projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: add a flags field to btrfs_fs_info
[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>
5a0e3ad6 28#include <linux/slab.h>
784b4e29 29#include <linux/migrate.h>
7a36ddec 30#include <linux/ratelimit.h>
6463fe58 31#include <linux/uuid.h>
803b2f54 32#include <linux/semaphore.h>
7e75bf3f 33#include <asm/unaligned.h>
eb60ceac
CM		 34#include "ctree.h"
35#include "disk-io.h"
0b947aff 36#include "hash.h"
e089f05c 37#include "transaction.h"
0f7d52f4 38#include "btrfs_inode.h"
0b86a832 39#include "volumes.h"
db94535d 40#include "print-tree.h"
925baedd 41#include "locking.h"
e02119d5 42#include "tree-log.h"
fa9c0d79 43#include "free-space-cache.h"
70f6d82e 44#include "free-space-tree.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"
ebb8765b 52#include "compression.h"
eb60ceac 53
de0022b9
JB		 54#ifdef CONFIG_X86
55#include <asm/cpufeature.h>
56#endif
57
319e4d06
QW		 58#define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\
59 BTRFS_HEADER_FLAG_RELOC |\
60 BTRFS_SUPER_FLAG_ERROR |\
61 BTRFS_SUPER_FLAG_SEEDING |\
62 BTRFS_SUPER_FLAG_METADUMP)
63
e8c9f186 64static const struct extent_io_ops btree_extent_io_ops;
8b712842 65static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 66static void free_fs_root(struct btrfs_root *root);
fcd1f065 67static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 68 int read_only);
143bede5 69static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
acce952b 70static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
71 struct btrfs_root *root);
143bede5 72static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
acce952b 73static int btrfs_destroy_marked_extents(struct btrfs_root *root,
74 struct extent_io_tree *dirty_pages,
75 int mark);
76static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
77 struct extent_io_tree *pinned_extents);
48a3b636
ES		 78static int btrfs_cleanup_transaction(struct btrfs_root *root);
79static void btrfs_error_commit_super(struct btrfs_root *root);
ce9adaa5 80
d352ac68 81/*
97eb6b69
DS		 82 * btrfs_end_io_wq structs are used to do processing in task context when an IO
83 * is complete. This is used during reads to verify checksums, and it is used
d352ac68
CM		 84 * by writes to insert metadata for new file extents after IO is complete.
85 */
97eb6b69 86struct btrfs_end_io_wq {
ce9adaa5
CM		 87 struct bio *bio;
88 bio_end_io_t *end_io;
89 void *private;
90 struct btrfs_fs_info *info;
91 int error;
bfebd8b5 92 enum btrfs_wq_endio_type metadata;
ce9adaa5 93 struct list_head list;
8b712842 94 struct btrfs_work work;
ce9adaa5 95};
0da5468f 96
97eb6b69
DS		 97static struct kmem_cache *btrfs_end_io_wq_cache;
98
99int __init btrfs_end_io_wq_init(void)
100{
101 btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq",
102 sizeof(struct btrfs_end_io_wq),
103 0,
fba4b697 104 SLAB_MEM_SPREAD,
97eb6b69
DS		 105 NULL);
106 if (!btrfs_end_io_wq_cache)
107 return -ENOMEM;
108 return 0;
109}
110
111void btrfs_end_io_wq_exit(void)
112{
5598e900 113 kmem_cache_destroy(btrfs_end_io_wq_cache);
97eb6b69
DS		 114}
115
d352ac68
CM		 116/*
117 * async submit bios are used to offload expensive checksumming
118 * onto the worker threads. They checksum file and metadata bios
119 * just before they are sent down the IO stack.
120 */
44b8bd7e
CM		 121struct async_submit_bio {
122 struct inode *inode;
123 struct bio *bio;
124 struct list_head list;
4a69a410
CM		 125 extent_submit_bio_hook_t *submit_bio_start;
126 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e 127 int mirror_num;
c8b97818 128 unsigned long bio_flags;
eaf25d93
CM		 129 /*
130 * bio_offset is optional, can be used if the pages in the bio
131 * can't tell us where in the file the bio should go
132 */
133 u64 bio_offset;
8b712842 134 struct btrfs_work work;
79787eaa 135 int error;
44b8bd7e
CM		 136};
137
85d4e461
CM		 138/*
139 * Lockdep class keys for extent_buffer->lock's in this root. For a given
140 * eb, the lockdep key is determined by the btrfs_root it belongs to and
141 * the level the eb occupies in the tree.
142 *
143 * Different roots are used for different purposes and may nest inside each
144 * other and they require separate keysets. As lockdep keys should be
145 * static, assign keysets according to the purpose of the root as indicated
146 * by btrfs_root->objectid. This ensures that all special purpose roots
147 * have separate keysets.
4008c04a 148 *
85d4e461
CM		 149 * Lock-nesting across peer nodes is always done with the immediate parent
150 * node locked thus preventing deadlock. As lockdep doesn't know this, use
151 * subclass to avoid triggering lockdep warning in such cases.
4008c04a 152 *
85d4e461
CM		 153 * The key is set by the readpage_end_io_hook after the buffer has passed
154 * csum validation but before the pages are unlocked. It is also set by
155 * btrfs_init_new_buffer on freshly allocated blocks.
4008c04a 156 *
85d4e461
CM		 157 * We also add a check to make sure the highest level of the tree is the
158 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
159 * needs update as well.
4008c04a
CM		 160 */
161#ifdef CONFIG_DEBUG_LOCK_ALLOC
162# if BTRFS_MAX_LEVEL != 8
163# error
164# endif
85d4e461
CM		 165
166static struct btrfs_lockdep_keyset {
167 u64 id; /* root objectid */
168 const char *name_stem; /* lock name stem */
169 char names[BTRFS_MAX_LEVEL + 1][20];
170 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
171} btrfs_lockdep_keysets[] = {
172 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
173 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
174 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
175 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
176 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
177 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
60b62978 178 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
85d4e461
CM		 179 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
180 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
181 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
13fd8da9 182 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
6b20e0ad 183 { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" },
85d4e461 184 { .id = 0, .name_stem = "tree" },
4008c04a 185};
85d4e461
CM		 186
187void __init btrfs_init_lockdep(void)
188{
189 int i, j;
190
191 /* initialize lockdep class names */
192 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
193 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
194
195 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
196 snprintf(ks->names[j], sizeof(ks->names[j]),
197 "btrfs-%s-%02d", ks->name_stem, j);
198 }
199}
200
201void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
202 int level)
203{
204 struct btrfs_lockdep_keyset *ks;
205
206 BUG_ON(level >= ARRAY_SIZE(ks->keys));
207
208 /* find the matching keyset, id 0 is the default entry */
209 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
210 if (ks->id == objectid)
211 break;
212
213 lockdep_set_class_and_name(&eb->lock,
214 &ks->keys[level], ks->names[level]);
215}
216
4008c04a
CM		 217#endif
218
d352ac68
CM		 219/*
220 * extents on the btree inode are pretty simple, there's one extent
221 * that covers the entire device
222 */
b2950863 223static struct extent_map *btree_get_extent(struct inode *inode,
306e16ce 224 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 225 int create)
7eccb903 226{
5f39d397
CM		 227 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
228 struct extent_map *em;
229 int ret;
230
890871be 231 read_lock(&em_tree->lock);
d1310b2e 232 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 233 if (em) {
234 em->bdev =
235 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 236 read_unlock(&em_tree->lock);
5f39d397 237 goto out;
a061fc8d 238 }
890871be 239 read_unlock(&em_tree->lock);
7b13b7b1 240
172ddd60 241 em = alloc_extent_map();
5f39d397
CM		 242 if (!em) {
243 em = ERR_PTR(-ENOMEM);
244 goto out;
245 }
246 em->start = 0;
0afbaf8c 247 em->len = (u64)-1;
c8b97818 248 em->block_len = (u64)-1;
5f39d397 249 em->block_start = 0;
a061fc8d 250 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 251
890871be 252 write_lock(&em_tree->lock);
09a2a8f9 253 ret = add_extent_mapping(em_tree, em, 0);
5f39d397
CM		 254 if (ret == -EEXIST) {
255 free_extent_map(em);
7b13b7b1 256 em = lookup_extent_mapping(em_tree, start, len);
b4f359ab 257 if (!em)
0433f20d 258 em = ERR_PTR(-EIO);
5f39d397 259 } else if (ret) {
7b13b7b1 260 free_extent_map(em);
0433f20d 261 em = ERR_PTR(ret);
5f39d397 262 }
890871be 263 write_unlock(&em_tree->lock);
7b13b7b1 264
5f39d397
CM		 265out:
266 return em;
7eccb903
CM		 267}
268
b0496686 269u32 btrfs_csum_data(char *data, u32 seed, size_t len)
19c00ddc 270{
0b947aff 271 return btrfs_crc32c(seed, data, len);
19c00ddc
CM		 272}
273
274void btrfs_csum_final(u32 crc, char *result)
275{
7e75bf3f 276 put_unaligned_le32(~crc, result);
19c00ddc
CM		 277}
278
d352ac68
CM		 279/*
280 * compute the csum for a btree block, and either verify it or write it
281 * into the csum field of the block.
282 */
01d58472
DD		 283static int csum_tree_block(struct btrfs_fs_info *fs_info,
284 struct extent_buffer *buf,
19c00ddc
CM		 285 int verify)
286{
01d58472 287 u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
607d432d 288 char *result = NULL;
19c00ddc
CM		 289 unsigned long len;
290 unsigned long cur_len;
291 unsigned long offset = BTRFS_CSUM_SIZE;
19c00ddc
CM		 292 char *kaddr;
293 unsigned long map_start;
294 unsigned long map_len;
295 int err;
296 u32 crc = ~(u32)0;
607d432d 297 unsigned long inline_result;
19c00ddc
CM		 298
299 len = buf->len - offset;
d397712b 300 while (len > 0) {
19c00ddc 301 err = map_private_extent_buffer(buf, offset, 32,
a6591715 302 &kaddr, &map_start, &map_len);
d397712b 303 if (err)
8bd98f0e 304 return err;
19c00ddc 305 cur_len = min(len, map_len - (offset - map_start));
b0496686 306 crc = btrfs_csum_data(kaddr + offset - map_start,
19c00ddc
CM		 307 crc, cur_len);
308 len -= cur_len;
309 offset += cur_len;
19c00ddc 310 }
607d432d 311 if (csum_size > sizeof(inline_result)) {
31e818fe 312 result = kzalloc(csum_size, GFP_NOFS);
607d432d 313 if (!result)
8bd98f0e 314 return -ENOMEM;
607d432d
JB		 315 } else {
316 result = (char *)&inline_result;
317 }
318
19c00ddc
CM		 319 btrfs_csum_final(crc, result);
320
321 if (verify) {
607d432d 322 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 323 u32 val;
324 u32 found = 0;
607d432d 325 memcpy(&found, result, csum_size);
e4204ded 326
607d432d 327 read_extent_buffer(buf, &val, 0, csum_size);
94647322
DS		 328 btrfs_warn_rl(fs_info,
329 "%s checksum verify failed on %llu wanted %X found %X "
330 "level %d",
01d58472 331 fs_info->sb->s_id, buf->start,
efe120a0 332 val, found, btrfs_header_level(buf));
607d432d
JB		 333 if (result != (char *)&inline_result)
334 kfree(result);
8bd98f0e 335 return -EUCLEAN;
19c00ddc
CM		 336 }
337 } else {
607d432d 338 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 339 }
607d432d
JB		 340 if (result != (char *)&inline_result)
341 kfree(result);
19c00ddc
CM		 342 return 0;
343}
344
d352ac68
CM		 345/*
346 * we can't consider a given block up to date unless the transid of the
347 * block matches the transid in the parent node's pointer. This is how we
348 * detect blocks that either didn't get written at all or got written
349 * in the wrong place.
350 */
1259ab75 351static int verify_parent_transid(struct extent_io_tree *io_tree,
b9fab919
CM		 352 struct extent_buffer *eb, u64 parent_transid,
353 int atomic)
1259ab75 354{
2ac55d41 355 struct extent_state *cached_state = NULL;
1259ab75 356 int ret;
2755a0de 357 bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB);
1259ab75
CM		 358
359 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
360 return 0;
361
b9fab919
CM		 362 if (atomic)
363 return -EAGAIN;
364
a26e8c9f
JB		 365 if (need_lock) {
366 btrfs_tree_read_lock(eb);
367 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
368 }
369
2ac55d41 370 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
ff13db41 371 &cached_state);
0b32f4bb 372 if (extent_buffer_uptodate(eb) &&
1259ab75
CM		 373 btrfs_header_generation(eb) == parent_transid) {
374 ret = 0;
375 goto out;
376 }
94647322
DS		 377 btrfs_err_rl(eb->fs_info,
378 "parent transid verify failed on %llu wanted %llu found %llu",
379 eb->start,
29549aec 380 parent_transid, btrfs_header_generation(eb));
1259ab75 381 ret = 1;
a26e8c9f
JB		 382
383 /*
384 * Things reading via commit roots that don't have normal protection,
385 * like send, can have a really old block in cache that may point at a
01327610 386 * block that has been freed and re-allocated. So don't clear uptodate
a26e8c9f
JB		 387 * if we find an eb that is under IO (dirty/writeback) because we could
388 * end up reading in the stale data and then writing it back out and
389 * making everybody very sad.
390 */
391 if (!extent_buffer_under_io(eb))
392 clear_extent_buffer_uptodate(eb);
33958dc6 393out:
2ac55d41
JB		 394 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
395 &cached_state, GFP_NOFS);
472b909f
JB		 396 if (need_lock)
397 btrfs_tree_read_unlock_blocking(eb);
1259ab75 398 return ret;
1259ab75
CM		 399}
400
1104a885
DS		 401/*
402 * Return 0 if the superblock checksum type matches the checksum value of that
403 * algorithm. Pass the raw disk superblock data.
404 */
405static int btrfs_check_super_csum(char *raw_disk_sb)
406{
407 struct btrfs_super_block *disk_sb =
408 (struct btrfs_super_block *)raw_disk_sb;
409 u16 csum_type = btrfs_super_csum_type(disk_sb);
410 int ret = 0;
411
412 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
413 u32 crc = ~(u32)0;
414 const int csum_size = sizeof(crc);
415 char result[csum_size];
416
417 /*
418 * The super_block structure does not span the whole
419 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
01327610 420 * is filled with zeros and is included in the checksum.
1104a885
DS		 421 */
422 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
423 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
424 btrfs_csum_final(crc, result);
425
426 if (memcmp(raw_disk_sb, result, csum_size))
427 ret = 1;
428 }
429
430 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
efe120a0 431 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
1104a885
DS		 432 csum_type);
433 ret = 1;
434 }
435
436 return ret;
437}
438
d352ac68
CM		 439/*
440 * helper to read a given tree block, doing retries as required when
441 * the checksums don't match and we have alternate mirrors to try.
442 */
f188591e
CM		 443static int btree_read_extent_buffer_pages(struct btrfs_root *root,
444 struct extent_buffer *eb,
ca7a79ad 445 u64 start, u64 parent_transid)
f188591e
CM		 446{
447 struct extent_io_tree *io_tree;
ea466794 448 int failed = 0;
f188591e
CM		 449 int ret;
450 int num_copies = 0;
451 int mirror_num = 0;
ea466794 452 int failed_mirror = 0;
f188591e 453
a826d6dc 454 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
f188591e
CM		 455 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
456 while (1) {
bb82ab88
AJ		 457 ret = read_extent_buffer_pages(io_tree, eb, start,
458 WAIT_COMPLETE,
f188591e 459 btree_get_extent, mirror_num);
256dd1bb
SB		 460 if (!ret) {
461 if (!verify_parent_transid(io_tree, eb,
b9fab919 462 parent_transid, 0))
256dd1bb
SB		 463 break;
464 else
465 ret = -EIO;
466 }
d397712b 467
a826d6dc
JB		 468 /*
469 * This buffer's crc is fine, but its contents are corrupted, so
470 * there is no reason to read the other copies, they won't be
471 * any less wrong.
472 */
473 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
ea466794
JB		 474 break;
475
5d964051 476 num_copies = btrfs_num_copies(root->fs_info,
f188591e 477 eb->start, eb->len);
4235298e 478 if (num_copies == 1)
ea466794 479 break;
4235298e 480
5cf1ab56
JB		 481 if (!failed_mirror) {
482 failed = 1;
483 failed_mirror = eb->read_mirror;
484 }
485
f188591e 486 mirror_num++;
ea466794
JB		 487 if (mirror_num == failed_mirror)
488 mirror_num++;
489
4235298e 490 if (mirror_num > num_copies)
ea466794 491 break;
f188591e 492 }
ea466794 493
c0901581 494 if (failed && !ret && failed_mirror)
ea466794
JB		 495 repair_eb_io_failure(root, eb, failed_mirror);
496
497 return ret;
f188591e 498}
19c00ddc 499
d352ac68 500/*
d397712b
CM		 501 * checksum a dirty tree block before IO. This has extra checks to make sure
502 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 503 */
d397712b 504
01d58472 505static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page)
19c00ddc 506{
4eee4fa4 507 u64 start = page_offset(page);
19c00ddc 508 u64 found_start;
19c00ddc 509 struct extent_buffer *eb;
f188591e 510
4f2de97a
JB		 511 eb = (struct extent_buffer *)page->private;
512 if (page != eb->pages[0])
513 return 0;
0f805531 514
19c00ddc 515 found_start = btrfs_header_bytenr(eb);
0f805531
AL		 516 /*
517 * Please do not consolidate these warnings into a single if.
518 * It is useful to know what went wrong.
519 */
520 if (WARN_ON(found_start != start))
521 return -EUCLEAN;
522 if (WARN_ON(!PageUptodate(page)))
523 return -EUCLEAN;
524
525 ASSERT(memcmp_extent_buffer(eb, fs_info->fsid,
526 btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0);
527
8bd98f0e 528 return csum_tree_block(fs_info, eb, 0);
19c00ddc
CM		 529}
530
01d58472 531static int check_tree_block_fsid(struct btrfs_fs_info *fs_info,
2b82032c
YZ		 532 struct extent_buffer *eb)
533{
01d58472 534 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2b82032c
YZ		 535 u8 fsid[BTRFS_UUID_SIZE];
536 int ret = 1;
537
0a4e5586 538 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
2b82032c
YZ		 539 while (fs_devices) {
540 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
541 ret = 0;
542 break;
543 }
544 fs_devices = fs_devices->seed;
545 }
546 return ret;
547}
548
a826d6dc 549#define CORRUPT(reason, eb, root, slot) \
efe120a0
FH		 550 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
551 "root=%llu, slot=%d", reason, \
c1c9ff7c 552 btrfs_header_bytenr(eb), root->objectid, slot)
a826d6dc
JB		 553
554static noinline int check_leaf(struct btrfs_root *root,
555 struct extent_buffer *leaf)
556{
557 struct btrfs_key key;
558 struct btrfs_key leaf_key;
559 u32 nritems = btrfs_header_nritems(leaf);
560 int slot;
561
1ba98d08
LB		 562 if (nritems == 0) {
563 struct btrfs_root *check_root;
564
565 key.objectid = btrfs_header_owner(leaf);
566 key.type = BTRFS_ROOT_ITEM_KEY;
567 key.offset = (u64)-1;
568
569 check_root = btrfs_get_fs_root(root->fs_info, &key, false);
570 /*
571 * The only reason we also check NULL here is that during
572 * open_ctree() some roots has not yet been set up.
573 */
574 if (!IS_ERR_OR_NULL(check_root)) {
575 /* if leaf is the root, then it's fine */
576 if (leaf->start !=
577 btrfs_root_bytenr(&check_root->root_item)) {
578 CORRUPT("non-root leaf's nritems is 0",
579 leaf, root, 0);
580 return -EIO;
581 }
582 }
a826d6dc 583 return 0;
1ba98d08 584 }
a826d6dc
JB		 585
586 /* Check the 0 item */
587 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
588 BTRFS_LEAF_DATA_SIZE(root)) {
589 CORRUPT("invalid item offset size pair", leaf, root, 0);
590 return -EIO;
591 }
592
593 /*
594 * Check to make sure each items keys are in the correct order and their
595 * offsets make sense. We only have to loop through nritems-1 because
596 * we check the current slot against the next slot, which verifies the
597 * next slot's offset+size makes sense and that the current's slot
598 * offset is correct.
599 */
600 for (slot = 0; slot < nritems - 1; slot++) {
601 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
602 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
603
604 /* Make sure the keys are in the right order */
605 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
606 CORRUPT("bad key order", leaf, root, slot);
607 return -EIO;
608 }
609
610 /*
611 * Make sure the offset and ends are right, remember that the
612 * item data starts at the end of the leaf and grows towards the
613 * front.
614 */
615 if (btrfs_item_offset_nr(leaf, slot) !=
616 btrfs_item_end_nr(leaf, slot + 1)) {
617 CORRUPT("slot offset bad", leaf, root, slot);
618 return -EIO;
619 }
620
621 /*
622 * Check to make sure that we don't point outside of the leaf,
01327610 623 * just in case all the items are consistent to each other, but
a826d6dc
JB		 624 * all point outside of the leaf.
625 */
626 if (btrfs_item_end_nr(leaf, slot) >
627 BTRFS_LEAF_DATA_SIZE(root)) {
628 CORRUPT("slot end outside of leaf", leaf, root, slot);
629 return -EIO;
630 }
631 }
632
633 return 0;
634}
635
053ab70f
LB		 636static int check_node(struct btrfs_root *root, struct extent_buffer *node)
637{
638 unsigned long nr = btrfs_header_nritems(node);
639
640 if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(root)) {
641 btrfs_crit(root->fs_info,
642 "corrupt node: block %llu root %llu nritems %lu",
643 node->start, root->objectid, nr);
644 return -EIO;
645 }
646 return 0;
647}
648
facc8a22
MX		 649static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
650 u64 phy_offset, struct page *page,
651 u64 start, u64 end, int mirror)
ce9adaa5 652{
ce9adaa5
CM		 653 u64 found_start;
654 int found_level;
ce9adaa5
CM		 655 struct extent_buffer *eb;
656 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
02873e43 657 struct btrfs_fs_info *fs_info = root->fs_info;
f188591e 658 int ret = 0;
727011e0 659 int reads_done;
ce9adaa5 660
ce9adaa5
CM		 661 if (!page->private)
662 goto out;
d397712b 663
4f2de97a 664 eb = (struct extent_buffer *)page->private;
d397712b 665
0b32f4bb
JB		 666 /* the pending IO might have been the only thing that kept this buffer
667 * in memory. Make sure we have a ref for all this other checks
668 */
669 extent_buffer_get(eb);
670
671 reads_done = atomic_dec_and_test(&eb->io_pages);
727011e0
CM		 672 if (!reads_done)
673 goto err;
f188591e 674
5cf1ab56 675 eb->read_mirror = mirror;
656f30db 676 if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) {
ea466794
JB		 677 ret = -EIO;
678 goto err;
679 }
680
ce9adaa5 681 found_start = btrfs_header_bytenr(eb);
727011e0 682 if (found_start != eb->start) {
02873e43
ZL		 683 btrfs_err_rl(fs_info, "bad tree block start %llu %llu",
684 found_start, eb->start);
f188591e 685 ret = -EIO;
ce9adaa5
CM		 686 goto err;
687 }
02873e43
ZL		 688 if (check_tree_block_fsid(fs_info, eb)) {
689 btrfs_err_rl(fs_info, "bad fsid on block %llu",
690 eb->start);
1259ab75
CM		 691 ret = -EIO;
692 goto err;
693 }
ce9adaa5 694 found_level = btrfs_header_level(eb);
1c24c3ce 695 if (found_level >= BTRFS_MAX_LEVEL) {
02873e43
ZL		 696 btrfs_err(fs_info, "bad tree block level %d",
697 (int)btrfs_header_level(eb));
1c24c3ce
JB		 698 ret = -EIO;
699 goto err;
700 }
ce9adaa5 701
85d4e461
CM		 702 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
703 eb, found_level);
4008c04a 704
02873e43 705 ret = csum_tree_block(fs_info, eb, 1);
8bd98f0e 706 if (ret)
a826d6dc 707 goto err;
a826d6dc
JB		 708
709 /*
710 * If this is a leaf block and it is corrupt, set the corrupt bit so
711 * that we don't try and read the other copies of this block, just
712 * return -EIO.
713 */
714 if (found_level == 0 && check_leaf(root, eb)) {
715 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
716 ret = -EIO;
717 }
ce9adaa5 718
053ab70f
LB		 719 if (found_level > 0 && check_node(root, eb))
720 ret = -EIO;
721
0b32f4bb
JB		 722 if (!ret)
723 set_extent_buffer_uptodate(eb);
ce9adaa5 724err:
79fb65a1
JB		 725 if (reads_done &&
726 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
02873e43 727 btree_readahead_hook(fs_info, eb, eb->start, ret);
4bb31e92 728
53b381b3
DW		 729 if (ret) {
730 /*
731 * our io error hook is going to dec the io pages
732 * again, we have to make sure it has something
733 * to decrement
734 */
735 atomic_inc(&eb->io_pages);
0b32f4bb 736 clear_extent_buffer_uptodate(eb);
53b381b3 737 }
0b32f4bb 738 free_extent_buffer(eb);
ce9adaa5 739out:
f188591e 740 return ret;
ce9adaa5
CM		 741}
742
ea466794 743static int btree_io_failed_hook(struct page *page, int failed_mirror)
4bb31e92 744{
4bb31e92 745 struct extent_buffer *eb;
4bb31e92 746
4f2de97a 747 eb = (struct extent_buffer *)page->private;
656f30db 748 set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
5cf1ab56 749 eb->read_mirror = failed_mirror;
53b381b3 750 atomic_dec(&eb->io_pages);
ea466794 751 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
02873e43 752 btree_readahead_hook(eb->fs_info, eb, eb->start, -EIO);
4bb31e92
AJ		 753 return -EIO; /* we fixed nothing */
754}
755
4246a0b6 756static void end_workqueue_bio(struct bio *bio)
ce9adaa5 757{
97eb6b69 758 struct btrfs_end_io_wq *end_io_wq = bio->bi_private;
ce9adaa5 759 struct btrfs_fs_info *fs_info;
9e0af237
LB		 760 struct btrfs_workqueue *wq;
761 btrfs_work_func_t func;
ce9adaa5 762
ce9adaa5 763 fs_info = end_io_wq->info;
4246a0b6 764 end_io_wq->error = bio->bi_error;
d20f7043 765
37226b21 766 if (bio_op(bio) == REQ_OP_WRITE) {
9e0af237
LB		 767 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) {
768 wq = fs_info->endio_meta_write_workers;
769 func = btrfs_endio_meta_write_helper;
770 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) {
771 wq = fs_info->endio_freespace_worker;
772 func = btrfs_freespace_write_helper;
773 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
774 wq = fs_info->endio_raid56_workers;
775 func = btrfs_endio_raid56_helper;
776 } else {
777 wq = fs_info->endio_write_workers;
778 func = btrfs_endio_write_helper;
779 }
d20f7043 780 } else {
8b110e39
MX		 781 if (unlikely(end_io_wq->metadata ==
782 BTRFS_WQ_ENDIO_DIO_REPAIR)) {
783 wq = fs_info->endio_repair_workers;
784 func = btrfs_endio_repair_helper;
785 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
9e0af237
LB		 786 wq = fs_info->endio_raid56_workers;
787 func = btrfs_endio_raid56_helper;
788 } else if (end_io_wq->metadata) {
789 wq = fs_info->endio_meta_workers;
790 func = btrfs_endio_meta_helper;
791 } else {
792 wq = fs_info->endio_workers;
793 func = btrfs_endio_helper;
794 }
d20f7043 795 }
9e0af237
LB		 796
797 btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL);
798 btrfs_queue_work(wq, &end_io_wq->work);
ce9adaa5
CM		 799}
800
22c59948 801int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
bfebd8b5 802 enum btrfs_wq_endio_type metadata)
0b86a832 803{
97eb6b69 804 struct btrfs_end_io_wq *end_io_wq;
8b110e39 805
97eb6b69 806 end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS);
ce9adaa5
CM		 807 if (!end_io_wq)
808 return -ENOMEM;
809
810 end_io_wq->private = bio->bi_private;
811 end_io_wq->end_io = bio->bi_end_io;
22c59948 812 end_io_wq->info = info;
ce9adaa5
CM		 813 end_io_wq->error = 0;
814 end_io_wq->bio = bio;
22c59948 815 end_io_wq->metadata = metadata;
ce9adaa5
CM		 816
817 bio->bi_private = end_io_wq;
818 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 819 return 0;
820}
821
b64a2851 822unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 823{
4854ddd0 824 unsigned long limit = min_t(unsigned long,
5cdc7ad3 825 info->thread_pool_size,
4854ddd0
CM		 826 info->fs_devices->open_devices);
827 return 256 * limit;
828}
0986fe9e 829
4a69a410
CM		 830static void run_one_async_start(struct btrfs_work *work)
831{
4a69a410 832 struct async_submit_bio *async;
79787eaa 833 int ret;
4a69a410
CM		 834
835 async = container_of(work, struct async_submit_bio, work);
81a75f67 836 ret = async->submit_bio_start(async->inode, async->bio,
79787eaa
JM		 837 async->mirror_num, async->bio_flags,
838 async->bio_offset);
839 if (ret)
840 async->error = ret;
4a69a410
CM		 841}
842
843static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 844{
845 struct btrfs_fs_info *fs_info;
846 struct async_submit_bio *async;
4854ddd0 847 int limit;
8b712842
CM		 848
849 async = container_of(work, struct async_submit_bio, work);
850 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 851
b64a2851 852 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 853 limit = limit * 2 / 3;
854
ee863954
DS		 855 /*
856 * atomic_dec_return implies a barrier for waitqueue_active
857 */
66657b31 858 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
b64a2851 859 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 860 wake_up(&fs_info->async_submit_wait);
861
bb7ab3b9 862 /* If an error occurred we just want to clean up the bio and move on */
79787eaa 863 if (async->error) {
4246a0b6
CH		 864 async->bio->bi_error = async->error;
865 bio_endio(async->bio);
79787eaa
JM		 866 return;
867 }
868
81a75f67
MC		 869 async->submit_bio_done(async->inode, async->bio, async->mirror_num,
870 async->bio_flags, async->bio_offset);
4a69a410
CM		 871}
872
873static void run_one_async_free(struct btrfs_work *work)
874{
875 struct async_submit_bio *async;
876
877 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 878 kfree(async);
879}
880
44b8bd7e 881int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
81a75f67 882 struct bio *bio, int mirror_num,
c8b97818 883 unsigned long bio_flags,
eaf25d93 884 u64 bio_offset,
4a69a410
CM		 885 extent_submit_bio_hook_t *submit_bio_start,
886 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 887{
888 struct async_submit_bio *async;
889
890 async = kmalloc(sizeof(*async), GFP_NOFS);
891 if (!async)
892 return -ENOMEM;
893
894 async->inode = inode;
44b8bd7e
CM		 895 async->bio = bio;
896 async->mirror_num = mirror_num;
4a69a410
CM		 897 async->submit_bio_start = submit_bio_start;
898 async->submit_bio_done = submit_bio_done;
899
9e0af237 900 btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start,
5cdc7ad3 901 run_one_async_done, run_one_async_free);
4a69a410 902
c8b97818 903 async->bio_flags = bio_flags;
eaf25d93 904 async->bio_offset = bio_offset;
8c8bee1d 905
79787eaa
JM		 906 async->error = 0;
907
cb03c743 908 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 909
1eff9d32 910 if (bio->bi_opf & REQ_SYNC)
5cdc7ad3 911 btrfs_set_work_high_priority(&async->work);
d313d7a3 912
5cdc7ad3 913 btrfs_queue_work(fs_info->workers, &async->work);
9473f16c 914
d397712b 915 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 916 atomic_read(&fs_info->nr_async_submits)) {
917 wait_event(fs_info->async_submit_wait,
918 (atomic_read(&fs_info->nr_async_submits) == 0));
919 }
920
44b8bd7e
CM		 921 return 0;
922}
923
ce3ed71a
CM		 924static int btree_csum_one_bio(struct bio *bio)
925{
2c30c71b 926 struct bio_vec *bvec;
ce3ed71a 927 struct btrfs_root *root;
2c30c71b 928 int i, ret = 0;
ce3ed71a 929
2c30c71b 930 bio_for_each_segment_all(bvec, bio, i) {
ce3ed71a 931 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
01d58472 932 ret = csum_dirty_buffer(root->fs_info, bvec->bv_page);
79787eaa
JM		 933 if (ret)
934 break;
ce3ed71a 935 }
2c30c71b 936
79787eaa 937 return ret;
ce3ed71a
CM		 938}
939
81a75f67
MC		 940static int __btree_submit_bio_start(struct inode *inode, struct bio *bio,
941 int mirror_num, unsigned long bio_flags,
eaf25d93 942 u64 bio_offset)
22c59948 943{
8b712842
CM		 944 /*
945 * when we're called for a write, we're already in the async
5443be45 946 * submission context. Just jump into btrfs_map_bio
8b712842 947 */
79787eaa 948 return btree_csum_one_bio(bio);
4a69a410 949}
22c59948 950
81a75f67 951static int __btree_submit_bio_done(struct inode *inode, struct bio *bio,
eaf25d93
CM		 952 int mirror_num, unsigned long bio_flags,
953 u64 bio_offset)
4a69a410 954{
61891923
SB		 955 int ret;
956
8b712842 957 /*
4a69a410
CM		 958 * when we're called for a write, we're already in the async
959 * submission context. Just jump into btrfs_map_bio
8b712842 960 */
81a75f67 961 ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 1);
4246a0b6
CH		 962 if (ret) {
963 bio->bi_error = ret;
964 bio_endio(bio);
965 }
61891923 966 return ret;
0b86a832
CM		 967}
968
de0022b9
JB		 969static int check_async_write(struct inode *inode, unsigned long bio_flags)
970{
971 if (bio_flags & EXTENT_BIO_TREE_LOG)
972 return 0;
973#ifdef CONFIG_X86
bc696ca0 974 if (static_cpu_has(X86_FEATURE_XMM4_2))
de0022b9
JB		 975 return 0;
976#endif
977 return 1;
978}
979
81a75f67 980static int btree_submit_bio_hook(struct inode *inode, struct bio *bio,
eaf25d93
CM		 981 int mirror_num, unsigned long bio_flags,
982 u64 bio_offset)
44b8bd7e 983{
de0022b9 984 int async = check_async_write(inode, bio_flags);
cad321ad
CM		 985 int ret;
986
37226b21 987 if (bio_op(bio) != REQ_OP_WRITE) {
4a69a410
CM		 988 /*
989 * called for a read, do the setup so that checksum validation
990 * can happen in the async kernel threads
991 */
f3f266ab 992 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
bfebd8b5 993 bio, BTRFS_WQ_ENDIO_METADATA);
1d4284bd 994 if (ret)
61891923 995 goto out_w_error;
81a75f67 996 ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 0);
de0022b9
JB		 997 } else if (!async) {
998 ret = btree_csum_one_bio(bio);
999 if (ret)
61891923 1000 goto out_w_error;
81a75f67 1001 ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 0);
61891923
SB		 1002 } else {
1003 /*
1004 * kthread helpers are used to submit writes so that
1005 * checksumming can happen in parallel across all CPUs
1006 */
1007 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
81a75f67 1008 inode, bio, mirror_num, 0,
61891923
SB		 1009 bio_offset,
1010 __btree_submit_bio_start,
1011 __btree_submit_bio_done);
44b8bd7e 1012 }
d313d7a3 1013
4246a0b6
CH		 1014 if (ret)
1015 goto out_w_error;
1016 return 0;
1017
61891923 1018out_w_error:
4246a0b6
CH		 1019 bio->bi_error = ret;
1020 bio_endio(bio);
61891923 1021 return ret;
44b8bd7e
CM		 1022}
1023
3dd1462e 1024#ifdef CONFIG_MIGRATION
784b4e29 1025static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG		 1026 struct page *newpage, struct page *page,
1027 enum migrate_mode mode)
784b4e29
CM		 1028{
1029 /*
1030 * we can't safely write a btree page from here,
1031 * we haven't done the locking hook
1032 */
1033 if (PageDirty(page))
1034 return -EAGAIN;
1035 /*
1036 * Buffers may be managed in a filesystem specific way.
1037 * We must have no buffers or drop them.
1038 */
1039 if (page_has_private(page) &&
1040 !try_to_release_page(page, GFP_KERNEL))
1041 return -EAGAIN;
a6bc32b8 1042 return migrate_page(mapping, newpage, page, mode);
784b4e29 1043}
3dd1462e 1044#endif
784b4e29 1045
0da5468f
CM		 1046
1047static int btree_writepages(struct address_space *mapping,
1048 struct writeback_control *wbc)
1049{
e2d84521
MX		 1050 struct btrfs_fs_info *fs_info;
1051 int ret;
1052
d8d5f3e1 1053 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM		 1054
1055 if (wbc->for_kupdate)
1056 return 0;
1057
e2d84521 1058 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 1059 /* this is a bit racy, but that's ok */
e2d84521
MX		 1060 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
1061 BTRFS_DIRTY_METADATA_THRESH);
1062 if (ret < 0)
793955bc 1063 return 0;
793955bc 1064 }
0b32f4bb 1065 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM		 1066}
1067
b2950863 1068static int btree_readpage(struct file *file, struct page *page)
5f39d397 1069{
d1310b2e
CM		 1070 struct extent_io_tree *tree;
1071 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 1072 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 1073}
22b0ebda 1074
70dec807 1075static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 1076{
98509cfc 1077 if (PageWriteback(page) || PageDirty(page))
d397712b 1078 return 0;
0c4e538b 1079
f7a52a40 1080 return try_release_extent_buffer(page);
d98237b3
CM		 1081}
1082
d47992f8
LC		 1083static void btree_invalidatepage(struct page *page, unsigned int offset,
1084 unsigned int length)
d98237b3 1085{
d1310b2e
CM		 1086 struct extent_io_tree *tree;
1087 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 1088 extent_invalidatepage(tree, page, offset);
1089 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 1090 if (PagePrivate(page)) {
efe120a0
FH		 1091 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1092 "page private not zero on page %llu",
1093 (unsigned long long)page_offset(page));
9ad6b7bc
CM		 1094 ClearPagePrivate(page);
1095 set_page_private(page, 0);
09cbfeaf 1096 put_page(page);
9ad6b7bc 1097 }
d98237b3
CM		 1098}
1099
0b32f4bb
JB		 1100static int btree_set_page_dirty(struct page *page)
1101{
bb146eb2 1102#ifdef DEBUG
0b32f4bb
JB		 1103 struct extent_buffer *eb;
1104
1105 BUG_ON(!PagePrivate(page));
1106 eb = (struct extent_buffer *)page->private;
1107 BUG_ON(!eb);
1108 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1109 BUG_ON(!atomic_read(&eb->refs));
1110 btrfs_assert_tree_locked(eb);
bb146eb2 1111#endif
0b32f4bb
JB		 1112 return __set_page_dirty_nobuffers(page);
1113}
1114
7f09410b 1115static const struct address_space_operations btree_aops = {
d98237b3 1116 .readpage = btree_readpage,
0da5468f 1117 .writepages = btree_writepages,
5f39d397
CM		 1118 .releasepage = btree_releasepage,
1119 .invalidatepage = btree_invalidatepage,
5a92bc88 1120#ifdef CONFIG_MIGRATION
784b4e29 1121 .migratepage = btree_migratepage,
5a92bc88 1122#endif
0b32f4bb 1123 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM		 1124};
1125
d3e46fea 1126void readahead_tree_block(struct btrfs_root *root, u64 bytenr)
090d1875 1127{
5f39d397
CM		 1128 struct extent_buffer *buf = NULL;
1129 struct inode *btree_inode = root->fs_info->btree_inode;
090d1875 1130
a83fffb7 1131 buf = btrfs_find_create_tree_block(root, bytenr);
c871b0f2 1132 if (IS_ERR(buf))
6197d86e 1133 return;
d1310b2e 1134 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
bb82ab88 1135 buf, 0, WAIT_NONE, btree_get_extent, 0);
5f39d397 1136 free_extent_buffer(buf);
090d1875
CM		 1137}
1138
c0dcaa4d 1139int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr,
ab0fff03
AJ		 1140 int mirror_num, struct extent_buffer **eb)
1141{
1142 struct extent_buffer *buf = NULL;
1143 struct inode *btree_inode = root->fs_info->btree_inode;
1144 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1145 int ret;
1146
a83fffb7 1147 buf = btrfs_find_create_tree_block(root, bytenr);
c871b0f2 1148 if (IS_ERR(buf))
ab0fff03
AJ		 1149 return 0;
1150
1151 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1152
1153 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1154 btree_get_extent, mirror_num);
1155 if (ret) {
1156 free_extent_buffer(buf);
1157 return ret;
1158 }
1159
1160 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1161 free_extent_buffer(buf);
1162 return -EIO;
0b32f4bb 1163 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ		 1164 *eb = buf;
1165 } else {
1166 free_extent_buffer(buf);
1167 }
1168 return 0;
1169}
1170
01d58472 1171struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info,
0308af44 1172 u64 bytenr)
0999df54 1173{
01d58472 1174 return find_extent_buffer(fs_info, bytenr);
0999df54
CM		 1175}
1176
1177struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
a83fffb7 1178 u64 bytenr)
0999df54 1179{
f5ee5c9a 1180 if (btrfs_is_testing(root->fs_info))
b9ef22de
FX		 1181 return alloc_test_extent_buffer(root->fs_info, bytenr,
1182 root->nodesize);
ce3e6984 1183 return alloc_extent_buffer(root->fs_info, bytenr);
0999df54
CM		 1184}
1185
1186
e02119d5
CM		 1187int btrfs_write_tree_block(struct extent_buffer *buf)
1188{
727011e0 1189 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1190 buf->start + buf->len - 1);
e02119d5
CM		 1191}
1192
1193int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1194{
727011e0 1195 return filemap_fdatawait_range(buf->pages[0]->mapping,
8aa38c31 1196 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1197}
1198
0999df54 1199struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ce86cd59 1200 u64 parent_transid)
0999df54
CM		 1201{
1202 struct extent_buffer *buf = NULL;
0999df54
CM		 1203 int ret;
1204
a83fffb7 1205 buf = btrfs_find_create_tree_block(root, bytenr);
c871b0f2
LB		 1206 if (IS_ERR(buf))
1207 return buf;
0999df54 1208
ca7a79ad 1209 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
0f0fe8f7
FDBM		 1210 if (ret) {
1211 free_extent_buffer(buf);
64c043de 1212 return ERR_PTR(ret);
0f0fe8f7 1213 }
5f39d397 1214 return buf;
ce9adaa5 1215
eb60ceac
CM		 1216}
1217
01d58472
DD		 1218void clean_tree_block(struct btrfs_trans_handle *trans,
1219 struct btrfs_fs_info *fs_info,
d5c13f92 1220 struct extent_buffer *buf)
ed2ff2cb 1221{
55c69072 1222 if (btrfs_header_generation(buf) ==
e2d84521 1223 fs_info->running_transaction->transid) {
b9447ef8 1224 btrfs_assert_tree_locked(buf);
b4ce94de 1225
b9473439 1226 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
e2d84521
MX		 1227 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1228 -buf->len,
1229 fs_info->dirty_metadata_batch);
ed7b63eb
JB		 1230 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1231 btrfs_set_lock_blocking(buf);
1232 clear_extent_buffer_dirty(buf);
1233 }
925baedd 1234 }
5f39d397
CM		 1235}
1236
8257b2dc
MX		 1237static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1238{
1239 struct btrfs_subvolume_writers *writers;
1240 int ret;
1241
1242 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1243 if (!writers)
1244 return ERR_PTR(-ENOMEM);
1245
908c7f19 1246 ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL);
8257b2dc
MX		 1247 if (ret < 0) {
1248 kfree(writers);
1249 return ERR_PTR(ret);
1250 }
1251
1252 init_waitqueue_head(&writers->wait);
1253 return writers;
1254}
1255
1256static void
1257btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1258{
1259 percpu_counter_destroy(&writers->counter);
1260 kfree(writers);
1261}
1262
707e8a07
DS		 1263static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize,
1264 struct btrfs_root *root, struct btrfs_fs_info *fs_info,
143bede5 1265 u64 objectid)
d97e63b6 1266{
7c0260ee 1267 bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state);
cfaa7295 1268 root->node = NULL;
a28ec197 1269 root->commit_root = NULL;
db94535d
CM		 1270 root->sectorsize = sectorsize;
1271 root->nodesize = nodesize;
87ee04eb 1272 root->stripesize = stripesize;
27cdeb70 1273 root->state = 0;
d68fc57b 1274 root->orphan_cleanup_state = 0;
0b86a832 1275
0f7d52f4
CM		 1276 root->objectid = objectid;
1277 root->last_trans = 0;
13a8a7c8 1278 root->highest_objectid = 0;
eb73c1b7 1279 root->nr_delalloc_inodes = 0;
199c2a9c 1280 root->nr_ordered_extents = 0;
58176a96 1281 root->name = NULL;
6bef4d31 1282 root->inode_tree = RB_ROOT;
16cdcec7 1283 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1284 root->block_rsv = NULL;
d68fc57b 1285 root->orphan_block_rsv = NULL;
0b86a832
CM		 1286
1287 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1288 INIT_LIST_HEAD(&root->root_list);
eb73c1b7
MX		 1289 INIT_LIST_HEAD(&root->delalloc_inodes);
1290 INIT_LIST_HEAD(&root->delalloc_root);
199c2a9c
MX		 1291 INIT_LIST_HEAD(&root->ordered_extents);
1292 INIT_LIST_HEAD(&root->ordered_root);
2ab28f32
JB		 1293 INIT_LIST_HEAD(&root->logged_list[0]);
1294 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1295 spin_lock_init(&root->orphan_lock);
5d4f98a2 1296 spin_lock_init(&root->inode_lock);
eb73c1b7 1297 spin_lock_init(&root->delalloc_lock);
199c2a9c 1298 spin_lock_init(&root->ordered_extent_lock);
f0486c68 1299 spin_lock_init(&root->accounting_lock);
2ab28f32
JB		 1300 spin_lock_init(&root->log_extents_lock[0]);
1301 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1302 mutex_init(&root->objectid_mutex);
e02119d5 1303 mutex_init(&root->log_mutex);
31f3d255 1304 mutex_init(&root->ordered_extent_mutex);
573bfb72 1305 mutex_init(&root->delalloc_mutex);
7237f183
YZ		 1306 init_waitqueue_head(&root->log_writer_wait);
1307 init_waitqueue_head(&root->log_commit_wait[0]);
1308 init_waitqueue_head(&root->log_commit_wait[1]);
8b050d35
MX		 1309 INIT_LIST_HEAD(&root->log_ctxs[0]);
1310 INIT_LIST_HEAD(&root->log_ctxs[1]);
7237f183
YZ		 1311 atomic_set(&root->log_commit[0], 0);
1312 atomic_set(&root->log_commit[1], 0);
1313 atomic_set(&root->log_writers, 0);
2ecb7923 1314 atomic_set(&root->log_batch, 0);
8a35d95f 1315 atomic_set(&root->orphan_inodes, 0);
b0feb9d9 1316 atomic_set(&root->refs, 1);
8257b2dc 1317 atomic_set(&root->will_be_snapshoted, 0);
55eeaf05 1318 atomic_set(&root->qgroup_meta_rsv, 0);
7237f183 1319 root->log_transid = 0;
d1433deb 1320 root->log_transid_committed = -1;
257c62e1 1321 root->last_log_commit = 0;
7c0260ee 1322 if (!dummy)
06ea65a3
JB		 1323 extent_io_tree_init(&root->dirty_log_pages,
1324 fs_info->btree_inode->i_mapping);
017e5369 1325
3768f368
CM		 1326 memset(&root->root_key, 0, sizeof(root->root_key));
1327 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1328 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
7c0260ee 1329 if (!dummy)
06ea65a3
JB		 1330 root->defrag_trans_start = fs_info->generation;
1331 else
1332 root->defrag_trans_start = 0;
4d775673 1333 root->root_key.objectid = objectid;
0ee5dc67 1334 root->anon_dev = 0;
8ea05e3a 1335
5f3ab90a 1336 spin_lock_init(&root->root_item_lock);
3768f368
CM		 1337}
1338
74e4d827
DS		 1339static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info,
1340 gfp_t flags)
6f07e42e 1341{
74e4d827 1342 struct btrfs_root *root = kzalloc(sizeof(*root), flags);
6f07e42e
AV		 1343 if (root)
1344 root->fs_info = fs_info;
1345 return root;
1346}
1347
06ea65a3
JB		 1348#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1349/* Should only be used by the testing infrastructure */
7c0260ee
JM		 1350struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info,
1351 u32 sectorsize, u32 nodesize)
06ea65a3
JB		 1352{
1353 struct btrfs_root *root;
1354
7c0260ee
JM		 1355 if (!fs_info)
1356 return ERR_PTR(-EINVAL);
1357
1358 root = btrfs_alloc_root(fs_info, GFP_KERNEL);
06ea65a3
JB		 1359 if (!root)
1360 return ERR_PTR(-ENOMEM);
b9ef22de 1361 /* We don't use the stripesize in selftest, set it as sectorsize */
7c0260ee 1362 __setup_root(nodesize, sectorsize, sectorsize, root, fs_info,
ef9f2db3 1363 BTRFS_ROOT_TREE_OBJECTID);
faa2dbf0 1364 root->alloc_bytenr = 0;
06ea65a3
JB		 1365
1366 return root;
1367}
1368#endif
1369
20897f5c
AJ		 1370struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1371 struct btrfs_fs_info *fs_info,
1372 u64 objectid)
1373{
1374 struct extent_buffer *leaf;
1375 struct btrfs_root *tree_root = fs_info->tree_root;
1376 struct btrfs_root *root;
1377 struct btrfs_key key;
1378 int ret = 0;
6463fe58 1379 uuid_le uuid;
20897f5c 1380
74e4d827 1381 root = btrfs_alloc_root(fs_info, GFP_KERNEL);
20897f5c
AJ		 1382 if (!root)
1383 return ERR_PTR(-ENOMEM);
1384
707e8a07
DS		 1385 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1386 tree_root->stripesize, root, fs_info, objectid);
20897f5c
AJ		 1387 root->root_key.objectid = objectid;
1388 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1389 root->root_key.offset = 0;
1390
4d75f8a9 1391 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
20897f5c
AJ		 1392 if (IS_ERR(leaf)) {
1393 ret = PTR_ERR(leaf);
1dd05682 1394 leaf = NULL;
20897f5c
AJ		 1395 goto fail;
1396 }
1397
20897f5c
AJ		 1398 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1399 btrfs_set_header_bytenr(leaf, leaf->start);
1400 btrfs_set_header_generation(leaf, trans->transid);
1401 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1402 btrfs_set_header_owner(leaf, objectid);
1403 root->node = leaf;
1404
0a4e5586 1405 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
20897f5c
AJ		 1406 BTRFS_FSID_SIZE);
1407 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
b308bc2f 1408 btrfs_header_chunk_tree_uuid(leaf),
20897f5c
AJ		 1409 BTRFS_UUID_SIZE);
1410 btrfs_mark_buffer_dirty(leaf);
1411
1412 root->commit_root = btrfs_root_node(root);
27cdeb70 1413 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
20897f5c
AJ		 1414
1415 root->root_item.flags = 0;
1416 root->root_item.byte_limit = 0;
1417 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1418 btrfs_set_root_generation(&root->root_item, trans->transid);
1419 btrfs_set_root_level(&root->root_item, 0);
1420 btrfs_set_root_refs(&root->root_item, 1);
1421 btrfs_set_root_used(&root->root_item, leaf->len);
1422 btrfs_set_root_last_snapshot(&root->root_item, 0);
1423 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB		 1424 uuid_le_gen(&uuid);
1425 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ		 1426 root->root_item.drop_level = 0;
1427
1428 key.objectid = objectid;
1429 key.type = BTRFS_ROOT_ITEM_KEY;
1430 key.offset = 0;
1431 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1432 if (ret)
1433 goto fail;
1434
1435 btrfs_tree_unlock(leaf);
1436
1dd05682
TI		 1437 return root;
1438
20897f5c 1439fail:
1dd05682
TI		 1440 if (leaf) {
1441 btrfs_tree_unlock(leaf);
59885b39 1442 free_extent_buffer(root->commit_root);
1dd05682
TI		 1443 free_extent_buffer(leaf);
1444 }
1445 kfree(root);
20897f5c 1446
1dd05682 1447 return ERR_PTR(ret);
20897f5c
AJ		 1448}
1449
7237f183
YZ		 1450static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1451 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1452{
1453 struct btrfs_root *root;
1454 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1455 struct extent_buffer *leaf;
e02119d5 1456
74e4d827 1457 root = btrfs_alloc_root(fs_info, GFP_NOFS);
e02119d5 1458 if (!root)
7237f183 1459 return ERR_PTR(-ENOMEM);
e02119d5 1460
707e8a07
DS		 1461 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1462 tree_root->stripesize, root, fs_info,
1463 BTRFS_TREE_LOG_OBJECTID);
e02119d5
CM		 1464
1465 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1466 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1467 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
27cdeb70 1468
7237f183 1469 /*
27cdeb70
MX		 1470 * DON'T set REF_COWS for log trees
1471 *
7237f183
YZ		 1472 * log trees do not get reference counted because they go away
1473 * before a real commit is actually done. They do store pointers
1474 * to file data extents, and those reference counts still get
1475 * updated (along with back refs to the log tree).
1476 */
e02119d5 1477
4d75f8a9
DS		 1478 leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID,
1479 NULL, 0, 0, 0);
7237f183
YZ		 1480 if (IS_ERR(leaf)) {
1481 kfree(root);
1482 return ERR_CAST(leaf);
1483 }
e02119d5 1484
5d4f98a2
YZ		 1485 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1486 btrfs_set_header_bytenr(leaf, leaf->start);
1487 btrfs_set_header_generation(leaf, trans->transid);
1488 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1489 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1490 root->node = leaf;
e02119d5
CM		 1491
1492 write_extent_buffer(root->node, root->fs_info->fsid,
0a4e5586 1493 btrfs_header_fsid(), BTRFS_FSID_SIZE);
e02119d5
CM		 1494 btrfs_mark_buffer_dirty(root->node);
1495 btrfs_tree_unlock(root->node);
7237f183
YZ		 1496 return root;
1497}
1498
1499int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1500 struct btrfs_fs_info *fs_info)
1501{
1502 struct btrfs_root *log_root;
1503
1504 log_root = alloc_log_tree(trans, fs_info);
1505 if (IS_ERR(log_root))
1506 return PTR_ERR(log_root);
1507 WARN_ON(fs_info->log_root_tree);
1508 fs_info->log_root_tree = log_root;
1509 return 0;
1510}
1511
1512int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1513 struct btrfs_root *root)
1514{
1515 struct btrfs_root *log_root;
1516 struct btrfs_inode_item *inode_item;
1517
1518 log_root = alloc_log_tree(trans, root->fs_info);
1519 if (IS_ERR(log_root))
1520 return PTR_ERR(log_root);
1521
1522 log_root->last_trans = trans->transid;
1523 log_root->root_key.offset = root->root_key.objectid;
1524
1525 inode_item = &log_root->root_item.inode;
3cae210f
QW		 1526 btrfs_set_stack_inode_generation(inode_item, 1);
1527 btrfs_set_stack_inode_size(inode_item, 3);
1528 btrfs_set_stack_inode_nlink(inode_item, 1);
707e8a07 1529 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
3cae210f 1530 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
7237f183 1531
5d4f98a2 1532 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1533
1534 WARN_ON(root->log_root);
1535 root->log_root = log_root;
1536 root->log_transid = 0;
d1433deb 1537 root->log_transid_committed = -1;
257c62e1 1538 root->last_log_commit = 0;
e02119d5
CM		 1539 return 0;
1540}
1541
35a3621b
SB		 1542static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1543 struct btrfs_key *key)
e02119d5
CM		 1544{
1545 struct btrfs_root *root;
1546 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1547 struct btrfs_path *path;
84234f3a 1548 u64 generation;
cb517eab 1549 int ret;
0f7d52f4 1550
cb517eab
MX		 1551 path = btrfs_alloc_path();
1552 if (!path)
0f7d52f4 1553 return ERR_PTR(-ENOMEM);
cb517eab 1554
74e4d827 1555 root = btrfs_alloc_root(fs_info, GFP_NOFS);
cb517eab
MX		 1556 if (!root) {
1557 ret = -ENOMEM;
1558 goto alloc_fail;
0f7d52f4
CM		 1559 }
1560
707e8a07
DS		 1561 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1562 tree_root->stripesize, root, fs_info, key->objectid);
0f7d52f4 1563
cb517eab
MX		 1564 ret = btrfs_find_root(tree_root, key, path,
1565 &root->root_item, &root->root_key);
0f7d52f4 1566 if (ret) {
13a8a7c8
YZ		 1567 if (ret > 0)
1568 ret = -ENOENT;
cb517eab 1569 goto find_fail;
0f7d52f4 1570 }
13a8a7c8 1571
84234f3a 1572 generation = btrfs_root_generation(&root->root_item);
db94535d 1573 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
ce86cd59 1574 generation);
64c043de
LB		 1575 if (IS_ERR(root->node)) {
1576 ret = PTR_ERR(root->node);
cb517eab
MX		 1577 goto find_fail;
1578 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1579 ret = -EIO;
64c043de
LB		 1580 free_extent_buffer(root->node);
1581 goto find_fail;
416bc658 1582 }
5d4f98a2 1583 root->commit_root = btrfs_root_node(root);
13a8a7c8 1584out:
cb517eab
MX		 1585 btrfs_free_path(path);
1586 return root;
1587
cb517eab
MX		 1588find_fail:
1589 kfree(root);
1590alloc_fail:
1591 root = ERR_PTR(ret);
1592 goto out;
1593}
1594
1595struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1596 struct btrfs_key *location)
1597{
1598 struct btrfs_root *root;
1599
1600 root = btrfs_read_tree_root(tree_root, location);
1601 if (IS_ERR(root))
1602 return root;
1603
1604 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
27cdeb70 1605 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
08fe4db1
LZ		 1606 btrfs_check_and_init_root_item(&root->root_item);
1607 }
13a8a7c8 1608
5eda7b5e
CM		 1609 return root;
1610}
1611
cb517eab
MX		 1612int btrfs_init_fs_root(struct btrfs_root *root)
1613{
1614 int ret;
8257b2dc 1615 struct btrfs_subvolume_writers *writers;
cb517eab
MX		 1616
1617 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1618 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1619 GFP_NOFS);
1620 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1621 ret = -ENOMEM;
1622 goto fail;
1623 }
1624
8257b2dc
MX		 1625 writers = btrfs_alloc_subvolume_writers();
1626 if (IS_ERR(writers)) {
1627 ret = PTR_ERR(writers);
1628 goto fail;
1629 }
1630 root->subv_writers = writers;
1631
cb517eab 1632 btrfs_init_free_ino_ctl(root);
57cdc8db
DS		 1633 spin_lock_init(&root->ino_cache_lock);
1634 init_waitqueue_head(&root->ino_cache_wait);
cb517eab
MX		 1635
1636 ret = get_anon_bdev(&root->anon_dev);
1637 if (ret)
876d2cf1 1638 goto fail;
f32e48e9
CR		 1639
1640 mutex_lock(&root->objectid_mutex);
1641 ret = btrfs_find_highest_objectid(root,
1642 &root->highest_objectid);
1643 if (ret) {
1644 mutex_unlock(&root->objectid_mutex);
876d2cf1 1645 goto fail;
f32e48e9
CR		 1646 }
1647
1648 ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID);
1649
1650 mutex_unlock(&root->objectid_mutex);
1651
cb517eab
MX		 1652 return 0;
1653fail:
876d2cf1 1654 /* the caller is responsible to call free_fs_root */
cb517eab
MX		 1655 return ret;
1656}
1657
35bbb97f
JM		 1658struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1659 u64 root_id)
cb517eab
MX		 1660{
1661 struct btrfs_root *root;
1662
1663 spin_lock(&fs_info->fs_roots_radix_lock);
1664 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1665 (unsigned long)root_id);
1666 spin_unlock(&fs_info->fs_roots_radix_lock);
1667 return root;
1668}
1669
1670int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1671 struct btrfs_root *root)
1672{
1673 int ret;
1674
e1860a77 1675 ret = radix_tree_preload(GFP_NOFS);
cb517eab
MX		 1676 if (ret)
1677 return ret;
1678
1679 spin_lock(&fs_info->fs_roots_radix_lock);
1680 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1681 (unsigned long)root->root_key.objectid,
1682 root);
1683 if (ret == 0)
27cdeb70 1684 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
cb517eab
MX		 1685 spin_unlock(&fs_info->fs_roots_radix_lock);
1686 radix_tree_preload_end();
1687
1688 return ret;
1689}
1690
c00869f1
MX		 1691struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1692 struct btrfs_key *location,
1693 bool check_ref)
5eda7b5e
CM		 1694{
1695 struct btrfs_root *root;
381cf658 1696 struct btrfs_path *path;
1d4c08e0 1697 struct btrfs_key key;
5eda7b5e
CM		 1698 int ret;
1699
edbd8d4e
CM		 1700 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1701 return fs_info->tree_root;
1702 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1703 return fs_info->extent_root;
8f18cf13
CM		 1704 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1705 return fs_info->chunk_root;
1706 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1707 return fs_info->dev_root;
0403e47e
YZ		 1708 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1709 return fs_info->csum_root;
bcef60f2
AJ		 1710 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1711 return fs_info->quota_root ? fs_info->quota_root :
1712 ERR_PTR(-ENOENT);
f7a81ea4
SB		 1713 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1714 return fs_info->uuid_root ? fs_info->uuid_root :
1715 ERR_PTR(-ENOENT);
70f6d82e
OS		 1716 if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID)
1717 return fs_info->free_space_root ? fs_info->free_space_root :
1718 ERR_PTR(-ENOENT);
4df27c4d 1719again:
cb517eab 1720 root = btrfs_lookup_fs_root(fs_info, location->objectid);
48475471 1721 if (root) {
c00869f1 1722 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
48475471 1723 return ERR_PTR(-ENOENT);
5eda7b5e 1724 return root;
48475471 1725 }
5eda7b5e 1726
cb517eab 1727 root = btrfs_read_fs_root(fs_info->tree_root, location);
5eda7b5e
CM		 1728 if (IS_ERR(root))
1729 return root;
3394e160 1730
c00869f1 1731 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
cb517eab 1732 ret = -ENOENT;
581bb050 1733 goto fail;
35a30d7c 1734 }
581bb050 1735
cb517eab 1736 ret = btrfs_init_fs_root(root);
ac08aedf
CM		 1737 if (ret)
1738 goto fail;
3394e160 1739
381cf658
DS		 1740 path = btrfs_alloc_path();
1741 if (!path) {
1742 ret = -ENOMEM;
1743 goto fail;
1744 }
1d4c08e0
DS		 1745 key.objectid = BTRFS_ORPHAN_OBJECTID;
1746 key.type = BTRFS_ORPHAN_ITEM_KEY;
1747 key.offset = location->objectid;
1748
1749 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
381cf658 1750 btrfs_free_path(path);
d68fc57b
YZ		 1751 if (ret < 0)
1752 goto fail;
1753 if (ret == 0)
27cdeb70 1754 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
d68fc57b 1755
cb517eab 1756 ret = btrfs_insert_fs_root(fs_info, root);
0f7d52f4 1757 if (ret) {
4df27c4d
YZ		 1758 if (ret == -EEXIST) {
1759 free_fs_root(root);
1760 goto again;
1761 }
1762 goto fail;
0f7d52f4 1763 }
edbd8d4e 1764 return root;
4df27c4d
YZ		 1765fail:
1766 free_fs_root(root);
1767 return ERR_PTR(ret);
edbd8d4e
CM		 1768}
1769
04160088
CM		 1770static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1771{
1772 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1773 int ret = 0;
04160088
CM		 1774 struct btrfs_device *device;
1775 struct backing_dev_info *bdi;
b7967db7 1776
1f78160c
XG		 1777 rcu_read_lock();
1778 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1779 if (!device->bdev)
1780 continue;
04160088 1781 bdi = blk_get_backing_dev_info(device->bdev);
ff9ea323 1782 if (bdi_congested(bdi, bdi_bits)) {
04160088
CM		 1783 ret = 1;
1784 break;
1785 }
1786 }
1f78160c 1787 rcu_read_unlock();
04160088
CM		 1788 return ret;
1789}
1790
04160088
CM		 1791static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1792{
ad081f14
JA		 1793 int err;
1794
b4caecd4 1795 err = bdi_setup_and_register(bdi, "btrfs");
ad081f14
JA		 1796 if (err)
1797 return err;
1798
09cbfeaf 1799 bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_SIZE;
04160088
CM		 1800 bdi->congested_fn = btrfs_congested_fn;
1801 bdi->congested_data = info;
da2f0f74 1802 bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK;
04160088
CM		 1803 return 0;
1804}
1805
8b712842
CM		 1806/*
1807 * called by the kthread helper functions to finally call the bio end_io
1808 * functions. This is where read checksum verification actually happens
1809 */
1810static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1811{
ce9adaa5 1812 struct bio *bio;
97eb6b69 1813 struct btrfs_end_io_wq *end_io_wq;
ce9adaa5 1814
97eb6b69 1815 end_io_wq = container_of(work, struct btrfs_end_io_wq, work);
8b712842 1816 bio = end_io_wq->bio;
ce9adaa5 1817
4246a0b6 1818 bio->bi_error = end_io_wq->error;
8b712842
CM		 1819 bio->bi_private = end_io_wq->private;
1820 bio->bi_end_io = end_io_wq->end_io;
97eb6b69 1821 kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq);
4246a0b6 1822 bio_endio(bio);
44b8bd7e
CM		 1823}
1824
a74a4b97
CM		 1825static int cleaner_kthread(void *arg)
1826{
1827 struct btrfs_root *root = arg;
d0278245 1828 int again;
da288d28 1829 struct btrfs_trans_handle *trans;
a74a4b97
CM		 1830
1831 do {
d0278245 1832 again = 0;
a74a4b97 1833
d0278245 1834 /* Make the cleaner go to sleep early. */
babbf170 1835 if (btrfs_need_cleaner_sleep(root))
d0278245
MX		 1836 goto sleep;
1837
90c711ab
ZB		 1838 /*
1839 * Do not do anything if we might cause open_ctree() to block
1840 * before we have finished mounting the filesystem.
1841 */
afcdd129 1842 if (!test_bit(BTRFS_FS_OPEN, &root->fs_info->flags))
90c711ab
ZB		 1843 goto sleep;
1844
d0278245
MX		 1845 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1846 goto sleep;
1847
dc7f370c
MX		 1848 /*
1849 * Avoid the problem that we change the status of the fs
1850 * during the above check and trylock.
1851 */
babbf170 1852 if (btrfs_need_cleaner_sleep(root)) {
dc7f370c
MX		 1853 mutex_unlock(&root->fs_info->cleaner_mutex);
1854 goto sleep;
76dda93c 1855 }
a74a4b97 1856
c2d6cb16 1857 mutex_lock(&root->fs_info->cleaner_delayed_iput_mutex);
d0278245 1858 btrfs_run_delayed_iputs(root);
c2d6cb16
FM		 1859 mutex_unlock(&root->fs_info->cleaner_delayed_iput_mutex);
1860
d0278245
MX		 1861 again = btrfs_clean_one_deleted_snapshot(root);
1862 mutex_unlock(&root->fs_info->cleaner_mutex);
1863
1864 /*
05323cd1
MX		 1865 * The defragger has dealt with the R/O remount and umount,
1866 * needn't do anything special here.
d0278245
MX		 1867 */
1868 btrfs_run_defrag_inodes(root->fs_info);
67c5e7d4
FM		 1869
1870 /*
1871 * Acquires fs_info->delete_unused_bgs_mutex to avoid racing
1872 * with relocation (btrfs_relocate_chunk) and relocation
1873 * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group)
1874 * after acquiring fs_info->delete_unused_bgs_mutex. So we
1875 * can't hold, nor need to, fs_info->cleaner_mutex when deleting
1876 * unused block groups.
1877 */
1878 btrfs_delete_unused_bgs(root->fs_info);
d0278245 1879sleep:
838fe188 1880 if (!again) {
a74a4b97 1881 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1882 if (!kthread_should_stop())
1883 schedule();
a74a4b97
CM		 1884 __set_current_state(TASK_RUNNING);
1885 }
1886 } while (!kthread_should_stop());
da288d28
FM		 1887
1888 /*
1889 * Transaction kthread is stopped before us and wakes us up.
1890 * However we might have started a new transaction and COWed some
1891 * tree blocks when deleting unused block groups for example. So
1892 * make sure we commit the transaction we started to have a clean
1893 * shutdown when evicting the btree inode - if it has dirty pages
1894 * when we do the final iput() on it, eviction will trigger a
1895 * writeback for it which will fail with null pointer dereferences
1896 * since work queues and other resources were already released and
1897 * destroyed by the time the iput/eviction/writeback is made.
1898 */
1899 trans = btrfs_attach_transaction(root);
1900 if (IS_ERR(trans)) {
1901 if (PTR_ERR(trans) != -ENOENT)
1902 btrfs_err(root->fs_info,
1903 "cleaner transaction attach returned %ld",
1904 PTR_ERR(trans));
1905 } else {
1906 int ret;
1907
1908 ret = btrfs_commit_transaction(trans, root);
1909 if (ret)
1910 btrfs_err(root->fs_info,
1911 "cleaner open transaction commit returned %d",
1912 ret);
1913 }
1914
a74a4b97
CM		 1915 return 0;
1916}
1917
1918static int transaction_kthread(void *arg)
1919{
1920 struct btrfs_root *root = arg;
1921 struct btrfs_trans_handle *trans;
1922 struct btrfs_transaction *cur;
8929ecfa 1923 u64 transid;
a74a4b97
CM		 1924 unsigned long now;
1925 unsigned long delay;
914b2007 1926 bool cannot_commit;
a74a4b97
CM		 1927
1928 do {
914b2007 1929 cannot_commit = false;
8b87dc17 1930 delay = HZ * root->fs_info->commit_interval;
a74a4b97
CM		 1931 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1932
a4abeea4 1933 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1934 cur = root->fs_info->running_transaction;
1935 if (!cur) {
a4abeea4 1936 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1937 goto sleep;
1938 }
31153d81 1939
a74a4b97 1940 now = get_seconds();
4a9d8bde 1941 if (cur->state < TRANS_STATE_BLOCKED &&
8b87dc17
DS		 1942 (now < cur->start_time ||
1943 now - cur->start_time < root->fs_info->commit_interval)) {
a4abeea4 1944 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1945 delay = HZ * 5;
1946 goto sleep;
1947 }
8929ecfa 1948 transid = cur->transid;
a4abeea4 1949 spin_unlock(&root->fs_info->trans_lock);
56bec294 1950
79787eaa 1951 /* If the file system is aborted, this will always fail. */
354aa0fb 1952 trans = btrfs_attach_transaction(root);
914b2007 1953 if (IS_ERR(trans)) {
354aa0fb
MX		 1954 if (PTR_ERR(trans) != -ENOENT)
1955 cannot_commit = true;
79787eaa 1956 goto sleep;
914b2007 1957 }
8929ecfa 1958 if (transid == trans->transid) {
79787eaa 1959 btrfs_commit_transaction(trans, root);
8929ecfa
YZ		 1960 } else {
1961 btrfs_end_transaction(trans, root);
1962 }
a74a4b97
CM		 1963sleep:
1964 wake_up_process(root->fs_info->cleaner_kthread);
1965 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1966
4e121c06
JB		 1967 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1968 &root->fs_info->fs_state)))
1969 btrfs_cleanup_transaction(root);
ce63f891
JK		 1970 set_current_state(TASK_INTERRUPTIBLE);
1971 if (!kthread_should_stop() &&
1972 (!btrfs_transaction_blocked(root->fs_info) ||
1973 cannot_commit))
1974 schedule_timeout(delay);
1975 __set_current_state(TASK_RUNNING);
a74a4b97
CM		 1976 } while (!kthread_should_stop());
1977 return 0;
1978}
1979
af31f5e5
CM		 1980/*
1981 * this will find the highest generation in the array of
1982 * root backups. The index of the highest array is returned,
1983 * or -1 if we can't find anything.
1984 *
1985 * We check to make sure the array is valid by comparing the
1986 * generation of the latest root in the array with the generation
1987 * in the super block. If they don't match we pitch it.
1988 */
1989static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1990{
1991 u64 cur;
1992 int newest_index = -1;
1993 struct btrfs_root_backup *root_backup;
1994 int i;
1995
1996 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1997 root_backup = info->super_copy->super_roots + i;
1998 cur = btrfs_backup_tree_root_gen(root_backup);
1999 if (cur == newest_gen)
2000 newest_index = i;
2001 }
2002
2003 /* check to see if we actually wrapped around */
2004 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
2005 root_backup = info->super_copy->super_roots;
2006 cur = btrfs_backup_tree_root_gen(root_backup);
2007 if (cur == newest_gen)
2008 newest_index = 0;
2009 }
2010 return newest_index;
2011}
2012
2013
2014/*
2015 * find the oldest backup so we know where to store new entries
2016 * in the backup array. This will set the backup_root_index
2017 * field in the fs_info struct
2018 */
2019static void find_oldest_super_backup(struct btrfs_fs_info *info,
2020 u64 newest_gen)
2021{
2022 int newest_index = -1;
2023
2024 newest_index = find_newest_super_backup(info, newest_gen);
2025 /* if there was garbage in there, just move along */
2026 if (newest_index == -1) {
2027 info->backup_root_index = 0;
2028 } else {
2029 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
2030 }
2031}
2032
2033/*
2034 * copy all the root pointers into the super backup array.
2035 * this will bump the backup pointer by one when it is
2036 * done
2037 */
2038static void backup_super_roots(struct btrfs_fs_info *info)
2039{
2040 int next_backup;
2041 struct btrfs_root_backup *root_backup;
2042 int last_backup;
2043
2044 next_backup = info->backup_root_index;
2045 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
2046 BTRFS_NUM_BACKUP_ROOTS;
2047
2048 /*
2049 * just overwrite the last backup if we're at the same generation
2050 * this happens only at umount
2051 */
2052 root_backup = info->super_for_commit->super_roots + last_backup;
2053 if (btrfs_backup_tree_root_gen(root_backup) ==
2054 btrfs_header_generation(info->tree_root->node))
2055 next_backup = last_backup;
2056
2057 root_backup = info->super_for_commit->super_roots + next_backup;
2058
2059 /*
2060 * make sure all of our padding and empty slots get zero filled
2061 * regardless of which ones we use today
2062 */
2063 memset(root_backup, 0, sizeof(*root_backup));
2064
2065 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
2066
2067 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
2068 btrfs_set_backup_tree_root_gen(root_backup,
2069 btrfs_header_generation(info->tree_root->node));
2070
2071 btrfs_set_backup_tree_root_level(root_backup,
2072 btrfs_header_level(info->tree_root->node));
2073
2074 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
2075 btrfs_set_backup_chunk_root_gen(root_backup,
2076 btrfs_header_generation(info->chunk_root->node));
2077 btrfs_set_backup_chunk_root_level(root_backup,
2078 btrfs_header_level(info->chunk_root->node));
2079
2080 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
2081 btrfs_set_backup_extent_root_gen(root_backup,
2082 btrfs_header_generation(info->extent_root->node));
2083 btrfs_set_backup_extent_root_level(root_backup,
2084 btrfs_header_level(info->extent_root->node));
2085
7c7e82a7
CM		 2086 /*
2087 * we might commit during log recovery, which happens before we set
2088 * the fs_root. Make sure it is valid before we fill it in.
2089 */
2090 if (info->fs_root && info->fs_root->node) {
2091 btrfs_set_backup_fs_root(root_backup,
2092 info->fs_root->node->start);
2093 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 2094 btrfs_header_generation(info->fs_root->node));
7c7e82a7 2095 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 2096 btrfs_header_level(info->fs_root->node));
7c7e82a7 2097 }
af31f5e5
CM		 2098
2099 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
2100 btrfs_set_backup_dev_root_gen(root_backup,
2101 btrfs_header_generation(info->dev_root->node));
2102 btrfs_set_backup_dev_root_level(root_backup,
2103 btrfs_header_level(info->dev_root->node));
2104
2105 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
2106 btrfs_set_backup_csum_root_gen(root_backup,
2107 btrfs_header_generation(info->csum_root->node));
2108 btrfs_set_backup_csum_root_level(root_backup,
2109 btrfs_header_level(info->csum_root->node));
2110
2111 btrfs_set_backup_total_bytes(root_backup,
2112 btrfs_super_total_bytes(info->super_copy));
2113 btrfs_set_backup_bytes_used(root_backup,
2114 btrfs_super_bytes_used(info->super_copy));
2115 btrfs_set_backup_num_devices(root_backup,
2116 btrfs_super_num_devices(info->super_copy));
2117
2118 /*
2119 * if we don't copy this out to the super_copy, it won't get remembered
2120 * for the next commit
2121 */
2122 memcpy(&info->super_copy->super_roots,
2123 &info->super_for_commit->super_roots,
2124 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
2125}
2126
2127/*
2128 * this copies info out of the root backup array and back into
2129 * the in-memory super block. It is meant to help iterate through
2130 * the array, so you send it the number of backups you've already
2131 * tried and the last backup index you used.
2132 *
2133 * this returns -1 when it has tried all the backups
2134 */
2135static noinline int next_root_backup(struct btrfs_fs_info *info,
2136 struct btrfs_super_block *super,
2137 int *num_backups_tried, int *backup_index)
2138{
2139 struct btrfs_root_backup *root_backup;
2140 int newest = *backup_index;
2141
2142 if (*num_backups_tried == 0) {
2143 u64 gen = btrfs_super_generation(super);
2144
2145 newest = find_newest_super_backup(info, gen);
2146 if (newest == -1)
2147 return -1;
2148
2149 *backup_index = newest;
2150 *num_backups_tried = 1;
2151 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2152 /* we've tried all the backups, all done */
2153 return -1;
2154 } else {
2155 /* jump to the next oldest backup */
2156 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2157 BTRFS_NUM_BACKUP_ROOTS;
2158 *backup_index = newest;
2159 *num_backups_tried += 1;
2160 }
2161 root_backup = super->super_roots + newest;
2162
2163 btrfs_set_super_generation(super,
2164 btrfs_backup_tree_root_gen(root_backup));
2165 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2166 btrfs_set_super_root_level(super,
2167 btrfs_backup_tree_root_level(root_backup));
2168 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2169
2170 /*
2171 * fixme: the total bytes and num_devices need to match or we should
2172 * need a fsck
2173 */
2174 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2175 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2176 return 0;
2177}
2178
7abadb64
LB		 2179/* helper to cleanup workers */
2180static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2181{
dc6e3209 2182 btrfs_destroy_workqueue(fs_info->fixup_workers);
afe3d242 2183 btrfs_destroy_workqueue(fs_info->delalloc_workers);
5cdc7ad3 2184 btrfs_destroy_workqueue(fs_info->workers);
fccb5d86
QW		 2185 btrfs_destroy_workqueue(fs_info->endio_workers);
2186 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2187 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
8b110e39 2188 btrfs_destroy_workqueue(fs_info->endio_repair_workers);
d05a33ac 2189 btrfs_destroy_workqueue(fs_info->rmw_workers);
fccb5d86
QW		 2190 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2191 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2192 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
a8c93d4e 2193 btrfs_destroy_workqueue(fs_info->submit_workers);
5b3bc44e 2194 btrfs_destroy_workqueue(fs_info->delayed_workers);
e66f0bb1 2195 btrfs_destroy_workqueue(fs_info->caching_workers);
736cfa15 2196 btrfs_destroy_workqueue(fs_info->readahead_workers);
a44903ab 2197 btrfs_destroy_workqueue(fs_info->flush_workers);
fc97fab0 2198 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
a79b7d4b 2199 btrfs_destroy_workqueue(fs_info->extent_workers);
7abadb64
LB		 2200}
2201
2e9f5954
R		 2202static void free_root_extent_buffers(struct btrfs_root *root)
2203{
2204 if (root) {
2205 free_extent_buffer(root->node);
2206 free_extent_buffer(root->commit_root);
2207 root->node = NULL;
2208 root->commit_root = NULL;
2209 }
2210}
2211
af31f5e5
CM		 2212/* helper to cleanup tree roots */
2213static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2214{
2e9f5954 2215 free_root_extent_buffers(info->tree_root);
655b09fe 2216
2e9f5954
R		 2217 free_root_extent_buffers(info->dev_root);
2218 free_root_extent_buffers(info->extent_root);
2219 free_root_extent_buffers(info->csum_root);
2220 free_root_extent_buffers(info->quota_root);
2221 free_root_extent_buffers(info->uuid_root);
2222 if (chunk_root)
2223 free_root_extent_buffers(info->chunk_root);
70f6d82e 2224 free_root_extent_buffers(info->free_space_root);
af31f5e5
CM		 2225}
2226
faa2dbf0 2227void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
171f6537
JB		 2228{
2229 int ret;
2230 struct btrfs_root *gang[8];
2231 int i;
2232
2233 while (!list_empty(&fs_info->dead_roots)) {
2234 gang[0] = list_entry(fs_info->dead_roots.next,
2235 struct btrfs_root, root_list);
2236 list_del(&gang[0]->root_list);
2237
27cdeb70 2238 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
cb517eab 2239 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
171f6537
JB		 2240 } else {
2241 free_extent_buffer(gang[0]->node);
2242 free_extent_buffer(gang[0]->commit_root);
b0feb9d9 2243 btrfs_put_fs_root(gang[0]);
171f6537
JB		 2244 }
2245 }
2246
2247 while (1) {
2248 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2249 (void **)gang, 0,
2250 ARRAY_SIZE(gang));
2251 if (!ret)
2252 break;
2253 for (i = 0; i < ret; i++)
cb517eab 2254 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
171f6537 2255 }
1a4319cc
LB		 2256
2257 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2258 btrfs_free_log_root_tree(NULL, fs_info);
2259 btrfs_destroy_pinned_extent(fs_info->tree_root,
2260 fs_info->pinned_extents);
2261 }
171f6537 2262}
af31f5e5 2263
638aa7ed
ES		 2264static void btrfs_init_scrub(struct btrfs_fs_info *fs_info)
2265{
2266 mutex_init(&fs_info->scrub_lock);
2267 atomic_set(&fs_info->scrubs_running, 0);
2268 atomic_set(&fs_info->scrub_pause_req, 0);
2269 atomic_set(&fs_info->scrubs_paused, 0);
2270 atomic_set(&fs_info->scrub_cancel_req, 0);
2271 init_waitqueue_head(&fs_info->scrub_pause_wait);
2272 fs_info->scrub_workers_refcnt = 0;
2273}
2274
779a65a4
ES		 2275static void btrfs_init_balance(struct btrfs_fs_info *fs_info)
2276{
2277 spin_lock_init(&fs_info->balance_lock);
2278 mutex_init(&fs_info->balance_mutex);
2279 atomic_set(&fs_info->balance_running, 0);
2280 atomic_set(&fs_info->balance_pause_req, 0);
2281 atomic_set(&fs_info->balance_cancel_req, 0);
2282 fs_info->balance_ctl = NULL;
2283 init_waitqueue_head(&fs_info->balance_wait_q);
2284}
2285
f37938e0
ES		 2286static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info,
2287 struct btrfs_root *tree_root)
2288{
2289 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
2290 set_nlink(fs_info->btree_inode, 1);
2291 /*
2292 * we set the i_size on the btree inode to the max possible int.
2293 * the real end of the address space is determined by all of
2294 * the devices in the system
2295 */
2296 fs_info->btree_inode->i_size = OFFSET_MAX;
2297 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
f37938e0
ES		 2298
2299 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
2300 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
2301 fs_info->btree_inode->i_mapping);
2302 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
2303 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
2304
2305 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2306
2307 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2308 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2309 sizeof(struct btrfs_key));
2310 set_bit(BTRFS_INODE_DUMMY,
2311 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
2312 btrfs_insert_inode_hash(fs_info->btree_inode);
2313}
2314
ad618368
ES		 2315static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info)
2316{
2317 fs_info->dev_replace.lock_owner = 0;
2318 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2319 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
73beece9
LB		 2320 rwlock_init(&fs_info->dev_replace.lock);
2321 atomic_set(&fs_info->dev_replace.read_locks, 0);
2322 atomic_set(&fs_info->dev_replace.blocking_readers, 0);
ad618368 2323 init_waitqueue_head(&fs_info->replace_wait);
73beece9 2324 init_waitqueue_head(&fs_info->dev_replace.read_lock_wq);
ad618368
ES		 2325}
2326
f9e92e40
ES		 2327static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info)
2328{
2329 spin_lock_init(&fs_info->qgroup_lock);
2330 mutex_init(&fs_info->qgroup_ioctl_lock);
2331 fs_info->qgroup_tree = RB_ROOT;
2332 fs_info->qgroup_op_tree = RB_ROOT;
2333 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2334 fs_info->qgroup_seq = 1;
f9e92e40 2335 fs_info->qgroup_ulist = NULL;
d2c609b8 2336 fs_info->qgroup_rescan_running = false;
f9e92e40
ES		 2337 mutex_init(&fs_info->qgroup_rescan_lock);
2338}
2339
2a458198
ES		 2340static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info,
2341 struct btrfs_fs_devices *fs_devices)
2342{
2343 int max_active = fs_info->thread_pool_size;
6f011058 2344 unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
2a458198
ES		 2345
2346 fs_info->workers =
cb001095
JM		 2347 btrfs_alloc_workqueue(fs_info, "worker",
2348 flags | WQ_HIGHPRI, max_active, 16);
2a458198
ES		 2349
2350 fs_info->delalloc_workers =
cb001095
JM		 2351 btrfs_alloc_workqueue(fs_info, "delalloc",
2352 flags, max_active, 2);
2a458198
ES		 2353
2354 fs_info->flush_workers =
cb001095
JM		 2355 btrfs_alloc_workqueue(fs_info, "flush_delalloc",
2356 flags, max_active, 0);
2a458198
ES		 2357
2358 fs_info->caching_workers =
cb001095 2359 btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0);
2a458198
ES		 2360
2361 /*
2362 * a higher idle thresh on the submit workers makes it much more
2363 * likely that bios will be send down in a sane order to the
2364 * devices
2365 */
2366 fs_info->submit_workers =
cb001095 2367 btrfs_alloc_workqueue(fs_info, "submit", flags,
2a458198
ES		 2368 min_t(u64, fs_devices->num_devices,
2369 max_active), 64);
2370
2371 fs_info->fixup_workers =
cb001095 2372 btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0);
2a458198
ES		 2373
2374 /*
2375 * endios are largely parallel and should have a very
2376 * low idle thresh
2377 */
2378 fs_info->endio_workers =
cb001095 2379 btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4);
2a458198 2380 fs_info->endio_meta_workers =
cb001095
JM		 2381 btrfs_alloc_workqueue(fs_info, "endio-meta", flags,
2382 max_active, 4);
2a458198 2383 fs_info->endio_meta_write_workers =
cb001095
JM		 2384 btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags,
2385 max_active, 2);
2a458198 2386 fs_info->endio_raid56_workers =
cb001095
JM		 2387 btrfs_alloc_workqueue(fs_info, "endio-raid56", flags,
2388 max_active, 4);
2a458198 2389 fs_info->endio_repair_workers =
cb001095 2390 btrfs_alloc_workqueue(fs_info, "endio-repair", flags, 1, 0);
2a458198 2391 fs_info->rmw_workers =
cb001095 2392 btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2);
2a458198 2393 fs_info->endio_write_workers =
cb001095
JM		 2394 btrfs_alloc_workqueue(fs_info, "endio-write", flags,
2395 max_active, 2);
2a458198 2396 fs_info->endio_freespace_worker =
cb001095
JM		 2397 btrfs_alloc_workqueue(fs_info, "freespace-write", flags,
2398 max_active, 0);
2a458198 2399 fs_info->delayed_workers =
cb001095
JM		 2400 btrfs_alloc_workqueue(fs_info, "delayed-meta", flags,
2401 max_active, 0);
2a458198 2402 fs_info->readahead_workers =
cb001095
JM		 2403 btrfs_alloc_workqueue(fs_info, "readahead", flags,
2404 max_active, 2);
2a458198 2405 fs_info->qgroup_rescan_workers =
cb001095 2406 btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0);
2a458198 2407 fs_info->extent_workers =
cb001095 2408 btrfs_alloc_workqueue(fs_info, "extent-refs", flags,
2a458198
ES		 2409 min_t(u64, fs_devices->num_devices,
2410 max_active), 8);
2411
2412 if (!(fs_info->workers && fs_info->delalloc_workers &&
2413 fs_info->submit_workers && fs_info->flush_workers &&
2414 fs_info->endio_workers && fs_info->endio_meta_workers &&
2415 fs_info->endio_meta_write_workers &&
2416 fs_info->endio_repair_workers &&
2417 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
2418 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
2419 fs_info->caching_workers && fs_info->readahead_workers &&
2420 fs_info->fixup_workers && fs_info->delayed_workers &&
2421 fs_info->extent_workers &&
2422 fs_info->qgroup_rescan_workers)) {
2423 return -ENOMEM;
2424 }
2425
2426 return 0;
2427}
2428
63443bf5
ES		 2429static int btrfs_replay_log(struct btrfs_fs_info *fs_info,
2430 struct btrfs_fs_devices *fs_devices)
2431{
2432 int ret;
2433 struct btrfs_root *tree_root = fs_info->tree_root;
2434 struct btrfs_root *log_tree_root;
2435 struct btrfs_super_block *disk_super = fs_info->super_copy;
2436 u64 bytenr = btrfs_super_log_root(disk_super);
2437
2438 if (fs_devices->rw_devices == 0) {
f14d104d 2439 btrfs_warn(fs_info, "log replay required on RO media");
63443bf5
ES		 2440 return -EIO;
2441 }
2442
74e4d827 2443 log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
63443bf5
ES		 2444 if (!log_tree_root)
2445 return -ENOMEM;
2446
2447 __setup_root(tree_root->nodesize, tree_root->sectorsize,
2448 tree_root->stripesize, log_tree_root, fs_info,
2449 BTRFS_TREE_LOG_OBJECTID);
2450
2451 log_tree_root->node = read_tree_block(tree_root, bytenr,
2452 fs_info->generation + 1);
64c043de 2453 if (IS_ERR(log_tree_root->node)) {
f14d104d 2454 btrfs_warn(fs_info, "failed to read log tree");
0eeff236 2455 ret = PTR_ERR(log_tree_root->node);
64c043de 2456 kfree(log_tree_root);
0eeff236 2457 return ret;
64c043de 2458 } else if (!extent_buffer_uptodate(log_tree_root->node)) {
f14d104d 2459 btrfs_err(fs_info, "failed to read log tree");
63443bf5
ES		 2460 free_extent_buffer(log_tree_root->node);
2461 kfree(log_tree_root);
2462 return -EIO;
2463 }
2464 /* returns with log_tree_root freed on success */
2465 ret = btrfs_recover_log_trees(log_tree_root);
2466 if (ret) {
34d97007 2467 btrfs_handle_fs_error(tree_root->fs_info, ret,
63443bf5
ES		 2468 "Failed to recover log tree");
2469 free_extent_buffer(log_tree_root->node);
2470 kfree(log_tree_root);
2471 return ret;
2472 }
2473
2474 if (fs_info->sb->s_flags & MS_RDONLY) {
2475 ret = btrfs_commit_super(tree_root);
2476 if (ret)
2477 return ret;
2478 }
2479
2480 return 0;
2481}
2482
4bbcaa64
ES		 2483static int btrfs_read_roots(struct btrfs_fs_info *fs_info,
2484 struct btrfs_root *tree_root)
2485{
a4f3d2c4 2486 struct btrfs_root *root;
4bbcaa64
ES		 2487 struct btrfs_key location;
2488 int ret;
2489
2490 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2491 location.type = BTRFS_ROOT_ITEM_KEY;
2492 location.offset = 0;
2493
a4f3d2c4
DS		 2494 root = btrfs_read_tree_root(tree_root, &location);
2495 if (IS_ERR(root))
2496 return PTR_ERR(root);
2497 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2498 fs_info->extent_root = root;
4bbcaa64
ES		 2499
2500 location.objectid = BTRFS_DEV_TREE_OBJECTID;
a4f3d2c4
DS		 2501 root = btrfs_read_tree_root(tree_root, &location);
2502 if (IS_ERR(root))
2503 return PTR_ERR(root);
2504 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2505 fs_info->dev_root = root;
4bbcaa64
ES		 2506 btrfs_init_devices_late(fs_info);
2507
2508 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
a4f3d2c4
DS		 2509 root = btrfs_read_tree_root(tree_root, &location);
2510 if (IS_ERR(root))
2511 return PTR_ERR(root);
2512 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2513 fs_info->csum_root = root;
4bbcaa64
ES		 2514
2515 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
a4f3d2c4
DS		 2516 root = btrfs_read_tree_root(tree_root, &location);
2517 if (!IS_ERR(root)) {
2518 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
afcdd129 2519 set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags);
a4f3d2c4 2520 fs_info->quota_root = root;
4bbcaa64
ES		 2521 }
2522
2523 location.objectid = BTRFS_UUID_TREE_OBJECTID;
a4f3d2c4
DS		 2524 root = btrfs_read_tree_root(tree_root, &location);
2525 if (IS_ERR(root)) {
2526 ret = PTR_ERR(root);
4bbcaa64
ES		 2527 if (ret != -ENOENT)
2528 return ret;
2529 } else {
a4f3d2c4
DS		 2530 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2531 fs_info->uuid_root = root;
4bbcaa64
ES		 2532 }
2533
70f6d82e
OS		 2534 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
2535 location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID;
2536 root = btrfs_read_tree_root(tree_root, &location);
2537 if (IS_ERR(root))
2538 return PTR_ERR(root);
2539 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2540 fs_info->free_space_root = root;
2541 }
2542
4bbcaa64
ES		 2543 return 0;
2544}
2545
ad2b2c80
AV		 2546int open_ctree(struct super_block *sb,
2547 struct btrfs_fs_devices *fs_devices,
2548 char *options)
2e635a27 2549{
db94535d
CM		 2550 u32 sectorsize;
2551 u32 nodesize;
87ee04eb 2552 u32 stripesize;
84234f3a 2553 u64 generation;
f2b636e8 2554 u64 features;
3de4586c 2555 struct btrfs_key location;
a061fc8d 2556 struct buffer_head *bh;
4d34b278 2557 struct btrfs_super_block *disk_super;
815745cf 2558 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2559 struct btrfs_root *tree_root;
4d34b278 2560 struct btrfs_root *chunk_root;
eb60ceac 2561 int ret;
e58ca020 2562 int err = -EINVAL;
af31f5e5
CM		 2563 int num_backups_tried = 0;
2564 int backup_index = 0;
5cdc7ad3 2565 int max_active;
4543df7e 2566
74e4d827
DS		 2567 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
2568 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
cb517eab 2569 if (!tree_root || !chunk_root) {
39279cc3
CM		 2570 err = -ENOMEM;
2571 goto fail;
2572 }
76dda93c
YZ		 2573
2574 ret = init_srcu_struct(&fs_info->subvol_srcu);
2575 if (ret) {
2576 err = ret;
2577 goto fail;
2578 }
2579
2580 ret = setup_bdi(fs_info, &fs_info->bdi);
2581 if (ret) {
2582 err = ret;
2583 goto fail_srcu;
2584 }
2585
908c7f19 2586 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL);
e2d84521
MX		 2587 if (ret) {
2588 err = ret;
2589 goto fail_bdi;
2590 }
09cbfeaf 2591 fs_info->dirty_metadata_batch = PAGE_SIZE *
e2d84521
MX		 2592 (1 + ilog2(nr_cpu_ids));
2593
908c7f19 2594 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL);
963d678b
MX		 2595 if (ret) {
2596 err = ret;
2597 goto fail_dirty_metadata_bytes;
2598 }
2599
908c7f19 2600 ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL);
c404e0dc
MX		 2601 if (ret) {
2602 err = ret;
2603 goto fail_delalloc_bytes;
2604 }
2605
76dda93c
YZ		 2606 fs_info->btree_inode = new_inode(sb);
2607 if (!fs_info->btree_inode) {
2608 err = -ENOMEM;
c404e0dc 2609 goto fail_bio_counter;
76dda93c
YZ		 2610 }
2611
a6591715 2612 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2613
76dda93c 2614 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
f28491e0 2615 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
8fd17795 2616 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2617 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2618 INIT_LIST_HEAD(&fs_info->delayed_iputs);
eb73c1b7 2619 INIT_LIST_HEAD(&fs_info->delalloc_roots);
11833d66 2620 INIT_LIST_HEAD(&fs_info->caching_block_groups);
eb73c1b7 2621 spin_lock_init(&fs_info->delalloc_root_lock);
a4abeea4 2622 spin_lock_init(&fs_info->trans_lock);
76dda93c 2623 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2624 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2625 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 2626 spin_lock_init(&fs_info->free_chunk_lock);
f29021b2 2627 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2628 spin_lock_init(&fs_info->super_lock);
fcebe456 2629 spin_lock_init(&fs_info->qgroup_op_lock);
f28491e0 2630 spin_lock_init(&fs_info->buffer_lock);
47ab2a6c 2631 spin_lock_init(&fs_info->unused_bgs_lock);
f29021b2 2632 rwlock_init(&fs_info->tree_mod_log_lock);
d7c15171 2633 mutex_init(&fs_info->unused_bg_unpin_mutex);
67c5e7d4 2634 mutex_init(&fs_info->delete_unused_bgs_mutex);
7585717f 2635 mutex_init(&fs_info->reloc_mutex);
573bfb72 2636 mutex_init(&fs_info->delalloc_root_mutex);
c2d6cb16 2637 mutex_init(&fs_info->cleaner_delayed_iput_mutex);
de98ced9 2638 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2639
0b86a832 2640 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2641 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2642 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
47ab2a6c 2643 INIT_LIST_HEAD(&fs_info->unused_bgs);
0b86a832 2644 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX		 2645 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2646 BTRFS_BLOCK_RSV_GLOBAL);
2647 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2648 BTRFS_BLOCK_RSV_DELALLOC);
2649 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2650 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2651 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2652 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2653 BTRFS_BLOCK_RSV_DELOPS);
cb03c743 2654 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 2655 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 2656 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 2657 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 2658 atomic_set(&fs_info->defrag_running, 0);
fcebe456 2659 atomic_set(&fs_info->qgroup_op_seq, 0);
2fefd558 2660 atomic_set(&fs_info->reada_works_cnt, 0);
fc36ed7e 2661 atomic64_set(&fs_info->tree_mod_seq, 0);
9e7cc91a 2662 fs_info->fs_frozen = 0;
e20d96d6 2663 fs_info->sb = sb;
95ac567a 2664 fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE;
9ed74f2d 2665 fs_info->metadata_ratio = 0;
4cb5300b 2666 fs_info->defrag_inodes = RB_ROOT;
2bf64758 2667 fs_info->free_chunk_space = 0;
f29021b2 2668 fs_info->tree_mod_log = RB_ROOT;
8b87dc17 2669 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
f8c269d7 2670 fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */
90519d66 2671 /* readahead state */
d0164adc 2672 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM);
90519d66 2673 spin_lock_init(&fs_info->reada_lock);
c8b97818 2674
b34b086c
CM		 2675 fs_info->thread_pool_size = min_t(unsigned long,
2676 num_online_cpus() + 2, 8);
0afbaf8c 2677
199c2a9c
MX		 2678 INIT_LIST_HEAD(&fs_info->ordered_roots);
2679 spin_lock_init(&fs_info->ordered_root_lock);
16cdcec7 2680 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
74e4d827 2681 GFP_KERNEL);
16cdcec7
MX		 2682 if (!fs_info->delayed_root) {
2683 err = -ENOMEM;
2684 goto fail_iput;
2685 }
2686 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2687
638aa7ed 2688 btrfs_init_scrub(fs_info);
21adbd5c
SB		 2689#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2690 fs_info->check_integrity_print_mask = 0;
2691#endif
779a65a4 2692 btrfs_init_balance(fs_info);
21c7e756 2693 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
a2de733c 2694
a061fc8d
CM		 2695 sb->s_blocksize = 4096;
2696 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 2697 sb->s_bdi = &fs_info->bdi;
a061fc8d 2698
f37938e0 2699 btrfs_init_btree_inode(fs_info, tree_root);
76dda93c 2700
0f9dd46c 2701 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2702 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2703 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2704
11833d66 2705 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 2706 fs_info->btree_inode->i_mapping);
11833d66 2707 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 2708 fs_info->btree_inode->i_mapping);
11833d66 2709 fs_info->pinned_extents = &fs_info->freed_extents[0];
afcdd129 2710 set_bit(BTRFS_FS_BARRIER, &fs_info->flags);
39279cc3 2711
5a3f23d5 2712 mutex_init(&fs_info->ordered_operations_mutex);
e02119d5 2713 mutex_init(&fs_info->tree_log_mutex);
925baedd 2714 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 2715 mutex_init(&fs_info->transaction_kthread_mutex);
2716 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2717 mutex_init(&fs_info->volume_mutex);
1bbc621e 2718 mutex_init(&fs_info->ro_block_group_mutex);
9e351cc8 2719 init_rwsem(&fs_info->commit_root_sem);
c71bf099 2720 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2721 init_rwsem(&fs_info->subvol_sem);
803b2f54 2722 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
fa9c0d79 2723
ad618368 2724 btrfs_init_dev_replace_locks(fs_info);
f9e92e40 2725 btrfs_init_qgroup(fs_info);
416ac51d 2726
fa9c0d79
CM		 2727 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2728 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2729
e6dcd2dc 2730 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2731 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2732 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2733 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2734
04216820
FM		 2735 INIT_LIST_HEAD(&fs_info->pinned_chunks);
2736
53b381b3
DW		 2737 ret = btrfs_alloc_stripe_hash_table(fs_info);
2738 if (ret) {
83c8266a 2739 err = ret;
53b381b3
DW		 2740 goto fail_alloc;
2741 }
2742
707e8a07 2743 __setup_root(4096, 4096, 4096, tree_root,
2c90e5d6 2744 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2745
3c4bb26b 2746 invalidate_bdev(fs_devices->latest_bdev);
1104a885
DS		 2747
2748 /*
2749 * Read super block and check the signature bytes only
2750 */
a512bbf8 2751 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
92fc03fb
AJ		 2752 if (IS_ERR(bh)) {
2753 err = PTR_ERR(bh);
16cdcec7 2754 goto fail_alloc;
20b45077 2755 }
39279cc3 2756
1104a885
DS		 2757 /*
2758 * We want to check superblock checksum, the type is stored inside.
2759 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2760 */
2761 if (btrfs_check_super_csum(bh->b_data)) {
05135f59 2762 btrfs_err(fs_info, "superblock checksum mismatch");
1104a885 2763 err = -EINVAL;
b2acdddf 2764 brelse(bh);
1104a885
DS		 2765 goto fail_alloc;
2766 }
2767
2768 /*
2769 * super_copy is zeroed at allocation time and we never touch the
2770 * following bytes up to INFO_SIZE, the checksum is calculated from
2771 * the whole block of INFO_SIZE
2772 */
6c41761f
DS		 2773 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2774 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2775 sizeof(*fs_info->super_for_commit));
a061fc8d 2776 brelse(bh);
5f39d397 2777
6c41761f 2778 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2779
1104a885
DS		 2780 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2781 if (ret) {
05135f59 2782 btrfs_err(fs_info, "superblock contains fatal errors");
1104a885
DS		 2783 err = -EINVAL;
2784 goto fail_alloc;
2785 }
2786
6c41761f 2787 disk_super = fs_info->super_copy;
0f7d52f4 2788 if (!btrfs_super_root(disk_super))
16cdcec7 2789 goto fail_alloc;
0f7d52f4 2790
acce952b 2791 /* check FS state, whether FS is broken. */
87533c47
MX		 2792 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2793 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2794
af31f5e5
CM		 2795 /*
2796 * run through our array of backup supers and setup
2797 * our ring pointer to the oldest one
2798 */
2799 generation = btrfs_super_generation(disk_super);
2800 find_oldest_super_backup(fs_info, generation);
2801
75e7cb7f
LB		 2802 /*
2803 * In the long term, we'll store the compression type in the super
2804 * block, and it'll be used for per file compression control.
2805 */
2806 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2807
96da0919 2808 ret = btrfs_parse_options(tree_root, options, sb->s_flags);
2b82032c
YZ		 2809 if (ret) {
2810 err = ret;
16cdcec7 2811 goto fail_alloc;
2b82032c 2812 }
dfe25020 2813
f2b636e8
JB		 2814 features = btrfs_super_incompat_flags(disk_super) &
2815 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2816 if (features) {
05135f59
DS		 2817 btrfs_err(fs_info,
2818 "cannot mount because of unsupported optional features (%llx)",
2819 features);
f2b636e8 2820 err = -EINVAL;
16cdcec7 2821 goto fail_alloc;
f2b636e8
JB		 2822 }
2823
5d4f98a2 2824 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2825 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
69e380d1 2826 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2827 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
727011e0 2828
3173a18f 2829 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
05135f59 2830 btrfs_info(fs_info, "has skinny extents");
3173a18f 2831
727011e0
CM		 2832 /*
2833 * flag our filesystem as having big metadata blocks if
2834 * they are bigger than the page size
2835 */
09cbfeaf 2836 if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) {
727011e0 2837 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
05135f59
DS		 2838 btrfs_info(fs_info,
2839 "flagging fs with big metadata feature");
727011e0
CM		 2840 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2841 }
2842
bc3f116f 2843 nodesize = btrfs_super_nodesize(disk_super);
bc3f116f 2844 sectorsize = btrfs_super_sectorsize(disk_super);
b7f67055 2845 stripesize = sectorsize;
707e8a07 2846 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids));
963d678b 2847 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f
CM		 2848
2849 /*
2850 * mixed block groups end up with duplicate but slightly offset
2851 * extent buffers for the same range. It leads to corruptions
2852 */
2853 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
707e8a07 2854 (sectorsize != nodesize)) {
05135f59
DS		 2855 btrfs_err(fs_info,
2856"unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups",
2857 nodesize, sectorsize);
bc3f116f
CM		 2858 goto fail_alloc;
2859 }
2860
ceda0864
MX		 2861 /*
2862 * Needn't use the lock because there is no other task which will
2863 * update the flag.
2864 */
a6fa6fae 2865 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2866
f2b636e8
JB		 2867 features = btrfs_super_compat_ro_flags(disk_super) &
2868 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2869 if (!(sb->s_flags & MS_RDONLY) && features) {
05135f59
DS		 2870 btrfs_err(fs_info,
2871 "cannot mount read-write because of unsupported optional features (%llx)",
c1c9ff7c 2872 features);
f2b636e8 2873 err = -EINVAL;
16cdcec7 2874 goto fail_alloc;
f2b636e8 2875 }
61d92c32 2876
5cdc7ad3 2877 max_active = fs_info->thread_pool_size;
61d92c32 2878
2a458198
ES		 2879 ret = btrfs_init_workqueues(fs_info, fs_devices);
2880 if (ret) {
2881 err = ret;
0dc3b84a
JB		 2882 goto fail_sb_buffer;
2883 }
4543df7e 2884
4575c9cc 2885 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818 2886 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
09cbfeaf 2887 SZ_4M / PAGE_SIZE);
4575c9cc 2888
db94535d 2889 tree_root->nodesize = nodesize;
db94535d 2890 tree_root->sectorsize = sectorsize;
87ee04eb 2891 tree_root->stripesize = stripesize;
a061fc8d
CM		 2892
2893 sb->s_blocksize = sectorsize;
2894 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 2895
925baedd 2896 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 2897 ret = btrfs_read_sys_array(tree_root);
925baedd 2898 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2899 if (ret) {
05135f59 2900 btrfs_err(fs_info, "failed to read the system array: %d", ret);
5d4f98a2 2901 goto fail_sb_buffer;
84eed90f 2902 }
0b86a832 2903
84234f3a 2904 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832 2905
707e8a07
DS		 2906 __setup_root(nodesize, sectorsize, stripesize, chunk_root,
2907 fs_info, BTRFS_CHUNK_TREE_OBJECTID);
0b86a832
CM		 2908
2909 chunk_root->node = read_tree_block(chunk_root,
2910 btrfs_super_chunk_root(disk_super),
ce86cd59 2911 generation);
64c043de
LB		 2912 if (IS_ERR(chunk_root->node) ||
2913 !extent_buffer_uptodate(chunk_root->node)) {
05135f59 2914 btrfs_err(fs_info, "failed to read chunk root");
e5fffbac 2915 if (!IS_ERR(chunk_root->node))
2916 free_extent_buffer(chunk_root->node);
95ab1f64 2917 chunk_root->node = NULL;
af31f5e5 2918 goto fail_tree_roots;
83121942 2919 }
5d4f98a2
YZ		 2920 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2921 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2922
e17cade2 2923 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
b308bc2f 2924 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
e17cade2 2925
0b86a832 2926 ret = btrfs_read_chunk_tree(chunk_root);
2b82032c 2927 if (ret) {
05135f59 2928 btrfs_err(fs_info, "failed to read chunk tree: %d", ret);
af31f5e5 2929 goto fail_tree_roots;
2b82032c 2930 }
0b86a832 2931
8dabb742
SB		 2932 /*
2933 * keep the device that is marked to be the target device for the
2934 * dev_replace procedure
2935 */
9eaed21e 2936 btrfs_close_extra_devices(fs_devices, 0);
dfe25020 2937
a6b0d5c8 2938 if (!fs_devices->latest_bdev) {
05135f59 2939 btrfs_err(fs_info, "failed to read devices");
a6b0d5c8
CM		 2940 goto fail_tree_roots;
2941 }
2942
af31f5e5 2943retry_root_backup:
84234f3a 2944 generation = btrfs_super_generation(disk_super);
0b86a832 2945
e20d96d6 2946 tree_root->node = read_tree_block(tree_root,
db94535d 2947 btrfs_super_root(disk_super),
ce86cd59 2948 generation);
64c043de
LB		 2949 if (IS_ERR(tree_root->node) ||
2950 !extent_buffer_uptodate(tree_root->node)) {
05135f59 2951 btrfs_warn(fs_info, "failed to read tree root");
e5fffbac 2952 if (!IS_ERR(tree_root->node))
2953 free_extent_buffer(tree_root->node);
95ab1f64 2954 tree_root->node = NULL;
af31f5e5 2955 goto recovery_tree_root;
83121942 2956 }
af31f5e5 2957
5d4f98a2
YZ		 2958 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2959 tree_root->commit_root = btrfs_root_node(tree_root);
69e9c6c6 2960 btrfs_set_root_refs(&tree_root->root_item, 1);
db94535d 2961
f32e48e9
CR		 2962 mutex_lock(&tree_root->objectid_mutex);
2963 ret = btrfs_find_highest_objectid(tree_root,
2964 &tree_root->highest_objectid);
2965 if (ret) {
2966 mutex_unlock(&tree_root->objectid_mutex);
2967 goto recovery_tree_root;
2968 }
2969
2970 ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID);
2971
2972 mutex_unlock(&tree_root->objectid_mutex);
2973
4bbcaa64
ES		 2974 ret = btrfs_read_roots(fs_info, tree_root);
2975 if (ret)
af31f5e5 2976 goto recovery_tree_root;
f7a81ea4 2977
8929ecfa
YZ		 2978 fs_info->generation = generation;
2979 fs_info->last_trans_committed = generation;
8929ecfa 2980
68310a5e
ID		 2981 ret = btrfs_recover_balance(fs_info);
2982 if (ret) {
05135f59 2983 btrfs_err(fs_info, "failed to recover balance: %d", ret);
68310a5e
ID		 2984 goto fail_block_groups;
2985 }
2986
733f4fbb
SB		 2987 ret = btrfs_init_dev_stats(fs_info);
2988 if (ret) {
05135f59 2989 btrfs_err(fs_info, "failed to init dev_stats: %d", ret);
733f4fbb
SB		 2990 goto fail_block_groups;
2991 }
2992
8dabb742
SB		 2993 ret = btrfs_init_dev_replace(fs_info);
2994 if (ret) {
05135f59 2995 btrfs_err(fs_info, "failed to init dev_replace: %d", ret);
8dabb742
SB		 2996 goto fail_block_groups;
2997 }
2998
9eaed21e 2999 btrfs_close_extra_devices(fs_devices, 1);
8dabb742 3000
b7c35e81
AJ		 3001 ret = btrfs_sysfs_add_fsid(fs_devices, NULL);
3002 if (ret) {
05135f59
DS		 3003 btrfs_err(fs_info, "failed to init sysfs fsid interface: %d",
3004 ret);
b7c35e81
AJ		 3005 goto fail_block_groups;
3006 }
3007
3008 ret = btrfs_sysfs_add_device(fs_devices);
3009 if (ret) {
05135f59
DS		 3010 btrfs_err(fs_info, "failed to init sysfs device interface: %d",
3011 ret);
b7c35e81
AJ		 3012 goto fail_fsdev_sysfs;
3013 }
3014
96f3136e 3015 ret = btrfs_sysfs_add_mounted(fs_info);
c59021f8 3016 if (ret) {
05135f59 3017 btrfs_err(fs_info, "failed to init sysfs interface: %d", ret);
b7c35e81 3018 goto fail_fsdev_sysfs;
c59021f8 3019 }
3020
c59021f8 3021 ret = btrfs_init_space_info(fs_info);
3022 if (ret) {
05135f59 3023 btrfs_err(fs_info, "failed to initialize space info: %d", ret);
2365dd3c 3024 goto fail_sysfs;
c59021f8 3025 }
3026
4bbcaa64 3027 ret = btrfs_read_block_groups(fs_info->extent_root);
1b1d1f66 3028 if (ret) {
05135f59 3029 btrfs_err(fs_info, "failed to read block groups: %d", ret);
2365dd3c 3030 goto fail_sysfs;
1b1d1f66 3031 }
5af3e8cc
SB		 3032 fs_info->num_tolerated_disk_barrier_failures =
3033 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
292fd7fc
SB		 3034 if (fs_info->fs_devices->missing_devices >
3035 fs_info->num_tolerated_disk_barrier_failures &&
3036 !(sb->s_flags & MS_RDONLY)) {
05135f59
DS		 3037 btrfs_warn(fs_info,
3038"missing devices (%llu) exceeds the limit (%d), writeable mount is not allowed",
78fa1770
ZL		 3039 fs_info->fs_devices->missing_devices,
3040 fs_info->num_tolerated_disk_barrier_failures);
2365dd3c 3041 goto fail_sysfs;
292fd7fc 3042 }
9078a3e1 3043
a74a4b97
CM		 3044 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
3045 "btrfs-cleaner");
57506d50 3046 if (IS_ERR(fs_info->cleaner_kthread))
2365dd3c 3047 goto fail_sysfs;
a74a4b97
CM		 3048
3049 fs_info->transaction_kthread = kthread_run(transaction_kthread,
3050 tree_root,
3051 "btrfs-transaction");
57506d50 3052 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 3053 goto fail_cleaner;
a74a4b97 3054
3cdde224
JM		 3055 if (!btrfs_test_opt(tree_root->fs_info, SSD) &&
3056 !btrfs_test_opt(tree_root->fs_info, NOSSD) &&
c289811c 3057 !fs_info->fs_devices->rotating) {
05135f59 3058 btrfs_info(fs_info, "detected SSD devices, enabling SSD mode");
c289811c
CM		 3059 btrfs_set_opt(fs_info->mount_opt, SSD);
3060 }
3061
572d9ab7 3062 /*
01327610 3063 * Mount does not set all options immediately, we can do it now and do
572d9ab7
DS		 3064 * not have to wait for transaction commit
3065 */
3066 btrfs_apply_pending_changes(fs_info);
3818aea2 3067
21adbd5c 3068#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3cdde224 3069 if (btrfs_test_opt(tree_root->fs_info, CHECK_INTEGRITY)) {
21adbd5c 3070 ret = btrfsic_mount(tree_root, fs_devices,
3cdde224 3071 btrfs_test_opt(tree_root->fs_info,
21adbd5c
SB		 3072 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
3073 1 : 0,
3074 fs_info->check_integrity_print_mask);
3075 if (ret)
05135f59
DS		 3076 btrfs_warn(fs_info,
3077 "failed to initialize integrity check module: %d",
3078 ret);
21adbd5c
SB		 3079 }
3080#endif
bcef60f2
AJ		 3081 ret = btrfs_read_qgroup_config(fs_info);
3082 if (ret)
3083 goto fail_trans_kthread;
21adbd5c 3084
96da0919
QW		 3085 /* do not make disk changes in broken FS or nologreplay is given */
3086 if (btrfs_super_log_root(disk_super) != 0 &&
3cdde224 3087 !btrfs_test_opt(tree_root->fs_info, NOLOGREPLAY)) {
63443bf5 3088 ret = btrfs_replay_log(fs_info, fs_devices);
79787eaa 3089 if (ret) {
63443bf5 3090 err = ret;
28c16cbb 3091 goto fail_qgroup;
79787eaa 3092 }
e02119d5 3093 }
1a40e23b 3094
76dda93c 3095 ret = btrfs_find_orphan_roots(tree_root);
79787eaa 3096 if (ret)
28c16cbb 3097 goto fail_qgroup;
76dda93c 3098
7c2ca468 3099 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b 3100 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2 3101 if (ret)
28c16cbb 3102 goto fail_qgroup;
90c711ab
ZB		 3103
3104 mutex_lock(&fs_info->cleaner_mutex);
5d4f98a2 3105 ret = btrfs_recover_relocation(tree_root);
90c711ab 3106 mutex_unlock(&fs_info->cleaner_mutex);
d7ce5843 3107 if (ret < 0) {
05135f59
DS		 3108 btrfs_warn(fs_info, "failed to recover relocation: %d",
3109 ret);
d7ce5843 3110 err = -EINVAL;
bcef60f2 3111 goto fail_qgroup;
d7ce5843 3112 }
7c2ca468 3113 }
1a40e23b 3114
3de4586c
CM		 3115 location.objectid = BTRFS_FS_TREE_OBJECTID;
3116 location.type = BTRFS_ROOT_ITEM_KEY;
cb517eab 3117 location.offset = 0;
3de4586c 3118
3de4586c 3119 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
3140c9a3
DC		 3120 if (IS_ERR(fs_info->fs_root)) {
3121 err = PTR_ERR(fs_info->fs_root);
bcef60f2 3122 goto fail_qgroup;
3140c9a3 3123 }
c289811c 3124
2b6ba629
ID		 3125 if (sb->s_flags & MS_RDONLY)
3126 return 0;
59641015 3127
3cdde224 3128 if (btrfs_test_opt(tree_root->fs_info, FREE_SPACE_TREE) &&
511711af 3129 !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
05135f59 3130 btrfs_info(fs_info, "creating free space tree");
511711af
CM		 3131 ret = btrfs_create_free_space_tree(fs_info);
3132 if (ret) {
05135f59
DS		 3133 btrfs_warn(fs_info,
3134 "failed to create free space tree: %d", ret);
511711af
CM		 3135 close_ctree(tree_root);
3136 return ret;
3137 }
3138 }
3139
2b6ba629
ID		 3140 down_read(&fs_info->cleanup_work_sem);
3141 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
3142 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 3143 up_read(&fs_info->cleanup_work_sem);
2b6ba629
ID		 3144 close_ctree(tree_root);
3145 return ret;
3146 }
3147 up_read(&fs_info->cleanup_work_sem);
59641015 3148
2b6ba629
ID		 3149 ret = btrfs_resume_balance_async(fs_info);
3150 if (ret) {
05135f59 3151 btrfs_warn(fs_info, "failed to resume balance: %d", ret);
2b6ba629
ID		 3152 close_ctree(tree_root);
3153 return ret;
e3acc2a6
JB		 3154 }
3155
8dabb742
SB		 3156 ret = btrfs_resume_dev_replace_async(fs_info);
3157 if (ret) {
05135f59 3158 btrfs_warn(fs_info, "failed to resume device replace: %d", ret);
8dabb742
SB		 3159 close_ctree(tree_root);
3160 return ret;
3161 }
3162
b382a324
JS		 3163 btrfs_qgroup_rescan_resume(fs_info);
3164
3cdde224 3165 if (btrfs_test_opt(tree_root->fs_info, CLEAR_CACHE) &&
70f6d82e 3166 btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
05135f59 3167 btrfs_info(fs_info, "clearing free space tree");
70f6d82e
OS		 3168 ret = btrfs_clear_free_space_tree(fs_info);
3169 if (ret) {
05135f59
DS		 3170 btrfs_warn(fs_info,
3171 "failed to clear free space tree: %d", ret);
70f6d82e
OS		 3172 close_ctree(tree_root);
3173 return ret;
3174 }
3175 }
3176
4bbcaa64 3177 if (!fs_info->uuid_root) {
05135f59 3178 btrfs_info(fs_info, "creating UUID tree");
f7a81ea4
SB		 3179 ret = btrfs_create_uuid_tree(fs_info);
3180 if (ret) {
05135f59
DS		 3181 btrfs_warn(fs_info,
3182 "failed to create the UUID tree: %d", ret);
f7a81ea4
SB		 3183 close_ctree(tree_root);
3184 return ret;
3185 }
3cdde224 3186 } else if (btrfs_test_opt(tree_root->fs_info, RESCAN_UUID_TREE) ||
4bbcaa64
ES		 3187 fs_info->generation !=
3188 btrfs_super_uuid_tree_generation(disk_super)) {
05135f59 3189 btrfs_info(fs_info, "checking UUID tree");
70f80175
SB		 3190 ret = btrfs_check_uuid_tree(fs_info);
3191 if (ret) {
05135f59
DS		 3192 btrfs_warn(fs_info,
3193 "failed to check the UUID tree: %d", ret);
70f80175
SB		 3194 close_ctree(tree_root);
3195 return ret;
3196 }
3197 } else {
afcdd129 3198 set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags);
f7a81ea4 3199 }
afcdd129 3200 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
47ab2a6c 3201
8dcddfa0
QW		 3202 /*
3203 * backuproot only affect mount behavior, and if open_ctree succeeded,
3204 * no need to keep the flag
3205 */
3206 btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
3207
ad2b2c80 3208 return 0;
39279cc3 3209
bcef60f2
AJ		 3210fail_qgroup:
3211 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM		 3212fail_trans_kthread:
3213 kthread_stop(fs_info->transaction_kthread);
54067ae9 3214 btrfs_cleanup_transaction(fs_info->tree_root);
faa2dbf0 3215 btrfs_free_fs_roots(fs_info);
3f157a2f 3216fail_cleaner:
a74a4b97 3217 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 3218
3219 /*
3220 * make sure we're done with the btree inode before we stop our
3221 * kthreads
3222 */
3223 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 3224
2365dd3c 3225fail_sysfs:
6618a59b 3226 btrfs_sysfs_remove_mounted(fs_info);
2365dd3c 3227
b7c35e81
AJ		 3228fail_fsdev_sysfs:
3229 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
3230
1b1d1f66 3231fail_block_groups:
54067ae9 3232 btrfs_put_block_group_cache(fs_info);
1b1d1f66 3233 btrfs_free_block_groups(fs_info);
af31f5e5
CM		 3234
3235fail_tree_roots:
3236 free_root_pointers(fs_info, 1);
2b8195bb 3237 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 3238
39279cc3 3239fail_sb_buffer:
7abadb64 3240 btrfs_stop_all_workers(fs_info);
16cdcec7 3241fail_alloc:
4543df7e 3242fail_iput:
586e46e2
ID		 3243 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3244
4543df7e 3245 iput(fs_info->btree_inode);
c404e0dc
MX		 3246fail_bio_counter:
3247 percpu_counter_destroy(&fs_info->bio_counter);
963d678b
MX		 3248fail_delalloc_bytes:
3249 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX		 3250fail_dirty_metadata_bytes:
3251 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
ad081f14 3252fail_bdi:
7e662854 3253 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 3254fail_srcu:
3255 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 3256fail:
53b381b3 3257 btrfs_free_stripe_hash_table(fs_info);
586e46e2 3258 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 3259 return err;
af31f5e5
CM		 3260
3261recovery_tree_root:
3cdde224 3262 if (!btrfs_test_opt(tree_root->fs_info, USEBACKUPROOT))
af31f5e5
CM		 3263 goto fail_tree_roots;
3264
3265 free_root_pointers(fs_info, 0);
3266
3267 /* don't use the log in recovery mode, it won't be valid */
3268 btrfs_set_super_log_root(disk_super, 0);
3269
3270 /* we can't trust the free space cache either */
3271 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3272
3273 ret = next_root_backup(fs_info, fs_info->super_copy,
3274 &num_backups_tried, &backup_index);
3275 if (ret == -1)
3276 goto fail_block_groups;
3277 goto retry_root_backup;
eb60ceac
CM		 3278}
3279
f2984462
CM		 3280static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3281{
f2984462
CM		 3282 if (uptodate) {
3283 set_buffer_uptodate(bh);
3284 } else {
442a4f63
SB		 3285 struct btrfs_device *device = (struct btrfs_device *)
3286 bh->b_private;
3287
b14af3b4
DS		 3288 btrfs_warn_rl_in_rcu(device->dev_root->fs_info,
3289 "lost page write due to IO error on %s",
606686ee 3290 rcu_str_deref(device->name));
01327610 3291 /* note, we don't set_buffer_write_io_error because we have
1259ab75
CM		 3292 * our own ways of dealing with the IO errors
3293 */
f2984462 3294 clear_buffer_uptodate(bh);
442a4f63 3295 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM		 3296 }
3297 unlock_buffer(bh);
3298 put_bh(bh);
3299}
3300
29c36d72
AJ		 3301int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num,
3302 struct buffer_head **bh_ret)
3303{
3304 struct buffer_head *bh;
3305 struct btrfs_super_block *super;
3306 u64 bytenr;
3307
3308 bytenr = btrfs_sb_offset(copy_num);
3309 if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode))
3310 return -EINVAL;
3311
3312 bh = __bread(bdev, bytenr / 4096, BTRFS_SUPER_INFO_SIZE);
3313 /*
3314 * If we fail to read from the underlying devices, as of now
3315 * the best option we have is to mark it EIO.
3316 */
3317 if (!bh)
3318 return -EIO;
3319
3320 super = (struct btrfs_super_block *)bh->b_data;
3321 if (btrfs_super_bytenr(super) != bytenr ||
3322 btrfs_super_magic(super) != BTRFS_MAGIC) {
3323 brelse(bh);
3324 return -EINVAL;
3325 }
3326
3327 *bh_ret = bh;
3328 return 0;
3329}
3330
3331
a512bbf8
YZ		 3332struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3333{
3334 struct buffer_head *bh;
3335 struct buffer_head *latest = NULL;
3336 struct btrfs_super_block *super;
3337 int i;
3338 u64 transid = 0;
92fc03fb 3339 int ret = -EINVAL;
a512bbf8
YZ		 3340
3341 /* we would like to check all the supers, but that would make
3342 * a btrfs mount succeed after a mkfs from a different FS.
3343 * So, we need to add a special mount option to scan for
3344 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3345 */
3346 for (i = 0; i < 1; i++) {
29c36d72
AJ		 3347 ret = btrfs_read_dev_one_super(bdev, i, &bh);
3348 if (ret)
a512bbf8
YZ		 3349 continue;
3350
3351 super = (struct btrfs_super_block *)bh->b_data;
a512bbf8
YZ		 3352
3353 if (!latest || btrfs_super_generation(super) > transid) {
3354 brelse(latest);
3355 latest = bh;
3356 transid = btrfs_super_generation(super);
3357 } else {
3358 brelse(bh);
3359 }
3360 }
92fc03fb
AJ		 3361
3362 if (!latest)
3363 return ERR_PTR(ret);
3364
a512bbf8
YZ		 3365 return latest;
3366}
3367
4eedeb75
HH		 3368/*
3369 * this should be called twice, once with wait == 0 and
3370 * once with wait == 1. When wait == 0 is done, all the buffer heads
3371 * we write are pinned.
3372 *
3373 * They are released when wait == 1 is done.
3374 * max_mirrors must be the same for both runs, and it indicates how
3375 * many supers on this one device should be written.
3376 *
3377 * max_mirrors == 0 means to write them all.
3378 */
a512bbf8
YZ		 3379static int write_dev_supers(struct btrfs_device *device,
3380 struct btrfs_super_block *sb,
3381 int do_barriers, int wait, int max_mirrors)
3382{
3383 struct buffer_head *bh;
3384 int i;
3385 int ret;
3386 int errors = 0;
3387 u32 crc;
3388 u64 bytenr;
a512bbf8
YZ		 3389
3390 if (max_mirrors == 0)
3391 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3392
a512bbf8
YZ		 3393 for (i = 0; i < max_mirrors; i++) {
3394 bytenr = btrfs_sb_offset(i);
935e5cc9
MX		 3395 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3396 device->commit_total_bytes)
a512bbf8
YZ		 3397 break;
3398
3399 if (wait) {
3400 bh = __find_get_block(device->bdev, bytenr / 4096,
3401 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3402 if (!bh) {
3403 errors++;
3404 continue;
3405 }
a512bbf8 3406 wait_on_buffer(bh);
4eedeb75
HH		 3407 if (!buffer_uptodate(bh))
3408 errors++;
3409
3410 /* drop our reference */
3411 brelse(bh);
3412
3413 /* drop the reference from the wait == 0 run */
3414 brelse(bh);
3415 continue;
a512bbf8
YZ		 3416 } else {
3417 btrfs_set_super_bytenr(sb, bytenr);
3418
3419 crc = ~(u32)0;
b0496686 3420 crc = btrfs_csum_data((char *)sb +
a512bbf8
YZ		 3421 BTRFS_CSUM_SIZE, crc,
3422 BTRFS_SUPER_INFO_SIZE -
3423 BTRFS_CSUM_SIZE);
3424 btrfs_csum_final(crc, sb->csum);
3425
4eedeb75
HH		 3426 /*
3427 * one reference for us, and we leave it for the
3428 * caller
3429 */
a512bbf8
YZ		 3430 bh = __getblk(device->bdev, bytenr / 4096,
3431 BTRFS_SUPER_INFO_SIZE);
634554dc 3432 if (!bh) {
f14d104d
DS		 3433 btrfs_err(device->dev_root->fs_info,
3434 "couldn't get super buffer head for bytenr %llu",
3435 bytenr);
634554dc
JB		 3436 errors++;
3437 continue;
3438 }
3439
a512bbf8
YZ		 3440 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3441
4eedeb75 3442 /* one reference for submit_bh */
a512bbf8 3443 get_bh(bh);
4eedeb75
HH		 3444
3445 set_buffer_uptodate(bh);
a512bbf8
YZ		 3446 lock_buffer(bh);
3447 bh->b_end_io = btrfs_end_buffer_write_sync;
442a4f63 3448 bh->b_private = device;
a512bbf8
YZ		 3449 }
3450
387125fc
CM		 3451 /*
3452 * we fua the first super. The others we allow
3453 * to go down lazy.
3454 */
e8117c26 3455 if (i == 0)
2a222ca9 3456 ret = btrfsic_submit_bh(REQ_OP_WRITE, WRITE_FUA, bh);
e8117c26 3457 else
2a222ca9 3458 ret = btrfsic_submit_bh(REQ_OP_WRITE, WRITE_SYNC, bh);
4eedeb75 3459 if (ret)
a512bbf8 3460 errors++;
a512bbf8
YZ		 3461 }
3462 return errors < i ? 0 : -1;
3463}
3464
387125fc
CM		 3465/*
3466 * endio for the write_dev_flush, this will wake anyone waiting
3467 * for the barrier when it is done
3468 */
4246a0b6 3469static void btrfs_end_empty_barrier(struct bio *bio)
387125fc 3470{
387125fc
CM		 3471 if (bio->bi_private)
3472 complete(bio->bi_private);
3473 bio_put(bio);
3474}
3475
3476/*
3477 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3478 * sent down. With wait == 1, it waits for the previous flush.
3479 *
3480 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3481 * capable
3482 */
3483static int write_dev_flush(struct btrfs_device *device, int wait)
3484{
3485 struct bio *bio;
3486 int ret = 0;
3487
3488 if (device->nobarriers)
3489 return 0;
3490
3491 if (wait) {
3492 bio = device->flush_bio;
3493 if (!bio)
3494 return 0;
3495
3496 wait_for_completion(&device->flush_wait);
3497
4246a0b6
CH		 3498 if (bio->bi_error) {
3499 ret = bio->bi_error;
5af3e8cc
SB		 3500 btrfs_dev_stat_inc_and_print(device,
3501 BTRFS_DEV_STAT_FLUSH_ERRS);
387125fc
CM		 3502 }
3503
3504 /* drop the reference from the wait == 0 run */
3505 bio_put(bio);
3506 device->flush_bio = NULL;
3507
3508 return ret;
3509 }
3510
3511 /*
3512 * one reference for us, and we leave it for the
3513 * caller
3514 */
9c017abc 3515 device->flush_bio = NULL;
9be3395b 3516 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
387125fc
CM		 3517 if (!bio)
3518 return -ENOMEM;
3519
3520 bio->bi_end_io = btrfs_end_empty_barrier;
3521 bio->bi_bdev = device->bdev;
37226b21 3522 bio_set_op_attrs(bio, REQ_OP_WRITE, WRITE_FLUSH);
387125fc
CM		 3523 init_completion(&device->flush_wait);
3524 bio->bi_private = &device->flush_wait;
3525 device->flush_bio = bio;
3526
3527 bio_get(bio);
4e49ea4a 3528 btrfsic_submit_bio(bio);
387125fc
CM		 3529
3530 return 0;
3531}
3532
3533/*
3534 * send an empty flush down to each device in parallel,
3535 * then wait for them
3536 */
3537static int barrier_all_devices(struct btrfs_fs_info *info)
3538{
3539 struct list_head *head;
3540 struct btrfs_device *dev;
5af3e8cc
SB		 3541 int errors_send = 0;
3542 int errors_wait = 0;
387125fc
CM		 3543 int ret;
3544
3545 /* send down all the barriers */
3546 head = &info->fs_devices->devices;
3547 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3548 if (dev->missing)
3549 continue;
387125fc 3550 if (!dev->bdev) {
5af3e8cc 3551 errors_send++;
387125fc
CM		 3552 continue;
3553 }
3554 if (!dev->in_fs_metadata || !dev->writeable)
3555 continue;
3556
3557 ret = write_dev_flush(dev, 0);
3558 if (ret)
5af3e8cc 3559 errors_send++;
387125fc
CM		 3560 }
3561
3562 /* wait for all the barriers */
3563 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3564 if (dev->missing)
3565 continue;
387125fc 3566 if (!dev->bdev) {
5af3e8cc 3567 errors_wait++;
387125fc
CM		 3568 continue;
3569 }
3570 if (!dev->in_fs_metadata || !dev->writeable)
3571 continue;
3572
3573 ret = write_dev_flush(dev, 1);
3574 if (ret)
5af3e8cc 3575 errors_wait++;
387125fc 3576 }
5af3e8cc
SB		 3577 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3578 errors_wait > info->num_tolerated_disk_barrier_failures)
387125fc
CM		 3579 return -EIO;
3580 return 0;
3581}
3582
943c6e99
ZL		 3583int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags)
3584{
8789f4fe
ZL		 3585 int raid_type;
3586 int min_tolerated = INT_MAX;
943c6e99 3587
8789f4fe
ZL		 3588 if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 ||
3589 (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE))
3590 min_tolerated = min(min_tolerated,
3591 btrfs_raid_array[BTRFS_RAID_SINGLE].
3592 tolerated_failures);
943c6e99 3593
8789f4fe
ZL		 3594 for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) {
3595 if (raid_type == BTRFS_RAID_SINGLE)
3596 continue;
3597 if (!(flags & btrfs_raid_group[raid_type]))
3598 continue;
3599 min_tolerated = min(min_tolerated,
3600 btrfs_raid_array[raid_type].
3601 tolerated_failures);
3602 }
943c6e99 3603
8789f4fe
ZL		 3604 if (min_tolerated == INT_MAX) {
3605 pr_warn("BTRFS: unknown raid flag: %llu\n", flags);
3606 min_tolerated = 0;
3607 }
3608
3609 return min_tolerated;
943c6e99
ZL		 3610}
3611
5af3e8cc
SB		 3612int btrfs_calc_num_tolerated_disk_barrier_failures(
3613 struct btrfs_fs_info *fs_info)
3614{
3615 struct btrfs_ioctl_space_info space;
3616 struct btrfs_space_info *sinfo;
3617 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3618 BTRFS_BLOCK_GROUP_SYSTEM,
3619 BTRFS_BLOCK_GROUP_METADATA,
3620 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
5af3e8cc
SB		 3621 int i;
3622 int c;
3623 int num_tolerated_disk_barrier_failures =
3624 (int)fs_info->fs_devices->num_devices;
3625
2c458045 3626 for (i = 0; i < ARRAY_SIZE(types); i++) {
5af3e8cc
SB		 3627 struct btrfs_space_info *tmp;
3628
3629 sinfo = NULL;
3630 rcu_read_lock();
3631 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3632 if (tmp->flags == types[i]) {
3633 sinfo = tmp;
3634 break;
3635 }
3636 }
3637 rcu_read_unlock();
3638
3639 if (!sinfo)
3640 continue;
3641
3642 down_read(&sinfo->groups_sem);
3643 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2c458045
ZL		 3644 u64 flags;
3645
3646 if (list_empty(&sinfo->block_groups[c]))
3647 continue;
3648
3649 btrfs_get_block_group_info(&sinfo->block_groups[c],
3650 &space);
3651 if (space.total_bytes == 0 || space.used_bytes == 0)
3652 continue;
3653 flags = space.flags;
943c6e99
ZL		 3654
3655 num_tolerated_disk_barrier_failures = min(
3656 num_tolerated_disk_barrier_failures,
3657 btrfs_get_num_tolerated_disk_barrier_failures(
3658 flags));
5af3e8cc
SB		 3659 }
3660 up_read(&sinfo->groups_sem);
3661 }
3662
3663 return num_tolerated_disk_barrier_failures;
3664}
3665
48a3b636 3666static int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 3667{
e5e9a520 3668 struct list_head *head;
f2984462 3669 struct btrfs_device *dev;
a061fc8d 3670 struct btrfs_super_block *sb;
f2984462 3671 struct btrfs_dev_item *dev_item;
f2984462
CM		 3672 int ret;
3673 int do_barriers;
a236aed1
CM		 3674 int max_errors;
3675 int total_errors = 0;
a061fc8d 3676 u64 flags;
f2984462 3677
3cdde224 3678 do_barriers = !btrfs_test_opt(root->fs_info, NOBARRIER);
af31f5e5 3679 backup_super_roots(root->fs_info);
f2984462 3680
6c41761f 3681 sb = root->fs_info->super_for_commit;
a061fc8d 3682 dev_item = &sb->dev_item;
e5e9a520 3683
174ba509 3684 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 3685 head = &root->fs_info->fs_devices->devices;
d7306801 3686 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
387125fc 3687
5af3e8cc
SB		 3688 if (do_barriers) {
3689 ret = barrier_all_devices(root->fs_info);
3690 if (ret) {
3691 mutex_unlock(
3692 &root->fs_info->fs_devices->device_list_mutex);
34d97007 3693 btrfs_handle_fs_error(root->fs_info, ret,
5af3e8cc
SB		 3694 "errors while submitting device barriers.");
3695 return ret;
3696 }
3697 }
387125fc 3698
1f78160c 3699 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3700 if (!dev->bdev) {
3701 total_errors++;
3702 continue;
3703 }
2b82032c 3704 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3705 continue;
3706
2b82032c 3707 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 3708 btrfs_set_stack_device_type(dev_item, dev->type);
3709 btrfs_set_stack_device_id(dev_item, dev->devid);
7df69d3e 3710 btrfs_set_stack_device_total_bytes(dev_item,
935e5cc9 3711 dev->commit_total_bytes);
ce7213c7
MX		 3712 btrfs_set_stack_device_bytes_used(dev_item,
3713 dev->commit_bytes_used);
a061fc8d
CM		 3714 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3715 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3716 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3717 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 3718 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 3719
a061fc8d
CM		 3720 flags = btrfs_super_flags(sb);
3721 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3722
a512bbf8 3723 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 3724 if (ret)
3725 total_errors++;
f2984462 3726 }
a236aed1 3727 if (total_errors > max_errors) {
efe120a0 3728 btrfs_err(root->fs_info, "%d errors while writing supers",
d397712b 3729 total_errors);
a724b436 3730 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
79787eaa 3731
9d565ba4 3732 /* FUA is masked off if unsupported and can't be the reason */
34d97007 3733 btrfs_handle_fs_error(root->fs_info, -EIO,
9d565ba4
SB		 3734 "%d errors while writing supers", total_errors);
3735 return -EIO;
a236aed1 3736 }
f2984462 3737
a512bbf8 3738 total_errors = 0;
1f78160c 3739 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3740 if (!dev->bdev)
3741 continue;
2b82032c 3742 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3743 continue;
3744
a512bbf8
YZ		 3745 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3746 if (ret)
3747 total_errors++;
f2984462 3748 }
174ba509 3749 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 3750 if (total_errors > max_errors) {
34d97007 3751 btrfs_handle_fs_error(root->fs_info, -EIO,
79787eaa
JM		 3752 "%d errors while writing supers", total_errors);
3753 return -EIO;
a236aed1 3754 }
f2984462
CM		 3755 return 0;
3756}
3757
a512bbf8
YZ		 3758int write_ctree_super(struct btrfs_trans_handle *trans,
3759 struct btrfs_root *root, int max_mirrors)
eb60ceac 3760{
f570e757 3761 return write_all_supers(root, max_mirrors);
cfaa7295
CM		 3762}
3763
cb517eab
MX		 3764/* Drop a fs root from the radix tree and free it. */
3765void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3766 struct btrfs_root *root)
2619ba1f 3767{
4df27c4d 3768 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 3769 radix_tree_delete(&fs_info->fs_roots_radix,
3770 (unsigned long)root->root_key.objectid);
4df27c4d 3771 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 3772
3773 if (btrfs_root_refs(&root->root_item) == 0)
3774 synchronize_srcu(&fs_info->subvol_srcu);
3775
1c1ea4f7 3776 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
3321719e 3777 btrfs_free_log(NULL, root);
1c1ea4f7
LB		 3778 if (root->reloc_root) {
3779 free_extent_buffer(root->reloc_root->node);
3780 free_extent_buffer(root->reloc_root->commit_root);
3781 btrfs_put_fs_root(root->reloc_root);
3782 root->reloc_root = NULL;
3783 }
3784 }
3321719e 3785
faa2dbf0
JB		 3786 if (root->free_ino_pinned)
3787 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3788 if (root->free_ino_ctl)
3789 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3790 free_fs_root(root);
4df27c4d
YZ		 3791}
3792
3793static void free_fs_root(struct btrfs_root *root)
3794{
57cdc8db 3795 iput(root->ino_cache_inode);
4df27c4d 3796 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
1cb048f5
FDBM		 3797 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3798 root->orphan_block_rsv = NULL;
0ee5dc67
AV		 3799 if (root->anon_dev)
3800 free_anon_bdev(root->anon_dev);
8257b2dc
MX		 3801 if (root->subv_writers)
3802 btrfs_free_subvolume_writers(root->subv_writers);
4df27c4d
YZ		 3803 free_extent_buffer(root->node);
3804 free_extent_buffer(root->commit_root);
581bb050
LZ		 3805 kfree(root->free_ino_ctl);
3806 kfree(root->free_ino_pinned);
d397712b 3807 kfree(root->name);
b0feb9d9 3808 btrfs_put_fs_root(root);
2619ba1f
CM		 3809}
3810
cb517eab
MX		 3811void btrfs_free_fs_root(struct btrfs_root *root)
3812{
3813 free_fs_root(root);
2619ba1f
CM		 3814}
3815
c146afad 3816int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3817{
c146afad
YZ		 3818 u64 root_objectid = 0;
3819 struct btrfs_root *gang[8];
65d33fd7
QW		 3820 int i = 0;
3821 int err = 0;
3822 unsigned int ret = 0;
3823 int index;
e089f05c 3824
c146afad 3825 while (1) {
65d33fd7 3826 index = srcu_read_lock(&fs_info->subvol_srcu);
c146afad
YZ		 3827 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3828 (void **)gang, root_objectid,
3829 ARRAY_SIZE(gang));
65d33fd7
QW		 3830 if (!ret) {
3831 srcu_read_unlock(&fs_info->subvol_srcu, index);
c146afad 3832 break;
65d33fd7 3833 }
5d4f98a2 3834 root_objectid = gang[ret - 1]->root_key.objectid + 1;
65d33fd7 3835
c146afad 3836 for (i = 0; i < ret; i++) {
65d33fd7
QW		 3837 /* Avoid to grab roots in dead_roots */
3838 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3839 gang[i] = NULL;
3840 continue;
3841 }
3842 /* grab all the search result for later use */
3843 gang[i] = btrfs_grab_fs_root(gang[i]);
3844 }
3845 srcu_read_unlock(&fs_info->subvol_srcu, index);
66b4ffd1 3846
65d33fd7
QW		 3847 for (i = 0; i < ret; i++) {
3848 if (!gang[i])
3849 continue;
c146afad 3850 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 3851 err = btrfs_orphan_cleanup(gang[i]);
3852 if (err)
65d33fd7
QW		 3853 break;
3854 btrfs_put_fs_root(gang[i]);
c146afad
YZ		 3855 }
3856 root_objectid++;
3857 }
65d33fd7
QW		 3858
3859 /* release the uncleaned roots due to error */
3860 for (; i < ret; i++) {
3861 if (gang[i])
3862 btrfs_put_fs_root(gang[i]);
3863 }
3864 return err;
c146afad 3865}
a2135011 3866
c146afad
YZ		 3867int btrfs_commit_super(struct btrfs_root *root)
3868{
3869 struct btrfs_trans_handle *trans;
a74a4b97 3870
c146afad 3871 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 3872 btrfs_run_delayed_iputs(root);
c146afad 3873 mutex_unlock(&root->fs_info->cleaner_mutex);
9d1a2a3a 3874 wake_up_process(root->fs_info->cleaner_kthread);
c71bf099
YZ		 3875
3876 /* wait until ongoing cleanup work done */
3877 down_write(&root->fs_info->cleanup_work_sem);
3878 up_write(&root->fs_info->cleanup_work_sem);
3879
7a7eaa40 3880 trans = btrfs_join_transaction(root);
3612b495
TI		 3881 if (IS_ERR(trans))
3882 return PTR_ERR(trans);
d52c1bcc 3883 return btrfs_commit_transaction(trans, root);
c146afad
YZ		 3884}
3885
3abdbd78 3886void close_ctree(struct btrfs_root *root)
c146afad
YZ		 3887{
3888 struct btrfs_fs_info *fs_info = root->fs_info;
3889 int ret;
3890
afcdd129 3891 set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags);
c146afad 3892
7343dd61 3893 /* wait for the qgroup rescan worker to stop */
d06f23d6 3894 btrfs_qgroup_wait_for_completion(fs_info, false);
7343dd61 3895
803b2f54
SB		 3896 /* wait for the uuid_scan task to finish */
3897 down(&fs_info->uuid_tree_rescan_sem);
3898 /* avoid complains from lockdep et al., set sem back to initial state */
3899 up(&fs_info->uuid_tree_rescan_sem);
3900
837d5b6e 3901 /* pause restriper - we want to resume on mount */
aa1b8cd4 3902 btrfs_pause_balance(fs_info);
837d5b6e 3903
8dabb742
SB		 3904 btrfs_dev_replace_suspend_for_unmount(fs_info);
3905
aa1b8cd4 3906 btrfs_scrub_cancel(fs_info);
4cb5300b
CM		 3907
3908 /* wait for any defraggers to finish */
3909 wait_event(fs_info->transaction_wait,
3910 (atomic_read(&fs_info->defrag_running) == 0));
3911
3912 /* clear out the rbtree of defraggable inodes */
26176e7c 3913 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3914
21c7e756
MX		 3915 cancel_work_sync(&fs_info->async_reclaim_work);
3916
c146afad 3917 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
e44163e1
JM		 3918 /*
3919 * If the cleaner thread is stopped and there are
3920 * block groups queued for removal, the deletion will be
3921 * skipped when we quit the cleaner thread.
3922 */
e44163e1 3923 btrfs_delete_unused_bgs(root->fs_info);
e44163e1 3924
acce952b 3925 ret = btrfs_commit_super(root);
3926 if (ret)
04892340 3927 btrfs_err(fs_info, "commit super ret %d", ret);
acce952b 3928 }
3929
87533c47 3930 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
68ce9682 3931 btrfs_error_commit_super(root);
0f7d52f4 3932
e3029d9f
AV		 3933 kthread_stop(fs_info->transaction_kthread);
3934 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3935
afcdd129 3936 set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags);
f25784b3 3937
04892340 3938 btrfs_free_qgroup_config(fs_info);
bcef60f2 3939
963d678b 3940 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
04892340 3941 btrfs_info(fs_info, "at unmount delalloc count %lld",
963d678b 3942 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3943 }
bcc63abb 3944
6618a59b 3945 btrfs_sysfs_remove_mounted(fs_info);
b7c35e81 3946 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
5ac1d209 3947
faa2dbf0 3948 btrfs_free_fs_roots(fs_info);
d10c5f31 3949
1a4319cc
LB		 3950 btrfs_put_block_group_cache(fs_info);
3951
2b1360da
JB		 3952 btrfs_free_block_groups(fs_info);
3953
de348ee0
WS		 3954 /*
3955 * we must make sure there is not any read request to
3956 * submit after we stopping all workers.
3957 */
3958 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
96192499
JB		 3959 btrfs_stop_all_workers(fs_info);
3960
afcdd129 3961 clear_bit(BTRFS_FS_OPEN, &fs_info->flags);
13e6c37b 3962 free_root_pointers(fs_info, 1);
9ad6b7bc 3963
13e6c37b 3964 iput(fs_info->btree_inode);
d6bfde87 3965
21adbd5c 3966#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3cdde224 3967 if (btrfs_test_opt(root->fs_info, CHECK_INTEGRITY))
21adbd5c
SB		 3968 btrfsic_unmount(root, fs_info->fs_devices);
3969#endif
3970
dfe25020 3971 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3972 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3973
e2d84521 3974 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 3975 percpu_counter_destroy(&fs_info->delalloc_bytes);
c404e0dc 3976 percpu_counter_destroy(&fs_info->bio_counter);
04160088 3977 bdi_destroy(&fs_info->bdi);
76dda93c 3978 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 3979
53b381b3
DW		 3980 btrfs_free_stripe_hash_table(fs_info);
3981
cdfb080e 3982 __btrfs_free_block_rsv(root->orphan_block_rsv);
1cb048f5 3983 root->orphan_block_rsv = NULL;
04216820
FM		 3984
3985 lock_chunks(root);
3986 while (!list_empty(&fs_info->pinned_chunks)) {
3987 struct extent_map *em;
3988
3989 em = list_first_entry(&fs_info->pinned_chunks,
3990 struct extent_map, list);
3991 list_del_init(&em->list);
3992 free_extent_map(em);
3993 }
3994 unlock_chunks(root);
eb60ceac
CM		 3995}
3996
b9fab919
CM		 3997int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3998 int atomic)
5f39d397 3999{
1259ab75 4000 int ret;
727011e0 4001 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 4002
0b32f4bb 4003 ret = extent_buffer_uptodate(buf);
1259ab75
CM		 4004 if (!ret)
4005 return ret;
4006
4007 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM		 4008 parent_transid, atomic);
4009 if (ret == -EAGAIN)
4010 return ret;
1259ab75 4011 return !ret;
5f39d397
CM		 4012}
4013
5f39d397
CM		 4014void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
4015{
06ea65a3 4016 struct btrfs_root *root;
5f39d397 4017 u64 transid = btrfs_header_generation(buf);
b9473439 4018 int was_dirty;
b4ce94de 4019
06ea65a3
JB		 4020#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
4021 /*
4022 * This is a fast path so only do this check if we have sanity tests
4023 * enabled. Normal people shouldn't be marking dummy buffers as dirty
4024 * outside of the sanity tests.
4025 */
4026 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
4027 return;
4028#endif
4029 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
b9447ef8 4030 btrfs_assert_tree_locked(buf);
31b1a2bd
JL		 4031 if (transid != root->fs_info->generation)
4032 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
d397712b 4033 "found %llu running %llu\n",
c1c9ff7c 4034 buf->start, transid, root->fs_info->generation);
0b32f4bb 4035 was_dirty = set_extent_buffer_dirty(buf);
e2d84521
MX		 4036 if (!was_dirty)
4037 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
4038 buf->len,
4039 root->fs_info->dirty_metadata_batch);
1f21ef0a
FM		 4040#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
4041 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
4042 btrfs_print_leaf(root, buf);
4043 ASSERT(0);
4044 }
4045#endif
eb60ceac
CM		 4046}
4047
b53d3f5d
LB		 4048static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
4049 int flush_delayed)
16cdcec7
MX		 4050{
4051 /*
4052 * looks as though older kernels can get into trouble with
4053 * this code, they end up stuck in balance_dirty_pages forever
4054 */
e2d84521 4055 int ret;
16cdcec7
MX		 4056
4057 if (current->flags & PF_MEMALLOC)
4058 return;
4059
b53d3f5d
LB		 4060 if (flush_delayed)
4061 btrfs_balance_delayed_items(root);
16cdcec7 4062
e2d84521
MX		 4063 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
4064 BTRFS_DIRTY_METADATA_THRESH);
4065 if (ret > 0) {
d0e1d66b
NJ		 4066 balance_dirty_pages_ratelimited(
4067 root->fs_info->btree_inode->i_mapping);
16cdcec7 4068 }
16cdcec7
MX		 4069}
4070
b53d3f5d 4071void btrfs_btree_balance_dirty(struct btrfs_root *root)
35b7e476 4072{
b53d3f5d
LB		 4073 __btrfs_btree_balance_dirty(root, 1);
4074}
585ad2c3 4075
b53d3f5d
LB		 4076void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
4077{
4078 __btrfs_btree_balance_dirty(root, 0);
35b7e476 4079}
6b80053d 4080
ca7a79ad 4081int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 4082{
727011e0 4083 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
0b32f4bb 4084 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
6b80053d 4085}
0da5468f 4086
fcd1f065 4087static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 4088 int read_only)
4089{
c926093e 4090 struct btrfs_super_block *sb = fs_info->super_copy;
319e4d06
QW		 4091 u64 nodesize = btrfs_super_nodesize(sb);
4092 u64 sectorsize = btrfs_super_sectorsize(sb);
c926093e
DS		 4093 int ret = 0;
4094
319e4d06
QW		 4095 if (btrfs_super_magic(sb) != BTRFS_MAGIC) {
4096 printk(KERN_ERR "BTRFS: no valid FS found\n");
4097 ret = -EINVAL;
4098 }
4099 if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP)
4100 printk(KERN_WARNING "BTRFS: unrecognized super flag: %llu\n",
4101 btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP);
21e7626b
DS		 4102 if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) {
4103 printk(KERN_ERR "BTRFS: tree_root level too big: %d >= %d\n",
4104 btrfs_super_root_level(sb), BTRFS_MAX_LEVEL);
c926093e
DS		 4105 ret = -EINVAL;
4106 }
21e7626b
DS		 4107 if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) {
4108 printk(KERN_ERR "BTRFS: chunk_root level too big: %d >= %d\n",
4109 btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL);
c926093e
DS		 4110 ret = -EINVAL;
4111 }
21e7626b
DS		 4112 if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) {
4113 printk(KERN_ERR "BTRFS: log_root level too big: %d >= %d\n",
4114 btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL);
c926093e
DS		 4115 ret = -EINVAL;
4116 }
4117
1104a885 4118 /*
319e4d06
QW		 4119 * Check sectorsize and nodesize first, other check will need it.
4120 * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here.
1104a885 4121 */
319e4d06
QW		 4122 if (!is_power_of_2(sectorsize) || sectorsize < 4096 ||
4123 sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) {
4124 printk(KERN_ERR "BTRFS: invalid sectorsize %llu\n", sectorsize);
4125 ret = -EINVAL;
4126 }
4127 /* Only PAGE SIZE is supported yet */
09cbfeaf 4128 if (sectorsize != PAGE_SIZE) {
319e4d06 4129 printk(KERN_ERR "BTRFS: sectorsize %llu not supported yet, only support %lu\n",
09cbfeaf 4130 sectorsize, PAGE_SIZE);
319e4d06
QW		 4131 ret = -EINVAL;
4132 }
4133 if (!is_power_of_2(nodesize) || nodesize < sectorsize ||
4134 nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
4135 printk(KERN_ERR "BTRFS: invalid nodesize %llu\n", nodesize);
4136 ret = -EINVAL;
4137 }
4138 if (nodesize != le32_to_cpu(sb->__unused_leafsize)) {
4139 printk(KERN_ERR "BTRFS: invalid leafsize %u, should be %llu\n",
4140 le32_to_cpu(sb->__unused_leafsize),
4141 nodesize);
4142 ret = -EINVAL;
4143 }
4144
4145 /* Root alignment check */
4146 if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) {
c926093e 4147 printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n",
cd743fac 4148 btrfs_super_root(sb));
319e4d06
QW		 4149 ret = -EINVAL;
4150 }
4151 if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) {
cd743fac
DS		 4152 printk(KERN_WARNING "BTRFS: chunk_root block unaligned: %llu\n",
4153 btrfs_super_chunk_root(sb));
75d6ad38
DS		 4154 ret = -EINVAL;
4155 }
319e4d06
QW		 4156 if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) {
4157 printk(KERN_WARNING "BTRFS: log_root block unaligned: %llu\n",
4158 btrfs_super_log_root(sb));
75d6ad38
DS		 4159 ret = -EINVAL;
4160 }
4161
c926093e
DS		 4162 if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) {
4163 printk(KERN_ERR "BTRFS: dev_item UUID does not match fsid: %pU != %pU\n",
4164 fs_info->fsid, sb->dev_item.fsid);
4165 ret = -EINVAL;
4166 }
4167
4168 /*
4169 * Hint to catch really bogus numbers, bitflips or so, more exact checks are
4170 * done later
4171 */
99e3ecfc
LB		 4172 if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) {
4173 btrfs_err(fs_info, "bytes_used is too small %llu",
4174 btrfs_super_bytes_used(sb));
4175 ret = -EINVAL;
4176 }
b7f67055 4177 if (!is_power_of_2(btrfs_super_stripesize(sb))) {
99e3ecfc
LB		 4178 btrfs_err(fs_info, "invalid stripesize %u",
4179 btrfs_super_stripesize(sb));
4180 ret = -EINVAL;
4181 }
21e7626b 4182 if (btrfs_super_num_devices(sb) > (1UL << 31))
c926093e 4183 printk(KERN_WARNING "BTRFS: suspicious number of devices: %llu\n",
21e7626b 4184 btrfs_super_num_devices(sb));
75d6ad38
DS		 4185 if (btrfs_super_num_devices(sb) == 0) {
4186 printk(KERN_ERR "BTRFS: number of devices is 0\n");
4187 ret = -EINVAL;
4188 }
c926093e 4189
21e7626b 4190 if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) {
c926093e 4191 printk(KERN_ERR "BTRFS: super offset mismatch %llu != %u\n",
21e7626b 4192 btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET);
c926093e
DS		 4193 ret = -EINVAL;
4194 }
4195
ce7fca5f
DS		 4196 /*
4197 * Obvious sys_chunk_array corruptions, it must hold at least one key
4198 * and one chunk
4199 */
4200 if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) {
4201 printk(KERN_ERR "BTRFS: system chunk array too big %u > %u\n",
4202 btrfs_super_sys_array_size(sb),
4203 BTRFS_SYSTEM_CHUNK_ARRAY_SIZE);
4204 ret = -EINVAL;
4205 }
4206 if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key)
4207 + sizeof(struct btrfs_chunk)) {
d2207129 4208 printk(KERN_ERR "BTRFS: system chunk array too small %u < %zu\n",
ce7fca5f
DS		 4209 btrfs_super_sys_array_size(sb),
4210 sizeof(struct btrfs_disk_key)
4211 + sizeof(struct btrfs_chunk));
4212 ret = -EINVAL;
4213 }
4214
c926093e
DS		 4215 /*
4216 * The generation is a global counter, we'll trust it more than the others
4217 * but it's still possible that it's the one that's wrong.
4218 */
21e7626b 4219 if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb))
c926093e
DS		 4220 printk(KERN_WARNING
4221 "BTRFS: suspicious: generation < chunk_root_generation: %llu < %llu\n",
21e7626b
DS		 4222 btrfs_super_generation(sb), btrfs_super_chunk_root_generation(sb));
4223 if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb)
4224 && btrfs_super_cache_generation(sb) != (u64)-1)
c926093e
DS		 4225 printk(KERN_WARNING
4226 "BTRFS: suspicious: generation < cache_generation: %llu < %llu\n",
21e7626b 4227 btrfs_super_generation(sb), btrfs_super_cache_generation(sb));
c926093e
DS		 4228
4229 return ret;
acce952b 4230}
4231
48a3b636 4232static void btrfs_error_commit_super(struct btrfs_root *root)
acce952b 4233{
acce952b 4234 mutex_lock(&root->fs_info->cleaner_mutex);
4235 btrfs_run_delayed_iputs(root);
4236 mutex_unlock(&root->fs_info->cleaner_mutex);
4237
4238 down_write(&root->fs_info->cleanup_work_sem);
4239 up_write(&root->fs_info->cleanup_work_sem);
4240
4241 /* cleanup FS via transaction */
4242 btrfs_cleanup_transaction(root);
acce952b 4243}
4244
143bede5 4245static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 4246{
acce952b 4247 struct btrfs_ordered_extent *ordered;
acce952b 4248
199c2a9c 4249 spin_lock(&root->ordered_extent_lock);
779880ef
JB		 4250 /*
4251 * This will just short circuit the ordered completion stuff which will
4252 * make sure the ordered extent gets properly cleaned up.
4253 */
199c2a9c 4254 list_for_each_entry(ordered, &root->ordered_extents,
779880ef
JB		 4255 root_extent_list)
4256 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
199c2a9c
MX		 4257 spin_unlock(&root->ordered_extent_lock);
4258}
4259
4260static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
4261{
4262 struct btrfs_root *root;
4263 struct list_head splice;
4264
4265 INIT_LIST_HEAD(&splice);
4266
4267 spin_lock(&fs_info->ordered_root_lock);
4268 list_splice_init(&fs_info->ordered_roots, &splice);
4269 while (!list_empty(&splice)) {
4270 root = list_first_entry(&splice, struct btrfs_root,
4271 ordered_root);
1de2cfde
JB		 4272 list_move_tail(&root->ordered_root,
4273 &fs_info->ordered_roots);
199c2a9c 4274
2a85d9ca 4275 spin_unlock(&fs_info->ordered_root_lock);
199c2a9c
MX		 4276 btrfs_destroy_ordered_extents(root);
4277
2a85d9ca
LB		 4278 cond_resched();
4279 spin_lock(&fs_info->ordered_root_lock);
199c2a9c
MX		 4280 }
4281 spin_unlock(&fs_info->ordered_root_lock);
acce952b 4282}
4283
35a3621b
SB		 4284static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
4285 struct btrfs_root *root)
acce952b 4286{
4287 struct rb_node *node;
4288 struct btrfs_delayed_ref_root *delayed_refs;
4289 struct btrfs_delayed_ref_node *ref;
4290 int ret = 0;
4291
4292 delayed_refs = &trans->delayed_refs;
4293
4294 spin_lock(&delayed_refs->lock);
d7df2c79 4295 if (atomic_read(&delayed_refs->num_entries) == 0) {
cfece4db 4296 spin_unlock(&delayed_refs->lock);
efe120a0 4297 btrfs_info(root->fs_info, "delayed_refs has NO entry");
acce952b 4298 return ret;
4299 }
4300
d7df2c79
JB		 4301 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
4302 struct btrfs_delayed_ref_head *head;
c6fc2454 4303 struct btrfs_delayed_ref_node *tmp;
e78417d1 4304 bool pin_bytes = false;
acce952b 4305
d7df2c79
JB		 4306 head = rb_entry(node, struct btrfs_delayed_ref_head,
4307 href_node);
4308 if (!mutex_trylock(&head->mutex)) {
4309 atomic_inc(&head->node.refs);
4310 spin_unlock(&delayed_refs->lock);
eb12db69 4311
d7df2c79 4312 mutex_lock(&head->mutex);
e78417d1 4313 mutex_unlock(&head->mutex);
d7df2c79
JB		 4314 btrfs_put_delayed_ref(&head->node);
4315 spin_lock(&delayed_refs->lock);
4316 continue;
4317 }
4318 spin_lock(&head->lock);
c6fc2454
QW		 4319 list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list,
4320 list) {
d7df2c79 4321 ref->in_tree = 0;
c6fc2454 4322 list_del(&ref->list);
d7df2c79
JB		 4323 atomic_dec(&delayed_refs->num_entries);
4324 btrfs_put_delayed_ref(ref);
e78417d1 4325 }
d7df2c79
JB		 4326 if (head->must_insert_reserved)
4327 pin_bytes = true;
4328 btrfs_free_delayed_extent_op(head->extent_op);
4329 delayed_refs->num_heads--;
4330 if (head->processing == 0)
4331 delayed_refs->num_heads_ready--;
4332 atomic_dec(&delayed_refs->num_entries);
4333 head->node.in_tree = 0;
4334 rb_erase(&head->href_node, &delayed_refs->href_root);
4335 spin_unlock(&head->lock);
4336 spin_unlock(&delayed_refs->lock);
4337 mutex_unlock(&head->mutex);
acce952b 4338
d7df2c79
JB		 4339 if (pin_bytes)
4340 btrfs_pin_extent(root, head->node.bytenr,
4341 head->node.num_bytes, 1);
4342 btrfs_put_delayed_ref(&head->node);
acce952b 4343 cond_resched();
4344 spin_lock(&delayed_refs->lock);
4345 }
4346
4347 spin_unlock(&delayed_refs->lock);
4348
4349 return ret;
4350}
4351
143bede5 4352static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 4353{
4354 struct btrfs_inode *btrfs_inode;
4355 struct list_head splice;
4356
4357 INIT_LIST_HEAD(&splice);
4358
eb73c1b7
MX		 4359 spin_lock(&root->delalloc_lock);
4360 list_splice_init(&root->delalloc_inodes, &splice);
acce952b 4361
4362 while (!list_empty(&splice)) {
eb73c1b7
MX		 4363 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
4364 delalloc_inodes);
acce952b 4365
4366 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX		 4367 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
4368 &btrfs_inode->runtime_flags);
eb73c1b7 4369 spin_unlock(&root->delalloc_lock);
acce952b 4370
4371 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 4372
eb73c1b7 4373 spin_lock(&root->delalloc_lock);
acce952b 4374 }
4375
eb73c1b7
MX		 4376 spin_unlock(&root->delalloc_lock);
4377}
4378
4379static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
4380{
4381 struct btrfs_root *root;
4382 struct list_head splice;
4383
4384 INIT_LIST_HEAD(&splice);
4385
4386 spin_lock(&fs_info->delalloc_root_lock);
4387 list_splice_init(&fs_info->delalloc_roots, &splice);
4388 while (!list_empty(&splice)) {
4389 root = list_first_entry(&splice, struct btrfs_root,
4390 delalloc_root);
4391 list_del_init(&root->delalloc_root);
4392 root = btrfs_grab_fs_root(root);
4393 BUG_ON(!root);
4394 spin_unlock(&fs_info->delalloc_root_lock);
4395
4396 btrfs_destroy_delalloc_inodes(root);
4397 btrfs_put_fs_root(root);
4398
4399 spin_lock(&fs_info->delalloc_root_lock);
4400 }
4401 spin_unlock(&fs_info->delalloc_root_lock);
acce952b 4402}
4403
4404static int btrfs_destroy_marked_extents(struct btrfs_root *root,
4405 struct extent_io_tree *dirty_pages,
4406 int mark)
4407{
4408 int ret;
acce952b 4409 struct extent_buffer *eb;
4410 u64 start = 0;
4411 u64 end;
acce952b 4412
4413 while (1) {
4414 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 4415 mark, NULL);
acce952b 4416 if (ret)
4417 break;
4418
91166212 4419 clear_extent_bits(dirty_pages, start, end, mark);
acce952b 4420 while (start <= end) {
01d58472 4421 eb = btrfs_find_tree_block(root->fs_info, start);
707e8a07 4422 start += root->nodesize;
fd8b2b61 4423 if (!eb)
acce952b 4424 continue;
fd8b2b61 4425 wait_on_extent_buffer_writeback(eb);
acce952b 4426
fd8b2b61
JB		 4427 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4428 &eb->bflags))
4429 clear_extent_buffer_dirty(eb);
4430 free_extent_buffer_stale(eb);
acce952b 4431 }
4432 }
4433
4434 return ret;
4435}
4436
4437static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4438 struct extent_io_tree *pinned_extents)
4439{
4440 struct extent_io_tree *unpin;
4441 u64 start;
4442 u64 end;
4443 int ret;
ed0eaa14 4444 bool loop = true;
acce952b 4445
4446 unpin = pinned_extents;
ed0eaa14 4447again:
acce952b 4448 while (1) {
4449 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 4450 EXTENT_DIRTY, NULL);
acce952b 4451 if (ret)
4452 break;
4453
af6f8f60 4454 clear_extent_dirty(unpin, start, end);
acce952b 4455 btrfs_error_unpin_extent_range(root, start, end);
4456 cond_resched();
4457 }
4458
ed0eaa14
LB		 4459 if (loop) {
4460 if (unpin == &root->fs_info->freed_extents[0])
4461 unpin = &root->fs_info->freed_extents[1];
4462 else
4463 unpin = &root->fs_info->freed_extents[0];
4464 loop = false;
4465 goto again;
4466 }
4467
acce952b 4468 return 0;
4469}
4470
c79a1751
LB		 4471static void btrfs_cleanup_bg_io(struct btrfs_block_group_cache *cache)
4472{
4473 struct inode *inode;
4474
4475 inode = cache->io_ctl.inode;
4476 if (inode) {
4477 invalidate_inode_pages2(inode->i_mapping);
4478 BTRFS_I(inode)->generation = 0;
4479 cache->io_ctl.inode = NULL;
4480 iput(inode);
4481 }
4482 btrfs_put_block_group(cache);
4483}
4484
4485void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans,
4486 struct btrfs_root *root)
4487{
4488 struct btrfs_block_group_cache *cache;
4489
4490 spin_lock(&cur_trans->dirty_bgs_lock);
4491 while (!list_empty(&cur_trans->dirty_bgs)) {
4492 cache = list_first_entry(&cur_trans->dirty_bgs,
4493 struct btrfs_block_group_cache,
4494 dirty_list);
4495 if (!cache) {
4496 btrfs_err(root->fs_info,
4497 "orphan block group dirty_bgs list");
4498 spin_unlock(&cur_trans->dirty_bgs_lock);
4499 return;
4500 }
4501
4502 if (!list_empty(&cache->io_list)) {
4503 spin_unlock(&cur_trans->dirty_bgs_lock);
4504 list_del_init(&cache->io_list);
4505 btrfs_cleanup_bg_io(cache);
4506 spin_lock(&cur_trans->dirty_bgs_lock);
4507 }
4508
4509 list_del_init(&cache->dirty_list);
4510 spin_lock(&cache->lock);
4511 cache->disk_cache_state = BTRFS_DC_ERROR;
4512 spin_unlock(&cache->lock);
4513
4514 spin_unlock(&cur_trans->dirty_bgs_lock);
4515 btrfs_put_block_group(cache);
4516 spin_lock(&cur_trans->dirty_bgs_lock);
4517 }
4518 spin_unlock(&cur_trans->dirty_bgs_lock);
4519
4520 while (!list_empty(&cur_trans->io_bgs)) {
4521 cache = list_first_entry(&cur_trans->io_bgs,
4522 struct btrfs_block_group_cache,
4523 io_list);
4524 if (!cache) {
4525 btrfs_err(root->fs_info,
4526 "orphan block group on io_bgs list");
4527 return;
4528 }
4529
4530 list_del_init(&cache->io_list);
4531 spin_lock(&cache->lock);
4532 cache->disk_cache_state = BTRFS_DC_ERROR;
4533 spin_unlock(&cache->lock);
4534 btrfs_cleanup_bg_io(cache);
4535 }
4536}
4537
49b25e05
JM		 4538void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4539 struct btrfs_root *root)
4540{
c79a1751
LB		 4541 btrfs_cleanup_dirty_bgs(cur_trans, root);
4542 ASSERT(list_empty(&cur_trans->dirty_bgs));
4543 ASSERT(list_empty(&cur_trans->io_bgs));
4544
49b25e05 4545 btrfs_destroy_delayed_refs(cur_trans, root);
49b25e05 4546
4a9d8bde 4547 cur_trans->state = TRANS_STATE_COMMIT_START;
d7096fc3 4548 wake_up(&root->fs_info->transaction_blocked_wait);
49b25e05 4549
4a9d8bde 4550 cur_trans->state = TRANS_STATE_UNBLOCKED;
d7096fc3 4551 wake_up(&root->fs_info->transaction_wait);
49b25e05 4552
67cde344
MX		 4553 btrfs_destroy_delayed_inodes(root);
4554 btrfs_assert_delayed_root_empty(root);
49b25e05 4555
49b25e05
JM		 4556 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4557 EXTENT_DIRTY);
6e841e32
LB		 4558 btrfs_destroy_pinned_extent(root,
4559 root->fs_info->pinned_extents);
49b25e05 4560
4a9d8bde
MX		 4561 cur_trans->state =TRANS_STATE_COMPLETED;
4562 wake_up(&cur_trans->commit_wait);
4563
49b25e05
JM		 4564 /*
4565 memset(cur_trans, 0, sizeof(*cur_trans));
4566 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4567 */
4568}
4569
48a3b636 4570static int btrfs_cleanup_transaction(struct btrfs_root *root)
acce952b 4571{
4572 struct btrfs_transaction *t;
acce952b 4573
acce952b 4574 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4575
a4abeea4 4576 spin_lock(&root->fs_info->trans_lock);
724e2315
JB		 4577 while (!list_empty(&root->fs_info->trans_list)) {
4578 t = list_first_entry(&root->fs_info->trans_list,
4579 struct btrfs_transaction, list);
4580 if (t->state >= TRANS_STATE_COMMIT_START) {
4581 atomic_inc(&t->use_count);
4582 spin_unlock(&root->fs_info->trans_lock);
4583 btrfs_wait_for_commit(root, t->transid);
4584 btrfs_put_transaction(t);
4585 spin_lock(&root->fs_info->trans_lock);
4586 continue;
4587 }
4588 if (t == root->fs_info->running_transaction) {
4589 t->state = TRANS_STATE_COMMIT_DOING;
4590 spin_unlock(&root->fs_info->trans_lock);
4591 /*
4592 * We wait for 0 num_writers since we don't hold a trans
4593 * handle open currently for this transaction.
4594 */
4595 wait_event(t->writer_wait,
4596 atomic_read(&t->num_writers) == 0);
4597 } else {
4598 spin_unlock(&root->fs_info->trans_lock);
4599 }
4600 btrfs_cleanup_one_transaction(t, root);
4a9d8bde 4601
724e2315
JB		 4602 spin_lock(&root->fs_info->trans_lock);
4603 if (t == root->fs_info->running_transaction)
4604 root->fs_info->running_transaction = NULL;
acce952b 4605 list_del_init(&t->list);
724e2315 4606 spin_unlock(&root->fs_info->trans_lock);
acce952b 4607
724e2315
JB		 4608 btrfs_put_transaction(t);
4609 trace_btrfs_transaction_commit(root);
4610 spin_lock(&root->fs_info->trans_lock);
4611 }
4612 spin_unlock(&root->fs_info->trans_lock);
4613 btrfs_destroy_all_ordered_extents(root->fs_info);
4614 btrfs_destroy_delayed_inodes(root);
4615 btrfs_assert_delayed_root_empty(root);
4616 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4617 btrfs_destroy_all_delalloc_inodes(root->fs_info);
acce952b 4618 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 4619
4620 return 0;
4621}
4622
e8c9f186 4623static const struct extent_io_ops btree_extent_io_ops = {
ce9adaa5 4624 .readpage_end_io_hook = btree_readpage_end_io_hook,
4bb31e92 4625 .readpage_io_failed_hook = btree_io_failed_hook,
0b86a832 4626 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 4627 /* note we're sharing with inode.c for the merge bio hook */
4628 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 4629};
Linux 6.x block layer and io_uring tree(s)