projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: factor btrfs_init_balance() out of open_ctree()
[linux-2.6-block.git] / fs / btrfs / disk-io.c
CommitLineData
6cbd5570
CM		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
e20d96d6 19#include <linux/fs.h>
d98237b3 20#include <linux/blkdev.h>
87cbda5c 21#include <linux/scatterlist.h>
22b0ebda 22#include <linux/swap.h>
0f7d52f4 23#include <linux/radix-tree.h>
35b7e476 24#include <linux/writeback.h>
d397712b 25#include <linux/buffer_head.h>
ce9adaa5 26#include <linux/workqueue.h>
a74a4b97 27#include <linux/kthread.h>
4b4e25f2 28#include <linux/freezer.h>
5a0e3ad6 29#include <linux/slab.h>
784b4e29 30#include <linux/migrate.h>
7a36ddec 31#include <linux/ratelimit.h>
6463fe58 32#include <linux/uuid.h>
803b2f54 33#include <linux/semaphore.h>
7e75bf3f 34#include <asm/unaligned.h>
eb60ceac
CM		 35#include "ctree.h"
36#include "disk-io.h"
0b947aff 37#include "hash.h"
e089f05c 38#include "transaction.h"
0f7d52f4 39#include "btrfs_inode.h"
0b86a832 40#include "volumes.h"
db94535d 41#include "print-tree.h"
925baedd 42#include "locking.h"
e02119d5 43#include "tree-log.h"
fa9c0d79 44#include "free-space-cache.h"
581bb050 45#include "inode-map.h"
21adbd5c 46#include "check-integrity.h"
606686ee 47#include "rcu-string.h"
8dabb742 48#include "dev-replace.h"
53b381b3 49#include "raid56.h"
5ac1d209 50#include "sysfs.h"
fcebe456 51#include "qgroup.h"
eb60ceac 52
de0022b9
JB		 53#ifdef CONFIG_X86
54#include <asm/cpufeature.h>
55#endif
56
e8c9f186 57static const struct extent_io_ops btree_extent_io_ops;
8b712842 58static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 59static void free_fs_root(struct btrfs_root *root);
fcd1f065 60static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 61 int read_only);
143bede5 62static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
acce952b 63static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
64 struct btrfs_root *root);
143bede5 65static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
acce952b 66static int btrfs_destroy_marked_extents(struct btrfs_root *root,
67 struct extent_io_tree *dirty_pages,
68 int mark);
69static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
70 struct extent_io_tree *pinned_extents);
48a3b636
ES		 71static int btrfs_cleanup_transaction(struct btrfs_root *root);
72static void btrfs_error_commit_super(struct btrfs_root *root);
ce9adaa5 73
d352ac68 74/*
97eb6b69
DS		 75 * btrfs_end_io_wq structs are used to do processing in task context when an IO
76 * is complete. This is used during reads to verify checksums, and it is used
d352ac68
CM		 77 * by writes to insert metadata for new file extents after IO is complete.
78 */
97eb6b69 79struct btrfs_end_io_wq {
ce9adaa5
CM		 80 struct bio *bio;
81 bio_end_io_t *end_io;
82 void *private;
83 struct btrfs_fs_info *info;
84 int error;
bfebd8b5 85 enum btrfs_wq_endio_type metadata;
ce9adaa5 86 struct list_head list;
8b712842 87 struct btrfs_work work;
ce9adaa5 88};
0da5468f 89
97eb6b69
DS		 90static struct kmem_cache *btrfs_end_io_wq_cache;
91
92int __init btrfs_end_io_wq_init(void)
93{
94 btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq",
95 sizeof(struct btrfs_end_io_wq),
96 0,
97 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
98 NULL);
99 if (!btrfs_end_io_wq_cache)
100 return -ENOMEM;
101 return 0;
102}
103
104void btrfs_end_io_wq_exit(void)
105{
106 if (btrfs_end_io_wq_cache)
107 kmem_cache_destroy(btrfs_end_io_wq_cache);
108}
109
d352ac68
CM		 110/*
111 * async submit bios are used to offload expensive checksumming
112 * onto the worker threads. They checksum file and metadata bios
113 * just before they are sent down the IO stack.
114 */
44b8bd7e
CM		 115struct async_submit_bio {
116 struct inode *inode;
117 struct bio *bio;
118 struct list_head list;
4a69a410
CM		 119 extent_submit_bio_hook_t *submit_bio_start;
120 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM		 121 int rw;
122 int mirror_num;
c8b97818 123 unsigned long bio_flags;
eaf25d93
CM		 124 /*
125 * bio_offset is optional, can be used if the pages in the bio
126 * can't tell us where in the file the bio should go
127 */
128 u64 bio_offset;
8b712842 129 struct btrfs_work work;
79787eaa 130 int error;
44b8bd7e
CM		 131};
132
85d4e461
CM		 133/*
134 * Lockdep class keys for extent_buffer->lock's in this root. For a given
135 * eb, the lockdep key is determined by the btrfs_root it belongs to and
136 * the level the eb occupies in the tree.
137 *
138 * Different roots are used for different purposes and may nest inside each
139 * other and they require separate keysets. As lockdep keys should be
140 * static, assign keysets according to the purpose of the root as indicated
141 * by btrfs_root->objectid. This ensures that all special purpose roots
142 * have separate keysets.
4008c04a 143 *
85d4e461
CM		 144 * Lock-nesting across peer nodes is always done with the immediate parent
145 * node locked thus preventing deadlock. As lockdep doesn't know this, use
146 * subclass to avoid triggering lockdep warning in such cases.
4008c04a 147 *
85d4e461
CM		 148 * The key is set by the readpage_end_io_hook after the buffer has passed
149 * csum validation but before the pages are unlocked. It is also set by
150 * btrfs_init_new_buffer on freshly allocated blocks.
4008c04a 151 *
85d4e461
CM		 152 * We also add a check to make sure the highest level of the tree is the
153 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
154 * needs update as well.
4008c04a
CM		 155 */
156#ifdef CONFIG_DEBUG_LOCK_ALLOC
157# if BTRFS_MAX_LEVEL != 8
158# error
159# endif
85d4e461
CM		 160
161static struct btrfs_lockdep_keyset {
162 u64 id; /* root objectid */
163 const char *name_stem; /* lock name stem */
164 char names[BTRFS_MAX_LEVEL + 1][20];
165 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
166} btrfs_lockdep_keysets[] = {
167 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
168 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
169 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
170 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
171 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
172 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
60b62978 173 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
85d4e461
CM		 174 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
175 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
176 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
13fd8da9 177 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
85d4e461 178 { .id = 0, .name_stem = "tree" },
4008c04a 179};
85d4e461
CM		 180
181void __init btrfs_init_lockdep(void)
182{
183 int i, j;
184
185 /* initialize lockdep class names */
186 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
187 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
188
189 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
190 snprintf(ks->names[j], sizeof(ks->names[j]),
191 "btrfs-%s-%02d", ks->name_stem, j);
192 }
193}
194
195void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
196 int level)
197{
198 struct btrfs_lockdep_keyset *ks;
199
200 BUG_ON(level >= ARRAY_SIZE(ks->keys));
201
202 /* find the matching keyset, id 0 is the default entry */
203 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
204 if (ks->id == objectid)
205 break;
206
207 lockdep_set_class_and_name(&eb->lock,
208 &ks->keys[level], ks->names[level]);
209}
210
4008c04a
CM		 211#endif
212
d352ac68
CM		 213/*
214 * extents on the btree inode are pretty simple, there's one extent
215 * that covers the entire device
216 */
b2950863 217static struct extent_map *btree_get_extent(struct inode *inode,
306e16ce 218 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 219 int create)
7eccb903 220{
5f39d397
CM		 221 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
222 struct extent_map *em;
223 int ret;
224
890871be 225 read_lock(&em_tree->lock);
d1310b2e 226 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 227 if (em) {
228 em->bdev =
229 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 230 read_unlock(&em_tree->lock);
5f39d397 231 goto out;
a061fc8d 232 }
890871be 233 read_unlock(&em_tree->lock);
7b13b7b1 234
172ddd60 235 em = alloc_extent_map();
5f39d397
CM		 236 if (!em) {
237 em = ERR_PTR(-ENOMEM);
238 goto out;
239 }
240 em->start = 0;
0afbaf8c 241 em->len = (u64)-1;
c8b97818 242 em->block_len = (u64)-1;
5f39d397 243 em->block_start = 0;
a061fc8d 244 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 245
890871be 246 write_lock(&em_tree->lock);
09a2a8f9 247 ret = add_extent_mapping(em_tree, em, 0);
5f39d397
CM		 248 if (ret == -EEXIST) {
249 free_extent_map(em);
7b13b7b1 250 em = lookup_extent_mapping(em_tree, start, len);
b4f359ab 251 if (!em)
0433f20d 252 em = ERR_PTR(-EIO);
5f39d397 253 } else if (ret) {
7b13b7b1 254 free_extent_map(em);
0433f20d 255 em = ERR_PTR(ret);
5f39d397 256 }
890871be 257 write_unlock(&em_tree->lock);
7b13b7b1 258
5f39d397
CM		 259out:
260 return em;
7eccb903
CM		 261}
262
b0496686 263u32 btrfs_csum_data(char *data, u32 seed, size_t len)
19c00ddc 264{
0b947aff 265 return btrfs_crc32c(seed, data, len);
19c00ddc
CM		 266}
267
268void btrfs_csum_final(u32 crc, char *result)
269{
7e75bf3f 270 put_unaligned_le32(~crc, result);
19c00ddc
CM		 271}
272
d352ac68
CM		 273/*
274 * compute the csum for a btree block, and either verify it or write it
275 * into the csum field of the block.
276 */
01d58472
DD		 277static int csum_tree_block(struct btrfs_fs_info *fs_info,
278 struct extent_buffer *buf,
19c00ddc
CM		 279 int verify)
280{
01d58472 281 u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
607d432d 282 char *result = NULL;
19c00ddc
CM		 283 unsigned long len;
284 unsigned long cur_len;
285 unsigned long offset = BTRFS_CSUM_SIZE;
19c00ddc
CM		 286 char *kaddr;
287 unsigned long map_start;
288 unsigned long map_len;
289 int err;
290 u32 crc = ~(u32)0;
607d432d 291 unsigned long inline_result;
19c00ddc
CM		 292
293 len = buf->len - offset;
d397712b 294 while (len > 0) {
19c00ddc 295 err = map_private_extent_buffer(buf, offset, 32,
a6591715 296 &kaddr, &map_start, &map_len);
d397712b 297 if (err)
19c00ddc 298 return 1;
19c00ddc 299 cur_len = min(len, map_len - (offset - map_start));
b0496686 300 crc = btrfs_csum_data(kaddr + offset - map_start,
19c00ddc
CM		 301 crc, cur_len);
302 len -= cur_len;
303 offset += cur_len;
19c00ddc 304 }
607d432d
JB		 305 if (csum_size > sizeof(inline_result)) {
306 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
307 if (!result)
308 return 1;
309 } else {
310 result = (char *)&inline_result;
311 }
312
19c00ddc
CM		 313 btrfs_csum_final(crc, result);
314
315 if (verify) {
607d432d 316 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 317 u32 val;
318 u32 found = 0;
607d432d 319 memcpy(&found, result, csum_size);
e4204ded 320
607d432d 321 read_extent_buffer(buf, &val, 0, csum_size);
f0954c66 322 printk_ratelimited(KERN_WARNING
efe120a0
FH		 323 "BTRFS: %s checksum verify failed on %llu wanted %X found %X "
324 "level %d\n",
01d58472 325 fs_info->sb->s_id, buf->start,
efe120a0 326 val, found, btrfs_header_level(buf));
607d432d
JB		 327 if (result != (char *)&inline_result)
328 kfree(result);
19c00ddc
CM		 329 return 1;
330 }
331 } else {
607d432d 332 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 333 }
607d432d
JB		 334 if (result != (char *)&inline_result)
335 kfree(result);
19c00ddc
CM		 336 return 0;
337}
338
d352ac68
CM		 339/*
340 * we can't consider a given block up to date unless the transid of the
341 * block matches the transid in the parent node's pointer. This is how we
342 * detect blocks that either didn't get written at all or got written
343 * in the wrong place.
344 */
1259ab75 345static int verify_parent_transid(struct extent_io_tree *io_tree,
b9fab919
CM		 346 struct extent_buffer *eb, u64 parent_transid,
347 int atomic)
1259ab75 348{
2ac55d41 349 struct extent_state *cached_state = NULL;
1259ab75 350 int ret;
2755a0de 351 bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB);
1259ab75
CM		 352
353 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
354 return 0;
355
b9fab919
CM		 356 if (atomic)
357 return -EAGAIN;
358
a26e8c9f
JB		 359 if (need_lock) {
360 btrfs_tree_read_lock(eb);
361 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
362 }
363
2ac55d41 364 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
d0082371 365 0, &cached_state);
0b32f4bb 366 if (extent_buffer_uptodate(eb) &&
1259ab75
CM		 367 btrfs_header_generation(eb) == parent_transid) {
368 ret = 0;
369 goto out;
370 }
68b663d1
DS		 371 printk_ratelimited(KERN_ERR
372 "BTRFS (device %s): parent transid verify failed on %llu wanted %llu found %llu\n",
29549aec
WS		 373 eb->fs_info->sb->s_id, eb->start,
374 parent_transid, btrfs_header_generation(eb));
1259ab75 375 ret = 1;
a26e8c9f
JB		 376
377 /*
378 * Things reading via commit roots that don't have normal protection,
379 * like send, can have a really old block in cache that may point at a
380 * block that has been free'd and re-allocated. So don't clear uptodate
381 * if we find an eb that is under IO (dirty/writeback) because we could
382 * end up reading in the stale data and then writing it back out and
383 * making everybody very sad.
384 */
385 if (!extent_buffer_under_io(eb))
386 clear_extent_buffer_uptodate(eb);
33958dc6 387out:
2ac55d41
JB		 388 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
389 &cached_state, GFP_NOFS);
472b909f
JB		 390 if (need_lock)
391 btrfs_tree_read_unlock_blocking(eb);
1259ab75 392 return ret;
1259ab75
CM		 393}
394
1104a885
DS		 395/*
396 * Return 0 if the superblock checksum type matches the checksum value of that
397 * algorithm. Pass the raw disk superblock data.
398 */
399static int btrfs_check_super_csum(char *raw_disk_sb)
400{
401 struct btrfs_super_block *disk_sb =
402 (struct btrfs_super_block *)raw_disk_sb;
403 u16 csum_type = btrfs_super_csum_type(disk_sb);
404 int ret = 0;
405
406 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
407 u32 crc = ~(u32)0;
408 const int csum_size = sizeof(crc);
409 char result[csum_size];
410
411 /*
412 * The super_block structure does not span the whole
413 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
414 * is filled with zeros and is included in the checkum.
415 */
416 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
417 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
418 btrfs_csum_final(crc, result);
419
420 if (memcmp(raw_disk_sb, result, csum_size))
421 ret = 1;
667e7d94
CM		 422
423 if (ret && btrfs_super_generation(disk_sb) < 10) {
efe120a0
FH		 424 printk(KERN_WARNING
425 "BTRFS: super block crcs don't match, older mkfs detected\n");
667e7d94
CM		 426 ret = 0;
427 }
1104a885
DS		 428 }
429
430 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
efe120a0 431 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
1104a885
DS		 432 csum_type);
433 ret = 1;
434 }
435
436 return ret;
437}
438
d352ac68
CM		 439/*
440 * helper to read a given tree block, doing retries as required when
441 * the checksums don't match and we have alternate mirrors to try.
442 */
f188591e
CM		 443static int btree_read_extent_buffer_pages(struct btrfs_root *root,
444 struct extent_buffer *eb,
ca7a79ad 445 u64 start, u64 parent_transid)
f188591e
CM		 446{
447 struct extent_io_tree *io_tree;
ea466794 448 int failed = 0;
f188591e
CM		 449 int ret;
450 int num_copies = 0;
451 int mirror_num = 0;
ea466794 452 int failed_mirror = 0;
f188591e 453
a826d6dc 454 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
f188591e
CM		 455 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
456 while (1) {
bb82ab88
AJ		 457 ret = read_extent_buffer_pages(io_tree, eb, start,
458 WAIT_COMPLETE,
f188591e 459 btree_get_extent, mirror_num);
256dd1bb
SB		 460 if (!ret) {
461 if (!verify_parent_transid(io_tree, eb,
b9fab919 462 parent_transid, 0))
256dd1bb
SB		 463 break;
464 else
465 ret = -EIO;
466 }
d397712b 467
a826d6dc
JB		 468 /*
469 * This buffer's crc is fine, but its contents are corrupted, so
470 * there is no reason to read the other copies, they won't be
471 * any less wrong.
472 */
473 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
ea466794
JB		 474 break;
475
5d964051 476 num_copies = btrfs_num_copies(root->fs_info,
f188591e 477 eb->start, eb->len);
4235298e 478 if (num_copies == 1)
ea466794 479 break;
4235298e 480
5cf1ab56
JB		 481 if (!failed_mirror) {
482 failed = 1;
483 failed_mirror = eb->read_mirror;
484 }
485
f188591e 486 mirror_num++;
ea466794
JB		 487 if (mirror_num == failed_mirror)
488 mirror_num++;
489
4235298e 490 if (mirror_num > num_copies)
ea466794 491 break;
f188591e 492 }
ea466794 493
c0901581 494 if (failed && !ret && failed_mirror)
ea466794
JB		 495 repair_eb_io_failure(root, eb, failed_mirror);
496
497 return ret;
f188591e 498}
19c00ddc 499
d352ac68 500/*
d397712b
CM		 501 * checksum a dirty tree block before IO. This has extra checks to make sure
502 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 503 */
d397712b 504
01d58472 505static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page)
19c00ddc 506{
4eee4fa4 507 u64 start = page_offset(page);
19c00ddc 508 u64 found_start;
19c00ddc 509 struct extent_buffer *eb;
f188591e 510
4f2de97a
JB		 511 eb = (struct extent_buffer *)page->private;
512 if (page != eb->pages[0])
513 return 0;
19c00ddc 514 found_start = btrfs_header_bytenr(eb);
fae7f21c 515 if (WARN_ON(found_start != start || !PageUptodate(page)))
4f2de97a 516 return 0;
01d58472 517 csum_tree_block(fs_info, eb, 0);
19c00ddc
CM		 518 return 0;
519}
520
01d58472 521static int check_tree_block_fsid(struct btrfs_fs_info *fs_info,
2b82032c
YZ		 522 struct extent_buffer *eb)
523{
01d58472 524 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2b82032c
YZ		 525 u8 fsid[BTRFS_UUID_SIZE];
526 int ret = 1;
527
0a4e5586 528 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
2b82032c
YZ		 529 while (fs_devices) {
530 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
531 ret = 0;
532 break;
533 }
534 fs_devices = fs_devices->seed;
535 }
536 return ret;
537}
538
a826d6dc 539#define CORRUPT(reason, eb, root, slot) \
efe120a0
FH		 540 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
541 "root=%llu, slot=%d", reason, \
c1c9ff7c 542 btrfs_header_bytenr(eb), root->objectid, slot)
a826d6dc
JB		 543
544static noinline int check_leaf(struct btrfs_root *root,
545 struct extent_buffer *leaf)
546{
547 struct btrfs_key key;
548 struct btrfs_key leaf_key;
549 u32 nritems = btrfs_header_nritems(leaf);
550 int slot;
551
552 if (nritems == 0)
553 return 0;
554
555 /* Check the 0 item */
556 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
557 BTRFS_LEAF_DATA_SIZE(root)) {
558 CORRUPT("invalid item offset size pair", leaf, root, 0);
559 return -EIO;
560 }
561
562 /*
563 * Check to make sure each items keys are in the correct order and their
564 * offsets make sense. We only have to loop through nritems-1 because
565 * we check the current slot against the next slot, which verifies the
566 * next slot's offset+size makes sense and that the current's slot
567 * offset is correct.
568 */
569 for (slot = 0; slot < nritems - 1; slot++) {
570 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
571 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
572
573 /* Make sure the keys are in the right order */
574 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
575 CORRUPT("bad key order", leaf, root, slot);
576 return -EIO;
577 }
578
579 /*
580 * Make sure the offset and ends are right, remember that the
581 * item data starts at the end of the leaf and grows towards the
582 * front.
583 */
584 if (btrfs_item_offset_nr(leaf, slot) !=
585 btrfs_item_end_nr(leaf, slot + 1)) {
586 CORRUPT("slot offset bad", leaf, root, slot);
587 return -EIO;
588 }
589
590 /*
591 * Check to make sure that we don't point outside of the leaf,
592 * just incase all the items are consistent to eachother, but
593 * all point outside of the leaf.
594 */
595 if (btrfs_item_end_nr(leaf, slot) >
596 BTRFS_LEAF_DATA_SIZE(root)) {
597 CORRUPT("slot end outside of leaf", leaf, root, slot);
598 return -EIO;
599 }
600 }
601
602 return 0;
603}
604
facc8a22
MX		 605static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
606 u64 phy_offset, struct page *page,
607 u64 start, u64 end, int mirror)
ce9adaa5 608{
ce9adaa5
CM		 609 u64 found_start;
610 int found_level;
ce9adaa5
CM		 611 struct extent_buffer *eb;
612 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 613 int ret = 0;
727011e0 614 int reads_done;
ce9adaa5 615
ce9adaa5
CM		 616 if (!page->private)
617 goto out;
d397712b 618
4f2de97a 619 eb = (struct extent_buffer *)page->private;
d397712b 620
0b32f4bb
JB		 621 /* the pending IO might have been the only thing that kept this buffer
622 * in memory. Make sure we have a ref for all this other checks
623 */
624 extent_buffer_get(eb);
625
626 reads_done = atomic_dec_and_test(&eb->io_pages);
727011e0
CM		 627 if (!reads_done)
628 goto err;
f188591e 629
5cf1ab56 630 eb->read_mirror = mirror;
656f30db 631 if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) {
ea466794
JB		 632 ret = -EIO;
633 goto err;
634 }
635
ce9adaa5 636 found_start = btrfs_header_bytenr(eb);
727011e0 637 if (found_start != eb->start) {
68b663d1 638 printk_ratelimited(KERN_ERR "BTRFS (device %s): bad tree block start "
193f284d 639 "%llu %llu\n",
29549aec 640 eb->fs_info->sb->s_id, found_start, eb->start);
f188591e 641 ret = -EIO;
ce9adaa5
CM		 642 goto err;
643 }
01d58472 644 if (check_tree_block_fsid(root->fs_info, eb)) {
68b663d1 645 printk_ratelimited(KERN_ERR "BTRFS (device %s): bad fsid on block %llu\n",
29549aec 646 eb->fs_info->sb->s_id, eb->start);
1259ab75
CM		 647 ret = -EIO;
648 goto err;
649 }
ce9adaa5 650 found_level = btrfs_header_level(eb);
1c24c3ce 651 if (found_level >= BTRFS_MAX_LEVEL) {
68b663d1 652 btrfs_err(root->fs_info, "bad tree block level %d",
1c24c3ce
JB		 653 (int)btrfs_header_level(eb));
654 ret = -EIO;
655 goto err;
656 }
ce9adaa5 657
85d4e461
CM		 658 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
659 eb, found_level);
4008c04a 660
01d58472 661 ret = csum_tree_block(root->fs_info, eb, 1);
a826d6dc 662 if (ret) {
f188591e 663 ret = -EIO;
a826d6dc
JB		 664 goto err;
665 }
666
667 /*
668 * If this is a leaf block and it is corrupt, set the corrupt bit so
669 * that we don't try and read the other copies of this block, just
670 * return -EIO.
671 */
672 if (found_level == 0 && check_leaf(root, eb)) {
673 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
674 ret = -EIO;
675 }
ce9adaa5 676
0b32f4bb
JB		 677 if (!ret)
678 set_extent_buffer_uptodate(eb);
ce9adaa5 679err:
79fb65a1
JB		 680 if (reads_done &&
681 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 682 btree_readahead_hook(root, eb, eb->start, ret);
4bb31e92 683
53b381b3
DW		 684 if (ret) {
685 /*
686 * our io error hook is going to dec the io pages
687 * again, we have to make sure it has something
688 * to decrement
689 */
690 atomic_inc(&eb->io_pages);
0b32f4bb 691 clear_extent_buffer_uptodate(eb);
53b381b3 692 }
0b32f4bb 693 free_extent_buffer(eb);
ce9adaa5 694out:
f188591e 695 return ret;
ce9adaa5
CM		 696}
697
ea466794 698static int btree_io_failed_hook(struct page *page, int failed_mirror)
4bb31e92 699{
4bb31e92
AJ		 700 struct extent_buffer *eb;
701 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
702
4f2de97a 703 eb = (struct extent_buffer *)page->private;
656f30db 704 set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
5cf1ab56 705 eb->read_mirror = failed_mirror;
53b381b3 706 atomic_dec(&eb->io_pages);
ea466794 707 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 708 btree_readahead_hook(root, eb, eb->start, -EIO);
4bb31e92
AJ		 709 return -EIO; /* we fixed nothing */
710}
711
ce9adaa5 712static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5 713{
97eb6b69 714 struct btrfs_end_io_wq *end_io_wq = bio->bi_private;
ce9adaa5 715 struct btrfs_fs_info *fs_info;
9e0af237
LB		 716 struct btrfs_workqueue *wq;
717 btrfs_work_func_t func;
ce9adaa5 718
ce9adaa5 719 fs_info = end_io_wq->info;
ce9adaa5 720 end_io_wq->error = err;
d20f7043 721
7b6d91da 722 if (bio->bi_rw & REQ_WRITE) {
9e0af237
LB		 723 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) {
724 wq = fs_info->endio_meta_write_workers;
725 func = btrfs_endio_meta_write_helper;
726 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) {
727 wq = fs_info->endio_freespace_worker;
728 func = btrfs_freespace_write_helper;
729 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
730 wq = fs_info->endio_raid56_workers;
731 func = btrfs_endio_raid56_helper;
732 } else {
733 wq = fs_info->endio_write_workers;
734 func = btrfs_endio_write_helper;
735 }
d20f7043 736 } else {
8b110e39
MX		 737 if (unlikely(end_io_wq->metadata ==
738 BTRFS_WQ_ENDIO_DIO_REPAIR)) {
739 wq = fs_info->endio_repair_workers;
740 func = btrfs_endio_repair_helper;
741 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
9e0af237
LB		 742 wq = fs_info->endio_raid56_workers;
743 func = btrfs_endio_raid56_helper;
744 } else if (end_io_wq->metadata) {
745 wq = fs_info->endio_meta_workers;
746 func = btrfs_endio_meta_helper;
747 } else {
748 wq = fs_info->endio_workers;
749 func = btrfs_endio_helper;
750 }
d20f7043 751 }
9e0af237
LB		 752
753 btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL);
754 btrfs_queue_work(wq, &end_io_wq->work);
ce9adaa5
CM		 755}
756
22c59948 757int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
bfebd8b5 758 enum btrfs_wq_endio_type metadata)
0b86a832 759{
97eb6b69 760 struct btrfs_end_io_wq *end_io_wq;
8b110e39 761
97eb6b69 762 end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS);
ce9adaa5
CM		 763 if (!end_io_wq)
764 return -ENOMEM;
765
766 end_io_wq->private = bio->bi_private;
767 end_io_wq->end_io = bio->bi_end_io;
22c59948 768 end_io_wq->info = info;
ce9adaa5
CM		 769 end_io_wq->error = 0;
770 end_io_wq->bio = bio;
22c59948 771 end_io_wq->metadata = metadata;
ce9adaa5
CM		 772
773 bio->bi_private = end_io_wq;
774 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 775 return 0;
776}
777
b64a2851 778unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 779{
4854ddd0 780 unsigned long limit = min_t(unsigned long,
5cdc7ad3 781 info->thread_pool_size,
4854ddd0
CM		 782 info->fs_devices->open_devices);
783 return 256 * limit;
784}
0986fe9e 785
4a69a410
CM		 786static void run_one_async_start(struct btrfs_work *work)
787{
4a69a410 788 struct async_submit_bio *async;
79787eaa 789 int ret;
4a69a410
CM		 790
791 async = container_of(work, struct async_submit_bio, work);
79787eaa
JM		 792 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
793 async->mirror_num, async->bio_flags,
794 async->bio_offset);
795 if (ret)
796 async->error = ret;
4a69a410
CM		 797}
798
799static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 800{
801 struct btrfs_fs_info *fs_info;
802 struct async_submit_bio *async;
4854ddd0 803 int limit;
8b712842
CM		 804
805 async = container_of(work, struct async_submit_bio, work);
806 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 807
b64a2851 808 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 809 limit = limit * 2 / 3;
810
66657b31 811 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
b64a2851 812 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 813 wake_up(&fs_info->async_submit_wait);
814
79787eaa
JM		 815 /* If an error occured we just want to clean up the bio and move on */
816 if (async->error) {
817 bio_endio(async->bio, async->error);
818 return;
819 }
820
4a69a410 821 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 822 async->mirror_num, async->bio_flags,
823 async->bio_offset);
4a69a410
CM		 824}
825
826static void run_one_async_free(struct btrfs_work *work)
827{
828 struct async_submit_bio *async;
829
830 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 831 kfree(async);
832}
833
44b8bd7e
CM		 834int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
835 int rw, struct bio *bio, int mirror_num,
c8b97818 836 unsigned long bio_flags,
eaf25d93 837 u64 bio_offset,
4a69a410
CM		 838 extent_submit_bio_hook_t *submit_bio_start,
839 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 840{
841 struct async_submit_bio *async;
842
843 async = kmalloc(sizeof(*async), GFP_NOFS);
844 if (!async)
845 return -ENOMEM;
846
847 async->inode = inode;
848 async->rw = rw;
849 async->bio = bio;
850 async->mirror_num = mirror_num;
4a69a410
CM		 851 async->submit_bio_start = submit_bio_start;
852 async->submit_bio_done = submit_bio_done;
853
9e0af237 854 btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start,
5cdc7ad3 855 run_one_async_done, run_one_async_free);
4a69a410 856
c8b97818 857 async->bio_flags = bio_flags;
eaf25d93 858 async->bio_offset = bio_offset;
8c8bee1d 859
79787eaa
JM		 860 async->error = 0;
861
cb03c743 862 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 863
7b6d91da 864 if (rw & REQ_SYNC)
5cdc7ad3 865 btrfs_set_work_high_priority(&async->work);
d313d7a3 866
5cdc7ad3 867 btrfs_queue_work(fs_info->workers, &async->work);
9473f16c 868
d397712b 869 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 870 atomic_read(&fs_info->nr_async_submits)) {
871 wait_event(fs_info->async_submit_wait,
872 (atomic_read(&fs_info->nr_async_submits) == 0));
873 }
874
44b8bd7e
CM		 875 return 0;
876}
877
ce3ed71a
CM		 878static int btree_csum_one_bio(struct bio *bio)
879{
2c30c71b 880 struct bio_vec *bvec;
ce3ed71a 881 struct btrfs_root *root;
2c30c71b 882 int i, ret = 0;
ce3ed71a 883
2c30c71b 884 bio_for_each_segment_all(bvec, bio, i) {
ce3ed71a 885 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
01d58472 886 ret = csum_dirty_buffer(root->fs_info, bvec->bv_page);
79787eaa
JM		 887 if (ret)
888 break;
ce3ed71a 889 }
2c30c71b 890
79787eaa 891 return ret;
ce3ed71a
CM		 892}
893
4a69a410
CM		 894static int __btree_submit_bio_start(struct inode *inode, int rw,
895 struct bio *bio, int mirror_num,
eaf25d93
CM		 896 unsigned long bio_flags,
897 u64 bio_offset)
22c59948 898{
8b712842
CM		 899 /*
900 * when we're called for a write, we're already in the async
5443be45 901 * submission context. Just jump into btrfs_map_bio
8b712842 902 */
79787eaa 903 return btree_csum_one_bio(bio);
4a69a410 904}
22c59948 905
4a69a410 906static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 907 int mirror_num, unsigned long bio_flags,
908 u64 bio_offset)
4a69a410 909{
61891923
SB		 910 int ret;
911
8b712842 912 /*
4a69a410
CM		 913 * when we're called for a write, we're already in the async
914 * submission context. Just jump into btrfs_map_bio
8b712842 915 */
61891923
SB		 916 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
917 if (ret)
918 bio_endio(bio, ret);
919 return ret;
0b86a832
CM		 920}
921
de0022b9
JB		 922static int check_async_write(struct inode *inode, unsigned long bio_flags)
923{
924 if (bio_flags & EXTENT_BIO_TREE_LOG)
925 return 0;
926#ifdef CONFIG_X86
927 if (cpu_has_xmm4_2)
928 return 0;
929#endif
930 return 1;
931}
932
44b8bd7e 933static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 934 int mirror_num, unsigned long bio_flags,
935 u64 bio_offset)
44b8bd7e 936{
de0022b9 937 int async = check_async_write(inode, bio_flags);
cad321ad
CM		 938 int ret;
939
7b6d91da 940 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 941 /*
942 * called for a read, do the setup so that checksum validation
943 * can happen in the async kernel threads
944 */
f3f266ab 945 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
bfebd8b5 946 bio, BTRFS_WQ_ENDIO_METADATA);
1d4284bd 947 if (ret)
61891923
SB		 948 goto out_w_error;
949 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
950 mirror_num, 0);
de0022b9
JB		 951 } else if (!async) {
952 ret = btree_csum_one_bio(bio);
953 if (ret)
61891923
SB		 954 goto out_w_error;
955 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
956 mirror_num, 0);
957 } else {
958 /*
959 * kthread helpers are used to submit writes so that
960 * checksumming can happen in parallel across all CPUs
961 */
962 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
963 inode, rw, bio, mirror_num, 0,
964 bio_offset,
965 __btree_submit_bio_start,
966 __btree_submit_bio_done);
44b8bd7e 967 }
d313d7a3 968
61891923
SB		 969 if (ret) {
970out_w_error:
971 bio_endio(bio, ret);
972 }
973 return ret;
44b8bd7e
CM		 974}
975
3dd1462e 976#ifdef CONFIG_MIGRATION
784b4e29 977static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG		 978 struct page *newpage, struct page *page,
979 enum migrate_mode mode)
784b4e29
CM		 980{
981 /*
982 * we can't safely write a btree page from here,
983 * we haven't done the locking hook
984 */
985 if (PageDirty(page))
986 return -EAGAIN;
987 /*
988 * Buffers may be managed in a filesystem specific way.
989 * We must have no buffers or drop them.
990 */
991 if (page_has_private(page) &&
992 !try_to_release_page(page, GFP_KERNEL))
993 return -EAGAIN;
a6bc32b8 994 return migrate_page(mapping, newpage, page, mode);
784b4e29 995}
3dd1462e 996#endif
784b4e29 997
0da5468f
CM		 998
999static int btree_writepages(struct address_space *mapping,
1000 struct writeback_control *wbc)
1001{
e2d84521
MX		 1002 struct btrfs_fs_info *fs_info;
1003 int ret;
1004
d8d5f3e1 1005 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM		 1006
1007 if (wbc->for_kupdate)
1008 return 0;
1009
e2d84521 1010 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 1011 /* this is a bit racy, but that's ok */
e2d84521
MX		 1012 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
1013 BTRFS_DIRTY_METADATA_THRESH);
1014 if (ret < 0)
793955bc 1015 return 0;
793955bc 1016 }
0b32f4bb 1017 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM		 1018}
1019
b2950863 1020static int btree_readpage(struct file *file, struct page *page)
5f39d397 1021{
d1310b2e
CM		 1022 struct extent_io_tree *tree;
1023 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 1024 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 1025}
22b0ebda 1026
70dec807 1027static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 1028{
98509cfc 1029 if (PageWriteback(page) || PageDirty(page))
d397712b 1030 return 0;
0c4e538b 1031
f7a52a40 1032 return try_release_extent_buffer(page);
d98237b3
CM		 1033}
1034
d47992f8
LC		 1035static void btree_invalidatepage(struct page *page, unsigned int offset,
1036 unsigned int length)
d98237b3 1037{
d1310b2e
CM		 1038 struct extent_io_tree *tree;
1039 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 1040 extent_invalidatepage(tree, page, offset);
1041 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 1042 if (PagePrivate(page)) {
efe120a0
FH		 1043 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1044 "page private not zero on page %llu",
1045 (unsigned long long)page_offset(page));
9ad6b7bc
CM		 1046 ClearPagePrivate(page);
1047 set_page_private(page, 0);
1048 page_cache_release(page);
1049 }
d98237b3
CM		 1050}
1051
0b32f4bb
JB		 1052static int btree_set_page_dirty(struct page *page)
1053{
bb146eb2 1054#ifdef DEBUG
0b32f4bb
JB		 1055 struct extent_buffer *eb;
1056
1057 BUG_ON(!PagePrivate(page));
1058 eb = (struct extent_buffer *)page->private;
1059 BUG_ON(!eb);
1060 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1061 BUG_ON(!atomic_read(&eb->refs));
1062 btrfs_assert_tree_locked(eb);
bb146eb2 1063#endif
0b32f4bb
JB		 1064 return __set_page_dirty_nobuffers(page);
1065}
1066
7f09410b 1067static const struct address_space_operations btree_aops = {
d98237b3 1068 .readpage = btree_readpage,
0da5468f 1069 .writepages = btree_writepages,
5f39d397
CM		 1070 .releasepage = btree_releasepage,
1071 .invalidatepage = btree_invalidatepage,
5a92bc88 1072#ifdef CONFIG_MIGRATION
784b4e29 1073 .migratepage = btree_migratepage,
5a92bc88 1074#endif
0b32f4bb 1075 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM		 1076};
1077
d3e46fea 1078void readahead_tree_block(struct btrfs_root *root, u64 bytenr)
090d1875 1079{
5f39d397
CM		 1080 struct extent_buffer *buf = NULL;
1081 struct inode *btree_inode = root->fs_info->btree_inode;
090d1875 1082
a83fffb7 1083 buf = btrfs_find_create_tree_block(root, bytenr);
5f39d397 1084 if (!buf)
6197d86e 1085 return;
d1310b2e 1086 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
bb82ab88 1087 buf, 0, WAIT_NONE, btree_get_extent, 0);
5f39d397 1088 free_extent_buffer(buf);
090d1875
CM		 1089}
1090
c0dcaa4d 1091int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr,
ab0fff03
AJ		 1092 int mirror_num, struct extent_buffer **eb)
1093{
1094 struct extent_buffer *buf = NULL;
1095 struct inode *btree_inode = root->fs_info->btree_inode;
1096 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1097 int ret;
1098
a83fffb7 1099 buf = btrfs_find_create_tree_block(root, bytenr);
ab0fff03
AJ		 1100 if (!buf)
1101 return 0;
1102
1103 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1104
1105 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1106 btree_get_extent, mirror_num);
1107 if (ret) {
1108 free_extent_buffer(buf);
1109 return ret;
1110 }
1111
1112 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1113 free_extent_buffer(buf);
1114 return -EIO;
0b32f4bb 1115 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ		 1116 *eb = buf;
1117 } else {
1118 free_extent_buffer(buf);
1119 }
1120 return 0;
1121}
1122
01d58472 1123struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info,
0308af44 1124 u64 bytenr)
0999df54 1125{
01d58472 1126 return find_extent_buffer(fs_info, bytenr);
0999df54
CM		 1127}
1128
1129struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
a83fffb7 1130 u64 bytenr)
0999df54 1131{
fccb84c9 1132 if (btrfs_test_is_dummy_root(root))
ce3e6984
DS		 1133 return alloc_test_extent_buffer(root->fs_info, bytenr);
1134 return alloc_extent_buffer(root->fs_info, bytenr);
0999df54
CM		 1135}
1136
1137
e02119d5
CM		 1138int btrfs_write_tree_block(struct extent_buffer *buf)
1139{
727011e0 1140 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1141 buf->start + buf->len - 1);
e02119d5
CM		 1142}
1143
1144int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1145{
727011e0 1146 return filemap_fdatawait_range(buf->pages[0]->mapping,
8aa38c31 1147 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1148}
1149
0999df54 1150struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ce86cd59 1151 u64 parent_transid)
0999df54
CM		 1152{
1153 struct extent_buffer *buf = NULL;
0999df54
CM		 1154 int ret;
1155
a83fffb7 1156 buf = btrfs_find_create_tree_block(root, bytenr);
0999df54
CM		 1157 if (!buf)
1158 return NULL;
0999df54 1159
ca7a79ad 1160 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
0f0fe8f7
FDBM		 1161 if (ret) {
1162 free_extent_buffer(buf);
1163 return NULL;
1164 }
5f39d397 1165 return buf;
ce9adaa5 1166
eb60ceac
CM		 1167}
1168
01d58472
DD		 1169void clean_tree_block(struct btrfs_trans_handle *trans,
1170 struct btrfs_fs_info *fs_info,
d5c13f92 1171 struct extent_buffer *buf)
ed2ff2cb 1172{
55c69072 1173 if (btrfs_header_generation(buf) ==
e2d84521 1174 fs_info->running_transaction->transid) {
b9447ef8 1175 btrfs_assert_tree_locked(buf);
b4ce94de 1176
b9473439 1177 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
e2d84521
MX		 1178 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1179 -buf->len,
1180 fs_info->dirty_metadata_batch);
ed7b63eb
JB		 1181 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1182 btrfs_set_lock_blocking(buf);
1183 clear_extent_buffer_dirty(buf);
1184 }
925baedd 1185 }
5f39d397
CM		 1186}
1187
8257b2dc
MX		 1188static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1189{
1190 struct btrfs_subvolume_writers *writers;
1191 int ret;
1192
1193 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1194 if (!writers)
1195 return ERR_PTR(-ENOMEM);
1196
908c7f19 1197 ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL);
8257b2dc
MX		 1198 if (ret < 0) {
1199 kfree(writers);
1200 return ERR_PTR(ret);
1201 }
1202
1203 init_waitqueue_head(&writers->wait);
1204 return writers;
1205}
1206
1207static void
1208btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1209{
1210 percpu_counter_destroy(&writers->counter);
1211 kfree(writers);
1212}
1213
707e8a07
DS		 1214static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize,
1215 struct btrfs_root *root, struct btrfs_fs_info *fs_info,
143bede5 1216 u64 objectid)
d97e63b6 1217{
cfaa7295 1218 root->node = NULL;
a28ec197 1219 root->commit_root = NULL;
db94535d
CM		 1220 root->sectorsize = sectorsize;
1221 root->nodesize = nodesize;
87ee04eb 1222 root->stripesize = stripesize;
27cdeb70 1223 root->state = 0;
d68fc57b 1224 root->orphan_cleanup_state = 0;
0b86a832 1225
0f7d52f4
CM		 1226 root->objectid = objectid;
1227 root->last_trans = 0;
13a8a7c8 1228 root->highest_objectid = 0;
eb73c1b7 1229 root->nr_delalloc_inodes = 0;
199c2a9c 1230 root->nr_ordered_extents = 0;
58176a96 1231 root->name = NULL;
6bef4d31 1232 root->inode_tree = RB_ROOT;
16cdcec7 1233 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1234 root->block_rsv = NULL;
d68fc57b 1235 root->orphan_block_rsv = NULL;
0b86a832
CM		 1236
1237 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1238 INIT_LIST_HEAD(&root->root_list);
eb73c1b7
MX		 1239 INIT_LIST_HEAD(&root->delalloc_inodes);
1240 INIT_LIST_HEAD(&root->delalloc_root);
199c2a9c
MX		 1241 INIT_LIST_HEAD(&root->ordered_extents);
1242 INIT_LIST_HEAD(&root->ordered_root);
2ab28f32
JB		 1243 INIT_LIST_HEAD(&root->logged_list[0]);
1244 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1245 spin_lock_init(&root->orphan_lock);
5d4f98a2 1246 spin_lock_init(&root->inode_lock);
eb73c1b7 1247 spin_lock_init(&root->delalloc_lock);
199c2a9c 1248 spin_lock_init(&root->ordered_extent_lock);
f0486c68 1249 spin_lock_init(&root->accounting_lock);
2ab28f32
JB		 1250 spin_lock_init(&root->log_extents_lock[0]);
1251 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1252 mutex_init(&root->objectid_mutex);
e02119d5 1253 mutex_init(&root->log_mutex);
31f3d255 1254 mutex_init(&root->ordered_extent_mutex);
573bfb72 1255 mutex_init(&root->delalloc_mutex);
7237f183
YZ		 1256 init_waitqueue_head(&root->log_writer_wait);
1257 init_waitqueue_head(&root->log_commit_wait[0]);
1258 init_waitqueue_head(&root->log_commit_wait[1]);
8b050d35
MX		 1259 INIT_LIST_HEAD(&root->log_ctxs[0]);
1260 INIT_LIST_HEAD(&root->log_ctxs[1]);
7237f183
YZ		 1261 atomic_set(&root->log_commit[0], 0);
1262 atomic_set(&root->log_commit[1], 0);
1263 atomic_set(&root->log_writers, 0);
2ecb7923 1264 atomic_set(&root->log_batch, 0);
8a35d95f 1265 atomic_set(&root->orphan_inodes, 0);
b0feb9d9 1266 atomic_set(&root->refs, 1);
8257b2dc 1267 atomic_set(&root->will_be_snapshoted, 0);
7237f183 1268 root->log_transid = 0;
d1433deb 1269 root->log_transid_committed = -1;
257c62e1 1270 root->last_log_commit = 0;
06ea65a3
JB		 1271 if (fs_info)
1272 extent_io_tree_init(&root->dirty_log_pages,
1273 fs_info->btree_inode->i_mapping);
017e5369 1274
3768f368
CM		 1275 memset(&root->root_key, 0, sizeof(root->root_key));
1276 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1277 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
06ea65a3
JB		 1278 if (fs_info)
1279 root->defrag_trans_start = fs_info->generation;
1280 else
1281 root->defrag_trans_start = 0;
4d775673 1282 root->root_key.objectid = objectid;
0ee5dc67 1283 root->anon_dev = 0;
8ea05e3a 1284
5f3ab90a 1285 spin_lock_init(&root->root_item_lock);
3768f368
CM		 1286}
1287
f84a8bd6 1288static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
6f07e42e
AV		 1289{
1290 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1291 if (root)
1292 root->fs_info = fs_info;
1293 return root;
1294}
1295
06ea65a3
JB		 1296#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1297/* Should only be used by the testing infrastructure */
1298struct btrfs_root *btrfs_alloc_dummy_root(void)
1299{
1300 struct btrfs_root *root;
1301
1302 root = btrfs_alloc_root(NULL);
1303 if (!root)
1304 return ERR_PTR(-ENOMEM);
707e8a07 1305 __setup_root(4096, 4096, 4096, root, NULL, 1);
27cdeb70 1306 set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state);
faa2dbf0 1307 root->alloc_bytenr = 0;
06ea65a3
JB		 1308
1309 return root;
1310}
1311#endif
1312
20897f5c
AJ		 1313struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1314 struct btrfs_fs_info *fs_info,
1315 u64 objectid)
1316{
1317 struct extent_buffer *leaf;
1318 struct btrfs_root *tree_root = fs_info->tree_root;
1319 struct btrfs_root *root;
1320 struct btrfs_key key;
1321 int ret = 0;
6463fe58 1322 uuid_le uuid;
20897f5c
AJ		 1323
1324 root = btrfs_alloc_root(fs_info);
1325 if (!root)
1326 return ERR_PTR(-ENOMEM);
1327
707e8a07
DS		 1328 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1329 tree_root->stripesize, root, fs_info, objectid);
20897f5c
AJ		 1330 root->root_key.objectid = objectid;
1331 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1332 root->root_key.offset = 0;
1333
4d75f8a9 1334 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
20897f5c
AJ		 1335 if (IS_ERR(leaf)) {
1336 ret = PTR_ERR(leaf);
1dd05682 1337 leaf = NULL;
20897f5c
AJ		 1338 goto fail;
1339 }
1340
20897f5c
AJ		 1341 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1342 btrfs_set_header_bytenr(leaf, leaf->start);
1343 btrfs_set_header_generation(leaf, trans->transid);
1344 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1345 btrfs_set_header_owner(leaf, objectid);
1346 root->node = leaf;
1347
0a4e5586 1348 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
20897f5c
AJ		 1349 BTRFS_FSID_SIZE);
1350 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
b308bc2f 1351 btrfs_header_chunk_tree_uuid(leaf),
20897f5c
AJ		 1352 BTRFS_UUID_SIZE);
1353 btrfs_mark_buffer_dirty(leaf);
1354
1355 root->commit_root = btrfs_root_node(root);
27cdeb70 1356 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
20897f5c
AJ		 1357
1358 root->root_item.flags = 0;
1359 root->root_item.byte_limit = 0;
1360 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1361 btrfs_set_root_generation(&root->root_item, trans->transid);
1362 btrfs_set_root_level(&root->root_item, 0);
1363 btrfs_set_root_refs(&root->root_item, 1);
1364 btrfs_set_root_used(&root->root_item, leaf->len);
1365 btrfs_set_root_last_snapshot(&root->root_item, 0);
1366 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB		 1367 uuid_le_gen(&uuid);
1368 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ		 1369 root->root_item.drop_level = 0;
1370
1371 key.objectid = objectid;
1372 key.type = BTRFS_ROOT_ITEM_KEY;
1373 key.offset = 0;
1374 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1375 if (ret)
1376 goto fail;
1377
1378 btrfs_tree_unlock(leaf);
1379
1dd05682
TI		 1380 return root;
1381
20897f5c 1382fail:
1dd05682
TI		 1383 if (leaf) {
1384 btrfs_tree_unlock(leaf);
59885b39 1385 free_extent_buffer(root->commit_root);
1dd05682
TI		 1386 free_extent_buffer(leaf);
1387 }
1388 kfree(root);
20897f5c 1389
1dd05682 1390 return ERR_PTR(ret);
20897f5c
AJ		 1391}
1392
7237f183
YZ		 1393static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1394 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1395{
1396 struct btrfs_root *root;
1397 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1398 struct extent_buffer *leaf;
e02119d5 1399
6f07e42e 1400 root = btrfs_alloc_root(fs_info);
e02119d5 1401 if (!root)
7237f183 1402 return ERR_PTR(-ENOMEM);
e02119d5 1403
707e8a07
DS		 1404 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1405 tree_root->stripesize, root, fs_info,
1406 BTRFS_TREE_LOG_OBJECTID);
e02119d5
CM		 1407
1408 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1409 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1410 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
27cdeb70 1411
7237f183 1412 /*
27cdeb70
MX		 1413 * DON'T set REF_COWS for log trees
1414 *
7237f183
YZ		 1415 * log trees do not get reference counted because they go away
1416 * before a real commit is actually done. They do store pointers
1417 * to file data extents, and those reference counts still get
1418 * updated (along with back refs to the log tree).
1419 */
e02119d5 1420
4d75f8a9
DS		 1421 leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID,
1422 NULL, 0, 0, 0);
7237f183
YZ		 1423 if (IS_ERR(leaf)) {
1424 kfree(root);
1425 return ERR_CAST(leaf);
1426 }
e02119d5 1427
5d4f98a2
YZ		 1428 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1429 btrfs_set_header_bytenr(leaf, leaf->start);
1430 btrfs_set_header_generation(leaf, trans->transid);
1431 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1432 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1433 root->node = leaf;
e02119d5
CM		 1434
1435 write_extent_buffer(root->node, root->fs_info->fsid,
0a4e5586 1436 btrfs_header_fsid(), BTRFS_FSID_SIZE);
e02119d5
CM		 1437 btrfs_mark_buffer_dirty(root->node);
1438 btrfs_tree_unlock(root->node);
7237f183
YZ		 1439 return root;
1440}
1441
1442int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1443 struct btrfs_fs_info *fs_info)
1444{
1445 struct btrfs_root *log_root;
1446
1447 log_root = alloc_log_tree(trans, fs_info);
1448 if (IS_ERR(log_root))
1449 return PTR_ERR(log_root);
1450 WARN_ON(fs_info->log_root_tree);
1451 fs_info->log_root_tree = log_root;
1452 return 0;
1453}
1454
1455int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1456 struct btrfs_root *root)
1457{
1458 struct btrfs_root *log_root;
1459 struct btrfs_inode_item *inode_item;
1460
1461 log_root = alloc_log_tree(trans, root->fs_info);
1462 if (IS_ERR(log_root))
1463 return PTR_ERR(log_root);
1464
1465 log_root->last_trans = trans->transid;
1466 log_root->root_key.offset = root->root_key.objectid;
1467
1468 inode_item = &log_root->root_item.inode;
3cae210f
QW		 1469 btrfs_set_stack_inode_generation(inode_item, 1);
1470 btrfs_set_stack_inode_size(inode_item, 3);
1471 btrfs_set_stack_inode_nlink(inode_item, 1);
707e8a07 1472 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
3cae210f 1473 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
7237f183 1474
5d4f98a2 1475 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1476
1477 WARN_ON(root->log_root);
1478 root->log_root = log_root;
1479 root->log_transid = 0;
d1433deb 1480 root->log_transid_committed = -1;
257c62e1 1481 root->last_log_commit = 0;
e02119d5
CM		 1482 return 0;
1483}
1484
35a3621b
SB		 1485static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1486 struct btrfs_key *key)
e02119d5
CM		 1487{
1488 struct btrfs_root *root;
1489 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1490 struct btrfs_path *path;
84234f3a 1491 u64 generation;
cb517eab 1492 int ret;
0f7d52f4 1493
cb517eab
MX		 1494 path = btrfs_alloc_path();
1495 if (!path)
0f7d52f4 1496 return ERR_PTR(-ENOMEM);
cb517eab
MX		 1497
1498 root = btrfs_alloc_root(fs_info);
1499 if (!root) {
1500 ret = -ENOMEM;
1501 goto alloc_fail;
0f7d52f4
CM		 1502 }
1503
707e8a07
DS		 1504 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1505 tree_root->stripesize, root, fs_info, key->objectid);
0f7d52f4 1506
cb517eab
MX		 1507 ret = btrfs_find_root(tree_root, key, path,
1508 &root->root_item, &root->root_key);
0f7d52f4 1509 if (ret) {
13a8a7c8
YZ		 1510 if (ret > 0)
1511 ret = -ENOENT;
cb517eab 1512 goto find_fail;
0f7d52f4 1513 }
13a8a7c8 1514
84234f3a 1515 generation = btrfs_root_generation(&root->root_item);
db94535d 1516 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
ce86cd59 1517 generation);
cb517eab
MX		 1518 if (!root->node) {
1519 ret = -ENOMEM;
1520 goto find_fail;
1521 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1522 ret = -EIO;
1523 goto read_fail;
416bc658 1524 }
5d4f98a2 1525 root->commit_root = btrfs_root_node(root);
13a8a7c8 1526out:
cb517eab
MX		 1527 btrfs_free_path(path);
1528 return root;
1529
1530read_fail:
1531 free_extent_buffer(root->node);
1532find_fail:
1533 kfree(root);
1534alloc_fail:
1535 root = ERR_PTR(ret);
1536 goto out;
1537}
1538
1539struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1540 struct btrfs_key *location)
1541{
1542 struct btrfs_root *root;
1543
1544 root = btrfs_read_tree_root(tree_root, location);
1545 if (IS_ERR(root))
1546 return root;
1547
1548 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
27cdeb70 1549 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
08fe4db1
LZ		 1550 btrfs_check_and_init_root_item(&root->root_item);
1551 }
13a8a7c8 1552
5eda7b5e
CM		 1553 return root;
1554}
1555
cb517eab
MX		 1556int btrfs_init_fs_root(struct btrfs_root *root)
1557{
1558 int ret;
8257b2dc 1559 struct btrfs_subvolume_writers *writers;
cb517eab
MX		 1560
1561 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1562 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1563 GFP_NOFS);
1564 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1565 ret = -ENOMEM;
1566 goto fail;
1567 }
1568
8257b2dc
MX		 1569 writers = btrfs_alloc_subvolume_writers();
1570 if (IS_ERR(writers)) {
1571 ret = PTR_ERR(writers);
1572 goto fail;
1573 }
1574 root->subv_writers = writers;
1575
cb517eab 1576 btrfs_init_free_ino_ctl(root);
57cdc8db
DS		 1577 spin_lock_init(&root->ino_cache_lock);
1578 init_waitqueue_head(&root->ino_cache_wait);
cb517eab
MX		 1579
1580 ret = get_anon_bdev(&root->anon_dev);
1581 if (ret)
8257b2dc 1582 goto free_writers;
cb517eab 1583 return 0;
8257b2dc
MX		 1584
1585free_writers:
1586 btrfs_free_subvolume_writers(root->subv_writers);
cb517eab
MX		 1587fail:
1588 kfree(root->free_ino_ctl);
1589 kfree(root->free_ino_pinned);
1590 return ret;
1591}
1592
171170c1
ST		 1593static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1594 u64 root_id)
cb517eab
MX		 1595{
1596 struct btrfs_root *root;
1597
1598 spin_lock(&fs_info->fs_roots_radix_lock);
1599 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1600 (unsigned long)root_id);
1601 spin_unlock(&fs_info->fs_roots_radix_lock);
1602 return root;
1603}
1604
1605int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1606 struct btrfs_root *root)
1607{
1608 int ret;
1609
1610 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1611 if (ret)
1612 return ret;
1613
1614 spin_lock(&fs_info->fs_roots_radix_lock);
1615 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1616 (unsigned long)root->root_key.objectid,
1617 root);
1618 if (ret == 0)
27cdeb70 1619 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
cb517eab
MX		 1620 spin_unlock(&fs_info->fs_roots_radix_lock);
1621 radix_tree_preload_end();
1622
1623 return ret;
1624}
1625
c00869f1
MX		 1626struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1627 struct btrfs_key *location,
1628 bool check_ref)
5eda7b5e
CM		 1629{
1630 struct btrfs_root *root;
381cf658 1631 struct btrfs_path *path;
1d4c08e0 1632 struct btrfs_key key;
5eda7b5e
CM		 1633 int ret;
1634
edbd8d4e
CM		 1635 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1636 return fs_info->tree_root;
1637 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1638 return fs_info->extent_root;
8f18cf13
CM		 1639 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1640 return fs_info->chunk_root;
1641 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1642 return fs_info->dev_root;
0403e47e
YZ		 1643 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1644 return fs_info->csum_root;
bcef60f2
AJ		 1645 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1646 return fs_info->quota_root ? fs_info->quota_root :
1647 ERR_PTR(-ENOENT);
f7a81ea4
SB		 1648 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1649 return fs_info->uuid_root ? fs_info->uuid_root :
1650 ERR_PTR(-ENOENT);
4df27c4d 1651again:
cb517eab 1652 root = btrfs_lookup_fs_root(fs_info, location->objectid);
48475471 1653 if (root) {
c00869f1 1654 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
48475471 1655 return ERR_PTR(-ENOENT);
5eda7b5e 1656 return root;
48475471 1657 }
5eda7b5e 1658
cb517eab 1659 root = btrfs_read_fs_root(fs_info->tree_root, location);
5eda7b5e
CM		 1660 if (IS_ERR(root))
1661 return root;
3394e160 1662
c00869f1 1663 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
cb517eab 1664 ret = -ENOENT;
581bb050 1665 goto fail;
35a30d7c 1666 }
581bb050 1667
cb517eab 1668 ret = btrfs_init_fs_root(root);
ac08aedf
CM		 1669 if (ret)
1670 goto fail;
3394e160 1671
381cf658
DS		 1672 path = btrfs_alloc_path();
1673 if (!path) {
1674 ret = -ENOMEM;
1675 goto fail;
1676 }
1d4c08e0
DS		 1677 key.objectid = BTRFS_ORPHAN_OBJECTID;
1678 key.type = BTRFS_ORPHAN_ITEM_KEY;
1679 key.offset = location->objectid;
1680
1681 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
381cf658 1682 btrfs_free_path(path);
d68fc57b
YZ		 1683 if (ret < 0)
1684 goto fail;
1685 if (ret == 0)
27cdeb70 1686 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
d68fc57b 1687
cb517eab 1688 ret = btrfs_insert_fs_root(fs_info, root);
0f7d52f4 1689 if (ret) {
4df27c4d
YZ		 1690 if (ret == -EEXIST) {
1691 free_fs_root(root);
1692 goto again;
1693 }
1694 goto fail;
0f7d52f4 1695 }
edbd8d4e 1696 return root;
4df27c4d
YZ		 1697fail:
1698 free_fs_root(root);
1699 return ERR_PTR(ret);
edbd8d4e
CM		 1700}
1701
04160088
CM		 1702static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1703{
1704 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1705 int ret = 0;
04160088
CM		 1706 struct btrfs_device *device;
1707 struct backing_dev_info *bdi;
b7967db7 1708
1f78160c
XG		 1709 rcu_read_lock();
1710 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1711 if (!device->bdev)
1712 continue;
04160088 1713 bdi = blk_get_backing_dev_info(device->bdev);
ff9ea323 1714 if (bdi_congested(bdi, bdi_bits)) {
04160088
CM		 1715 ret = 1;
1716 break;
1717 }
1718 }
1f78160c 1719 rcu_read_unlock();
04160088
CM		 1720 return ret;
1721}
1722
04160088
CM		 1723static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1724{
ad081f14
JA		 1725 int err;
1726
1727 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1728 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1729 if (err)
1730 return err;
1731
4575c9cc 1732 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1733 bdi->congested_fn = btrfs_congested_fn;
1734 bdi->congested_data = info;
1735 return 0;
1736}
1737
8b712842
CM		 1738/*
1739 * called by the kthread helper functions to finally call the bio end_io
1740 * functions. This is where read checksum verification actually happens
1741 */
1742static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1743{
ce9adaa5 1744 struct bio *bio;
97eb6b69 1745 struct btrfs_end_io_wq *end_io_wq;
ce9adaa5 1746 int error;
ce9adaa5 1747
97eb6b69 1748 end_io_wq = container_of(work, struct btrfs_end_io_wq, work);
8b712842 1749 bio = end_io_wq->bio;
ce9adaa5 1750
8b712842
CM		 1751 error = end_io_wq->error;
1752 bio->bi_private = end_io_wq->private;
1753 bio->bi_end_io = end_io_wq->end_io;
97eb6b69 1754 kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq);
bc1e79ac 1755 bio_endio_nodec(bio, error);
44b8bd7e
CM		 1756}
1757
a74a4b97
CM		 1758static int cleaner_kthread(void *arg)
1759{
1760 struct btrfs_root *root = arg;
d0278245 1761 int again;
a74a4b97
CM		 1762
1763 do {
d0278245 1764 again = 0;
a74a4b97 1765
d0278245 1766 /* Make the cleaner go to sleep early. */
babbf170 1767 if (btrfs_need_cleaner_sleep(root))
d0278245
MX		 1768 goto sleep;
1769
1770 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1771 goto sleep;
1772
dc7f370c
MX		 1773 /*
1774 * Avoid the problem that we change the status of the fs
1775 * during the above check and trylock.
1776 */
babbf170 1777 if (btrfs_need_cleaner_sleep(root)) {
dc7f370c
MX		 1778 mutex_unlock(&root->fs_info->cleaner_mutex);
1779 goto sleep;
76dda93c 1780 }
a74a4b97 1781
d0278245 1782 btrfs_run_delayed_iputs(root);
47ab2a6c 1783 btrfs_delete_unused_bgs(root->fs_info);
d0278245
MX		 1784 again = btrfs_clean_one_deleted_snapshot(root);
1785 mutex_unlock(&root->fs_info->cleaner_mutex);
1786
1787 /*
05323cd1
MX		 1788 * The defragger has dealt with the R/O remount and umount,
1789 * needn't do anything special here.
d0278245
MX		 1790 */
1791 btrfs_run_defrag_inodes(root->fs_info);
1792sleep:
9d1a2a3a 1793 if (!try_to_freeze() && !again) {
a74a4b97 1794 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1795 if (!kthread_should_stop())
1796 schedule();
a74a4b97
CM		 1797 __set_current_state(TASK_RUNNING);
1798 }
1799 } while (!kthread_should_stop());
1800 return 0;
1801}
1802
1803static int transaction_kthread(void *arg)
1804{
1805 struct btrfs_root *root = arg;
1806 struct btrfs_trans_handle *trans;
1807 struct btrfs_transaction *cur;
8929ecfa 1808 u64 transid;
a74a4b97
CM		 1809 unsigned long now;
1810 unsigned long delay;
914b2007 1811 bool cannot_commit;
a74a4b97
CM		 1812
1813 do {
914b2007 1814 cannot_commit = false;
8b87dc17 1815 delay = HZ * root->fs_info->commit_interval;
a74a4b97
CM		 1816 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1817
a4abeea4 1818 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1819 cur = root->fs_info->running_transaction;
1820 if (!cur) {
a4abeea4 1821 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1822 goto sleep;
1823 }
31153d81 1824
a74a4b97 1825 now = get_seconds();
4a9d8bde 1826 if (cur->state < TRANS_STATE_BLOCKED &&
8b87dc17
DS		 1827 (now < cur->start_time ||
1828 now - cur->start_time < root->fs_info->commit_interval)) {
a4abeea4 1829 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1830 delay = HZ * 5;
1831 goto sleep;
1832 }
8929ecfa 1833 transid = cur->transid;
a4abeea4 1834 spin_unlock(&root->fs_info->trans_lock);
56bec294 1835
79787eaa 1836 /* If the file system is aborted, this will always fail. */
354aa0fb 1837 trans = btrfs_attach_transaction(root);
914b2007 1838 if (IS_ERR(trans)) {
354aa0fb
MX		 1839 if (PTR_ERR(trans) != -ENOENT)
1840 cannot_commit = true;
79787eaa 1841 goto sleep;
914b2007 1842 }
8929ecfa 1843 if (transid == trans->transid) {
79787eaa 1844 btrfs_commit_transaction(trans, root);
8929ecfa
YZ		 1845 } else {
1846 btrfs_end_transaction(trans, root);
1847 }
a74a4b97
CM		 1848sleep:
1849 wake_up_process(root->fs_info->cleaner_kthread);
1850 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1851
4e121c06
JB		 1852 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1853 &root->fs_info->fs_state)))
1854 btrfs_cleanup_transaction(root);
a0acae0e 1855 if (!try_to_freeze()) {
a74a4b97 1856 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa 1857 if (!kthread_should_stop() &&
914b2007
JK		 1858 (!btrfs_transaction_blocked(root->fs_info) ||
1859 cannot_commit))
8929ecfa 1860 schedule_timeout(delay);
a74a4b97
CM		 1861 __set_current_state(TASK_RUNNING);
1862 }
1863 } while (!kthread_should_stop());
1864 return 0;
1865}
1866
af31f5e5
CM		 1867/*
1868 * this will find the highest generation in the array of
1869 * root backups. The index of the highest array is returned,
1870 * or -1 if we can't find anything.
1871 *
1872 * We check to make sure the array is valid by comparing the
1873 * generation of the latest root in the array with the generation
1874 * in the super block. If they don't match we pitch it.
1875 */
1876static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1877{
1878 u64 cur;
1879 int newest_index = -1;
1880 struct btrfs_root_backup *root_backup;
1881 int i;
1882
1883 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1884 root_backup = info->super_copy->super_roots + i;
1885 cur = btrfs_backup_tree_root_gen(root_backup);
1886 if (cur == newest_gen)
1887 newest_index = i;
1888 }
1889
1890 /* check to see if we actually wrapped around */
1891 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1892 root_backup = info->super_copy->super_roots;
1893 cur = btrfs_backup_tree_root_gen(root_backup);
1894 if (cur == newest_gen)
1895 newest_index = 0;
1896 }
1897 return newest_index;
1898}
1899
1900
1901/*
1902 * find the oldest backup so we know where to store new entries
1903 * in the backup array. This will set the backup_root_index
1904 * field in the fs_info struct
1905 */
1906static void find_oldest_super_backup(struct btrfs_fs_info *info,
1907 u64 newest_gen)
1908{
1909 int newest_index = -1;
1910
1911 newest_index = find_newest_super_backup(info, newest_gen);
1912 /* if there was garbage in there, just move along */
1913 if (newest_index == -1) {
1914 info->backup_root_index = 0;
1915 } else {
1916 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1917 }
1918}
1919
1920/*
1921 * copy all the root pointers into the super backup array.
1922 * this will bump the backup pointer by one when it is
1923 * done
1924 */
1925static void backup_super_roots(struct btrfs_fs_info *info)
1926{
1927 int next_backup;
1928 struct btrfs_root_backup *root_backup;
1929 int last_backup;
1930
1931 next_backup = info->backup_root_index;
1932 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1933 BTRFS_NUM_BACKUP_ROOTS;
1934
1935 /*
1936 * just overwrite the last backup if we're at the same generation
1937 * this happens only at umount
1938 */
1939 root_backup = info->super_for_commit->super_roots + last_backup;
1940 if (btrfs_backup_tree_root_gen(root_backup) ==
1941 btrfs_header_generation(info->tree_root->node))
1942 next_backup = last_backup;
1943
1944 root_backup = info->super_for_commit->super_roots + next_backup;
1945
1946 /*
1947 * make sure all of our padding and empty slots get zero filled
1948 * regardless of which ones we use today
1949 */
1950 memset(root_backup, 0, sizeof(*root_backup));
1951
1952 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1953
1954 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1955 btrfs_set_backup_tree_root_gen(root_backup,
1956 btrfs_header_generation(info->tree_root->node));
1957
1958 btrfs_set_backup_tree_root_level(root_backup,
1959 btrfs_header_level(info->tree_root->node));
1960
1961 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1962 btrfs_set_backup_chunk_root_gen(root_backup,
1963 btrfs_header_generation(info->chunk_root->node));
1964 btrfs_set_backup_chunk_root_level(root_backup,
1965 btrfs_header_level(info->chunk_root->node));
1966
1967 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1968 btrfs_set_backup_extent_root_gen(root_backup,
1969 btrfs_header_generation(info->extent_root->node));
1970 btrfs_set_backup_extent_root_level(root_backup,
1971 btrfs_header_level(info->extent_root->node));
1972
7c7e82a7
CM		 1973 /*
1974 * we might commit during log recovery, which happens before we set
1975 * the fs_root. Make sure it is valid before we fill it in.
1976 */
1977 if (info->fs_root && info->fs_root->node) {
1978 btrfs_set_backup_fs_root(root_backup,
1979 info->fs_root->node->start);
1980 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 1981 btrfs_header_generation(info->fs_root->node));
7c7e82a7 1982 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 1983 btrfs_header_level(info->fs_root->node));
7c7e82a7 1984 }
af31f5e5
CM		 1985
1986 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1987 btrfs_set_backup_dev_root_gen(root_backup,
1988 btrfs_header_generation(info->dev_root->node));
1989 btrfs_set_backup_dev_root_level(root_backup,
1990 btrfs_header_level(info->dev_root->node));
1991
1992 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1993 btrfs_set_backup_csum_root_gen(root_backup,
1994 btrfs_header_generation(info->csum_root->node));
1995 btrfs_set_backup_csum_root_level(root_backup,
1996 btrfs_header_level(info->csum_root->node));
1997
1998 btrfs_set_backup_total_bytes(root_backup,
1999 btrfs_super_total_bytes(info->super_copy));
2000 btrfs_set_backup_bytes_used(root_backup,
2001 btrfs_super_bytes_used(info->super_copy));
2002 btrfs_set_backup_num_devices(root_backup,
2003 btrfs_super_num_devices(info->super_copy));
2004
2005 /*
2006 * if we don't copy this out to the super_copy, it won't get remembered
2007 * for the next commit
2008 */
2009 memcpy(&info->super_copy->super_roots,
2010 &info->super_for_commit->super_roots,
2011 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
2012}
2013
2014/*
2015 * this copies info out of the root backup array and back into
2016 * the in-memory super block. It is meant to help iterate through
2017 * the array, so you send it the number of backups you've already
2018 * tried and the last backup index you used.
2019 *
2020 * this returns -1 when it has tried all the backups
2021 */
2022static noinline int next_root_backup(struct btrfs_fs_info *info,
2023 struct btrfs_super_block *super,
2024 int *num_backups_tried, int *backup_index)
2025{
2026 struct btrfs_root_backup *root_backup;
2027 int newest = *backup_index;
2028
2029 if (*num_backups_tried == 0) {
2030 u64 gen = btrfs_super_generation(super);
2031
2032 newest = find_newest_super_backup(info, gen);
2033 if (newest == -1)
2034 return -1;
2035
2036 *backup_index = newest;
2037 *num_backups_tried = 1;
2038 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2039 /* we've tried all the backups, all done */
2040 return -1;
2041 } else {
2042 /* jump to the next oldest backup */
2043 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2044 BTRFS_NUM_BACKUP_ROOTS;
2045 *backup_index = newest;
2046 *num_backups_tried += 1;
2047 }
2048 root_backup = super->super_roots + newest;
2049
2050 btrfs_set_super_generation(super,
2051 btrfs_backup_tree_root_gen(root_backup));
2052 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2053 btrfs_set_super_root_level(super,
2054 btrfs_backup_tree_root_level(root_backup));
2055 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2056
2057 /*
2058 * fixme: the total bytes and num_devices need to match or we should
2059 * need a fsck
2060 */
2061 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2062 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2063 return 0;
2064}
2065
7abadb64
LB		 2066/* helper to cleanup workers */
2067static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2068{
dc6e3209 2069 btrfs_destroy_workqueue(fs_info->fixup_workers);
afe3d242 2070 btrfs_destroy_workqueue(fs_info->delalloc_workers);
5cdc7ad3 2071 btrfs_destroy_workqueue(fs_info->workers);
fccb5d86
QW		 2072 btrfs_destroy_workqueue(fs_info->endio_workers);
2073 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2074 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
8b110e39 2075 btrfs_destroy_workqueue(fs_info->endio_repair_workers);
d05a33ac 2076 btrfs_destroy_workqueue(fs_info->rmw_workers);
fccb5d86
QW		 2077 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2078 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2079 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
a8c93d4e 2080 btrfs_destroy_workqueue(fs_info->submit_workers);
5b3bc44e 2081 btrfs_destroy_workqueue(fs_info->delayed_workers);
e66f0bb1 2082 btrfs_destroy_workqueue(fs_info->caching_workers);
736cfa15 2083 btrfs_destroy_workqueue(fs_info->readahead_workers);
a44903ab 2084 btrfs_destroy_workqueue(fs_info->flush_workers);
fc97fab0 2085 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
a79b7d4b 2086 btrfs_destroy_workqueue(fs_info->extent_workers);
7abadb64
LB		 2087}
2088
2e9f5954
R		 2089static void free_root_extent_buffers(struct btrfs_root *root)
2090{
2091 if (root) {
2092 free_extent_buffer(root->node);
2093 free_extent_buffer(root->commit_root);
2094 root->node = NULL;
2095 root->commit_root = NULL;
2096 }
2097}
2098
af31f5e5
CM		 2099/* helper to cleanup tree roots */
2100static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2101{
2e9f5954 2102 free_root_extent_buffers(info->tree_root);
655b09fe 2103
2e9f5954
R		 2104 free_root_extent_buffers(info->dev_root);
2105 free_root_extent_buffers(info->extent_root);
2106 free_root_extent_buffers(info->csum_root);
2107 free_root_extent_buffers(info->quota_root);
2108 free_root_extent_buffers(info->uuid_root);
2109 if (chunk_root)
2110 free_root_extent_buffers(info->chunk_root);
af31f5e5
CM		 2111}
2112
faa2dbf0 2113void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
171f6537
JB		 2114{
2115 int ret;
2116 struct btrfs_root *gang[8];
2117 int i;
2118
2119 while (!list_empty(&fs_info->dead_roots)) {
2120 gang[0] = list_entry(fs_info->dead_roots.next,
2121 struct btrfs_root, root_list);
2122 list_del(&gang[0]->root_list);
2123
27cdeb70 2124 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
cb517eab 2125 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
171f6537
JB		 2126 } else {
2127 free_extent_buffer(gang[0]->node);
2128 free_extent_buffer(gang[0]->commit_root);
b0feb9d9 2129 btrfs_put_fs_root(gang[0]);
171f6537
JB		 2130 }
2131 }
2132
2133 while (1) {
2134 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2135 (void **)gang, 0,
2136 ARRAY_SIZE(gang));
2137 if (!ret)
2138 break;
2139 for (i = 0; i < ret; i++)
cb517eab 2140 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
171f6537 2141 }
1a4319cc
LB		 2142
2143 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2144 btrfs_free_log_root_tree(NULL, fs_info);
2145 btrfs_destroy_pinned_extent(fs_info->tree_root,
2146 fs_info->pinned_extents);
2147 }
171f6537 2148}
af31f5e5 2149
638aa7ed
ES		 2150static void btrfs_init_scrub(struct btrfs_fs_info *fs_info)
2151{
2152 mutex_init(&fs_info->scrub_lock);
2153 atomic_set(&fs_info->scrubs_running, 0);
2154 atomic_set(&fs_info->scrub_pause_req, 0);
2155 atomic_set(&fs_info->scrubs_paused, 0);
2156 atomic_set(&fs_info->scrub_cancel_req, 0);
2157 init_waitqueue_head(&fs_info->scrub_pause_wait);
2158 fs_info->scrub_workers_refcnt = 0;
2159}
2160
779a65a4
ES		 2161static void btrfs_init_balance(struct btrfs_fs_info *fs_info)
2162{
2163 spin_lock_init(&fs_info->balance_lock);
2164 mutex_init(&fs_info->balance_mutex);
2165 atomic_set(&fs_info->balance_running, 0);
2166 atomic_set(&fs_info->balance_pause_req, 0);
2167 atomic_set(&fs_info->balance_cancel_req, 0);
2168 fs_info->balance_ctl = NULL;
2169 init_waitqueue_head(&fs_info->balance_wait_q);
2170}
2171
ad2b2c80
AV		 2172int open_ctree(struct super_block *sb,
2173 struct btrfs_fs_devices *fs_devices,
2174 char *options)
2e635a27 2175{
db94535d
CM		 2176 u32 sectorsize;
2177 u32 nodesize;
87ee04eb 2178 u32 stripesize;
84234f3a 2179 u64 generation;
f2b636e8 2180 u64 features;
3de4586c 2181 struct btrfs_key location;
a061fc8d 2182 struct buffer_head *bh;
4d34b278 2183 struct btrfs_super_block *disk_super;
815745cf 2184 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2185 struct btrfs_root *tree_root;
4d34b278
ID		 2186 struct btrfs_root *extent_root;
2187 struct btrfs_root *csum_root;
2188 struct btrfs_root *chunk_root;
2189 struct btrfs_root *dev_root;
bcef60f2 2190 struct btrfs_root *quota_root;
f7a81ea4 2191 struct btrfs_root *uuid_root;
e02119d5 2192 struct btrfs_root *log_tree_root;
eb60ceac 2193 int ret;
e58ca020 2194 int err = -EINVAL;
af31f5e5
CM		 2195 int num_backups_tried = 0;
2196 int backup_index = 0;
5cdc7ad3
QW		 2197 int max_active;
2198 int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
70f80175
SB		 2199 bool create_uuid_tree;
2200 bool check_uuid_tree;
4543df7e 2201
f84a8bd6 2202 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
6f07e42e 2203 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
cb517eab 2204 if (!tree_root || !chunk_root) {
39279cc3
CM		 2205 err = -ENOMEM;
2206 goto fail;
2207 }
76dda93c
YZ		 2208
2209 ret = init_srcu_struct(&fs_info->subvol_srcu);
2210 if (ret) {
2211 err = ret;
2212 goto fail;
2213 }
2214
2215 ret = setup_bdi(fs_info, &fs_info->bdi);
2216 if (ret) {
2217 err = ret;
2218 goto fail_srcu;
2219 }
2220
908c7f19 2221 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL);
e2d84521
MX		 2222 if (ret) {
2223 err = ret;
2224 goto fail_bdi;
2225 }
2226 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2227 (1 + ilog2(nr_cpu_ids));
2228
908c7f19 2229 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL);
963d678b
MX		 2230 if (ret) {
2231 err = ret;
2232 goto fail_dirty_metadata_bytes;
2233 }
2234
908c7f19 2235 ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL);
c404e0dc
MX		 2236 if (ret) {
2237 err = ret;
2238 goto fail_delalloc_bytes;
2239 }
2240
76dda93c
YZ		 2241 fs_info->btree_inode = new_inode(sb);
2242 if (!fs_info->btree_inode) {
2243 err = -ENOMEM;
c404e0dc 2244 goto fail_bio_counter;
76dda93c
YZ		 2245 }
2246
a6591715 2247 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2248
76dda93c 2249 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
f28491e0 2250 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
8fd17795 2251 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2252 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2253 INIT_LIST_HEAD(&fs_info->delayed_iputs);
eb73c1b7 2254 INIT_LIST_HEAD(&fs_info->delalloc_roots);
11833d66 2255 INIT_LIST_HEAD(&fs_info->caching_block_groups);
eb73c1b7 2256 spin_lock_init(&fs_info->delalloc_root_lock);
a4abeea4 2257 spin_lock_init(&fs_info->trans_lock);
76dda93c 2258 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2259 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2260 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 2261 spin_lock_init(&fs_info->free_chunk_lock);
f29021b2 2262 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2263 spin_lock_init(&fs_info->super_lock);
fcebe456 2264 spin_lock_init(&fs_info->qgroup_op_lock);
f28491e0 2265 spin_lock_init(&fs_info->buffer_lock);
47ab2a6c 2266 spin_lock_init(&fs_info->unused_bgs_lock);
d4b450cd 2267 mutex_init(&fs_info->unused_bg_unpin_mutex);
f29021b2 2268 rwlock_init(&fs_info->tree_mod_log_lock);
7585717f 2269 mutex_init(&fs_info->reloc_mutex);
573bfb72 2270 mutex_init(&fs_info->delalloc_root_mutex);
de98ced9 2271 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2272
58176a96 2273 init_completion(&fs_info->kobj_unregister);
0b86a832 2274 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2275 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2276 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
47ab2a6c 2277 INIT_LIST_HEAD(&fs_info->unused_bgs);
0b86a832 2278 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX		 2279 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2280 BTRFS_BLOCK_RSV_GLOBAL);
2281 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2282 BTRFS_BLOCK_RSV_DELALLOC);
2283 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2284 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2285 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2286 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2287 BTRFS_BLOCK_RSV_DELOPS);
cb03c743 2288 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 2289 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 2290 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 2291 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 2292 atomic_set(&fs_info->defrag_running, 0);
fcebe456 2293 atomic_set(&fs_info->qgroup_op_seq, 0);
fc36ed7e 2294 atomic64_set(&fs_info->tree_mod_seq, 0);
e20d96d6 2295 fs_info->sb = sb;
95ac567a 2296 fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE;
9ed74f2d 2297 fs_info->metadata_ratio = 0;
4cb5300b 2298 fs_info->defrag_inodes = RB_ROOT;
2bf64758 2299 fs_info->free_chunk_space = 0;
f29021b2 2300 fs_info->tree_mod_log = RB_ROOT;
8b87dc17 2301 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
0a2b2a84 2302 fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
90519d66
AJ		 2303 /* readahead state */
2304 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2305 spin_lock_init(&fs_info->reada_lock);
c8b97818 2306
b34b086c
CM		 2307 fs_info->thread_pool_size = min_t(unsigned long,
2308 num_online_cpus() + 2, 8);
0afbaf8c 2309
199c2a9c
MX		 2310 INIT_LIST_HEAD(&fs_info->ordered_roots);
2311 spin_lock_init(&fs_info->ordered_root_lock);
16cdcec7
MX		 2312 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2313 GFP_NOFS);
2314 if (!fs_info->delayed_root) {
2315 err = -ENOMEM;
2316 goto fail_iput;
2317 }
2318 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2319
638aa7ed 2320 btrfs_init_scrub(fs_info);
c404e0dc 2321 init_waitqueue_head(&fs_info->replace_wait);
21adbd5c
SB		 2322#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2323 fs_info->check_integrity_print_mask = 0;
2324#endif
779a65a4 2325 btrfs_init_balance(fs_info);
21c7e756 2326 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
a2de733c 2327
a061fc8d
CM		 2328 sb->s_blocksize = 4096;
2329 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 2330 sb->s_bdi = &fs_info->bdi;
a061fc8d 2331
76dda93c 2332 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
bfe86848 2333 set_nlink(fs_info->btree_inode, 1);
0afbaf8c
CM		 2334 /*
2335 * we set the i_size on the btree inode to the max possible int.
2336 * the real end of the address space is determined by all of
2337 * the devices in the system
2338 */
2339 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 2340 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 2341 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2342
5d4f98a2 2343 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 2344 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 2345 fs_info->btree_inode->i_mapping);
0b32f4bb 2346 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
a8067e02 2347 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM		 2348
2349 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 2350
76dda93c
YZ		 2351 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2352 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2353 sizeof(struct btrfs_key));
72ac3c0d
JB		 2354 set_bit(BTRFS_INODE_DUMMY,
2355 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
778ba82b 2356 btrfs_insert_inode_hash(fs_info->btree_inode);
76dda93c 2357
0f9dd46c 2358 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2359 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2360 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2361
11833d66 2362 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 2363 fs_info->btree_inode->i_mapping);
11833d66 2364 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 2365 fs_info->btree_inode->i_mapping);
11833d66 2366 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 2367 fs_info->do_barriers = 1;
e18e4809 2368
39279cc3 2369
5a3f23d5 2370 mutex_init(&fs_info->ordered_operations_mutex);
9ffba8cd 2371 mutex_init(&fs_info->ordered_extent_flush_mutex);
e02119d5 2372 mutex_init(&fs_info->tree_log_mutex);
925baedd 2373 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 2374 mutex_init(&fs_info->transaction_kthread_mutex);
2375 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2376 mutex_init(&fs_info->volume_mutex);
9e351cc8 2377 init_rwsem(&fs_info->commit_root_sem);
c71bf099 2378 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2379 init_rwsem(&fs_info->subvol_sem);
803b2f54 2380 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
e922e087
SB		 2381 fs_info->dev_replace.lock_owner = 0;
2382 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2383 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2384 mutex_init(&fs_info->dev_replace.lock_management_lock);
2385 mutex_init(&fs_info->dev_replace.lock);
fa9c0d79 2386
416ac51d 2387 spin_lock_init(&fs_info->qgroup_lock);
f2f6ed3d 2388 mutex_init(&fs_info->qgroup_ioctl_lock);
416ac51d 2389 fs_info->qgroup_tree = RB_ROOT;
fcebe456 2390 fs_info->qgroup_op_tree = RB_ROOT;
416ac51d
AJ		 2391 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2392 fs_info->qgroup_seq = 1;
2393 fs_info->quota_enabled = 0;
2394 fs_info->pending_quota_state = 0;
1e8f9158 2395 fs_info->qgroup_ulist = NULL;
2f232036 2396 mutex_init(&fs_info->qgroup_rescan_lock);
416ac51d 2397
fa9c0d79
CM		 2398 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2399 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2400
e6dcd2dc 2401 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2402 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2403 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2404 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2405
04216820
FM		 2406 INIT_LIST_HEAD(&fs_info->pinned_chunks);
2407
53b381b3
DW		 2408 ret = btrfs_alloc_stripe_hash_table(fs_info);
2409 if (ret) {
83c8266a 2410 err = ret;
53b381b3
DW		 2411 goto fail_alloc;
2412 }
2413
707e8a07 2414 __setup_root(4096, 4096, 4096, tree_root,
2c90e5d6 2415 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2416
3c4bb26b 2417 invalidate_bdev(fs_devices->latest_bdev);
1104a885
DS		 2418
2419 /*
2420 * Read super block and check the signature bytes only
2421 */
a512bbf8 2422 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY		 2423 if (!bh) {
2424 err = -EINVAL;
16cdcec7 2425 goto fail_alloc;
20b45077 2426 }
39279cc3 2427
1104a885
DS		 2428 /*
2429 * We want to check superblock checksum, the type is stored inside.
2430 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2431 */
2432 if (btrfs_check_super_csum(bh->b_data)) {
efe120a0 2433 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
1104a885
DS		 2434 err = -EINVAL;
2435 goto fail_alloc;
2436 }
2437
2438 /*
2439 * super_copy is zeroed at allocation time and we never touch the
2440 * following bytes up to INFO_SIZE, the checksum is calculated from
2441 * the whole block of INFO_SIZE
2442 */
6c41761f
DS		 2443 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2444 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2445 sizeof(*fs_info->super_for_commit));
a061fc8d 2446 brelse(bh);
5f39d397 2447
6c41761f 2448 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2449
1104a885
DS		 2450 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2451 if (ret) {
efe120a0 2452 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
1104a885
DS		 2453 err = -EINVAL;
2454 goto fail_alloc;
2455 }
2456
6c41761f 2457 disk_super = fs_info->super_copy;
0f7d52f4 2458 if (!btrfs_super_root(disk_super))
16cdcec7 2459 goto fail_alloc;
0f7d52f4 2460
acce952b 2461 /* check FS state, whether FS is broken. */
87533c47
MX		 2462 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2463 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2464
af31f5e5
CM		 2465 /*
2466 * run through our array of backup supers and setup
2467 * our ring pointer to the oldest one
2468 */
2469 generation = btrfs_super_generation(disk_super);
2470 find_oldest_super_backup(fs_info, generation);
2471
75e7cb7f
LB		 2472 /*
2473 * In the long term, we'll store the compression type in the super
2474 * block, and it'll be used for per file compression control.
2475 */
2476 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2477
2b82032c
YZ		 2478 ret = btrfs_parse_options(tree_root, options);
2479 if (ret) {
2480 err = ret;
16cdcec7 2481 goto fail_alloc;
2b82032c 2482 }
dfe25020 2483
f2b636e8
JB		 2484 features = btrfs_super_incompat_flags(disk_super) &
2485 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2486 if (features) {
2487 printk(KERN_ERR "BTRFS: couldn't mount because of "
2488 "unsupported optional features (%Lx).\n",
c1c9ff7c 2489 features);
f2b636e8 2490 err = -EINVAL;
16cdcec7 2491 goto fail_alloc;
f2b636e8
JB		 2492 }
2493
707e8a07
DS		 2494 /*
2495 * Leafsize and nodesize were always equal, this is only a sanity check.
2496 */
2497 if (le32_to_cpu(disk_super->__unused_leafsize) !=
727011e0
CM		 2498 btrfs_super_nodesize(disk_super)) {
2499 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2500 "blocksizes don't match. node %d leaf %d\n",
2501 btrfs_super_nodesize(disk_super),
707e8a07 2502 le32_to_cpu(disk_super->__unused_leafsize));
727011e0
CM		 2503 err = -EINVAL;
2504 goto fail_alloc;
2505 }
707e8a07 2506 if (btrfs_super_nodesize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
727011e0
CM		 2507 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2508 "blocksize (%d) was too large\n",
707e8a07 2509 btrfs_super_nodesize(disk_super));
727011e0
CM		 2510 err = -EINVAL;
2511 goto fail_alloc;
2512 }
2513
5d4f98a2 2514 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2515 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
69e380d1 2516 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2517 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
727011e0 2518
3173a18f 2519 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
5efa0490 2520 printk(KERN_INFO "BTRFS: has skinny extents\n");
3173a18f 2521
727011e0
CM		 2522 /*
2523 * flag our filesystem as having big metadata blocks if
2524 * they are bigger than the page size
2525 */
707e8a07 2526 if (btrfs_super_nodesize(disk_super) > PAGE_CACHE_SIZE) {
727011e0 2527 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
efe120a0 2528 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
727011e0
CM		 2529 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2530 }
2531
bc3f116f 2532 nodesize = btrfs_super_nodesize(disk_super);
bc3f116f
CM		 2533 sectorsize = btrfs_super_sectorsize(disk_super);
2534 stripesize = btrfs_super_stripesize(disk_super);
707e8a07 2535 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids));
963d678b 2536 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f
CM		 2537
2538 /*
2539 * mixed block groups end up with duplicate but slightly offset
2540 * extent buffers for the same range. It leads to corruptions
2541 */
2542 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
707e8a07 2543 (sectorsize != nodesize)) {
aa8ee312 2544 printk(KERN_ERR "BTRFS: unequal leaf/node/sector sizes "
bc3f116f
CM		 2545 "are not allowed for mixed block groups on %s\n",
2546 sb->s_id);
2547 goto fail_alloc;
2548 }
2549
ceda0864
MX		 2550 /*
2551 * Needn't use the lock because there is no other task which will
2552 * update the flag.
2553 */
a6fa6fae 2554 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2555
f2b636e8
JB		 2556 features = btrfs_super_compat_ro_flags(disk_super) &
2557 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2558 if (!(sb->s_flags & MS_RDONLY) && features) {
2559 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2560 "unsupported option features (%Lx).\n",
c1c9ff7c 2561 features);
f2b636e8 2562 err = -EINVAL;
16cdcec7 2563 goto fail_alloc;
f2b636e8 2564 }
61d92c32 2565
5cdc7ad3 2566 max_active = fs_info->thread_pool_size;
61d92c32 2567
5cdc7ad3
QW		 2568 fs_info->workers =
2569 btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2570 max_active, 16);
c8b97818 2571
afe3d242
QW		 2572 fs_info->delalloc_workers =
2573 btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
771ed689 2574
a44903ab
QW		 2575 fs_info->flush_workers =
2576 btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
8ccf6f19 2577
e66f0bb1
QW		 2578 fs_info->caching_workers =
2579 btrfs_alloc_workqueue("cache", flags, max_active, 0);
61b49440 2580
a8c93d4e
QW		 2581 /*
2582 * a higher idle thresh on the submit workers makes it much more
61b49440
CM		 2583 * likely that bios will be send down in a sane order to the
2584 * devices
2585 */
a8c93d4e
QW		 2586 fs_info->submit_workers =
2587 btrfs_alloc_workqueue("submit", flags,
2588 min_t(u64, fs_devices->num_devices,
2589 max_active), 64);
53863232 2590
dc6e3209
QW		 2591 fs_info->fixup_workers =
2592 btrfs_alloc_workqueue("fixup", flags, 1, 0);
61b49440
CM		 2593
2594 /*
2595 * endios are largely parallel and should have a very
2596 * low idle thresh
2597 */
fccb5d86
QW		 2598 fs_info->endio_workers =
2599 btrfs_alloc_workqueue("endio", flags, max_active, 4);
2600 fs_info->endio_meta_workers =
2601 btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2602 fs_info->endio_meta_write_workers =
2603 btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2604 fs_info->endio_raid56_workers =
2605 btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
8b110e39
MX		 2606 fs_info->endio_repair_workers =
2607 btrfs_alloc_workqueue("endio-repair", flags, 1, 0);
d05a33ac
QW		 2608 fs_info->rmw_workers =
2609 btrfs_alloc_workqueue("rmw", flags, max_active, 2);
fccb5d86
QW		 2610 fs_info->endio_write_workers =
2611 btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2612 fs_info->endio_freespace_worker =
2613 btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
5b3bc44e
QW		 2614 fs_info->delayed_workers =
2615 btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
736cfa15
QW		 2616 fs_info->readahead_workers =
2617 btrfs_alloc_workqueue("readahead", flags, max_active, 2);
fc97fab0
QW		 2618 fs_info->qgroup_rescan_workers =
2619 btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
a79b7d4b
CM		 2620 fs_info->extent_workers =
2621 btrfs_alloc_workqueue("extent-refs", flags,
2622 min_t(u64, fs_devices->num_devices,
2623 max_active), 8);
61b49440 2624
a8c93d4e 2625 if (!(fs_info->workers && fs_info->delalloc_workers &&
fccb5d86
QW		 2626 fs_info->submit_workers && fs_info->flush_workers &&
2627 fs_info->endio_workers && fs_info->endio_meta_workers &&
2628 fs_info->endio_meta_write_workers &&
8b110e39 2629 fs_info->endio_repair_workers &&
fccb5d86 2630 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
e66f0bb1 2631 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
dc6e3209 2632 fs_info->caching_workers && fs_info->readahead_workers &&
fc97fab0 2633 fs_info->fixup_workers && fs_info->delayed_workers &&
56094eec 2634 fs_info->extent_workers &&
fc97fab0 2635 fs_info->qgroup_rescan_workers)) {
fed425c7 2636 err = -ENOMEM;
0dc3b84a
JB		 2637 goto fail_sb_buffer;
2638 }
4543df7e 2639
4575c9cc 2640 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 2641 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2642 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 2643
db94535d 2644 tree_root->nodesize = nodesize;
db94535d 2645 tree_root->sectorsize = sectorsize;
87ee04eb 2646 tree_root->stripesize = stripesize;
a061fc8d
CM		 2647
2648 sb->s_blocksize = sectorsize;
2649 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 2650
3cae210f 2651 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
aa8ee312 2652 printk(KERN_ERR "BTRFS: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 2653 goto fail_sb_buffer;
2654 }
19c00ddc 2655
8d082fb7 2656 if (sectorsize != PAGE_SIZE) {
aa8ee312 2657 printk(KERN_ERR "BTRFS: incompatible sector size (%lu) "
8d082fb7 2658 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
941b2ddf
KM		 2659 goto fail_sb_buffer;
2660 }
2661
925baedd 2662 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 2663 ret = btrfs_read_sys_array(tree_root);
925baedd 2664 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2665 if (ret) {
aa8ee312 2666 printk(KERN_ERR "BTRFS: failed to read the system "
d397712b 2667 "array on %s\n", sb->s_id);
5d4f98a2 2668 goto fail_sb_buffer;
84eed90f 2669 }
0b86a832 2670
84234f3a 2671 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832 2672
707e8a07
DS		 2673 __setup_root(nodesize, sectorsize, stripesize, chunk_root,
2674 fs_info, BTRFS_CHUNK_TREE_OBJECTID);
0b86a832
CM		 2675
2676 chunk_root->node = read_tree_block(chunk_root,
2677 btrfs_super_chunk_root(disk_super),
ce86cd59 2678 generation);
416bc658
JB		 2679 if (!chunk_root->node ||
2680 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
aa8ee312 2681 printk(KERN_ERR "BTRFS: failed to read chunk root on %s\n",
83121942 2682 sb->s_id);
af31f5e5 2683 goto fail_tree_roots;
83121942 2684 }
5d4f98a2
YZ		 2685 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2686 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2687
e17cade2 2688 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
b308bc2f 2689 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
e17cade2 2690
0b86a832 2691 ret = btrfs_read_chunk_tree(chunk_root);
2b82032c 2692 if (ret) {
aa8ee312 2693 printk(KERN_ERR "BTRFS: failed to read chunk tree on %s\n",
d397712b 2694 sb->s_id);
af31f5e5 2695 goto fail_tree_roots;
2b82032c 2696 }
0b86a832 2697
8dabb742
SB		 2698 /*
2699 * keep the device that is marked to be the target device for the
2700 * dev_replace procedure
2701 */
9eaed21e 2702 btrfs_close_extra_devices(fs_devices, 0);
dfe25020 2703
a6b0d5c8 2704 if (!fs_devices->latest_bdev) {
aa8ee312 2705 printk(KERN_ERR "BTRFS: failed to read devices on %s\n",
a6b0d5c8
CM		 2706 sb->s_id);
2707 goto fail_tree_roots;
2708 }
2709
af31f5e5 2710retry_root_backup:
84234f3a 2711 generation = btrfs_super_generation(disk_super);
0b86a832 2712
e20d96d6 2713 tree_root->node = read_tree_block(tree_root,
db94535d 2714 btrfs_super_root(disk_super),
ce86cd59 2715 generation);
af31f5e5
CM		 2716 if (!tree_root->node ||
2717 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
efe120a0 2718 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
83121942 2719 sb->s_id);
af31f5e5
CM		 2720
2721 goto recovery_tree_root;
83121942 2722 }
af31f5e5 2723
5d4f98a2
YZ		 2724 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2725 tree_root->commit_root = btrfs_root_node(tree_root);
69e9c6c6 2726 btrfs_set_root_refs(&tree_root->root_item, 1);
db94535d 2727
cb517eab
MX		 2728 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2729 location.type = BTRFS_ROOT_ITEM_KEY;
2730 location.offset = 0;
2731
2732 extent_root = btrfs_read_tree_root(tree_root, &location);
2733 if (IS_ERR(extent_root)) {
2734 ret = PTR_ERR(extent_root);
af31f5e5 2735 goto recovery_tree_root;
cb517eab 2736 }
27cdeb70 2737 set_bit(BTRFS_ROOT_TRACK_DIRTY, &extent_root->state);
cb517eab 2738 fs_info->extent_root = extent_root;
0b86a832 2739
cb517eab
MX		 2740 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2741 dev_root = btrfs_read_tree_root(tree_root, &location);
2742 if (IS_ERR(dev_root)) {
2743 ret = PTR_ERR(dev_root);
af31f5e5 2744 goto recovery_tree_root;
cb517eab 2745 }
27cdeb70 2746 set_bit(BTRFS_ROOT_TRACK_DIRTY, &dev_root->state);
cb517eab
MX		 2747 fs_info->dev_root = dev_root;
2748 btrfs_init_devices_late(fs_info);
3768f368 2749
cb517eab
MX		 2750 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2751 csum_root = btrfs_read_tree_root(tree_root, &location);
2752 if (IS_ERR(csum_root)) {
2753 ret = PTR_ERR(csum_root);
af31f5e5 2754 goto recovery_tree_root;
cb517eab 2755 }
27cdeb70 2756 set_bit(BTRFS_ROOT_TRACK_DIRTY, &csum_root->state);
cb517eab 2757 fs_info->csum_root = csum_root;
d20f7043 2758
cb517eab
MX		 2759 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2760 quota_root = btrfs_read_tree_root(tree_root, &location);
2761 if (!IS_ERR(quota_root)) {
27cdeb70 2762 set_bit(BTRFS_ROOT_TRACK_DIRTY, &quota_root->state);
bcef60f2
AJ		 2763 fs_info->quota_enabled = 1;
2764 fs_info->pending_quota_state = 1;
cb517eab 2765 fs_info->quota_root = quota_root;
bcef60f2
AJ		 2766 }
2767
f7a81ea4
SB		 2768 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2769 uuid_root = btrfs_read_tree_root(tree_root, &location);
2770 if (IS_ERR(uuid_root)) {
2771 ret = PTR_ERR(uuid_root);
2772 if (ret != -ENOENT)
2773 goto recovery_tree_root;
2774 create_uuid_tree = true;
70f80175 2775 check_uuid_tree = false;
f7a81ea4 2776 } else {
27cdeb70 2777 set_bit(BTRFS_ROOT_TRACK_DIRTY, &uuid_root->state);
f7a81ea4 2778 fs_info->uuid_root = uuid_root;
70f80175
SB		 2779 create_uuid_tree = false;
2780 check_uuid_tree =
2781 generation != btrfs_super_uuid_tree_generation(disk_super);
f7a81ea4
SB		 2782 }
2783
8929ecfa
YZ		 2784 fs_info->generation = generation;
2785 fs_info->last_trans_committed = generation;
8929ecfa 2786
68310a5e
ID		 2787 ret = btrfs_recover_balance(fs_info);
2788 if (ret) {
aa8ee312 2789 printk(KERN_ERR "BTRFS: failed to recover balance\n");
68310a5e
ID		 2790 goto fail_block_groups;
2791 }
2792
733f4fbb
SB		 2793 ret = btrfs_init_dev_stats(fs_info);
2794 if (ret) {
efe120a0 2795 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
733f4fbb
SB		 2796 ret);
2797 goto fail_block_groups;
2798 }
2799
8dabb742
SB		 2800 ret = btrfs_init_dev_replace(fs_info);
2801 if (ret) {
efe120a0 2802 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
8dabb742
SB		 2803 goto fail_block_groups;
2804 }
2805
9eaed21e 2806 btrfs_close_extra_devices(fs_devices, 1);
8dabb742 2807
5ac1d209 2808 ret = btrfs_sysfs_add_one(fs_info);
c59021f8 2809 if (ret) {
efe120a0 2810 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
c59021f8 2811 goto fail_block_groups;
2812 }
2813
c59021f8 2814 ret = btrfs_init_space_info(fs_info);
2815 if (ret) {
efe120a0 2816 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
2365dd3c 2817 goto fail_sysfs;
c59021f8 2818 }
2819
1b1d1f66
JB		 2820 ret = btrfs_read_block_groups(extent_root);
2821 if (ret) {
efe120a0 2822 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
2365dd3c 2823 goto fail_sysfs;
1b1d1f66 2824 }
5af3e8cc
SB		 2825 fs_info->num_tolerated_disk_barrier_failures =
2826 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
292fd7fc
SB		 2827 if (fs_info->fs_devices->missing_devices >
2828 fs_info->num_tolerated_disk_barrier_failures &&
2829 !(sb->s_flags & MS_RDONLY)) {
efe120a0
FH		 2830 printk(KERN_WARNING "BTRFS: "
2831 "too many missing devices, writeable mount is not allowed\n");
2365dd3c 2832 goto fail_sysfs;
292fd7fc 2833 }
9078a3e1 2834
a74a4b97
CM		 2835 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2836 "btrfs-cleaner");
57506d50 2837 if (IS_ERR(fs_info->cleaner_kthread))
2365dd3c 2838 goto fail_sysfs;
a74a4b97
CM		 2839
2840 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2841 tree_root,
2842 "btrfs-transaction");
57506d50 2843 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2844 goto fail_cleaner;
a74a4b97 2845
c289811c
CM		 2846 if (!btrfs_test_opt(tree_root, SSD) &&
2847 !btrfs_test_opt(tree_root, NOSSD) &&
2848 !fs_info->fs_devices->rotating) {
efe120a0 2849 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
c289811c
CM		 2850 "mode\n");
2851 btrfs_set_opt(fs_info->mount_opt, SSD);
2852 }
2853
572d9ab7
DS		 2854 /*
2855 * Mount does not set all options immediatelly, we can do it now and do
2856 * not have to wait for transaction commit
2857 */
2858 btrfs_apply_pending_changes(fs_info);
3818aea2 2859
21adbd5c
SB		 2860#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2861 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2862 ret = btrfsic_mount(tree_root, fs_devices,
2863 btrfs_test_opt(tree_root,
2864 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2865 1 : 0,
2866 fs_info->check_integrity_print_mask);
2867 if (ret)
efe120a0 2868 printk(KERN_WARNING "BTRFS: failed to initialize"
21adbd5c
SB		 2869 " integrity check module %s\n", sb->s_id);
2870 }
2871#endif
bcef60f2
AJ		 2872 ret = btrfs_read_qgroup_config(fs_info);
2873 if (ret)
2874 goto fail_trans_kthread;
21adbd5c 2875
acce952b 2876 /* do not make disk changes in broken FS */
68ce9682 2877 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM		 2878 u64 bytenr = btrfs_super_log_root(disk_super);
2879
7c2ca468 2880 if (fs_devices->rw_devices == 0) {
efe120a0 2881 printk(KERN_WARNING "BTRFS: log replay required "
d397712b 2882 "on RO media\n");
7c2ca468 2883 err = -EIO;
bcef60f2 2884 goto fail_qgroup;
7c2ca468 2885 }
d18a2c44 2886
6f07e42e 2887 log_tree_root = btrfs_alloc_root(fs_info);
676e4c86
DC		 2888 if (!log_tree_root) {
2889 err = -ENOMEM;
bcef60f2 2890 goto fail_qgroup;
676e4c86 2891 }
e02119d5 2892
707e8a07 2893 __setup_root(nodesize, sectorsize, stripesize,
e02119d5
CM		 2894 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2895
2896 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a 2897 generation + 1);
416bc658
JB		 2898 if (!log_tree_root->node ||
2899 !extent_buffer_uptodate(log_tree_root->node)) {
efe120a0 2900 printk(KERN_ERR "BTRFS: failed to read log tree\n");
416bc658
JB		 2901 free_extent_buffer(log_tree_root->node);
2902 kfree(log_tree_root);
28c16cbb 2903 goto fail_qgroup;
416bc658 2904 }
79787eaa 2905 /* returns with log_tree_root freed on success */
e02119d5 2906 ret = btrfs_recover_log_trees(log_tree_root);
79787eaa
JM		 2907 if (ret) {
2908 btrfs_error(tree_root->fs_info, ret,
2909 "Failed to recover log tree");
2910 free_extent_buffer(log_tree_root->node);
2911 kfree(log_tree_root);
28c16cbb 2912 goto fail_qgroup;
79787eaa 2913 }
e556ce2c
YZ		 2914
2915 if (sb->s_flags & MS_RDONLY) {
79787eaa
JM		 2916 ret = btrfs_commit_super(tree_root);
2917 if (ret)
28c16cbb 2918 goto fail_qgroup;
e556ce2c 2919 }
e02119d5 2920 }
1a40e23b 2921
76dda93c 2922 ret = btrfs_find_orphan_roots(tree_root);
79787eaa 2923 if (ret)
28c16cbb 2924 goto fail_qgroup;
76dda93c 2925
7c2ca468 2926 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b 2927 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2 2928 if (ret)
28c16cbb 2929 goto fail_qgroup;
d68fc57b 2930
5f316481 2931 mutex_lock(&fs_info->cleaner_mutex);
5d4f98a2 2932 ret = btrfs_recover_relocation(tree_root);
5f316481 2933 mutex_unlock(&fs_info->cleaner_mutex);
d7ce5843
MX		 2934 if (ret < 0) {
2935 printk(KERN_WARNING
efe120a0 2936 "BTRFS: failed to recover relocation\n");
d7ce5843 2937 err = -EINVAL;
bcef60f2 2938 goto fail_qgroup;
d7ce5843 2939 }
7c2ca468 2940 }
1a40e23b 2941
3de4586c
CM		 2942 location.objectid = BTRFS_FS_TREE_OBJECTID;
2943 location.type = BTRFS_ROOT_ITEM_KEY;
cb517eab 2944 location.offset = 0;
3de4586c 2945
3de4586c 2946 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
3140c9a3
DC		 2947 if (IS_ERR(fs_info->fs_root)) {
2948 err = PTR_ERR(fs_info->fs_root);
bcef60f2 2949 goto fail_qgroup;
3140c9a3 2950 }
c289811c 2951
2b6ba629
ID		 2952 if (sb->s_flags & MS_RDONLY)
2953 return 0;
59641015 2954
2b6ba629
ID		 2955 down_read(&fs_info->cleanup_work_sem);
2956 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2957 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 2958 up_read(&fs_info->cleanup_work_sem);
2b6ba629
ID		 2959 close_ctree(tree_root);
2960 return ret;
2961 }
2962 up_read(&fs_info->cleanup_work_sem);
59641015 2963
2b6ba629
ID		 2964 ret = btrfs_resume_balance_async(fs_info);
2965 if (ret) {
efe120a0 2966 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
2b6ba629
ID		 2967 close_ctree(tree_root);
2968 return ret;
e3acc2a6
JB		 2969 }
2970
8dabb742
SB		 2971 ret = btrfs_resume_dev_replace_async(fs_info);
2972 if (ret) {
efe120a0 2973 pr_warn("BTRFS: failed to resume dev_replace\n");
8dabb742
SB		 2974 close_ctree(tree_root);
2975 return ret;
2976 }
2977
b382a324
JS		 2978 btrfs_qgroup_rescan_resume(fs_info);
2979
f7a81ea4 2980 if (create_uuid_tree) {
efe120a0 2981 pr_info("BTRFS: creating UUID tree\n");
f7a81ea4
SB		 2982 ret = btrfs_create_uuid_tree(fs_info);
2983 if (ret) {
efe120a0 2984 pr_warn("BTRFS: failed to create the UUID tree %d\n",
f7a81ea4
SB		 2985 ret);
2986 close_ctree(tree_root);
2987 return ret;
2988 }
f420ee1e
SB		 2989 } else if (check_uuid_tree ||
2990 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
efe120a0 2991 pr_info("BTRFS: checking UUID tree\n");
70f80175
SB		 2992 ret = btrfs_check_uuid_tree(fs_info);
2993 if (ret) {
efe120a0 2994 pr_warn("BTRFS: failed to check the UUID tree %d\n",
70f80175
SB		 2995 ret);
2996 close_ctree(tree_root);
2997 return ret;
2998 }
2999 } else {
3000 fs_info->update_uuid_tree_gen = 1;
f7a81ea4
SB		 3001 }
3002
47ab2a6c
JB		 3003 fs_info->open = 1;
3004
ad2b2c80 3005 return 0;
39279cc3 3006
bcef60f2
AJ		 3007fail_qgroup:
3008 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM		 3009fail_trans_kthread:
3010 kthread_stop(fs_info->transaction_kthread);
54067ae9 3011 btrfs_cleanup_transaction(fs_info->tree_root);
faa2dbf0 3012 btrfs_free_fs_roots(fs_info);
3f157a2f 3013fail_cleaner:
a74a4b97 3014 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 3015
3016 /*
3017 * make sure we're done with the btree inode before we stop our
3018 * kthreads
3019 */
3020 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 3021
2365dd3c
AJ		 3022fail_sysfs:
3023 btrfs_sysfs_remove_one(fs_info);
3024
1b1d1f66 3025fail_block_groups:
54067ae9 3026 btrfs_put_block_group_cache(fs_info);
1b1d1f66 3027 btrfs_free_block_groups(fs_info);
af31f5e5
CM		 3028
3029fail_tree_roots:
3030 free_root_pointers(fs_info, 1);
2b8195bb 3031 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 3032
39279cc3 3033fail_sb_buffer:
7abadb64 3034 btrfs_stop_all_workers(fs_info);
16cdcec7 3035fail_alloc:
4543df7e 3036fail_iput:
586e46e2
ID		 3037 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3038
4543df7e 3039 iput(fs_info->btree_inode);
c404e0dc
MX		 3040fail_bio_counter:
3041 percpu_counter_destroy(&fs_info->bio_counter);
963d678b
MX		 3042fail_delalloc_bytes:
3043 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX		 3044fail_dirty_metadata_bytes:
3045 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
ad081f14 3046fail_bdi:
7e662854 3047 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 3048fail_srcu:
3049 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 3050fail:
53b381b3 3051 btrfs_free_stripe_hash_table(fs_info);
586e46e2 3052 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 3053 return err;
af31f5e5
CM		 3054
3055recovery_tree_root:
af31f5e5
CM		 3056 if (!btrfs_test_opt(tree_root, RECOVERY))
3057 goto fail_tree_roots;
3058
3059 free_root_pointers(fs_info, 0);
3060
3061 /* don't use the log in recovery mode, it won't be valid */
3062 btrfs_set_super_log_root(disk_super, 0);
3063
3064 /* we can't trust the free space cache either */
3065 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3066
3067 ret = next_root_backup(fs_info, fs_info->super_copy,
3068 &num_backups_tried, &backup_index);
3069 if (ret == -1)
3070 goto fail_block_groups;
3071 goto retry_root_backup;
eb60ceac
CM		 3072}
3073
f2984462
CM		 3074static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3075{
f2984462
CM		 3076 if (uptodate) {
3077 set_buffer_uptodate(bh);
3078 } else {
442a4f63
SB		 3079 struct btrfs_device *device = (struct btrfs_device *)
3080 bh->b_private;
3081
efe120a0 3082 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
606686ee
JB		 3083 "I/O error on %s\n",
3084 rcu_str_deref(device->name));
1259ab75
CM		 3085 /* note, we dont' set_buffer_write_io_error because we have
3086 * our own ways of dealing with the IO errors
3087 */
f2984462 3088 clear_buffer_uptodate(bh);
442a4f63 3089 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM		 3090 }
3091 unlock_buffer(bh);
3092 put_bh(bh);
3093}
3094
a512bbf8
YZ		 3095struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3096{
3097 struct buffer_head *bh;
3098 struct buffer_head *latest = NULL;
3099 struct btrfs_super_block *super;
3100 int i;
3101 u64 transid = 0;
3102 u64 bytenr;
3103
3104 /* we would like to check all the supers, but that would make
3105 * a btrfs mount succeed after a mkfs from a different FS.
3106 * So, we need to add a special mount option to scan for
3107 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3108 */
3109 for (i = 0; i < 1; i++) {
3110 bytenr = btrfs_sb_offset(i);
8068a47e
AJ		 3111 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3112 i_size_read(bdev->bd_inode))
a512bbf8 3113 break;
8068a47e
AJ		 3114 bh = __bread(bdev, bytenr / 4096,
3115 BTRFS_SUPER_INFO_SIZE);
a512bbf8
YZ		 3116 if (!bh)
3117 continue;
3118
3119 super = (struct btrfs_super_block *)bh->b_data;
3120 if (btrfs_super_bytenr(super) != bytenr ||
3cae210f 3121 btrfs_super_magic(super) != BTRFS_MAGIC) {
a512bbf8
YZ		 3122 brelse(bh);
3123 continue;
3124 }
3125
3126 if (!latest || btrfs_super_generation(super) > transid) {
3127 brelse(latest);
3128 latest = bh;
3129 transid = btrfs_super_generation(super);
3130 } else {
3131 brelse(bh);
3132 }
3133 }
3134 return latest;
3135}
3136
4eedeb75
HH		 3137/*
3138 * this should be called twice, once with wait == 0 and
3139 * once with wait == 1. When wait == 0 is done, all the buffer heads
3140 * we write are pinned.
3141 *
3142 * They are released when wait == 1 is done.
3143 * max_mirrors must be the same for both runs, and it indicates how
3144 * many supers on this one device should be written.
3145 *
3146 * max_mirrors == 0 means to write them all.
3147 */
a512bbf8
YZ		 3148static int write_dev_supers(struct btrfs_device *device,
3149 struct btrfs_super_block *sb,
3150 int do_barriers, int wait, int max_mirrors)
3151{
3152 struct buffer_head *bh;
3153 int i;
3154 int ret;
3155 int errors = 0;
3156 u32 crc;
3157 u64 bytenr;
a512bbf8
YZ		 3158
3159 if (max_mirrors == 0)
3160 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3161
a512bbf8
YZ		 3162 for (i = 0; i < max_mirrors; i++) {
3163 bytenr = btrfs_sb_offset(i);
935e5cc9
MX		 3164 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3165 device->commit_total_bytes)
a512bbf8
YZ		 3166 break;
3167
3168 if (wait) {
3169 bh = __find_get_block(device->bdev, bytenr / 4096,
3170 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3171 if (!bh) {
3172 errors++;
3173 continue;
3174 }
a512bbf8 3175 wait_on_buffer(bh);
4eedeb75
HH		 3176 if (!buffer_uptodate(bh))
3177 errors++;
3178
3179 /* drop our reference */
3180 brelse(bh);
3181
3182 /* drop the reference from the wait == 0 run */
3183 brelse(bh);
3184 continue;
a512bbf8
YZ		 3185 } else {
3186 btrfs_set_super_bytenr(sb, bytenr);
3187
3188 crc = ~(u32)0;
b0496686 3189 crc = btrfs_csum_data((char *)sb +
a512bbf8
YZ		 3190 BTRFS_CSUM_SIZE, crc,
3191 BTRFS_SUPER_INFO_SIZE -
3192 BTRFS_CSUM_SIZE);
3193 btrfs_csum_final(crc, sb->csum);
3194
4eedeb75
HH		 3195 /*
3196 * one reference for us, and we leave it for the
3197 * caller
3198 */
a512bbf8
YZ		 3199 bh = __getblk(device->bdev, bytenr / 4096,
3200 BTRFS_SUPER_INFO_SIZE);
634554dc 3201 if (!bh) {
efe120a0 3202 printk(KERN_ERR "BTRFS: couldn't get super "
634554dc
JB		 3203 "buffer head for bytenr %Lu\n", bytenr);
3204 errors++;
3205 continue;
3206 }
3207
a512bbf8
YZ		 3208 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3209
4eedeb75 3210 /* one reference for submit_bh */
a512bbf8 3211 get_bh(bh);
4eedeb75
HH		 3212
3213 set_buffer_uptodate(bh);
a512bbf8
YZ		 3214 lock_buffer(bh);
3215 bh->b_end_io = btrfs_end_buffer_write_sync;
442a4f63 3216 bh->b_private = device;
a512bbf8
YZ		 3217 }
3218
387125fc
CM		 3219 /*
3220 * we fua the first super. The others we allow
3221 * to go down lazy.
3222 */
e8117c26
WS		 3223 if (i == 0)
3224 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3225 else
3226 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
4eedeb75 3227 if (ret)
a512bbf8 3228 errors++;
a512bbf8
YZ		 3229 }
3230 return errors < i ? 0 : -1;
3231}
3232
387125fc
CM		 3233/*
3234 * endio for the write_dev_flush, this will wake anyone waiting
3235 * for the barrier when it is done
3236 */
3237static void btrfs_end_empty_barrier(struct bio *bio, int err)
3238{
3239 if (err) {
3240 if (err == -EOPNOTSUPP)
3241 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3242 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3243 }
3244 if (bio->bi_private)
3245 complete(bio->bi_private);
3246 bio_put(bio);
3247}
3248
3249/*
3250 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3251 * sent down. With wait == 1, it waits for the previous flush.
3252 *
3253 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3254 * capable
3255 */
3256static int write_dev_flush(struct btrfs_device *device, int wait)
3257{
3258 struct bio *bio;
3259 int ret = 0;
3260
3261 if (device->nobarriers)
3262 return 0;
3263
3264 if (wait) {
3265 bio = device->flush_bio;
3266 if (!bio)
3267 return 0;
3268
3269 wait_for_completion(&device->flush_wait);
3270
3271 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
efe120a0 3272 printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
606686ee 3273 rcu_str_deref(device->name));
387125fc 3274 device->nobarriers = 1;
5af3e8cc 3275 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
387125fc 3276 ret = -EIO;
5af3e8cc
SB		 3277 btrfs_dev_stat_inc_and_print(device,
3278 BTRFS_DEV_STAT_FLUSH_ERRS);
387125fc
CM		 3279 }
3280
3281 /* drop the reference from the wait == 0 run */
3282 bio_put(bio);
3283 device->flush_bio = NULL;
3284
3285 return ret;
3286 }
3287
3288 /*
3289 * one reference for us, and we leave it for the
3290 * caller
3291 */
9c017abc 3292 device->flush_bio = NULL;
9be3395b 3293 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
387125fc
CM		 3294 if (!bio)
3295 return -ENOMEM;
3296
3297 bio->bi_end_io = btrfs_end_empty_barrier;
3298 bio->bi_bdev = device->bdev;
3299 init_completion(&device->flush_wait);
3300 bio->bi_private = &device->flush_wait;
3301 device->flush_bio = bio;
3302
3303 bio_get(bio);
21adbd5c 3304 btrfsic_submit_bio(WRITE_FLUSH, bio);
387125fc
CM		 3305
3306 return 0;
3307}
3308
3309/*
3310 * send an empty flush down to each device in parallel,
3311 * then wait for them
3312 */
3313static int barrier_all_devices(struct btrfs_fs_info *info)
3314{
3315 struct list_head *head;
3316 struct btrfs_device *dev;
5af3e8cc
SB		 3317 int errors_send = 0;
3318 int errors_wait = 0;
387125fc
CM		 3319 int ret;
3320
3321 /* send down all the barriers */
3322 head = &info->fs_devices->devices;
3323 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3324 if (dev->missing)
3325 continue;
387125fc 3326 if (!dev->bdev) {
5af3e8cc 3327 errors_send++;
387125fc
CM		 3328 continue;
3329 }
3330 if (!dev->in_fs_metadata || !dev->writeable)
3331 continue;
3332
3333 ret = write_dev_flush(dev, 0);
3334 if (ret)
5af3e8cc 3335 errors_send++;
387125fc
CM		 3336 }
3337
3338 /* wait for all the barriers */
3339 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3340 if (dev->missing)
3341 continue;
387125fc 3342 if (!dev->bdev) {
5af3e8cc 3343 errors_wait++;
387125fc
CM		 3344 continue;
3345 }
3346 if (!dev->in_fs_metadata || !dev->writeable)
3347 continue;
3348
3349 ret = write_dev_flush(dev, 1);
3350 if (ret)
5af3e8cc 3351 errors_wait++;
387125fc 3352 }
5af3e8cc
SB		 3353 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3354 errors_wait > info->num_tolerated_disk_barrier_failures)
387125fc
CM		 3355 return -EIO;
3356 return 0;
3357}
3358
5af3e8cc
SB		 3359int btrfs_calc_num_tolerated_disk_barrier_failures(
3360 struct btrfs_fs_info *fs_info)
3361{
3362 struct btrfs_ioctl_space_info space;
3363 struct btrfs_space_info *sinfo;
3364 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3365 BTRFS_BLOCK_GROUP_SYSTEM,
3366 BTRFS_BLOCK_GROUP_METADATA,
3367 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3368 int num_types = 4;
3369 int i;
3370 int c;
3371 int num_tolerated_disk_barrier_failures =
3372 (int)fs_info->fs_devices->num_devices;
3373
3374 for (i = 0; i < num_types; i++) {
3375 struct btrfs_space_info *tmp;
3376
3377 sinfo = NULL;
3378 rcu_read_lock();
3379 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3380 if (tmp->flags == types[i]) {
3381 sinfo = tmp;
3382 break;
3383 }
3384 }
3385 rcu_read_unlock();
3386
3387 if (!sinfo)
3388 continue;
3389
3390 down_read(&sinfo->groups_sem);
3391 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3392 if (!list_empty(&sinfo->block_groups[c])) {
3393 u64 flags;
3394
3395 btrfs_get_block_group_info(
3396 &sinfo->block_groups[c], &space);
3397 if (space.total_bytes == 0 ||
3398 space.used_bytes == 0)
3399 continue;
3400 flags = space.flags;
3401 /*
3402 * return
3403 * 0: if dup, single or RAID0 is configured for
3404 * any of metadata, system or data, else
3405 * 1: if RAID5 is configured, or if RAID1 or
3406 * RAID10 is configured and only two mirrors
3407 * are used, else
3408 * 2: if RAID6 is configured, else
3409 * num_mirrors - 1: if RAID1 or RAID10 is
3410 * configured and more than
3411 * 2 mirrors are used.
3412 */
3413 if (num_tolerated_disk_barrier_failures > 0 &&
3414 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3415 BTRFS_BLOCK_GROUP_RAID0)) ||
3416 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3417 == 0)))
3418 num_tolerated_disk_barrier_failures = 0;
53b381b3
DW		 3419 else if (num_tolerated_disk_barrier_failures > 1) {
3420 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3421 BTRFS_BLOCK_GROUP_RAID5 |
3422 BTRFS_BLOCK_GROUP_RAID10)) {
3423 num_tolerated_disk_barrier_failures = 1;
3424 } else if (flags &
15b0a89d 3425 BTRFS_BLOCK_GROUP_RAID6) {
53b381b3
DW		 3426 num_tolerated_disk_barrier_failures = 2;
3427 }
3428 }
5af3e8cc
SB		 3429 }
3430 }
3431 up_read(&sinfo->groups_sem);
3432 }
3433
3434 return num_tolerated_disk_barrier_failures;
3435}
3436
48a3b636 3437static int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 3438{
e5e9a520 3439 struct list_head *head;
f2984462 3440 struct btrfs_device *dev;
a061fc8d 3441 struct btrfs_super_block *sb;
f2984462 3442 struct btrfs_dev_item *dev_item;
f2984462
CM		 3443 int ret;
3444 int do_barriers;
a236aed1
CM		 3445 int max_errors;
3446 int total_errors = 0;
a061fc8d 3447 u64 flags;
f2984462
CM		 3448
3449 do_barriers = !btrfs_test_opt(root, NOBARRIER);
af31f5e5 3450 backup_super_roots(root->fs_info);
f2984462 3451
6c41761f 3452 sb = root->fs_info->super_for_commit;
a061fc8d 3453 dev_item = &sb->dev_item;
e5e9a520 3454
174ba509 3455 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 3456 head = &root->fs_info->fs_devices->devices;
d7306801 3457 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
387125fc 3458
5af3e8cc
SB		 3459 if (do_barriers) {
3460 ret = barrier_all_devices(root->fs_info);
3461 if (ret) {
3462 mutex_unlock(
3463 &root->fs_info->fs_devices->device_list_mutex);
3464 btrfs_error(root->fs_info, ret,
3465 "errors while submitting device barriers.");
3466 return ret;
3467 }
3468 }
387125fc 3469
1f78160c 3470 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3471 if (!dev->bdev) {
3472 total_errors++;
3473 continue;
3474 }
2b82032c 3475 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3476 continue;
3477
2b82032c 3478 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 3479 btrfs_set_stack_device_type(dev_item, dev->type);
3480 btrfs_set_stack_device_id(dev_item, dev->devid);
7df69d3e 3481 btrfs_set_stack_device_total_bytes(dev_item,
935e5cc9 3482 dev->commit_total_bytes);
ce7213c7
MX		 3483 btrfs_set_stack_device_bytes_used(dev_item,
3484 dev->commit_bytes_used);
a061fc8d
CM		 3485 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3486 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3487 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3488 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 3489 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 3490
a061fc8d
CM		 3491 flags = btrfs_super_flags(sb);
3492 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3493
a512bbf8 3494 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 3495 if (ret)
3496 total_errors++;
f2984462 3497 }
a236aed1 3498 if (total_errors > max_errors) {
efe120a0 3499 btrfs_err(root->fs_info, "%d errors while writing supers",
d397712b 3500 total_errors);
a724b436 3501 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
79787eaa 3502
9d565ba4
SB		 3503 /* FUA is masked off if unsupported and can't be the reason */
3504 btrfs_error(root->fs_info, -EIO,
3505 "%d errors while writing supers", total_errors);
3506 return -EIO;
a236aed1 3507 }
f2984462 3508
a512bbf8 3509 total_errors = 0;
1f78160c 3510 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3511 if (!dev->bdev)
3512 continue;
2b82032c 3513 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3514 continue;
3515
a512bbf8
YZ		 3516 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3517 if (ret)
3518 total_errors++;
f2984462 3519 }
174ba509 3520 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 3521 if (total_errors > max_errors) {
79787eaa
JM		 3522 btrfs_error(root->fs_info, -EIO,
3523 "%d errors while writing supers", total_errors);
3524 return -EIO;
a236aed1 3525 }
f2984462
CM		 3526 return 0;
3527}
3528
a512bbf8
YZ		 3529int write_ctree_super(struct btrfs_trans_handle *trans,
3530 struct btrfs_root *root, int max_mirrors)
eb60ceac 3531{
f570e757 3532 return write_all_supers(root, max_mirrors);
cfaa7295
CM		 3533}
3534
cb517eab
MX		 3535/* Drop a fs root from the radix tree and free it. */
3536void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3537 struct btrfs_root *root)
2619ba1f 3538{
4df27c4d 3539 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 3540 radix_tree_delete(&fs_info->fs_roots_radix,
3541 (unsigned long)root->root_key.objectid);
4df27c4d 3542 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 3543
3544 if (btrfs_root_refs(&root->root_item) == 0)
3545 synchronize_srcu(&fs_info->subvol_srcu);
3546
1a4319cc 3547 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
3321719e 3548 btrfs_free_log(NULL, root);
3321719e 3549
faa2dbf0
JB		 3550 if (root->free_ino_pinned)
3551 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3552 if (root->free_ino_ctl)
3553 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3554 free_fs_root(root);
4df27c4d
YZ		 3555}
3556
3557static void free_fs_root(struct btrfs_root *root)
3558{
57cdc8db 3559 iput(root->ino_cache_inode);
4df27c4d 3560 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
1cb048f5
FDBM		 3561 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3562 root->orphan_block_rsv = NULL;
0ee5dc67
AV		 3563 if (root->anon_dev)
3564 free_anon_bdev(root->anon_dev);
8257b2dc
MX		 3565 if (root->subv_writers)
3566 btrfs_free_subvolume_writers(root->subv_writers);
4df27c4d
YZ		 3567 free_extent_buffer(root->node);
3568 free_extent_buffer(root->commit_root);
581bb050
LZ		 3569 kfree(root->free_ino_ctl);
3570 kfree(root->free_ino_pinned);
d397712b 3571 kfree(root->name);
b0feb9d9 3572 btrfs_put_fs_root(root);
2619ba1f
CM		 3573}
3574
cb517eab
MX		 3575void btrfs_free_fs_root(struct btrfs_root *root)
3576{
3577 free_fs_root(root);
2619ba1f
CM		 3578}
3579
c146afad 3580int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3581{
c146afad
YZ		 3582 u64 root_objectid = 0;
3583 struct btrfs_root *gang[8];
65d33fd7
QW		 3584 int i = 0;
3585 int err = 0;
3586 unsigned int ret = 0;
3587 int index;
e089f05c 3588
c146afad 3589 while (1) {
65d33fd7 3590 index = srcu_read_lock(&fs_info->subvol_srcu);
c146afad
YZ		 3591 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3592 (void **)gang, root_objectid,
3593 ARRAY_SIZE(gang));
65d33fd7
QW		 3594 if (!ret) {
3595 srcu_read_unlock(&fs_info->subvol_srcu, index);
c146afad 3596 break;
65d33fd7 3597 }
5d4f98a2 3598 root_objectid = gang[ret - 1]->root_key.objectid + 1;
65d33fd7 3599
c146afad 3600 for (i = 0; i < ret; i++) {
65d33fd7
QW		 3601 /* Avoid to grab roots in dead_roots */
3602 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3603 gang[i] = NULL;
3604 continue;
3605 }
3606 /* grab all the search result for later use */
3607 gang[i] = btrfs_grab_fs_root(gang[i]);
3608 }
3609 srcu_read_unlock(&fs_info->subvol_srcu, index);
66b4ffd1 3610
65d33fd7
QW		 3611 for (i = 0; i < ret; i++) {
3612 if (!gang[i])
3613 continue;
c146afad 3614 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 3615 err = btrfs_orphan_cleanup(gang[i]);
3616 if (err)
65d33fd7
QW		 3617 break;
3618 btrfs_put_fs_root(gang[i]);
c146afad
YZ		 3619 }
3620 root_objectid++;
3621 }
65d33fd7
QW		 3622
3623 /* release the uncleaned roots due to error */
3624 for (; i < ret; i++) {
3625 if (gang[i])
3626 btrfs_put_fs_root(gang[i]);
3627 }
3628 return err;
c146afad 3629}
a2135011 3630
c146afad
YZ		 3631int btrfs_commit_super(struct btrfs_root *root)
3632{
3633 struct btrfs_trans_handle *trans;
a74a4b97 3634
c146afad 3635 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 3636 btrfs_run_delayed_iputs(root);
c146afad 3637 mutex_unlock(&root->fs_info->cleaner_mutex);
9d1a2a3a 3638 wake_up_process(root->fs_info->cleaner_kthread);
c71bf099
YZ		 3639
3640 /* wait until ongoing cleanup work done */
3641 down_write(&root->fs_info->cleanup_work_sem);
3642 up_write(&root->fs_info->cleanup_work_sem);
3643
7a7eaa40 3644 trans = btrfs_join_transaction(root);
3612b495
TI		 3645 if (IS_ERR(trans))
3646 return PTR_ERR(trans);
d52c1bcc 3647 return btrfs_commit_transaction(trans, root);
c146afad
YZ		 3648}
3649
3abdbd78 3650void close_ctree(struct btrfs_root *root)
c146afad
YZ		 3651{
3652 struct btrfs_fs_info *fs_info = root->fs_info;
3653 int ret;
3654
3655 fs_info->closing = 1;
3656 smp_mb();
3657
803b2f54
SB		 3658 /* wait for the uuid_scan task to finish */
3659 down(&fs_info->uuid_tree_rescan_sem);
3660 /* avoid complains from lockdep et al., set sem back to initial state */
3661 up(&fs_info->uuid_tree_rescan_sem);
3662
837d5b6e 3663 /* pause restriper - we want to resume on mount */
aa1b8cd4 3664 btrfs_pause_balance(fs_info);
837d5b6e 3665
8dabb742
SB		 3666 btrfs_dev_replace_suspend_for_unmount(fs_info);
3667
aa1b8cd4 3668 btrfs_scrub_cancel(fs_info);
4cb5300b
CM		 3669
3670 /* wait for any defraggers to finish */
3671 wait_event(fs_info->transaction_wait,
3672 (atomic_read(&fs_info->defrag_running) == 0));
3673
3674 /* clear out the rbtree of defraggable inodes */
26176e7c 3675 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3676
21c7e756
MX		 3677 cancel_work_sync(&fs_info->async_reclaim_work);
3678
c146afad 3679 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 3680 ret = btrfs_commit_super(root);
3681 if (ret)
04892340 3682 btrfs_err(fs_info, "commit super ret %d", ret);
acce952b 3683 }
3684
87533c47 3685 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
68ce9682 3686 btrfs_error_commit_super(root);
0f7d52f4 3687
e3029d9f
AV		 3688 kthread_stop(fs_info->transaction_kthread);
3689 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3690
f25784b3
YZ		 3691 fs_info->closing = 2;
3692 smp_mb();
3693
04892340 3694 btrfs_free_qgroup_config(fs_info);
bcef60f2 3695
963d678b 3696 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
04892340 3697 btrfs_info(fs_info, "at unmount delalloc count %lld",
963d678b 3698 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3699 }
bcc63abb 3700
5ac1d209
JM		 3701 btrfs_sysfs_remove_one(fs_info);
3702
faa2dbf0 3703 btrfs_free_fs_roots(fs_info);
d10c5f31 3704
1a4319cc
LB		 3705 btrfs_put_block_group_cache(fs_info);
3706
2b1360da
JB		 3707 btrfs_free_block_groups(fs_info);
3708
de348ee0
WS		 3709 /*
3710 * we must make sure there is not any read request to
3711 * submit after we stopping all workers.
3712 */
3713 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
96192499
JB		 3714 btrfs_stop_all_workers(fs_info);
3715
47ab2a6c 3716 fs_info->open = 0;
13e6c37b 3717 free_root_pointers(fs_info, 1);
9ad6b7bc 3718
13e6c37b 3719 iput(fs_info->btree_inode);
d6bfde87 3720
21adbd5c
SB		 3721#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3722 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3723 btrfsic_unmount(root, fs_info->fs_devices);
3724#endif
3725
dfe25020 3726 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3727 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3728
e2d84521 3729 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 3730 percpu_counter_destroy(&fs_info->delalloc_bytes);
c404e0dc 3731 percpu_counter_destroy(&fs_info->bio_counter);
04160088 3732 bdi_destroy(&fs_info->bdi);
76dda93c 3733 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 3734
53b381b3
DW		 3735 btrfs_free_stripe_hash_table(fs_info);
3736
1cb048f5
FDBM		 3737 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3738 root->orphan_block_rsv = NULL;
04216820
FM		 3739
3740 lock_chunks(root);
3741 while (!list_empty(&fs_info->pinned_chunks)) {
3742 struct extent_map *em;
3743
3744 em = list_first_entry(&fs_info->pinned_chunks,
3745 struct extent_map, list);
3746 list_del_init(&em->list);
3747 free_extent_map(em);
3748 }
3749 unlock_chunks(root);
eb60ceac
CM		 3750}
3751
b9fab919
CM		 3752int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3753 int atomic)
5f39d397 3754{
1259ab75 3755 int ret;
727011e0 3756 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 3757
0b32f4bb 3758 ret = extent_buffer_uptodate(buf);
1259ab75
CM		 3759 if (!ret)
3760 return ret;
3761
3762 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM		 3763 parent_transid, atomic);
3764 if (ret == -EAGAIN)
3765 return ret;
1259ab75 3766 return !ret;
5f39d397
CM		 3767}
3768
3769int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 3770{
0b32f4bb 3771 return set_extent_buffer_uptodate(buf);
5f39d397 3772}
6702ed49 3773
5f39d397
CM		 3774void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3775{
06ea65a3 3776 struct btrfs_root *root;
5f39d397 3777 u64 transid = btrfs_header_generation(buf);
b9473439 3778 int was_dirty;
b4ce94de 3779
06ea65a3
JB		 3780#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3781 /*
3782 * This is a fast path so only do this check if we have sanity tests
3783 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3784 * outside of the sanity tests.
3785 */
3786 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3787 return;
3788#endif
3789 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
b9447ef8 3790 btrfs_assert_tree_locked(buf);
31b1a2bd
JL		 3791 if (transid != root->fs_info->generation)
3792 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
d397712b 3793 "found %llu running %llu\n",
c1c9ff7c 3794 buf->start, transid, root->fs_info->generation);
0b32f4bb 3795 was_dirty = set_extent_buffer_dirty(buf);
e2d84521
MX		 3796 if (!was_dirty)
3797 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3798 buf->len,
3799 root->fs_info->dirty_metadata_batch);
1f21ef0a
FM		 3800#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3801 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
3802 btrfs_print_leaf(root, buf);
3803 ASSERT(0);
3804 }
3805#endif
eb60ceac
CM		 3806}
3807
b53d3f5d
LB		 3808static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3809 int flush_delayed)
16cdcec7
MX		 3810{
3811 /*
3812 * looks as though older kernels can get into trouble with
3813 * this code, they end up stuck in balance_dirty_pages forever
3814 */
e2d84521 3815 int ret;
16cdcec7
MX		 3816
3817 if (current->flags & PF_MEMALLOC)
3818 return;
3819
b53d3f5d
LB		 3820 if (flush_delayed)
3821 btrfs_balance_delayed_items(root);
16cdcec7 3822
e2d84521
MX		 3823 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3824 BTRFS_DIRTY_METADATA_THRESH);
3825 if (ret > 0) {
d0e1d66b
NJ		 3826 balance_dirty_pages_ratelimited(
3827 root->fs_info->btree_inode->i_mapping);
16cdcec7
MX		 3828 }
3829 return;
3830}
3831
b53d3f5d 3832void btrfs_btree_balance_dirty(struct btrfs_root *root)
35b7e476 3833{
b53d3f5d
LB		 3834 __btrfs_btree_balance_dirty(root, 1);
3835}
585ad2c3 3836
b53d3f5d
LB		 3837void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3838{
3839 __btrfs_btree_balance_dirty(root, 0);
35b7e476 3840}
6b80053d 3841
ca7a79ad 3842int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 3843{
727011e0 3844 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
0b32f4bb 3845 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
6b80053d 3846}
0da5468f 3847
fcd1f065 3848static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 3849 int read_only)
3850{
c926093e
DS		 3851 struct btrfs_super_block *sb = fs_info->super_copy;
3852 int ret = 0;
3853
21e7626b
DS		 3854 if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) {
3855 printk(KERN_ERR "BTRFS: tree_root level too big: %d >= %d\n",
3856 btrfs_super_root_level(sb), BTRFS_MAX_LEVEL);
c926093e
DS		 3857 ret = -EINVAL;
3858 }
21e7626b
DS		 3859 if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) {
3860 printk(KERN_ERR "BTRFS: chunk_root level too big: %d >= %d\n",
3861 btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL);
c926093e
DS		 3862 ret = -EINVAL;
3863 }
21e7626b
DS		 3864 if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) {
3865 printk(KERN_ERR "BTRFS: log_root level too big: %d >= %d\n",
3866 btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL);
c926093e
DS		 3867 ret = -EINVAL;
3868 }
3869
1104a885 3870 /*
c926093e
DS		 3871 * The common minimum, we don't know if we can trust the nodesize/sectorsize
3872 * items yet, they'll be verified later. Issue just a warning.
1104a885 3873 */
21e7626b 3874 if (!IS_ALIGNED(btrfs_super_root(sb), 4096))
c926093e 3875 printk(KERN_WARNING "BTRFS: tree_root block unaligned: %llu\n",
cd743fac 3876 btrfs_super_root(sb));
21e7626b 3877 if (!IS_ALIGNED(btrfs_super_chunk_root(sb), 4096))
cd743fac
DS		 3878 printk(KERN_WARNING "BTRFS: chunk_root block unaligned: %llu\n",
3879 btrfs_super_chunk_root(sb));
21e7626b 3880 if (!IS_ALIGNED(btrfs_super_log_root(sb), 4096))
cd743fac 3881 printk(KERN_WARNING "BTRFS: log_root block unaligned: %llu\n",
21e7626b 3882 btrfs_super_log_root(sb));
c926093e 3883
75d6ad38
DS		 3884 /*
3885 * Check the lower bound, the alignment and other constraints are
3886 * checked later.
3887 */
3888 if (btrfs_super_nodesize(sb) < 4096) {
3889 printk(KERN_ERR "BTRFS: nodesize too small: %u < 4096\n",
3890 btrfs_super_nodesize(sb));
3891 ret = -EINVAL;
3892 }
3893 if (btrfs_super_sectorsize(sb) < 4096) {
3894 printk(KERN_ERR "BTRFS: sectorsize too small: %u < 4096\n",
3895 btrfs_super_sectorsize(sb));
3896 ret = -EINVAL;
3897 }
3898
c926093e
DS		 3899 if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) {
3900 printk(KERN_ERR "BTRFS: dev_item UUID does not match fsid: %pU != %pU\n",
3901 fs_info->fsid, sb->dev_item.fsid);
3902 ret = -EINVAL;
3903 }
3904
3905 /*
3906 * Hint to catch really bogus numbers, bitflips or so, more exact checks are
3907 * done later
3908 */
21e7626b 3909 if (btrfs_super_num_devices(sb) > (1UL << 31))
c926093e 3910 printk(KERN_WARNING "BTRFS: suspicious number of devices: %llu\n",
21e7626b 3911 btrfs_super_num_devices(sb));
75d6ad38
DS		 3912 if (btrfs_super_num_devices(sb) == 0) {
3913 printk(KERN_ERR "BTRFS: number of devices is 0\n");
3914 ret = -EINVAL;
3915 }
c926093e 3916
21e7626b 3917 if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) {
c926093e 3918 printk(KERN_ERR "BTRFS: super offset mismatch %llu != %u\n",
21e7626b 3919 btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET);
c926093e
DS		 3920 ret = -EINVAL;
3921 }
3922
ce7fca5f
DS		 3923 /*
3924 * Obvious sys_chunk_array corruptions, it must hold at least one key
3925 * and one chunk
3926 */
3927 if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) {
3928 printk(KERN_ERR "BTRFS: system chunk array too big %u > %u\n",
3929 btrfs_super_sys_array_size(sb),
3930 BTRFS_SYSTEM_CHUNK_ARRAY_SIZE);
3931 ret = -EINVAL;
3932 }
3933 if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key)
3934 + sizeof(struct btrfs_chunk)) {
41d6b13e 3935 printk(KERN_ERR "BTRFS: system chunk array too small %u < %zu\n",
ce7fca5f
DS		 3936 btrfs_super_sys_array_size(sb),
3937 sizeof(struct btrfs_disk_key)
3938 + sizeof(struct btrfs_chunk));
3939 ret = -EINVAL;
3940 }
3941
c926093e
DS		 3942 /*
3943 * The generation is a global counter, we'll trust it more than the others
3944 * but it's still possible that it's the one that's wrong.
3945 */
21e7626b 3946 if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb))
c926093e
DS		 3947 printk(KERN_WARNING
3948 "BTRFS: suspicious: generation < chunk_root_generation: %llu < %llu\n",
21e7626b
DS		 3949 btrfs_super_generation(sb), btrfs_super_chunk_root_generation(sb));
3950 if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb)
3951 && btrfs_super_cache_generation(sb) != (u64)-1)
c926093e
DS		 3952 printk(KERN_WARNING
3953 "BTRFS: suspicious: generation < cache_generation: %llu < %llu\n",
21e7626b 3954 btrfs_super_generation(sb), btrfs_super_cache_generation(sb));
c926093e
DS		 3955
3956 return ret;
acce952b 3957}
3958
48a3b636 3959static void btrfs_error_commit_super(struct btrfs_root *root)
acce952b 3960{
acce952b 3961 mutex_lock(&root->fs_info->cleaner_mutex);
3962 btrfs_run_delayed_iputs(root);
3963 mutex_unlock(&root->fs_info->cleaner_mutex);
3964
3965 down_write(&root->fs_info->cleanup_work_sem);
3966 up_write(&root->fs_info->cleanup_work_sem);
3967
3968 /* cleanup FS via transaction */
3969 btrfs_cleanup_transaction(root);
acce952b 3970}
3971
143bede5 3972static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 3973{
acce952b 3974 struct btrfs_ordered_extent *ordered;
acce952b 3975
199c2a9c 3976 spin_lock(&root->ordered_extent_lock);
779880ef
JB		 3977 /*
3978 * This will just short circuit the ordered completion stuff which will
3979 * make sure the ordered extent gets properly cleaned up.
3980 */
199c2a9c 3981 list_for_each_entry(ordered, &root->ordered_extents,
779880ef
JB		 3982 root_extent_list)
3983 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
199c2a9c
MX		 3984 spin_unlock(&root->ordered_extent_lock);
3985}
3986
3987static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3988{
3989 struct btrfs_root *root;
3990 struct list_head splice;
3991
3992 INIT_LIST_HEAD(&splice);
3993
3994 spin_lock(&fs_info->ordered_root_lock);
3995 list_splice_init(&fs_info->ordered_roots, &splice);
3996 while (!list_empty(&splice)) {
3997 root = list_first_entry(&splice, struct btrfs_root,
3998 ordered_root);
1de2cfde
JB		 3999 list_move_tail(&root->ordered_root,
4000 &fs_info->ordered_roots);
199c2a9c 4001
2a85d9ca 4002 spin_unlock(&fs_info->ordered_root_lock);
199c2a9c
MX		 4003 btrfs_destroy_ordered_extents(root);
4004
2a85d9ca
LB		 4005 cond_resched();
4006 spin_lock(&fs_info->ordered_root_lock);
199c2a9c
MX		 4007 }
4008 spin_unlock(&fs_info->ordered_root_lock);
acce952b 4009}
4010
35a3621b
SB		 4011static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
4012 struct btrfs_root *root)
acce952b 4013{
4014 struct rb_node *node;
4015 struct btrfs_delayed_ref_root *delayed_refs;
4016 struct btrfs_delayed_ref_node *ref;
4017 int ret = 0;
4018
4019 delayed_refs = &trans->delayed_refs;
4020
4021 spin_lock(&delayed_refs->lock);
d7df2c79 4022 if (atomic_read(&delayed_refs->num_entries) == 0) {
cfece4db 4023 spin_unlock(&delayed_refs->lock);
efe120a0 4024 btrfs_info(root->fs_info, "delayed_refs has NO entry");
acce952b 4025 return ret;
4026 }
4027
d7df2c79
JB		 4028 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
4029 struct btrfs_delayed_ref_head *head;
e78417d1 4030 bool pin_bytes = false;
acce952b 4031
d7df2c79
JB		 4032 head = rb_entry(node, struct btrfs_delayed_ref_head,
4033 href_node);
4034 if (!mutex_trylock(&head->mutex)) {
4035 atomic_inc(&head->node.refs);
4036 spin_unlock(&delayed_refs->lock);
eb12db69 4037
d7df2c79 4038 mutex_lock(&head->mutex);
e78417d1 4039 mutex_unlock(&head->mutex);
d7df2c79
JB		 4040 btrfs_put_delayed_ref(&head->node);
4041 spin_lock(&delayed_refs->lock);
4042 continue;
4043 }
4044 spin_lock(&head->lock);
4045 while ((node = rb_first(&head->ref_root)) != NULL) {
4046 ref = rb_entry(node, struct btrfs_delayed_ref_node,
4047 rb_node);
4048 ref->in_tree = 0;
4049 rb_erase(&ref->rb_node, &head->ref_root);
4050 atomic_dec(&delayed_refs->num_entries);
4051 btrfs_put_delayed_ref(ref);
e78417d1 4052 }
d7df2c79
JB		 4053 if (head->must_insert_reserved)
4054 pin_bytes = true;
4055 btrfs_free_delayed_extent_op(head->extent_op);
4056 delayed_refs->num_heads--;
4057 if (head->processing == 0)
4058 delayed_refs->num_heads_ready--;
4059 atomic_dec(&delayed_refs->num_entries);
4060 head->node.in_tree = 0;
4061 rb_erase(&head->href_node, &delayed_refs->href_root);
4062 spin_unlock(&head->lock);
4063 spin_unlock(&delayed_refs->lock);
4064 mutex_unlock(&head->mutex);
acce952b 4065
d7df2c79
JB		 4066 if (pin_bytes)
4067 btrfs_pin_extent(root, head->node.bytenr,
4068 head->node.num_bytes, 1);
4069 btrfs_put_delayed_ref(&head->node);
acce952b 4070 cond_resched();
4071 spin_lock(&delayed_refs->lock);
4072 }
4073
4074 spin_unlock(&delayed_refs->lock);
4075
4076 return ret;
4077}
4078
143bede5 4079static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 4080{
4081 struct btrfs_inode *btrfs_inode;
4082 struct list_head splice;
4083
4084 INIT_LIST_HEAD(&splice);
4085
eb73c1b7
MX		 4086 spin_lock(&root->delalloc_lock);
4087 list_splice_init(&root->delalloc_inodes, &splice);
acce952b 4088
4089 while (!list_empty(&splice)) {
eb73c1b7
MX		 4090 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
4091 delalloc_inodes);
acce952b 4092
4093 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX		 4094 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
4095 &btrfs_inode->runtime_flags);
eb73c1b7 4096 spin_unlock(&root->delalloc_lock);
acce952b 4097
4098 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 4099
eb73c1b7 4100 spin_lock(&root->delalloc_lock);
acce952b 4101 }
4102
eb73c1b7
MX		 4103 spin_unlock(&root->delalloc_lock);
4104}
4105
4106static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
4107{
4108 struct btrfs_root *root;
4109 struct list_head splice;
4110
4111 INIT_LIST_HEAD(&splice);
4112
4113 spin_lock(&fs_info->delalloc_root_lock);
4114 list_splice_init(&fs_info->delalloc_roots, &splice);
4115 while (!list_empty(&splice)) {
4116 root = list_first_entry(&splice, struct btrfs_root,
4117 delalloc_root);
4118 list_del_init(&root->delalloc_root);
4119 root = btrfs_grab_fs_root(root);
4120 BUG_ON(!root);
4121 spin_unlock(&fs_info->delalloc_root_lock);
4122
4123 btrfs_destroy_delalloc_inodes(root);
4124 btrfs_put_fs_root(root);
4125
4126 spin_lock(&fs_info->delalloc_root_lock);
4127 }
4128 spin_unlock(&fs_info->delalloc_root_lock);
acce952b 4129}
4130
4131static int btrfs_destroy_marked_extents(struct btrfs_root *root,
4132 struct extent_io_tree *dirty_pages,
4133 int mark)
4134{
4135 int ret;
acce952b 4136 struct extent_buffer *eb;
4137 u64 start = 0;
4138 u64 end;
acce952b 4139
4140 while (1) {
4141 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 4142 mark, NULL);
acce952b 4143 if (ret)
4144 break;
4145
4146 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
4147 while (start <= end) {
01d58472 4148 eb = btrfs_find_tree_block(root->fs_info, start);
707e8a07 4149 start += root->nodesize;
fd8b2b61 4150 if (!eb)
acce952b 4151 continue;
fd8b2b61 4152 wait_on_extent_buffer_writeback(eb);
acce952b 4153
fd8b2b61
JB		 4154 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4155 &eb->bflags))
4156 clear_extent_buffer_dirty(eb);
4157 free_extent_buffer_stale(eb);
acce952b 4158 }
4159 }
4160
4161 return ret;
4162}
4163
4164static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4165 struct extent_io_tree *pinned_extents)
4166{
4167 struct extent_io_tree *unpin;
4168 u64 start;
4169 u64 end;
4170 int ret;
ed0eaa14 4171 bool loop = true;
acce952b 4172
4173 unpin = pinned_extents;
ed0eaa14 4174again:
acce952b 4175 while (1) {
4176 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 4177 EXTENT_DIRTY, NULL);
acce952b 4178 if (ret)
4179 break;
4180
acce952b 4181 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4182 btrfs_error_unpin_extent_range(root, start, end);
4183 cond_resched();
4184 }
4185
ed0eaa14
LB		 4186 if (loop) {
4187 if (unpin == &root->fs_info->freed_extents[0])
4188 unpin = &root->fs_info->freed_extents[1];
4189 else
4190 unpin = &root->fs_info->freed_extents[0];
4191 loop = false;
4192 goto again;
4193 }
4194
acce952b 4195 return 0;
4196}
4197
50d9aa99
JB		 4198static void btrfs_free_pending_ordered(struct btrfs_transaction *cur_trans,
4199 struct btrfs_fs_info *fs_info)
4200{
4201 struct btrfs_ordered_extent *ordered;
4202
4203 spin_lock(&fs_info->trans_lock);
4204 while (!list_empty(&cur_trans->pending_ordered)) {
4205 ordered = list_first_entry(&cur_trans->pending_ordered,
4206 struct btrfs_ordered_extent,
4207 trans_list);
4208 list_del_init(&ordered->trans_list);
4209 spin_unlock(&fs_info->trans_lock);
4210
4211 btrfs_put_ordered_extent(ordered);
4212 spin_lock(&fs_info->trans_lock);
4213 }
4214 spin_unlock(&fs_info->trans_lock);
4215}
4216
49b25e05
JM		 4217void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4218 struct btrfs_root *root)
4219{
4220 btrfs_destroy_delayed_refs(cur_trans, root);
49b25e05 4221
4a9d8bde 4222 cur_trans->state = TRANS_STATE_COMMIT_START;
d7096fc3 4223 wake_up(&root->fs_info->transaction_blocked_wait);
49b25e05 4224
4a9d8bde 4225 cur_trans->state = TRANS_STATE_UNBLOCKED;
d7096fc3 4226 wake_up(&root->fs_info->transaction_wait);
49b25e05 4227
50d9aa99 4228 btrfs_free_pending_ordered(cur_trans, root->fs_info);
67cde344
MX		 4229 btrfs_destroy_delayed_inodes(root);
4230 btrfs_assert_delayed_root_empty(root);
49b25e05 4231
49b25e05
JM		 4232 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4233 EXTENT_DIRTY);
6e841e32
LB		 4234 btrfs_destroy_pinned_extent(root,
4235 root->fs_info->pinned_extents);
49b25e05 4236
4a9d8bde
MX		 4237 cur_trans->state =TRANS_STATE_COMPLETED;
4238 wake_up(&cur_trans->commit_wait);
4239
49b25e05
JM		 4240 /*
4241 memset(cur_trans, 0, sizeof(*cur_trans));
4242 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4243 */
4244}
4245
48a3b636 4246static int btrfs_cleanup_transaction(struct btrfs_root *root)
acce952b 4247{
4248 struct btrfs_transaction *t;
acce952b 4249
acce952b 4250 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4251
a4abeea4 4252 spin_lock(&root->fs_info->trans_lock);
724e2315
JB		 4253 while (!list_empty(&root->fs_info->trans_list)) {
4254 t = list_first_entry(&root->fs_info->trans_list,
4255 struct btrfs_transaction, list);
4256 if (t->state >= TRANS_STATE_COMMIT_START) {
4257 atomic_inc(&t->use_count);
4258 spin_unlock(&root->fs_info->trans_lock);
4259 btrfs_wait_for_commit(root, t->transid);
4260 btrfs_put_transaction(t);
4261 spin_lock(&root->fs_info->trans_lock);
4262 continue;
4263 }
4264 if (t == root->fs_info->running_transaction) {
4265 t->state = TRANS_STATE_COMMIT_DOING;
4266 spin_unlock(&root->fs_info->trans_lock);
4267 /*
4268 * We wait for 0 num_writers since we don't hold a trans
4269 * handle open currently for this transaction.
4270 */
4271 wait_event(t->writer_wait,
4272 atomic_read(&t->num_writers) == 0);
4273 } else {
4274 spin_unlock(&root->fs_info->trans_lock);
4275 }
4276 btrfs_cleanup_one_transaction(t, root);
4a9d8bde 4277
724e2315
JB		 4278 spin_lock(&root->fs_info->trans_lock);
4279 if (t == root->fs_info->running_transaction)
4280 root->fs_info->running_transaction = NULL;
acce952b 4281 list_del_init(&t->list);
724e2315 4282 spin_unlock(&root->fs_info->trans_lock);
acce952b 4283
724e2315
JB		 4284 btrfs_put_transaction(t);
4285 trace_btrfs_transaction_commit(root);
4286 spin_lock(&root->fs_info->trans_lock);
4287 }
4288 spin_unlock(&root->fs_info->trans_lock);
4289 btrfs_destroy_all_ordered_extents(root->fs_info);
4290 btrfs_destroy_delayed_inodes(root);
4291 btrfs_assert_delayed_root_empty(root);
4292 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4293 btrfs_destroy_all_delalloc_inodes(root->fs_info);
acce952b 4294 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 4295
4296 return 0;
4297}
4298
e8c9f186 4299static const struct extent_io_ops btree_extent_io_ops = {
ce9adaa5 4300 .readpage_end_io_hook = btree_readpage_end_io_hook,
4bb31e92 4301 .readpage_io_failed_hook = btree_io_failed_hook,
0b86a832 4302 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 4303 /* note we're sharing with inode.c for the merge bio hook */
4304 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 4305};
Linux 6.x block layer and io_uring tree(s)