projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: introduce helpers for updating eb uuids
[linux-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 328 btrfs_warn_rl(fs_info,
5d163e0e 329 "%s checksum verify failed on %llu wanted %X found %X level %d",
01d58472 330 fs_info->sb->s_id, buf->start,
efe120a0 331 val, found, btrfs_header_level(buf));
607d432d
JB		 332 if (result != (char *)&inline_result)
333 kfree(result);
8bd98f0e 334 return -EUCLEAN;
19c00ddc
CM		 335 }
336 } else {
607d432d 337 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 338 }
607d432d
JB		 339 if (result != (char *)&inline_result)
340 kfree(result);
19c00ddc
CM		 341 return 0;
342}
343
d352ac68
CM		 344/*
345 * we can't consider a given block up to date unless the transid of the
346 * block matches the transid in the parent node's pointer. This is how we
347 * detect blocks that either didn't get written at all or got written
348 * in the wrong place.
349 */
1259ab75 350static int verify_parent_transid(struct extent_io_tree *io_tree,
b9fab919
CM		 351 struct extent_buffer *eb, u64 parent_transid,
352 int atomic)
1259ab75 353{
2ac55d41 354 struct extent_state *cached_state = NULL;
1259ab75 355 int ret;
2755a0de 356 bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB);
1259ab75
CM		 357
358 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
359 return 0;
360
b9fab919
CM		 361 if (atomic)
362 return -EAGAIN;
363
a26e8c9f
JB		 364 if (need_lock) {
365 btrfs_tree_read_lock(eb);
366 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
367 }
368
2ac55d41 369 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
ff13db41 370 &cached_state);
0b32f4bb 371 if (extent_buffer_uptodate(eb) &&
1259ab75
CM		 372 btrfs_header_generation(eb) == parent_transid) {
373 ret = 0;
374 goto out;
375 }
94647322
DS		 376 btrfs_err_rl(eb->fs_info,
377 "parent transid verify failed on %llu wanted %llu found %llu",
378 eb->start,
29549aec 379 parent_transid, btrfs_header_generation(eb));
1259ab75 380 ret = 1;
a26e8c9f
JB		 381
382 /*
383 * Things reading via commit roots that don't have normal protection,
384 * like send, can have a really old block in cache that may point at a
01327610 385 * block that has been freed and re-allocated. So don't clear uptodate
a26e8c9f
JB		 386 * if we find an eb that is under IO (dirty/writeback) because we could
387 * end up reading in the stale data and then writing it back out and
388 * making everybody very sad.
389 */
390 if (!extent_buffer_under_io(eb))
391 clear_extent_buffer_uptodate(eb);
33958dc6 392out:
2ac55d41
JB		 393 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
394 &cached_state, GFP_NOFS);
472b909f
JB		 395 if (need_lock)
396 btrfs_tree_read_unlock_blocking(eb);
1259ab75 397 return ret;
1259ab75
CM		 398}
399
1104a885
DS		 400/*
401 * Return 0 if the superblock checksum type matches the checksum value of that
402 * algorithm. Pass the raw disk superblock data.
403 */
ab8d0fc4
JM		 404static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info,
405 char *raw_disk_sb)
1104a885
DS		 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)) {
ab8d0fc4 431 btrfs_err(fs_info, "unsupported checksum algorithm %u",
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,
8436ea91 445 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) {
8436ea91 457 ret = read_extent_buffer_pages(io_tree, eb, WAIT_COMPLETE,
f188591e 458 btree_get_extent, mirror_num);
256dd1bb
SB		 459 if (!ret) {
460 if (!verify_parent_transid(io_tree, eb,
b9fab919 461 parent_transid, 0))
256dd1bb
SB		 462 break;
463 else
464 ret = -EIO;
465 }
d397712b 466
a826d6dc
JB		 467 /*
468 * This buffer's crc is fine, but its contents are corrupted, so
469 * there is no reason to read the other copies, they won't be
470 * any less wrong.
471 */
472 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
ea466794
JB		 473 break;
474
5d964051 475 num_copies = btrfs_num_copies(root->fs_info,
f188591e 476 eb->start, eb->len);
4235298e 477 if (num_copies == 1)
ea466794 478 break;
4235298e 479
5cf1ab56
JB		 480 if (!failed_mirror) {
481 failed = 1;
482 failed_mirror = eb->read_mirror;
483 }
484
f188591e 485 mirror_num++;
ea466794
JB		 486 if (mirror_num == failed_mirror)
487 mirror_num++;
488
4235298e 489 if (mirror_num > num_copies)
ea466794 490 break;
f188591e 491 }
ea466794 492
c0901581 493 if (failed && !ret && failed_mirror)
ea466794
JB		 494 repair_eb_io_failure(root, eb, failed_mirror);
495
496 return ret;
f188591e 497}
19c00ddc 498
d352ac68 499/*
d397712b
CM		 500 * checksum a dirty tree block before IO. This has extra checks to make sure
501 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 502 */
d397712b 503
01d58472 504static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page)
19c00ddc 505{
4eee4fa4 506 u64 start = page_offset(page);
19c00ddc 507 u64 found_start;
19c00ddc 508 struct extent_buffer *eb;
f188591e 509
4f2de97a
JB		 510 eb = (struct extent_buffer *)page->private;
511 if (page != eb->pages[0])
512 return 0;
0f805531 513
19c00ddc 514 found_start = btrfs_header_bytenr(eb);
0f805531
AL		 515 /*
516 * Please do not consolidate these warnings into a single if.
517 * It is useful to know what went wrong.
518 */
519 if (WARN_ON(found_start != start))
520 return -EUCLEAN;
521 if (WARN_ON(!PageUptodate(page)))
522 return -EUCLEAN;
523
524 ASSERT(memcmp_extent_buffer(eb, fs_info->fsid,
525 btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0);
526
8bd98f0e 527 return csum_tree_block(fs_info, eb, 0);
19c00ddc
CM		 528}
529
01d58472 530static int check_tree_block_fsid(struct btrfs_fs_info *fs_info,
2b82032c
YZ		 531 struct extent_buffer *eb)
532{
01d58472 533 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2b82032c
YZ		 534 u8 fsid[BTRFS_UUID_SIZE];
535 int ret = 1;
536
0a4e5586 537 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
2b82032c
YZ		 538 while (fs_devices) {
539 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
540 ret = 0;
541 break;
542 }
543 fs_devices = fs_devices->seed;
544 }
545 return ret;
546}
547
a826d6dc 548#define CORRUPT(reason, eb, root, slot) \
6b722c17
LB		 549 btrfs_crit(root->fs_info, "corrupt %s, %s: block=%llu," \
550 " root=%llu, slot=%d", \
551 btrfs_header_level(eb) == 0 ? "leaf" : "node",\
552 reason, 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);
6b722c17
LB		 639 struct btrfs_key key, next_key;
640 int slot;
641 u64 bytenr;
642 int ret = 0;
053ab70f
LB		 643
644 if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(root)) {
645 btrfs_crit(root->fs_info,
646 "corrupt node: block %llu root %llu nritems %lu",
647 node->start, root->objectid, nr);
648 return -EIO;
649 }
6b722c17
LB		 650
651 for (slot = 0; slot < nr - 1; slot++) {
652 bytenr = btrfs_node_blockptr(node, slot);
653 btrfs_node_key_to_cpu(node, &key, slot);
654 btrfs_node_key_to_cpu(node, &next_key, slot + 1);
655
656 if (!bytenr) {
657 CORRUPT("invalid item slot", node, root, slot);
658 ret = -EIO;
659 goto out;
660 }
661
662 if (btrfs_comp_cpu_keys(&key, &next_key) >= 0) {
663 CORRUPT("bad key order", node, root, slot);
664 ret = -EIO;
665 goto out;
666 }
667 }
668out:
669 return ret;
053ab70f
LB		 670}
671
facc8a22
MX		 672static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
673 u64 phy_offset, struct page *page,
674 u64 start, u64 end, int mirror)
ce9adaa5 675{
ce9adaa5
CM		 676 u64 found_start;
677 int found_level;
ce9adaa5
CM		 678 struct extent_buffer *eb;
679 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
02873e43 680 struct btrfs_fs_info *fs_info = root->fs_info;
f188591e 681 int ret = 0;
727011e0 682 int reads_done;
ce9adaa5 683
ce9adaa5
CM		 684 if (!page->private)
685 goto out;
d397712b 686
4f2de97a 687 eb = (struct extent_buffer *)page->private;
d397712b 688
0b32f4bb
JB		 689 /* the pending IO might have been the only thing that kept this buffer
690 * in memory. Make sure we have a ref for all this other checks
691 */
692 extent_buffer_get(eb);
693
694 reads_done = atomic_dec_and_test(&eb->io_pages);
727011e0
CM		 695 if (!reads_done)
696 goto err;
f188591e 697
5cf1ab56 698 eb->read_mirror = mirror;
656f30db 699 if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) {
ea466794
JB		 700 ret = -EIO;
701 goto err;
702 }
703
ce9adaa5 704 found_start = btrfs_header_bytenr(eb);
727011e0 705 if (found_start != eb->start) {
02873e43
ZL		 706 btrfs_err_rl(fs_info, "bad tree block start %llu %llu",
707 found_start, eb->start);
f188591e 708 ret = -EIO;
ce9adaa5
CM		 709 goto err;
710 }
02873e43
ZL		 711 if (check_tree_block_fsid(fs_info, eb)) {
712 btrfs_err_rl(fs_info, "bad fsid on block %llu",
713 eb->start);
1259ab75
CM		 714 ret = -EIO;
715 goto err;
716 }
ce9adaa5 717 found_level = btrfs_header_level(eb);
1c24c3ce 718 if (found_level >= BTRFS_MAX_LEVEL) {
02873e43
ZL		 719 btrfs_err(fs_info, "bad tree block level %d",
720 (int)btrfs_header_level(eb));
1c24c3ce
JB		 721 ret = -EIO;
722 goto err;
723 }
ce9adaa5 724
85d4e461
CM		 725 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
726 eb, found_level);
4008c04a 727
02873e43 728 ret = csum_tree_block(fs_info, eb, 1);
8bd98f0e 729 if (ret)
a826d6dc 730 goto err;
a826d6dc
JB		 731
732 /*
733 * If this is a leaf block and it is corrupt, set the corrupt bit so
734 * that we don't try and read the other copies of this block, just
735 * return -EIO.
736 */
737 if (found_level == 0 && check_leaf(root, eb)) {
738 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
739 ret = -EIO;
740 }
ce9adaa5 741
053ab70f
LB		 742 if (found_level > 0 && check_node(root, eb))
743 ret = -EIO;
744
0b32f4bb
JB		 745 if (!ret)
746 set_extent_buffer_uptodate(eb);
ce9adaa5 747err:
79fb65a1
JB		 748 if (reads_done &&
749 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
fc2e901f 750 btree_readahead_hook(fs_info, eb, ret);
4bb31e92 751
53b381b3
DW		 752 if (ret) {
753 /*
754 * our io error hook is going to dec the io pages
755 * again, we have to make sure it has something
756 * to decrement
757 */
758 atomic_inc(&eb->io_pages);
0b32f4bb 759 clear_extent_buffer_uptodate(eb);
53b381b3 760 }
0b32f4bb 761 free_extent_buffer(eb);
ce9adaa5 762out:
f188591e 763 return ret;
ce9adaa5
CM		 764}
765
ea466794 766static int btree_io_failed_hook(struct page *page, int failed_mirror)
4bb31e92 767{
4bb31e92 768 struct extent_buffer *eb;
4bb31e92 769
4f2de97a 770 eb = (struct extent_buffer *)page->private;
656f30db 771 set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
5cf1ab56 772 eb->read_mirror = failed_mirror;
53b381b3 773 atomic_dec(&eb->io_pages);
ea466794 774 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
fc2e901f 775 btree_readahead_hook(eb->fs_info, eb, -EIO);
4bb31e92
AJ		 776 return -EIO; /* we fixed nothing */
777}
778
4246a0b6 779static void end_workqueue_bio(struct bio *bio)
ce9adaa5 780{
97eb6b69 781 struct btrfs_end_io_wq *end_io_wq = bio->bi_private;
ce9adaa5 782 struct btrfs_fs_info *fs_info;
9e0af237
LB		 783 struct btrfs_workqueue *wq;
784 btrfs_work_func_t func;
ce9adaa5 785
ce9adaa5 786 fs_info = end_io_wq->info;
4246a0b6 787 end_io_wq->error = bio->bi_error;
d20f7043 788
37226b21 789 if (bio_op(bio) == REQ_OP_WRITE) {
9e0af237
LB		 790 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) {
791 wq = fs_info->endio_meta_write_workers;
792 func = btrfs_endio_meta_write_helper;
793 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) {
794 wq = fs_info->endio_freespace_worker;
795 func = btrfs_freespace_write_helper;
796 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
797 wq = fs_info->endio_raid56_workers;
798 func = btrfs_endio_raid56_helper;
799 } else {
800 wq = fs_info->endio_write_workers;
801 func = btrfs_endio_write_helper;
802 }
d20f7043 803 } else {
8b110e39
MX		 804 if (unlikely(end_io_wq->metadata ==
805 BTRFS_WQ_ENDIO_DIO_REPAIR)) {
806 wq = fs_info->endio_repair_workers;
807 func = btrfs_endio_repair_helper;
808 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
9e0af237
LB		 809 wq = fs_info->endio_raid56_workers;
810 func = btrfs_endio_raid56_helper;
811 } else if (end_io_wq->metadata) {
812 wq = fs_info->endio_meta_workers;
813 func = btrfs_endio_meta_helper;
814 } else {
815 wq = fs_info->endio_workers;
816 func = btrfs_endio_helper;
817 }
d20f7043 818 }
9e0af237
LB		 819
820 btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL);
821 btrfs_queue_work(wq, &end_io_wq->work);
ce9adaa5
CM		 822}
823
22c59948 824int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
bfebd8b5 825 enum btrfs_wq_endio_type metadata)
0b86a832 826{
97eb6b69 827 struct btrfs_end_io_wq *end_io_wq;
8b110e39 828
97eb6b69 829 end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS);
ce9adaa5
CM		 830 if (!end_io_wq)
831 return -ENOMEM;
832
833 end_io_wq->private = bio->bi_private;
834 end_io_wq->end_io = bio->bi_end_io;
22c59948 835 end_io_wq->info = info;
ce9adaa5
CM		 836 end_io_wq->error = 0;
837 end_io_wq->bio = bio;
22c59948 838 end_io_wq->metadata = metadata;
ce9adaa5
CM		 839
840 bio->bi_private = end_io_wq;
841 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 842 return 0;
843}
844
b64a2851 845unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 846{
4854ddd0 847 unsigned long limit = min_t(unsigned long,
5cdc7ad3 848 info->thread_pool_size,
4854ddd0
CM		 849 info->fs_devices->open_devices);
850 return 256 * limit;
851}
0986fe9e 852
4a69a410
CM		 853static void run_one_async_start(struct btrfs_work *work)
854{
4a69a410 855 struct async_submit_bio *async;
79787eaa 856 int ret;
4a69a410
CM		 857
858 async = container_of(work, struct async_submit_bio, work);
81a75f67 859 ret = async->submit_bio_start(async->inode, async->bio,
79787eaa
JM		 860 async->mirror_num, async->bio_flags,
861 async->bio_offset);
862 if (ret)
863 async->error = ret;
4a69a410
CM		 864}
865
866static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 867{
868 struct btrfs_fs_info *fs_info;
869 struct async_submit_bio *async;
4854ddd0 870 int limit;
8b712842
CM		 871
872 async = container_of(work, struct async_submit_bio, work);
873 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 874
b64a2851 875 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 876 limit = limit * 2 / 3;
877
ee863954
DS		 878 /*
879 * atomic_dec_return implies a barrier for waitqueue_active
880 */
66657b31 881 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
b64a2851 882 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 883 wake_up(&fs_info->async_submit_wait);
884
bb7ab3b9 885 /* If an error occurred we just want to clean up the bio and move on */
79787eaa 886 if (async->error) {
4246a0b6
CH		 887 async->bio->bi_error = async->error;
888 bio_endio(async->bio);
79787eaa
JM		 889 return;
890 }
891
81a75f67
MC		 892 async->submit_bio_done(async->inode, async->bio, async->mirror_num,
893 async->bio_flags, async->bio_offset);
4a69a410
CM		 894}
895
896static void run_one_async_free(struct btrfs_work *work)
897{
898 struct async_submit_bio *async;
899
900 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 901 kfree(async);
902}
903
44b8bd7e 904int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
81a75f67 905 struct bio *bio, int mirror_num,
c8b97818 906 unsigned long bio_flags,
eaf25d93 907 u64 bio_offset,
4a69a410
CM		 908 extent_submit_bio_hook_t *submit_bio_start,
909 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 910{
911 struct async_submit_bio *async;
912
913 async = kmalloc(sizeof(*async), GFP_NOFS);
914 if (!async)
915 return -ENOMEM;
916
917 async->inode = inode;
44b8bd7e
CM		 918 async->bio = bio;
919 async->mirror_num = mirror_num;
4a69a410
CM		 920 async->submit_bio_start = submit_bio_start;
921 async->submit_bio_done = submit_bio_done;
922
9e0af237 923 btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start,
5cdc7ad3 924 run_one_async_done, run_one_async_free);
4a69a410 925
c8b97818 926 async->bio_flags = bio_flags;
eaf25d93 927 async->bio_offset = bio_offset;
8c8bee1d 928
79787eaa
JM		 929 async->error = 0;
930
cb03c743 931 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 932
1eff9d32 933 if (bio->bi_opf & REQ_SYNC)
5cdc7ad3 934 btrfs_set_work_high_priority(&async->work);
d313d7a3 935
5cdc7ad3 936 btrfs_queue_work(fs_info->workers, &async->work);
9473f16c 937
d397712b 938 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 939 atomic_read(&fs_info->nr_async_submits)) {
940 wait_event(fs_info->async_submit_wait,
941 (atomic_read(&fs_info->nr_async_submits) == 0));
942 }
943
44b8bd7e
CM		 944 return 0;
945}
946
ce3ed71a
CM		 947static int btree_csum_one_bio(struct bio *bio)
948{
2c30c71b 949 struct bio_vec *bvec;
ce3ed71a 950 struct btrfs_root *root;
2c30c71b 951 int i, ret = 0;
ce3ed71a 952
2c30c71b 953 bio_for_each_segment_all(bvec, bio, i) {
ce3ed71a 954 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
01d58472 955 ret = csum_dirty_buffer(root->fs_info, bvec->bv_page);
79787eaa
JM		 956 if (ret)
957 break;
ce3ed71a 958 }
2c30c71b 959
79787eaa 960 return ret;
ce3ed71a
CM		 961}
962
81a75f67
MC		 963static int __btree_submit_bio_start(struct inode *inode, struct bio *bio,
964 int mirror_num, unsigned long bio_flags,
eaf25d93 965 u64 bio_offset)
22c59948 966{
8b712842
CM		 967 /*
968 * when we're called for a write, we're already in the async
5443be45 969 * submission context. Just jump into btrfs_map_bio
8b712842 970 */
79787eaa 971 return btree_csum_one_bio(bio);
4a69a410 972}
22c59948 973
81a75f67 974static int __btree_submit_bio_done(struct inode *inode, struct bio *bio,
eaf25d93
CM		 975 int mirror_num, unsigned long bio_flags,
976 u64 bio_offset)
4a69a410 977{
61891923
SB		 978 int ret;
979
8b712842 980 /*
4a69a410
CM		 981 * when we're called for a write, we're already in the async
982 * submission context. Just jump into btrfs_map_bio
8b712842 983 */
81a75f67 984 ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 1);
4246a0b6
CH		 985 if (ret) {
986 bio->bi_error = ret;
987 bio_endio(bio);
988 }
61891923 989 return ret;
0b86a832
CM		 990}
991
de0022b9
JB		 992static int check_async_write(struct inode *inode, unsigned long bio_flags)
993{
994 if (bio_flags & EXTENT_BIO_TREE_LOG)
995 return 0;
996#ifdef CONFIG_X86
bc696ca0 997 if (static_cpu_has(X86_FEATURE_XMM4_2))
de0022b9
JB		 998 return 0;
999#endif
1000 return 1;
1001}
1002
81a75f67 1003static int btree_submit_bio_hook(struct inode *inode, struct bio *bio,
eaf25d93
CM		 1004 int mirror_num, unsigned long bio_flags,
1005 u64 bio_offset)
44b8bd7e 1006{
de0022b9 1007 int async = check_async_write(inode, bio_flags);
cad321ad
CM		 1008 int ret;
1009
37226b21 1010 if (bio_op(bio) != REQ_OP_WRITE) {
4a69a410
CM		 1011 /*
1012 * called for a read, do the setup so that checksum validation
1013 * can happen in the async kernel threads
1014 */
f3f266ab 1015 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
bfebd8b5 1016 bio, BTRFS_WQ_ENDIO_METADATA);
1d4284bd 1017 if (ret)
61891923 1018 goto out_w_error;
81a75f67 1019 ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 0);
de0022b9
JB		 1020 } else if (!async) {
1021 ret = btree_csum_one_bio(bio);
1022 if (ret)
61891923 1023 goto out_w_error;
81a75f67 1024 ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 0);
61891923
SB		 1025 } else {
1026 /*
1027 * kthread helpers are used to submit writes so that
1028 * checksumming can happen in parallel across all CPUs
1029 */
1030 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
81a75f67 1031 inode, bio, mirror_num, 0,
61891923
SB		 1032 bio_offset,
1033 __btree_submit_bio_start,
1034 __btree_submit_bio_done);
44b8bd7e 1035 }
d313d7a3 1036
4246a0b6
CH		 1037 if (ret)
1038 goto out_w_error;
1039 return 0;
1040
61891923 1041out_w_error:
4246a0b6
CH		 1042 bio->bi_error = ret;
1043 bio_endio(bio);
61891923 1044 return ret;
44b8bd7e
CM		 1045}
1046
3dd1462e 1047#ifdef CONFIG_MIGRATION
784b4e29 1048static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG		 1049 struct page *newpage, struct page *page,
1050 enum migrate_mode mode)
784b4e29
CM		 1051{
1052 /*
1053 * we can't safely write a btree page from here,
1054 * we haven't done the locking hook
1055 */
1056 if (PageDirty(page))
1057 return -EAGAIN;
1058 /*
1059 * Buffers may be managed in a filesystem specific way.
1060 * We must have no buffers or drop them.
1061 */
1062 if (page_has_private(page) &&
1063 !try_to_release_page(page, GFP_KERNEL))
1064 return -EAGAIN;
a6bc32b8 1065 return migrate_page(mapping, newpage, page, mode);
784b4e29 1066}
3dd1462e 1067#endif
784b4e29 1068
0da5468f
CM		 1069
1070static int btree_writepages(struct address_space *mapping,
1071 struct writeback_control *wbc)
1072{
e2d84521
MX		 1073 struct btrfs_fs_info *fs_info;
1074 int ret;
1075
d8d5f3e1 1076 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM		 1077
1078 if (wbc->for_kupdate)
1079 return 0;
1080
e2d84521 1081 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 1082 /* this is a bit racy, but that's ok */
e2d84521
MX		 1083 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
1084 BTRFS_DIRTY_METADATA_THRESH);
1085 if (ret < 0)
793955bc 1086 return 0;
793955bc 1087 }
0b32f4bb 1088 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM		 1089}
1090
b2950863 1091static int btree_readpage(struct file *file, struct page *page)
5f39d397 1092{
d1310b2e
CM		 1093 struct extent_io_tree *tree;
1094 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 1095 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 1096}
22b0ebda 1097
70dec807 1098static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 1099{
98509cfc 1100 if (PageWriteback(page) || PageDirty(page))
d397712b 1101 return 0;
0c4e538b 1102
f7a52a40 1103 return try_release_extent_buffer(page);
d98237b3
CM		 1104}
1105
d47992f8
LC		 1106static void btree_invalidatepage(struct page *page, unsigned int offset,
1107 unsigned int length)
d98237b3 1108{
d1310b2e
CM		 1109 struct extent_io_tree *tree;
1110 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 1111 extent_invalidatepage(tree, page, offset);
1112 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 1113 if (PagePrivate(page)) {
efe120a0
FH		 1114 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1115 "page private not zero on page %llu",
1116 (unsigned long long)page_offset(page));
9ad6b7bc
CM		 1117 ClearPagePrivate(page);
1118 set_page_private(page, 0);
09cbfeaf 1119 put_page(page);
9ad6b7bc 1120 }
d98237b3
CM		 1121}
1122
0b32f4bb
JB		 1123static int btree_set_page_dirty(struct page *page)
1124{
bb146eb2 1125#ifdef DEBUG
0b32f4bb
JB		 1126 struct extent_buffer *eb;
1127
1128 BUG_ON(!PagePrivate(page));
1129 eb = (struct extent_buffer *)page->private;
1130 BUG_ON(!eb);
1131 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1132 BUG_ON(!atomic_read(&eb->refs));
1133 btrfs_assert_tree_locked(eb);
bb146eb2 1134#endif
0b32f4bb
JB		 1135 return __set_page_dirty_nobuffers(page);
1136}
1137
7f09410b 1138static const struct address_space_operations btree_aops = {
d98237b3 1139 .readpage = btree_readpage,
0da5468f 1140 .writepages = btree_writepages,
5f39d397
CM		 1141 .releasepage = btree_releasepage,
1142 .invalidatepage = btree_invalidatepage,
5a92bc88 1143#ifdef CONFIG_MIGRATION
784b4e29 1144 .migratepage = btree_migratepage,
5a92bc88 1145#endif
0b32f4bb 1146 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM		 1147};
1148
d3e46fea 1149void readahead_tree_block(struct btrfs_root *root, u64 bytenr)
090d1875 1150{
5f39d397
CM		 1151 struct extent_buffer *buf = NULL;
1152 struct inode *btree_inode = root->fs_info->btree_inode;
090d1875 1153
a83fffb7 1154 buf = btrfs_find_create_tree_block(root, bytenr);
c871b0f2 1155 if (IS_ERR(buf))
6197d86e 1156 return;
d1310b2e 1157 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
8436ea91 1158 buf, WAIT_NONE, btree_get_extent, 0);
5f39d397 1159 free_extent_buffer(buf);
090d1875
CM		 1160}
1161
c0dcaa4d 1162int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr,
ab0fff03
AJ		 1163 int mirror_num, struct extent_buffer **eb)
1164{
1165 struct extent_buffer *buf = NULL;
1166 struct inode *btree_inode = root->fs_info->btree_inode;
1167 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1168 int ret;
1169
a83fffb7 1170 buf = btrfs_find_create_tree_block(root, bytenr);
c871b0f2 1171 if (IS_ERR(buf))
ab0fff03
AJ		 1172 return 0;
1173
1174 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1175
8436ea91 1176 ret = read_extent_buffer_pages(io_tree, buf, WAIT_PAGE_LOCK,
ab0fff03
AJ		 1177 btree_get_extent, mirror_num);
1178 if (ret) {
1179 free_extent_buffer(buf);
1180 return ret;
1181 }
1182
1183 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1184 free_extent_buffer(buf);
1185 return -EIO;
0b32f4bb 1186 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ		 1187 *eb = buf;
1188 } else {
1189 free_extent_buffer(buf);
1190 }
1191 return 0;
1192}
1193
0999df54 1194struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
a83fffb7 1195 u64 bytenr)
0999df54 1196{
f5ee5c9a 1197 if (btrfs_is_testing(root->fs_info))
b9ef22de
FX		 1198 return alloc_test_extent_buffer(root->fs_info, bytenr,
1199 root->nodesize);
ce3e6984 1200 return alloc_extent_buffer(root->fs_info, bytenr);
0999df54
CM		 1201}
1202
1203
e02119d5
CM		 1204int btrfs_write_tree_block(struct extent_buffer *buf)
1205{
727011e0 1206 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1207 buf->start + buf->len - 1);
e02119d5
CM		 1208}
1209
1210int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1211{
727011e0 1212 return filemap_fdatawait_range(buf->pages[0]->mapping,
8aa38c31 1213 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1214}
1215
0999df54 1216struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ce86cd59 1217 u64 parent_transid)
0999df54
CM		 1218{
1219 struct extent_buffer *buf = NULL;
0999df54
CM		 1220 int ret;
1221
a83fffb7 1222 buf = btrfs_find_create_tree_block(root, bytenr);
c871b0f2
LB		 1223 if (IS_ERR(buf))
1224 return buf;
0999df54 1225
8436ea91 1226 ret = btree_read_extent_buffer_pages(root, buf, parent_transid);
0f0fe8f7
FDBM		 1227 if (ret) {
1228 free_extent_buffer(buf);
64c043de 1229 return ERR_PTR(ret);
0f0fe8f7 1230 }
5f39d397 1231 return buf;
ce9adaa5 1232
eb60ceac
CM		 1233}
1234
01d58472
DD		 1235void clean_tree_block(struct btrfs_trans_handle *trans,
1236 struct btrfs_fs_info *fs_info,
d5c13f92 1237 struct extent_buffer *buf)
ed2ff2cb 1238{
55c69072 1239 if (btrfs_header_generation(buf) ==
e2d84521 1240 fs_info->running_transaction->transid) {
b9447ef8 1241 btrfs_assert_tree_locked(buf);
b4ce94de 1242
b9473439 1243 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
e2d84521
MX		 1244 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1245 -buf->len,
1246 fs_info->dirty_metadata_batch);
ed7b63eb
JB		 1247 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1248 btrfs_set_lock_blocking(buf);
1249 clear_extent_buffer_dirty(buf);
1250 }
925baedd 1251 }
5f39d397
CM		 1252}
1253
8257b2dc
MX		 1254static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1255{
1256 struct btrfs_subvolume_writers *writers;
1257 int ret;
1258
1259 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1260 if (!writers)
1261 return ERR_PTR(-ENOMEM);
1262
908c7f19 1263 ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL);
8257b2dc
MX		 1264 if (ret < 0) {
1265 kfree(writers);
1266 return ERR_PTR(ret);
1267 }
1268
1269 init_waitqueue_head(&writers->wait);
1270 return writers;
1271}
1272
1273static void
1274btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1275{
1276 percpu_counter_destroy(&writers->counter);
1277 kfree(writers);
1278}
1279
707e8a07
DS		 1280static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize,
1281 struct btrfs_root *root, struct btrfs_fs_info *fs_info,
143bede5 1282 u64 objectid)
d97e63b6 1283{
7c0260ee 1284 bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state);
cfaa7295 1285 root->node = NULL;
a28ec197 1286 root->commit_root = NULL;
db94535d
CM		 1287 root->sectorsize = sectorsize;
1288 root->nodesize = nodesize;
87ee04eb 1289 root->stripesize = stripesize;
27cdeb70 1290 root->state = 0;
d68fc57b 1291 root->orphan_cleanup_state = 0;
0b86a832 1292
0f7d52f4
CM		 1293 root->objectid = objectid;
1294 root->last_trans = 0;
13a8a7c8 1295 root->highest_objectid = 0;
eb73c1b7 1296 root->nr_delalloc_inodes = 0;
199c2a9c 1297 root->nr_ordered_extents = 0;
58176a96 1298 root->name = NULL;
6bef4d31 1299 root->inode_tree = RB_ROOT;
16cdcec7 1300 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1301 root->block_rsv = NULL;
d68fc57b 1302 root->orphan_block_rsv = NULL;
0b86a832
CM		 1303
1304 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1305 INIT_LIST_HEAD(&root->root_list);
eb73c1b7
MX		 1306 INIT_LIST_HEAD(&root->delalloc_inodes);
1307 INIT_LIST_HEAD(&root->delalloc_root);
199c2a9c
MX		 1308 INIT_LIST_HEAD(&root->ordered_extents);
1309 INIT_LIST_HEAD(&root->ordered_root);
2ab28f32
JB		 1310 INIT_LIST_HEAD(&root->logged_list[0]);
1311 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1312 spin_lock_init(&root->orphan_lock);
5d4f98a2 1313 spin_lock_init(&root->inode_lock);
eb73c1b7 1314 spin_lock_init(&root->delalloc_lock);
199c2a9c 1315 spin_lock_init(&root->ordered_extent_lock);
f0486c68 1316 spin_lock_init(&root->accounting_lock);
2ab28f32
JB		 1317 spin_lock_init(&root->log_extents_lock[0]);
1318 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1319 mutex_init(&root->objectid_mutex);
e02119d5 1320 mutex_init(&root->log_mutex);
31f3d255 1321 mutex_init(&root->ordered_extent_mutex);
573bfb72 1322 mutex_init(&root->delalloc_mutex);
7237f183
YZ		 1323 init_waitqueue_head(&root->log_writer_wait);
1324 init_waitqueue_head(&root->log_commit_wait[0]);
1325 init_waitqueue_head(&root->log_commit_wait[1]);
8b050d35
MX		 1326 INIT_LIST_HEAD(&root->log_ctxs[0]);
1327 INIT_LIST_HEAD(&root->log_ctxs[1]);
7237f183
YZ		 1328 atomic_set(&root->log_commit[0], 0);
1329 atomic_set(&root->log_commit[1], 0);
1330 atomic_set(&root->log_writers, 0);
2ecb7923 1331 atomic_set(&root->log_batch, 0);
8a35d95f 1332 atomic_set(&root->orphan_inodes, 0);
b0feb9d9 1333 atomic_set(&root->refs, 1);
8257b2dc 1334 atomic_set(&root->will_be_snapshoted, 0);
55eeaf05 1335 atomic_set(&root->qgroup_meta_rsv, 0);
7237f183 1336 root->log_transid = 0;
d1433deb 1337 root->log_transid_committed = -1;
257c62e1 1338 root->last_log_commit = 0;
7c0260ee 1339 if (!dummy)
06ea65a3
JB		 1340 extent_io_tree_init(&root->dirty_log_pages,
1341 fs_info->btree_inode->i_mapping);
017e5369 1342
3768f368
CM		 1343 memset(&root->root_key, 0, sizeof(root->root_key));
1344 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1345 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
7c0260ee 1346 if (!dummy)
06ea65a3
JB		 1347 root->defrag_trans_start = fs_info->generation;
1348 else
1349 root->defrag_trans_start = 0;
4d775673 1350 root->root_key.objectid = objectid;
0ee5dc67 1351 root->anon_dev = 0;
8ea05e3a 1352
5f3ab90a 1353 spin_lock_init(&root->root_item_lock);
3768f368
CM		 1354}
1355
74e4d827
DS		 1356static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info,
1357 gfp_t flags)
6f07e42e 1358{
74e4d827 1359 struct btrfs_root *root = kzalloc(sizeof(*root), flags);
6f07e42e
AV		 1360 if (root)
1361 root->fs_info = fs_info;
1362 return root;
1363}
1364
06ea65a3
JB		 1365#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1366/* Should only be used by the testing infrastructure */
7c0260ee
JM		 1367struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info,
1368 u32 sectorsize, u32 nodesize)
06ea65a3
JB		 1369{
1370 struct btrfs_root *root;
1371
7c0260ee
JM		 1372 if (!fs_info)
1373 return ERR_PTR(-EINVAL);
1374
1375 root = btrfs_alloc_root(fs_info, GFP_KERNEL);
06ea65a3
JB		 1376 if (!root)
1377 return ERR_PTR(-ENOMEM);
b9ef22de 1378 /* We don't use the stripesize in selftest, set it as sectorsize */
7c0260ee 1379 __setup_root(nodesize, sectorsize, sectorsize, root, fs_info,
ef9f2db3 1380 BTRFS_ROOT_TREE_OBJECTID);
faa2dbf0 1381 root->alloc_bytenr = 0;
06ea65a3
JB		 1382
1383 return root;
1384}
1385#endif
1386
20897f5c
AJ		 1387struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1388 struct btrfs_fs_info *fs_info,
1389 u64 objectid)
1390{
1391 struct extent_buffer *leaf;
1392 struct btrfs_root *tree_root = fs_info->tree_root;
1393 struct btrfs_root *root;
1394 struct btrfs_key key;
1395 int ret = 0;
6463fe58 1396 uuid_le uuid;
20897f5c 1397
74e4d827 1398 root = btrfs_alloc_root(fs_info, GFP_KERNEL);
20897f5c
AJ		 1399 if (!root)
1400 return ERR_PTR(-ENOMEM);
1401
707e8a07
DS		 1402 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1403 tree_root->stripesize, root, fs_info, objectid);
20897f5c
AJ		 1404 root->root_key.objectid = objectid;
1405 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1406 root->root_key.offset = 0;
1407
4d75f8a9 1408 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
20897f5c
AJ		 1409 if (IS_ERR(leaf)) {
1410 ret = PTR_ERR(leaf);
1dd05682 1411 leaf = NULL;
20897f5c
AJ		 1412 goto fail;
1413 }
1414
20897f5c
AJ		 1415 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1416 btrfs_set_header_bytenr(leaf, leaf->start);
1417 btrfs_set_header_generation(leaf, trans->transid);
1418 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1419 btrfs_set_header_owner(leaf, objectid);
1420 root->node = leaf;
1421
0a4e5586 1422 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
20897f5c
AJ		 1423 BTRFS_FSID_SIZE);
1424 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
b308bc2f 1425 btrfs_header_chunk_tree_uuid(leaf),
20897f5c
AJ		 1426 BTRFS_UUID_SIZE);
1427 btrfs_mark_buffer_dirty(leaf);
1428
1429 root->commit_root = btrfs_root_node(root);
27cdeb70 1430 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
20897f5c
AJ		 1431
1432 root->root_item.flags = 0;
1433 root->root_item.byte_limit = 0;
1434 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1435 btrfs_set_root_generation(&root->root_item, trans->transid);
1436 btrfs_set_root_level(&root->root_item, 0);
1437 btrfs_set_root_refs(&root->root_item, 1);
1438 btrfs_set_root_used(&root->root_item, leaf->len);
1439 btrfs_set_root_last_snapshot(&root->root_item, 0);
1440 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB		 1441 uuid_le_gen(&uuid);
1442 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ		 1443 root->root_item.drop_level = 0;
1444
1445 key.objectid = objectid;
1446 key.type = BTRFS_ROOT_ITEM_KEY;
1447 key.offset = 0;
1448 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1449 if (ret)
1450 goto fail;
1451
1452 btrfs_tree_unlock(leaf);
1453
1dd05682
TI		 1454 return root;
1455
20897f5c 1456fail:
1dd05682
TI		 1457 if (leaf) {
1458 btrfs_tree_unlock(leaf);
59885b39 1459 free_extent_buffer(root->commit_root);
1dd05682
TI		 1460 free_extent_buffer(leaf);
1461 }
1462 kfree(root);
20897f5c 1463
1dd05682 1464 return ERR_PTR(ret);
20897f5c
AJ		 1465}
1466
7237f183
YZ		 1467static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1468 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1469{
1470 struct btrfs_root *root;
1471 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1472 struct extent_buffer *leaf;
e02119d5 1473
74e4d827 1474 root = btrfs_alloc_root(fs_info, GFP_NOFS);
e02119d5 1475 if (!root)
7237f183 1476 return ERR_PTR(-ENOMEM);
e02119d5 1477
707e8a07
DS		 1478 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1479 tree_root->stripesize, root, fs_info,
1480 BTRFS_TREE_LOG_OBJECTID);
e02119d5
CM		 1481
1482 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1483 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1484 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
27cdeb70 1485
7237f183 1486 /*
27cdeb70
MX		 1487 * DON'T set REF_COWS for log trees
1488 *
7237f183
YZ		 1489 * log trees do not get reference counted because they go away
1490 * before a real commit is actually done. They do store pointers
1491 * to file data extents, and those reference counts still get
1492 * updated (along with back refs to the log tree).
1493 */
e02119d5 1494
4d75f8a9
DS		 1495 leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID,
1496 NULL, 0, 0, 0);
7237f183
YZ		 1497 if (IS_ERR(leaf)) {
1498 kfree(root);
1499 return ERR_CAST(leaf);
1500 }
e02119d5 1501
5d4f98a2
YZ		 1502 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1503 btrfs_set_header_bytenr(leaf, leaf->start);
1504 btrfs_set_header_generation(leaf, trans->transid);
1505 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1506 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1507 root->node = leaf;
e02119d5
CM		 1508
1509 write_extent_buffer(root->node, root->fs_info->fsid,
0a4e5586 1510 btrfs_header_fsid(), BTRFS_FSID_SIZE);
e02119d5
CM		 1511 btrfs_mark_buffer_dirty(root->node);
1512 btrfs_tree_unlock(root->node);
7237f183
YZ		 1513 return root;
1514}
1515
1516int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1517 struct btrfs_fs_info *fs_info)
1518{
1519 struct btrfs_root *log_root;
1520
1521 log_root = alloc_log_tree(trans, fs_info);
1522 if (IS_ERR(log_root))
1523 return PTR_ERR(log_root);
1524 WARN_ON(fs_info->log_root_tree);
1525 fs_info->log_root_tree = log_root;
1526 return 0;
1527}
1528
1529int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1530 struct btrfs_root *root)
1531{
1532 struct btrfs_root *log_root;
1533 struct btrfs_inode_item *inode_item;
1534
1535 log_root = alloc_log_tree(trans, root->fs_info);
1536 if (IS_ERR(log_root))
1537 return PTR_ERR(log_root);
1538
1539 log_root->last_trans = trans->transid;
1540 log_root->root_key.offset = root->root_key.objectid;
1541
1542 inode_item = &log_root->root_item.inode;
3cae210f
QW		 1543 btrfs_set_stack_inode_generation(inode_item, 1);
1544 btrfs_set_stack_inode_size(inode_item, 3);
1545 btrfs_set_stack_inode_nlink(inode_item, 1);
707e8a07 1546 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
3cae210f 1547 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
7237f183 1548
5d4f98a2 1549 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1550
1551 WARN_ON(root->log_root);
1552 root->log_root = log_root;
1553 root->log_transid = 0;
d1433deb 1554 root->log_transid_committed = -1;
257c62e1 1555 root->last_log_commit = 0;
e02119d5
CM		 1556 return 0;
1557}
1558
35a3621b
SB		 1559static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1560 struct btrfs_key *key)
e02119d5
CM		 1561{
1562 struct btrfs_root *root;
1563 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1564 struct btrfs_path *path;
84234f3a 1565 u64 generation;
cb517eab 1566 int ret;
0f7d52f4 1567
cb517eab
MX		 1568 path = btrfs_alloc_path();
1569 if (!path)
0f7d52f4 1570 return ERR_PTR(-ENOMEM);
cb517eab 1571
74e4d827 1572 root = btrfs_alloc_root(fs_info, GFP_NOFS);
cb517eab
MX		 1573 if (!root) {
1574 ret = -ENOMEM;
1575 goto alloc_fail;
0f7d52f4
CM		 1576 }
1577
707e8a07
DS		 1578 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1579 tree_root->stripesize, root, fs_info, key->objectid);
0f7d52f4 1580
cb517eab
MX		 1581 ret = btrfs_find_root(tree_root, key, path,
1582 &root->root_item, &root->root_key);
0f7d52f4 1583 if (ret) {
13a8a7c8
YZ		 1584 if (ret > 0)
1585 ret = -ENOENT;
cb517eab 1586 goto find_fail;
0f7d52f4 1587 }
13a8a7c8 1588
84234f3a 1589 generation = btrfs_root_generation(&root->root_item);
db94535d 1590 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
ce86cd59 1591 generation);
64c043de
LB		 1592 if (IS_ERR(root->node)) {
1593 ret = PTR_ERR(root->node);
cb517eab
MX		 1594 goto find_fail;
1595 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1596 ret = -EIO;
64c043de
LB		 1597 free_extent_buffer(root->node);
1598 goto find_fail;
416bc658 1599 }
5d4f98a2 1600 root->commit_root = btrfs_root_node(root);
13a8a7c8 1601out:
cb517eab
MX		 1602 btrfs_free_path(path);
1603 return root;
1604
cb517eab
MX		 1605find_fail:
1606 kfree(root);
1607alloc_fail:
1608 root = ERR_PTR(ret);
1609 goto out;
1610}
1611
1612struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1613 struct btrfs_key *location)
1614{
1615 struct btrfs_root *root;
1616
1617 root = btrfs_read_tree_root(tree_root, location);
1618 if (IS_ERR(root))
1619 return root;
1620
1621 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
27cdeb70 1622 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
08fe4db1
LZ		 1623 btrfs_check_and_init_root_item(&root->root_item);
1624 }
13a8a7c8 1625
5eda7b5e
CM		 1626 return root;
1627}
1628
cb517eab
MX		 1629int btrfs_init_fs_root(struct btrfs_root *root)
1630{
1631 int ret;
8257b2dc 1632 struct btrfs_subvolume_writers *writers;
cb517eab
MX		 1633
1634 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1635 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1636 GFP_NOFS);
1637 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1638 ret = -ENOMEM;
1639 goto fail;
1640 }
1641
8257b2dc
MX		 1642 writers = btrfs_alloc_subvolume_writers();
1643 if (IS_ERR(writers)) {
1644 ret = PTR_ERR(writers);
1645 goto fail;
1646 }
1647 root->subv_writers = writers;
1648
cb517eab 1649 btrfs_init_free_ino_ctl(root);
57cdc8db
DS		 1650 spin_lock_init(&root->ino_cache_lock);
1651 init_waitqueue_head(&root->ino_cache_wait);
cb517eab
MX		 1652
1653 ret = get_anon_bdev(&root->anon_dev);
1654 if (ret)
876d2cf1 1655 goto fail;
f32e48e9
CR		 1656
1657 mutex_lock(&root->objectid_mutex);
1658 ret = btrfs_find_highest_objectid(root,
1659 &root->highest_objectid);
1660 if (ret) {
1661 mutex_unlock(&root->objectid_mutex);
876d2cf1 1662 goto fail;
f32e48e9
CR		 1663 }
1664
1665 ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID);
1666
1667 mutex_unlock(&root->objectid_mutex);
1668
cb517eab
MX		 1669 return 0;
1670fail:
876d2cf1 1671 /* the caller is responsible to call free_fs_root */
cb517eab
MX		 1672 return ret;
1673}
1674
35bbb97f
JM		 1675struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1676 u64 root_id)
cb517eab
MX		 1677{
1678 struct btrfs_root *root;
1679
1680 spin_lock(&fs_info->fs_roots_radix_lock);
1681 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1682 (unsigned long)root_id);
1683 spin_unlock(&fs_info->fs_roots_radix_lock);
1684 return root;
1685}
1686
1687int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1688 struct btrfs_root *root)
1689{
1690 int ret;
1691
e1860a77 1692 ret = radix_tree_preload(GFP_NOFS);
cb517eab
MX		 1693 if (ret)
1694 return ret;
1695
1696 spin_lock(&fs_info->fs_roots_radix_lock);
1697 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1698 (unsigned long)root->root_key.objectid,
1699 root);
1700 if (ret == 0)
27cdeb70 1701 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
cb517eab
MX		 1702 spin_unlock(&fs_info->fs_roots_radix_lock);
1703 radix_tree_preload_end();
1704
1705 return ret;
1706}
1707
c00869f1
MX		 1708struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1709 struct btrfs_key *location,
1710 bool check_ref)
5eda7b5e
CM		 1711{
1712 struct btrfs_root *root;
381cf658 1713 struct btrfs_path *path;
1d4c08e0 1714 struct btrfs_key key;
5eda7b5e
CM		 1715 int ret;
1716
edbd8d4e
CM		 1717 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1718 return fs_info->tree_root;
1719 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1720 return fs_info->extent_root;
8f18cf13
CM		 1721 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1722 return fs_info->chunk_root;
1723 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1724 return fs_info->dev_root;
0403e47e
YZ		 1725 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1726 return fs_info->csum_root;
bcef60f2
AJ		 1727 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1728 return fs_info->quota_root ? fs_info->quota_root :
1729 ERR_PTR(-ENOENT);
f7a81ea4
SB		 1730 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1731 return fs_info->uuid_root ? fs_info->uuid_root :
1732 ERR_PTR(-ENOENT);
70f6d82e
OS		 1733 if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID)
1734 return fs_info->free_space_root ? fs_info->free_space_root :
1735 ERR_PTR(-ENOENT);
4df27c4d 1736again:
cb517eab 1737 root = btrfs_lookup_fs_root(fs_info, location->objectid);
48475471 1738 if (root) {
c00869f1 1739 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
48475471 1740 return ERR_PTR(-ENOENT);
5eda7b5e 1741 return root;
48475471 1742 }
5eda7b5e 1743
cb517eab 1744 root = btrfs_read_fs_root(fs_info->tree_root, location);
5eda7b5e
CM		 1745 if (IS_ERR(root))
1746 return root;
3394e160 1747
c00869f1 1748 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
cb517eab 1749 ret = -ENOENT;
581bb050 1750 goto fail;
35a30d7c 1751 }
581bb050 1752
cb517eab 1753 ret = btrfs_init_fs_root(root);
ac08aedf
CM		 1754 if (ret)
1755 goto fail;
3394e160 1756
381cf658
DS		 1757 path = btrfs_alloc_path();
1758 if (!path) {
1759 ret = -ENOMEM;
1760 goto fail;
1761 }
1d4c08e0
DS		 1762 key.objectid = BTRFS_ORPHAN_OBJECTID;
1763 key.type = BTRFS_ORPHAN_ITEM_KEY;
1764 key.offset = location->objectid;
1765
1766 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
381cf658 1767 btrfs_free_path(path);
d68fc57b
YZ		 1768 if (ret < 0)
1769 goto fail;
1770 if (ret == 0)
27cdeb70 1771 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
d68fc57b 1772
cb517eab 1773 ret = btrfs_insert_fs_root(fs_info, root);
0f7d52f4 1774 if (ret) {
4df27c4d
YZ		 1775 if (ret == -EEXIST) {
1776 free_fs_root(root);
1777 goto again;
1778 }
1779 goto fail;
0f7d52f4 1780 }
edbd8d4e 1781 return root;
4df27c4d
YZ		 1782fail:
1783 free_fs_root(root);
1784 return ERR_PTR(ret);
edbd8d4e
CM		 1785}
1786
04160088
CM		 1787static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1788{
1789 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1790 int ret = 0;
04160088
CM		 1791 struct btrfs_device *device;
1792 struct backing_dev_info *bdi;
b7967db7 1793
1f78160c
XG		 1794 rcu_read_lock();
1795 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1796 if (!device->bdev)
1797 continue;
04160088 1798 bdi = blk_get_backing_dev_info(device->bdev);
ff9ea323 1799 if (bdi_congested(bdi, bdi_bits)) {
04160088
CM		 1800 ret = 1;
1801 break;
1802 }
1803 }
1f78160c 1804 rcu_read_unlock();
04160088
CM		 1805 return ret;
1806}
1807
04160088
CM		 1808static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1809{
ad081f14
JA		 1810 int err;
1811
b4caecd4 1812 err = bdi_setup_and_register(bdi, "btrfs");
ad081f14
JA		 1813 if (err)
1814 return err;
1815
09cbfeaf 1816 bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_SIZE;
04160088
CM		 1817 bdi->congested_fn = btrfs_congested_fn;
1818 bdi->congested_data = info;
da2f0f74 1819 bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK;
04160088
CM		 1820 return 0;
1821}
1822
8b712842
CM		 1823/*
1824 * called by the kthread helper functions to finally call the bio end_io
1825 * functions. This is where read checksum verification actually happens
1826 */
1827static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1828{
ce9adaa5 1829 struct bio *bio;
97eb6b69 1830 struct btrfs_end_io_wq *end_io_wq;
ce9adaa5 1831
97eb6b69 1832 end_io_wq = container_of(work, struct btrfs_end_io_wq, work);
8b712842 1833 bio = end_io_wq->bio;
ce9adaa5 1834
4246a0b6 1835 bio->bi_error = end_io_wq->error;
8b712842
CM		 1836 bio->bi_private = end_io_wq->private;
1837 bio->bi_end_io = end_io_wq->end_io;
97eb6b69 1838 kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq);
4246a0b6 1839 bio_endio(bio);
44b8bd7e
CM		 1840}
1841
a74a4b97
CM		 1842static int cleaner_kthread(void *arg)
1843{
1844 struct btrfs_root *root = arg;
d0278245 1845 int again;
da288d28 1846 struct btrfs_trans_handle *trans;
a74a4b97
CM		 1847
1848 do {
d0278245 1849 again = 0;
a74a4b97 1850
d0278245 1851 /* Make the cleaner go to sleep early. */
babbf170 1852 if (btrfs_need_cleaner_sleep(root))
d0278245
MX		 1853 goto sleep;
1854
90c711ab
ZB		 1855 /*
1856 * Do not do anything if we might cause open_ctree() to block
1857 * before we have finished mounting the filesystem.
1858 */
afcdd129 1859 if (!test_bit(BTRFS_FS_OPEN, &root->fs_info->flags))
90c711ab
ZB		 1860 goto sleep;
1861
d0278245
MX		 1862 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1863 goto sleep;
1864
dc7f370c
MX		 1865 /*
1866 * Avoid the problem that we change the status of the fs
1867 * during the above check and trylock.
1868 */
babbf170 1869 if (btrfs_need_cleaner_sleep(root)) {
dc7f370c
MX		 1870 mutex_unlock(&root->fs_info->cleaner_mutex);
1871 goto sleep;
76dda93c 1872 }
a74a4b97 1873
c2d6cb16 1874 mutex_lock(&root->fs_info->cleaner_delayed_iput_mutex);
d0278245 1875 btrfs_run_delayed_iputs(root);
c2d6cb16
FM		 1876 mutex_unlock(&root->fs_info->cleaner_delayed_iput_mutex);
1877
d0278245
MX		 1878 again = btrfs_clean_one_deleted_snapshot(root);
1879 mutex_unlock(&root->fs_info->cleaner_mutex);
1880
1881 /*
05323cd1
MX		 1882 * The defragger has dealt with the R/O remount and umount,
1883 * needn't do anything special here.
d0278245
MX		 1884 */
1885 btrfs_run_defrag_inodes(root->fs_info);
67c5e7d4
FM		 1886
1887 /*
1888 * Acquires fs_info->delete_unused_bgs_mutex to avoid racing
1889 * with relocation (btrfs_relocate_chunk) and relocation
1890 * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group)
1891 * after acquiring fs_info->delete_unused_bgs_mutex. So we
1892 * can't hold, nor need to, fs_info->cleaner_mutex when deleting
1893 * unused block groups.
1894 */
1895 btrfs_delete_unused_bgs(root->fs_info);
d0278245 1896sleep:
838fe188 1897 if (!again) {
a74a4b97 1898 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1899 if (!kthread_should_stop())
1900 schedule();
a74a4b97
CM		 1901 __set_current_state(TASK_RUNNING);
1902 }
1903 } while (!kthread_should_stop());
da288d28
FM		 1904
1905 /*
1906 * Transaction kthread is stopped before us and wakes us up.
1907 * However we might have started a new transaction and COWed some
1908 * tree blocks when deleting unused block groups for example. So
1909 * make sure we commit the transaction we started to have a clean
1910 * shutdown when evicting the btree inode - if it has dirty pages
1911 * when we do the final iput() on it, eviction will trigger a
1912 * writeback for it which will fail with null pointer dereferences
1913 * since work queues and other resources were already released and
1914 * destroyed by the time the iput/eviction/writeback is made.
1915 */
1916 trans = btrfs_attach_transaction(root);
1917 if (IS_ERR(trans)) {
1918 if (PTR_ERR(trans) != -ENOENT)
1919 btrfs_err(root->fs_info,
1920 "cleaner transaction attach returned %ld",
1921 PTR_ERR(trans));
1922 } else {
1923 int ret;
1924
1925 ret = btrfs_commit_transaction(trans, root);
1926 if (ret)
1927 btrfs_err(root->fs_info,
1928 "cleaner open transaction commit returned %d",
1929 ret);
1930 }
1931
a74a4b97
CM		 1932 return 0;
1933}
1934
1935static int transaction_kthread(void *arg)
1936{
1937 struct btrfs_root *root = arg;
1938 struct btrfs_trans_handle *trans;
1939 struct btrfs_transaction *cur;
8929ecfa 1940 u64 transid;
a74a4b97
CM		 1941 unsigned long now;
1942 unsigned long delay;
914b2007 1943 bool cannot_commit;
a74a4b97
CM		 1944
1945 do {
914b2007 1946 cannot_commit = false;
8b87dc17 1947 delay = HZ * root->fs_info->commit_interval;
a74a4b97
CM		 1948 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1949
a4abeea4 1950 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1951 cur = root->fs_info->running_transaction;
1952 if (!cur) {
a4abeea4 1953 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1954 goto sleep;
1955 }
31153d81 1956
a74a4b97 1957 now = get_seconds();
4a9d8bde 1958 if (cur->state < TRANS_STATE_BLOCKED &&
8b87dc17
DS		 1959 (now < cur->start_time ||
1960 now - cur->start_time < root->fs_info->commit_interval)) {
a4abeea4 1961 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1962 delay = HZ * 5;
1963 goto sleep;
1964 }
8929ecfa 1965 transid = cur->transid;
a4abeea4 1966 spin_unlock(&root->fs_info->trans_lock);
56bec294 1967
79787eaa 1968 /* If the file system is aborted, this will always fail. */
354aa0fb 1969 trans = btrfs_attach_transaction(root);
914b2007 1970 if (IS_ERR(trans)) {
354aa0fb
MX		 1971 if (PTR_ERR(trans) != -ENOENT)
1972 cannot_commit = true;
79787eaa 1973 goto sleep;
914b2007 1974 }
8929ecfa 1975 if (transid == trans->transid) {
79787eaa 1976 btrfs_commit_transaction(trans, root);
8929ecfa
YZ		 1977 } else {
1978 btrfs_end_transaction(trans, root);
1979 }
a74a4b97
CM		 1980sleep:
1981 wake_up_process(root->fs_info->cleaner_kthread);
1982 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1983
4e121c06
JB		 1984 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1985 &root->fs_info->fs_state)))
1986 btrfs_cleanup_transaction(root);
ce63f891
JK		 1987 set_current_state(TASK_INTERRUPTIBLE);
1988 if (!kthread_should_stop() &&
1989 (!btrfs_transaction_blocked(root->fs_info) ||
1990 cannot_commit))
1991 schedule_timeout(delay);
1992 __set_current_state(TASK_RUNNING);
a74a4b97
CM		 1993 } while (!kthread_should_stop());
1994 return 0;
1995}
1996
af31f5e5
CM		 1997/*
1998 * this will find the highest generation in the array of
1999 * root backups. The index of the highest array is returned,
2000 * or -1 if we can't find anything.
2001 *
2002 * We check to make sure the array is valid by comparing the
2003 * generation of the latest root in the array with the generation
2004 * in the super block. If they don't match we pitch it.
2005 */
2006static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
2007{
2008 u64 cur;
2009 int newest_index = -1;
2010 struct btrfs_root_backup *root_backup;
2011 int i;
2012
2013 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
2014 root_backup = info->super_copy->super_roots + i;
2015 cur = btrfs_backup_tree_root_gen(root_backup);
2016 if (cur == newest_gen)
2017 newest_index = i;
2018 }
2019
2020 /* check to see if we actually wrapped around */
2021 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
2022 root_backup = info->super_copy->super_roots;
2023 cur = btrfs_backup_tree_root_gen(root_backup);
2024 if (cur == newest_gen)
2025 newest_index = 0;
2026 }
2027 return newest_index;
2028}
2029
2030
2031/*
2032 * find the oldest backup so we know where to store new entries
2033 * in the backup array. This will set the backup_root_index
2034 * field in the fs_info struct
2035 */
2036static void find_oldest_super_backup(struct btrfs_fs_info *info,
2037 u64 newest_gen)
2038{
2039 int newest_index = -1;
2040
2041 newest_index = find_newest_super_backup(info, newest_gen);
2042 /* if there was garbage in there, just move along */
2043 if (newest_index == -1) {
2044 info->backup_root_index = 0;
2045 } else {
2046 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
2047 }
2048}
2049
2050/*
2051 * copy all the root pointers into the super backup array.
2052 * this will bump the backup pointer by one when it is
2053 * done
2054 */
2055static void backup_super_roots(struct btrfs_fs_info *info)
2056{
2057 int next_backup;
2058 struct btrfs_root_backup *root_backup;
2059 int last_backup;
2060
2061 next_backup = info->backup_root_index;
2062 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
2063 BTRFS_NUM_BACKUP_ROOTS;
2064
2065 /*
2066 * just overwrite the last backup if we're at the same generation
2067 * this happens only at umount
2068 */
2069 root_backup = info->super_for_commit->super_roots + last_backup;
2070 if (btrfs_backup_tree_root_gen(root_backup) ==
2071 btrfs_header_generation(info->tree_root->node))
2072 next_backup = last_backup;
2073
2074 root_backup = info->super_for_commit->super_roots + next_backup;
2075
2076 /*
2077 * make sure all of our padding and empty slots get zero filled
2078 * regardless of which ones we use today
2079 */
2080 memset(root_backup, 0, sizeof(*root_backup));
2081
2082 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
2083
2084 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
2085 btrfs_set_backup_tree_root_gen(root_backup,
2086 btrfs_header_generation(info->tree_root->node));
2087
2088 btrfs_set_backup_tree_root_level(root_backup,
2089 btrfs_header_level(info->tree_root->node));
2090
2091 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
2092 btrfs_set_backup_chunk_root_gen(root_backup,
2093 btrfs_header_generation(info->chunk_root->node));
2094 btrfs_set_backup_chunk_root_level(root_backup,
2095 btrfs_header_level(info->chunk_root->node));
2096
2097 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
2098 btrfs_set_backup_extent_root_gen(root_backup,
2099 btrfs_header_generation(info->extent_root->node));
2100 btrfs_set_backup_extent_root_level(root_backup,
2101 btrfs_header_level(info->extent_root->node));
2102
7c7e82a7
CM		 2103 /*
2104 * we might commit during log recovery, which happens before we set
2105 * the fs_root. Make sure it is valid before we fill it in.
2106 */
2107 if (info->fs_root && info->fs_root->node) {
2108 btrfs_set_backup_fs_root(root_backup,
2109 info->fs_root->node->start);
2110 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 2111 btrfs_header_generation(info->fs_root->node));
7c7e82a7 2112 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 2113 btrfs_header_level(info->fs_root->node));
7c7e82a7 2114 }
af31f5e5
CM		 2115
2116 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
2117 btrfs_set_backup_dev_root_gen(root_backup,
2118 btrfs_header_generation(info->dev_root->node));
2119 btrfs_set_backup_dev_root_level(root_backup,
2120 btrfs_header_level(info->dev_root->node));
2121
2122 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
2123 btrfs_set_backup_csum_root_gen(root_backup,
2124 btrfs_header_generation(info->csum_root->node));
2125 btrfs_set_backup_csum_root_level(root_backup,
2126 btrfs_header_level(info->csum_root->node));
2127
2128 btrfs_set_backup_total_bytes(root_backup,
2129 btrfs_super_total_bytes(info->super_copy));
2130 btrfs_set_backup_bytes_used(root_backup,
2131 btrfs_super_bytes_used(info->super_copy));
2132 btrfs_set_backup_num_devices(root_backup,
2133 btrfs_super_num_devices(info->super_copy));
2134
2135 /*
2136 * if we don't copy this out to the super_copy, it won't get remembered
2137 * for the next commit
2138 */
2139 memcpy(&info->super_copy->super_roots,
2140 &info->super_for_commit->super_roots,
2141 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
2142}
2143
2144/*
2145 * this copies info out of the root backup array and back into
2146 * the in-memory super block. It is meant to help iterate through
2147 * the array, so you send it the number of backups you've already
2148 * tried and the last backup index you used.
2149 *
2150 * this returns -1 when it has tried all the backups
2151 */
2152static noinline int next_root_backup(struct btrfs_fs_info *info,
2153 struct btrfs_super_block *super,
2154 int *num_backups_tried, int *backup_index)
2155{
2156 struct btrfs_root_backup *root_backup;
2157 int newest = *backup_index;
2158
2159 if (*num_backups_tried == 0) {
2160 u64 gen = btrfs_super_generation(super);
2161
2162 newest = find_newest_super_backup(info, gen);
2163 if (newest == -1)
2164 return -1;
2165
2166 *backup_index = newest;
2167 *num_backups_tried = 1;
2168 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2169 /* we've tried all the backups, all done */
2170 return -1;
2171 } else {
2172 /* jump to the next oldest backup */
2173 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2174 BTRFS_NUM_BACKUP_ROOTS;
2175 *backup_index = newest;
2176 *num_backups_tried += 1;
2177 }
2178 root_backup = super->super_roots + newest;
2179
2180 btrfs_set_super_generation(super,
2181 btrfs_backup_tree_root_gen(root_backup));
2182 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2183 btrfs_set_super_root_level(super,
2184 btrfs_backup_tree_root_level(root_backup));
2185 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2186
2187 /*
2188 * fixme: the total bytes and num_devices need to match or we should
2189 * need a fsck
2190 */
2191 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2192 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2193 return 0;
2194}
2195
7abadb64
LB		 2196/* helper to cleanup workers */
2197static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2198{
dc6e3209 2199 btrfs_destroy_workqueue(fs_info->fixup_workers);
afe3d242 2200 btrfs_destroy_workqueue(fs_info->delalloc_workers);
5cdc7ad3 2201 btrfs_destroy_workqueue(fs_info->workers);
fccb5d86
QW		 2202 btrfs_destroy_workqueue(fs_info->endio_workers);
2203 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2204 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
8b110e39 2205 btrfs_destroy_workqueue(fs_info->endio_repair_workers);
d05a33ac 2206 btrfs_destroy_workqueue(fs_info->rmw_workers);
fccb5d86
QW		 2207 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2208 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2209 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
a8c93d4e 2210 btrfs_destroy_workqueue(fs_info->submit_workers);
5b3bc44e 2211 btrfs_destroy_workqueue(fs_info->delayed_workers);
e66f0bb1 2212 btrfs_destroy_workqueue(fs_info->caching_workers);
736cfa15 2213 btrfs_destroy_workqueue(fs_info->readahead_workers);
a44903ab 2214 btrfs_destroy_workqueue(fs_info->flush_workers);
fc97fab0 2215 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
a79b7d4b 2216 btrfs_destroy_workqueue(fs_info->extent_workers);
7abadb64
LB		 2217}
2218
2e9f5954
R		 2219static void free_root_extent_buffers(struct btrfs_root *root)
2220{
2221 if (root) {
2222 free_extent_buffer(root->node);
2223 free_extent_buffer(root->commit_root);
2224 root->node = NULL;
2225 root->commit_root = NULL;
2226 }
2227}
2228
af31f5e5
CM		 2229/* helper to cleanup tree roots */
2230static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2231{
2e9f5954 2232 free_root_extent_buffers(info->tree_root);
655b09fe 2233
2e9f5954
R		 2234 free_root_extent_buffers(info->dev_root);
2235 free_root_extent_buffers(info->extent_root);
2236 free_root_extent_buffers(info->csum_root);
2237 free_root_extent_buffers(info->quota_root);
2238 free_root_extent_buffers(info->uuid_root);
2239 if (chunk_root)
2240 free_root_extent_buffers(info->chunk_root);
70f6d82e 2241 free_root_extent_buffers(info->free_space_root);
af31f5e5
CM		 2242}
2243
faa2dbf0 2244void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
171f6537
JB		 2245{
2246 int ret;
2247 struct btrfs_root *gang[8];
2248 int i;
2249
2250 while (!list_empty(&fs_info->dead_roots)) {
2251 gang[0] = list_entry(fs_info->dead_roots.next,
2252 struct btrfs_root, root_list);
2253 list_del(&gang[0]->root_list);
2254
27cdeb70 2255 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
cb517eab 2256 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
171f6537
JB		 2257 } else {
2258 free_extent_buffer(gang[0]->node);
2259 free_extent_buffer(gang[0]->commit_root);
b0feb9d9 2260 btrfs_put_fs_root(gang[0]);
171f6537
JB		 2261 }
2262 }
2263
2264 while (1) {
2265 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2266 (void **)gang, 0,
2267 ARRAY_SIZE(gang));
2268 if (!ret)
2269 break;
2270 for (i = 0; i < ret; i++)
cb517eab 2271 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
171f6537 2272 }
1a4319cc
LB		 2273
2274 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2275 btrfs_free_log_root_tree(NULL, fs_info);
2276 btrfs_destroy_pinned_extent(fs_info->tree_root,
2277 fs_info->pinned_extents);
2278 }
171f6537 2279}
af31f5e5 2280
638aa7ed
ES		 2281static void btrfs_init_scrub(struct btrfs_fs_info *fs_info)
2282{
2283 mutex_init(&fs_info->scrub_lock);
2284 atomic_set(&fs_info->scrubs_running, 0);
2285 atomic_set(&fs_info->scrub_pause_req, 0);
2286 atomic_set(&fs_info->scrubs_paused, 0);
2287 atomic_set(&fs_info->scrub_cancel_req, 0);
2288 init_waitqueue_head(&fs_info->scrub_pause_wait);
2289 fs_info->scrub_workers_refcnt = 0;
2290}
2291
779a65a4
ES		 2292static void btrfs_init_balance(struct btrfs_fs_info *fs_info)
2293{
2294 spin_lock_init(&fs_info->balance_lock);
2295 mutex_init(&fs_info->balance_mutex);
2296 atomic_set(&fs_info->balance_running, 0);
2297 atomic_set(&fs_info->balance_pause_req, 0);
2298 atomic_set(&fs_info->balance_cancel_req, 0);
2299 fs_info->balance_ctl = NULL;
2300 init_waitqueue_head(&fs_info->balance_wait_q);
2301}
2302
f37938e0
ES		 2303static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info,
2304 struct btrfs_root *tree_root)
2305{
2306 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
2307 set_nlink(fs_info->btree_inode, 1);
2308 /*
2309 * we set the i_size on the btree inode to the max possible int.
2310 * the real end of the address space is determined by all of
2311 * the devices in the system
2312 */
2313 fs_info->btree_inode->i_size = OFFSET_MAX;
2314 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
f37938e0
ES		 2315
2316 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
2317 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
2318 fs_info->btree_inode->i_mapping);
2319 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
2320 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
2321
2322 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2323
2324 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2325 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2326 sizeof(struct btrfs_key));
2327 set_bit(BTRFS_INODE_DUMMY,
2328 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
2329 btrfs_insert_inode_hash(fs_info->btree_inode);
2330}
2331
ad618368
ES		 2332static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info)
2333{
2334 fs_info->dev_replace.lock_owner = 0;
2335 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2336 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
73beece9
LB		 2337 rwlock_init(&fs_info->dev_replace.lock);
2338 atomic_set(&fs_info->dev_replace.read_locks, 0);
2339 atomic_set(&fs_info->dev_replace.blocking_readers, 0);
ad618368 2340 init_waitqueue_head(&fs_info->replace_wait);
73beece9 2341 init_waitqueue_head(&fs_info->dev_replace.read_lock_wq);
ad618368
ES		 2342}
2343
f9e92e40
ES		 2344static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info)
2345{
2346 spin_lock_init(&fs_info->qgroup_lock);
2347 mutex_init(&fs_info->qgroup_ioctl_lock);
2348 fs_info->qgroup_tree = RB_ROOT;
2349 fs_info->qgroup_op_tree = RB_ROOT;
2350 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2351 fs_info->qgroup_seq = 1;
f9e92e40 2352 fs_info->qgroup_ulist = NULL;
d2c609b8 2353 fs_info->qgroup_rescan_running = false;
f9e92e40
ES		 2354 mutex_init(&fs_info->qgroup_rescan_lock);
2355}
2356
2a458198
ES		 2357static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info,
2358 struct btrfs_fs_devices *fs_devices)
2359{
2360 int max_active = fs_info->thread_pool_size;
6f011058 2361 unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
2a458198
ES		 2362
2363 fs_info->workers =
cb001095
JM		 2364 btrfs_alloc_workqueue(fs_info, "worker",
2365 flags | WQ_HIGHPRI, max_active, 16);
2a458198
ES		 2366
2367 fs_info->delalloc_workers =
cb001095
JM		 2368 btrfs_alloc_workqueue(fs_info, "delalloc",
2369 flags, max_active, 2);
2a458198
ES		 2370
2371 fs_info->flush_workers =
cb001095
JM		 2372 btrfs_alloc_workqueue(fs_info, "flush_delalloc",
2373 flags, max_active, 0);
2a458198
ES		 2374
2375 fs_info->caching_workers =
cb001095 2376 btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0);
2a458198
ES		 2377
2378 /*
2379 * a higher idle thresh on the submit workers makes it much more
2380 * likely that bios will be send down in a sane order to the
2381 * devices
2382 */
2383 fs_info->submit_workers =
cb001095 2384 btrfs_alloc_workqueue(fs_info, "submit", flags,
2a458198
ES		 2385 min_t(u64, fs_devices->num_devices,
2386 max_active), 64);
2387
2388 fs_info->fixup_workers =
cb001095 2389 btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0);
2a458198
ES		 2390
2391 /*
2392 * endios are largely parallel and should have a very
2393 * low idle thresh
2394 */
2395 fs_info->endio_workers =
cb001095 2396 btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4);
2a458198 2397 fs_info->endio_meta_workers =
cb001095
JM		 2398 btrfs_alloc_workqueue(fs_info, "endio-meta", flags,
2399 max_active, 4);
2a458198 2400 fs_info->endio_meta_write_workers =
cb001095
JM		 2401 btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags,
2402 max_active, 2);
2a458198 2403 fs_info->endio_raid56_workers =
cb001095
JM		 2404 btrfs_alloc_workqueue(fs_info, "endio-raid56", flags,
2405 max_active, 4);
2a458198 2406 fs_info->endio_repair_workers =
cb001095 2407 btrfs_alloc_workqueue(fs_info, "endio-repair", flags, 1, 0);
2a458198 2408 fs_info->rmw_workers =
cb001095 2409 btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2);
2a458198 2410 fs_info->endio_write_workers =
cb001095
JM		 2411 btrfs_alloc_workqueue(fs_info, "endio-write", flags,
2412 max_active, 2);
2a458198 2413 fs_info->endio_freespace_worker =
cb001095
JM		 2414 btrfs_alloc_workqueue(fs_info, "freespace-write", flags,
2415 max_active, 0);
2a458198 2416 fs_info->delayed_workers =
cb001095
JM		 2417 btrfs_alloc_workqueue(fs_info, "delayed-meta", flags,
2418 max_active, 0);
2a458198 2419 fs_info->readahead_workers =
cb001095
JM		 2420 btrfs_alloc_workqueue(fs_info, "readahead", flags,
2421 max_active, 2);
2a458198 2422 fs_info->qgroup_rescan_workers =
cb001095 2423 btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0);
2a458198 2424 fs_info->extent_workers =
cb001095 2425 btrfs_alloc_workqueue(fs_info, "extent-refs", flags,
2a458198
ES		 2426 min_t(u64, fs_devices->num_devices,
2427 max_active), 8);
2428
2429 if (!(fs_info->workers && fs_info->delalloc_workers &&
2430 fs_info->submit_workers && fs_info->flush_workers &&
2431 fs_info->endio_workers && fs_info->endio_meta_workers &&
2432 fs_info->endio_meta_write_workers &&
2433 fs_info->endio_repair_workers &&
2434 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
2435 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
2436 fs_info->caching_workers && fs_info->readahead_workers &&
2437 fs_info->fixup_workers && fs_info->delayed_workers &&
2438 fs_info->extent_workers &&
2439 fs_info->qgroup_rescan_workers)) {
2440 return -ENOMEM;
2441 }
2442
2443 return 0;
2444}
2445
63443bf5
ES		 2446static int btrfs_replay_log(struct btrfs_fs_info *fs_info,
2447 struct btrfs_fs_devices *fs_devices)
2448{
2449 int ret;
2450 struct btrfs_root *tree_root = fs_info->tree_root;
2451 struct btrfs_root *log_tree_root;
2452 struct btrfs_super_block *disk_super = fs_info->super_copy;
2453 u64 bytenr = btrfs_super_log_root(disk_super);
2454
2455 if (fs_devices->rw_devices == 0) {
f14d104d 2456 btrfs_warn(fs_info, "log replay required on RO media");
63443bf5
ES		 2457 return -EIO;
2458 }
2459
74e4d827 2460 log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
63443bf5
ES		 2461 if (!log_tree_root)
2462 return -ENOMEM;
2463
2464 __setup_root(tree_root->nodesize, tree_root->sectorsize,
2465 tree_root->stripesize, log_tree_root, fs_info,
2466 BTRFS_TREE_LOG_OBJECTID);
2467
2468 log_tree_root->node = read_tree_block(tree_root, bytenr,
2469 fs_info->generation + 1);
64c043de 2470 if (IS_ERR(log_tree_root->node)) {
f14d104d 2471 btrfs_warn(fs_info, "failed to read log tree");
0eeff236 2472 ret = PTR_ERR(log_tree_root->node);
64c043de 2473 kfree(log_tree_root);
0eeff236 2474 return ret;
64c043de 2475 } else if (!extent_buffer_uptodate(log_tree_root->node)) {
f14d104d 2476 btrfs_err(fs_info, "failed to read log tree");
63443bf5
ES		 2477 free_extent_buffer(log_tree_root->node);
2478 kfree(log_tree_root);
2479 return -EIO;
2480 }
2481 /* returns with log_tree_root freed on success */
2482 ret = btrfs_recover_log_trees(log_tree_root);
2483 if (ret) {
34d97007 2484 btrfs_handle_fs_error(tree_root->fs_info, ret,
63443bf5
ES		 2485 "Failed to recover log tree");
2486 free_extent_buffer(log_tree_root->node);
2487 kfree(log_tree_root);
2488 return ret;
2489 }
2490
2491 if (fs_info->sb->s_flags & MS_RDONLY) {
2492 ret = btrfs_commit_super(tree_root);
2493 if (ret)
2494 return ret;
2495 }
2496
2497 return 0;
2498}
2499
4bbcaa64
ES		 2500static int btrfs_read_roots(struct btrfs_fs_info *fs_info,
2501 struct btrfs_root *tree_root)
2502{
a4f3d2c4 2503 struct btrfs_root *root;
4bbcaa64
ES		 2504 struct btrfs_key location;
2505 int ret;
2506
2507 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2508 location.type = BTRFS_ROOT_ITEM_KEY;
2509 location.offset = 0;
2510
a4f3d2c4
DS		 2511 root = btrfs_read_tree_root(tree_root, &location);
2512 if (IS_ERR(root))
2513 return PTR_ERR(root);
2514 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2515 fs_info->extent_root = root;
4bbcaa64
ES		 2516
2517 location.objectid = BTRFS_DEV_TREE_OBJECTID;
a4f3d2c4
DS		 2518 root = btrfs_read_tree_root(tree_root, &location);
2519 if (IS_ERR(root))
2520 return PTR_ERR(root);
2521 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2522 fs_info->dev_root = root;
4bbcaa64
ES		 2523 btrfs_init_devices_late(fs_info);
2524
2525 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
a4f3d2c4
DS		 2526 root = btrfs_read_tree_root(tree_root, &location);
2527 if (IS_ERR(root))
2528 return PTR_ERR(root);
2529 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2530 fs_info->csum_root = root;
4bbcaa64
ES		 2531
2532 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
a4f3d2c4
DS		 2533 root = btrfs_read_tree_root(tree_root, &location);
2534 if (!IS_ERR(root)) {
2535 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
afcdd129 2536 set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags);
a4f3d2c4 2537 fs_info->quota_root = root;
4bbcaa64
ES		 2538 }
2539
2540 location.objectid = BTRFS_UUID_TREE_OBJECTID;
a4f3d2c4
DS		 2541 root = btrfs_read_tree_root(tree_root, &location);
2542 if (IS_ERR(root)) {
2543 ret = PTR_ERR(root);
4bbcaa64
ES		 2544 if (ret != -ENOENT)
2545 return ret;
2546 } else {
a4f3d2c4
DS		 2547 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2548 fs_info->uuid_root = root;
4bbcaa64
ES		 2549 }
2550
70f6d82e
OS		 2551 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
2552 location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID;
2553 root = btrfs_read_tree_root(tree_root, &location);
2554 if (IS_ERR(root))
2555 return PTR_ERR(root);
2556 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2557 fs_info->free_space_root = root;
2558 }
2559
4bbcaa64
ES		 2560 return 0;
2561}
2562
ad2b2c80
AV		 2563int open_ctree(struct super_block *sb,
2564 struct btrfs_fs_devices *fs_devices,
2565 char *options)
2e635a27 2566{
db94535d
CM		 2567 u32 sectorsize;
2568 u32 nodesize;
87ee04eb 2569 u32 stripesize;
84234f3a 2570 u64 generation;
f2b636e8 2571 u64 features;
3de4586c 2572 struct btrfs_key location;
a061fc8d 2573 struct buffer_head *bh;
4d34b278 2574 struct btrfs_super_block *disk_super;
815745cf 2575 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2576 struct btrfs_root *tree_root;
4d34b278 2577 struct btrfs_root *chunk_root;
eb60ceac 2578 int ret;
e58ca020 2579 int err = -EINVAL;
af31f5e5
CM		 2580 int num_backups_tried = 0;
2581 int backup_index = 0;
5cdc7ad3 2582 int max_active;
6675df31 2583 int clear_free_space_tree = 0;
4543df7e 2584
74e4d827
DS		 2585 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
2586 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
cb517eab 2587 if (!tree_root || !chunk_root) {
39279cc3
CM		 2588 err = -ENOMEM;
2589 goto fail;
2590 }
76dda93c
YZ		 2591
2592 ret = init_srcu_struct(&fs_info->subvol_srcu);
2593 if (ret) {
2594 err = ret;
2595 goto fail;
2596 }
2597
2598 ret = setup_bdi(fs_info, &fs_info->bdi);
2599 if (ret) {
2600 err = ret;
2601 goto fail_srcu;
2602 }
2603
908c7f19 2604 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL);
e2d84521
MX		 2605 if (ret) {
2606 err = ret;
2607 goto fail_bdi;
2608 }
09cbfeaf 2609 fs_info->dirty_metadata_batch = PAGE_SIZE *
e2d84521
MX		 2610 (1 + ilog2(nr_cpu_ids));
2611
908c7f19 2612 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL);
963d678b
MX		 2613 if (ret) {
2614 err = ret;
2615 goto fail_dirty_metadata_bytes;
2616 }
2617
908c7f19 2618 ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL);
c404e0dc
MX		 2619 if (ret) {
2620 err = ret;
2621 goto fail_delalloc_bytes;
2622 }
2623
76dda93c
YZ		 2624 fs_info->btree_inode = new_inode(sb);
2625 if (!fs_info->btree_inode) {
2626 err = -ENOMEM;
c404e0dc 2627 goto fail_bio_counter;
76dda93c
YZ		 2628 }
2629
a6591715 2630 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2631
76dda93c 2632 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
f28491e0 2633 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
8fd17795 2634 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2635 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2636 INIT_LIST_HEAD(&fs_info->delayed_iputs);
eb73c1b7 2637 INIT_LIST_HEAD(&fs_info->delalloc_roots);
11833d66 2638 INIT_LIST_HEAD(&fs_info->caching_block_groups);
eb73c1b7 2639 spin_lock_init(&fs_info->delalloc_root_lock);
a4abeea4 2640 spin_lock_init(&fs_info->trans_lock);
76dda93c 2641 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2642 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2643 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 2644 spin_lock_init(&fs_info->free_chunk_lock);
f29021b2 2645 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2646 spin_lock_init(&fs_info->super_lock);
fcebe456 2647 spin_lock_init(&fs_info->qgroup_op_lock);
f28491e0 2648 spin_lock_init(&fs_info->buffer_lock);
47ab2a6c 2649 spin_lock_init(&fs_info->unused_bgs_lock);
f29021b2 2650 rwlock_init(&fs_info->tree_mod_log_lock);
d7c15171 2651 mutex_init(&fs_info->unused_bg_unpin_mutex);
67c5e7d4 2652 mutex_init(&fs_info->delete_unused_bgs_mutex);
7585717f 2653 mutex_init(&fs_info->reloc_mutex);
573bfb72 2654 mutex_init(&fs_info->delalloc_root_mutex);
c2d6cb16 2655 mutex_init(&fs_info->cleaner_delayed_iput_mutex);
de98ced9 2656 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2657
0b86a832 2658 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2659 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2660 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
47ab2a6c 2661 INIT_LIST_HEAD(&fs_info->unused_bgs);
0b86a832 2662 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX		 2663 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2664 BTRFS_BLOCK_RSV_GLOBAL);
2665 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2666 BTRFS_BLOCK_RSV_DELALLOC);
2667 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2668 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2669 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2670 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2671 BTRFS_BLOCK_RSV_DELOPS);
cb03c743 2672 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 2673 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 2674 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 2675 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 2676 atomic_set(&fs_info->defrag_running, 0);
fcebe456 2677 atomic_set(&fs_info->qgroup_op_seq, 0);
2fefd558 2678 atomic_set(&fs_info->reada_works_cnt, 0);
fc36ed7e 2679 atomic64_set(&fs_info->tree_mod_seq, 0);
9e7cc91a 2680 fs_info->fs_frozen = 0;
e20d96d6 2681 fs_info->sb = sb;
95ac567a 2682 fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE;
9ed74f2d 2683 fs_info->metadata_ratio = 0;
4cb5300b 2684 fs_info->defrag_inodes = RB_ROOT;
2bf64758 2685 fs_info->free_chunk_space = 0;
f29021b2 2686 fs_info->tree_mod_log = RB_ROOT;
8b87dc17 2687 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
f8c269d7 2688 fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */
90519d66 2689 /* readahead state */
d0164adc 2690 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM);
90519d66 2691 spin_lock_init(&fs_info->reada_lock);
c8b97818 2692
b34b086c
CM		 2693 fs_info->thread_pool_size = min_t(unsigned long,
2694 num_online_cpus() + 2, 8);
0afbaf8c 2695
199c2a9c
MX		 2696 INIT_LIST_HEAD(&fs_info->ordered_roots);
2697 spin_lock_init(&fs_info->ordered_root_lock);
16cdcec7 2698 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
74e4d827 2699 GFP_KERNEL);
16cdcec7
MX		 2700 if (!fs_info->delayed_root) {
2701 err = -ENOMEM;
2702 goto fail_iput;
2703 }
2704 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2705
638aa7ed 2706 btrfs_init_scrub(fs_info);
21adbd5c
SB		 2707#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2708 fs_info->check_integrity_print_mask = 0;
2709#endif
779a65a4 2710 btrfs_init_balance(fs_info);
21c7e756 2711 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
a2de733c 2712
a061fc8d
CM		 2713 sb->s_blocksize = 4096;
2714 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 2715 sb->s_bdi = &fs_info->bdi;
a061fc8d 2716
f37938e0 2717 btrfs_init_btree_inode(fs_info, tree_root);
76dda93c 2718
0f9dd46c 2719 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2720 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2721 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2722
11833d66 2723 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 2724 fs_info->btree_inode->i_mapping);
11833d66 2725 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 2726 fs_info->btree_inode->i_mapping);
11833d66 2727 fs_info->pinned_extents = &fs_info->freed_extents[0];
afcdd129 2728 set_bit(BTRFS_FS_BARRIER, &fs_info->flags);
39279cc3 2729
5a3f23d5 2730 mutex_init(&fs_info->ordered_operations_mutex);
e02119d5 2731 mutex_init(&fs_info->tree_log_mutex);
925baedd 2732 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 2733 mutex_init(&fs_info->transaction_kthread_mutex);
2734 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2735 mutex_init(&fs_info->volume_mutex);
1bbc621e 2736 mutex_init(&fs_info->ro_block_group_mutex);
9e351cc8 2737 init_rwsem(&fs_info->commit_root_sem);
c71bf099 2738 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2739 init_rwsem(&fs_info->subvol_sem);
803b2f54 2740 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
fa9c0d79 2741
ad618368 2742 btrfs_init_dev_replace_locks(fs_info);
f9e92e40 2743 btrfs_init_qgroup(fs_info);
416ac51d 2744
fa9c0d79
CM		 2745 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2746 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2747
e6dcd2dc 2748 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2749 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2750 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2751 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2752
04216820
FM		 2753 INIT_LIST_HEAD(&fs_info->pinned_chunks);
2754
53b381b3
DW		 2755 ret = btrfs_alloc_stripe_hash_table(fs_info);
2756 if (ret) {
83c8266a 2757 err = ret;
53b381b3
DW		 2758 goto fail_alloc;
2759 }
2760
707e8a07 2761 __setup_root(4096, 4096, 4096, tree_root,
2c90e5d6 2762 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2763
3c4bb26b 2764 invalidate_bdev(fs_devices->latest_bdev);
1104a885
DS		 2765
2766 /*
2767 * Read super block and check the signature bytes only
2768 */
a512bbf8 2769 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
92fc03fb
AJ		 2770 if (IS_ERR(bh)) {
2771 err = PTR_ERR(bh);
16cdcec7 2772 goto fail_alloc;
20b45077 2773 }
39279cc3 2774
1104a885
DS		 2775 /*
2776 * We want to check superblock checksum, the type is stored inside.
2777 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2778 */
ab8d0fc4 2779 if (btrfs_check_super_csum(fs_info, bh->b_data)) {
05135f59 2780 btrfs_err(fs_info, "superblock checksum mismatch");
1104a885 2781 err = -EINVAL;
b2acdddf 2782 brelse(bh);
1104a885
DS		 2783 goto fail_alloc;
2784 }
2785
2786 /*
2787 * super_copy is zeroed at allocation time and we never touch the
2788 * following bytes up to INFO_SIZE, the checksum is calculated from
2789 * the whole block of INFO_SIZE
2790 */
6c41761f
DS		 2791 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2792 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2793 sizeof(*fs_info->super_for_commit));
a061fc8d 2794 brelse(bh);
5f39d397 2795
6c41761f 2796 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2797
1104a885
DS		 2798 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2799 if (ret) {
05135f59 2800 btrfs_err(fs_info, "superblock contains fatal errors");
1104a885
DS		 2801 err = -EINVAL;
2802 goto fail_alloc;
2803 }
2804
6c41761f 2805 disk_super = fs_info->super_copy;
0f7d52f4 2806 if (!btrfs_super_root(disk_super))
16cdcec7 2807 goto fail_alloc;
0f7d52f4 2808
acce952b 2809 /* check FS state, whether FS is broken. */
87533c47
MX		 2810 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2811 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2812
af31f5e5
CM		 2813 /*
2814 * run through our array of backup supers and setup
2815 * our ring pointer to the oldest one
2816 */
2817 generation = btrfs_super_generation(disk_super);
2818 find_oldest_super_backup(fs_info, generation);
2819
75e7cb7f
LB		 2820 /*
2821 * In the long term, we'll store the compression type in the super
2822 * block, and it'll be used for per file compression control.
2823 */
2824 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2825
96da0919 2826 ret = btrfs_parse_options(tree_root, options, sb->s_flags);
2b82032c
YZ		 2827 if (ret) {
2828 err = ret;
16cdcec7 2829 goto fail_alloc;
2b82032c 2830 }
dfe25020 2831
f2b636e8
JB		 2832 features = btrfs_super_incompat_flags(disk_super) &
2833 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2834 if (features) {
05135f59
DS		 2835 btrfs_err(fs_info,
2836 "cannot mount because of unsupported optional features (%llx)",
2837 features);
f2b636e8 2838 err = -EINVAL;
16cdcec7 2839 goto fail_alloc;
f2b636e8
JB		 2840 }
2841
5d4f98a2 2842 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2843 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
69e380d1 2844 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2845 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
727011e0 2846
3173a18f 2847 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
05135f59 2848 btrfs_info(fs_info, "has skinny extents");
3173a18f 2849
727011e0
CM		 2850 /*
2851 * flag our filesystem as having big metadata blocks if
2852 * they are bigger than the page size
2853 */
09cbfeaf 2854 if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) {
727011e0 2855 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
05135f59
DS		 2856 btrfs_info(fs_info,
2857 "flagging fs with big metadata feature");
727011e0
CM		 2858 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2859 }
2860
bc3f116f 2861 nodesize = btrfs_super_nodesize(disk_super);
bc3f116f 2862 sectorsize = btrfs_super_sectorsize(disk_super);
b7f67055 2863 stripesize = sectorsize;
707e8a07 2864 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids));
963d678b 2865 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f
CM		 2866
2867 /*
2868 * mixed block groups end up with duplicate but slightly offset
2869 * extent buffers for the same range. It leads to corruptions
2870 */
2871 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
707e8a07 2872 (sectorsize != nodesize)) {
05135f59
DS		 2873 btrfs_err(fs_info,
2874"unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups",
2875 nodesize, sectorsize);
bc3f116f
CM		 2876 goto fail_alloc;
2877 }
2878
ceda0864
MX		 2879 /*
2880 * Needn't use the lock because there is no other task which will
2881 * update the flag.
2882 */
a6fa6fae 2883 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2884
f2b636e8
JB		 2885 features = btrfs_super_compat_ro_flags(disk_super) &
2886 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2887 if (!(sb->s_flags & MS_RDONLY) && features) {
05135f59
DS		 2888 btrfs_err(fs_info,
2889 "cannot mount read-write because of unsupported optional features (%llx)",
c1c9ff7c 2890 features);
f2b636e8 2891 err = -EINVAL;
16cdcec7 2892 goto fail_alloc;
f2b636e8 2893 }
61d92c32 2894
5cdc7ad3 2895 max_active = fs_info->thread_pool_size;
61d92c32 2896
2a458198
ES		 2897 ret = btrfs_init_workqueues(fs_info, fs_devices);
2898 if (ret) {
2899 err = ret;
0dc3b84a
JB		 2900 goto fail_sb_buffer;
2901 }
4543df7e 2902
4575c9cc 2903 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818 2904 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
09cbfeaf 2905 SZ_4M / PAGE_SIZE);
4575c9cc 2906
db94535d 2907 tree_root->nodesize = nodesize;
db94535d 2908 tree_root->sectorsize = sectorsize;
87ee04eb 2909 tree_root->stripesize = stripesize;
a061fc8d
CM		 2910
2911 sb->s_blocksize = sectorsize;
2912 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 2913
925baedd 2914 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 2915 ret = btrfs_read_sys_array(tree_root);
925baedd 2916 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2917 if (ret) {
05135f59 2918 btrfs_err(fs_info, "failed to read the system array: %d", ret);
5d4f98a2 2919 goto fail_sb_buffer;
84eed90f 2920 }
0b86a832 2921
84234f3a 2922 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832 2923
707e8a07
DS		 2924 __setup_root(nodesize, sectorsize, stripesize, chunk_root,
2925 fs_info, BTRFS_CHUNK_TREE_OBJECTID);
0b86a832
CM		 2926
2927 chunk_root->node = read_tree_block(chunk_root,
2928 btrfs_super_chunk_root(disk_super),
ce86cd59 2929 generation);
64c043de
LB		 2930 if (IS_ERR(chunk_root->node) ||
2931 !extent_buffer_uptodate(chunk_root->node)) {
05135f59 2932 btrfs_err(fs_info, "failed to read chunk root");
e5fffbac 2933 if (!IS_ERR(chunk_root->node))
2934 free_extent_buffer(chunk_root->node);
95ab1f64 2935 chunk_root->node = NULL;
af31f5e5 2936 goto fail_tree_roots;
83121942 2937 }
5d4f98a2
YZ		 2938 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2939 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2940
e17cade2 2941 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
b308bc2f 2942 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
e17cade2 2943
0b86a832 2944 ret = btrfs_read_chunk_tree(chunk_root);
2b82032c 2945 if (ret) {
05135f59 2946 btrfs_err(fs_info, "failed to read chunk tree: %d", ret);
af31f5e5 2947 goto fail_tree_roots;
2b82032c 2948 }
0b86a832 2949
8dabb742
SB		 2950 /*
2951 * keep the device that is marked to be the target device for the
2952 * dev_replace procedure
2953 */
9eaed21e 2954 btrfs_close_extra_devices(fs_devices, 0);
dfe25020 2955
a6b0d5c8 2956 if (!fs_devices->latest_bdev) {
05135f59 2957 btrfs_err(fs_info, "failed to read devices");
a6b0d5c8
CM		 2958 goto fail_tree_roots;
2959 }
2960
af31f5e5 2961retry_root_backup:
84234f3a 2962 generation = btrfs_super_generation(disk_super);
0b86a832 2963
e20d96d6 2964 tree_root->node = read_tree_block(tree_root,
db94535d 2965 btrfs_super_root(disk_super),
ce86cd59 2966 generation);
64c043de
LB		 2967 if (IS_ERR(tree_root->node) ||
2968 !extent_buffer_uptodate(tree_root->node)) {
05135f59 2969 btrfs_warn(fs_info, "failed to read tree root");
e5fffbac 2970 if (!IS_ERR(tree_root->node))
2971 free_extent_buffer(tree_root->node);
95ab1f64 2972 tree_root->node = NULL;
af31f5e5 2973 goto recovery_tree_root;
83121942 2974 }
af31f5e5 2975
5d4f98a2
YZ		 2976 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2977 tree_root->commit_root = btrfs_root_node(tree_root);
69e9c6c6 2978 btrfs_set_root_refs(&tree_root->root_item, 1);
db94535d 2979
f32e48e9
CR		 2980 mutex_lock(&tree_root->objectid_mutex);
2981 ret = btrfs_find_highest_objectid(tree_root,
2982 &tree_root->highest_objectid);
2983 if (ret) {
2984 mutex_unlock(&tree_root->objectid_mutex);
2985 goto recovery_tree_root;
2986 }
2987
2988 ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID);
2989
2990 mutex_unlock(&tree_root->objectid_mutex);
2991
4bbcaa64
ES		 2992 ret = btrfs_read_roots(fs_info, tree_root);
2993 if (ret)
af31f5e5 2994 goto recovery_tree_root;
f7a81ea4 2995
8929ecfa
YZ		 2996 fs_info->generation = generation;
2997 fs_info->last_trans_committed = generation;
8929ecfa 2998
68310a5e
ID		 2999 ret = btrfs_recover_balance(fs_info);
3000 if (ret) {
05135f59 3001 btrfs_err(fs_info, "failed to recover balance: %d", ret);
68310a5e
ID		 3002 goto fail_block_groups;
3003 }
3004
733f4fbb
SB		 3005 ret = btrfs_init_dev_stats(fs_info);
3006 if (ret) {
05135f59 3007 btrfs_err(fs_info, "failed to init dev_stats: %d", ret);
733f4fbb
SB		 3008 goto fail_block_groups;
3009 }
3010
8dabb742
SB		 3011 ret = btrfs_init_dev_replace(fs_info);
3012 if (ret) {
05135f59 3013 btrfs_err(fs_info, "failed to init dev_replace: %d", ret);
8dabb742
SB		 3014 goto fail_block_groups;
3015 }
3016
9eaed21e 3017 btrfs_close_extra_devices(fs_devices, 1);
8dabb742 3018
b7c35e81
AJ		 3019 ret = btrfs_sysfs_add_fsid(fs_devices, NULL);
3020 if (ret) {
05135f59
DS		 3021 btrfs_err(fs_info, "failed to init sysfs fsid interface: %d",
3022 ret);
b7c35e81
AJ		 3023 goto fail_block_groups;
3024 }
3025
3026 ret = btrfs_sysfs_add_device(fs_devices);
3027 if (ret) {
05135f59
DS		 3028 btrfs_err(fs_info, "failed to init sysfs device interface: %d",
3029 ret);
b7c35e81
AJ		 3030 goto fail_fsdev_sysfs;
3031 }
3032
96f3136e 3033 ret = btrfs_sysfs_add_mounted(fs_info);
c59021f8 3034 if (ret) {
05135f59 3035 btrfs_err(fs_info, "failed to init sysfs interface: %d", ret);
b7c35e81 3036 goto fail_fsdev_sysfs;
c59021f8 3037 }
3038
c59021f8 3039 ret = btrfs_init_space_info(fs_info);
3040 if (ret) {
05135f59 3041 btrfs_err(fs_info, "failed to initialize space info: %d", ret);
2365dd3c 3042 goto fail_sysfs;
c59021f8 3043 }
3044
4bbcaa64 3045 ret = btrfs_read_block_groups(fs_info->extent_root);
1b1d1f66 3046 if (ret) {
05135f59 3047 btrfs_err(fs_info, "failed to read block groups: %d", ret);
2365dd3c 3048 goto fail_sysfs;
1b1d1f66 3049 }
5af3e8cc
SB		 3050 fs_info->num_tolerated_disk_barrier_failures =
3051 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
292fd7fc
SB		 3052 if (fs_info->fs_devices->missing_devices >
3053 fs_info->num_tolerated_disk_barrier_failures &&
3054 !(sb->s_flags & MS_RDONLY)) {
05135f59
DS		 3055 btrfs_warn(fs_info,
3056"missing devices (%llu) exceeds the limit (%d), writeable mount is not allowed",
78fa1770
ZL		 3057 fs_info->fs_devices->missing_devices,
3058 fs_info->num_tolerated_disk_barrier_failures);
2365dd3c 3059 goto fail_sysfs;
292fd7fc 3060 }
9078a3e1 3061
a74a4b97
CM		 3062 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
3063 "btrfs-cleaner");
57506d50 3064 if (IS_ERR(fs_info->cleaner_kthread))
2365dd3c 3065 goto fail_sysfs;
a74a4b97
CM		 3066
3067 fs_info->transaction_kthread = kthread_run(transaction_kthread,
3068 tree_root,
3069 "btrfs-transaction");
57506d50 3070 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 3071 goto fail_cleaner;
a74a4b97 3072
3cdde224
JM		 3073 if (!btrfs_test_opt(tree_root->fs_info, SSD) &&
3074 !btrfs_test_opt(tree_root->fs_info, NOSSD) &&
c289811c 3075 !fs_info->fs_devices->rotating) {
05135f59 3076 btrfs_info(fs_info, "detected SSD devices, enabling SSD mode");
c289811c
CM		 3077 btrfs_set_opt(fs_info->mount_opt, SSD);
3078 }
3079
572d9ab7 3080 /*
01327610 3081 * Mount does not set all options immediately, we can do it now and do
572d9ab7
DS		 3082 * not have to wait for transaction commit
3083 */
3084 btrfs_apply_pending_changes(fs_info);
3818aea2 3085
21adbd5c 3086#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3cdde224 3087 if (btrfs_test_opt(tree_root->fs_info, CHECK_INTEGRITY)) {
21adbd5c 3088 ret = btrfsic_mount(tree_root, fs_devices,
3cdde224 3089 btrfs_test_opt(tree_root->fs_info,
21adbd5c
SB		 3090 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
3091 1 : 0,
3092 fs_info->check_integrity_print_mask);
3093 if (ret)
05135f59
DS		 3094 btrfs_warn(fs_info,
3095 "failed to initialize integrity check module: %d",
3096 ret);
21adbd5c
SB		 3097 }
3098#endif
bcef60f2
AJ		 3099 ret = btrfs_read_qgroup_config(fs_info);
3100 if (ret)
3101 goto fail_trans_kthread;
21adbd5c 3102
96da0919
QW		 3103 /* do not make disk changes in broken FS or nologreplay is given */
3104 if (btrfs_super_log_root(disk_super) != 0 &&
3cdde224 3105 !btrfs_test_opt(tree_root->fs_info, NOLOGREPLAY)) {
63443bf5 3106 ret = btrfs_replay_log(fs_info, fs_devices);
79787eaa 3107 if (ret) {
63443bf5 3108 err = ret;
28c16cbb 3109 goto fail_qgroup;
79787eaa 3110 }
e02119d5 3111 }
1a40e23b 3112
76dda93c 3113 ret = btrfs_find_orphan_roots(tree_root);
79787eaa 3114 if (ret)
28c16cbb 3115 goto fail_qgroup;
76dda93c 3116
7c2ca468 3117 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b 3118 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2 3119 if (ret)
28c16cbb 3120 goto fail_qgroup;
90c711ab
ZB		 3121
3122 mutex_lock(&fs_info->cleaner_mutex);
5d4f98a2 3123 ret = btrfs_recover_relocation(tree_root);
90c711ab 3124 mutex_unlock(&fs_info->cleaner_mutex);
d7ce5843 3125 if (ret < 0) {
05135f59
DS		 3126 btrfs_warn(fs_info, "failed to recover relocation: %d",
3127 ret);
d7ce5843 3128 err = -EINVAL;
bcef60f2 3129 goto fail_qgroup;
d7ce5843 3130 }
7c2ca468 3131 }
1a40e23b 3132
3de4586c
CM		 3133 location.objectid = BTRFS_FS_TREE_OBJECTID;
3134 location.type = BTRFS_ROOT_ITEM_KEY;
cb517eab 3135 location.offset = 0;
3de4586c 3136
3de4586c 3137 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
3140c9a3
DC		 3138 if (IS_ERR(fs_info->fs_root)) {
3139 err = PTR_ERR(fs_info->fs_root);
bcef60f2 3140 goto fail_qgroup;
3140c9a3 3141 }
c289811c 3142
2b6ba629
ID		 3143 if (sb->s_flags & MS_RDONLY)
3144 return 0;
59641015 3145
f8d468a1
OS		 3146 if (btrfs_test_opt(fs_info, CLEAR_CACHE) &&
3147 btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
6675df31
OS		 3148 clear_free_space_tree = 1;
3149 } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
3150 !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) {
3151 btrfs_warn(fs_info, "free space tree is invalid");
3152 clear_free_space_tree = 1;
3153 }
3154
3155 if (clear_free_space_tree) {
f8d468a1
OS		 3156 btrfs_info(fs_info, "clearing free space tree");
3157 ret = btrfs_clear_free_space_tree(fs_info);
3158 if (ret) {
3159 btrfs_warn(fs_info,
3160 "failed to clear free space tree: %d", ret);
3161 close_ctree(tree_root);
3162 return ret;
3163 }
3164 }
3165
3cdde224 3166 if (btrfs_test_opt(tree_root->fs_info, FREE_SPACE_TREE) &&
511711af 3167 !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
05135f59 3168 btrfs_info(fs_info, "creating free space tree");
511711af
CM		 3169 ret = btrfs_create_free_space_tree(fs_info);
3170 if (ret) {
05135f59
DS		 3171 btrfs_warn(fs_info,
3172 "failed to create free space tree: %d", ret);
511711af
CM		 3173 close_ctree(tree_root);
3174 return ret;
3175 }
3176 }
3177
2b6ba629
ID		 3178 down_read(&fs_info->cleanup_work_sem);
3179 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
3180 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 3181 up_read(&fs_info->cleanup_work_sem);
2b6ba629
ID		 3182 close_ctree(tree_root);
3183 return ret;
3184 }
3185 up_read(&fs_info->cleanup_work_sem);
59641015 3186
2b6ba629
ID		 3187 ret = btrfs_resume_balance_async(fs_info);
3188 if (ret) {
05135f59 3189 btrfs_warn(fs_info, "failed to resume balance: %d", ret);
2b6ba629
ID		 3190 close_ctree(tree_root);
3191 return ret;
e3acc2a6
JB		 3192 }
3193
8dabb742
SB		 3194 ret = btrfs_resume_dev_replace_async(fs_info);
3195 if (ret) {
05135f59 3196 btrfs_warn(fs_info, "failed to resume device replace: %d", ret);
8dabb742
SB		 3197 close_ctree(tree_root);
3198 return ret;
3199 }
3200
b382a324
JS		 3201 btrfs_qgroup_rescan_resume(fs_info);
3202
4bbcaa64 3203 if (!fs_info->uuid_root) {
05135f59 3204 btrfs_info(fs_info, "creating UUID tree");
f7a81ea4
SB		 3205 ret = btrfs_create_uuid_tree(fs_info);
3206 if (ret) {
05135f59
DS		 3207 btrfs_warn(fs_info,
3208 "failed to create the UUID tree: %d", ret);
f7a81ea4
SB		 3209 close_ctree(tree_root);
3210 return ret;
3211 }
3cdde224 3212 } else if (btrfs_test_opt(tree_root->fs_info, RESCAN_UUID_TREE) ||
4bbcaa64
ES		 3213 fs_info->generation !=
3214 btrfs_super_uuid_tree_generation(disk_super)) {
05135f59 3215 btrfs_info(fs_info, "checking UUID tree");
70f80175
SB		 3216 ret = btrfs_check_uuid_tree(fs_info);
3217 if (ret) {
05135f59
DS		 3218 btrfs_warn(fs_info,
3219 "failed to check the UUID tree: %d", ret);
70f80175
SB		 3220 close_ctree(tree_root);
3221 return ret;
3222 }
3223 } else {
afcdd129 3224 set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags);
f7a81ea4 3225 }
afcdd129 3226 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
47ab2a6c 3227
8dcddfa0
QW		 3228 /*
3229 * backuproot only affect mount behavior, and if open_ctree succeeded,
3230 * no need to keep the flag
3231 */
3232 btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
3233
ad2b2c80 3234 return 0;
39279cc3 3235
bcef60f2
AJ		 3236fail_qgroup:
3237 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM		 3238fail_trans_kthread:
3239 kthread_stop(fs_info->transaction_kthread);
54067ae9 3240 btrfs_cleanup_transaction(fs_info->tree_root);
faa2dbf0 3241 btrfs_free_fs_roots(fs_info);
3f157a2f 3242fail_cleaner:
a74a4b97 3243 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 3244
3245 /*
3246 * make sure we're done with the btree inode before we stop our
3247 * kthreads
3248 */
3249 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 3250
2365dd3c 3251fail_sysfs:
6618a59b 3252 btrfs_sysfs_remove_mounted(fs_info);
2365dd3c 3253
b7c35e81
AJ		 3254fail_fsdev_sysfs:
3255 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
3256
1b1d1f66 3257fail_block_groups:
54067ae9 3258 btrfs_put_block_group_cache(fs_info);
1b1d1f66 3259 btrfs_free_block_groups(fs_info);
af31f5e5
CM		 3260
3261fail_tree_roots:
3262 free_root_pointers(fs_info, 1);
2b8195bb 3263 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 3264
39279cc3 3265fail_sb_buffer:
7abadb64 3266 btrfs_stop_all_workers(fs_info);
16cdcec7 3267fail_alloc:
4543df7e 3268fail_iput:
586e46e2
ID		 3269 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3270
4543df7e 3271 iput(fs_info->btree_inode);
c404e0dc
MX		 3272fail_bio_counter:
3273 percpu_counter_destroy(&fs_info->bio_counter);
963d678b
MX		 3274fail_delalloc_bytes:
3275 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX		 3276fail_dirty_metadata_bytes:
3277 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
ad081f14 3278fail_bdi:
7e662854 3279 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 3280fail_srcu:
3281 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 3282fail:
53b381b3 3283 btrfs_free_stripe_hash_table(fs_info);
586e46e2 3284 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 3285 return err;
af31f5e5
CM		 3286
3287recovery_tree_root:
3cdde224 3288 if (!btrfs_test_opt(tree_root->fs_info, USEBACKUPROOT))
af31f5e5
CM		 3289 goto fail_tree_roots;
3290
3291 free_root_pointers(fs_info, 0);
3292
3293 /* don't use the log in recovery mode, it won't be valid */
3294 btrfs_set_super_log_root(disk_super, 0);
3295
3296 /* we can't trust the free space cache either */
3297 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3298
3299 ret = next_root_backup(fs_info, fs_info->super_copy,
3300 &num_backups_tried, &backup_index);
3301 if (ret == -1)
3302 goto fail_block_groups;
3303 goto retry_root_backup;
eb60ceac
CM		 3304}
3305
f2984462
CM		 3306static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3307{
f2984462
CM		 3308 if (uptodate) {
3309 set_buffer_uptodate(bh);
3310 } else {
442a4f63
SB		 3311 struct btrfs_device *device = (struct btrfs_device *)
3312 bh->b_private;
3313
b14af3b4
DS		 3314 btrfs_warn_rl_in_rcu(device->dev_root->fs_info,
3315 "lost page write due to IO error on %s",
606686ee 3316 rcu_str_deref(device->name));
01327610 3317 /* note, we don't set_buffer_write_io_error because we have
1259ab75
CM		 3318 * our own ways of dealing with the IO errors
3319 */
f2984462 3320 clear_buffer_uptodate(bh);
442a4f63 3321 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM		 3322 }
3323 unlock_buffer(bh);
3324 put_bh(bh);
3325}
3326
29c36d72
AJ		 3327int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num,
3328 struct buffer_head **bh_ret)
3329{
3330 struct buffer_head *bh;
3331 struct btrfs_super_block *super;
3332 u64 bytenr;
3333
3334 bytenr = btrfs_sb_offset(copy_num);
3335 if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode))
3336 return -EINVAL;
3337
3338 bh = __bread(bdev, bytenr / 4096, BTRFS_SUPER_INFO_SIZE);
3339 /*
3340 * If we fail to read from the underlying devices, as of now
3341 * the best option we have is to mark it EIO.
3342 */
3343 if (!bh)
3344 return -EIO;
3345
3346 super = (struct btrfs_super_block *)bh->b_data;
3347 if (btrfs_super_bytenr(super) != bytenr ||
3348 btrfs_super_magic(super) != BTRFS_MAGIC) {
3349 brelse(bh);
3350 return -EINVAL;
3351 }
3352
3353 *bh_ret = bh;
3354 return 0;
3355}
3356
3357
a512bbf8
YZ		 3358struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3359{
3360 struct buffer_head *bh;
3361 struct buffer_head *latest = NULL;
3362 struct btrfs_super_block *super;
3363 int i;
3364 u64 transid = 0;
92fc03fb 3365 int ret = -EINVAL;
a512bbf8
YZ		 3366
3367 /* we would like to check all the supers, but that would make
3368 * a btrfs mount succeed after a mkfs from a different FS.
3369 * So, we need to add a special mount option to scan for
3370 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3371 */
3372 for (i = 0; i < 1; i++) {
29c36d72
AJ		 3373 ret = btrfs_read_dev_one_super(bdev, i, &bh);
3374 if (ret)
a512bbf8
YZ		 3375 continue;
3376
3377 super = (struct btrfs_super_block *)bh->b_data;
a512bbf8
YZ		 3378
3379 if (!latest || btrfs_super_generation(super) > transid) {
3380 brelse(latest);
3381 latest = bh;
3382 transid = btrfs_super_generation(super);
3383 } else {
3384 brelse(bh);
3385 }
3386 }
92fc03fb
AJ		 3387
3388 if (!latest)
3389 return ERR_PTR(ret);
3390
a512bbf8
YZ		 3391 return latest;
3392}
3393
4eedeb75
HH		 3394/*
3395 * this should be called twice, once with wait == 0 and
3396 * once with wait == 1. When wait == 0 is done, all the buffer heads
3397 * we write are pinned.
3398 *
3399 * They are released when wait == 1 is done.
3400 * max_mirrors must be the same for both runs, and it indicates how
3401 * many supers on this one device should be written.
3402 *
3403 * max_mirrors == 0 means to write them all.
3404 */
a512bbf8
YZ		 3405static int write_dev_supers(struct btrfs_device *device,
3406 struct btrfs_super_block *sb,
3407 int do_barriers, int wait, int max_mirrors)
3408{
3409 struct buffer_head *bh;
3410 int i;
3411 int ret;
3412 int errors = 0;
3413 u32 crc;
3414 u64 bytenr;
a512bbf8
YZ		 3415
3416 if (max_mirrors == 0)
3417 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3418
a512bbf8
YZ		 3419 for (i = 0; i < max_mirrors; i++) {
3420 bytenr = btrfs_sb_offset(i);
935e5cc9
MX		 3421 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3422 device->commit_total_bytes)
a512bbf8
YZ		 3423 break;
3424
3425 if (wait) {
3426 bh = __find_get_block(device->bdev, bytenr / 4096,
3427 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3428 if (!bh) {
3429 errors++;
3430 continue;
3431 }
a512bbf8 3432 wait_on_buffer(bh);
4eedeb75
HH		 3433 if (!buffer_uptodate(bh))
3434 errors++;
3435
3436 /* drop our reference */
3437 brelse(bh);
3438
3439 /* drop the reference from the wait == 0 run */
3440 brelse(bh);
3441 continue;
a512bbf8
YZ		 3442 } else {
3443 btrfs_set_super_bytenr(sb, bytenr);
3444
3445 crc = ~(u32)0;
b0496686 3446 crc = btrfs_csum_data((char *)sb +
a512bbf8
YZ		 3447 BTRFS_CSUM_SIZE, crc,
3448 BTRFS_SUPER_INFO_SIZE -
3449 BTRFS_CSUM_SIZE);
3450 btrfs_csum_final(crc, sb->csum);
3451
4eedeb75
HH		 3452 /*
3453 * one reference for us, and we leave it for the
3454 * caller
3455 */
a512bbf8
YZ		 3456 bh = __getblk(device->bdev, bytenr / 4096,
3457 BTRFS_SUPER_INFO_SIZE);
634554dc 3458 if (!bh) {
f14d104d
DS		 3459 btrfs_err(device->dev_root->fs_info,
3460 "couldn't get super buffer head for bytenr %llu",
3461 bytenr);
634554dc
JB		 3462 errors++;
3463 continue;
3464 }
3465
a512bbf8
YZ		 3466 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3467
4eedeb75 3468 /* one reference for submit_bh */
a512bbf8 3469 get_bh(bh);
4eedeb75
HH		 3470
3471 set_buffer_uptodate(bh);
a512bbf8
YZ		 3472 lock_buffer(bh);
3473 bh->b_end_io = btrfs_end_buffer_write_sync;
442a4f63 3474 bh->b_private = device;
a512bbf8
YZ		 3475 }
3476
387125fc
CM		 3477 /*
3478 * we fua the first super. The others we allow
3479 * to go down lazy.
3480 */
e8117c26 3481 if (i == 0)
2a222ca9 3482 ret = btrfsic_submit_bh(REQ_OP_WRITE, WRITE_FUA, bh);
e8117c26 3483 else
2a222ca9 3484 ret = btrfsic_submit_bh(REQ_OP_WRITE, WRITE_SYNC, bh);
4eedeb75 3485 if (ret)
a512bbf8 3486 errors++;
a512bbf8
YZ		 3487 }
3488 return errors < i ? 0 : -1;
3489}
3490
387125fc
CM		 3491/*
3492 * endio for the write_dev_flush, this will wake anyone waiting
3493 * for the barrier when it is done
3494 */
4246a0b6 3495static void btrfs_end_empty_barrier(struct bio *bio)
387125fc 3496{
387125fc
CM		 3497 if (bio->bi_private)
3498 complete(bio->bi_private);
3499 bio_put(bio);
3500}
3501
3502/*
3503 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3504 * sent down. With wait == 1, it waits for the previous flush.
3505 *
3506 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3507 * capable
3508 */
3509static int write_dev_flush(struct btrfs_device *device, int wait)
3510{
3511 struct bio *bio;
3512 int ret = 0;
3513
3514 if (device->nobarriers)
3515 return 0;
3516
3517 if (wait) {
3518 bio = device->flush_bio;
3519 if (!bio)
3520 return 0;
3521
3522 wait_for_completion(&device->flush_wait);
3523
4246a0b6
CH		 3524 if (bio->bi_error) {
3525 ret = bio->bi_error;
5af3e8cc
SB		 3526 btrfs_dev_stat_inc_and_print(device,
3527 BTRFS_DEV_STAT_FLUSH_ERRS);
387125fc
CM		 3528 }
3529
3530 /* drop the reference from the wait == 0 run */
3531 bio_put(bio);
3532 device->flush_bio = NULL;
3533
3534 return ret;
3535 }
3536
3537 /*
3538 * one reference for us, and we leave it for the
3539 * caller
3540 */
9c017abc 3541 device->flush_bio = NULL;
9be3395b 3542 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
387125fc
CM		 3543 if (!bio)
3544 return -ENOMEM;
3545
3546 bio->bi_end_io = btrfs_end_empty_barrier;
3547 bio->bi_bdev = device->bdev;
37226b21 3548 bio_set_op_attrs(bio, REQ_OP_WRITE, WRITE_FLUSH);
387125fc
CM		 3549 init_completion(&device->flush_wait);
3550 bio->bi_private = &device->flush_wait;
3551 device->flush_bio = bio;
3552
3553 bio_get(bio);
4e49ea4a 3554 btrfsic_submit_bio(bio);
387125fc
CM		 3555
3556 return 0;
3557}
3558
3559/*
3560 * send an empty flush down to each device in parallel,
3561 * then wait for them
3562 */
3563static int barrier_all_devices(struct btrfs_fs_info *info)
3564{
3565 struct list_head *head;
3566 struct btrfs_device *dev;
5af3e8cc
SB		 3567 int errors_send = 0;
3568 int errors_wait = 0;
387125fc
CM		 3569 int ret;
3570
3571 /* send down all the barriers */
3572 head = &info->fs_devices->devices;
3573 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3574 if (dev->missing)
3575 continue;
387125fc 3576 if (!dev->bdev) {
5af3e8cc 3577 errors_send++;
387125fc
CM		 3578 continue;
3579 }
3580 if (!dev->in_fs_metadata || !dev->writeable)
3581 continue;
3582
3583 ret = write_dev_flush(dev, 0);
3584 if (ret)
5af3e8cc 3585 errors_send++;
387125fc
CM		 3586 }
3587
3588 /* wait for all the barriers */
3589 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3590 if (dev->missing)
3591 continue;
387125fc 3592 if (!dev->bdev) {
5af3e8cc 3593 errors_wait++;
387125fc
CM		 3594 continue;
3595 }
3596 if (!dev->in_fs_metadata || !dev->writeable)
3597 continue;
3598
3599 ret = write_dev_flush(dev, 1);
3600 if (ret)
5af3e8cc 3601 errors_wait++;
387125fc 3602 }
5af3e8cc
SB		 3603 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3604 errors_wait > info->num_tolerated_disk_barrier_failures)
387125fc
CM		 3605 return -EIO;
3606 return 0;
3607}
3608
943c6e99
ZL		 3609int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags)
3610{
8789f4fe
ZL		 3611 int raid_type;
3612 int min_tolerated = INT_MAX;
943c6e99 3613
8789f4fe
ZL		 3614 if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 ||
3615 (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE))
3616 min_tolerated = min(min_tolerated,
3617 btrfs_raid_array[BTRFS_RAID_SINGLE].
3618 tolerated_failures);
943c6e99 3619
8789f4fe
ZL		 3620 for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) {
3621 if (raid_type == BTRFS_RAID_SINGLE)
3622 continue;
3623 if (!(flags & btrfs_raid_group[raid_type]))
3624 continue;
3625 min_tolerated = min(min_tolerated,
3626 btrfs_raid_array[raid_type].
3627 tolerated_failures);
3628 }
943c6e99 3629
8789f4fe 3630 if (min_tolerated == INT_MAX) {
ab8d0fc4 3631 pr_warn("BTRFS: unknown raid flag: %llu", flags);
8789f4fe
ZL		 3632 min_tolerated = 0;
3633 }
3634
3635 return min_tolerated;
943c6e99
ZL		 3636}
3637
5af3e8cc
SB		 3638int btrfs_calc_num_tolerated_disk_barrier_failures(
3639 struct btrfs_fs_info *fs_info)
3640{
3641 struct btrfs_ioctl_space_info space;
3642 struct btrfs_space_info *sinfo;
3643 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3644 BTRFS_BLOCK_GROUP_SYSTEM,
3645 BTRFS_BLOCK_GROUP_METADATA,
3646 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
5af3e8cc
SB		 3647 int i;
3648 int c;
3649 int num_tolerated_disk_barrier_failures =
3650 (int)fs_info->fs_devices->num_devices;
3651
2c458045 3652 for (i = 0; i < ARRAY_SIZE(types); i++) {
5af3e8cc
SB		 3653 struct btrfs_space_info *tmp;
3654
3655 sinfo = NULL;
3656 rcu_read_lock();
3657 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3658 if (tmp->flags == types[i]) {
3659 sinfo = tmp;
3660 break;
3661 }
3662 }
3663 rcu_read_unlock();
3664
3665 if (!sinfo)
3666 continue;
3667
3668 down_read(&sinfo->groups_sem);
3669 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2c458045
ZL		 3670 u64 flags;
3671
3672 if (list_empty(&sinfo->block_groups[c]))
3673 continue;
3674
3675 btrfs_get_block_group_info(&sinfo->block_groups[c],
3676 &space);
3677 if (space.total_bytes == 0 || space.used_bytes == 0)
3678 continue;
3679 flags = space.flags;
943c6e99
ZL		 3680
3681 num_tolerated_disk_barrier_failures = min(
3682 num_tolerated_disk_barrier_failures,
3683 btrfs_get_num_tolerated_disk_barrier_failures(
3684 flags));
5af3e8cc
SB		 3685 }
3686 up_read(&sinfo->groups_sem);
3687 }
3688
3689 return num_tolerated_disk_barrier_failures;
3690}
3691
48a3b636 3692static int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 3693{
e5e9a520 3694 struct list_head *head;
f2984462 3695 struct btrfs_device *dev;
a061fc8d 3696 struct btrfs_super_block *sb;
f2984462 3697 struct btrfs_dev_item *dev_item;
f2984462
CM		 3698 int ret;
3699 int do_barriers;
a236aed1
CM		 3700 int max_errors;
3701 int total_errors = 0;
a061fc8d 3702 u64 flags;
f2984462 3703
3cdde224 3704 do_barriers = !btrfs_test_opt(root->fs_info, NOBARRIER);
af31f5e5 3705 backup_super_roots(root->fs_info);
f2984462 3706
6c41761f 3707 sb = root->fs_info->super_for_commit;
a061fc8d 3708 dev_item = &sb->dev_item;
e5e9a520 3709
174ba509 3710 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 3711 head = &root->fs_info->fs_devices->devices;
d7306801 3712 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
387125fc 3713
5af3e8cc
SB		 3714 if (do_barriers) {
3715 ret = barrier_all_devices(root->fs_info);
3716 if (ret) {
3717 mutex_unlock(
3718 &root->fs_info->fs_devices->device_list_mutex);
34d97007 3719 btrfs_handle_fs_error(root->fs_info, ret,
5af3e8cc
SB		 3720 "errors while submitting device barriers.");
3721 return ret;
3722 }
3723 }
387125fc 3724
1f78160c 3725 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3726 if (!dev->bdev) {
3727 total_errors++;
3728 continue;
3729 }
2b82032c 3730 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3731 continue;
3732
2b82032c 3733 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 3734 btrfs_set_stack_device_type(dev_item, dev->type);
3735 btrfs_set_stack_device_id(dev_item, dev->devid);
7df69d3e 3736 btrfs_set_stack_device_total_bytes(dev_item,
935e5cc9 3737 dev->commit_total_bytes);
ce7213c7
MX		 3738 btrfs_set_stack_device_bytes_used(dev_item,
3739 dev->commit_bytes_used);
a061fc8d
CM		 3740 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3741 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3742 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3743 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 3744 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 3745
a061fc8d
CM		 3746 flags = btrfs_super_flags(sb);
3747 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3748
a512bbf8 3749 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 3750 if (ret)
3751 total_errors++;
f2984462 3752 }
a236aed1 3753 if (total_errors > max_errors) {
efe120a0 3754 btrfs_err(root->fs_info, "%d errors while writing supers",
d397712b 3755 total_errors);
a724b436 3756 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
79787eaa 3757
9d565ba4 3758 /* FUA is masked off if unsupported and can't be the reason */
34d97007 3759 btrfs_handle_fs_error(root->fs_info, -EIO,
9d565ba4
SB		 3760 "%d errors while writing supers", total_errors);
3761 return -EIO;
a236aed1 3762 }
f2984462 3763
a512bbf8 3764 total_errors = 0;
1f78160c 3765 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3766 if (!dev->bdev)
3767 continue;
2b82032c 3768 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3769 continue;
3770
a512bbf8
YZ		 3771 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3772 if (ret)
3773 total_errors++;
f2984462 3774 }
174ba509 3775 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 3776 if (total_errors > max_errors) {
34d97007 3777 btrfs_handle_fs_error(root->fs_info, -EIO,
79787eaa
JM		 3778 "%d errors while writing supers", total_errors);
3779 return -EIO;
a236aed1 3780 }
f2984462
CM		 3781 return 0;
3782}
3783
a512bbf8
YZ		 3784int write_ctree_super(struct btrfs_trans_handle *trans,
3785 struct btrfs_root *root, int max_mirrors)
eb60ceac 3786{
f570e757 3787 return write_all_supers(root, max_mirrors);
cfaa7295
CM		 3788}
3789
cb517eab
MX		 3790/* Drop a fs root from the radix tree and free it. */
3791void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3792 struct btrfs_root *root)
2619ba1f 3793{
4df27c4d 3794 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 3795 radix_tree_delete(&fs_info->fs_roots_radix,
3796 (unsigned long)root->root_key.objectid);
4df27c4d 3797 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 3798
3799 if (btrfs_root_refs(&root->root_item) == 0)
3800 synchronize_srcu(&fs_info->subvol_srcu);
3801
1c1ea4f7 3802 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
3321719e 3803 btrfs_free_log(NULL, root);
1c1ea4f7
LB		 3804 if (root->reloc_root) {
3805 free_extent_buffer(root->reloc_root->node);
3806 free_extent_buffer(root->reloc_root->commit_root);
3807 btrfs_put_fs_root(root->reloc_root);
3808 root->reloc_root = NULL;
3809 }
3810 }
3321719e 3811
faa2dbf0
JB		 3812 if (root->free_ino_pinned)
3813 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3814 if (root->free_ino_ctl)
3815 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3816 free_fs_root(root);
4df27c4d
YZ		 3817}
3818
3819static void free_fs_root(struct btrfs_root *root)
3820{
57cdc8db 3821 iput(root->ino_cache_inode);
4df27c4d 3822 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
1cb048f5
FDBM		 3823 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3824 root->orphan_block_rsv = NULL;
0ee5dc67
AV		 3825 if (root->anon_dev)
3826 free_anon_bdev(root->anon_dev);
8257b2dc
MX		 3827 if (root->subv_writers)
3828 btrfs_free_subvolume_writers(root->subv_writers);
4df27c4d
YZ		 3829 free_extent_buffer(root->node);
3830 free_extent_buffer(root->commit_root);
581bb050
LZ		 3831 kfree(root->free_ino_ctl);
3832 kfree(root->free_ino_pinned);
d397712b 3833 kfree(root->name);
b0feb9d9 3834 btrfs_put_fs_root(root);
2619ba1f
CM		 3835}
3836
cb517eab
MX		 3837void btrfs_free_fs_root(struct btrfs_root *root)
3838{
3839 free_fs_root(root);
2619ba1f
CM		 3840}
3841
c146afad 3842int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3843{
c146afad
YZ		 3844 u64 root_objectid = 0;
3845 struct btrfs_root *gang[8];
65d33fd7
QW		 3846 int i = 0;
3847 int err = 0;
3848 unsigned int ret = 0;
3849 int index;
e089f05c 3850
c146afad 3851 while (1) {
65d33fd7 3852 index = srcu_read_lock(&fs_info->subvol_srcu);
c146afad
YZ		 3853 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3854 (void **)gang, root_objectid,
3855 ARRAY_SIZE(gang));
65d33fd7
QW		 3856 if (!ret) {
3857 srcu_read_unlock(&fs_info->subvol_srcu, index);
c146afad 3858 break;
65d33fd7 3859 }
5d4f98a2 3860 root_objectid = gang[ret - 1]->root_key.objectid + 1;
65d33fd7 3861
c146afad 3862 for (i = 0; i < ret; i++) {
65d33fd7
QW		 3863 /* Avoid to grab roots in dead_roots */
3864 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3865 gang[i] = NULL;
3866 continue;
3867 }
3868 /* grab all the search result for later use */
3869 gang[i] = btrfs_grab_fs_root(gang[i]);
3870 }
3871 srcu_read_unlock(&fs_info->subvol_srcu, index);
66b4ffd1 3872
65d33fd7
QW		 3873 for (i = 0; i < ret; i++) {
3874 if (!gang[i])
3875 continue;
c146afad 3876 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 3877 err = btrfs_orphan_cleanup(gang[i]);
3878 if (err)
65d33fd7
QW		 3879 break;
3880 btrfs_put_fs_root(gang[i]);
c146afad
YZ		 3881 }
3882 root_objectid++;
3883 }
65d33fd7
QW		 3884
3885 /* release the uncleaned roots due to error */
3886 for (; i < ret; i++) {
3887 if (gang[i])
3888 btrfs_put_fs_root(gang[i]);
3889 }
3890 return err;
c146afad 3891}
a2135011 3892
c146afad
YZ		 3893int btrfs_commit_super(struct btrfs_root *root)
3894{
3895 struct btrfs_trans_handle *trans;
a74a4b97 3896
c146afad 3897 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 3898 btrfs_run_delayed_iputs(root);
c146afad 3899 mutex_unlock(&root->fs_info->cleaner_mutex);
9d1a2a3a 3900 wake_up_process(root->fs_info->cleaner_kthread);
c71bf099
YZ		 3901
3902 /* wait until ongoing cleanup work done */
3903 down_write(&root->fs_info->cleanup_work_sem);
3904 up_write(&root->fs_info->cleanup_work_sem);
3905
7a7eaa40 3906 trans = btrfs_join_transaction(root);
3612b495
TI		 3907 if (IS_ERR(trans))
3908 return PTR_ERR(trans);
d52c1bcc 3909 return btrfs_commit_transaction(trans, root);
c146afad
YZ		 3910}
3911
3abdbd78 3912void close_ctree(struct btrfs_root *root)
c146afad
YZ		 3913{
3914 struct btrfs_fs_info *fs_info = root->fs_info;
3915 int ret;
3916
afcdd129 3917 set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags);
c146afad 3918
7343dd61 3919 /* wait for the qgroup rescan worker to stop */
d06f23d6 3920 btrfs_qgroup_wait_for_completion(fs_info, false);
7343dd61 3921
803b2f54
SB		 3922 /* wait for the uuid_scan task to finish */
3923 down(&fs_info->uuid_tree_rescan_sem);
3924 /* avoid complains from lockdep et al., set sem back to initial state */
3925 up(&fs_info->uuid_tree_rescan_sem);
3926
837d5b6e 3927 /* pause restriper - we want to resume on mount */
aa1b8cd4 3928 btrfs_pause_balance(fs_info);
837d5b6e 3929
8dabb742
SB		 3930 btrfs_dev_replace_suspend_for_unmount(fs_info);
3931
aa1b8cd4 3932 btrfs_scrub_cancel(fs_info);
4cb5300b
CM		 3933
3934 /* wait for any defraggers to finish */
3935 wait_event(fs_info->transaction_wait,
3936 (atomic_read(&fs_info->defrag_running) == 0));
3937
3938 /* clear out the rbtree of defraggable inodes */
26176e7c 3939 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3940
21c7e756
MX		 3941 cancel_work_sync(&fs_info->async_reclaim_work);
3942
c146afad 3943 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
e44163e1
JM		 3944 /*
3945 * If the cleaner thread is stopped and there are
3946 * block groups queued for removal, the deletion will be
3947 * skipped when we quit the cleaner thread.
3948 */
e44163e1 3949 btrfs_delete_unused_bgs(root->fs_info);
e44163e1 3950
acce952b 3951 ret = btrfs_commit_super(root);
3952 if (ret)
04892340 3953 btrfs_err(fs_info, "commit super ret %d", ret);
acce952b 3954 }
3955
87533c47 3956 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
68ce9682 3957 btrfs_error_commit_super(root);
0f7d52f4 3958
e3029d9f
AV		 3959 kthread_stop(fs_info->transaction_kthread);
3960 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3961
afcdd129 3962 set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags);
f25784b3 3963
04892340 3964 btrfs_free_qgroup_config(fs_info);
bcef60f2 3965
963d678b 3966 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
04892340 3967 btrfs_info(fs_info, "at unmount delalloc count %lld",
963d678b 3968 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3969 }
bcc63abb 3970
6618a59b 3971 btrfs_sysfs_remove_mounted(fs_info);
b7c35e81 3972 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
5ac1d209 3973
faa2dbf0 3974 btrfs_free_fs_roots(fs_info);
d10c5f31 3975
1a4319cc
LB		 3976 btrfs_put_block_group_cache(fs_info);
3977
2b1360da
JB		 3978 btrfs_free_block_groups(fs_info);
3979
de348ee0
WS		 3980 /*
3981 * we must make sure there is not any read request to
3982 * submit after we stopping all workers.
3983 */
3984 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
96192499
JB		 3985 btrfs_stop_all_workers(fs_info);
3986
afcdd129 3987 clear_bit(BTRFS_FS_OPEN, &fs_info->flags);
13e6c37b 3988 free_root_pointers(fs_info, 1);
9ad6b7bc 3989
13e6c37b 3990 iput(fs_info->btree_inode);
d6bfde87 3991
21adbd5c 3992#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3cdde224 3993 if (btrfs_test_opt(root->fs_info, CHECK_INTEGRITY))
21adbd5c
SB		 3994 btrfsic_unmount(root, fs_info->fs_devices);
3995#endif
3996
dfe25020 3997 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3998 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3999
e2d84521 4000 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 4001 percpu_counter_destroy(&fs_info->delalloc_bytes);
c404e0dc 4002 percpu_counter_destroy(&fs_info->bio_counter);
04160088 4003 bdi_destroy(&fs_info->bdi);
76dda93c 4004 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 4005
53b381b3
DW		 4006 btrfs_free_stripe_hash_table(fs_info);
4007
cdfb080e 4008 __btrfs_free_block_rsv(root->orphan_block_rsv);
1cb048f5 4009 root->orphan_block_rsv = NULL;
04216820
FM		 4010
4011 lock_chunks(root);
4012 while (!list_empty(&fs_info->pinned_chunks)) {
4013 struct extent_map *em;
4014
4015 em = list_first_entry(&fs_info->pinned_chunks,
4016 struct extent_map, list);
4017 list_del_init(&em->list);
4018 free_extent_map(em);
4019 }
4020 unlock_chunks(root);
eb60ceac
CM		 4021}
4022
b9fab919
CM		 4023int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
4024 int atomic)
5f39d397 4025{
1259ab75 4026 int ret;
727011e0 4027 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 4028
0b32f4bb 4029 ret = extent_buffer_uptodate(buf);
1259ab75
CM		 4030 if (!ret)
4031 return ret;
4032
4033 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM		 4034 parent_transid, atomic);
4035 if (ret == -EAGAIN)
4036 return ret;
1259ab75 4037 return !ret;
5f39d397
CM		 4038}
4039
5f39d397
CM		 4040void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
4041{
06ea65a3 4042 struct btrfs_root *root;
5f39d397 4043 u64 transid = btrfs_header_generation(buf);
b9473439 4044 int was_dirty;
b4ce94de 4045
06ea65a3
JB		 4046#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
4047 /*
4048 * This is a fast path so only do this check if we have sanity tests
4049 * enabled. Normal people shouldn't be marking dummy buffers as dirty
4050 * outside of the sanity tests.
4051 */
4052 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
4053 return;
4054#endif
4055 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
b9447ef8 4056 btrfs_assert_tree_locked(buf);
31b1a2bd 4057 if (transid != root->fs_info->generation)
5d163e0e 4058 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n",
c1c9ff7c 4059 buf->start, transid, root->fs_info->generation);
0b32f4bb 4060 was_dirty = set_extent_buffer_dirty(buf);
e2d84521
MX		 4061 if (!was_dirty)
4062 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
4063 buf->len,
4064 root->fs_info->dirty_metadata_batch);
1f21ef0a
FM		 4065#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
4066 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
4067 btrfs_print_leaf(root, buf);
4068 ASSERT(0);
4069 }
4070#endif
eb60ceac
CM		 4071}
4072
b53d3f5d
LB		 4073static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
4074 int flush_delayed)
16cdcec7
MX		 4075{
4076 /*
4077 * looks as though older kernels can get into trouble with
4078 * this code, they end up stuck in balance_dirty_pages forever
4079 */
e2d84521 4080 int ret;
16cdcec7
MX		 4081
4082 if (current->flags & PF_MEMALLOC)
4083 return;
4084
b53d3f5d
LB		 4085 if (flush_delayed)
4086 btrfs_balance_delayed_items(root);
16cdcec7 4087
e2d84521
MX		 4088 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
4089 BTRFS_DIRTY_METADATA_THRESH);
4090 if (ret > 0) {
d0e1d66b
NJ		 4091 balance_dirty_pages_ratelimited(
4092 root->fs_info->btree_inode->i_mapping);
16cdcec7 4093 }
16cdcec7
MX		 4094}
4095
b53d3f5d 4096void btrfs_btree_balance_dirty(struct btrfs_root *root)
35b7e476 4097{
b53d3f5d
LB		 4098 __btrfs_btree_balance_dirty(root, 1);
4099}
585ad2c3 4100
b53d3f5d
LB		 4101void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
4102{
4103 __btrfs_btree_balance_dirty(root, 0);
35b7e476 4104}
6b80053d 4105
ca7a79ad 4106int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 4107{
727011e0 4108 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
8436ea91 4109 return btree_read_extent_buffer_pages(root, buf, parent_transid);
6b80053d 4110}
0da5468f 4111
fcd1f065 4112static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 4113 int read_only)
4114{
c926093e 4115 struct btrfs_super_block *sb = fs_info->super_copy;
319e4d06
QW		 4116 u64 nodesize = btrfs_super_nodesize(sb);
4117 u64 sectorsize = btrfs_super_sectorsize(sb);
c926093e
DS		 4118 int ret = 0;
4119
319e4d06 4120 if (btrfs_super_magic(sb) != BTRFS_MAGIC) {
ab8d0fc4 4121 btrfs_err(fs_info, "no valid FS found");
319e4d06
QW		 4122 ret = -EINVAL;
4123 }
4124 if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP)
ab8d0fc4 4125 btrfs_warn(fs_info, "unrecognized super flag: %llu",
319e4d06 4126 btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP);
21e7626b 4127 if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) {
ab8d0fc4 4128 btrfs_err(fs_info, "tree_root level too big: %d >= %d",
21e7626b 4129 btrfs_super_root_level(sb), BTRFS_MAX_LEVEL);
c926093e
DS		 4130 ret = -EINVAL;
4131 }
21e7626b 4132 if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) {
ab8d0fc4 4133 btrfs_err(fs_info, "chunk_root level too big: %d >= %d",
21e7626b 4134 btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL);
c926093e
DS		 4135 ret = -EINVAL;
4136 }
21e7626b 4137 if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) {
ab8d0fc4 4138 btrfs_err(fs_info, "log_root level too big: %d >= %d",
21e7626b 4139 btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL);
c926093e
DS		 4140 ret = -EINVAL;
4141 }
4142
1104a885 4143 /*
319e4d06
QW		 4144 * Check sectorsize and nodesize first, other check will need it.
4145 * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here.
1104a885 4146 */
319e4d06
QW		 4147 if (!is_power_of_2(sectorsize) || sectorsize < 4096 ||
4148 sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) {
ab8d0fc4 4149 btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize);
319e4d06
QW		 4150 ret = -EINVAL;
4151 }
4152 /* Only PAGE SIZE is supported yet */
09cbfeaf 4153 if (sectorsize != PAGE_SIZE) {
ab8d0fc4
JM		 4154 btrfs_err(fs_info,
4155 "sectorsize %llu not supported yet, only support %lu",
4156 sectorsize, PAGE_SIZE);
319e4d06
QW		 4157 ret = -EINVAL;
4158 }
4159 if (!is_power_of_2(nodesize) || nodesize < sectorsize ||
4160 nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
ab8d0fc4 4161 btrfs_err(fs_info, "invalid nodesize %llu", nodesize);
319e4d06
QW		 4162 ret = -EINVAL;
4163 }
4164 if (nodesize != le32_to_cpu(sb->__unused_leafsize)) {
ab8d0fc4
JM		 4165 btrfs_err(fs_info, "invalid leafsize %u, should be %llu",
4166 le32_to_cpu(sb->__unused_leafsize), nodesize);
319e4d06
QW		 4167 ret = -EINVAL;
4168 }
4169
4170 /* Root alignment check */
4171 if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) {
ab8d0fc4
JM		 4172 btrfs_warn(fs_info, "tree_root block unaligned: %llu",
4173 btrfs_super_root(sb));
319e4d06
QW		 4174 ret = -EINVAL;
4175 }
4176 if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) {
ab8d0fc4
JM		 4177 btrfs_warn(fs_info, "chunk_root block unaligned: %llu",
4178 btrfs_super_chunk_root(sb));
75d6ad38
DS		 4179 ret = -EINVAL;
4180 }
319e4d06 4181 if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) {
ab8d0fc4
JM		 4182 btrfs_warn(fs_info, "log_root block unaligned: %llu",
4183 btrfs_super_log_root(sb));
75d6ad38
DS		 4184 ret = -EINVAL;
4185 }
4186
c926093e 4187 if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) {
ab8d0fc4
JM		 4188 btrfs_err(fs_info,
4189 "dev_item UUID does not match fsid: %pU != %pU",
4190 fs_info->fsid, sb->dev_item.fsid);
c926093e
DS		 4191 ret = -EINVAL;
4192 }
4193
4194 /*
4195 * Hint to catch really bogus numbers, bitflips or so, more exact checks are
4196 * done later
4197 */
99e3ecfc
LB		 4198 if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) {
4199 btrfs_err(fs_info, "bytes_used is too small %llu",
ab8d0fc4 4200 btrfs_super_bytes_used(sb));
99e3ecfc
LB		 4201 ret = -EINVAL;
4202 }
b7f67055 4203 if (!is_power_of_2(btrfs_super_stripesize(sb))) {
99e3ecfc 4204 btrfs_err(fs_info, "invalid stripesize %u",
ab8d0fc4 4205 btrfs_super_stripesize(sb));
99e3ecfc
LB		 4206 ret = -EINVAL;
4207 }
21e7626b 4208 if (btrfs_super_num_devices(sb) > (1UL << 31))
ab8d0fc4
JM		 4209 btrfs_warn(fs_info, "suspicious number of devices: %llu",
4210 btrfs_super_num_devices(sb));
75d6ad38 4211 if (btrfs_super_num_devices(sb) == 0) {
ab8d0fc4 4212 btrfs_err(fs_info, "number of devices is 0");
75d6ad38
DS		 4213 ret = -EINVAL;
4214 }
c926093e 4215
21e7626b 4216 if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) {
ab8d0fc4
JM		 4217 btrfs_err(fs_info, "super offset mismatch %llu != %u",
4218 btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET);
c926093e
DS		 4219 ret = -EINVAL;
4220 }
4221
ce7fca5f
DS		 4222 /*
4223 * Obvious sys_chunk_array corruptions, it must hold at least one key
4224 * and one chunk
4225 */
4226 if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) {
ab8d0fc4
JM		 4227 btrfs_err(fs_info, "system chunk array too big %u > %u",
4228 btrfs_super_sys_array_size(sb),
4229 BTRFS_SYSTEM_CHUNK_ARRAY_SIZE);
ce7fca5f
DS		 4230 ret = -EINVAL;
4231 }
4232 if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key)
4233 + sizeof(struct btrfs_chunk)) {
ab8d0fc4
JM		 4234 btrfs_err(fs_info, "system chunk array too small %u < %zu",
4235 btrfs_super_sys_array_size(sb),
4236 sizeof(struct btrfs_disk_key)
4237 + sizeof(struct btrfs_chunk));
ce7fca5f
DS		 4238 ret = -EINVAL;
4239 }
4240
c926093e
DS		 4241 /*
4242 * The generation is a global counter, we'll trust it more than the others
4243 * but it's still possible that it's the one that's wrong.
4244 */
21e7626b 4245 if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb))
ab8d0fc4
JM		 4246 btrfs_warn(fs_info,
4247 "suspicious: generation < chunk_root_generation: %llu < %llu",
4248 btrfs_super_generation(sb),
4249 btrfs_super_chunk_root_generation(sb));
21e7626b
DS		 4250 if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb)
4251 && btrfs_super_cache_generation(sb) != (u64)-1)
ab8d0fc4
JM		 4252 btrfs_warn(fs_info,
4253 "suspicious: generation < cache_generation: %llu < %llu",
4254 btrfs_super_generation(sb),
4255 btrfs_super_cache_generation(sb));
c926093e
DS		 4256
4257 return ret;
acce952b 4258}
4259
48a3b636 4260static void btrfs_error_commit_super(struct btrfs_root *root)
acce952b 4261{
acce952b 4262 mutex_lock(&root->fs_info->cleaner_mutex);
4263 btrfs_run_delayed_iputs(root);
4264 mutex_unlock(&root->fs_info->cleaner_mutex);
4265
4266 down_write(&root->fs_info->cleanup_work_sem);
4267 up_write(&root->fs_info->cleanup_work_sem);
4268
4269 /* cleanup FS via transaction */
4270 btrfs_cleanup_transaction(root);
acce952b 4271}
4272
143bede5 4273static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 4274{
acce952b 4275 struct btrfs_ordered_extent *ordered;
acce952b 4276
199c2a9c 4277 spin_lock(&root->ordered_extent_lock);
779880ef
JB		 4278 /*
4279 * This will just short circuit the ordered completion stuff which will
4280 * make sure the ordered extent gets properly cleaned up.
4281 */
199c2a9c 4282 list_for_each_entry(ordered, &root->ordered_extents,
779880ef
JB		 4283 root_extent_list)
4284 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
199c2a9c
MX		 4285 spin_unlock(&root->ordered_extent_lock);
4286}
4287
4288static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
4289{
4290 struct btrfs_root *root;
4291 struct list_head splice;
4292
4293 INIT_LIST_HEAD(&splice);
4294
4295 spin_lock(&fs_info->ordered_root_lock);
4296 list_splice_init(&fs_info->ordered_roots, &splice);
4297 while (!list_empty(&splice)) {
4298 root = list_first_entry(&splice, struct btrfs_root,
4299 ordered_root);
1de2cfde
JB		 4300 list_move_tail(&root->ordered_root,
4301 &fs_info->ordered_roots);
199c2a9c 4302
2a85d9ca 4303 spin_unlock(&fs_info->ordered_root_lock);
199c2a9c
MX		 4304 btrfs_destroy_ordered_extents(root);
4305
2a85d9ca
LB		 4306 cond_resched();
4307 spin_lock(&fs_info->ordered_root_lock);
199c2a9c
MX		 4308 }
4309 spin_unlock(&fs_info->ordered_root_lock);
acce952b 4310}
4311
35a3621b
SB		 4312static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
4313 struct btrfs_root *root)
acce952b 4314{
4315 struct rb_node *node;
4316 struct btrfs_delayed_ref_root *delayed_refs;
4317 struct btrfs_delayed_ref_node *ref;
4318 int ret = 0;
4319
4320 delayed_refs = &trans->delayed_refs;
4321
4322 spin_lock(&delayed_refs->lock);
d7df2c79 4323 if (atomic_read(&delayed_refs->num_entries) == 0) {
cfece4db 4324 spin_unlock(&delayed_refs->lock);
efe120a0 4325 btrfs_info(root->fs_info, "delayed_refs has NO entry");
acce952b 4326 return ret;
4327 }
4328
d7df2c79
JB		 4329 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
4330 struct btrfs_delayed_ref_head *head;
c6fc2454 4331 struct btrfs_delayed_ref_node *tmp;
e78417d1 4332 bool pin_bytes = false;
acce952b 4333
d7df2c79
JB		 4334 head = rb_entry(node, struct btrfs_delayed_ref_head,
4335 href_node);
4336 if (!mutex_trylock(&head->mutex)) {
4337 atomic_inc(&head->node.refs);
4338 spin_unlock(&delayed_refs->lock);
eb12db69 4339
d7df2c79 4340 mutex_lock(&head->mutex);
e78417d1 4341 mutex_unlock(&head->mutex);
d7df2c79
JB		 4342 btrfs_put_delayed_ref(&head->node);
4343 spin_lock(&delayed_refs->lock);
4344 continue;
4345 }
4346 spin_lock(&head->lock);
c6fc2454
QW		 4347 list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list,
4348 list) {
d7df2c79 4349 ref->in_tree = 0;
c6fc2454 4350 list_del(&ref->list);
d7df2c79
JB		 4351 atomic_dec(&delayed_refs->num_entries);
4352 btrfs_put_delayed_ref(ref);
e78417d1 4353 }
d7df2c79
JB		 4354 if (head->must_insert_reserved)
4355 pin_bytes = true;
4356 btrfs_free_delayed_extent_op(head->extent_op);
4357 delayed_refs->num_heads--;
4358 if (head->processing == 0)
4359 delayed_refs->num_heads_ready--;
4360 atomic_dec(&delayed_refs->num_entries);
4361 head->node.in_tree = 0;
4362 rb_erase(&head->href_node, &delayed_refs->href_root);
4363 spin_unlock(&head->lock);
4364 spin_unlock(&delayed_refs->lock);
4365 mutex_unlock(&head->mutex);
acce952b 4366
d7df2c79
JB		 4367 if (pin_bytes)
4368 btrfs_pin_extent(root, head->node.bytenr,
4369 head->node.num_bytes, 1);
4370 btrfs_put_delayed_ref(&head->node);
acce952b 4371 cond_resched();
4372 spin_lock(&delayed_refs->lock);
4373 }
4374
4375 spin_unlock(&delayed_refs->lock);
4376
4377 return ret;
4378}
4379
143bede5 4380static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 4381{
4382 struct btrfs_inode *btrfs_inode;
4383 struct list_head splice;
4384
4385 INIT_LIST_HEAD(&splice);
4386
eb73c1b7
MX		 4387 spin_lock(&root->delalloc_lock);
4388 list_splice_init(&root->delalloc_inodes, &splice);
acce952b 4389
4390 while (!list_empty(&splice)) {
eb73c1b7
MX		 4391 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
4392 delalloc_inodes);
acce952b 4393
4394 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX		 4395 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
4396 &btrfs_inode->runtime_flags);
eb73c1b7 4397 spin_unlock(&root->delalloc_lock);
acce952b 4398
4399 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 4400
eb73c1b7 4401 spin_lock(&root->delalloc_lock);
acce952b 4402 }
4403
eb73c1b7
MX		 4404 spin_unlock(&root->delalloc_lock);
4405}
4406
4407static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
4408{
4409 struct btrfs_root *root;
4410 struct list_head splice;
4411
4412 INIT_LIST_HEAD(&splice);
4413
4414 spin_lock(&fs_info->delalloc_root_lock);
4415 list_splice_init(&fs_info->delalloc_roots, &splice);
4416 while (!list_empty(&splice)) {
4417 root = list_first_entry(&splice, struct btrfs_root,
4418 delalloc_root);
4419 list_del_init(&root->delalloc_root);
4420 root = btrfs_grab_fs_root(root);
4421 BUG_ON(!root);
4422 spin_unlock(&fs_info->delalloc_root_lock);
4423
4424 btrfs_destroy_delalloc_inodes(root);
4425 btrfs_put_fs_root(root);
4426
4427 spin_lock(&fs_info->delalloc_root_lock);
4428 }
4429 spin_unlock(&fs_info->delalloc_root_lock);
acce952b 4430}
4431
4432static int btrfs_destroy_marked_extents(struct btrfs_root *root,
4433 struct extent_io_tree *dirty_pages,
4434 int mark)
4435{
4436 int ret;
acce952b 4437 struct extent_buffer *eb;
4438 u64 start = 0;
4439 u64 end;
acce952b 4440
4441 while (1) {
4442 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 4443 mark, NULL);
acce952b 4444 if (ret)
4445 break;
4446
91166212 4447 clear_extent_bits(dirty_pages, start, end, mark);
acce952b 4448 while (start <= end) {
62d1f9fe 4449 eb = find_extent_buffer(root->fs_info, start);
707e8a07 4450 start += root->nodesize;
fd8b2b61 4451 if (!eb)
acce952b 4452 continue;
fd8b2b61 4453 wait_on_extent_buffer_writeback(eb);
acce952b 4454
fd8b2b61
JB		 4455 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4456 &eb->bflags))
4457 clear_extent_buffer_dirty(eb);
4458 free_extent_buffer_stale(eb);
acce952b 4459 }
4460 }
4461
4462 return ret;
4463}
4464
4465static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4466 struct extent_io_tree *pinned_extents)
4467{
4468 struct extent_io_tree *unpin;
4469 u64 start;
4470 u64 end;
4471 int ret;
ed0eaa14 4472 bool loop = true;
acce952b 4473
4474 unpin = pinned_extents;
ed0eaa14 4475again:
acce952b 4476 while (1) {
4477 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 4478 EXTENT_DIRTY, NULL);
acce952b 4479 if (ret)
4480 break;
4481
af6f8f60 4482 clear_extent_dirty(unpin, start, end);
acce952b 4483 btrfs_error_unpin_extent_range(root, start, end);
4484 cond_resched();
4485 }
4486
ed0eaa14
LB		 4487 if (loop) {
4488 if (unpin == &root->fs_info->freed_extents[0])
4489 unpin = &root->fs_info->freed_extents[1];
4490 else
4491 unpin = &root->fs_info->freed_extents[0];
4492 loop = false;
4493 goto again;
4494 }
4495
acce952b 4496 return 0;
4497}
4498
c79a1751
LB		 4499static void btrfs_cleanup_bg_io(struct btrfs_block_group_cache *cache)
4500{
4501 struct inode *inode;
4502
4503 inode = cache->io_ctl.inode;
4504 if (inode) {
4505 invalidate_inode_pages2(inode->i_mapping);
4506 BTRFS_I(inode)->generation = 0;
4507 cache->io_ctl.inode = NULL;
4508 iput(inode);
4509 }
4510 btrfs_put_block_group(cache);
4511}
4512
4513void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans,
4514 struct btrfs_root *root)
4515{
4516 struct btrfs_block_group_cache *cache;
4517
4518 spin_lock(&cur_trans->dirty_bgs_lock);
4519 while (!list_empty(&cur_trans->dirty_bgs)) {
4520 cache = list_first_entry(&cur_trans->dirty_bgs,
4521 struct btrfs_block_group_cache,
4522 dirty_list);
4523 if (!cache) {
4524 btrfs_err(root->fs_info,
4525 "orphan block group dirty_bgs list");
4526 spin_unlock(&cur_trans->dirty_bgs_lock);
4527 return;
4528 }
4529
4530 if (!list_empty(&cache->io_list)) {
4531 spin_unlock(&cur_trans->dirty_bgs_lock);
4532 list_del_init(&cache->io_list);
4533 btrfs_cleanup_bg_io(cache);
4534 spin_lock(&cur_trans->dirty_bgs_lock);
4535 }
4536
4537 list_del_init(&cache->dirty_list);
4538 spin_lock(&cache->lock);
4539 cache->disk_cache_state = BTRFS_DC_ERROR;
4540 spin_unlock(&cache->lock);
4541
4542 spin_unlock(&cur_trans->dirty_bgs_lock);
4543 btrfs_put_block_group(cache);
4544 spin_lock(&cur_trans->dirty_bgs_lock);
4545 }
4546 spin_unlock(&cur_trans->dirty_bgs_lock);
4547
4548 while (!list_empty(&cur_trans->io_bgs)) {
4549 cache = list_first_entry(&cur_trans->io_bgs,
4550 struct btrfs_block_group_cache,
4551 io_list);
4552 if (!cache) {
4553 btrfs_err(root->fs_info,
4554 "orphan block group on io_bgs list");
4555 return;
4556 }
4557
4558 list_del_init(&cache->io_list);
4559 spin_lock(&cache->lock);
4560 cache->disk_cache_state = BTRFS_DC_ERROR;
4561 spin_unlock(&cache->lock);
4562 btrfs_cleanup_bg_io(cache);
4563 }
4564}
4565
49b25e05
JM		 4566void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4567 struct btrfs_root *root)
4568{
c79a1751
LB		 4569 btrfs_cleanup_dirty_bgs(cur_trans, root);
4570 ASSERT(list_empty(&cur_trans->dirty_bgs));
4571 ASSERT(list_empty(&cur_trans->io_bgs));
4572
49b25e05 4573 btrfs_destroy_delayed_refs(cur_trans, root);
49b25e05 4574
4a9d8bde 4575 cur_trans->state = TRANS_STATE_COMMIT_START;
d7096fc3 4576 wake_up(&root->fs_info->transaction_blocked_wait);
49b25e05 4577
4a9d8bde 4578 cur_trans->state = TRANS_STATE_UNBLOCKED;
d7096fc3 4579 wake_up(&root->fs_info->transaction_wait);
49b25e05 4580
67cde344
MX		 4581 btrfs_destroy_delayed_inodes(root);
4582 btrfs_assert_delayed_root_empty(root);
49b25e05 4583
49b25e05
JM		 4584 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4585 EXTENT_DIRTY);
6e841e32
LB		 4586 btrfs_destroy_pinned_extent(root,
4587 root->fs_info->pinned_extents);
49b25e05 4588
4a9d8bde
MX		 4589 cur_trans->state =TRANS_STATE_COMPLETED;
4590 wake_up(&cur_trans->commit_wait);
4591
49b25e05
JM		 4592 /*
4593 memset(cur_trans, 0, sizeof(*cur_trans));
4594 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4595 */
4596}
4597
48a3b636 4598static int btrfs_cleanup_transaction(struct btrfs_root *root)
acce952b 4599{
4600 struct btrfs_transaction *t;
acce952b 4601
acce952b 4602 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4603
a4abeea4 4604 spin_lock(&root->fs_info->trans_lock);
724e2315
JB		 4605 while (!list_empty(&root->fs_info->trans_list)) {
4606 t = list_first_entry(&root->fs_info->trans_list,
4607 struct btrfs_transaction, list);
4608 if (t->state >= TRANS_STATE_COMMIT_START) {
4609 atomic_inc(&t->use_count);
4610 spin_unlock(&root->fs_info->trans_lock);
4611 btrfs_wait_for_commit(root, t->transid);
4612 btrfs_put_transaction(t);
4613 spin_lock(&root->fs_info->trans_lock);
4614 continue;
4615 }
4616 if (t == root->fs_info->running_transaction) {
4617 t->state = TRANS_STATE_COMMIT_DOING;
4618 spin_unlock(&root->fs_info->trans_lock);
4619 /*
4620 * We wait for 0 num_writers since we don't hold a trans
4621 * handle open currently for this transaction.
4622 */
4623 wait_event(t->writer_wait,
4624 atomic_read(&t->num_writers) == 0);
4625 } else {
4626 spin_unlock(&root->fs_info->trans_lock);
4627 }
4628 btrfs_cleanup_one_transaction(t, root);
4a9d8bde 4629
724e2315
JB		 4630 spin_lock(&root->fs_info->trans_lock);
4631 if (t == root->fs_info->running_transaction)
4632 root->fs_info->running_transaction = NULL;
acce952b 4633 list_del_init(&t->list);
724e2315 4634 spin_unlock(&root->fs_info->trans_lock);
acce952b 4635
724e2315
JB		 4636 btrfs_put_transaction(t);
4637 trace_btrfs_transaction_commit(root);
4638 spin_lock(&root->fs_info->trans_lock);
4639 }
4640 spin_unlock(&root->fs_info->trans_lock);
4641 btrfs_destroy_all_ordered_extents(root->fs_info);
4642 btrfs_destroy_delayed_inodes(root);
4643 btrfs_assert_delayed_root_empty(root);
4644 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4645 btrfs_destroy_all_delalloc_inodes(root->fs_info);
acce952b 4646 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 4647
4648 return 0;
4649}
4650
e8c9f186 4651static const struct extent_io_ops btree_extent_io_ops = {
ce9adaa5 4652 .readpage_end_io_hook = btree_readpage_end_io_hook,
4bb31e92 4653 .readpage_io_failed_hook = btree_io_failed_hook,
0b86a832 4654 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 4655 /* note we're sharing with inode.c for the merge bio hook */
4656 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 4657};
Linux 6.x block layer and io_uring tree(s)