809 2340
39279cc3 2341
5a3f23d5 2342 mutex_init(&fs_info->ordered_operations_mutex);
9ffba8cd 2343 mutex_init(&fs_info->ordered_extent_flush_mutex);
e02119d5 2344 mutex_init(&fs_info->tree_log_mutex);
925baedd 2345 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 2346 mutex_init(&fs_info->transaction_kthread_mutex);
2347 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2348 mutex_init(&fs_info->volume_mutex);
9e351cc8 2349 init_rwsem(&fs_info->commit_root_sem);
c71bf099 2350 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2351 init_rwsem(&fs_info->subvol_sem);
803b2f54 2352 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
e922e087
SB		 2353 fs_info->dev_replace.lock_owner = 0;
2354 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2355 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2356 mutex_init(&fs_info->dev_replace.lock_management_lock);
2357 mutex_init(&fs_info->dev_replace.lock);
fa9c0d79 2358
416ac51d 2359 spin_lock_init(&fs_info->qgroup_lock);
f2f6ed3d 2360 mutex_init(&fs_info->qgroup_ioctl_lock);
416ac51d 2361 fs_info->qgroup_tree = RB_ROOT;
fcebe456 2362 fs_info->qgroup_op_tree = RB_ROOT;
416ac51d
AJ		 2363 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2364 fs_info->qgroup_seq = 1;
2365 fs_info->quota_enabled = 0;
2366 fs_info->pending_quota_state = 0;
1e8f9158 2367 fs_info->qgroup_ulist = NULL;
2f232036 2368 mutex_init(&fs_info->qgroup_rescan_lock);
416ac51d 2369
fa9c0d79
CM		 2370 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2371 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2372
e6dcd2dc 2373 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2374 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2375 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2376 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2377
53b381b3
DW		 2378 ret = btrfs_alloc_stripe_hash_table(fs_info);
2379 if (ret) {
83c8266a 2380 err = ret;
53b381b3
DW		 2381 goto fail_alloc;
2382 }
2383
707e8a07 2384 __setup_root(4096, 4096, 4096, tree_root,
2c90e5d6 2385 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2386
3c4bb26b 2387 invalidate_bdev(fs_devices->latest_bdev);
1104a885
DS		 2388
2389 /*
2390 * Read super block and check the signature bytes only
2391 */
a512bbf8 2392 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY		 2393 if (!bh) {
2394 err = -EINVAL;
16cdcec7 2395 goto fail_alloc;
20b45077 2396 }
39279cc3 2397
1104a885
DS		 2398 /*
2399 * We want to check superblock checksum, the type is stored inside.
2400 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2401 */
2402 if (btrfs_check_super_csum(bh->b_data)) {
efe120a0 2403 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
1104a885
DS		 2404 err = -EINVAL;
2405 goto fail_alloc;
2406 }
2407
2408 /*
2409 * super_copy is zeroed at allocation time and we never touch the
2410 * following bytes up to INFO_SIZE, the checksum is calculated from
2411 * the whole block of INFO_SIZE
2412 */
6c41761f
DS		 2413 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2414 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2415 sizeof(*fs_info->super_for_commit));
a061fc8d 2416 brelse(bh);
5f39d397 2417
6c41761f 2418 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2419
1104a885
DS		 2420 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2421 if (ret) {
efe120a0 2422 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
1104a885
DS		 2423 err = -EINVAL;
2424 goto fail_alloc;
2425 }
2426
6c41761f 2427 disk_super = fs_info->super_copy;
0f7d52f4 2428 if (!btrfs_super_root(disk_super))
16cdcec7 2429 goto fail_alloc;
0f7d52f4 2430
acce952b 2431 /* check FS state, whether FS is broken. */
87533c47
MX		 2432 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2433 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2434
af31f5e5
CM		 2435 /*
2436 * run through our array of backup supers and setup
2437 * our ring pointer to the oldest one
2438 */
2439 generation = btrfs_super_generation(disk_super);
2440 find_oldest_super_backup(fs_info, generation);
2441
75e7cb7f
LB		 2442 /*
2443 * In the long term, we'll store the compression type in the super
2444 * block, and it'll be used for per file compression control.
2445 */
2446 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2447
2b82032c
YZ		 2448 ret = btrfs_parse_options(tree_root, options);
2449 if (ret) {
2450 err = ret;
16cdcec7 2451 goto fail_alloc;
2b82032c 2452 }
dfe25020 2453
f2b636e8
JB		 2454 features = btrfs_super_incompat_flags(disk_super) &
2455 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2456 if (features) {
2457 printk(KERN_ERR "BTRFS: couldn't mount because of "
2458 "unsupported optional features (%Lx).\n",
c1c9ff7c 2459 features);
f2b636e8 2460 err = -EINVAL;
16cdcec7 2461 goto fail_alloc;
f2b636e8
JB		 2462 }
2463
707e8a07
DS		 2464 /*
2465 * Leafsize and nodesize were always equal, this is only a sanity check.
2466 */
2467 if (le32_to_cpu(disk_super->__unused_leafsize) !=
727011e0
CM		 2468 btrfs_super_nodesize(disk_super)) {
2469 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2470 "blocksizes don't match. node %d leaf %d\n",
2471 btrfs_super_nodesize(disk_super),
707e8a07 2472 le32_to_cpu(disk_super->__unused_leafsize));
727011e0
CM		 2473 err = -EINVAL;
2474 goto fail_alloc;
2475 }
707e8a07 2476 if (btrfs_super_nodesize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
727011e0
CM		 2477 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2478 "blocksize (%d) was too large\n",
707e8a07 2479 btrfs_super_nodesize(disk_super));
727011e0
CM		 2480 err = -EINVAL;
2481 goto fail_alloc;
2482 }
2483
5d4f98a2 2484 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2485 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
69e380d1 2486 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2487 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
727011e0 2488
3173a18f 2489 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
efe120a0 2490 printk(KERN_ERR "BTRFS: has skinny extents\n");
3173a18f 2491
727011e0
CM		 2492 /*
2493 * flag our filesystem as having big metadata blocks if
2494 * they are bigger than the page size
2495 */
707e8a07 2496 if (btrfs_super_nodesize(disk_super) > PAGE_CACHE_SIZE) {
727011e0 2497 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
efe120a0 2498 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
727011e0
CM		 2499 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2500 }
2501
bc3f116f 2502 nodesize = btrfs_super_nodesize(disk_super);
bc3f116f
CM		 2503 sectorsize = btrfs_super_sectorsize(disk_super);
2504 stripesize = btrfs_super_stripesize(disk_super);
707e8a07 2505 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids));
963d678b 2506 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f
CM		 2507
2508 /*
2509 * mixed block groups end up with duplicate but slightly offset
2510 * extent buffers for the same range. It leads to corruptions
2511 */
2512 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
707e8a07 2513 (sectorsize != nodesize)) {
efe120a0 2514 printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes "
bc3f116f
CM		 2515 "are not allowed for mixed block groups on %s\n",
2516 sb->s_id);
2517 goto fail_alloc;
2518 }
2519
ceda0864
MX		 2520 /*
2521 * Needn't use the lock because there is no other task which will
2522 * update the flag.
2523 */
a6fa6fae 2524 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2525
f2b636e8
JB		 2526 features = btrfs_super_compat_ro_flags(disk_super) &
2527 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2528 if (!(sb->s_flags & MS_RDONLY) && features) {
2529 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2530 "unsupported option features (%Lx).\n",
c1c9ff7c 2531 features);
f2b636e8 2532 err = -EINVAL;
16cdcec7 2533 goto fail_alloc;
f2b636e8 2534 }
61d92c32 2535
5cdc7ad3 2536 max_active = fs_info->thread_pool_size;
61d92c32 2537
5cdc7ad3
QW		 2538 fs_info->workers =
2539 btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2540 max_active, 16);
c8b97818 2541
afe3d242
QW		 2542 fs_info->delalloc_workers =
2543 btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
771ed689 2544
a44903ab
QW		 2545 fs_info->flush_workers =
2546 btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
8ccf6f19 2547
e66f0bb1
QW		 2548 fs_info->caching_workers =
2549 btrfs_alloc_workqueue("cache", flags, max_active, 0);
61b49440 2550
a8c93d4e
QW		 2551 /*
2552 * a higher idle thresh on the submit workers makes it much more
61b49440
CM		 2553 * likely that bios will be send down in a sane order to the
2554 * devices
2555 */
a8c93d4e
QW		 2556 fs_info->submit_workers =
2557 btrfs_alloc_workqueue("submit", flags,
2558 min_t(u64, fs_devices->num_devices,
2559 max_active), 64);
53863232 2560
dc6e3209
QW		 2561 fs_info->fixup_workers =
2562 btrfs_alloc_workqueue("fixup", flags, 1, 0);
61b49440
CM		 2563
2564 /*
2565 * endios are largely parallel and should have a very
2566 * low idle thresh
2567 */
fccb5d86
QW		 2568 fs_info->endio_workers =
2569 btrfs_alloc_workqueue("endio", flags, max_active, 4);
2570 fs_info->endio_meta_workers =
2571 btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2572 fs_info->endio_meta_write_workers =
2573 btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2574 fs_info->endio_raid56_workers =
2575 btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
8b110e39
MX		 2576 fs_info->endio_repair_workers =
2577 btrfs_alloc_workqueue("endio-repair", flags, 1, 0);
d05a33ac
QW		 2578 fs_info->rmw_workers =
2579 btrfs_alloc_workqueue("rmw", flags, max_active, 2);
fccb5d86
QW		 2580 fs_info->endio_write_workers =
2581 btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2582 fs_info->endio_freespace_worker =
2583 btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
5b3bc44e
QW		 2584 fs_info->delayed_workers =
2585 btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
736cfa15
QW		 2586 fs_info->readahead_workers =
2587 btrfs_alloc_workqueue("readahead", flags, max_active, 2);
fc97fab0
QW		 2588 fs_info->qgroup_rescan_workers =
2589 btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
a79b7d4b
CM		 2590 fs_info->extent_workers =
2591 btrfs_alloc_workqueue("extent-refs", flags,
2592 min_t(u64, fs_devices->num_devices,
2593 max_active), 8);
61b49440 2594
a8c93d4e 2595 if (!(fs_info->workers && fs_info->delalloc_workers &&
fccb5d86
QW		 2596 fs_info->submit_workers && fs_info->flush_workers &&
2597 fs_info->endio_workers && fs_info->endio_meta_workers &&
2598 fs_info->endio_meta_write_workers &&
8b110e39 2599 fs_info->endio_repair_workers &&
fccb5d86 2600 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
e66f0bb1 2601 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
dc6e3209 2602 fs_info->caching_workers && fs_info->readahead_workers &&
fc97fab0 2603 fs_info->fixup_workers && fs_info->delayed_workers &&
56094eec 2604 fs_info->extent_workers &&
fc97fab0 2605 fs_info->qgroup_rescan_workers)) {
fed425c7 2606 err = -ENOMEM;
0dc3b84a
JB		 2607 goto fail_sb_buffer;
2608 }
4543df7e 2609
4575c9cc 2610 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 2611 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2612 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 2613
db94535d 2614 tree_root->nodesize = nodesize;
db94535d 2615 tree_root->sectorsize = sectorsize;
87ee04eb 2616 tree_root->stripesize = stripesize;
a061fc8d
CM		 2617
2618 sb->s_blocksize = sectorsize;
2619 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 2620
3cae210f 2621 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
efe120a0 2622 printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 2623 goto fail_sb_buffer;
2624 }
19c00ddc 2625
8d082fb7 2626 if (sectorsize != PAGE_SIZE) {
efe120a0 2627 printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) "
8d082fb7 2628 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
941b2ddf
KM		 2629 goto fail_sb_buffer;
2630 }
2631
925baedd 2632 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 2633 ret = btrfs_read_sys_array(tree_root);
925baedd 2634 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2635 if (ret) {
efe120a0 2636 printk(KERN_WARNING "BTRFS: failed to read the system "
d397712b 2637 "array on %s\n", sb->s_id);
5d4f98a2 2638 goto fail_sb_buffer;
84eed90f 2639 }
0b86a832 2640
707e8a07 2641 blocksize = tree_root->nodesize;
84234f3a 2642 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832 2643
707e8a07
DS		 2644 __setup_root(nodesize, sectorsize, stripesize, chunk_root,
2645 fs_info, BTRFS_CHUNK_TREE_OBJECTID);
0b86a832
CM		 2646
2647 chunk_root->node = read_tree_block(chunk_root,
2648 btrfs_super_chunk_root(disk_super),
84234f3a 2649 blocksize, generation);
416bc658
JB		 2650 if (!chunk_root->node ||
2651 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
efe120a0 2652 printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n",
83121942 2653 sb->s_id);
af31f5e5 2654 goto fail_tree_roots;
83121942 2655 }
5d4f98a2
YZ		 2656 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2657 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2658
e17cade2 2659 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
b308bc2f 2660 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
e17cade2 2661
0b86a832 2662 ret = btrfs_read_chunk_tree(chunk_root);
2b82032c 2663 if (ret) {
efe120a0 2664 printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n",
d397712b 2665 sb->s_id);
af31f5e5 2666 goto fail_tree_roots;
2b82032c 2667 }
0b86a832 2668
8dabb742
SB		 2669 /*
2670 * keep the device that is marked to be the target device for the
2671 * dev_replace procedure
2672 */
2673 btrfs_close_extra_devices(fs_info, fs_devices, 0);
dfe25020 2674
a6b0d5c8 2675 if (!fs_devices->latest_bdev) {
efe120a0 2676 printk(KERN_CRIT "BTRFS: failed to read devices on %s\n",
a6b0d5c8
CM		 2677 sb->s_id);
2678 goto fail_tree_roots;
2679 }
2680
af31f5e5 2681retry_root_backup:
707e8a07 2682 blocksize = tree_root->nodesize;
84234f3a 2683 generation = btrfs_super_generation(disk_super);
0b86a832 2684
e20d96d6 2685 tree_root->node = read_tree_block(tree_root,
db94535d 2686 btrfs_super_root(disk_super),
84234f3a 2687 blocksize, generation);
af31f5e5
CM		 2688 if (!tree_root->node ||
2689 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
efe120a0 2690 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
83121942 2691 sb->s_id);
af31f5e5
CM		 2692
2693 goto recovery_tree_root;
83121942 2694 }
af31f5e5 2695
5d4f98a2
YZ		 2696 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2697 tree_root->commit_root = btrfs_root_node(tree_root);
69e9c6c6 2698 btrfs_set_root_refs(&tree_root->root_item, 1);
db94535d 2699
cb517eab
MX		 2700 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2701 location.type = BTRFS_ROOT_ITEM_KEY;
2702 location.offset = 0;
2703
2704 extent_root = btrfs_read_tree_root(tree_root, &location);
2705 if (IS_ERR(extent_root)) {
2706 ret = PTR_ERR(extent_root);
af31f5e5 2707 goto recovery_tree_root;
cb517eab 2708 }
27cdeb70 2709 set_bit(BTRFS_ROOT_TRACK_DIRTY, &extent_root->state);
cb517eab 2710 fs_info->extent_root = extent_root;
0b86a832 2711
cb517eab
MX		 2712 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2713 dev_root = btrfs_read_tree_root(tree_root, &location);
2714 if (IS_ERR(dev_root)) {
2715 ret = PTR_ERR(dev_root);
af31f5e5 2716 goto recovery_tree_root;
cb517eab 2717 }
27cdeb70 2718 set_bit(BTRFS_ROOT_TRACK_DIRTY, &dev_root->state);
cb517eab
MX		 2719 fs_info->dev_root = dev_root;
2720 btrfs_init_devices_late(fs_info);
3768f368 2721
cb517eab
MX		 2722 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2723 csum_root = btrfs_read_tree_root(tree_root, &location);
2724 if (IS_ERR(csum_root)) {
2725 ret = PTR_ERR(csum_root);
af31f5e5 2726 goto recovery_tree_root;
cb517eab 2727 }
27cdeb70 2728 set_bit(BTRFS_ROOT_TRACK_DIRTY, &csum_root->state);
cb517eab 2729 fs_info->csum_root = csum_root;
d20f7043 2730
cb517eab
MX		 2731 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2732 quota_root = btrfs_read_tree_root(tree_root, &location);
2733 if (!IS_ERR(quota_root)) {
27cdeb70 2734 set_bit(BTRFS_ROOT_TRACK_DIRTY, &quota_root->state);
bcef60f2
AJ		 2735 fs_info->quota_enabled = 1;
2736 fs_info->pending_quota_state = 1;
cb517eab 2737 fs_info->quota_root = quota_root;
bcef60f2
AJ		 2738 }
2739
f7a81ea4
SB		 2740 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2741 uuid_root = btrfs_read_tree_root(tree_root, &location);
2742 if (IS_ERR(uuid_root)) {
2743 ret = PTR_ERR(uuid_root);
2744 if (ret != -ENOENT)
2745 goto recovery_tree_root;
2746 create_uuid_tree = true;
70f80175 2747 check_uuid_tree = false;
f7a81ea4 2748 } else {
27cdeb70 2749 set_bit(BTRFS_ROOT_TRACK_DIRTY, &uuid_root->state);
f7a81ea4 2750 fs_info->uuid_root = uuid_root;
70f80175
SB		 2751 create_uuid_tree = false;
2752 check_uuid_tree =
2753 generation != btrfs_super_uuid_tree_generation(disk_super);
f7a81ea4
SB		 2754 }
2755
8929ecfa
YZ		 2756 fs_info->generation = generation;
2757 fs_info->last_trans_committed = generation;
8929ecfa 2758
68310a5e
ID		 2759 ret = btrfs_recover_balance(fs_info);
2760 if (ret) {
efe120a0 2761 printk(KERN_WARNING "BTRFS: failed to recover balance\n");
68310a5e
ID		 2762 goto fail_block_groups;
2763 }
2764
733f4fbb
SB		 2765 ret = btrfs_init_dev_stats(fs_info);
2766 if (ret) {
efe120a0 2767 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
733f4fbb
SB		 2768 ret);
2769 goto fail_block_groups;
2770 }
2771
8dabb742
SB		 2772 ret = btrfs_init_dev_replace(fs_info);
2773 if (ret) {
efe120a0 2774 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
8dabb742
SB		 2775 goto fail_block_groups;
2776 }
2777
2778 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2779
5ac1d209 2780 ret = btrfs_sysfs_add_one(fs_info);
c59021f8 2781 if (ret) {
efe120a0 2782 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
c59021f8 2783 goto fail_block_groups;
2784 }
2785
c59021f8 2786 ret = btrfs_init_space_info(fs_info);
2787 if (ret) {
efe120a0 2788 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
2365dd3c 2789 goto fail_sysfs;
c59021f8 2790 }
2791
1b1d1f66
JB		 2792 ret = btrfs_read_block_groups(extent_root);
2793 if (ret) {
efe120a0 2794 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
2365dd3c 2795 goto fail_sysfs;
1b1d1f66 2796 }
5af3e8cc
SB		 2797 fs_info->num_tolerated_disk_barrier_failures =
2798 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
292fd7fc
SB		 2799 if (fs_info->fs_devices->missing_devices >
2800 fs_info->num_tolerated_disk_barrier_failures &&
2801 !(sb->s_flags & MS_RDONLY)) {
efe120a0
FH		 2802 printk(KERN_WARNING "BTRFS: "
2803 "too many missing devices, writeable mount is not allowed\n");
2365dd3c 2804 goto fail_sysfs;
292fd7fc 2805 }
9078a3e1 2806
a74a4b97
CM		 2807 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2808 "btrfs-cleaner");
57506d50 2809 if (IS_ERR(fs_info->cleaner_kthread))
2365dd3c 2810 goto fail_sysfs;
a74a4b97
CM		 2811
2812 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2813 tree_root,
2814 "btrfs-transaction");
57506d50 2815 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2816 goto fail_cleaner;
a74a4b97 2817
c289811c
CM		 2818 if (!btrfs_test_opt(tree_root, SSD) &&
2819 !btrfs_test_opt(tree_root, NOSSD) &&
2820 !fs_info->fs_devices->rotating) {
efe120a0 2821 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
c289811c
CM		 2822 "mode\n");
2823 btrfs_set_opt(fs_info->mount_opt, SSD);
2824 }
2825
3818aea2
QW		 2826 /* Set the real inode map cache flag */
2827 if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE))
2828 btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE);
2829
21adbd5c
SB		 2830#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2831 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2832 ret = btrfsic_mount(tree_root, fs_devices,
2833 btrfs_test_opt(tree_root,
2834 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2835 1 : 0,
2836 fs_info->check_integrity_print_mask);
2837 if (ret)
efe120a0 2838 printk(KERN_WARNING "BTRFS: failed to initialize"
21adbd5c
SB		 2839 " integrity check module %s\n", sb->s_id);
2840 }
2841#endif
bcef60f2
AJ		 2842 ret = btrfs_read_qgroup_config(fs_info);
2843 if (ret)
2844 goto fail_trans_kthread;
21adbd5c 2845
acce952b 2846 /* do not make disk changes in broken FS */
68ce9682 2847 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM		 2848 u64 bytenr = btrfs_super_log_root(disk_super);
2849
7c2ca468 2850 if (fs_devices->rw_devices == 0) {
efe120a0 2851 printk(KERN_WARNING "BTRFS: log replay required "
d397712b 2852 "on RO media\n");
7c2ca468 2853 err = -EIO;
bcef60f2 2854 goto fail_qgroup;
7c2ca468 2855 }
707e8a07 2856 blocksize = tree_root->nodesize;
d18a2c44 2857
6f07e42e 2858 log_tree_root = btrfs_alloc_root(fs_info);
676e4c86
DC		 2859 if (!log_tree_root) {
2860 err = -ENOMEM;
bcef60f2 2861 goto fail_qgroup;
676e4c86 2862 }
e02119d5 2863
707e8a07 2864 __setup_root(nodesize, sectorsize, stripesize,
e02119d5
CM		 2865 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2866
2867 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 2868 blocksize,
2869 generation + 1);
416bc658
JB		 2870 if (!log_tree_root->node ||
2871 !extent_buffer_uptodate(log_tree_root->node)) {
efe120a0 2872 printk(KERN_ERR "BTRFS: failed to read log tree\n");
416bc658
JB		 2873 free_extent_buffer(log_tree_root->node);
2874 kfree(log_tree_root);
28c16cbb 2875 goto fail_qgroup;
416bc658 2876 }
79787eaa 2877 /* returns with log_tree_root freed on success */
e02119d5 2878 ret = btrfs_recover_log_trees(log_tree_root);
79787eaa
JM		 2879 if (ret) {
2880 btrfs_error(tree_root->fs_info, ret,
2881 "Failed to recover log tree");
2882 free_extent_buffer(log_tree_root->node);
2883 kfree(log_tree_root);
28c16cbb 2884 goto fail_qgroup;
79787eaa 2885 }
e556ce2c
YZ		 2886
2887 if (sb->s_flags & MS_RDONLY) {
79787eaa
JM		 2888 ret = btrfs_commit_super(tree_root);
2889 if (ret)
28c16cbb 2890 goto fail_qgroup;
e556ce2c 2891 }
e02119d5 2892 }
1a40e23b 2893
76dda93c 2894 ret = btrfs_find_orphan_roots(tree_root);
79787eaa 2895 if (ret)
28c16cbb 2896 goto fail_qgroup;
76dda93c 2897
7c2ca468 2898 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b 2899 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2 2900 if (ret)
28c16cbb 2901 goto fail_qgroup;
d68fc57b 2902
5f316481 2903 mutex_lock(&fs_info->cleaner_mutex);
5d4f98a2 2904 ret = btrfs_recover_relocation(tree_root);
5f316481 2905 mutex_unlock(&fs_info->cleaner_mutex);
d7ce5843
MX		 2906 if (ret < 0) {
2907 printk(KERN_WARNING
efe120a0 2908 "BTRFS: failed to recover relocation\n");
d7ce5843 2909 err = -EINVAL;
bcef60f2 2910 goto fail_qgroup;
d7ce5843 2911 }
7c2ca468 2912 }
1a40e23b 2913
3de4586c
CM		 2914 location.objectid = BTRFS_FS_TREE_OBJECTID;
2915 location.type = BTRFS_ROOT_ITEM_KEY;
cb517eab 2916 location.offset = 0;
3de4586c 2917
3de4586c 2918 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
3140c9a3
DC		 2919 if (IS_ERR(fs_info->fs_root)) {
2920 err = PTR_ERR(fs_info->fs_root);
bcef60f2 2921 goto fail_qgroup;
3140c9a3 2922 }
c289811c 2923
2b6ba629
ID		 2924 if (sb->s_flags & MS_RDONLY)
2925 return 0;
59641015 2926
2b6ba629
ID		 2927 down_read(&fs_info->cleanup_work_sem);
2928 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2929 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 2930 up_read(&fs_info->cleanup_work_sem);
2b6ba629
ID		 2931 close_ctree(tree_root);
2932 return ret;
2933 }
2934 up_read(&fs_info->cleanup_work_sem);
59641015 2935
2b6ba629
ID		 2936 ret = btrfs_resume_balance_async(fs_info);
2937 if (ret) {
efe120a0 2938 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
2b6ba629
ID		 2939 close_ctree(tree_root);
2940 return ret;
e3acc2a6
JB		 2941 }
2942
8dabb742
SB		 2943 ret = btrfs_resume_dev_replace_async(fs_info);
2944 if (ret) {
efe120a0 2945 pr_warn("BTRFS: failed to resume dev_replace\n");
8dabb742
SB		 2946 close_ctree(tree_root);
2947 return ret;
2948 }
2949
b382a324
JS		 2950 btrfs_qgroup_rescan_resume(fs_info);
2951
f7a81ea4 2952 if (create_uuid_tree) {
efe120a0 2953 pr_info("BTRFS: creating UUID tree\n");
f7a81ea4
SB		 2954 ret = btrfs_create_uuid_tree(fs_info);
2955 if (ret) {
efe120a0 2956 pr_warn("BTRFS: failed to create the UUID tree %d\n",
f7a81ea4
SB		 2957 ret);
2958 close_ctree(tree_root);
2959 return ret;
2960 }
f420ee1e
SB		 2961 } else if (check_uuid_tree ||
2962 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
efe120a0 2963 pr_info("BTRFS: checking UUID tree\n");
70f80175
SB		 2964 ret = btrfs_check_uuid_tree(fs_info);
2965 if (ret) {
efe120a0 2966 pr_warn("BTRFS: failed to check the UUID tree %d\n",
70f80175
SB		 2967 ret);
2968 close_ctree(tree_root);
2969 return ret;
2970 }
2971 } else {
2972 fs_info->update_uuid_tree_gen = 1;
f7a81ea4
SB		 2973 }
2974
47ab2a6c
JB		 2975 fs_info->open = 1;
2976
ad2b2c80 2977 return 0;
39279cc3 2978
bcef60f2
AJ		 2979fail_qgroup:
2980 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM		 2981fail_trans_kthread:
2982 kthread_stop(fs_info->transaction_kthread);
54067ae9 2983 btrfs_cleanup_transaction(fs_info->tree_root);
faa2dbf0 2984 btrfs_free_fs_roots(fs_info);
3f157a2f 2985fail_cleaner:
a74a4b97 2986 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2987
2988 /*
2989 * make sure we're done with the btree inode before we stop our
2990 * kthreads
2991 */
2992 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 2993
2365dd3c
AJ		 2994fail_sysfs:
2995 btrfs_sysfs_remove_one(fs_info);
2996
1b1d1f66 2997fail_block_groups:
54067ae9 2998 btrfs_put_block_group_cache(fs_info);
1b1d1f66 2999 btrfs_free_block_groups(fs_info);
af31f5e5
CM		 3000
3001fail_tree_roots:
3002 free_root_pointers(fs_info, 1);
2b8195bb 3003 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 3004
39279cc3 3005fail_sb_buffer:
7abadb64 3006 btrfs_stop_all_workers(fs_info);
16cdcec7 3007fail_alloc:
4543df7e 3008fail_iput:
586e46e2
ID		 3009 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3010
4543df7e 3011 iput(fs_info->btree_inode);
c404e0dc
MX		 3012fail_bio_counter:
3013 percpu_counter_destroy(&fs_info->bio_counter);
963d678b
MX		 3014fail_delalloc_bytes:
3015 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX		 3016fail_dirty_metadata_bytes:
3017 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
ad081f14 3018fail_bdi:
7e662854 3019 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 3020fail_srcu:
3021 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 3022fail:
53b381b3 3023 btrfs_free_stripe_hash_table(fs_info);
586e46e2 3024 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 3025 return err;
af31f5e5
CM		 3026
3027recovery_tree_root:
af31f5e5
CM		 3028 if (!btrfs_test_opt(tree_root, RECOVERY))
3029 goto fail_tree_roots;
3030
3031 free_root_pointers(fs_info, 0);
3032
3033 /* don't use the log in recovery mode, it won't be valid */
3034 btrfs_set_super_log_root(disk_super, 0);
3035
3036 /* we can't trust the free space cache either */
3037 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3038
3039 ret = next_root_backup(fs_info, fs_info->super_copy,
3040 &num_backups_tried, &backup_index);
3041 if (ret == -1)
3042 goto fail_block_groups;
3043 goto retry_root_backup;
eb60ceac
CM		 3044}
3045
f2984462
CM		 3046static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3047{
f2984462
CM		 3048 if (uptodate) {
3049 set_buffer_uptodate(bh);
3050 } else {
442a4f63
SB		 3051 struct btrfs_device *device = (struct btrfs_device *)
3052 bh->b_private;
3053
efe120a0 3054 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
606686ee
JB		 3055 "I/O error on %s\n",
3056 rcu_str_deref(device->name));
1259ab75
CM		 3057 /* note, we dont' set_buffer_write_io_error because we have
3058 * our own ways of dealing with the IO errors
3059 */
f2984462 3060 clear_buffer_uptodate(bh);
442a4f63 3061 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM		 3062 }
3063 unlock_buffer(bh);
3064 put_bh(bh);
3065}
3066
a512bbf8
YZ		 3067struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3068{
3069 struct buffer_head *bh;
3070 struct buffer_head *latest = NULL;
3071 struct btrfs_super_block *super;
3072 int i;
3073 u64 transid = 0;
3074 u64 bytenr;
3075
3076 /* we would like to check all the supers, but that would make
3077 * a btrfs mount succeed after a mkfs from a different FS.
3078 * So, we need to add a special mount option to scan for
3079 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3080 */
3081 for (i = 0; i < 1; i++) {
3082 bytenr = btrfs_sb_offset(i);
8068a47e
AJ		 3083 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3084 i_size_read(bdev->bd_inode))
a512bbf8 3085 break;
8068a47e
AJ		 3086 bh = __bread(bdev, bytenr / 4096,
3087 BTRFS_SUPER_INFO_SIZE);
a512bbf8
YZ		 3088 if (!bh)
3089 continue;
3090
3091 super = (struct btrfs_super_block *)bh->b_data;
3092 if (btrfs_super_bytenr(super) != bytenr ||
3cae210f 3093 btrfs_super_magic(super) != BTRFS_MAGIC) {
a512bbf8
YZ		 3094 brelse(bh);
3095 continue;
3096 }
3097
3098 if (!latest || btrfs_super_generation(super) > transid) {
3099 brelse(latest);
3100 latest = bh;
3101 transid = btrfs_super_generation(super);
3102 } else {
3103 brelse(bh);
3104 }
3105 }
3106 return latest;
3107}
3108
4eedeb75
HH		 3109/*
3110 * this should be called twice, once with wait == 0 and
3111 * once with wait == 1. When wait == 0 is done, all the buffer heads
3112 * we write are pinned.
3113 *
3114 * They are released when wait == 1 is done.
3115 * max_mirrors must be the same for both runs, and it indicates how
3116 * many supers on this one device should be written.
3117 *
3118 * max_mirrors == 0 means to write them all.
3119 */
a512bbf8
YZ		 3120static int write_dev_supers(struct btrfs_device *device,
3121 struct btrfs_super_block *sb,
3122 int do_barriers, int wait, int max_mirrors)
3123{
3124 struct buffer_head *bh;
3125 int i;
3126 int ret;
3127 int errors = 0;
3128 u32 crc;
3129 u64 bytenr;
a512bbf8
YZ		 3130
3131 if (max_mirrors == 0)
3132 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3133
a512bbf8
YZ		 3134 for (i = 0; i < max_mirrors; i++) {
3135 bytenr = btrfs_sb_offset(i);
935e5cc9
MX		 3136 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3137 device->commit_total_bytes)
a512bbf8
YZ		 3138 break;
3139
3140 if (wait) {
3141 bh = __find_get_block(device->bdev, bytenr / 4096,
3142 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3143 if (!bh) {
3144 errors++;
3145 continue;
3146 }
a512bbf8 3147 wait_on_buffer(bh);
4eedeb75
HH		 3148 if (!buffer_uptodate(bh))
3149 errors++;
3150
3151 /* drop our reference */
3152 brelse(bh);
3153
3154 /* drop the reference from the wait == 0 run */
3155 brelse(bh);
3156 continue;
a512bbf8
YZ		 3157 } else {
3158 btrfs_set_super_bytenr(sb, bytenr);
3159
3160 crc = ~(u32)0;
b0496686 3161 crc = btrfs_csum_data((char *)sb +
a512bbf8
YZ		 3162 BTRFS_CSUM_SIZE, crc,
3163 BTRFS_SUPER_INFO_SIZE -
3164 BTRFS_CSUM_SIZE);
3165 btrfs_csum_final(crc, sb->csum);
3166
4eedeb75
HH		 3167 /*
3168 * one reference for us, and we leave it for the
3169 * caller
3170 */
a512bbf8
YZ		 3171 bh = __getblk(device->bdev, bytenr / 4096,
3172 BTRFS_SUPER_INFO_SIZE);
634554dc 3173 if (!bh) {
efe120a0 3174 printk(KERN_ERR "BTRFS: couldn't get super "
634554dc
JB		 3175 "buffer head for bytenr %Lu\n", bytenr);
3176 errors++;
3177 continue;
3178 }
3179
a512bbf8
YZ		 3180 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3181
4eedeb75 3182 /* one reference for submit_bh */
a512bbf8 3183 get_bh(bh);
4eedeb75
HH		 3184
3185 set_buffer_uptodate(bh);
a512bbf8
YZ		 3186 lock_buffer(bh);
3187 bh->b_end_io = btrfs_end_buffer_write_sync;
442a4f63 3188 bh->b_private = device;
a512bbf8
YZ		 3189 }
3190
387125fc
CM		 3191 /*
3192 * we fua the first super. The others we allow
3193 * to go down lazy.
3194 */
e8117c26
WS		 3195 if (i == 0)
3196 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3197 else
3198 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
4eedeb75 3199 if (ret)
a512bbf8 3200 errors++;
a512bbf8
YZ		 3201 }
3202 return errors < i ? 0 : -1;
3203}
3204
387125fc
CM		 3205/*
3206 * endio for the write_dev_flush, this will wake anyone waiting
3207 * for the barrier when it is done
3208 */
3209static void btrfs_end_empty_barrier(struct bio *bio, int err)
3210{
3211 if (err) {
3212 if (err == -EOPNOTSUPP)
3213 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3214 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3215 }
3216 if (bio->bi_private)
3217 complete(bio->bi_private);
3218 bio_put(bio);
3219}
3220
3221/*
3222 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3223 * sent down. With wait == 1, it waits for the previous flush.
3224 *
3225 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3226 * capable
3227 */
3228static int write_dev_flush(struct btrfs_device *device, int wait)
3229{
3230 struct bio *bio;
3231 int ret = 0;
3232
3233 if (device->nobarriers)
3234 return 0;
3235
3236 if (wait) {
3237 bio = device->flush_bio;
3238 if (!bio)
3239 return 0;
3240
3241 wait_for_completion(&device->flush_wait);
3242
3243 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
efe120a0 3244 printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
606686ee 3245 rcu_str_deref(device->name));
387125fc 3246 device->nobarriers = 1;
5af3e8cc 3247 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
387125fc 3248 ret = -EIO;
5af3e8cc
SB		 3249 btrfs_dev_stat_inc_and_print(device,
3250 BTRFS_DEV_STAT_FLUSH_ERRS);
387125fc
CM		 3251 }
3252
3253 /* drop the reference from the wait == 0 run */
3254 bio_put(bio);
3255 device->flush_bio = NULL;
3256
3257 return ret;
3258 }
3259
3260 /*
3261 * one reference for us, and we leave it for the
3262 * caller
3263 */
9c017abc 3264 device->flush_bio = NULL;
9be3395b 3265 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
387125fc
CM		 3266 if (!bio)
3267 return -ENOMEM;
3268
3269 bio->bi_end_io = btrfs_end_empty_barrier;
3270 bio->bi_bdev = device->bdev;
3271 init_completion(&device->flush_wait);
3272 bio->bi_private = &device->flush_wait;
3273 device->flush_bio = bio;
3274
3275 bio_get(bio);
21adbd5c 3276 btrfsic_submit_bio(WRITE_FLUSH, bio);
387125fc
CM		 3277
3278 return 0;
3279}
3280
3281/*
3282 * send an empty flush down to each device in parallel,
3283 * then wait for them
3284 */
3285static int barrier_all_devices(struct btrfs_fs_info *info)
3286{
3287 struct list_head *head;
3288 struct btrfs_device *dev;
5af3e8cc
SB		 3289 int errors_send = 0;
3290 int errors_wait = 0;
387125fc
CM		 3291 int ret;
3292
3293 /* send down all the barriers */
3294 head = &info->fs_devices->devices;
3295 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3296 if (dev->missing)
3297 continue;
387125fc 3298 if (!dev->bdev) {
5af3e8cc 3299 errors_send++;
387125fc
CM		 3300 continue;
3301 }
3302 if (!dev->in_fs_metadata || !dev->writeable)
3303 continue;
3304
3305 ret = write_dev_flush(dev, 0);
3306 if (ret)
5af3e8cc 3307 errors_send++;
387125fc
CM		 3308 }
3309
3310 /* wait for all the barriers */
3311 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3312 if (dev->missing)
3313 continue;
387125fc 3314 if (!dev->bdev) {
5af3e8cc 3315 errors_wait++;
387125fc
CM		 3316 continue;
3317 }
3318 if (!dev->in_fs_metadata || !dev->writeable)
3319 continue;
3320
3321 ret = write_dev_flush(dev, 1);
3322 if (ret)
5af3e8cc 3323 errors_wait++;
387125fc 3324 }
5af3e8cc
SB		 3325 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3326 errors_wait > info->num_tolerated_disk_barrier_failures)
387125fc
CM		 3327 return -EIO;
3328 return 0;
3329}
3330
5af3e8cc
SB		 3331int btrfs_calc_num_tolerated_disk_barrier_failures(
3332 struct btrfs_fs_info *fs_info)
3333{
3334 struct btrfs_ioctl_space_info space;
3335 struct btrfs_space_info *sinfo;
3336 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3337 BTRFS_BLOCK_GROUP_SYSTEM,
3338 BTRFS_BLOCK_GROUP_METADATA,
3339 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3340 int num_types = 4;
3341 int i;
3342 int c;
3343 int num_tolerated_disk_barrier_failures =
3344 (int)fs_info->fs_devices->num_devices;
3345
3346 for (i = 0; i < num_types; i++) {
3347 struct btrfs_space_info *tmp;
3348
3349 sinfo = NULL;
3350 rcu_read_lock();
3351 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3352 if (tmp->flags == types[i]) {
3353 sinfo = tmp;
3354 break;
3355 }
3356 }
3357 rcu_read_unlock();
3358
3359 if (!sinfo)
3360 continue;
3361
3362 down_read(&sinfo->groups_sem);
3363 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3364 if (!list_empty(&sinfo->block_groups[c])) {
3365 u64 flags;
3366
3367 btrfs_get_block_group_info(
3368 &sinfo->block_groups[c], &space);
3369 if (space.total_bytes == 0 ||
3370 space.used_bytes == 0)
3371 continue;
3372 flags = space.flags;
3373 /*
3374 * return
3375 * 0: if dup, single or RAID0 is configured for
3376 * any of metadata, system or data, else
3377 * 1: if RAID5 is configured, or if RAID1 or
3378 * RAID10 is configured and only two mirrors
3379 * are used, else
3380 * 2: if RAID6 is configured, else
3381 * num_mirrors - 1: if RAID1 or RAID10 is
3382 * configured and more than
3383 * 2 mirrors are used.
3384 */
3385 if (num_tolerated_disk_barrier_failures > 0 &&
3386 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3387 BTRFS_BLOCK_GROUP_RAID0)) ||
3388 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3389 == 0)))
3390 num_tolerated_disk_barrier_failures = 0;
53b381b3
DW		 3391 else if (num_tolerated_disk_barrier_failures > 1) {
3392 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3393 BTRFS_BLOCK_GROUP_RAID5 |
3394 BTRFS_BLOCK_GROUP_RAID10)) {
3395 num_tolerated_disk_barrier_failures = 1;
3396 } else if (flags &
15b0a89d 3397 BTRFS_BLOCK_GROUP_RAID6) {
53b381b3
DW		 3398 num_tolerated_disk_barrier_failures = 2;
3399 }
3400 }
5af3e8cc
SB		 3401 }
3402 }
3403 up_read(&sinfo->groups_sem);
3404 }
3405
3406 return num_tolerated_disk_barrier_failures;
3407}
3408
48a3b636 3409static int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 3410{
e5e9a520 3411 struct list_head *head;
f2984462 3412 struct btrfs_device *dev;
a061fc8d 3413 struct btrfs_super_block *sb;
f2984462 3414 struct btrfs_dev_item *dev_item;
f2984462
CM		 3415 int ret;
3416 int do_barriers;
a236aed1
CM		 3417 int max_errors;
3418 int total_errors = 0;
a061fc8d 3419 u64 flags;
f2984462
CM		 3420
3421 do_barriers = !btrfs_test_opt(root, NOBARRIER);
af31f5e5 3422 backup_super_roots(root->fs_info);
f2984462 3423
6c41761f 3424 sb = root->fs_info->super_for_commit;
a061fc8d 3425 dev_item = &sb->dev_item;
e5e9a520 3426
174ba509 3427 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 3428 head = &root->fs_info->fs_devices->devices;
d7306801 3429 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
387125fc 3430
5af3e8cc
SB		 3431 if (do_barriers) {
3432 ret = barrier_all_devices(root->fs_info);
3433 if (ret) {
3434 mutex_unlock(
3435 &root->fs_info->fs_devices->device_list_mutex);
3436 btrfs_error(root->fs_info, ret,
3437 "errors while submitting device barriers.");
3438 return ret;
3439 }
3440 }
387125fc 3441
1f78160c 3442 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3443 if (!dev->bdev) {
3444 total_errors++;
3445 continue;
3446 }
2b82032c 3447 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3448 continue;
3449
2b82032c 3450 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 3451 btrfs_set_stack_device_type(dev_item, dev->type);
3452 btrfs_set_stack_device_id(dev_item, dev->devid);
7df69d3e 3453 btrfs_set_stack_device_total_bytes(dev_item,
935e5cc9 3454 dev->commit_total_bytes);
ce7213c7
MX		 3455 btrfs_set_stack_device_bytes_used(dev_item,
3456 dev->commit_bytes_used);
a061fc8d
CM		 3457 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3458 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3459 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3460 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 3461 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 3462
a061fc8d
CM		 3463 flags = btrfs_super_flags(sb);
3464 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3465
a512bbf8 3466 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 3467 if (ret)
3468 total_errors++;
f2984462 3469 }
a236aed1 3470 if (total_errors > max_errors) {
efe120a0 3471 btrfs_err(root->fs_info, "%d errors while writing supers",
d397712b 3472 total_errors);
a724b436 3473 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
79787eaa 3474
9d565ba4
SB		 3475 /* FUA is masked off if unsupported and can't be the reason */
3476 btrfs_error(root->fs_info, -EIO,
3477 "%d errors while writing supers", total_errors);
3478 return -EIO;
a236aed1 3479 }
f2984462 3480
a512bbf8 3481 total_errors = 0;
1f78160c 3482 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3483 if (!dev->bdev)
3484 continue;
2b82032c 3485 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3486 continue;
3487
a512bbf8
YZ		 3488 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3489 if (ret)
3490 total_errors++;
f2984462 3491 }
174ba509 3492 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 3493 if (total_errors > max_errors) {
79787eaa
JM		 3494 btrfs_error(root->fs_info, -EIO,
3495 "%d errors while writing supers", total_errors);
3496 return -EIO;
a236aed1 3497 }
f2984462
CM		 3498 return 0;
3499}
3500
a512bbf8
YZ		 3501int write_ctree_super(struct btrfs_trans_handle *trans,
3502 struct btrfs_root *root, int max_mirrors)
eb60ceac 3503{
f570e757 3504 return write_all_supers(root, max_mirrors);
cfaa7295
CM		 3505}
3506
cb517eab
MX		 3507/* Drop a fs root from the radix tree and free it. */
3508void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3509 struct btrfs_root *root)
2619ba1f 3510{
4df27c4d 3511 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 3512 radix_tree_delete(&fs_info->fs_roots_radix,
3513 (unsigned long)root->root_key.objectid);
4df27c4d 3514 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 3515
3516 if (btrfs_root_refs(&root->root_item) == 0)
3517 synchronize_srcu(&fs_info->subvol_srcu);
3518
1a4319cc 3519 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
3321719e 3520 btrfs_free_log(NULL, root);
3321719e 3521
faa2dbf0
JB		 3522 if (root->free_ino_pinned)
3523 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3524 if (root->free_ino_ctl)
3525 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3526 free_fs_root(root);
4df27c4d
YZ		 3527}
3528
3529static void free_fs_root(struct btrfs_root *root)
3530{
57cdc8db 3531 iput(root->ino_cache_inode);
4df27c4d 3532 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
1cb048f5
FDBM		 3533 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3534 root->orphan_block_rsv = NULL;
0ee5dc67
AV		 3535 if (root->anon_dev)
3536 free_anon_bdev(root->anon_dev);
8257b2dc
MX		 3537 if (root->subv_writers)
3538 btrfs_free_subvolume_writers(root->subv_writers);
4df27c4d
YZ		 3539 free_extent_buffer(root->node);
3540 free_extent_buffer(root->commit_root);
581bb050
LZ		 3541 kfree(root->free_ino_ctl);
3542 kfree(root->free_ino_pinned);
d397712b 3543 kfree(root->name);
b0feb9d9 3544 btrfs_put_fs_root(root);
2619ba1f
CM		 3545}
3546
cb517eab
MX		 3547void btrfs_free_fs_root(struct btrfs_root *root)
3548{
3549 free_fs_root(root);
2619ba1f
CM		 3550}
3551
c146afad 3552int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3553{
c146afad
YZ		 3554 u64 root_objectid = 0;
3555 struct btrfs_root *gang[8];
65d33fd7
QW		 3556 int i = 0;
3557 int err = 0;
3558 unsigned int ret = 0;
3559 int index;
e089f05c 3560
c146afad 3561 while (1) {
65d33fd7 3562 index = srcu_read_lock(&fs_info->subvol_srcu);
c146afad
YZ		 3563 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3564 (void **)gang, root_objectid,
3565 ARRAY_SIZE(gang));
65d33fd7
QW		 3566 if (!ret) {
3567 srcu_read_unlock(&fs_info->subvol_srcu, index);
c146afad 3568 break;
65d33fd7 3569 }
5d4f98a2 3570 root_objectid = gang[ret - 1]->root_key.objectid + 1;
65d33fd7 3571
c146afad 3572 for (i = 0; i < ret; i++) {
65d33fd7
QW		 3573 /* Avoid to grab roots in dead_roots */
3574 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3575 gang[i] = NULL;
3576 continue;
3577 }
3578 /* grab all the search result for later use */
3579 gang[i] = btrfs_grab_fs_root(gang[i]);
3580 }
3581 srcu_read_unlock(&fs_info->subvol_srcu, index);
66b4ffd1 3582
65d33fd7
QW		 3583 for (i = 0; i < ret; i++) {
3584 if (!gang[i])
3585 continue;
c146afad 3586 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 3587 err = btrfs_orphan_cleanup(gang[i]);
3588 if (err)
65d33fd7
QW		 3589 break;
3590 btrfs_put_fs_root(gang[i]);
c146afad
YZ		 3591 }
3592 root_objectid++;
3593 }
65d33fd7
QW		 3594
3595 /* release the uncleaned roots due to error */
3596 for (; i < ret; i++) {
3597 if (gang[i])
3598 btrfs_put_fs_root(gang[i]);
3599 }
3600 return err;
c146afad 3601}
a2135011 3602
c146afad
YZ		 3603int btrfs_commit_super(struct btrfs_root *root)
3604{
3605 struct btrfs_trans_handle *trans;
a74a4b97 3606
c146afad 3607 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 3608 btrfs_run_delayed_iputs(root);
c146afad 3609 mutex_unlock(&root->fs_info->cleaner_mutex);
9d1a2a3a 3610 wake_up_process(root->fs_info->cleaner_kthread);
c71bf099
YZ		 3611
3612 /* wait until ongoing cleanup work done */
3613 down_write(&root->fs_info->cleanup_work_sem);
3614 up_write(&root->fs_info->cleanup_work_sem);
3615
7a7eaa40 3616 trans = btrfs_join_transaction(root);
3612b495
TI		 3617 if (IS_ERR(trans))
3618 return PTR_ERR(trans);
d52c1bcc 3619 return btrfs_commit_transaction(trans, root);
c146afad
YZ		 3620}
3621
3abdbd78 3622void close_ctree(struct btrfs_root *root)
c146afad
YZ		 3623{
3624 struct btrfs_fs_info *fs_info = root->fs_info;
3625 int ret;
3626
3627 fs_info->closing = 1;
3628 smp_mb();
3629
803b2f54
SB		 3630 /* wait for the uuid_scan task to finish */
3631 down(&fs_info->uuid_tree_rescan_sem);
3632 /* avoid complains from lockdep et al., set sem back to initial state */
3633 up(&fs_info->uuid_tree_rescan_sem);
3634
837d5b6e 3635 /* pause restriper - we want to resume on mount */
aa1b8cd4 3636 btrfs_pause_balance(fs_info);
837d5b6e 3637
8dabb742
SB		 3638 btrfs_dev_replace_suspend_for_unmount(fs_info);
3639
aa1b8cd4 3640 btrfs_scrub_cancel(fs_info);
4cb5300b
CM		 3641
3642 /* wait for any defraggers to finish */
3643 wait_event(fs_info->transaction_wait,
3644 (atomic_read(&fs_info->defrag_running) == 0));
3645
3646 /* clear out the rbtree of defraggable inodes */
26176e7c 3647 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3648
21c7e756
MX		 3649 cancel_work_sync(&fs_info->async_reclaim_work);
3650
c146afad 3651 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 3652 ret = btrfs_commit_super(root);
3653 if (ret)
efe120a0 3654 btrfs_err(root->fs_info, "commit super ret %d", ret);
acce952b 3655 }
3656
87533c47 3657 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
68ce9682 3658 btrfs_error_commit_super(root);
0f7d52f4 3659
e3029d9f
AV		 3660 kthread_stop(fs_info->transaction_kthread);
3661 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3662
f25784b3
YZ		 3663 fs_info->closing = 2;
3664 smp_mb();
3665
bcef60f2
AJ		 3666 btrfs_free_qgroup_config(root->fs_info);
3667
963d678b 3668 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
efe120a0 3669 btrfs_info(root->fs_info, "at unmount delalloc count %lld",
963d678b 3670 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3671 }
bcc63abb 3672
5ac1d209
JM		 3673 btrfs_sysfs_remove_one(fs_info);
3674
faa2dbf0 3675 btrfs_free_fs_roots(fs_info);
d10c5f31 3676
1a4319cc
LB		 3677 btrfs_put_block_group_cache(fs_info);
3678
2b1360da
JB		 3679 btrfs_free_block_groups(fs_info);
3680
de348ee0
WS		 3681 /*
3682 * we must make sure there is not any read request to
3683 * submit after we stopping all workers.
3684 */
3685 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
96192499
JB		 3686 btrfs_stop_all_workers(fs_info);
3687
47ab2a6c 3688 fs_info->open = 0;
13e6c37b 3689 free_root_pointers(fs_info, 1);
9ad6b7bc 3690
13e6c37b 3691 iput(fs_info->btree_inode);
d6bfde87 3692
21adbd5c
SB		 3693#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3694 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3695 btrfsic_unmount(root, fs_info->fs_devices);
3696#endif
3697
dfe25020 3698 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3699 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3700
e2d84521 3701 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 3702 percpu_counter_destroy(&fs_info->delalloc_bytes);
c404e0dc 3703 percpu_counter_destroy(&fs_info->bio_counter);
04160088 3704 bdi_destroy(&fs_info->bdi);
76dda93c 3705 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 3706
53b381b3
DW		 3707 btrfs_free_stripe_hash_table(fs_info);
3708
1cb048f5
FDBM		 3709 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3710 root->orphan_block_rsv = NULL;
eb60ceac
CM		 3711}
3712
b9fab919
CM		 3713int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3714 int atomic)
5f39d397 3715{
1259ab75 3716 int ret;
727011e0 3717 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 3718
0b32f4bb 3719 ret = extent_buffer_uptodate(buf);
1259ab75
CM		 3720 if (!ret)
3721 return ret;
3722
3723 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM		 3724 parent_transid, atomic);
3725 if (ret == -EAGAIN)
3726 return ret;
1259ab75 3727 return !ret;
5f39d397
CM		 3728}
3729
3730int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 3731{
0b32f4bb 3732 return set_extent_buffer_uptodate(buf);
5f39d397 3733}
6702ed49 3734
5f39d397
CM		 3735void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3736{
06ea65a3 3737 struct btrfs_root *root;
5f39d397 3738 u64 transid = btrfs_header_generation(buf);
b9473439 3739 int was_dirty;
b4ce94de 3740
06ea65a3
JB		 3741#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3742 /*
3743 * This is a fast path so only do this check if we have sanity tests
3744 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3745 * outside of the sanity tests.
3746 */
3747 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3748 return;
3749#endif
3750 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
b9447ef8 3751 btrfs_assert_tree_locked(buf);
31b1a2bd
JL		 3752 if (transid != root->fs_info->generation)
3753 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
d397712b 3754 "found %llu running %llu\n",
c1c9ff7c 3755 buf->start, transid, root->fs_info->generation);
0b32f4bb 3756 was_dirty = set_extent_buffer_dirty(buf);
e2d84521
MX		 3757 if (!was_dirty)
3758 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3759 buf->len,
3760 root->fs_info->dirty_metadata_batch);
1f21ef0a
FM		 3761#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3762 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
3763 btrfs_print_leaf(root, buf);
3764 ASSERT(0);
3765 }
3766#endif
eb60ceac
CM		 3767}
3768
b53d3f5d
LB		 3769static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3770 int flush_delayed)
16cdcec7
MX		 3771{
3772 /*
3773 * looks as though older kernels can get into trouble with
3774 * this code, they end up stuck in balance_dirty_pages forever
3775 */
e2d84521 3776 int ret;
16cdcec7
MX		 3777
3778 if (current->flags & PF_MEMALLOC)
3779 return;
3780
b53d3f5d
LB		 3781 if (flush_delayed)
3782 btrfs_balance_delayed_items(root);
16cdcec7 3783
e2d84521
MX		 3784 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3785 BTRFS_DIRTY_METADATA_THRESH);
3786 if (ret > 0) {
d0e1d66b
NJ		 3787 balance_dirty_pages_ratelimited(
3788 root->fs_info->btree_inode->i_mapping);
16cdcec7
MX		 3789 }
3790 return;
3791}
3792
b53d3f5d 3793void btrfs_btree_balance_dirty(struct btrfs_root *root)
35b7e476 3794{
b53d3f5d
LB		 3795 __btrfs_btree_balance_dirty(root, 1);
3796}
585ad2c3 3797
b53d3f5d
LB		 3798void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3799{
3800 __btrfs_btree_balance_dirty(root, 0);
35b7e476 3801}
6b80053d 3802
ca7a79ad 3803int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 3804{
727011e0 3805 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
0b32f4bb 3806 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
6b80053d 3807}
0da5468f 3808
fcd1f065 3809static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 3810 int read_only)
3811{
1104a885
DS		 3812 /*
3813 * Placeholder for checks
3814 */
fcd1f065 3815 return 0;
acce952b 3816}
3817
48a3b636 3818static void btrfs_error_commit_super(struct btrfs_root *root)
acce952b 3819{
acce952b 3820 mutex_lock(&root->fs_info->cleaner_mutex);
3821 btrfs_run_delayed_iputs(root);
3822 mutex_unlock(&root->fs_info->cleaner_mutex);
3823
3824 down_write(&root->fs_info->cleanup_work_sem);
3825 up_write(&root->fs_info->cleanup_work_sem);
3826
3827 /* cleanup FS via transaction */
3828 btrfs_cleanup_transaction(root);
acce952b 3829}
3830
143bede5 3831static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 3832{
acce952b 3833 struct btrfs_ordered_extent *ordered;
acce952b 3834
199c2a9c 3835 spin_lock(&root->ordered_extent_lock);
779880ef
JB		 3836 /*
3837 * This will just short circuit the ordered completion stuff which will
3838 * make sure the ordered extent gets properly cleaned up.
3839 */
199c2a9c 3840 list_for_each_entry(ordered, &root->ordered_extents,
779880ef
JB		 3841 root_extent_list)
3842 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
199c2a9c
MX		 3843 spin_unlock(&root->ordered_extent_lock);
3844}
3845
3846static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3847{
3848 struct btrfs_root *root;
3849 struct list_head splice;
3850
3851 INIT_LIST_HEAD(&splice);
3852
3853 spin_lock(&fs_info->ordered_root_lock);
3854 list_splice_init(&fs_info->ordered_roots, &splice);
3855 while (!list_empty(&splice)) {
3856 root = list_first_entry(&splice, struct btrfs_root,
3857 ordered_root);
1de2cfde
JB		 3858 list_move_tail(&root->ordered_root,
3859 &fs_info->ordered_roots);
199c2a9c 3860
2a85d9ca 3861 spin_unlock(&fs_info->ordered_root_lock);
199c2a9c
MX		 3862 btrfs_destroy_ordered_extents(root);
3863
2a85d9ca
LB		 3864 cond_resched();
3865 spin_lock(&fs_info->ordered_root_lock);
199c2a9c
MX		 3866 }
3867 spin_unlock(&fs_info->ordered_root_lock);
acce952b 3868}
3869
35a3621b
SB		 3870static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3871 struct btrfs_root *root)
acce952b 3872{
3873 struct rb_node *node;
3874 struct btrfs_delayed_ref_root *delayed_refs;
3875 struct btrfs_delayed_ref_node *ref;
3876 int ret = 0;
3877
3878 delayed_refs = &trans->delayed_refs;
3879
3880 spin_lock(&delayed_refs->lock);
d7df2c79 3881 if (atomic_read(&delayed_refs->num_entries) == 0) {
cfece4db 3882 spin_unlock(&delayed_refs->lock);
efe120a0 3883 btrfs_info(root->fs_info, "delayed_refs has NO entry");
acce952b 3884 return ret;
3885 }
3886
d7df2c79
JB		 3887 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
3888 struct btrfs_delayed_ref_head *head;
e78417d1 3889 bool pin_bytes = false;
acce952b 3890
d7df2c79
JB		 3891 head = rb_entry(node, struct btrfs_delayed_ref_head,
3892 href_node);
3893 if (!mutex_trylock(&head->mutex)) {
3894 atomic_inc(&head->node.refs);
3895 spin_unlock(&delayed_refs->lock);
eb12db69 3896
d7df2c79 3897 mutex_lock(&head->mutex);
e78417d1 3898 mutex_unlock(&head->mutex);
d7df2c79
JB		 3899 btrfs_put_delayed_ref(&head->node);
3900 spin_lock(&delayed_refs->lock);
3901 continue;
3902 }
3903 spin_lock(&head->lock);
3904 while ((node = rb_first(&head->ref_root)) != NULL) {
3905 ref = rb_entry(node, struct btrfs_delayed_ref_node,
3906 rb_node);
3907 ref->in_tree = 0;
3908 rb_erase(&ref->rb_node, &head->ref_root);
3909 atomic_dec(&delayed_refs->num_entries);
3910 btrfs_put_delayed_ref(ref);
e78417d1 3911 }
d7df2c79
JB		 3912 if (head->must_insert_reserved)
3913 pin_bytes = true;
3914 btrfs_free_delayed_extent_op(head->extent_op);
3915 delayed_refs->num_heads--;
3916 if (head->processing == 0)
3917 delayed_refs->num_heads_ready--;
3918 atomic_dec(&delayed_refs->num_entries);
3919 head->node.in_tree = 0;
3920 rb_erase(&head->href_node, &delayed_refs->href_root);
3921 spin_unlock(&head->lock);
3922 spin_unlock(&delayed_refs->lock);
3923 mutex_unlock(&head->mutex);
acce952b 3924
d7df2c79
JB		 3925 if (pin_bytes)
3926 btrfs_pin_extent(root, head->node.bytenr,
3927 head->node.num_bytes, 1);
3928 btrfs_put_delayed_ref(&head->node);
acce952b 3929 cond_resched();
3930 spin_lock(&delayed_refs->lock);
3931 }
3932
3933 spin_unlock(&delayed_refs->lock);
3934
3935 return ret;
3936}
3937
143bede5 3938static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 3939{
3940 struct btrfs_inode *btrfs_inode;
3941 struct list_head splice;
3942
3943 INIT_LIST_HEAD(&splice);
3944
eb73c1b7
MX		 3945 spin_lock(&root->delalloc_lock);
3946 list_splice_init(&root->delalloc_inodes, &splice);
acce952b 3947
3948 while (!list_empty(&splice)) {
eb73c1b7
MX		 3949 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3950 delalloc_inodes);
acce952b 3951
3952 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX		 3953 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3954 &btrfs_inode->runtime_flags);
eb73c1b7 3955 spin_unlock(&root->delalloc_lock);
acce952b 3956
3957 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3958
eb73c1b7 3959 spin_lock(&root->delalloc_lock);
acce952b 3960 }
3961
eb73c1b7
MX		 3962 spin_unlock(&root->delalloc_lock);
3963}
3964
3965static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3966{
3967 struct btrfs_root *root;
3968 struct list_head splice;
3969
3970 INIT_LIST_HEAD(&splice);
3971
3972 spin_lock(&fs_info->delalloc_root_lock);
3973 list_splice_init(&fs_info->delalloc_roots, &splice);
3974 while (!list_empty(&splice)) {
3975 root = list_first_entry(&splice, struct btrfs_root,
3976 delalloc_root);
3977 list_del_init(&root->delalloc_root);
3978 root = btrfs_grab_fs_root(root);
3979 BUG_ON(!root);
3980 spin_unlock(&fs_info->delalloc_root_lock);
3981
3982 btrfs_destroy_delalloc_inodes(root);
3983 btrfs_put_fs_root(root);
3984
3985 spin_lock(&fs_info->delalloc_root_lock);
3986 }
3987 spin_unlock(&fs_info->delalloc_root_lock);
acce952b 3988}
3989
3990static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3991 struct extent_io_tree *dirty_pages,
3992 int mark)
3993{
3994 int ret;
acce952b 3995 struct extent_buffer *eb;
3996 u64 start = 0;
3997 u64 end;
acce952b 3998
3999 while (1) {
4000 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 4001 mark, NULL);
acce952b 4002 if (ret)
4003 break;
4004
4005 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
4006 while (start <= end) {
fd8b2b61 4007 eb = btrfs_find_tree_block(root, start,
707e8a07
DS		 4008 root->nodesize);
4009 start += root->nodesize;
fd8b2b61 4010 if (!eb)
acce952b 4011 continue;
fd8b2b61 4012 wait_on_extent_buffer_writeback(eb);
acce952b 4013
fd8b2b61
JB		 4014 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4015 &eb->bflags))
4016 clear_extent_buffer_dirty(eb);
4017 free_extent_buffer_stale(eb);
acce952b 4018 }
4019 }
4020
4021 return ret;
4022}
4023
4024static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4025 struct extent_io_tree *pinned_extents)
4026{
4027 struct extent_io_tree *unpin;
4028 u64 start;
4029 u64 end;
4030 int ret;
ed0eaa14 4031 bool loop = true;
acce952b 4032
4033 unpin = pinned_extents;
ed0eaa14 4034again:
acce952b 4035 while (1) {
4036 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 4037 EXTENT_DIRTY, NULL);
acce952b 4038 if (ret)
4039 break;
4040
4041 /* opt_discard */
5378e607
LD		 4042 if (btrfs_test_opt(root, DISCARD))
4043 ret = btrfs_error_discard_extent(root, start,
4044 end + 1 - start,
4045 NULL);
acce952b 4046
4047 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4048 btrfs_error_unpin_extent_range(root, start, end);
4049 cond_resched();
4050 }
4051
ed0eaa14
LB		 4052 if (loop) {
4053 if (unpin == &root->fs_info->freed_extents[0])
4054 unpin = &root->fs_info->freed_extents[1];
4055 else
4056 unpin = &root->fs_info->freed_extents[0];
4057 loop = false;
4058 goto again;
4059 }
4060
acce952b 4061 return 0;
4062}
4063
49b25e05
JM		 4064void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4065 struct btrfs_root *root)
4066{
4067 btrfs_destroy_delayed_refs(cur_trans, root);
49b25e05 4068
4a9d8bde 4069 cur_trans->state = TRANS_STATE_COMMIT_START;
d7096fc3 4070 wake_up(&root->fs_info->transaction_blocked_wait);
49b25e05 4071
4a9d8bde 4072 cur_trans->state = TRANS_STATE_UNBLOCKED;
d7096fc3 4073 wake_up(&root->fs_info->transaction_wait);
49b25e05 4074
67cde344
MX		 4075 btrfs_destroy_delayed_inodes(root);
4076 btrfs_assert_delayed_root_empty(root);
49b25e05 4077
49b25e05
JM		 4078 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4079 EXTENT_DIRTY);
6e841e32
LB		 4080 btrfs_destroy_pinned_extent(root,
4081 root->fs_info->pinned_extents);
49b25e05 4082
4a9d8bde
MX		 4083 cur_trans->state =TRANS_STATE_COMPLETED;
4084 wake_up(&cur_trans->commit_wait);
4085
49b25e05
JM		 4086 /*
4087 memset(cur_trans, 0, sizeof(*cur_trans));
4088 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4089 */
4090}
4091
48a3b636 4092static int btrfs_cleanup_transaction(struct btrfs_root *root)
acce952b 4093{
4094 struct btrfs_transaction *t;
acce952b 4095
acce952b 4096 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4097
a4abeea4 4098 spin_lock(&root->fs_info->trans_lock);
724e2315
JB		 4099 while (!list_empty(&root->fs_info->trans_list)) {
4100 t = list_first_entry(&root->fs_info->trans_list,
4101 struct btrfs_transaction, list);
4102 if (t->state >= TRANS_STATE_COMMIT_START) {
4103 atomic_inc(&t->use_count);
4104 spin_unlock(&root->fs_info->trans_lock);
4105 btrfs_wait_for_commit(root, t->transid);
4106 btrfs_put_transaction(t);
4107 spin_lock(&root->fs_info->trans_lock);
4108 continue;
4109 }
4110 if (t == root->fs_info->running_transaction) {
4111 t->state = TRANS_STATE_COMMIT_DOING;
4112 spin_unlock(&root->fs_info->trans_lock);
4113 /*
4114 * We wait for 0 num_writers since we don't hold a trans
4115 * handle open currently for this transaction.
4116 */
4117 wait_event(t->writer_wait,
4118 atomic_read(&t->num_writers) == 0);
4119 } else {
4120 spin_unlock(&root->fs_info->trans_lock);
4121 }
4122 btrfs_cleanup_one_transaction(t, root);
4a9d8bde 4123
724e2315
JB		 4124 spin_lock(&root->fs_info->trans_lock);
4125 if (t == root->fs_info->running_transaction)
4126 root->fs_info->running_transaction = NULL;
acce952b 4127 list_del_init(&t->list);
724e2315 4128 spin_unlock(&root->fs_info->trans_lock);
acce952b 4129
724e2315
JB		 4130 btrfs_put_transaction(t);
4131 trace_btrfs_transaction_commit(root);
4132 spin_lock(&root->fs_info->trans_lock);
4133 }
4134 spin_unlock(&root->fs_info->trans_lock);
4135 btrfs_destroy_all_ordered_extents(root->fs_info);
4136 btrfs_destroy_delayed_inodes(root);
4137 btrfs_assert_delayed_root_empty(root);
4138 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4139 btrfs_destroy_all_delalloc_inodes(root->fs_info);
acce952b 4140 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 4141
4142 return 0;
4143}
4144
d1310b2e 4145static struct extent_io_ops btree_extent_io_ops = {
ce9adaa5 4146 .readpage_end_io_hook = btree_readpage_end_io_hook,
4bb31e92 4147 .readpage_io_failed_hook = btree_io_failed_hook,
0b86a832 4148 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 4149 /* note we're sharing with inode.c for the merge bio hook */
4150 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 4151};
Linux 6.x block layer and io_uring tree(s)