projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: pass root to various extent ref mod functions
[linux-2.6-block.git] / fs / btrfs / disk-io.c
CommitLineData
6cbd5570
CM		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
e20d96d6 19#include <linux/fs.h>
d98237b3 20#include <linux/blkdev.h>
87cbda5c 21#include <linux/scatterlist.h>
22b0ebda 22#include <linux/swap.h>
0f7d52f4 23#include <linux/radix-tree.h>
35b7e476 24#include <linux/writeback.h>
d397712b 25#include <linux/buffer_head.h>
ce9adaa5 26#include <linux/workqueue.h>
a74a4b97 27#include <linux/kthread.h>
5a0e3ad6 28#include <linux/slab.h>
784b4e29 29#include <linux/migrate.h>
7a36ddec 30#include <linux/ratelimit.h>
6463fe58 31#include <linux/uuid.h>
803b2f54 32#include <linux/semaphore.h>
7e75bf3f 33#include <asm/unaligned.h>
eb60ceac
CM		 34#include "ctree.h"
35#include "disk-io.h"
0b947aff 36#include "hash.h"
e089f05c 37#include "transaction.h"
0f7d52f4 38#include "btrfs_inode.h"
0b86a832 39#include "volumes.h"
db94535d 40#include "print-tree.h"
925baedd 41#include "locking.h"
e02119d5 42#include "tree-log.h"
fa9c0d79 43#include "free-space-cache.h"
70f6d82e 44#include "free-space-tree.h"
581bb050 45#include "inode-map.h"
21adbd5c 46#include "check-integrity.h"
606686ee 47#include "rcu-string.h"
8dabb742 48#include "dev-replace.h"
53b381b3 49#include "raid56.h"
5ac1d209 50#include "sysfs.h"
fcebe456 51#include "qgroup.h"
ebb8765b 52#include "compression.h"
557ea5dd 53#include "tree-checker.h"
eb60ceac 54
de0022b9
JB		 55#ifdef CONFIG_X86
56#include <asm/cpufeature.h>
57#endif
58
319e4d06
QW		 59#define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\
60 BTRFS_HEADER_FLAG_RELOC |\
61 BTRFS_SUPER_FLAG_ERROR |\
62 BTRFS_SUPER_FLAG_SEEDING |\
63 BTRFS_SUPER_FLAG_METADUMP)
64
e8c9f186 65static const struct extent_io_ops btree_extent_io_ops;
8b712842 66static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 67static void free_fs_root(struct btrfs_root *root);
3d3a126a 68static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info);
143bede5 69static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
acce952b 70static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
2ff7e61e 71 struct btrfs_fs_info *fs_info);
143bede5 72static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
2ff7e61e 73static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info,
acce952b 74 struct extent_io_tree *dirty_pages,
75 int mark);
2ff7e61e 76static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info,
acce952b 77 struct extent_io_tree *pinned_extents);
2ff7e61e
JM		 78static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info);
79static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info);
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;
4e4cbee9 91 blk_status_t status;
bfebd8b5 92 enum btrfs_wq_endio_type metadata;
8b712842 93 struct btrfs_work work;
ce9adaa5 94};
0da5468f 95
97eb6b69
DS		 96static struct kmem_cache *btrfs_end_io_wq_cache;
97
98int __init btrfs_end_io_wq_init(void)
99{
100 btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq",
101 sizeof(struct btrfs_end_io_wq),
102 0,
fba4b697 103 SLAB_MEM_SPREAD,
97eb6b69
DS		 104 NULL);
105 if (!btrfs_end_io_wq_cache)
106 return -ENOMEM;
107 return 0;
108}
109
110void btrfs_end_io_wq_exit(void)
111{
5598e900 112 kmem_cache_destroy(btrfs_end_io_wq_cache);
97eb6b69
DS		 113}
114
d352ac68
CM		 115/*
116 * async submit bios are used to offload expensive checksumming
117 * onto the worker threads. They checksum file and metadata bios
118 * just before they are sent down the IO stack.
119 */
44b8bd7e 120struct async_submit_bio {
c6100a4b
JB		 121 void *private_data;
122 struct btrfs_fs_info *fs_info;
44b8bd7e 123 struct bio *bio;
4a69a410
CM		 124 extent_submit_bio_hook_t *submit_bio_start;
125 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e 126 int mirror_num;
c8b97818 127 unsigned long bio_flags;
eaf25d93
CM		 128 /*
129 * bio_offset is optional, can be used if the pages in the bio
130 * can't tell us where in the file the bio should go
131 */
132 u64 bio_offset;
8b712842 133 struct btrfs_work work;
4e4cbee9 134 blk_status_t status;
44b8bd7e
CM		 135};
136
85d4e461
CM		 137/*
138 * Lockdep class keys for extent_buffer->lock's in this root. For a given
139 * eb, the lockdep key is determined by the btrfs_root it belongs to and
140 * the level the eb occupies in the tree.
141 *
142 * Different roots are used for different purposes and may nest inside each
143 * other and they require separate keysets. As lockdep keys should be
144 * static, assign keysets according to the purpose of the root as indicated
145 * by btrfs_root->objectid. This ensures that all special purpose roots
146 * have separate keysets.
4008c04a 147 *
85d4e461
CM		 148 * Lock-nesting across peer nodes is always done with the immediate parent
149 * node locked thus preventing deadlock. As lockdep doesn't know this, use
150 * subclass to avoid triggering lockdep warning in such cases.
4008c04a 151 *
85d4e461
CM		 152 * The key is set by the readpage_end_io_hook after the buffer has passed
153 * csum validation but before the pages are unlocked. It is also set by
154 * btrfs_init_new_buffer on freshly allocated blocks.
4008c04a 155 *
85d4e461
CM		 156 * We also add a check to make sure the highest level of the tree is the
157 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
158 * needs update as well.
4008c04a
CM		 159 */
160#ifdef CONFIG_DEBUG_LOCK_ALLOC
161# if BTRFS_MAX_LEVEL != 8
162# error
163# endif
85d4e461
CM		 164
165static struct btrfs_lockdep_keyset {
166 u64 id; /* root objectid */
167 const char *name_stem; /* lock name stem */
168 char names[BTRFS_MAX_LEVEL + 1][20];
169 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
170} btrfs_lockdep_keysets[] = {
171 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
172 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
173 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
174 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
175 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
176 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
60b62978 177 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
85d4e461
CM		 178 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
179 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
180 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
13fd8da9 181 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
6b20e0ad 182 { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" },
85d4e461 183 { .id = 0, .name_stem = "tree" },
4008c04a 184};
85d4e461
CM		 185
186void __init btrfs_init_lockdep(void)
187{
188 int i, j;
189
190 /* initialize lockdep class names */
191 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
192 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
193
194 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
195 snprintf(ks->names[j], sizeof(ks->names[j]),
196 "btrfs-%s-%02d", ks->name_stem, j);
197 }
198}
199
200void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
201 int level)
202{
203 struct btrfs_lockdep_keyset *ks;
204
205 BUG_ON(level >= ARRAY_SIZE(ks->keys));
206
207 /* find the matching keyset, id 0 is the default entry */
208 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
209 if (ks->id == objectid)
210 break;
211
212 lockdep_set_class_and_name(&eb->lock,
213 &ks->keys[level], ks->names[level]);
214}
215
4008c04a
CM		 216#endif
217
d352ac68
CM		 218/*
219 * extents on the btree inode are pretty simple, there's one extent
220 * that covers the entire device
221 */
fc4f21b1 222static struct extent_map *btree_get_extent(struct btrfs_inode *inode,
306e16ce 223 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 224 int create)
7eccb903 225{
fc4f21b1
NB		 226 struct btrfs_fs_info *fs_info = btrfs_sb(inode->vfs_inode.i_sb);
227 struct extent_map_tree *em_tree = &inode->extent_tree;
5f39d397
CM		 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 233 if (em) {
0b246afa 234 em->bdev = fs_info->fs_devices->latest_bdev;
890871be 235 read_unlock(&em_tree->lock);
5f39d397 236 goto out;
a061fc8d 237 }
890871be 238 read_unlock(&em_tree->lock);
7b13b7b1 239
172ddd60 240 em = alloc_extent_map();
5f39d397
CM		 241 if (!em) {
242 em = ERR_PTR(-ENOMEM);
243 goto out;
244 }
245 em->start = 0;
0afbaf8c 246 em->len = (u64)-1;
c8b97818 247 em->block_len = (u64)-1;
5f39d397 248 em->block_start = 0;
0b246afa 249 em->bdev = fs_info->fs_devices->latest_bdev;
d1310b2e 250
890871be 251 write_lock(&em_tree->lock);
09a2a8f9 252 ret = add_extent_mapping(em_tree, em, 0);
5f39d397
CM		 253 if (ret == -EEXIST) {
254 free_extent_map(em);
7b13b7b1 255 em = lookup_extent_mapping(em_tree, start, len);
b4f359ab 256 if (!em)
0433f20d 257 em = ERR_PTR(-EIO);
5f39d397 258 } else if (ret) {
7b13b7b1 259 free_extent_map(em);
0433f20d 260 em = ERR_PTR(ret);
5f39d397 261 }
890871be 262 write_unlock(&em_tree->lock);
7b13b7b1 263
5f39d397
CM		 264out:
265 return em;
7eccb903
CM		 266}
267
9ed57367 268u32 btrfs_csum_data(const char *data, u32 seed, size_t len)
19c00ddc 269{
0b947aff 270 return btrfs_crc32c(seed, data, len);
19c00ddc
CM		 271}
272
0b5e3daf 273void btrfs_csum_final(u32 crc, u8 *result)
19c00ddc 274{
7e75bf3f 275 put_unaligned_le32(~crc, result);
19c00ddc
CM		 276}
277
d352ac68
CM		 278/*
279 * compute the csum for a btree block, and either verify it or write it
280 * into the csum field of the block.
281 */
01d58472
DD		 282static int csum_tree_block(struct btrfs_fs_info *fs_info,
283 struct extent_buffer *buf,
19c00ddc
CM		 284 int verify)
285{
01d58472 286 u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
607d432d 287 char *result = NULL;
19c00ddc
CM		 288 unsigned long len;
289 unsigned long cur_len;
290 unsigned long offset = BTRFS_CSUM_SIZE;
19c00ddc
CM		 291 char *kaddr;
292 unsigned long map_start;
293 unsigned long map_len;
294 int err;
295 u32 crc = ~(u32)0;
607d432d 296 unsigned long inline_result;
19c00ddc
CM		 297
298 len = buf->len - offset;
d397712b 299 while (len > 0) {
19c00ddc 300 err = map_private_extent_buffer(buf, offset, 32,
a6591715 301 &kaddr, &map_start, &map_len);
d397712b 302 if (err)
8bd98f0e 303 return err;
19c00ddc 304 cur_len = min(len, map_len - (offset - map_start));
b0496686 305 crc = btrfs_csum_data(kaddr + offset - map_start,
19c00ddc
CM		 306 crc, cur_len);
307 len -= cur_len;
308 offset += cur_len;
19c00ddc 309 }
607d432d 310 if (csum_size > sizeof(inline_result)) {
31e818fe 311 result = kzalloc(csum_size, GFP_NOFS);
607d432d 312 if (!result)
8bd98f0e 313 return -ENOMEM;
607d432d
JB		 314 } else {
315 result = (char *)&inline_result;
316 }
317
19c00ddc
CM		 318 btrfs_csum_final(crc, result);
319
320 if (verify) {
607d432d 321 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 322 u32 val;
323 u32 found = 0;
607d432d 324 memcpy(&found, result, csum_size);
e4204ded 325
607d432d 326 read_extent_buffer(buf, &val, 0, csum_size);
94647322 327 btrfs_warn_rl(fs_info,
5d163e0e 328 "%s checksum verify failed on %llu wanted %X found %X level %d",
01d58472 329 fs_info->sb->s_id, buf->start,
efe120a0 330 val, found, btrfs_header_level(buf));
607d432d
JB		 331 if (result != (char *)&inline_result)
332 kfree(result);
8bd98f0e 333 return -EUCLEAN;
19c00ddc
CM		 334 }
335 } else {
607d432d 336 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 337 }
607d432d
JB		 338 if (result != (char *)&inline_result)
339 kfree(result);
19c00ddc
CM		 340 return 0;
341}
342
d352ac68
CM		 343/*
344 * we can't consider a given block up to date unless the transid of the
345 * block matches the transid in the parent node's pointer. This is how we
346 * detect blocks that either didn't get written at all or got written
347 * in the wrong place.
348 */
1259ab75 349static int verify_parent_transid(struct extent_io_tree *io_tree,
b9fab919
CM		 350 struct extent_buffer *eb, u64 parent_transid,
351 int atomic)
1259ab75 352{
2ac55d41 353 struct extent_state *cached_state = NULL;
1259ab75 354 int ret;
2755a0de 355 bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB);
1259ab75
CM		 356
357 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
358 return 0;
359
b9fab919
CM		 360 if (atomic)
361 return -EAGAIN;
362
a26e8c9f
JB		 363 if (need_lock) {
364 btrfs_tree_read_lock(eb);
365 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
366 }
367
2ac55d41 368 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
ff13db41 369 &cached_state);
0b32f4bb 370 if (extent_buffer_uptodate(eb) &&
1259ab75
CM		 371 btrfs_header_generation(eb) == parent_transid) {
372 ret = 0;
373 goto out;
374 }
94647322
DS		 375 btrfs_err_rl(eb->fs_info,
376 "parent transid verify failed on %llu wanted %llu found %llu",
377 eb->start,
29549aec 378 parent_transid, btrfs_header_generation(eb));
1259ab75 379 ret = 1;
a26e8c9f
JB		 380
381 /*
382 * Things reading via commit roots that don't have normal protection,
383 * like send, can have a really old block in cache that may point at a
01327610 384 * block that has been freed and re-allocated. So don't clear uptodate
a26e8c9f
JB		 385 * if we find an eb that is under IO (dirty/writeback) because we could
386 * end up reading in the stale data and then writing it back out and
387 * making everybody very sad.
388 */
389 if (!extent_buffer_under_io(eb))
390 clear_extent_buffer_uptodate(eb);
33958dc6 391out:
2ac55d41
JB		 392 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
393 &cached_state, GFP_NOFS);
472b909f
JB		 394 if (need_lock)
395 btrfs_tree_read_unlock_blocking(eb);
1259ab75 396 return ret;
1259ab75
CM		 397}
398
1104a885
DS		 399/*
400 * Return 0 if the superblock checksum type matches the checksum value of that
401 * algorithm. Pass the raw disk superblock data.
402 */
ab8d0fc4
JM		 403static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info,
404 char *raw_disk_sb)
1104a885
DS		 405{
406 struct btrfs_super_block *disk_sb =
407 (struct btrfs_super_block *)raw_disk_sb;
408 u16 csum_type = btrfs_super_csum_type(disk_sb);
409 int ret = 0;
410
411 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
412 u32 crc = ~(u32)0;
413 const int csum_size = sizeof(crc);
414 char result[csum_size];
415
416 /*
417 * The super_block structure does not span the whole
418 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
01327610 419 * is filled with zeros and is included in the checksum.
1104a885
DS		 420 */
421 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
422 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
423 btrfs_csum_final(crc, result);
424
425 if (memcmp(raw_disk_sb, result, csum_size))
426 ret = 1;
427 }
428
429 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
ab8d0fc4 430 btrfs_err(fs_info, "unsupported checksum algorithm %u",
1104a885
DS		 431 csum_type);
432 ret = 1;
433 }
434
435 return ret;
436}
437
d352ac68
CM		 438/*
439 * helper to read a given tree block, doing retries as required when
440 * the checksums don't match and we have alternate mirrors to try.
441 */
2ff7e61e 442static int btree_read_extent_buffer_pages(struct btrfs_fs_info *fs_info,
f188591e 443 struct extent_buffer *eb,
8436ea91 444 u64 parent_transid)
f188591e
CM		 445{
446 struct extent_io_tree *io_tree;
ea466794 447 int failed = 0;
f188591e
CM		 448 int ret;
449 int num_copies = 0;
450 int mirror_num = 0;
ea466794 451 int failed_mirror = 0;
f188591e 452
a826d6dc 453 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
0b246afa 454 io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree;
f188591e 455 while (1) {
8436ea91 456 ret = read_extent_buffer_pages(io_tree, eb, WAIT_COMPLETE,
f188591e 457 btree_get_extent, mirror_num);
256dd1bb
SB		 458 if (!ret) {
459 if (!verify_parent_transid(io_tree, eb,
b9fab919 460 parent_transid, 0))
256dd1bb
SB		 461 break;
462 else
463 ret = -EIO;
464 }
d397712b 465
a826d6dc
JB		 466 /*
467 * This buffer's crc is fine, but its contents are corrupted, so
468 * there is no reason to read the other copies, they won't be
469 * any less wrong.
470 */
471 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
ea466794
JB		 472 break;
473
0b246afa 474 num_copies = btrfs_num_copies(fs_info,
f188591e 475 eb->start, eb->len);
4235298e 476 if (num_copies == 1)
ea466794 477 break;
4235298e 478
5cf1ab56
JB		 479 if (!failed_mirror) {
480 failed = 1;
481 failed_mirror = eb->read_mirror;
482 }
483
f188591e 484 mirror_num++;
ea466794
JB		 485 if (mirror_num == failed_mirror)
486 mirror_num++;
487
4235298e 488 if (mirror_num > num_copies)
ea466794 489 break;
f188591e 490 }
ea466794 491
c0901581 492 if (failed && !ret && failed_mirror)
2ff7e61e 493 repair_eb_io_failure(fs_info, eb, failed_mirror);
ea466794
JB		 494
495 return ret;
f188591e 496}
19c00ddc 497
d352ac68 498/*
d397712b
CM		 499 * checksum a dirty tree block before IO. This has extra checks to make sure
500 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 501 */
d397712b 502
01d58472 503static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page)
19c00ddc 504{
4eee4fa4 505 u64 start = page_offset(page);
19c00ddc 506 u64 found_start;
19c00ddc 507 struct extent_buffer *eb;
f188591e 508
4f2de97a
JB		 509 eb = (struct extent_buffer *)page->private;
510 if (page != eb->pages[0])
511 return 0;
0f805531 512
19c00ddc 513 found_start = btrfs_header_bytenr(eb);
0f805531
AL		 514 /*
515 * Please do not consolidate these warnings into a single if.
516 * It is useful to know what went wrong.
517 */
518 if (WARN_ON(found_start != start))
519 return -EUCLEAN;
520 if (WARN_ON(!PageUptodate(page)))
521 return -EUCLEAN;
522
523 ASSERT(memcmp_extent_buffer(eb, fs_info->fsid,
524 btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0);
525
8bd98f0e 526 return csum_tree_block(fs_info, eb, 0);
19c00ddc
CM		 527}
528
01d58472 529static int check_tree_block_fsid(struct btrfs_fs_info *fs_info,
2b82032c
YZ		 530 struct extent_buffer *eb)
531{
01d58472 532 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
44880fdc 533 u8 fsid[BTRFS_FSID_SIZE];
2b82032c
YZ		 534 int ret = 1;
535
0a4e5586 536 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
2b82032c
YZ		 537 while (fs_devices) {
538 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
539 ret = 0;
540 break;
541 }
542 fs_devices = fs_devices->seed;
543 }
544 return ret;
545}
546
facc8a22
MX		 547static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
548 u64 phy_offset, struct page *page,
549 u64 start, u64 end, int mirror)
ce9adaa5 550{
ce9adaa5
CM		 551 u64 found_start;
552 int found_level;
ce9adaa5
CM		 553 struct extent_buffer *eb;
554 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
02873e43 555 struct btrfs_fs_info *fs_info = root->fs_info;
f188591e 556 int ret = 0;
727011e0 557 int reads_done;
ce9adaa5 558
ce9adaa5
CM		 559 if (!page->private)
560 goto out;
d397712b 561
4f2de97a 562 eb = (struct extent_buffer *)page->private;
d397712b 563
0b32f4bb
JB		 564 /* the pending IO might have been the only thing that kept this buffer
565 * in memory. Make sure we have a ref for all this other checks
566 */
567 extent_buffer_get(eb);
568
569 reads_done = atomic_dec_and_test(&eb->io_pages);
727011e0
CM		 570 if (!reads_done)
571 goto err;
f188591e 572
5cf1ab56 573 eb->read_mirror = mirror;
656f30db 574 if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) {
ea466794
JB		 575 ret = -EIO;
576 goto err;
577 }
578
ce9adaa5 579 found_start = btrfs_header_bytenr(eb);
727011e0 580 if (found_start != eb->start) {
02873e43
ZL		 581 btrfs_err_rl(fs_info, "bad tree block start %llu %llu",
582 found_start, eb->start);
f188591e 583 ret = -EIO;
ce9adaa5
CM		 584 goto err;
585 }
02873e43
ZL		 586 if (check_tree_block_fsid(fs_info, eb)) {
587 btrfs_err_rl(fs_info, "bad fsid on block %llu",
588 eb->start);
1259ab75
CM		 589 ret = -EIO;
590 goto err;
591 }
ce9adaa5 592 found_level = btrfs_header_level(eb);
1c24c3ce 593 if (found_level >= BTRFS_MAX_LEVEL) {
02873e43
ZL		 594 btrfs_err(fs_info, "bad tree block level %d",
595 (int)btrfs_header_level(eb));
1c24c3ce
JB		 596 ret = -EIO;
597 goto err;
598 }
ce9adaa5 599
85d4e461
CM		 600 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
601 eb, found_level);
4008c04a 602
02873e43 603 ret = csum_tree_block(fs_info, eb, 1);
8bd98f0e 604 if (ret)
a826d6dc 605 goto err;
a826d6dc
JB		 606
607 /*
608 * If this is a leaf block and it is corrupt, set the corrupt bit so
609 * that we don't try and read the other copies of this block, just
610 * return -EIO.
611 */
557ea5dd 612 if (found_level == 0 && btrfs_check_leaf(root, eb)) {
a826d6dc
JB		 613 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
614 ret = -EIO;
615 }
ce9adaa5 616
557ea5dd 617 if (found_level > 0 && btrfs_check_node(root, eb))
053ab70f
LB		 618 ret = -EIO;
619
0b32f4bb
JB		 620 if (!ret)
621 set_extent_buffer_uptodate(eb);
ce9adaa5 622err:
79fb65a1
JB		 623 if (reads_done &&
624 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
d48d71aa 625 btree_readahead_hook(eb, ret);
4bb31e92 626
53b381b3
DW		 627 if (ret) {
628 /*
629 * our io error hook is going to dec the io pages
630 * again, we have to make sure it has something
631 * to decrement
632 */
633 atomic_inc(&eb->io_pages);
0b32f4bb 634 clear_extent_buffer_uptodate(eb);
53b381b3 635 }
0b32f4bb 636 free_extent_buffer(eb);
ce9adaa5 637out:
f188591e 638 return ret;
ce9adaa5
CM		 639}
640
ea466794 641static int btree_io_failed_hook(struct page *page, int failed_mirror)
4bb31e92 642{
4bb31e92 643 struct extent_buffer *eb;
4bb31e92 644
4f2de97a 645 eb = (struct extent_buffer *)page->private;
656f30db 646 set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
5cf1ab56 647 eb->read_mirror = failed_mirror;
53b381b3 648 atomic_dec(&eb->io_pages);
ea466794 649 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
d48d71aa 650 btree_readahead_hook(eb, -EIO);
4bb31e92
AJ		 651 return -EIO; /* we fixed nothing */
652}
653
4246a0b6 654static void end_workqueue_bio(struct bio *bio)
ce9adaa5 655{
97eb6b69 656 struct btrfs_end_io_wq *end_io_wq = bio->bi_private;
ce9adaa5 657 struct btrfs_fs_info *fs_info;
9e0af237
LB		 658 struct btrfs_workqueue *wq;
659 btrfs_work_func_t func;
ce9adaa5 660
ce9adaa5 661 fs_info = end_io_wq->info;
4e4cbee9 662 end_io_wq->status = bio->bi_status;
d20f7043 663
37226b21 664 if (bio_op(bio) == REQ_OP_WRITE) {
9e0af237
LB		 665 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) {
666 wq = fs_info->endio_meta_write_workers;
667 func = btrfs_endio_meta_write_helper;
668 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) {
669 wq = fs_info->endio_freespace_worker;
670 func = btrfs_freespace_write_helper;
671 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
672 wq = fs_info->endio_raid56_workers;
673 func = btrfs_endio_raid56_helper;
674 } else {
675 wq = fs_info->endio_write_workers;
676 func = btrfs_endio_write_helper;
677 }
d20f7043 678 } else {
8b110e39
MX		 679 if (unlikely(end_io_wq->metadata ==
680 BTRFS_WQ_ENDIO_DIO_REPAIR)) {
681 wq = fs_info->endio_repair_workers;
682 func = btrfs_endio_repair_helper;
683 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
9e0af237
LB		 684 wq = fs_info->endio_raid56_workers;
685 func = btrfs_endio_raid56_helper;
686 } else if (end_io_wq->metadata) {
687 wq = fs_info->endio_meta_workers;
688 func = btrfs_endio_meta_helper;
689 } else {
690 wq = fs_info->endio_workers;
691 func = btrfs_endio_helper;
692 }
d20f7043 693 }
9e0af237
LB		 694
695 btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL);
696 btrfs_queue_work(wq, &end_io_wq->work);
ce9adaa5
CM		 697}
698
4e4cbee9 699blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
bfebd8b5 700 enum btrfs_wq_endio_type metadata)
0b86a832 701{
97eb6b69 702 struct btrfs_end_io_wq *end_io_wq;
8b110e39 703
97eb6b69 704 end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS);
ce9adaa5 705 if (!end_io_wq)
4e4cbee9 706 return BLK_STS_RESOURCE;
ce9adaa5
CM		 707
708 end_io_wq->private = bio->bi_private;
709 end_io_wq->end_io = bio->bi_end_io;
22c59948 710 end_io_wq->info = info;
4e4cbee9 711 end_io_wq->status = 0;
ce9adaa5 712 end_io_wq->bio = bio;
22c59948 713 end_io_wq->metadata = metadata;
ce9adaa5
CM		 714
715 bio->bi_private = end_io_wq;
716 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 717 return 0;
718}
719
b64a2851 720unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 721{
4854ddd0 722 unsigned long limit = min_t(unsigned long,
5cdc7ad3 723 info->thread_pool_size,
4854ddd0
CM		 724 info->fs_devices->open_devices);
725 return 256 * limit;
726}
0986fe9e 727
4a69a410
CM		 728static void run_one_async_start(struct btrfs_work *work)
729{
4a69a410 730 struct async_submit_bio *async;
4e4cbee9 731 blk_status_t ret;
4a69a410
CM		 732
733 async = container_of(work, struct async_submit_bio, work);
c6100a4b 734 ret = async->submit_bio_start(async->private_data, async->bio,
79787eaa
JM		 735 async->mirror_num, async->bio_flags,
736 async->bio_offset);
737 if (ret)
4e4cbee9 738 async->status = ret;
4a69a410
CM		 739}
740
741static void run_one_async_done(struct btrfs_work *work)
8b712842 742{
8b712842
CM		 743 struct async_submit_bio *async;
744
745 async = container_of(work, struct async_submit_bio, work);
4854ddd0 746
bb7ab3b9 747 /* If an error occurred we just want to clean up the bio and move on */
4e4cbee9
CH		 748 if (async->status) {
749 async->bio->bi_status = async->status;
4246a0b6 750 bio_endio(async->bio);
79787eaa
JM		 751 return;
752 }
753
c6100a4b 754 async->submit_bio_done(async->private_data, async->bio, async->mirror_num,
81a75f67 755 async->bio_flags, async->bio_offset);
4a69a410
CM		 756}
757
758static void run_one_async_free(struct btrfs_work *work)
759{
760 struct async_submit_bio *async;
761
762 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 763 kfree(async);
764}
765
8c27cb35
LT		 766blk_status_t btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
767 int mirror_num, unsigned long bio_flags,
768 u64 bio_offset, void *private_data,
769 extent_submit_bio_hook_t *submit_bio_start,
770 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 771{
772 struct async_submit_bio *async;
773
774 async = kmalloc(sizeof(*async), GFP_NOFS);
775 if (!async)
4e4cbee9 776 return BLK_STS_RESOURCE;
44b8bd7e 777
c6100a4b
JB		 778 async->private_data = private_data;
779 async->fs_info = fs_info;
44b8bd7e
CM		 780 async->bio = bio;
781 async->mirror_num = mirror_num;
4a69a410
CM		 782 async->submit_bio_start = submit_bio_start;
783 async->submit_bio_done = submit_bio_done;
784
9e0af237 785 btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start,
5cdc7ad3 786 run_one_async_done, run_one_async_free);
4a69a410 787
c8b97818 788 async->bio_flags = bio_flags;
eaf25d93 789 async->bio_offset = bio_offset;
8c8bee1d 790
4e4cbee9 791 async->status = 0;
79787eaa 792
67f055c7 793 if (op_is_sync(bio->bi_opf))
5cdc7ad3 794 btrfs_set_work_high_priority(&async->work);
d313d7a3 795
5cdc7ad3 796 btrfs_queue_work(fs_info->workers, &async->work);
44b8bd7e
CM		 797 return 0;
798}
799
4e4cbee9 800static blk_status_t btree_csum_one_bio(struct bio *bio)
ce3ed71a 801{
2c30c71b 802 struct bio_vec *bvec;
ce3ed71a 803 struct btrfs_root *root;
2c30c71b 804 int i, ret = 0;
ce3ed71a 805
c09abff8 806 ASSERT(!bio_flagged(bio, BIO_CLONED));
2c30c71b 807 bio_for_each_segment_all(bvec, bio, i) {
ce3ed71a 808 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
01d58472 809 ret = csum_dirty_buffer(root->fs_info, bvec->bv_page);
79787eaa
JM		 810 if (ret)
811 break;
ce3ed71a 812 }
2c30c71b 813
4e4cbee9 814 return errno_to_blk_status(ret);
ce3ed71a
CM		 815}
816
8c27cb35
LT		 817static blk_status_t __btree_submit_bio_start(void *private_data, struct bio *bio,
818 int mirror_num, unsigned long bio_flags,
819 u64 bio_offset)
22c59948 820{
8b712842
CM		 821 /*
822 * when we're called for a write, we're already in the async
5443be45 823 * submission context. Just jump into btrfs_map_bio
8b712842 824 */
79787eaa 825 return btree_csum_one_bio(bio);
4a69a410 826}
22c59948 827
8c27cb35
LT		 828static blk_status_t __btree_submit_bio_done(void *private_data, struct bio *bio,
829 int mirror_num, unsigned long bio_flags,
830 u64 bio_offset)
4a69a410 831{
c6100a4b 832 struct inode *inode = private_data;
4e4cbee9 833 blk_status_t ret;
61891923 834
8b712842 835 /*
4a69a410
CM		 836 * when we're called for a write, we're already in the async
837 * submission context. Just jump into btrfs_map_bio
8b712842 838 */
2ff7e61e 839 ret = btrfs_map_bio(btrfs_sb(inode->i_sb), bio, mirror_num, 1);
4246a0b6 840 if (ret) {
4e4cbee9 841 bio->bi_status = ret;
4246a0b6
CH		 842 bio_endio(bio);
843 }
61891923 844 return ret;
0b86a832
CM		 845}
846
18fdc679 847static int check_async_write(struct btrfs_inode *bi)
de0022b9 848{
6300463b
LB		 849 if (atomic_read(&bi->sync_writers))
850 return 0;
de0022b9 851#ifdef CONFIG_X86
bc696ca0 852 if (static_cpu_has(X86_FEATURE_XMM4_2))
de0022b9
JB		 853 return 0;
854#endif
855 return 1;
856}
857
8c27cb35
LT		 858static blk_status_t btree_submit_bio_hook(void *private_data, struct bio *bio,
859 int mirror_num, unsigned long bio_flags,
860 u64 bio_offset)
44b8bd7e 861{
c6100a4b 862 struct inode *inode = private_data;
0b246afa 863 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
18fdc679 864 int async = check_async_write(BTRFS_I(inode));
4e4cbee9 865 blk_status_t ret;
cad321ad 866
37226b21 867 if (bio_op(bio) != REQ_OP_WRITE) {
4a69a410
CM		 868 /*
869 * called for a read, do the setup so that checksum validation
870 * can happen in the async kernel threads
871 */
0b246afa
JM		 872 ret = btrfs_bio_wq_end_io(fs_info, bio,
873 BTRFS_WQ_ENDIO_METADATA);
1d4284bd 874 if (ret)
61891923 875 goto out_w_error;
2ff7e61e 876 ret = btrfs_map_bio(fs_info, bio, mirror_num, 0);
de0022b9
JB		 877 } else if (!async) {
878 ret = btree_csum_one_bio(bio);
879 if (ret)
61891923 880 goto out_w_error;
2ff7e61e 881 ret = btrfs_map_bio(fs_info, bio, mirror_num, 0);
61891923
SB		 882 } else {
883 /*
884 * kthread helpers are used to submit writes so that
885 * checksumming can happen in parallel across all CPUs
886 */
c6100a4b
JB		 887 ret = btrfs_wq_submit_bio(fs_info, bio, mirror_num, 0,
888 bio_offset, private_data,
61891923
SB		 889 __btree_submit_bio_start,
890 __btree_submit_bio_done);
44b8bd7e 891 }
d313d7a3 892
4246a0b6
CH		 893 if (ret)
894 goto out_w_error;
895 return 0;
896
61891923 897out_w_error:
4e4cbee9 898 bio->bi_status = ret;
4246a0b6 899 bio_endio(bio);
61891923 900 return ret;
44b8bd7e
CM		 901}
902
3dd1462e 903#ifdef CONFIG_MIGRATION
784b4e29 904static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG		 905 struct page *newpage, struct page *page,
906 enum migrate_mode mode)
784b4e29
CM		 907{
908 /*
909 * we can't safely write a btree page from here,
910 * we haven't done the locking hook
911 */
912 if (PageDirty(page))
913 return -EAGAIN;
914 /*
915 * Buffers may be managed in a filesystem specific way.
916 * We must have no buffers or drop them.
917 */
918 if (page_has_private(page) &&
919 !try_to_release_page(page, GFP_KERNEL))
920 return -EAGAIN;
a6bc32b8 921 return migrate_page(mapping, newpage, page, mode);
784b4e29 922}
3dd1462e 923#endif
784b4e29 924
0da5468f
CM		 925
926static int btree_writepages(struct address_space *mapping,
927 struct writeback_control *wbc)
928{
e2d84521
MX		 929 struct btrfs_fs_info *fs_info;
930 int ret;
931
d8d5f3e1 932 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM		 933
934 if (wbc->for_kupdate)
935 return 0;
936
e2d84521 937 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 938 /* this is a bit racy, but that's ok */
e2d84521
MX		 939 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
940 BTRFS_DIRTY_METADATA_THRESH);
941 if (ret < 0)
793955bc 942 return 0;
793955bc 943 }
0b32f4bb 944 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM		 945}
946
b2950863 947static int btree_readpage(struct file *file, struct page *page)
5f39d397 948{
d1310b2e
CM		 949 struct extent_io_tree *tree;
950 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 951 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 952}
22b0ebda 953
70dec807 954static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 955{
98509cfc 956 if (PageWriteback(page) || PageDirty(page))
d397712b 957 return 0;
0c4e538b 958
f7a52a40 959 return try_release_extent_buffer(page);
d98237b3
CM		 960}
961
d47992f8
LC		 962static void btree_invalidatepage(struct page *page, unsigned int offset,
963 unsigned int length)
d98237b3 964{
d1310b2e
CM		 965 struct extent_io_tree *tree;
966 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 967 extent_invalidatepage(tree, page, offset);
968 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 969 if (PagePrivate(page)) {
efe120a0
FH		 970 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
971 "page private not zero on page %llu",
972 (unsigned long long)page_offset(page));
9ad6b7bc
CM		 973 ClearPagePrivate(page);
974 set_page_private(page, 0);
09cbfeaf 975 put_page(page);
9ad6b7bc 976 }
d98237b3
CM		 977}
978
0b32f4bb
JB		 979static int btree_set_page_dirty(struct page *page)
980{
bb146eb2 981#ifdef DEBUG
0b32f4bb
JB		 982 struct extent_buffer *eb;
983
984 BUG_ON(!PagePrivate(page));
985 eb = (struct extent_buffer *)page->private;
986 BUG_ON(!eb);
987 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
988 BUG_ON(!atomic_read(&eb->refs));
989 btrfs_assert_tree_locked(eb);
bb146eb2 990#endif
0b32f4bb
JB		 991 return __set_page_dirty_nobuffers(page);
992}
993
7f09410b 994static const struct address_space_operations btree_aops = {
d98237b3 995 .readpage = btree_readpage,
0da5468f 996 .writepages = btree_writepages,
5f39d397
CM		 997 .releasepage = btree_releasepage,
998 .invalidatepage = btree_invalidatepage,
5a92bc88 999#ifdef CONFIG_MIGRATION
784b4e29 1000 .migratepage = btree_migratepage,
5a92bc88 1001#endif
0b32f4bb 1002 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM		 1003};
1004
2ff7e61e 1005void readahead_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr)
090d1875 1006{
5f39d397 1007 struct extent_buffer *buf = NULL;
2ff7e61e 1008 struct inode *btree_inode = fs_info->btree_inode;
090d1875 1009
2ff7e61e 1010 buf = btrfs_find_create_tree_block(fs_info, bytenr);
c871b0f2 1011 if (IS_ERR(buf))
6197d86e 1012 return;
d1310b2e 1013 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
8436ea91 1014 buf, WAIT_NONE, btree_get_extent, 0);
5f39d397 1015 free_extent_buffer(buf);
090d1875
CM		 1016}
1017
2ff7e61e 1018int reada_tree_block_flagged(struct btrfs_fs_info *fs_info, u64 bytenr,
ab0fff03
AJ		 1019 int mirror_num, struct extent_buffer **eb)
1020{
1021 struct extent_buffer *buf = NULL;
2ff7e61e 1022 struct inode *btree_inode = fs_info->btree_inode;
ab0fff03
AJ		 1023 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1024 int ret;
1025
2ff7e61e 1026 buf = btrfs_find_create_tree_block(fs_info, bytenr);
c871b0f2 1027 if (IS_ERR(buf))
ab0fff03
AJ		 1028 return 0;
1029
1030 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1031
8436ea91 1032 ret = read_extent_buffer_pages(io_tree, buf, WAIT_PAGE_LOCK,
ab0fff03
AJ		 1033 btree_get_extent, mirror_num);
1034 if (ret) {
1035 free_extent_buffer(buf);
1036 return ret;
1037 }
1038
1039 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1040 free_extent_buffer(buf);
1041 return -EIO;
0b32f4bb 1042 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ		 1043 *eb = buf;
1044 } else {
1045 free_extent_buffer(buf);
1046 }
1047 return 0;
1048}
1049
2ff7e61e
JM		 1050struct extent_buffer *btrfs_find_create_tree_block(
1051 struct btrfs_fs_info *fs_info,
1052 u64 bytenr)
0999df54 1053{
0b246afa
JM		 1054 if (btrfs_is_testing(fs_info))
1055 return alloc_test_extent_buffer(fs_info, bytenr);
1056 return alloc_extent_buffer(fs_info, bytenr);
0999df54
CM		 1057}
1058
1059
e02119d5
CM		 1060int btrfs_write_tree_block(struct extent_buffer *buf)
1061{
727011e0 1062 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1063 buf->start + buf->len - 1);
e02119d5
CM		 1064}
1065
3189ff77 1066void btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
e02119d5 1067{
3189ff77
JL		 1068 filemap_fdatawait_range(buf->pages[0]->mapping,
1069 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1070}
1071
2ff7e61e 1072struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr,
ce86cd59 1073 u64 parent_transid)
0999df54
CM		 1074{
1075 struct extent_buffer *buf = NULL;
0999df54
CM		 1076 int ret;
1077
2ff7e61e 1078 buf = btrfs_find_create_tree_block(fs_info, bytenr);
c871b0f2
LB		 1079 if (IS_ERR(buf))
1080 return buf;
0999df54 1081
2ff7e61e 1082 ret = btree_read_extent_buffer_pages(fs_info, buf, parent_transid);
0f0fe8f7
FDBM		 1083 if (ret) {
1084 free_extent_buffer(buf);
64c043de 1085 return ERR_PTR(ret);
0f0fe8f7 1086 }
5f39d397 1087 return buf;
ce9adaa5 1088
eb60ceac
CM		 1089}
1090
7c302b49 1091void clean_tree_block(struct btrfs_fs_info *fs_info,
d5c13f92 1092 struct extent_buffer *buf)
ed2ff2cb 1093{
55c69072 1094 if (btrfs_header_generation(buf) ==
e2d84521 1095 fs_info->running_transaction->transid) {
b9447ef8 1096 btrfs_assert_tree_locked(buf);
b4ce94de 1097
b9473439 1098 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
104b4e51
NB		 1099 percpu_counter_add_batch(&fs_info->dirty_metadata_bytes,
1100 -buf->len,
1101 fs_info->dirty_metadata_batch);
ed7b63eb
JB		 1102 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1103 btrfs_set_lock_blocking(buf);
1104 clear_extent_buffer_dirty(buf);
1105 }
925baedd 1106 }
5f39d397
CM		 1107}
1108
8257b2dc
MX		 1109static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1110{
1111 struct btrfs_subvolume_writers *writers;
1112 int ret;
1113
1114 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1115 if (!writers)
1116 return ERR_PTR(-ENOMEM);
1117
908c7f19 1118 ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL);
8257b2dc
MX		 1119 if (ret < 0) {
1120 kfree(writers);
1121 return ERR_PTR(ret);
1122 }
1123
1124 init_waitqueue_head(&writers->wait);
1125 return writers;
1126}
1127
1128static void
1129btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1130{
1131 percpu_counter_destroy(&writers->counter);
1132 kfree(writers);
1133}
1134
da17066c 1135static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info,
143bede5 1136 u64 objectid)
d97e63b6 1137{
7c0260ee 1138 bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state);
cfaa7295 1139 root->node = NULL;
a28ec197 1140 root->commit_root = NULL;
27cdeb70 1141 root->state = 0;
d68fc57b 1142 root->orphan_cleanup_state = 0;
0b86a832 1143
0f7d52f4
CM		 1144 root->objectid = objectid;
1145 root->last_trans = 0;
13a8a7c8 1146 root->highest_objectid = 0;
eb73c1b7 1147 root->nr_delalloc_inodes = 0;
199c2a9c 1148 root->nr_ordered_extents = 0;
58176a96 1149 root->name = NULL;
6bef4d31 1150 root->inode_tree = RB_ROOT;
16cdcec7 1151 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1152 root->block_rsv = NULL;
d68fc57b 1153 root->orphan_block_rsv = NULL;
0b86a832
CM		 1154
1155 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1156 INIT_LIST_HEAD(&root->root_list);
eb73c1b7
MX		 1157 INIT_LIST_HEAD(&root->delalloc_inodes);
1158 INIT_LIST_HEAD(&root->delalloc_root);
199c2a9c
MX		 1159 INIT_LIST_HEAD(&root->ordered_extents);
1160 INIT_LIST_HEAD(&root->ordered_root);
2ab28f32
JB		 1161 INIT_LIST_HEAD(&root->logged_list[0]);
1162 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1163 spin_lock_init(&root->orphan_lock);
5d4f98a2 1164 spin_lock_init(&root->inode_lock);
eb73c1b7 1165 spin_lock_init(&root->delalloc_lock);
199c2a9c 1166 spin_lock_init(&root->ordered_extent_lock);
f0486c68 1167 spin_lock_init(&root->accounting_lock);
2ab28f32
JB		 1168 spin_lock_init(&root->log_extents_lock[0]);
1169 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1170 mutex_init(&root->objectid_mutex);
e02119d5 1171 mutex_init(&root->log_mutex);
31f3d255 1172 mutex_init(&root->ordered_extent_mutex);
573bfb72 1173 mutex_init(&root->delalloc_mutex);
7237f183
YZ		 1174 init_waitqueue_head(&root->log_writer_wait);
1175 init_waitqueue_head(&root->log_commit_wait[0]);
1176 init_waitqueue_head(&root->log_commit_wait[1]);
8b050d35
MX		 1177 INIT_LIST_HEAD(&root->log_ctxs[0]);
1178 INIT_LIST_HEAD(&root->log_ctxs[1]);
7237f183
YZ		 1179 atomic_set(&root->log_commit[0], 0);
1180 atomic_set(&root->log_commit[1], 0);
1181 atomic_set(&root->log_writers, 0);
2ecb7923 1182 atomic_set(&root->log_batch, 0);
8a35d95f 1183 atomic_set(&root->orphan_inodes, 0);
0700cea7 1184 refcount_set(&root->refs, 1);
ea14b57f 1185 atomic_set(&root->will_be_snapshotted, 0);
ce0dcee6 1186 atomic64_set(&root->qgroup_meta_rsv, 0);
7237f183 1187 root->log_transid = 0;
d1433deb 1188 root->log_transid_committed = -1;
257c62e1 1189 root->last_log_commit = 0;
7c0260ee 1190 if (!dummy)
c6100a4b 1191 extent_io_tree_init(&root->dirty_log_pages, NULL);
017e5369 1192
3768f368
CM		 1193 memset(&root->root_key, 0, sizeof(root->root_key));
1194 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1195 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
7c0260ee 1196 if (!dummy)
06ea65a3
JB		 1197 root->defrag_trans_start = fs_info->generation;
1198 else
1199 root->defrag_trans_start = 0;
4d775673 1200 root->root_key.objectid = objectid;
0ee5dc67 1201 root->anon_dev = 0;
8ea05e3a 1202
5f3ab90a 1203 spin_lock_init(&root->root_item_lock);
3768f368
CM		 1204}
1205
74e4d827
DS		 1206static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info,
1207 gfp_t flags)
6f07e42e 1208{
74e4d827 1209 struct btrfs_root *root = kzalloc(sizeof(*root), flags);
6f07e42e
AV		 1210 if (root)
1211 root->fs_info = fs_info;
1212 return root;
1213}
1214
06ea65a3
JB		 1215#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1216/* Should only be used by the testing infrastructure */
da17066c 1217struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info)
06ea65a3
JB		 1218{
1219 struct btrfs_root *root;
1220
7c0260ee
JM		 1221 if (!fs_info)
1222 return ERR_PTR(-EINVAL);
1223
1224 root = btrfs_alloc_root(fs_info, GFP_KERNEL);
06ea65a3
JB		 1225 if (!root)
1226 return ERR_PTR(-ENOMEM);
da17066c 1227
b9ef22de 1228 /* We don't use the stripesize in selftest, set it as sectorsize */
da17066c 1229 __setup_root(root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
faa2dbf0 1230 root->alloc_bytenr = 0;
06ea65a3
JB		 1231
1232 return root;
1233}
1234#endif
1235
20897f5c
AJ		 1236struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1237 struct btrfs_fs_info *fs_info,
1238 u64 objectid)
1239{
1240 struct extent_buffer *leaf;
1241 struct btrfs_root *tree_root = fs_info->tree_root;
1242 struct btrfs_root *root;
1243 struct btrfs_key key;
1244 int ret = 0;
6463fe58 1245 uuid_le uuid;
20897f5c 1246
74e4d827 1247 root = btrfs_alloc_root(fs_info, GFP_KERNEL);
20897f5c
AJ		 1248 if (!root)
1249 return ERR_PTR(-ENOMEM);
1250
da17066c 1251 __setup_root(root, fs_info, objectid);
20897f5c
AJ		 1252 root->root_key.objectid = objectid;
1253 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1254 root->root_key.offset = 0;
1255
4d75f8a9 1256 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
20897f5c
AJ		 1257 if (IS_ERR(leaf)) {
1258 ret = PTR_ERR(leaf);
1dd05682 1259 leaf = NULL;
20897f5c
AJ		 1260 goto fail;
1261 }
1262
b159fa28 1263 memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header));
20897f5c
AJ		 1264 btrfs_set_header_bytenr(leaf, leaf->start);
1265 btrfs_set_header_generation(leaf, trans->transid);
1266 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1267 btrfs_set_header_owner(leaf, objectid);
1268 root->node = leaf;
1269
d24ee97b
DS		 1270 write_extent_buffer_fsid(leaf, fs_info->fsid);
1271 write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid);
20897f5c
AJ		 1272 btrfs_mark_buffer_dirty(leaf);
1273
1274 root->commit_root = btrfs_root_node(root);
27cdeb70 1275 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
20897f5c
AJ		 1276
1277 root->root_item.flags = 0;
1278 root->root_item.byte_limit = 0;
1279 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1280 btrfs_set_root_generation(&root->root_item, trans->transid);
1281 btrfs_set_root_level(&root->root_item, 0);
1282 btrfs_set_root_refs(&root->root_item, 1);
1283 btrfs_set_root_used(&root->root_item, leaf->len);
1284 btrfs_set_root_last_snapshot(&root->root_item, 0);
1285 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB		 1286 uuid_le_gen(&uuid);
1287 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ		 1288 root->root_item.drop_level = 0;
1289
1290 key.objectid = objectid;
1291 key.type = BTRFS_ROOT_ITEM_KEY;
1292 key.offset = 0;
1293 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1294 if (ret)
1295 goto fail;
1296
1297 btrfs_tree_unlock(leaf);
1298
1dd05682
TI		 1299 return root;
1300
20897f5c 1301fail:
1dd05682
TI		 1302 if (leaf) {
1303 btrfs_tree_unlock(leaf);
59885b39 1304 free_extent_buffer(root->commit_root);
1dd05682
TI		 1305 free_extent_buffer(leaf);
1306 }
1307 kfree(root);
20897f5c 1308
1dd05682 1309 return ERR_PTR(ret);
20897f5c
AJ		 1310}
1311
7237f183
YZ		 1312static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1313 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1314{
1315 struct btrfs_root *root;
7237f183 1316 struct extent_buffer *leaf;
e02119d5 1317
74e4d827 1318 root = btrfs_alloc_root(fs_info, GFP_NOFS);
e02119d5 1319 if (!root)
7237f183 1320 return ERR_PTR(-ENOMEM);
e02119d5 1321
da17066c 1322 __setup_root(root, fs_info, BTRFS_TREE_LOG_OBJECTID);
e02119d5
CM		 1323
1324 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1325 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1326 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
27cdeb70 1327
7237f183 1328 /*
27cdeb70
MX		 1329 * DON'T set REF_COWS for log trees
1330 *
7237f183
YZ		 1331 * log trees do not get reference counted because they go away
1332 * before a real commit is actually done. They do store pointers
1333 * to file data extents, and those reference counts still get
1334 * updated (along with back refs to the log tree).
1335 */
e02119d5 1336
4d75f8a9
DS		 1337 leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID,
1338 NULL, 0, 0, 0);
7237f183
YZ		 1339 if (IS_ERR(leaf)) {
1340 kfree(root);
1341 return ERR_CAST(leaf);
1342 }
e02119d5 1343
b159fa28 1344 memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header));
5d4f98a2
YZ		 1345 btrfs_set_header_bytenr(leaf, leaf->start);
1346 btrfs_set_header_generation(leaf, trans->transid);
1347 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1348 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1349 root->node = leaf;
e02119d5 1350
0b246afa 1351 write_extent_buffer_fsid(root->node, fs_info->fsid);
e02119d5
CM		 1352 btrfs_mark_buffer_dirty(root->node);
1353 btrfs_tree_unlock(root->node);
7237f183
YZ		 1354 return root;
1355}
1356
1357int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1358 struct btrfs_fs_info *fs_info)
1359{
1360 struct btrfs_root *log_root;
1361
1362 log_root = alloc_log_tree(trans, fs_info);
1363 if (IS_ERR(log_root))
1364 return PTR_ERR(log_root);
1365 WARN_ON(fs_info->log_root_tree);
1366 fs_info->log_root_tree = log_root;
1367 return 0;
1368}
1369
1370int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1371 struct btrfs_root *root)
1372{
0b246afa 1373 struct btrfs_fs_info *fs_info = root->fs_info;
7237f183
YZ		 1374 struct btrfs_root *log_root;
1375 struct btrfs_inode_item *inode_item;
1376
0b246afa 1377 log_root = alloc_log_tree(trans, fs_info);
7237f183
YZ		 1378 if (IS_ERR(log_root))
1379 return PTR_ERR(log_root);
1380
1381 log_root->last_trans = trans->transid;
1382 log_root->root_key.offset = root->root_key.objectid;
1383
1384 inode_item = &log_root->root_item.inode;
3cae210f
QW		 1385 btrfs_set_stack_inode_generation(inode_item, 1);
1386 btrfs_set_stack_inode_size(inode_item, 3);
1387 btrfs_set_stack_inode_nlink(inode_item, 1);
da17066c 1388 btrfs_set_stack_inode_nbytes(inode_item,
0b246afa 1389 fs_info->nodesize);
3cae210f 1390 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
7237f183 1391
5d4f98a2 1392 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1393
1394 WARN_ON(root->log_root);
1395 root->log_root = log_root;
1396 root->log_transid = 0;
d1433deb 1397 root->log_transid_committed = -1;
257c62e1 1398 root->last_log_commit = 0;
e02119d5
CM		 1399 return 0;
1400}
1401
35a3621b
SB		 1402static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1403 struct btrfs_key *key)
e02119d5
CM		 1404{
1405 struct btrfs_root *root;
1406 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1407 struct btrfs_path *path;
84234f3a 1408 u64 generation;
cb517eab 1409 int ret;
0f7d52f4 1410
cb517eab
MX		 1411 path = btrfs_alloc_path();
1412 if (!path)
0f7d52f4 1413 return ERR_PTR(-ENOMEM);
cb517eab 1414
74e4d827 1415 root = btrfs_alloc_root(fs_info, GFP_NOFS);
cb517eab
MX		 1416 if (!root) {
1417 ret = -ENOMEM;
1418 goto alloc_fail;
0f7d52f4
CM		 1419 }
1420
da17066c 1421 __setup_root(root, fs_info, key->objectid);
0f7d52f4 1422
cb517eab
MX		 1423 ret = btrfs_find_root(tree_root, key, path,
1424 &root->root_item, &root->root_key);
0f7d52f4 1425 if (ret) {
13a8a7c8
YZ		 1426 if (ret > 0)
1427 ret = -ENOENT;
cb517eab 1428 goto find_fail;
0f7d52f4 1429 }
13a8a7c8 1430
84234f3a 1431 generation = btrfs_root_generation(&root->root_item);
2ff7e61e
JM		 1432 root->node = read_tree_block(fs_info,
1433 btrfs_root_bytenr(&root->root_item),
ce86cd59 1434 generation);
64c043de
LB		 1435 if (IS_ERR(root->node)) {
1436 ret = PTR_ERR(root->node);
cb517eab
MX		 1437 goto find_fail;
1438 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1439 ret = -EIO;
64c043de
LB		 1440 free_extent_buffer(root->node);
1441 goto find_fail;
416bc658 1442 }
5d4f98a2 1443 root->commit_root = btrfs_root_node(root);
13a8a7c8 1444out:
cb517eab
MX		 1445 btrfs_free_path(path);
1446 return root;
1447
cb517eab
MX		 1448find_fail:
1449 kfree(root);
1450alloc_fail:
1451 root = ERR_PTR(ret);
1452 goto out;
1453}
1454
1455struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1456 struct btrfs_key *location)
1457{
1458 struct btrfs_root *root;
1459
1460 root = btrfs_read_tree_root(tree_root, location);
1461 if (IS_ERR(root))
1462 return root;
1463
1464 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
27cdeb70 1465 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
08fe4db1
LZ		 1466 btrfs_check_and_init_root_item(&root->root_item);
1467 }
13a8a7c8 1468
5eda7b5e
CM		 1469 return root;
1470}
1471
cb517eab
MX		 1472int btrfs_init_fs_root(struct btrfs_root *root)
1473{
1474 int ret;
8257b2dc 1475 struct btrfs_subvolume_writers *writers;
cb517eab
MX		 1476
1477 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1478 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1479 GFP_NOFS);
1480 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1481 ret = -ENOMEM;
1482 goto fail;
1483 }
1484
8257b2dc
MX		 1485 writers = btrfs_alloc_subvolume_writers();
1486 if (IS_ERR(writers)) {
1487 ret = PTR_ERR(writers);
1488 goto fail;
1489 }
1490 root->subv_writers = writers;
1491
cb517eab 1492 btrfs_init_free_ino_ctl(root);
57cdc8db
DS		 1493 spin_lock_init(&root->ino_cache_lock);
1494 init_waitqueue_head(&root->ino_cache_wait);
cb517eab
MX		 1495
1496 ret = get_anon_bdev(&root->anon_dev);
1497 if (ret)
876d2cf1 1498 goto fail;
f32e48e9
CR		 1499
1500 mutex_lock(&root->objectid_mutex);
1501 ret = btrfs_find_highest_objectid(root,
1502 &root->highest_objectid);
1503 if (ret) {
1504 mutex_unlock(&root->objectid_mutex);
876d2cf1 1505 goto fail;
f32e48e9
CR		 1506 }
1507
1508 ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID);
1509
1510 mutex_unlock(&root->objectid_mutex);
1511
cb517eab
MX		 1512 return 0;
1513fail:
876d2cf1 1514 /* the caller is responsible to call free_fs_root */
cb517eab
MX		 1515 return ret;
1516}
1517
35bbb97f
JM		 1518struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1519 u64 root_id)
cb517eab
MX		 1520{
1521 struct btrfs_root *root;
1522
1523 spin_lock(&fs_info->fs_roots_radix_lock);
1524 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1525 (unsigned long)root_id);
1526 spin_unlock(&fs_info->fs_roots_radix_lock);
1527 return root;
1528}
1529
1530int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1531 struct btrfs_root *root)
1532{
1533 int ret;
1534
e1860a77 1535 ret = radix_tree_preload(GFP_NOFS);
cb517eab
MX		 1536 if (ret)
1537 return ret;
1538
1539 spin_lock(&fs_info->fs_roots_radix_lock);
1540 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1541 (unsigned long)root->root_key.objectid,
1542 root);
1543 if (ret == 0)
27cdeb70 1544 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
cb517eab
MX		 1545 spin_unlock(&fs_info->fs_roots_radix_lock);
1546 radix_tree_preload_end();
1547
1548 return ret;
1549}
1550
c00869f1
MX		 1551struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1552 struct btrfs_key *location,
1553 bool check_ref)
5eda7b5e
CM		 1554{
1555 struct btrfs_root *root;
381cf658 1556 struct btrfs_path *path;
1d4c08e0 1557 struct btrfs_key key;
5eda7b5e
CM		 1558 int ret;
1559
edbd8d4e
CM		 1560 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1561 return fs_info->tree_root;
1562 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1563 return fs_info->extent_root;
8f18cf13
CM		 1564 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1565 return fs_info->chunk_root;
1566 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1567 return fs_info->dev_root;
0403e47e
YZ		 1568 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1569 return fs_info->csum_root;
bcef60f2
AJ		 1570 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1571 return fs_info->quota_root ? fs_info->quota_root :
1572 ERR_PTR(-ENOENT);
f7a81ea4
SB		 1573 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1574 return fs_info->uuid_root ? fs_info->uuid_root :
1575 ERR_PTR(-ENOENT);
70f6d82e
OS		 1576 if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID)
1577 return fs_info->free_space_root ? fs_info->free_space_root :
1578 ERR_PTR(-ENOENT);
4df27c4d 1579again:
cb517eab 1580 root = btrfs_lookup_fs_root(fs_info, location->objectid);
48475471 1581 if (root) {
c00869f1 1582 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
48475471 1583 return ERR_PTR(-ENOENT);
5eda7b5e 1584 return root;
48475471 1585 }
5eda7b5e 1586
cb517eab 1587 root = btrfs_read_fs_root(fs_info->tree_root, location);
5eda7b5e
CM		 1588 if (IS_ERR(root))
1589 return root;
3394e160 1590
c00869f1 1591 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
cb517eab 1592 ret = -ENOENT;
581bb050 1593 goto fail;
35a30d7c 1594 }
581bb050 1595
cb517eab 1596 ret = btrfs_init_fs_root(root);
ac08aedf
CM		 1597 if (ret)
1598 goto fail;
3394e160 1599
381cf658
DS		 1600 path = btrfs_alloc_path();
1601 if (!path) {
1602 ret = -ENOMEM;
1603 goto fail;
1604 }
1d4c08e0
DS		 1605 key.objectid = BTRFS_ORPHAN_OBJECTID;
1606 key.type = BTRFS_ORPHAN_ITEM_KEY;
1607 key.offset = location->objectid;
1608
1609 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
381cf658 1610 btrfs_free_path(path);
d68fc57b
YZ		 1611 if (ret < 0)
1612 goto fail;
1613 if (ret == 0)
27cdeb70 1614 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
d68fc57b 1615
cb517eab 1616 ret = btrfs_insert_fs_root(fs_info, root);
0f7d52f4 1617 if (ret) {
4df27c4d
YZ		 1618 if (ret == -EEXIST) {
1619 free_fs_root(root);
1620 goto again;
1621 }
1622 goto fail;
0f7d52f4 1623 }
edbd8d4e 1624 return root;
4df27c4d
YZ		 1625fail:
1626 free_fs_root(root);
1627 return ERR_PTR(ret);
edbd8d4e
CM		 1628}
1629
04160088
CM		 1630static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1631{
1632 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1633 int ret = 0;
04160088
CM		 1634 struct btrfs_device *device;
1635 struct backing_dev_info *bdi;
b7967db7 1636
1f78160c
XG		 1637 rcu_read_lock();
1638 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1639 if (!device->bdev)
1640 continue;
efa7c9f9 1641 bdi = device->bdev->bd_bdi;
ff9ea323 1642 if (bdi_congested(bdi, bdi_bits)) {
04160088
CM		 1643 ret = 1;
1644 break;
1645 }
1646 }
1f78160c 1647 rcu_read_unlock();
04160088
CM		 1648 return ret;
1649}
1650
8b712842
CM		 1651/*
1652 * called by the kthread helper functions to finally call the bio end_io
1653 * functions. This is where read checksum verification actually happens
1654 */
1655static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1656{
ce9adaa5 1657 struct bio *bio;
97eb6b69 1658 struct btrfs_end_io_wq *end_io_wq;
ce9adaa5 1659
97eb6b69 1660 end_io_wq = container_of(work, struct btrfs_end_io_wq, work);
8b712842 1661 bio = end_io_wq->bio;
ce9adaa5 1662
4e4cbee9 1663 bio->bi_status = end_io_wq->status;
8b712842
CM		 1664 bio->bi_private = end_io_wq->private;
1665 bio->bi_end_io = end_io_wq->end_io;
97eb6b69 1666 kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq);
4246a0b6 1667 bio_endio(bio);
44b8bd7e
CM		 1668}
1669
a74a4b97
CM		 1670static int cleaner_kthread(void *arg)
1671{
1672 struct btrfs_root *root = arg;
0b246afa 1673 struct btrfs_fs_info *fs_info = root->fs_info;
d0278245 1674 int again;
da288d28 1675 struct btrfs_trans_handle *trans;
a74a4b97
CM		 1676
1677 do {
d0278245 1678 again = 0;
a74a4b97 1679
d0278245 1680 /* Make the cleaner go to sleep early. */
2ff7e61e 1681 if (btrfs_need_cleaner_sleep(fs_info))
d0278245
MX		 1682 goto sleep;
1683
90c711ab
ZB		 1684 /*
1685 * Do not do anything if we might cause open_ctree() to block
1686 * before we have finished mounting the filesystem.
1687 */
0b246afa 1688 if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags))
90c711ab
ZB		 1689 goto sleep;
1690
0b246afa 1691 if (!mutex_trylock(&fs_info->cleaner_mutex))
d0278245
MX		 1692 goto sleep;
1693
dc7f370c
MX		 1694 /*
1695 * Avoid the problem that we change the status of the fs
1696 * during the above check and trylock.
1697 */
2ff7e61e 1698 if (btrfs_need_cleaner_sleep(fs_info)) {
0b246afa 1699 mutex_unlock(&fs_info->cleaner_mutex);
dc7f370c 1700 goto sleep;
76dda93c 1701 }
a74a4b97 1702
0b246afa 1703 mutex_lock(&fs_info->cleaner_delayed_iput_mutex);
2ff7e61e 1704 btrfs_run_delayed_iputs(fs_info);
0b246afa 1705 mutex_unlock(&fs_info->cleaner_delayed_iput_mutex);
c2d6cb16 1706
d0278245 1707 again = btrfs_clean_one_deleted_snapshot(root);
0b246afa 1708 mutex_unlock(&fs_info->cleaner_mutex);
d0278245
MX		 1709
1710 /*
05323cd1
MX		 1711 * The defragger has dealt with the R/O remount and umount,
1712 * needn't do anything special here.
d0278245 1713 */
0b246afa 1714 btrfs_run_defrag_inodes(fs_info);
67c5e7d4
FM		 1715
1716 /*
1717 * Acquires fs_info->delete_unused_bgs_mutex to avoid racing
1718 * with relocation (btrfs_relocate_chunk) and relocation
1719 * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group)
1720 * after acquiring fs_info->delete_unused_bgs_mutex. So we
1721 * can't hold, nor need to, fs_info->cleaner_mutex when deleting
1722 * unused block groups.
1723 */
0b246afa 1724 btrfs_delete_unused_bgs(fs_info);
d0278245 1725sleep:
838fe188 1726 if (!again) {
a74a4b97 1727 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1728 if (!kthread_should_stop())
1729 schedule();
a74a4b97
CM		 1730 __set_current_state(TASK_RUNNING);
1731 }
1732 } while (!kthread_should_stop());
da288d28
FM		 1733
1734 /*
1735 * Transaction kthread is stopped before us and wakes us up.
1736 * However we might have started a new transaction and COWed some
1737 * tree blocks when deleting unused block groups for example. So
1738 * make sure we commit the transaction we started to have a clean
1739 * shutdown when evicting the btree inode - if it has dirty pages
1740 * when we do the final iput() on it, eviction will trigger a
1741 * writeback for it which will fail with null pointer dereferences
1742 * since work queues and other resources were already released and
1743 * destroyed by the time the iput/eviction/writeback is made.
1744 */
1745 trans = btrfs_attach_transaction(root);
1746 if (IS_ERR(trans)) {
1747 if (PTR_ERR(trans) != -ENOENT)
0b246afa 1748 btrfs_err(fs_info,
da288d28
FM		 1749 "cleaner transaction attach returned %ld",
1750 PTR_ERR(trans));
1751 } else {
1752 int ret;
1753
3a45bb20 1754 ret = btrfs_commit_transaction(trans);
da288d28 1755 if (ret)
0b246afa 1756 btrfs_err(fs_info,
da288d28
FM		 1757 "cleaner open transaction commit returned %d",
1758 ret);
1759 }
1760
a74a4b97
CM		 1761 return 0;
1762}
1763
1764static int transaction_kthread(void *arg)
1765{
1766 struct btrfs_root *root = arg;
0b246afa 1767 struct btrfs_fs_info *fs_info = root->fs_info;
a74a4b97
CM		 1768 struct btrfs_trans_handle *trans;
1769 struct btrfs_transaction *cur;
8929ecfa 1770 u64 transid;
a74a4b97
CM		 1771 unsigned long now;
1772 unsigned long delay;
914b2007 1773 bool cannot_commit;
a74a4b97
CM		 1774
1775 do {
914b2007 1776 cannot_commit = false;
0b246afa
JM		 1777 delay = HZ * fs_info->commit_interval;
1778 mutex_lock(&fs_info->transaction_kthread_mutex);
a74a4b97 1779
0b246afa
JM		 1780 spin_lock(&fs_info->trans_lock);
1781 cur = fs_info->running_transaction;
a74a4b97 1782 if (!cur) {
0b246afa 1783 spin_unlock(&fs_info->trans_lock);
a74a4b97
CM		 1784 goto sleep;
1785 }
31153d81 1786
a74a4b97 1787 now = get_seconds();
4a9d8bde 1788 if (cur->state < TRANS_STATE_BLOCKED &&
8b87dc17 1789 (now < cur->start_time ||
0b246afa
JM		 1790 now - cur->start_time < fs_info->commit_interval)) {
1791 spin_unlock(&fs_info->trans_lock);
a74a4b97
CM		 1792 delay = HZ * 5;
1793 goto sleep;
1794 }
8929ecfa 1795 transid = cur->transid;
0b246afa 1796 spin_unlock(&fs_info->trans_lock);
56bec294 1797
79787eaa 1798 /* If the file system is aborted, this will always fail. */
354aa0fb 1799 trans = btrfs_attach_transaction(root);
914b2007 1800 if (IS_ERR(trans)) {
354aa0fb
MX		 1801 if (PTR_ERR(trans) != -ENOENT)
1802 cannot_commit = true;
79787eaa 1803 goto sleep;
914b2007 1804 }
8929ecfa 1805 if (transid == trans->transid) {
3a45bb20 1806 btrfs_commit_transaction(trans);
8929ecfa 1807 } else {
3a45bb20 1808 btrfs_end_transaction(trans);
8929ecfa 1809 }
a74a4b97 1810sleep:
0b246afa
JM		 1811 wake_up_process(fs_info->cleaner_kthread);
1812 mutex_unlock(&fs_info->transaction_kthread_mutex);
a74a4b97 1813
4e121c06 1814 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
0b246afa 1815 &fs_info->fs_state)))
2ff7e61e 1816 btrfs_cleanup_transaction(fs_info);
ce63f891
JK		 1817 set_current_state(TASK_INTERRUPTIBLE);
1818 if (!kthread_should_stop() &&
0b246afa 1819 (!btrfs_transaction_blocked(fs_info) ||
ce63f891
JK		 1820 cannot_commit))
1821 schedule_timeout(delay);
1822 __set_current_state(TASK_RUNNING);
a74a4b97
CM		 1823 } while (!kthread_should_stop());
1824 return 0;
1825}
1826
af31f5e5
CM		 1827/*
1828 * this will find the highest generation in the array of
1829 * root backups. The index of the highest array is returned,
1830 * or -1 if we can't find anything.
1831 *
1832 * We check to make sure the array is valid by comparing the
1833 * generation of the latest root in the array with the generation
1834 * in the super block. If they don't match we pitch it.
1835 */
1836static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1837{
1838 u64 cur;
1839 int newest_index = -1;
1840 struct btrfs_root_backup *root_backup;
1841 int i;
1842
1843 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1844 root_backup = info->super_copy->super_roots + i;
1845 cur = btrfs_backup_tree_root_gen(root_backup);
1846 if (cur == newest_gen)
1847 newest_index = i;
1848 }
1849
1850 /* check to see if we actually wrapped around */
1851 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1852 root_backup = info->super_copy->super_roots;
1853 cur = btrfs_backup_tree_root_gen(root_backup);
1854 if (cur == newest_gen)
1855 newest_index = 0;
1856 }
1857 return newest_index;
1858}
1859
1860
1861/*
1862 * find the oldest backup so we know where to store new entries
1863 * in the backup array. This will set the backup_root_index
1864 * field in the fs_info struct
1865 */
1866static void find_oldest_super_backup(struct btrfs_fs_info *info,
1867 u64 newest_gen)
1868{
1869 int newest_index = -1;
1870
1871 newest_index = find_newest_super_backup(info, newest_gen);
1872 /* if there was garbage in there, just move along */
1873 if (newest_index == -1) {
1874 info->backup_root_index = 0;
1875 } else {
1876 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1877 }
1878}
1879
1880/*
1881 * copy all the root pointers into the super backup array.
1882 * this will bump the backup pointer by one when it is
1883 * done
1884 */
1885static void backup_super_roots(struct btrfs_fs_info *info)
1886{
1887 int next_backup;
1888 struct btrfs_root_backup *root_backup;
1889 int last_backup;
1890
1891 next_backup = info->backup_root_index;
1892 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1893 BTRFS_NUM_BACKUP_ROOTS;
1894
1895 /*
1896 * just overwrite the last backup if we're at the same generation
1897 * this happens only at umount
1898 */
1899 root_backup = info->super_for_commit->super_roots + last_backup;
1900 if (btrfs_backup_tree_root_gen(root_backup) ==
1901 btrfs_header_generation(info->tree_root->node))
1902 next_backup = last_backup;
1903
1904 root_backup = info->super_for_commit->super_roots + next_backup;
1905
1906 /*
1907 * make sure all of our padding and empty slots get zero filled
1908 * regardless of which ones we use today
1909 */
1910 memset(root_backup, 0, sizeof(*root_backup));
1911
1912 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1913
1914 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1915 btrfs_set_backup_tree_root_gen(root_backup,
1916 btrfs_header_generation(info->tree_root->node));
1917
1918 btrfs_set_backup_tree_root_level(root_backup,
1919 btrfs_header_level(info->tree_root->node));
1920
1921 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1922 btrfs_set_backup_chunk_root_gen(root_backup,
1923 btrfs_header_generation(info->chunk_root->node));
1924 btrfs_set_backup_chunk_root_level(root_backup,
1925 btrfs_header_level(info->chunk_root->node));
1926
1927 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1928 btrfs_set_backup_extent_root_gen(root_backup,
1929 btrfs_header_generation(info->extent_root->node));
1930 btrfs_set_backup_extent_root_level(root_backup,
1931 btrfs_header_level(info->extent_root->node));
1932
7c7e82a7
CM		 1933 /*
1934 * we might commit during log recovery, which happens before we set
1935 * the fs_root. Make sure it is valid before we fill it in.
1936 */
1937 if (info->fs_root && info->fs_root->node) {
1938 btrfs_set_backup_fs_root(root_backup,
1939 info->fs_root->node->start);
1940 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 1941 btrfs_header_generation(info->fs_root->node));
7c7e82a7 1942 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 1943 btrfs_header_level(info->fs_root->node));
7c7e82a7 1944 }
af31f5e5
CM		 1945
1946 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1947 btrfs_set_backup_dev_root_gen(root_backup,
1948 btrfs_header_generation(info->dev_root->node));
1949 btrfs_set_backup_dev_root_level(root_backup,
1950 btrfs_header_level(info->dev_root->node));
1951
1952 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1953 btrfs_set_backup_csum_root_gen(root_backup,
1954 btrfs_header_generation(info->csum_root->node));
1955 btrfs_set_backup_csum_root_level(root_backup,
1956 btrfs_header_level(info->csum_root->node));
1957
1958 btrfs_set_backup_total_bytes(root_backup,
1959 btrfs_super_total_bytes(info->super_copy));
1960 btrfs_set_backup_bytes_used(root_backup,
1961 btrfs_super_bytes_used(info->super_copy));
1962 btrfs_set_backup_num_devices(root_backup,
1963 btrfs_super_num_devices(info->super_copy));
1964
1965 /*
1966 * if we don't copy this out to the super_copy, it won't get remembered
1967 * for the next commit
1968 */
1969 memcpy(&info->super_copy->super_roots,
1970 &info->super_for_commit->super_roots,
1971 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1972}
1973
1974/*
1975 * this copies info out of the root backup array and back into
1976 * the in-memory super block. It is meant to help iterate through
1977 * the array, so you send it the number of backups you've already
1978 * tried and the last backup index you used.
1979 *
1980 * this returns -1 when it has tried all the backups
1981 */
1982static noinline int next_root_backup(struct btrfs_fs_info *info,
1983 struct btrfs_super_block *super,
1984 int *num_backups_tried, int *backup_index)
1985{
1986 struct btrfs_root_backup *root_backup;
1987 int newest = *backup_index;
1988
1989 if (*num_backups_tried == 0) {
1990 u64 gen = btrfs_super_generation(super);
1991
1992 newest = find_newest_super_backup(info, gen);
1993 if (newest == -1)
1994 return -1;
1995
1996 *backup_index = newest;
1997 *num_backups_tried = 1;
1998 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
1999 /* we've tried all the backups, all done */
2000 return -1;
2001 } else {
2002 /* jump to the next oldest backup */
2003 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2004 BTRFS_NUM_BACKUP_ROOTS;
2005 *backup_index = newest;
2006 *num_backups_tried += 1;
2007 }
2008 root_backup = super->super_roots + newest;
2009
2010 btrfs_set_super_generation(super,
2011 btrfs_backup_tree_root_gen(root_backup));
2012 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2013 btrfs_set_super_root_level(super,
2014 btrfs_backup_tree_root_level(root_backup));
2015 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2016
2017 /*
2018 * fixme: the total bytes and num_devices need to match or we should
2019 * need a fsck
2020 */
2021 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2022 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2023 return 0;
2024}
2025
7abadb64
LB		 2026/* helper to cleanup workers */
2027static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2028{
dc6e3209 2029 btrfs_destroy_workqueue(fs_info->fixup_workers);
afe3d242 2030 btrfs_destroy_workqueue(fs_info->delalloc_workers);
5cdc7ad3 2031 btrfs_destroy_workqueue(fs_info->workers);
fccb5d86 2032 btrfs_destroy_workqueue(fs_info->endio_workers);
fccb5d86 2033 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
8b110e39 2034 btrfs_destroy_workqueue(fs_info->endio_repair_workers);
d05a33ac 2035 btrfs_destroy_workqueue(fs_info->rmw_workers);
fccb5d86
QW		 2036 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2037 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
a8c93d4e 2038 btrfs_destroy_workqueue(fs_info->submit_workers);
5b3bc44e 2039 btrfs_destroy_workqueue(fs_info->delayed_workers);
e66f0bb1 2040 btrfs_destroy_workqueue(fs_info->caching_workers);
736cfa15 2041 btrfs_destroy_workqueue(fs_info->readahead_workers);
a44903ab 2042 btrfs_destroy_workqueue(fs_info->flush_workers);
fc97fab0 2043 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
a79b7d4b 2044 btrfs_destroy_workqueue(fs_info->extent_workers);
a9b9477d
FM		 2045 /*
2046 * Now that all other work queues are destroyed, we can safely destroy
2047 * the queues used for metadata I/O, since tasks from those other work
2048 * queues can do metadata I/O operations.
2049 */
2050 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2051 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
7abadb64
LB		 2052}
2053
2e9f5954
R		 2054static void free_root_extent_buffers(struct btrfs_root *root)
2055{
2056 if (root) {
2057 free_extent_buffer(root->node);
2058 free_extent_buffer(root->commit_root);
2059 root->node = NULL;
2060 root->commit_root = NULL;
2061 }
2062}
2063
af31f5e5
CM		 2064/* helper to cleanup tree roots */
2065static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2066{
2e9f5954 2067 free_root_extent_buffers(info->tree_root);
655b09fe 2068
2e9f5954
R		 2069 free_root_extent_buffers(info->dev_root);
2070 free_root_extent_buffers(info->extent_root);
2071 free_root_extent_buffers(info->csum_root);
2072 free_root_extent_buffers(info->quota_root);
2073 free_root_extent_buffers(info->uuid_root);
2074 if (chunk_root)
2075 free_root_extent_buffers(info->chunk_root);
70f6d82e 2076 free_root_extent_buffers(info->free_space_root);
af31f5e5
CM		 2077}
2078
faa2dbf0 2079void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
171f6537
JB		 2080{
2081 int ret;
2082 struct btrfs_root *gang[8];
2083 int i;
2084
2085 while (!list_empty(&fs_info->dead_roots)) {
2086 gang[0] = list_entry(fs_info->dead_roots.next,
2087 struct btrfs_root, root_list);
2088 list_del(&gang[0]->root_list);
2089
27cdeb70 2090 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
cb517eab 2091 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
171f6537
JB		 2092 } else {
2093 free_extent_buffer(gang[0]->node);
2094 free_extent_buffer(gang[0]->commit_root);
b0feb9d9 2095 btrfs_put_fs_root(gang[0]);
171f6537
JB		 2096 }
2097 }
2098
2099 while (1) {
2100 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2101 (void **)gang, 0,
2102 ARRAY_SIZE(gang));
2103 if (!ret)
2104 break;
2105 for (i = 0; i < ret; i++)
cb517eab 2106 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
171f6537 2107 }
1a4319cc
LB		 2108
2109 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2110 btrfs_free_log_root_tree(NULL, fs_info);
2ff7e61e 2111 btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents);
1a4319cc 2112 }
171f6537 2113}
af31f5e5 2114
638aa7ed
ES		 2115static void btrfs_init_scrub(struct btrfs_fs_info *fs_info)
2116{
2117 mutex_init(&fs_info->scrub_lock);
2118 atomic_set(&fs_info->scrubs_running, 0);
2119 atomic_set(&fs_info->scrub_pause_req, 0);
2120 atomic_set(&fs_info->scrubs_paused, 0);
2121 atomic_set(&fs_info->scrub_cancel_req, 0);
2122 init_waitqueue_head(&fs_info->scrub_pause_wait);
2123 fs_info->scrub_workers_refcnt = 0;
2124}
2125
779a65a4
ES		 2126static void btrfs_init_balance(struct btrfs_fs_info *fs_info)
2127{
2128 spin_lock_init(&fs_info->balance_lock);
2129 mutex_init(&fs_info->balance_mutex);
2130 atomic_set(&fs_info->balance_running, 0);
2131 atomic_set(&fs_info->balance_pause_req, 0);
2132 atomic_set(&fs_info->balance_cancel_req, 0);
2133 fs_info->balance_ctl = NULL;
2134 init_waitqueue_head(&fs_info->balance_wait_q);
2135}
2136
6bccf3ab 2137static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info)
f37938e0 2138{
2ff7e61e
JM		 2139 struct inode *inode = fs_info->btree_inode;
2140
2141 inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
2142 set_nlink(inode, 1);
f37938e0
ES		 2143 /*
2144 * we set the i_size on the btree inode to the max possible int.
2145 * the real end of the address space is determined by all of
2146 * the devices in the system
2147 */
2ff7e61e
JM		 2148 inode->i_size = OFFSET_MAX;
2149 inode->i_mapping->a_ops = &btree_aops;
f37938e0 2150
2ff7e61e 2151 RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node);
c6100a4b 2152 extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode);
2ff7e61e
JM		 2153 BTRFS_I(inode)->io_tree.track_uptodate = 0;
2154 extent_map_tree_init(&BTRFS_I(inode)->extent_tree);
f37938e0 2155
2ff7e61e 2156 BTRFS_I(inode)->io_tree.ops = &btree_extent_io_ops;
f37938e0 2157
2ff7e61e
JM		 2158 BTRFS_I(inode)->root = fs_info->tree_root;
2159 memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key));
2160 set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags);
2161 btrfs_insert_inode_hash(inode);
f37938e0
ES		 2162}
2163
ad618368
ES		 2164static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info)
2165{
2166 fs_info->dev_replace.lock_owner = 0;
2167 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2168 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
73beece9
LB		 2169 rwlock_init(&fs_info->dev_replace.lock);
2170 atomic_set(&fs_info->dev_replace.read_locks, 0);
2171 atomic_set(&fs_info->dev_replace.blocking_readers, 0);
ad618368 2172 init_waitqueue_head(&fs_info->replace_wait);
73beece9 2173 init_waitqueue_head(&fs_info->dev_replace.read_lock_wq);
ad618368
ES		 2174}
2175
f9e92e40
ES		 2176static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info)
2177{
2178 spin_lock_init(&fs_info->qgroup_lock);
2179 mutex_init(&fs_info->qgroup_ioctl_lock);
2180 fs_info->qgroup_tree = RB_ROOT;
2181 fs_info->qgroup_op_tree = RB_ROOT;
2182 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2183 fs_info->qgroup_seq = 1;
f9e92e40 2184 fs_info->qgroup_ulist = NULL;
d2c609b8 2185 fs_info->qgroup_rescan_running = false;
f9e92e40
ES		 2186 mutex_init(&fs_info->qgroup_rescan_lock);
2187}
2188
2a458198
ES		 2189static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info,
2190 struct btrfs_fs_devices *fs_devices)
2191{
2192 int max_active = fs_info->thread_pool_size;
6f011058 2193 unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
2a458198
ES		 2194
2195 fs_info->workers =
cb001095
JM		 2196 btrfs_alloc_workqueue(fs_info, "worker",
2197 flags | WQ_HIGHPRI, max_active, 16);
2a458198
ES		 2198
2199 fs_info->delalloc_workers =
cb001095
JM		 2200 btrfs_alloc_workqueue(fs_info, "delalloc",
2201 flags, max_active, 2);
2a458198
ES		 2202
2203 fs_info->flush_workers =
cb001095
JM		 2204 btrfs_alloc_workqueue(fs_info, "flush_delalloc",
2205 flags, max_active, 0);
2a458198
ES		 2206
2207 fs_info->caching_workers =
cb001095 2208 btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0);
2a458198
ES		 2209
2210 /*
2211 * a higher idle thresh on the submit workers makes it much more
2212 * likely that bios will be send down in a sane order to the
2213 * devices
2214 */
2215 fs_info->submit_workers =
cb001095 2216 btrfs_alloc_workqueue(fs_info, "submit", flags,
2a458198
ES		 2217 min_t(u64, fs_devices->num_devices,
2218 max_active), 64);
2219
2220 fs_info->fixup_workers =
cb001095 2221 btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0);
2a458198
ES		 2222
2223 /*
2224 * endios are largely parallel and should have a very
2225 * low idle thresh
2226 */
2227 fs_info->endio_workers =
cb001095 2228 btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4);
2a458198 2229 fs_info->endio_meta_workers =
cb001095
JM		 2230 btrfs_alloc_workqueue(fs_info, "endio-meta", flags,
2231 max_active, 4);
2a458198 2232 fs_info->endio_meta_write_workers =
cb001095
JM		 2233 btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags,
2234 max_active, 2);
2a458198 2235 fs_info->endio_raid56_workers =
cb001095
JM		 2236 btrfs_alloc_workqueue(fs_info, "endio-raid56", flags,
2237 max_active, 4);
2a458198 2238 fs_info->endio_repair_workers =
cb001095 2239 btrfs_alloc_workqueue(fs_info, "endio-repair", flags, 1, 0);
2a458198 2240 fs_info->rmw_workers =
cb001095 2241 btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2);
2a458198 2242 fs_info->endio_write_workers =
cb001095
JM		 2243 btrfs_alloc_workqueue(fs_info, "endio-write", flags,
2244 max_active, 2);
2a458198 2245 fs_info->endio_freespace_worker =
cb001095
JM		 2246 btrfs_alloc_workqueue(fs_info, "freespace-write", flags,
2247 max_active, 0);
2a458198 2248 fs_info->delayed_workers =
cb001095
JM		 2249 btrfs_alloc_workqueue(fs_info, "delayed-meta", flags,
2250 max_active, 0);
2a458198 2251 fs_info->readahead_workers =
cb001095
JM		 2252 btrfs_alloc_workqueue(fs_info, "readahead", flags,
2253 max_active, 2);
2a458198 2254 fs_info->qgroup_rescan_workers =
cb001095 2255 btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0);
2a458198 2256 fs_info->extent_workers =
cb001095 2257 btrfs_alloc_workqueue(fs_info, "extent-refs", flags,
2a458198
ES		 2258 min_t(u64, fs_devices->num_devices,
2259 max_active), 8);
2260
2261 if (!(fs_info->workers && fs_info->delalloc_workers &&
2262 fs_info->submit_workers && fs_info->flush_workers &&
2263 fs_info->endio_workers && fs_info->endio_meta_workers &&
2264 fs_info->endio_meta_write_workers &&
2265 fs_info->endio_repair_workers &&
2266 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
2267 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
2268 fs_info->caching_workers && fs_info->readahead_workers &&
2269 fs_info->fixup_workers && fs_info->delayed_workers &&
2270 fs_info->extent_workers &&
2271 fs_info->qgroup_rescan_workers)) {
2272 return -ENOMEM;
2273 }
2274
2275 return 0;
2276}
2277
63443bf5
ES		 2278static int btrfs_replay_log(struct btrfs_fs_info *fs_info,
2279 struct btrfs_fs_devices *fs_devices)
2280{
2281 int ret;
63443bf5
ES		 2282 struct btrfs_root *log_tree_root;
2283 struct btrfs_super_block *disk_super = fs_info->super_copy;
2284 u64 bytenr = btrfs_super_log_root(disk_super);
2285
2286 if (fs_devices->rw_devices == 0) {
f14d104d 2287 btrfs_warn(fs_info, "log replay required on RO media");
63443bf5
ES		 2288 return -EIO;
2289 }
2290
74e4d827 2291 log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
63443bf5
ES		 2292 if (!log_tree_root)
2293 return -ENOMEM;
2294
da17066c 2295 __setup_root(log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
63443bf5 2296
2ff7e61e
JM		 2297 log_tree_root->node = read_tree_block(fs_info, bytenr,
2298 fs_info->generation + 1);
64c043de 2299 if (IS_ERR(log_tree_root->node)) {
f14d104d 2300 btrfs_warn(fs_info, "failed to read log tree");
0eeff236 2301 ret = PTR_ERR(log_tree_root->node);
64c043de 2302 kfree(log_tree_root);
0eeff236 2303 return ret;
64c043de 2304 } else if (!extent_buffer_uptodate(log_tree_root->node)) {
f14d104d 2305 btrfs_err(fs_info, "failed to read log tree");
63443bf5
ES		 2306 free_extent_buffer(log_tree_root->node);
2307 kfree(log_tree_root);
2308 return -EIO;
2309 }
2310 /* returns with log_tree_root freed on success */
2311 ret = btrfs_recover_log_trees(log_tree_root);
2312 if (ret) {
0b246afa
JM		 2313 btrfs_handle_fs_error(fs_info, ret,
2314 "Failed to recover log tree");
63443bf5
ES		 2315 free_extent_buffer(log_tree_root->node);
2316 kfree(log_tree_root);
2317 return ret;
2318 }
2319
bc98a42c 2320 if (sb_rdonly(fs_info->sb)) {
6bccf3ab 2321 ret = btrfs_commit_super(fs_info);
63443bf5
ES		 2322 if (ret)
2323 return ret;
2324 }
2325
2326 return 0;
2327}
2328
6bccf3ab 2329static int btrfs_read_roots(struct btrfs_fs_info *fs_info)
4bbcaa64 2330{
6bccf3ab 2331 struct btrfs_root *tree_root = fs_info->tree_root;
a4f3d2c4 2332 struct btrfs_root *root;
4bbcaa64
ES		 2333 struct btrfs_key location;
2334 int ret;
2335
6bccf3ab
JM		 2336 BUG_ON(!fs_info->tree_root);
2337
4bbcaa64
ES		 2338 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2339 location.type = BTRFS_ROOT_ITEM_KEY;
2340 location.offset = 0;
2341
a4f3d2c4
DS		 2342 root = btrfs_read_tree_root(tree_root, &location);
2343 if (IS_ERR(root))
2344 return PTR_ERR(root);
2345 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2346 fs_info->extent_root = root;
4bbcaa64
ES		 2347
2348 location.objectid = BTRFS_DEV_TREE_OBJECTID;
a4f3d2c4
DS		 2349 root = btrfs_read_tree_root(tree_root, &location);
2350 if (IS_ERR(root))
2351 return PTR_ERR(root);
2352 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2353 fs_info->dev_root = root;
4bbcaa64
ES		 2354 btrfs_init_devices_late(fs_info);
2355
2356 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
a4f3d2c4
DS		 2357 root = btrfs_read_tree_root(tree_root, &location);
2358 if (IS_ERR(root))
2359 return PTR_ERR(root);
2360 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2361 fs_info->csum_root = root;
4bbcaa64
ES		 2362
2363 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
a4f3d2c4
DS		 2364 root = btrfs_read_tree_root(tree_root, &location);
2365 if (!IS_ERR(root)) {
2366 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
afcdd129 2367 set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags);
a4f3d2c4 2368 fs_info->quota_root = root;
4bbcaa64
ES		 2369 }
2370
2371 location.objectid = BTRFS_UUID_TREE_OBJECTID;
a4f3d2c4
DS		 2372 root = btrfs_read_tree_root(tree_root, &location);
2373 if (IS_ERR(root)) {
2374 ret = PTR_ERR(root);
4bbcaa64
ES		 2375 if (ret != -ENOENT)
2376 return ret;
2377 } else {
a4f3d2c4
DS		 2378 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2379 fs_info->uuid_root = root;
4bbcaa64
ES		 2380 }
2381
70f6d82e
OS		 2382 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
2383 location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID;
2384 root = btrfs_read_tree_root(tree_root, &location);
2385 if (IS_ERR(root))
2386 return PTR_ERR(root);
2387 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
2388 fs_info->free_space_root = root;
2389 }
2390
4bbcaa64
ES		 2391 return 0;
2392}
2393
ad2b2c80
AV		 2394int open_ctree(struct super_block *sb,
2395 struct btrfs_fs_devices *fs_devices,
2396 char *options)
2e635a27 2397{
db94535d
CM		 2398 u32 sectorsize;
2399 u32 nodesize;
87ee04eb 2400 u32 stripesize;
84234f3a 2401 u64 generation;
f2b636e8 2402 u64 features;
3de4586c 2403 struct btrfs_key location;
a061fc8d 2404 struct buffer_head *bh;
4d34b278 2405 struct btrfs_super_block *disk_super;
815745cf 2406 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2407 struct btrfs_root *tree_root;
4d34b278 2408 struct btrfs_root *chunk_root;
eb60ceac 2409 int ret;
e58ca020 2410 int err = -EINVAL;
af31f5e5
CM		 2411 int num_backups_tried = 0;
2412 int backup_index = 0;
5cdc7ad3 2413 int max_active;
6675df31 2414 int clear_free_space_tree = 0;
4543df7e 2415
74e4d827
DS		 2416 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
2417 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL);
cb517eab 2418 if (!tree_root || !chunk_root) {
39279cc3
CM		 2419 err = -ENOMEM;
2420 goto fail;
2421 }
76dda93c
YZ		 2422
2423 ret = init_srcu_struct(&fs_info->subvol_srcu);
2424 if (ret) {
2425 err = ret;
2426 goto fail;
2427 }
2428
908c7f19 2429 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL);
e2d84521
MX		 2430 if (ret) {
2431 err = ret;
9e11ceee 2432 goto fail_srcu;
e2d84521 2433 }
09cbfeaf 2434 fs_info->dirty_metadata_batch = PAGE_SIZE *
e2d84521
MX		 2435 (1 + ilog2(nr_cpu_ids));
2436
908c7f19 2437 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL);
963d678b
MX		 2438 if (ret) {
2439 err = ret;
2440 goto fail_dirty_metadata_bytes;
2441 }
2442
908c7f19 2443 ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL);
c404e0dc
MX		 2444 if (ret) {
2445 err = ret;
2446 goto fail_delalloc_bytes;
2447 }
2448
76dda93c
YZ		 2449 fs_info->btree_inode = new_inode(sb);
2450 if (!fs_info->btree_inode) {
2451 err = -ENOMEM;
c404e0dc 2452 goto fail_bio_counter;
76dda93c
YZ		 2453 }
2454
a6591715 2455 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2456
76dda93c 2457 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
f28491e0 2458 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
8fd17795 2459 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2460 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2461 INIT_LIST_HEAD(&fs_info->delayed_iputs);
eb73c1b7 2462 INIT_LIST_HEAD(&fs_info->delalloc_roots);
11833d66 2463 INIT_LIST_HEAD(&fs_info->caching_block_groups);
eb73c1b7 2464 spin_lock_init(&fs_info->delalloc_root_lock);
a4abeea4 2465 spin_lock_init(&fs_info->trans_lock);
76dda93c 2466 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2467 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2468 spin_lock_init(&fs_info->defrag_inodes_lock);
f29021b2 2469 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2470 spin_lock_init(&fs_info->super_lock);
fcebe456 2471 spin_lock_init(&fs_info->qgroup_op_lock);
f28491e0 2472 spin_lock_init(&fs_info->buffer_lock);
47ab2a6c 2473 spin_lock_init(&fs_info->unused_bgs_lock);
f29021b2 2474 rwlock_init(&fs_info->tree_mod_log_lock);
d7c15171 2475 mutex_init(&fs_info->unused_bg_unpin_mutex);
67c5e7d4 2476 mutex_init(&fs_info->delete_unused_bgs_mutex);
7585717f 2477 mutex_init(&fs_info->reloc_mutex);
573bfb72 2478 mutex_init(&fs_info->delalloc_root_mutex);
c2d6cb16 2479 mutex_init(&fs_info->cleaner_delayed_iput_mutex);
de98ced9 2480 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2481
0b86a832 2482 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2483 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2484 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
47ab2a6c 2485 INIT_LIST_HEAD(&fs_info->unused_bgs);
0b86a832 2486 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX		 2487 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2488 BTRFS_BLOCK_RSV_GLOBAL);
2489 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2490 BTRFS_BLOCK_RSV_DELALLOC);
2491 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2492 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2493 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2494 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2495 BTRFS_BLOCK_RSV_DELOPS);
771ed689 2496 atomic_set(&fs_info->async_delalloc_pages, 0);
4cb5300b 2497 atomic_set(&fs_info->defrag_running, 0);
fcebe456 2498 atomic_set(&fs_info->qgroup_op_seq, 0);
2fefd558 2499 atomic_set(&fs_info->reada_works_cnt, 0);
fc36ed7e 2500 atomic64_set(&fs_info->tree_mod_seq, 0);
e20d96d6 2501 fs_info->sb = sb;
95ac567a 2502 fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE;
9ed74f2d 2503 fs_info->metadata_ratio = 0;
4cb5300b 2504 fs_info->defrag_inodes = RB_ROOT;
a5ed45f8 2505 atomic64_set(&fs_info->free_chunk_space, 0);
f29021b2 2506 fs_info->tree_mod_log = RB_ROOT;
8b87dc17 2507 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
f8c269d7 2508 fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */
90519d66 2509 /* readahead state */
d0164adc 2510 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM);
90519d66 2511 spin_lock_init(&fs_info->reada_lock);
c8b97818 2512
b34b086c
CM		 2513 fs_info->thread_pool_size = min_t(unsigned long,
2514 num_online_cpus() + 2, 8);
0afbaf8c 2515
199c2a9c
MX		 2516 INIT_LIST_HEAD(&fs_info->ordered_roots);
2517 spin_lock_init(&fs_info->ordered_root_lock);
16cdcec7 2518 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
74e4d827 2519 GFP_KERNEL);
16cdcec7
MX		 2520 if (!fs_info->delayed_root) {
2521 err = -ENOMEM;
2522 goto fail_iput;
2523 }
2524 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2525
638aa7ed 2526 btrfs_init_scrub(fs_info);
21adbd5c
SB		 2527#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2528 fs_info->check_integrity_print_mask = 0;
2529#endif
779a65a4 2530 btrfs_init_balance(fs_info);
21c7e756 2531 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
a2de733c 2532
9f6d2510
DS		 2533 sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE;
2534 sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE);
a061fc8d 2535
6bccf3ab 2536 btrfs_init_btree_inode(fs_info);
76dda93c 2537
0f9dd46c 2538 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2539 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2540 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2541
c6100a4b
JB		 2542 extent_io_tree_init(&fs_info->freed_extents[0], NULL);
2543 extent_io_tree_init(&fs_info->freed_extents[1], NULL);
11833d66 2544 fs_info->pinned_extents = &fs_info->freed_extents[0];
afcdd129 2545 set_bit(BTRFS_FS_BARRIER, &fs_info->flags);
39279cc3 2546
5a3f23d5 2547 mutex_init(&fs_info->ordered_operations_mutex);
e02119d5 2548 mutex_init(&fs_info->tree_log_mutex);
925baedd 2549 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 2550 mutex_init(&fs_info->transaction_kthread_mutex);
2551 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2552 mutex_init(&fs_info->volume_mutex);
1bbc621e 2553 mutex_init(&fs_info->ro_block_group_mutex);
9e351cc8 2554 init_rwsem(&fs_info->commit_root_sem);
c71bf099 2555 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2556 init_rwsem(&fs_info->subvol_sem);
803b2f54 2557 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
fa9c0d79 2558
ad618368 2559 btrfs_init_dev_replace_locks(fs_info);
f9e92e40 2560 btrfs_init_qgroup(fs_info);
416ac51d 2561
fa9c0d79
CM		 2562 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2563 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2564
e6dcd2dc 2565 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2566 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2567 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2568 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2569
04216820
FM		 2570 INIT_LIST_HEAD(&fs_info->pinned_chunks);
2571
da17066c
JM		 2572 /* Usable values until the real ones are cached from the superblock */
2573 fs_info->nodesize = 4096;
2574 fs_info->sectorsize = 4096;
2575 fs_info->stripesize = 4096;
2576
53b381b3
DW		 2577 ret = btrfs_alloc_stripe_hash_table(fs_info);
2578 if (ret) {
83c8266a 2579 err = ret;
53b381b3
DW		 2580 goto fail_alloc;
2581 }
2582
da17066c 2583 __setup_root(tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2584
3c4bb26b 2585 invalidate_bdev(fs_devices->latest_bdev);
1104a885
DS		 2586
2587 /*
2588 * Read super block and check the signature bytes only
2589 */
a512bbf8 2590 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
92fc03fb
AJ		 2591 if (IS_ERR(bh)) {
2592 err = PTR_ERR(bh);
16cdcec7 2593 goto fail_alloc;
20b45077 2594 }
39279cc3 2595
1104a885
DS		 2596 /*
2597 * We want to check superblock checksum, the type is stored inside.
2598 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2599 */
ab8d0fc4 2600 if (btrfs_check_super_csum(fs_info, bh->b_data)) {
05135f59 2601 btrfs_err(fs_info, "superblock checksum mismatch");
1104a885 2602 err = -EINVAL;
b2acdddf 2603 brelse(bh);
1104a885
DS		 2604 goto fail_alloc;
2605 }
2606
2607 /*
2608 * super_copy is zeroed at allocation time and we never touch the
2609 * following bytes up to INFO_SIZE, the checksum is calculated from
2610 * the whole block of INFO_SIZE
2611 */
6c41761f
DS		 2612 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2613 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2614 sizeof(*fs_info->super_for_commit));
a061fc8d 2615 brelse(bh);
5f39d397 2616
6c41761f 2617 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2618
3d3a126a 2619 ret = btrfs_check_super_valid(fs_info);
1104a885 2620 if (ret) {
05135f59 2621 btrfs_err(fs_info, "superblock contains fatal errors");
1104a885
DS		 2622 err = -EINVAL;
2623 goto fail_alloc;
2624 }
2625
6c41761f 2626 disk_super = fs_info->super_copy;
0f7d52f4 2627 if (!btrfs_super_root(disk_super))
16cdcec7 2628 goto fail_alloc;
0f7d52f4 2629
acce952b 2630 /* check FS state, whether FS is broken. */
87533c47
MX		 2631 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2632 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2633
af31f5e5
CM		 2634 /*
2635 * run through our array of backup supers and setup
2636 * our ring pointer to the oldest one
2637 */
2638 generation = btrfs_super_generation(disk_super);
2639 find_oldest_super_backup(fs_info, generation);
2640
75e7cb7f
LB		 2641 /*
2642 * In the long term, we'll store the compression type in the super
2643 * block, and it'll be used for per file compression control.
2644 */
2645 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2646
2ff7e61e 2647 ret = btrfs_parse_options(fs_info, options, sb->s_flags);
2b82032c
YZ		 2648 if (ret) {
2649 err = ret;
16cdcec7 2650 goto fail_alloc;
2b82032c 2651 }
dfe25020 2652
f2b636e8
JB		 2653 features = btrfs_super_incompat_flags(disk_super) &
2654 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2655 if (features) {
05135f59
DS		 2656 btrfs_err(fs_info,
2657 "cannot mount because of unsupported optional features (%llx)",
2658 features);
f2b636e8 2659 err = -EINVAL;
16cdcec7 2660 goto fail_alloc;
f2b636e8
JB		 2661 }
2662
5d4f98a2 2663 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2664 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
0b246afa 2665 if (fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2666 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
5c1aab1d
NT		 2667 else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD)
2668 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD;
727011e0 2669
3173a18f 2670 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
05135f59 2671 btrfs_info(fs_info, "has skinny extents");
3173a18f 2672
727011e0
CM		 2673 /*
2674 * flag our filesystem as having big metadata blocks if
2675 * they are bigger than the page size
2676 */
09cbfeaf 2677 if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) {
727011e0 2678 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
05135f59
DS		 2679 btrfs_info(fs_info,
2680 "flagging fs with big metadata feature");
727011e0
CM		 2681 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2682 }
2683
bc3f116f 2684 nodesize = btrfs_super_nodesize(disk_super);
bc3f116f 2685 sectorsize = btrfs_super_sectorsize(disk_super);
b7f67055 2686 stripesize = sectorsize;
707e8a07 2687 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids));
963d678b 2688 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f 2689
da17066c
JM		 2690 /* Cache block sizes */
2691 fs_info->nodesize = nodesize;
2692 fs_info->sectorsize = sectorsize;
2693 fs_info->stripesize = stripesize;
2694
bc3f116f
CM		 2695 /*
2696 * mixed block groups end up with duplicate but slightly offset
2697 * extent buffers for the same range. It leads to corruptions
2698 */
2699 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
707e8a07 2700 (sectorsize != nodesize)) {
05135f59
DS		 2701 btrfs_err(fs_info,
2702"unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups",
2703 nodesize, sectorsize);
bc3f116f
CM		 2704 goto fail_alloc;
2705 }
2706
ceda0864
MX		 2707 /*
2708 * Needn't use the lock because there is no other task which will
2709 * update the flag.
2710 */
a6fa6fae 2711 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2712
f2b636e8
JB		 2713 features = btrfs_super_compat_ro_flags(disk_super) &
2714 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
bc98a42c 2715 if (!sb_rdonly(sb) && features) {
05135f59
DS		 2716 btrfs_err(fs_info,
2717 "cannot mount read-write because of unsupported optional features (%llx)",
c1c9ff7c 2718 features);
f2b636e8 2719 err = -EINVAL;
16cdcec7 2720 goto fail_alloc;
f2b636e8 2721 }
61d92c32 2722
5cdc7ad3 2723 max_active = fs_info->thread_pool_size;
61d92c32 2724
2a458198
ES		 2725 ret = btrfs_init_workqueues(fs_info, fs_devices);
2726 if (ret) {
2727 err = ret;
0dc3b84a
JB		 2728 goto fail_sb_buffer;
2729 }
4543df7e 2730
9e11ceee
JK		 2731 sb->s_bdi->congested_fn = btrfs_congested_fn;
2732 sb->s_bdi->congested_data = fs_info;
2733 sb->s_bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK;
2734 sb->s_bdi->ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_SIZE;
2735 sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super);
2736 sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE);
4575c9cc 2737
a061fc8d
CM		 2738 sb->s_blocksize = sectorsize;
2739 sb->s_blocksize_bits = blksize_bits(sectorsize);
ee87cf5e 2740 memcpy(&sb->s_uuid, fs_info->fsid, BTRFS_FSID_SIZE);
db94535d 2741
925baedd 2742 mutex_lock(&fs_info->chunk_mutex);
6bccf3ab 2743 ret = btrfs_read_sys_array(fs_info);
925baedd 2744 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2745 if (ret) {
05135f59 2746 btrfs_err(fs_info, "failed to read the system array: %d", ret);
5d4f98a2 2747 goto fail_sb_buffer;
84eed90f 2748 }
0b86a832 2749
84234f3a 2750 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832 2751
da17066c 2752 __setup_root(chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
0b86a832 2753
2ff7e61e 2754 chunk_root->node = read_tree_block(fs_info,
0b86a832 2755 btrfs_super_chunk_root(disk_super),
ce86cd59 2756 generation);
64c043de
LB		 2757 if (IS_ERR(chunk_root->node) ||
2758 !extent_buffer_uptodate(chunk_root->node)) {
05135f59 2759 btrfs_err(fs_info, "failed to read chunk root");
e5fffbac 2760 if (!IS_ERR(chunk_root->node))
2761 free_extent_buffer(chunk_root->node);
95ab1f64 2762 chunk_root->node = NULL;
af31f5e5 2763 goto fail_tree_roots;
83121942 2764 }
5d4f98a2
YZ		 2765 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2766 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2767
e17cade2 2768 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
b308bc2f 2769 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
e17cade2 2770
5b4aacef 2771 ret = btrfs_read_chunk_tree(fs_info);
2b82032c 2772 if (ret) {
05135f59 2773 btrfs_err(fs_info, "failed to read chunk tree: %d", ret);
af31f5e5 2774 goto fail_tree_roots;
2b82032c 2775 }
0b86a832 2776
8dabb742
SB		 2777 /*
2778 * keep the device that is marked to be the target device for the
2779 * dev_replace procedure
2780 */
9eaed21e 2781 btrfs_close_extra_devices(fs_devices, 0);
dfe25020 2782
a6b0d5c8 2783 if (!fs_devices->latest_bdev) {
05135f59 2784 btrfs_err(fs_info, "failed to read devices");
a6b0d5c8
CM		 2785 goto fail_tree_roots;
2786 }
2787
af31f5e5 2788retry_root_backup:
84234f3a 2789 generation = btrfs_super_generation(disk_super);
0b86a832 2790
2ff7e61e 2791 tree_root->node = read_tree_block(fs_info,
db94535d 2792 btrfs_super_root(disk_super),
ce86cd59 2793 generation);
64c043de
LB		 2794 if (IS_ERR(tree_root->node) ||
2795 !extent_buffer_uptodate(tree_root->node)) {
05135f59 2796 btrfs_warn(fs_info, "failed to read tree root");
e5fffbac 2797 if (!IS_ERR(tree_root->node))
2798 free_extent_buffer(tree_root->node);
95ab1f64 2799 tree_root->node = NULL;
af31f5e5 2800 goto recovery_tree_root;
83121942 2801 }
af31f5e5 2802
5d4f98a2
YZ		 2803 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2804 tree_root->commit_root = btrfs_root_node(tree_root);
69e9c6c6 2805 btrfs_set_root_refs(&tree_root->root_item, 1);
db94535d 2806
f32e48e9
CR		 2807 mutex_lock(&tree_root->objectid_mutex);
2808 ret = btrfs_find_highest_objectid(tree_root,
2809 &tree_root->highest_objectid);
2810 if (ret) {
2811 mutex_unlock(&tree_root->objectid_mutex);
2812 goto recovery_tree_root;
2813 }
2814
2815 ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID);
2816
2817 mutex_unlock(&tree_root->objectid_mutex);
2818
6bccf3ab 2819 ret = btrfs_read_roots(fs_info);
4bbcaa64 2820 if (ret)
af31f5e5 2821 goto recovery_tree_root;
f7a81ea4 2822
8929ecfa
YZ		 2823 fs_info->generation = generation;
2824 fs_info->last_trans_committed = generation;
8929ecfa 2825
68310a5e
ID		 2826 ret = btrfs_recover_balance(fs_info);
2827 if (ret) {
05135f59 2828 btrfs_err(fs_info, "failed to recover balance: %d", ret);
68310a5e
ID		 2829 goto fail_block_groups;
2830 }
2831
733f4fbb
SB		 2832 ret = btrfs_init_dev_stats(fs_info);
2833 if (ret) {
05135f59 2834 btrfs_err(fs_info, "failed to init dev_stats: %d", ret);
733f4fbb
SB		 2835 goto fail_block_groups;
2836 }
2837
8dabb742
SB		 2838 ret = btrfs_init_dev_replace(fs_info);
2839 if (ret) {
05135f59 2840 btrfs_err(fs_info, "failed to init dev_replace: %d", ret);
8dabb742
SB		 2841 goto fail_block_groups;
2842 }
2843
9eaed21e 2844 btrfs_close_extra_devices(fs_devices, 1);
8dabb742 2845
b7c35e81
AJ		 2846 ret = btrfs_sysfs_add_fsid(fs_devices, NULL);
2847 if (ret) {
05135f59
DS		 2848 btrfs_err(fs_info, "failed to init sysfs fsid interface: %d",
2849 ret);
b7c35e81
AJ		 2850 goto fail_block_groups;
2851 }
2852
2853 ret = btrfs_sysfs_add_device(fs_devices);
2854 if (ret) {
05135f59
DS		 2855 btrfs_err(fs_info, "failed to init sysfs device interface: %d",
2856 ret);
b7c35e81
AJ		 2857 goto fail_fsdev_sysfs;
2858 }
2859
96f3136e 2860 ret = btrfs_sysfs_add_mounted(fs_info);
c59021f8 2861 if (ret) {
05135f59 2862 btrfs_err(fs_info, "failed to init sysfs interface: %d", ret);
b7c35e81 2863 goto fail_fsdev_sysfs;
c59021f8 2864 }
2865
c59021f8 2866 ret = btrfs_init_space_info(fs_info);
2867 if (ret) {
05135f59 2868 btrfs_err(fs_info, "failed to initialize space info: %d", ret);
2365dd3c 2869 goto fail_sysfs;
c59021f8 2870 }
2871
5b4aacef 2872 ret = btrfs_read_block_groups(fs_info);
1b1d1f66 2873 if (ret) {
05135f59 2874 btrfs_err(fs_info, "failed to read block groups: %d", ret);
2365dd3c 2875 goto fail_sysfs;
1b1d1f66 2876 }
4330e183 2877
0f0d1272 2878 if (!sb_rdonly(sb) && !btrfs_check_rw_degradable(fs_info)) {
05135f59 2879 btrfs_warn(fs_info,
4330e183 2880 "writeable mount is not allowed due to too many missing devices");
2365dd3c 2881 goto fail_sysfs;
292fd7fc 2882 }
9078a3e1 2883
a74a4b97
CM		 2884 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2885 "btrfs-cleaner");
57506d50 2886 if (IS_ERR(fs_info->cleaner_kthread))
2365dd3c 2887 goto fail_sysfs;
a74a4b97
CM		 2888
2889 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2890 tree_root,
2891 "btrfs-transaction");
57506d50 2892 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2893 goto fail_cleaner;
a74a4b97 2894
583b7231 2895 if (!btrfs_test_opt(fs_info, NOSSD) &&
c289811c 2896 !fs_info->fs_devices->rotating) {
583b7231 2897 btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations");
c289811c
CM		 2898 }
2899
572d9ab7 2900 /*
01327610 2901 * Mount does not set all options immediately, we can do it now and do
572d9ab7
DS		 2902 * not have to wait for transaction commit
2903 */
2904 btrfs_apply_pending_changes(fs_info);
3818aea2 2905
21adbd5c 2906#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
0b246afa 2907 if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) {
2ff7e61e 2908 ret = btrfsic_mount(fs_info, fs_devices,
0b246afa 2909 btrfs_test_opt(fs_info,
21adbd5c
SB		 2910 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2911 1 : 0,
2912 fs_info->check_integrity_print_mask);
2913 if (ret)
05135f59
DS		 2914 btrfs_warn(fs_info,
2915 "failed to initialize integrity check module: %d",
2916 ret);
21adbd5c
SB		 2917 }
2918#endif
bcef60f2
AJ		 2919 ret = btrfs_read_qgroup_config(fs_info);
2920 if (ret)
2921 goto fail_trans_kthread;
21adbd5c 2922
96da0919
QW		 2923 /* do not make disk changes in broken FS or nologreplay is given */
2924 if (btrfs_super_log_root(disk_super) != 0 &&
0b246afa 2925 !btrfs_test_opt(fs_info, NOLOGREPLAY)) {
63443bf5 2926 ret = btrfs_replay_log(fs_info, fs_devices);
79787eaa 2927 if (ret) {
63443bf5 2928 err = ret;
28c16cbb 2929 goto fail_qgroup;
79787eaa 2930 }
e02119d5 2931 }
1a40e23b 2932
6bccf3ab 2933 ret = btrfs_find_orphan_roots(fs_info);
79787eaa 2934 if (ret)
28c16cbb 2935 goto fail_qgroup;
76dda93c 2936
bc98a42c 2937 if (!sb_rdonly(sb)) {
d68fc57b 2938 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2 2939 if (ret)
28c16cbb 2940 goto fail_qgroup;
90c711ab
ZB		 2941
2942 mutex_lock(&fs_info->cleaner_mutex);
5d4f98a2 2943 ret = btrfs_recover_relocation(tree_root);
90c711ab 2944 mutex_unlock(&fs_info->cleaner_mutex);
d7ce5843 2945 if (ret < 0) {
05135f59
DS		 2946 btrfs_warn(fs_info, "failed to recover relocation: %d",
2947 ret);
d7ce5843 2948 err = -EINVAL;
bcef60f2 2949 goto fail_qgroup;
d7ce5843 2950 }
7c2ca468 2951 }
1a40e23b 2952
3de4586c
CM		 2953 location.objectid = BTRFS_FS_TREE_OBJECTID;
2954 location.type = BTRFS_ROOT_ITEM_KEY;
cb517eab 2955 location.offset = 0;
3de4586c 2956
3de4586c 2957 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
3140c9a3
DC		 2958 if (IS_ERR(fs_info->fs_root)) {
2959 err = PTR_ERR(fs_info->fs_root);
bcef60f2 2960 goto fail_qgroup;
3140c9a3 2961 }
c289811c 2962
bc98a42c 2963 if (sb_rdonly(sb))
2b6ba629 2964 return 0;
59641015 2965
f8d468a1
OS		 2966 if (btrfs_test_opt(fs_info, CLEAR_CACHE) &&
2967 btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
6675df31
OS		 2968 clear_free_space_tree = 1;
2969 } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
2970 !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) {
2971 btrfs_warn(fs_info, "free space tree is invalid");
2972 clear_free_space_tree = 1;
2973 }
2974
2975 if (clear_free_space_tree) {
f8d468a1
OS		 2976 btrfs_info(fs_info, "clearing free space tree");
2977 ret = btrfs_clear_free_space_tree(fs_info);
2978 if (ret) {
2979 btrfs_warn(fs_info,
2980 "failed to clear free space tree: %d", ret);
6bccf3ab 2981 close_ctree(fs_info);
f8d468a1
OS		 2982 return ret;
2983 }
2984 }
2985
0b246afa 2986 if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) &&
511711af 2987 !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
05135f59 2988 btrfs_info(fs_info, "creating free space tree");
511711af
CM		 2989 ret = btrfs_create_free_space_tree(fs_info);
2990 if (ret) {
05135f59
DS		 2991 btrfs_warn(fs_info,
2992 "failed to create free space tree: %d", ret);
6bccf3ab 2993 close_ctree(fs_info);
511711af
CM		 2994 return ret;
2995 }
2996 }
2997
2b6ba629
ID		 2998 down_read(&fs_info->cleanup_work_sem);
2999 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
3000 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 3001 up_read(&fs_info->cleanup_work_sem);
6bccf3ab 3002 close_ctree(fs_info);
2b6ba629
ID		 3003 return ret;
3004 }
3005 up_read(&fs_info->cleanup_work_sem);
59641015 3006
2b6ba629
ID		 3007 ret = btrfs_resume_balance_async(fs_info);
3008 if (ret) {
05135f59 3009 btrfs_warn(fs_info, "failed to resume balance: %d", ret);
6bccf3ab 3010 close_ctree(fs_info);
2b6ba629 3011 return ret;
e3acc2a6
JB		 3012 }
3013
8dabb742
SB		 3014 ret = btrfs_resume_dev_replace_async(fs_info);
3015 if (ret) {
05135f59 3016 btrfs_warn(fs_info, "failed to resume device replace: %d", ret);
6bccf3ab 3017 close_ctree(fs_info);
8dabb742
SB		 3018 return ret;
3019 }
3020
b382a324
JS		 3021 btrfs_qgroup_rescan_resume(fs_info);
3022
4bbcaa64 3023 if (!fs_info->uuid_root) {
05135f59 3024 btrfs_info(fs_info, "creating UUID tree");
f7a81ea4
SB		 3025 ret = btrfs_create_uuid_tree(fs_info);
3026 if (ret) {
05135f59
DS		 3027 btrfs_warn(fs_info,
3028 "failed to create the UUID tree: %d", ret);
6bccf3ab 3029 close_ctree(fs_info);
f7a81ea4
SB		 3030 return ret;
3031 }
0b246afa 3032 } else if (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) ||
4bbcaa64
ES		 3033 fs_info->generation !=
3034 btrfs_super_uuid_tree_generation(disk_super)) {
05135f59 3035 btrfs_info(fs_info, "checking UUID tree");
70f80175
SB		 3036 ret = btrfs_check_uuid_tree(fs_info);
3037 if (ret) {
05135f59
DS		 3038 btrfs_warn(fs_info,
3039 "failed to check the UUID tree: %d", ret);
6bccf3ab 3040 close_ctree(fs_info);
70f80175
SB		 3041 return ret;
3042 }
3043 } else {
afcdd129 3044 set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags);
f7a81ea4 3045 }
afcdd129 3046 set_bit(BTRFS_FS_OPEN, &fs_info->flags);
47ab2a6c 3047
8dcddfa0
QW		 3048 /*
3049 * backuproot only affect mount behavior, and if open_ctree succeeded,
3050 * no need to keep the flag
3051 */
3052 btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
3053
ad2b2c80 3054 return 0;
39279cc3 3055
bcef60f2
AJ		 3056fail_qgroup:
3057 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM		 3058fail_trans_kthread:
3059 kthread_stop(fs_info->transaction_kthread);
2ff7e61e 3060 btrfs_cleanup_transaction(fs_info);
faa2dbf0 3061 btrfs_free_fs_roots(fs_info);
3f157a2f 3062fail_cleaner:
a74a4b97 3063 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 3064
3065 /*
3066 * make sure we're done with the btree inode before we stop our
3067 * kthreads
3068 */
3069 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 3070
2365dd3c 3071fail_sysfs:
6618a59b 3072 btrfs_sysfs_remove_mounted(fs_info);
2365dd3c 3073
b7c35e81
AJ		 3074fail_fsdev_sysfs:
3075 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
3076
1b1d1f66 3077fail_block_groups:
54067ae9 3078 btrfs_put_block_group_cache(fs_info);
af31f5e5
CM		 3079
3080fail_tree_roots:
3081 free_root_pointers(fs_info, 1);
2b8195bb 3082 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 3083
39279cc3 3084fail_sb_buffer:
7abadb64 3085 btrfs_stop_all_workers(fs_info);
5cdd7db6 3086 btrfs_free_block_groups(fs_info);
16cdcec7 3087fail_alloc:
4543df7e 3088fail_iput:
586e46e2
ID		 3089 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3090
4543df7e 3091 iput(fs_info->btree_inode);
c404e0dc
MX		 3092fail_bio_counter:
3093 percpu_counter_destroy(&fs_info->bio_counter);
963d678b
MX		 3094fail_delalloc_bytes:
3095 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX		 3096fail_dirty_metadata_bytes:
3097 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
76dda93c
YZ		 3098fail_srcu:
3099 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 3100fail:
53b381b3 3101 btrfs_free_stripe_hash_table(fs_info);
586e46e2 3102 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 3103 return err;
af31f5e5
CM		 3104
3105recovery_tree_root:
0b246afa 3106 if (!btrfs_test_opt(fs_info, USEBACKUPROOT))
af31f5e5
CM		 3107 goto fail_tree_roots;
3108
3109 free_root_pointers(fs_info, 0);
3110
3111 /* don't use the log in recovery mode, it won't be valid */
3112 btrfs_set_super_log_root(disk_super, 0);
3113
3114 /* we can't trust the free space cache either */
3115 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3116
3117 ret = next_root_backup(fs_info, fs_info->super_copy,
3118 &num_backups_tried, &backup_index);
3119 if (ret == -1)
3120 goto fail_block_groups;
3121 goto retry_root_backup;
eb60ceac
CM		 3122}
3123
f2984462
CM		 3124static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3125{
f2984462
CM		 3126 if (uptodate) {
3127 set_buffer_uptodate(bh);
3128 } else {
442a4f63
SB		 3129 struct btrfs_device *device = (struct btrfs_device *)
3130 bh->b_private;
3131
fb456252 3132 btrfs_warn_rl_in_rcu(device->fs_info,
b14af3b4 3133 "lost page write due to IO error on %s",
606686ee 3134 rcu_str_deref(device->name));
01327610 3135 /* note, we don't set_buffer_write_io_error because we have
1259ab75
CM		 3136 * our own ways of dealing with the IO errors
3137 */
f2984462 3138 clear_buffer_uptodate(bh);
442a4f63 3139 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM		 3140 }
3141 unlock_buffer(bh);
3142 put_bh(bh);
3143}
3144
29c36d72
AJ		 3145int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num,
3146 struct buffer_head **bh_ret)
3147{
3148 struct buffer_head *bh;
3149 struct btrfs_super_block *super;
3150 u64 bytenr;
3151
3152 bytenr = btrfs_sb_offset(copy_num);
3153 if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode))
3154 return -EINVAL;
3155
9f6d2510 3156 bh = __bread(bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, BTRFS_SUPER_INFO_SIZE);
29c36d72
AJ		 3157 /*
3158 * If we fail to read from the underlying devices, as of now
3159 * the best option we have is to mark it EIO.
3160 */
3161 if (!bh)
3162 return -EIO;
3163
3164 super = (struct btrfs_super_block *)bh->b_data;
3165 if (btrfs_super_bytenr(super) != bytenr ||
3166 btrfs_super_magic(super) != BTRFS_MAGIC) {
3167 brelse(bh);
3168 return -EINVAL;
3169 }
3170
3171 *bh_ret = bh;
3172 return 0;
3173}
3174
3175
a512bbf8
YZ		 3176struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3177{
3178 struct buffer_head *bh;
3179 struct buffer_head *latest = NULL;
3180 struct btrfs_super_block *super;
3181 int i;
3182 u64 transid = 0;
92fc03fb 3183 int ret = -EINVAL;
a512bbf8
YZ		 3184
3185 /* we would like to check all the supers, but that would make
3186 * a btrfs mount succeed after a mkfs from a different FS.
3187 * So, we need to add a special mount option to scan for
3188 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3189 */
3190 for (i = 0; i < 1; i++) {
29c36d72
AJ		 3191 ret = btrfs_read_dev_one_super(bdev, i, &bh);
3192 if (ret)
a512bbf8
YZ		 3193 continue;
3194
3195 super = (struct btrfs_super_block *)bh->b_data;
a512bbf8
YZ		 3196
3197 if (!latest || btrfs_super_generation(super) > transid) {
3198 brelse(latest);
3199 latest = bh;
3200 transid = btrfs_super_generation(super);
3201 } else {
3202 brelse(bh);
3203 }
3204 }
92fc03fb
AJ		 3205
3206 if (!latest)
3207 return ERR_PTR(ret);
3208
a512bbf8
YZ		 3209 return latest;
3210}
3211
4eedeb75 3212/*
abbb3b8e
DS		 3213 * Write superblock @sb to the @device. Do not wait for completion, all the
3214 * buffer heads we write are pinned.
4eedeb75 3215 *
abbb3b8e
DS		 3216 * Write @max_mirrors copies of the superblock, where 0 means default that fit
3217 * the expected device size at commit time. Note that max_mirrors must be
3218 * same for write and wait phases.
4eedeb75 3219 *
abbb3b8e 3220 * Return number of errors when buffer head is not found or submission fails.
4eedeb75 3221 */
a512bbf8 3222static int write_dev_supers(struct btrfs_device *device,
abbb3b8e 3223 struct btrfs_super_block *sb, int max_mirrors)
a512bbf8
YZ		 3224{
3225 struct buffer_head *bh;
3226 int i;
3227 int ret;
3228 int errors = 0;
3229 u32 crc;
3230 u64 bytenr;
a512bbf8
YZ		 3231
3232 if (max_mirrors == 0)
3233 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3234
a512bbf8
YZ		 3235 for (i = 0; i < max_mirrors; i++) {
3236 bytenr = btrfs_sb_offset(i);
935e5cc9
MX		 3237 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3238 device->commit_total_bytes)
a512bbf8
YZ		 3239 break;
3240
abbb3b8e 3241 btrfs_set_super_bytenr(sb, bytenr);
4eedeb75 3242
abbb3b8e
DS		 3243 crc = ~(u32)0;
3244 crc = btrfs_csum_data((const char *)sb + BTRFS_CSUM_SIZE, crc,
3245 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
3246 btrfs_csum_final(crc, sb->csum);
4eedeb75 3247
abbb3b8e 3248 /* One reference for us, and we leave it for the caller */
9f6d2510 3249 bh = __getblk(device->bdev, bytenr / BTRFS_BDEV_BLOCKSIZE,
abbb3b8e
DS		 3250 BTRFS_SUPER_INFO_SIZE);
3251 if (!bh) {
3252 btrfs_err(device->fs_info,
3253 "couldn't get super buffer head for bytenr %llu",
3254 bytenr);
3255 errors++;
4eedeb75 3256 continue;
abbb3b8e 3257 }
634554dc 3258
abbb3b8e 3259 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
a512bbf8 3260
abbb3b8e
DS		 3261 /* one reference for submit_bh */
3262 get_bh(bh);
4eedeb75 3263
abbb3b8e
DS		 3264 set_buffer_uptodate(bh);
3265 lock_buffer(bh);
3266 bh->b_end_io = btrfs_end_buffer_write_sync;
3267 bh->b_private = device;
a512bbf8 3268
387125fc
CM		 3269 /*
3270 * we fua the first super. The others we allow
3271 * to go down lazy.
3272 */
8d910125
JK		 3273 if (i == 0) {
3274 ret = btrfsic_submit_bh(REQ_OP_WRITE,
db95c876 3275 REQ_SYNC | REQ_FUA | REQ_META | REQ_PRIO, bh);
8d910125 3276 } else {
db95c876
DS		 3277 ret = btrfsic_submit_bh(REQ_OP_WRITE,
3278 REQ_SYNC | REQ_META | REQ_PRIO, bh);
8d910125 3279 }
4eedeb75 3280 if (ret)
a512bbf8 3281 errors++;
a512bbf8
YZ		 3282 }
3283 return errors < i ? 0 : -1;
3284}
3285
abbb3b8e
DS		 3286/*
3287 * Wait for write completion of superblocks done by write_dev_supers,
3288 * @max_mirrors same for write and wait phases.
3289 *
3290 * Return number of errors when buffer head is not found or not marked up to
3291 * date.
3292 */
3293static int wait_dev_supers(struct btrfs_device *device, int max_mirrors)
3294{
3295 struct buffer_head *bh;
3296 int i;
3297 int errors = 0;
3298 u64 bytenr;
3299
3300 if (max_mirrors == 0)
3301 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3302
3303 for (i = 0; i < max_mirrors; i++) {
3304 bytenr = btrfs_sb_offset(i);
3305 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3306 device->commit_total_bytes)
3307 break;
3308
9f6d2510
DS		 3309 bh = __find_get_block(device->bdev,
3310 bytenr / BTRFS_BDEV_BLOCKSIZE,
abbb3b8e
DS		 3311 BTRFS_SUPER_INFO_SIZE);
3312 if (!bh) {
3313 errors++;
3314 continue;
3315 }
3316 wait_on_buffer(bh);
3317 if (!buffer_uptodate(bh))
3318 errors++;
3319
3320 /* drop our reference */
3321 brelse(bh);
3322
3323 /* drop the reference from the writing run */
3324 brelse(bh);
3325 }
3326
3327 return errors < i ? 0 : -1;
3328}
3329
387125fc
CM		 3330/*
3331 * endio for the write_dev_flush, this will wake anyone waiting
3332 * for the barrier when it is done
3333 */
4246a0b6 3334static void btrfs_end_empty_barrier(struct bio *bio)
387125fc 3335{
e0ae9994 3336 complete(bio->bi_private);
387125fc
CM		 3337}
3338
3339/*
4fc6441a
AJ		 3340 * Submit a flush request to the device if it supports it. Error handling is
3341 * done in the waiting counterpart.
387125fc 3342 */
4fc6441a 3343static void write_dev_flush(struct btrfs_device *device)
387125fc 3344{
c2a9c7ab 3345 struct request_queue *q = bdev_get_queue(device->bdev);
e0ae9994 3346 struct bio *bio = device->flush_bio;
387125fc 3347
c2a9c7ab 3348 if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags))
4fc6441a 3349 return;
387125fc 3350
e0ae9994 3351 bio_reset(bio);
387125fc 3352 bio->bi_end_io = btrfs_end_empty_barrier;
74d46992 3353 bio_set_dev(bio, device->bdev);
8d910125 3354 bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH;
387125fc
CM		 3355 init_completion(&device->flush_wait);
3356 bio->bi_private = &device->flush_wait;
387125fc 3357
43a01111 3358 btrfsic_submit_bio(bio);
e0ae9994 3359 device->flush_bio_sent = 1;
4fc6441a 3360}
387125fc 3361
4fc6441a
AJ		 3362/*
3363 * If the flush bio has been submitted by write_dev_flush, wait for it.
3364 */
8c27cb35 3365static blk_status_t wait_dev_flush(struct btrfs_device *device)
4fc6441a 3366{
4fc6441a 3367 struct bio *bio = device->flush_bio;
387125fc 3368
e0ae9994 3369 if (!device->flush_bio_sent)
58efbc9f 3370 return BLK_STS_OK;
387125fc 3371
e0ae9994 3372 device->flush_bio_sent = 0;
2980d574 3373 wait_for_completion_io(&device->flush_wait);
387125fc 3374
8c27cb35 3375 return bio->bi_status;
387125fc 3376}
387125fc 3377
d10b82fe 3378static int check_barrier_error(struct btrfs_fs_info *fs_info)
401b41e5 3379{
d10b82fe 3380 if (!btrfs_check_rw_degradable(fs_info))
401b41e5 3381 return -EIO;
387125fc
CM		 3382 return 0;
3383}
3384
3385/*
3386 * send an empty flush down to each device in parallel,
3387 * then wait for them
3388 */
3389static int barrier_all_devices(struct btrfs_fs_info *info)
3390{
3391 struct list_head *head;
3392 struct btrfs_device *dev;
5af3e8cc 3393 int errors_wait = 0;
4e4cbee9 3394 blk_status_t ret;
387125fc
CM		 3395
3396 /* send down all the barriers */
3397 head = &info->fs_devices->devices;
3398 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3399 if (dev->missing)
3400 continue;
cea7c8bf 3401 if (!dev->bdev)
387125fc 3402 continue;
387125fc
CM		 3403 if (!dev->in_fs_metadata || !dev->writeable)
3404 continue;
3405
4fc6441a 3406 write_dev_flush(dev);
58efbc9f 3407 dev->last_flush_error = BLK_STS_OK;
387125fc
CM		 3408 }
3409
3410 /* wait for all the barriers */
3411 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3412 if (dev->missing)
3413 continue;
387125fc 3414 if (!dev->bdev) {
5af3e8cc 3415 errors_wait++;
387125fc
CM		 3416 continue;
3417 }
3418 if (!dev->in_fs_metadata || !dev->writeable)
3419 continue;
3420
4fc6441a 3421 ret = wait_dev_flush(dev);
401b41e5
AJ		 3422 if (ret) {
3423 dev->last_flush_error = ret;
66b4993e
DS		 3424 btrfs_dev_stat_inc_and_print(dev,
3425 BTRFS_DEV_STAT_FLUSH_ERRS);
5af3e8cc 3426 errors_wait++;
401b41e5
AJ		 3427 }
3428 }
3429
cea7c8bf 3430 if (errors_wait) {
401b41e5
AJ		 3431 /*
3432 * At some point we need the status of all disks
3433 * to arrive at the volume status. So error checking
3434 * is being pushed to a separate loop.
3435 */
d10b82fe 3436 return check_barrier_error(info);
387125fc 3437 }
387125fc
CM		 3438 return 0;
3439}
3440
943c6e99
ZL		 3441int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags)
3442{
8789f4fe
ZL		 3443 int raid_type;
3444 int min_tolerated = INT_MAX;
943c6e99 3445
8789f4fe
ZL		 3446 if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 ||
3447 (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE))
3448 min_tolerated = min(min_tolerated,
3449 btrfs_raid_array[BTRFS_RAID_SINGLE].
3450 tolerated_failures);
943c6e99 3451
8789f4fe
ZL		 3452 for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) {
3453 if (raid_type == BTRFS_RAID_SINGLE)
3454 continue;
3455 if (!(flags & btrfs_raid_group[raid_type]))
3456 continue;
3457 min_tolerated = min(min_tolerated,
3458 btrfs_raid_array[raid_type].
3459 tolerated_failures);
3460 }
943c6e99 3461
8789f4fe 3462 if (min_tolerated == INT_MAX) {
ab8d0fc4 3463 pr_warn("BTRFS: unknown raid flag: %llu", flags);
8789f4fe
ZL		 3464 min_tolerated = 0;
3465 }
3466
3467 return min_tolerated;
943c6e99
ZL		 3468}
3469
eece6a9c 3470int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors)
f2984462 3471{
e5e9a520 3472 struct list_head *head;
f2984462 3473 struct btrfs_device *dev;
a061fc8d 3474 struct btrfs_super_block *sb;
f2984462 3475 struct btrfs_dev_item *dev_item;
f2984462
CM		 3476 int ret;
3477 int do_barriers;
a236aed1
CM		 3478 int max_errors;
3479 int total_errors = 0;
a061fc8d 3480 u64 flags;
f2984462 3481
0b246afa 3482 do_barriers = !btrfs_test_opt(fs_info, NOBARRIER);
fed3b381
LB		 3483
3484 /*
3485 * max_mirrors == 0 indicates we're from commit_transaction,
3486 * not from fsync where the tree roots in fs_info have not
3487 * been consistent on disk.
3488 */
3489 if (max_mirrors == 0)
3490 backup_super_roots(fs_info);
f2984462 3491
0b246afa 3492 sb = fs_info->super_for_commit;
a061fc8d 3493 dev_item = &sb->dev_item;
e5e9a520 3494
0b246afa
JM		 3495 mutex_lock(&fs_info->fs_devices->device_list_mutex);
3496 head = &fs_info->fs_devices->devices;
3497 max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1;
387125fc 3498
5af3e8cc 3499 if (do_barriers) {
0b246afa 3500 ret = barrier_all_devices(fs_info);
5af3e8cc
SB		 3501 if (ret) {
3502 mutex_unlock(
0b246afa
JM		 3503 &fs_info->fs_devices->device_list_mutex);
3504 btrfs_handle_fs_error(fs_info, ret,
3505 "errors while submitting device barriers.");
5af3e8cc
SB		 3506 return ret;
3507 }
3508 }
387125fc 3509
1f78160c 3510 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3511 if (!dev->bdev) {
3512 total_errors++;
3513 continue;
3514 }
2b82032c 3515 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3516 continue;
3517
2b82032c 3518 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 3519 btrfs_set_stack_device_type(dev_item, dev->type);
3520 btrfs_set_stack_device_id(dev_item, dev->devid);
7df69d3e 3521 btrfs_set_stack_device_total_bytes(dev_item,
935e5cc9 3522 dev->commit_total_bytes);
ce7213c7
MX		 3523 btrfs_set_stack_device_bytes_used(dev_item,
3524 dev->commit_bytes_used);
a061fc8d
CM		 3525 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3526 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3527 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3528 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
44880fdc 3529 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_FSID_SIZE);
a512bbf8 3530
a061fc8d
CM		 3531 flags = btrfs_super_flags(sb);
3532 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3533
abbb3b8e 3534 ret = write_dev_supers(dev, sb, max_mirrors);
a236aed1
CM		 3535 if (ret)
3536 total_errors++;
f2984462 3537 }
a236aed1 3538 if (total_errors > max_errors) {
0b246afa
JM		 3539 btrfs_err(fs_info, "%d errors while writing supers",
3540 total_errors);
3541 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
79787eaa 3542
9d565ba4 3543 /* FUA is masked off if unsupported and can't be the reason */
0b246afa
JM		 3544 btrfs_handle_fs_error(fs_info, -EIO,
3545 "%d errors while writing supers",
3546 total_errors);
9d565ba4 3547 return -EIO;
a236aed1 3548 }
f2984462 3549
a512bbf8 3550 total_errors = 0;
1f78160c 3551 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3552 if (!dev->bdev)
3553 continue;
2b82032c 3554 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3555 continue;
3556
abbb3b8e 3557 ret = wait_dev_supers(dev, max_mirrors);
a512bbf8
YZ		 3558 if (ret)
3559 total_errors++;
f2984462 3560 }
0b246afa 3561 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
a236aed1 3562 if (total_errors > max_errors) {
0b246afa
JM		 3563 btrfs_handle_fs_error(fs_info, -EIO,
3564 "%d errors while writing supers",
3565 total_errors);
79787eaa 3566 return -EIO;
a236aed1 3567 }
f2984462
CM		 3568 return 0;
3569}
3570
cb517eab
MX		 3571/* Drop a fs root from the radix tree and free it. */
3572void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3573 struct btrfs_root *root)
2619ba1f 3574{
4df27c4d 3575 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 3576 radix_tree_delete(&fs_info->fs_roots_radix,
3577 (unsigned long)root->root_key.objectid);
4df27c4d 3578 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 3579
3580 if (btrfs_root_refs(&root->root_item) == 0)
3581 synchronize_srcu(&fs_info->subvol_srcu);
3582
1c1ea4f7 3583 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
3321719e 3584 btrfs_free_log(NULL, root);
1c1ea4f7
LB		 3585 if (root->reloc_root) {
3586 free_extent_buffer(root->reloc_root->node);
3587 free_extent_buffer(root->reloc_root->commit_root);
3588 btrfs_put_fs_root(root->reloc_root);
3589 root->reloc_root = NULL;
3590 }
3591 }
3321719e 3592
faa2dbf0
JB		 3593 if (root->free_ino_pinned)
3594 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3595 if (root->free_ino_ctl)
3596 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3597 free_fs_root(root);
4df27c4d
YZ		 3598}
3599
3600static void free_fs_root(struct btrfs_root *root)
3601{
57cdc8db 3602 iput(root->ino_cache_inode);
4df27c4d 3603 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
2ff7e61e 3604 btrfs_free_block_rsv(root->fs_info, root->orphan_block_rsv);
1cb048f5 3605 root->orphan_block_rsv = NULL;
0ee5dc67
AV		 3606 if (root->anon_dev)
3607 free_anon_bdev(root->anon_dev);
8257b2dc
MX		 3608 if (root->subv_writers)
3609 btrfs_free_subvolume_writers(root->subv_writers);
4df27c4d
YZ		 3610 free_extent_buffer(root->node);
3611 free_extent_buffer(root->commit_root);
581bb050
LZ		 3612 kfree(root->free_ino_ctl);
3613 kfree(root->free_ino_pinned);
d397712b 3614 kfree(root->name);
b0feb9d9 3615 btrfs_put_fs_root(root);
2619ba1f
CM		 3616}
3617
cb517eab
MX		 3618void btrfs_free_fs_root(struct btrfs_root *root)
3619{
3620 free_fs_root(root);
2619ba1f
CM		 3621}
3622
c146afad 3623int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3624{
c146afad
YZ		 3625 u64 root_objectid = 0;
3626 struct btrfs_root *gang[8];
65d33fd7
QW		 3627 int i = 0;
3628 int err = 0;
3629 unsigned int ret = 0;
3630 int index;
e089f05c 3631
c146afad 3632 while (1) {
65d33fd7 3633 index = srcu_read_lock(&fs_info->subvol_srcu);
c146afad
YZ		 3634 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3635 (void **)gang, root_objectid,
3636 ARRAY_SIZE(gang));
65d33fd7
QW		 3637 if (!ret) {
3638 srcu_read_unlock(&fs_info->subvol_srcu, index);
c146afad 3639 break;
65d33fd7 3640 }
5d4f98a2 3641 root_objectid = gang[ret - 1]->root_key.objectid + 1;
65d33fd7 3642
c146afad 3643 for (i = 0; i < ret; i++) {
65d33fd7
QW		 3644 /* Avoid to grab roots in dead_roots */
3645 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3646 gang[i] = NULL;
3647 continue;
3648 }
3649 /* grab all the search result for later use */
3650 gang[i] = btrfs_grab_fs_root(gang[i]);
3651 }
3652 srcu_read_unlock(&fs_info->subvol_srcu, index);
66b4ffd1 3653
65d33fd7
QW		 3654 for (i = 0; i < ret; i++) {
3655 if (!gang[i])
3656 continue;
c146afad 3657 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 3658 err = btrfs_orphan_cleanup(gang[i]);
3659 if (err)
65d33fd7
QW		 3660 break;
3661 btrfs_put_fs_root(gang[i]);
c146afad
YZ		 3662 }
3663 root_objectid++;
3664 }
65d33fd7
QW		 3665
3666 /* release the uncleaned roots due to error */
3667 for (; i < ret; i++) {
3668 if (gang[i])
3669 btrfs_put_fs_root(gang[i]);
3670 }
3671 return err;
c146afad 3672}
a2135011 3673
6bccf3ab 3674int btrfs_commit_super(struct btrfs_fs_info *fs_info)
c146afad 3675{
6bccf3ab 3676 struct btrfs_root *root = fs_info->tree_root;
c146afad 3677 struct btrfs_trans_handle *trans;
a74a4b97 3678
0b246afa 3679 mutex_lock(&fs_info->cleaner_mutex);
2ff7e61e 3680 btrfs_run_delayed_iputs(fs_info);
0b246afa
JM		 3681 mutex_unlock(&fs_info->cleaner_mutex);
3682 wake_up_process(fs_info->cleaner_kthread);
c71bf099
YZ		 3683
3684 /* wait until ongoing cleanup work done */
0b246afa
JM		 3685 down_write(&fs_info->cleanup_work_sem);
3686 up_write(&fs_info->cleanup_work_sem);
c71bf099 3687
7a7eaa40 3688 trans = btrfs_join_transaction(root);
3612b495
TI		 3689 if (IS_ERR(trans))
3690 return PTR_ERR(trans);
3a45bb20 3691 return btrfs_commit_transaction(trans);
c146afad
YZ		 3692}
3693
6bccf3ab 3694void close_ctree(struct btrfs_fs_info *fs_info)
c146afad 3695{
6bccf3ab 3696 struct btrfs_root *root = fs_info->tree_root;
c146afad
YZ		 3697 int ret;
3698
afcdd129 3699 set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags);
c146afad 3700
7343dd61 3701 /* wait for the qgroup rescan worker to stop */
d06f23d6 3702 btrfs_qgroup_wait_for_completion(fs_info, false);
7343dd61 3703
803b2f54
SB		 3704 /* wait for the uuid_scan task to finish */
3705 down(&fs_info->uuid_tree_rescan_sem);
3706 /* avoid complains from lockdep et al., set sem back to initial state */
3707 up(&fs_info->uuid_tree_rescan_sem);
3708
837d5b6e 3709 /* pause restriper - we want to resume on mount */
aa1b8cd4 3710 btrfs_pause_balance(fs_info);
837d5b6e 3711
8dabb742
SB		 3712 btrfs_dev_replace_suspend_for_unmount(fs_info);
3713
aa1b8cd4 3714 btrfs_scrub_cancel(fs_info);
4cb5300b
CM		 3715
3716 /* wait for any defraggers to finish */
3717 wait_event(fs_info->transaction_wait,
3718 (atomic_read(&fs_info->defrag_running) == 0));
3719
3720 /* clear out the rbtree of defraggable inodes */
26176e7c 3721 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3722
21c7e756
MX		 3723 cancel_work_sync(&fs_info->async_reclaim_work);
3724
bc98a42c 3725 if (!sb_rdonly(fs_info->sb)) {
e44163e1
JM		 3726 /*
3727 * If the cleaner thread is stopped and there are
3728 * block groups queued for removal, the deletion will be
3729 * skipped when we quit the cleaner thread.
3730 */
0b246afa 3731 btrfs_delete_unused_bgs(fs_info);
e44163e1 3732
6bccf3ab 3733 ret = btrfs_commit_super(fs_info);
acce952b 3734 if (ret)
04892340 3735 btrfs_err(fs_info, "commit super ret %d", ret);
acce952b 3736 }
3737
87533c47 3738 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
2ff7e61e 3739 btrfs_error_commit_super(fs_info);
0f7d52f4 3740
e3029d9f
AV		 3741 kthread_stop(fs_info->transaction_kthread);
3742 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3743
afcdd129 3744 set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags);
f25784b3 3745
04892340 3746 btrfs_free_qgroup_config(fs_info);
bcef60f2 3747
963d678b 3748 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
04892340 3749 btrfs_info(fs_info, "at unmount delalloc count %lld",
963d678b 3750 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3751 }
bcc63abb 3752
6618a59b 3753 btrfs_sysfs_remove_mounted(fs_info);
b7c35e81 3754 btrfs_sysfs_remove_fsid(fs_info->fs_devices);
5ac1d209 3755
faa2dbf0 3756 btrfs_free_fs_roots(fs_info);
d10c5f31 3757
1a4319cc
LB		 3758 btrfs_put_block_group_cache(fs_info);
3759
de348ee0
WS		 3760 /*
3761 * we must make sure there is not any read request to
3762 * submit after we stopping all workers.
3763 */
3764 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
96192499
JB		 3765 btrfs_stop_all_workers(fs_info);
3766
5cdd7db6
FM		 3767 btrfs_free_block_groups(fs_info);
3768
afcdd129 3769 clear_bit(BTRFS_FS_OPEN, &fs_info->flags);
13e6c37b 3770 free_root_pointers(fs_info, 1);
9ad6b7bc 3771
13e6c37b 3772 iput(fs_info->btree_inode);
d6bfde87 3773
21adbd5c 3774#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
0b246afa 3775 if (btrfs_test_opt(fs_info, CHECK_INTEGRITY))
2ff7e61e 3776 btrfsic_unmount(fs_info->fs_devices);
21adbd5c
SB		 3777#endif
3778
dfe25020 3779 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3780 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3781
e2d84521 3782 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 3783 percpu_counter_destroy(&fs_info->delalloc_bytes);
c404e0dc 3784 percpu_counter_destroy(&fs_info->bio_counter);
76dda93c 3785 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 3786
53b381b3
DW		 3787 btrfs_free_stripe_hash_table(fs_info);
3788
cdfb080e 3789 __btrfs_free_block_rsv(root->orphan_block_rsv);
1cb048f5 3790 root->orphan_block_rsv = NULL;
04216820 3791
04216820
FM		 3792 while (!list_empty(&fs_info->pinned_chunks)) {
3793 struct extent_map *em;
3794
3795 em = list_first_entry(&fs_info->pinned_chunks,
3796 struct extent_map, list);
3797 list_del_init(&em->list);
3798 free_extent_map(em);
3799 }
eb60ceac
CM		 3800}
3801
b9fab919
CM		 3802int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3803 int atomic)
5f39d397 3804{
1259ab75 3805 int ret;
727011e0 3806 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 3807
0b32f4bb 3808 ret = extent_buffer_uptodate(buf);
1259ab75
CM		 3809 if (!ret)
3810 return ret;
3811
3812 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM		 3813 parent_transid, atomic);
3814 if (ret == -EAGAIN)
3815 return ret;
1259ab75 3816 return !ret;
5f39d397
CM		 3817}
3818
5f39d397
CM		 3819void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3820{
0b246afa 3821 struct btrfs_fs_info *fs_info;
06ea65a3 3822 struct btrfs_root *root;
5f39d397 3823 u64 transid = btrfs_header_generation(buf);
b9473439 3824 int was_dirty;
b4ce94de 3825
06ea65a3
JB		 3826#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3827 /*
3828 * This is a fast path so only do this check if we have sanity tests
3829 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3830 * outside of the sanity tests.
3831 */
3832 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3833 return;
3834#endif
3835 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
0b246afa 3836 fs_info = root->fs_info;
b9447ef8 3837 btrfs_assert_tree_locked(buf);
0b246afa 3838 if (transid != fs_info->generation)
5d163e0e 3839 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n",
0b246afa 3840 buf->start, transid, fs_info->generation);
0b32f4bb 3841 was_dirty = set_extent_buffer_dirty(buf);
e2d84521 3842 if (!was_dirty)
104b4e51
NB		 3843 percpu_counter_add_batch(&fs_info->dirty_metadata_bytes,
3844 buf->len,
3845 fs_info->dirty_metadata_batch);
1f21ef0a 3846#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
557ea5dd 3847 if (btrfs_header_level(buf) == 0 && btrfs_check_leaf(root, buf)) {
a4f78750 3848 btrfs_print_leaf(buf);
1f21ef0a
FM		 3849 ASSERT(0);
3850 }
3851#endif
eb60ceac
CM		 3852}
3853
2ff7e61e 3854static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info,
b53d3f5d 3855 int flush_delayed)
16cdcec7
MX		 3856{
3857 /*
3858 * looks as though older kernels can get into trouble with
3859 * this code, they end up stuck in balance_dirty_pages forever
3860 */
e2d84521 3861 int ret;
16cdcec7
MX		 3862
3863 if (current->flags & PF_MEMALLOC)
3864 return;
3865
b53d3f5d 3866 if (flush_delayed)
2ff7e61e 3867 btrfs_balance_delayed_items(fs_info);
16cdcec7 3868
0b246afa 3869 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
e2d84521
MX		 3870 BTRFS_DIRTY_METADATA_THRESH);
3871 if (ret > 0) {
0b246afa 3872 balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping);
16cdcec7 3873 }
16cdcec7
MX		 3874}
3875
2ff7e61e 3876void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info)
35b7e476 3877{
2ff7e61e 3878 __btrfs_btree_balance_dirty(fs_info, 1);
b53d3f5d 3879}
585ad2c3 3880
2ff7e61e 3881void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info)
b53d3f5d 3882{
2ff7e61e 3883 __btrfs_btree_balance_dirty(fs_info, 0);
35b7e476 3884}
6b80053d 3885
ca7a79ad 3886int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 3887{
727011e0 3888 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
2ff7e61e
JM		 3889 struct btrfs_fs_info *fs_info = root->fs_info;
3890
3891 return btree_read_extent_buffer_pages(fs_info, buf, parent_transid);
6b80053d 3892}
0da5468f 3893
3d3a126a 3894static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info)
acce952b 3895{
c926093e 3896 struct btrfs_super_block *sb = fs_info->super_copy;
319e4d06
QW		 3897 u64 nodesize = btrfs_super_nodesize(sb);
3898 u64 sectorsize = btrfs_super_sectorsize(sb);
c926093e
DS		 3899 int ret = 0;
3900
319e4d06 3901 if (btrfs_super_magic(sb) != BTRFS_MAGIC) {
ab8d0fc4 3902 btrfs_err(fs_info, "no valid FS found");
319e4d06
QW		 3903 ret = -EINVAL;
3904 }
3905 if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP)
ab8d0fc4 3906 btrfs_warn(fs_info, "unrecognized super flag: %llu",
319e4d06 3907 btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP);
21e7626b 3908 if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) {
ab8d0fc4 3909 btrfs_err(fs_info, "tree_root level too big: %d >= %d",
21e7626b 3910 btrfs_super_root_level(sb), BTRFS_MAX_LEVEL);
c926093e
DS		 3911 ret = -EINVAL;
3912 }
21e7626b 3913 if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) {
ab8d0fc4 3914 btrfs_err(fs_info, "chunk_root level too big: %d >= %d",
21e7626b 3915 btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL);
c926093e
DS		 3916 ret = -EINVAL;
3917 }
21e7626b 3918 if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) {
ab8d0fc4 3919 btrfs_err(fs_info, "log_root level too big: %d >= %d",
21e7626b 3920 btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL);
c926093e
DS		 3921 ret = -EINVAL;
3922 }
3923
1104a885 3924 /*
319e4d06
QW		 3925 * Check sectorsize and nodesize first, other check will need it.
3926 * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here.
1104a885 3927 */
319e4d06
QW		 3928 if (!is_power_of_2(sectorsize) || sectorsize < 4096 ||
3929 sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) {
ab8d0fc4 3930 btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize);
319e4d06
QW		 3931 ret = -EINVAL;
3932 }
3933 /* Only PAGE SIZE is supported yet */
09cbfeaf 3934 if (sectorsize != PAGE_SIZE) {
ab8d0fc4
JM		 3935 btrfs_err(fs_info,
3936 "sectorsize %llu not supported yet, only support %lu",
3937 sectorsize, PAGE_SIZE);
319e4d06
QW		 3938 ret = -EINVAL;
3939 }
3940 if (!is_power_of_2(nodesize) || nodesize < sectorsize ||
3941 nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
ab8d0fc4 3942 btrfs_err(fs_info, "invalid nodesize %llu", nodesize);
319e4d06
QW		 3943 ret = -EINVAL;
3944 }
3945 if (nodesize != le32_to_cpu(sb->__unused_leafsize)) {
ab8d0fc4
JM		 3946 btrfs_err(fs_info, "invalid leafsize %u, should be %llu",
3947 le32_to_cpu(sb->__unused_leafsize), nodesize);
319e4d06
QW		 3948 ret = -EINVAL;
3949 }
3950
3951 /* Root alignment check */
3952 if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) {
ab8d0fc4
JM		 3953 btrfs_warn(fs_info, "tree_root block unaligned: %llu",
3954 btrfs_super_root(sb));
319e4d06
QW		 3955 ret = -EINVAL;
3956 }
3957 if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) {
ab8d0fc4
JM		 3958 btrfs_warn(fs_info, "chunk_root block unaligned: %llu",
3959 btrfs_super_chunk_root(sb));
75d6ad38
DS		 3960 ret = -EINVAL;
3961 }
319e4d06 3962 if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) {
ab8d0fc4
JM		 3963 btrfs_warn(fs_info, "log_root block unaligned: %llu",
3964 btrfs_super_log_root(sb));
75d6ad38
DS		 3965 ret = -EINVAL;
3966 }
3967
44880fdc 3968 if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_FSID_SIZE) != 0) {
ab8d0fc4
JM		 3969 btrfs_err(fs_info,
3970 "dev_item UUID does not match fsid: %pU != %pU",
3971 fs_info->fsid, sb->dev_item.fsid);
c926093e
DS		 3972 ret = -EINVAL;
3973 }
3974
3975 /*
3976 * Hint to catch really bogus numbers, bitflips or so, more exact checks are
3977 * done later
3978 */
99e3ecfc
LB		 3979 if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) {
3980 btrfs_err(fs_info, "bytes_used is too small %llu",
ab8d0fc4 3981 btrfs_super_bytes_used(sb));
99e3ecfc
LB		 3982 ret = -EINVAL;
3983 }
b7f67055 3984 if (!is_power_of_2(btrfs_super_stripesize(sb))) {
99e3ecfc 3985 btrfs_err(fs_info, "invalid stripesize %u",
ab8d0fc4 3986 btrfs_super_stripesize(sb));
99e3ecfc
LB		 3987 ret = -EINVAL;
3988 }
21e7626b 3989 if (btrfs_super_num_devices(sb) > (1UL << 31))
ab8d0fc4
JM		 3990 btrfs_warn(fs_info, "suspicious number of devices: %llu",
3991 btrfs_super_num_devices(sb));
75d6ad38 3992 if (btrfs_super_num_devices(sb) == 0) {
ab8d0fc4 3993 btrfs_err(fs_info, "number of devices is 0");
75d6ad38
DS		 3994 ret = -EINVAL;
3995 }
c926093e 3996
21e7626b 3997 if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) {
ab8d0fc4
JM		 3998 btrfs_err(fs_info, "super offset mismatch %llu != %u",
3999 btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET);
c926093e
DS		 4000 ret = -EINVAL;
4001 }
4002
ce7fca5f
DS		 4003 /*
4004 * Obvious sys_chunk_array corruptions, it must hold at least one key
4005 * and one chunk
4006 */
4007 if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) {
ab8d0fc4
JM		 4008 btrfs_err(fs_info, "system chunk array too big %u > %u",
4009 btrfs_super_sys_array_size(sb),
4010 BTRFS_SYSTEM_CHUNK_ARRAY_SIZE);
ce7fca5f
DS		 4011 ret = -EINVAL;
4012 }
4013 if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key)
4014 + sizeof(struct btrfs_chunk)) {
ab8d0fc4
JM		 4015 btrfs_err(fs_info, "system chunk array too small %u < %zu",
4016 btrfs_super_sys_array_size(sb),
4017 sizeof(struct btrfs_disk_key)
4018 + sizeof(struct btrfs_chunk));
ce7fca5f
DS		 4019 ret = -EINVAL;
4020 }
4021
c926093e
DS		 4022 /*
4023 * The generation is a global counter, we'll trust it more than the others
4024 * but it's still possible that it's the one that's wrong.
4025 */
21e7626b 4026 if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb))
ab8d0fc4
JM		 4027 btrfs_warn(fs_info,
4028 "suspicious: generation < chunk_root_generation: %llu < %llu",
4029 btrfs_super_generation(sb),
4030 btrfs_super_chunk_root_generation(sb));
21e7626b
DS		 4031 if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb)
4032 && btrfs_super_cache_generation(sb) != (u64)-1)
ab8d0fc4
JM		 4033 btrfs_warn(fs_info,
4034 "suspicious: generation < cache_generation: %llu < %llu",
4035 btrfs_super_generation(sb),
4036 btrfs_super_cache_generation(sb));
c926093e
DS		 4037
4038 return ret;
acce952b 4039}
4040
2ff7e61e 4041static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info)
acce952b 4042{
0b246afa 4043 mutex_lock(&fs_info->cleaner_mutex);
2ff7e61e 4044 btrfs_run_delayed_iputs(fs_info);
0b246afa 4045 mutex_unlock(&fs_info->cleaner_mutex);
acce952b 4046
0b246afa
JM		 4047 down_write(&fs_info->cleanup_work_sem);
4048 up_write(&fs_info->cleanup_work_sem);
acce952b 4049
4050 /* cleanup FS via transaction */
2ff7e61e 4051 btrfs_cleanup_transaction(fs_info);
acce952b 4052}
4053
143bede5 4054static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 4055{
acce952b 4056 struct btrfs_ordered_extent *ordered;
acce952b 4057
199c2a9c 4058 spin_lock(&root->ordered_extent_lock);
779880ef
JB		 4059 /*
4060 * This will just short circuit the ordered completion stuff which will
4061 * make sure the ordered extent gets properly cleaned up.
4062 */
199c2a9c 4063 list_for_each_entry(ordered, &root->ordered_extents,
779880ef
JB		 4064 root_extent_list)
4065 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
199c2a9c
MX		 4066 spin_unlock(&root->ordered_extent_lock);
4067}
4068
4069static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
4070{
4071 struct btrfs_root *root;
4072 struct list_head splice;
4073
4074 INIT_LIST_HEAD(&splice);
4075
4076 spin_lock(&fs_info->ordered_root_lock);
4077 list_splice_init(&fs_info->ordered_roots, &splice);
4078 while (!list_empty(&splice)) {
4079 root = list_first_entry(&splice, struct btrfs_root,
4080 ordered_root);
1de2cfde
JB		 4081 list_move_tail(&root->ordered_root,
4082 &fs_info->ordered_roots);
199c2a9c 4083
2a85d9ca 4084 spin_unlock(&fs_info->ordered_root_lock);
199c2a9c
MX		 4085 btrfs_destroy_ordered_extents(root);
4086
2a85d9ca
LB		 4087 cond_resched();
4088 spin_lock(&fs_info->ordered_root_lock);
199c2a9c
MX		 4089 }
4090 spin_unlock(&fs_info->ordered_root_lock);
acce952b 4091}
4092
35a3621b 4093static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
2ff7e61e 4094 struct btrfs_fs_info *fs_info)
acce952b 4095{
4096 struct rb_node *node;
4097 struct btrfs_delayed_ref_root *delayed_refs;
4098 struct btrfs_delayed_ref_node *ref;
4099 int ret = 0;
4100
4101 delayed_refs = &trans->delayed_refs;
4102
4103 spin_lock(&delayed_refs->lock);
d7df2c79 4104 if (atomic_read(&delayed_refs->num_entries) == 0) {
cfece4db 4105 spin_unlock(&delayed_refs->lock);
0b246afa 4106 btrfs_info(fs_info, "delayed_refs has NO entry");
acce952b 4107 return ret;
4108 }
4109
d7df2c79
JB		 4110 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
4111 struct btrfs_delayed_ref_head *head;
c6fc2454 4112 struct btrfs_delayed_ref_node *tmp;
e78417d1 4113 bool pin_bytes = false;
acce952b 4114
d7df2c79
JB		 4115 head = rb_entry(node, struct btrfs_delayed_ref_head,
4116 href_node);
4117 if (!mutex_trylock(&head->mutex)) {
6df8cdf5 4118 refcount_inc(&head->node.refs);
d7df2c79 4119 spin_unlock(&delayed_refs->lock);
eb12db69 4120
d7df2c79 4121 mutex_lock(&head->mutex);
e78417d1 4122 mutex_unlock(&head->mutex);
d7df2c79
JB		 4123 btrfs_put_delayed_ref(&head->node);
4124 spin_lock(&delayed_refs->lock);
4125 continue;
4126 }
4127 spin_lock(&head->lock);
c6fc2454
QW		 4128 list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list,
4129 list) {
d7df2c79 4130 ref->in_tree = 0;
c6fc2454 4131 list_del(&ref->list);
1d57ee94
WX		 4132 if (!list_empty(&ref->add_list))
4133 list_del(&ref->add_list);
d7df2c79
JB		 4134 atomic_dec(&delayed_refs->num_entries);
4135 btrfs_put_delayed_ref(ref);
e78417d1 4136 }
d7df2c79
JB		 4137 if (head->must_insert_reserved)
4138 pin_bytes = true;
4139 btrfs_free_delayed_extent_op(head->extent_op);
4140 delayed_refs->num_heads--;
4141 if (head->processing == 0)
4142 delayed_refs->num_heads_ready--;
4143 atomic_dec(&delayed_refs->num_entries);
4144 head->node.in_tree = 0;
4145 rb_erase(&head->href_node, &delayed_refs->href_root);
4146 spin_unlock(&head->lock);
4147 spin_unlock(&delayed_refs->lock);
4148 mutex_unlock(&head->mutex);
acce952b 4149
d7df2c79 4150 if (pin_bytes)
2ff7e61e 4151 btrfs_pin_extent(fs_info, head->node.bytenr,
d7df2c79
JB		 4152 head->node.num_bytes, 1);
4153 btrfs_put_delayed_ref(&head->node);
acce952b 4154 cond_resched();
4155 spin_lock(&delayed_refs->lock);
4156 }
4157
4158 spin_unlock(&delayed_refs->lock);
4159
4160 return ret;
4161}
4162
143bede5 4163static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 4164{
4165 struct btrfs_inode *btrfs_inode;
4166 struct list_head splice;
4167
4168 INIT_LIST_HEAD(&splice);
4169
eb73c1b7
MX		 4170 spin_lock(&root->delalloc_lock);
4171 list_splice_init(&root->delalloc_inodes, &splice);
acce952b 4172
4173 while (!list_empty(&splice)) {
eb73c1b7
MX		 4174 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
4175 delalloc_inodes);
acce952b 4176
4177 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX		 4178 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
4179 &btrfs_inode->runtime_flags);
eb73c1b7 4180 spin_unlock(&root->delalloc_lock);
acce952b 4181
4182 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 4183
eb73c1b7 4184 spin_lock(&root->delalloc_lock);
acce952b 4185 }
4186
eb73c1b7
MX		 4187 spin_unlock(&root->delalloc_lock);
4188}
4189
4190static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
4191{
4192 struct btrfs_root *root;
4193 struct list_head splice;
4194
4195 INIT_LIST_HEAD(&splice);
4196
4197 spin_lock(&fs_info->delalloc_root_lock);
4198 list_splice_init(&fs_info->delalloc_roots, &splice);
4199 while (!list_empty(&splice)) {
4200 root = list_first_entry(&splice, struct btrfs_root,
4201 delalloc_root);
4202 list_del_init(&root->delalloc_root);
4203 root = btrfs_grab_fs_root(root);
4204 BUG_ON(!root);
4205 spin_unlock(&fs_info->delalloc_root_lock);
4206
4207 btrfs_destroy_delalloc_inodes(root);
4208 btrfs_put_fs_root(root);
4209
4210 spin_lock(&fs_info->delalloc_root_lock);
4211 }
4212 spin_unlock(&fs_info->delalloc_root_lock);
acce952b 4213}
4214
2ff7e61e 4215static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info,
acce952b 4216 struct extent_io_tree *dirty_pages,
4217 int mark)
4218{
4219 int ret;
acce952b 4220 struct extent_buffer *eb;
4221 u64 start = 0;
4222 u64 end;
acce952b 4223
4224 while (1) {
4225 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 4226 mark, NULL);
acce952b 4227 if (ret)
4228 break;
4229
91166212 4230 clear_extent_bits(dirty_pages, start, end, mark);
acce952b 4231 while (start <= end) {
0b246afa
JM		 4232 eb = find_extent_buffer(fs_info, start);
4233 start += fs_info->nodesize;
fd8b2b61 4234 if (!eb)
acce952b 4235 continue;
fd8b2b61 4236 wait_on_extent_buffer_writeback(eb);
acce952b 4237
fd8b2b61
JB		 4238 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4239 &eb->bflags))
4240 clear_extent_buffer_dirty(eb);
4241 free_extent_buffer_stale(eb);
acce952b 4242 }
4243 }
4244
4245 return ret;
4246}
4247
2ff7e61e 4248static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info,
acce952b 4249 struct extent_io_tree *pinned_extents)
4250{
4251 struct extent_io_tree *unpin;
4252 u64 start;
4253 u64 end;
4254 int ret;
ed0eaa14 4255 bool loop = true;
acce952b 4256
4257 unpin = pinned_extents;
ed0eaa14 4258again:
acce952b 4259 while (1) {
4260 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 4261 EXTENT_DIRTY, NULL);
acce952b 4262 if (ret)
4263 break;
4264
af6f8f60 4265 clear_extent_dirty(unpin, start, end);
2ff7e61e 4266 btrfs_error_unpin_extent_range(fs_info, start, end);
acce952b 4267 cond_resched();
4268 }
4269
ed0eaa14 4270 if (loop) {
0b246afa
JM		 4271 if (unpin == &fs_info->freed_extents[0])
4272 unpin = &fs_info->freed_extents[1];
ed0eaa14 4273 else
0b246afa 4274 unpin = &fs_info->freed_extents[0];
ed0eaa14
LB		 4275 loop = false;
4276 goto again;
4277 }
4278
acce952b 4279 return 0;
4280}
4281
c79a1751
LB		 4282static void btrfs_cleanup_bg_io(struct btrfs_block_group_cache *cache)
4283{
4284 struct inode *inode;
4285
4286 inode = cache->io_ctl.inode;
4287 if (inode) {
4288 invalidate_inode_pages2(inode->i_mapping);
4289 BTRFS_I(inode)->generation = 0;
4290 cache->io_ctl.inode = NULL;
4291 iput(inode);
4292 }
4293 btrfs_put_block_group(cache);
4294}
4295
4296void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans,
2ff7e61e 4297 struct btrfs_fs_info *fs_info)
c79a1751
LB		 4298{
4299 struct btrfs_block_group_cache *cache;
4300
4301 spin_lock(&cur_trans->dirty_bgs_lock);
4302 while (!list_empty(&cur_trans->dirty_bgs)) {
4303 cache = list_first_entry(&cur_trans->dirty_bgs,
4304 struct btrfs_block_group_cache,
4305 dirty_list);
4306 if (!cache) {
0b246afa 4307 btrfs_err(fs_info, "orphan block group dirty_bgs list");
c79a1751
LB		 4308 spin_unlock(&cur_trans->dirty_bgs_lock);
4309 return;
4310 }
4311
4312 if (!list_empty(&cache->io_list)) {
4313 spin_unlock(&cur_trans->dirty_bgs_lock);
4314 list_del_init(&cache->io_list);
4315 btrfs_cleanup_bg_io(cache);
4316 spin_lock(&cur_trans->dirty_bgs_lock);
4317 }
4318
4319 list_del_init(&cache->dirty_list);
4320 spin_lock(&cache->lock);
4321 cache->disk_cache_state = BTRFS_DC_ERROR;
4322 spin_unlock(&cache->lock);
4323
4324 spin_unlock(&cur_trans->dirty_bgs_lock);
4325 btrfs_put_block_group(cache);
4326 spin_lock(&cur_trans->dirty_bgs_lock);
4327 }
4328 spin_unlock(&cur_trans->dirty_bgs_lock);
4329
4330 while (!list_empty(&cur_trans->io_bgs)) {
4331 cache = list_first_entry(&cur_trans->io_bgs,
4332 struct btrfs_block_group_cache,
4333 io_list);
4334 if (!cache) {
0b246afa 4335 btrfs_err(fs_info, "orphan block group on io_bgs list");
c79a1751
LB		 4336 return;
4337 }
4338
4339 list_del_init(&cache->io_list);
4340 spin_lock(&cache->lock);
4341 cache->disk_cache_state = BTRFS_DC_ERROR;
4342 spin_unlock(&cache->lock);
4343 btrfs_cleanup_bg_io(cache);
4344 }
4345}
4346
49b25e05 4347void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
2ff7e61e 4348 struct btrfs_fs_info *fs_info)
49b25e05 4349{
2ff7e61e 4350 btrfs_cleanup_dirty_bgs(cur_trans, fs_info);
c79a1751
LB		 4351 ASSERT(list_empty(&cur_trans->dirty_bgs));
4352 ASSERT(list_empty(&cur_trans->io_bgs));
4353
2ff7e61e 4354 btrfs_destroy_delayed_refs(cur_trans, fs_info);
49b25e05 4355
4a9d8bde 4356 cur_trans->state = TRANS_STATE_COMMIT_START;
0b246afa 4357 wake_up(&fs_info->transaction_blocked_wait);
49b25e05 4358
4a9d8bde 4359 cur_trans->state = TRANS_STATE_UNBLOCKED;
0b246afa 4360 wake_up(&fs_info->transaction_wait);
49b25e05 4361
ccdf9b30
JM		 4362 btrfs_destroy_delayed_inodes(fs_info);
4363 btrfs_assert_delayed_root_empty(fs_info);
49b25e05 4364
2ff7e61e 4365 btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages,
49b25e05 4366 EXTENT_DIRTY);
2ff7e61e 4367 btrfs_destroy_pinned_extent(fs_info,
0b246afa 4368 fs_info->pinned_extents);
49b25e05 4369
4a9d8bde
MX		 4370 cur_trans->state =TRANS_STATE_COMPLETED;
4371 wake_up(&cur_trans->commit_wait);
49b25e05
JM		 4372}
4373
2ff7e61e 4374static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info)
acce952b 4375{
4376 struct btrfs_transaction *t;
acce952b 4377
0b246afa 4378 mutex_lock(&fs_info->transaction_kthread_mutex);
acce952b 4379
0b246afa
JM		 4380 spin_lock(&fs_info->trans_lock);
4381 while (!list_empty(&fs_info->trans_list)) {
4382 t = list_first_entry(&fs_info->trans_list,
724e2315
JB		 4383 struct btrfs_transaction, list);
4384 if (t->state >= TRANS_STATE_COMMIT_START) {
9b64f57d 4385 refcount_inc(&t->use_count);
0b246afa 4386 spin_unlock(&fs_info->trans_lock);
2ff7e61e 4387 btrfs_wait_for_commit(fs_info, t->transid);
724e2315 4388 btrfs_put_transaction(t);
0b246afa 4389 spin_lock(&fs_info->trans_lock);
724e2315
JB		 4390 continue;
4391 }
0b246afa 4392 if (t == fs_info->running_transaction) {
724e2315 4393 t->state = TRANS_STATE_COMMIT_DOING;
0b246afa 4394 spin_unlock(&fs_info->trans_lock);
724e2315
JB		 4395 /*
4396 * We wait for 0 num_writers since we don't hold a trans
4397 * handle open currently for this transaction.
4398 */
4399 wait_event(t->writer_wait,
4400 atomic_read(&t->num_writers) == 0);
4401 } else {
0b246afa 4402 spin_unlock(&fs_info->trans_lock);
724e2315 4403 }
2ff7e61e 4404 btrfs_cleanup_one_transaction(t, fs_info);
4a9d8bde 4405
0b246afa
JM		 4406 spin_lock(&fs_info->trans_lock);
4407 if (t == fs_info->running_transaction)
4408 fs_info->running_transaction = NULL;
acce952b 4409 list_del_init(&t->list);
0b246afa 4410 spin_unlock(&fs_info->trans_lock);
acce952b 4411
724e2315 4412 btrfs_put_transaction(t);
2ff7e61e 4413 trace_btrfs_transaction_commit(fs_info->tree_root);
0b246afa 4414 spin_lock(&fs_info->trans_lock);
724e2315 4415 }
0b246afa
JM		 4416 spin_unlock(&fs_info->trans_lock);
4417 btrfs_destroy_all_ordered_extents(fs_info);
ccdf9b30
JM		 4418 btrfs_destroy_delayed_inodes(fs_info);
4419 btrfs_assert_delayed_root_empty(fs_info);
2ff7e61e 4420 btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents);
0b246afa
JM		 4421 btrfs_destroy_all_delalloc_inodes(fs_info);
4422 mutex_unlock(&fs_info->transaction_kthread_mutex);
acce952b 4423
4424 return 0;
4425}
4426
c6100a4b
JB		 4427static struct btrfs_fs_info *btree_fs_info(void *private_data)
4428{
4429 struct inode *inode = private_data;
4430 return btrfs_sb(inode->i_sb);
4431}
4432
e8c9f186 4433static const struct extent_io_ops btree_extent_io_ops = {
4d53dddb 4434 /* mandatory callbacks */
0b86a832 4435 .submit_bio_hook = btree_submit_bio_hook,
4d53dddb 4436 .readpage_end_io_hook = btree_readpage_end_io_hook,
239b14b3
CM		 4437 /* note we're sharing with inode.c for the merge bio hook */
4438 .merge_bio_hook = btrfs_merge_bio_hook,
20a7db8a 4439 .readpage_io_failed_hook = btree_io_failed_hook,
c6100a4b
JB		 4440 .set_range_writeback = btrfs_set_range_writeback,
4441 .tree_fs_info = btree_fs_info,
4d53dddb
DS		 4442
4443 /* optional callbacks */
0da5468f 4444};
Linux 6.x block layer and io_uring tree(s)