projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: remove parameter blocksize from read_tree_block
[linux-2.6-block.git] / fs / btrfs / disk-io.c
CommitLineData
6cbd5570
CM		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
e20d96d6 19#include <linux/fs.h>
d98237b3 20#include <linux/blkdev.h>
87cbda5c 21#include <linux/scatterlist.h>
22b0ebda 22#include <linux/swap.h>
0f7d52f4 23#include <linux/radix-tree.h>
35b7e476 24#include <linux/writeback.h>
d397712b 25#include <linux/buffer_head.h>
ce9adaa5 26#include <linux/workqueue.h>
a74a4b97 27#include <linux/kthread.h>
4b4e25f2 28#include <linux/freezer.h>
5a0e3ad6 29#include <linux/slab.h>
784b4e29 30#include <linux/migrate.h>
7a36ddec 31#include <linux/ratelimit.h>
6463fe58 32#include <linux/uuid.h>
803b2f54 33#include <linux/semaphore.h>
7e75bf3f 34#include <asm/unaligned.h>
eb60ceac
CM		 35#include "ctree.h"
36#include "disk-io.h"
0b947aff 37#include "hash.h"
e089f05c 38#include "transaction.h"
0f7d52f4 39#include "btrfs_inode.h"
0b86a832 40#include "volumes.h"
db94535d 41#include "print-tree.h"
925baedd 42#include "locking.h"
e02119d5 43#include "tree-log.h"
fa9c0d79 44#include "free-space-cache.h"
581bb050 45#include "inode-map.h"
21adbd5c 46#include "check-integrity.h"
606686ee 47#include "rcu-string.h"
8dabb742 48#include "dev-replace.h"
53b381b3 49#include "raid56.h"
5ac1d209 50#include "sysfs.h"
fcebe456 51#include "qgroup.h"
eb60ceac 52
de0022b9
JB		 53#ifdef CONFIG_X86
54#include <asm/cpufeature.h>
55#endif
56
d1310b2e 57static struct extent_io_ops btree_extent_io_ops;
8b712842 58static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 59static void free_fs_root(struct btrfs_root *root);
fcd1f065 60static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 61 int read_only);
143bede5 62static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
acce952b 63static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
64 struct btrfs_root *root);
143bede5 65static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
acce952b 66static int btrfs_destroy_marked_extents(struct btrfs_root *root,
67 struct extent_io_tree *dirty_pages,
68 int mark);
69static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
70 struct extent_io_tree *pinned_extents);
48a3b636
ES		 71static int btrfs_cleanup_transaction(struct btrfs_root *root);
72static void btrfs_error_commit_super(struct btrfs_root *root);
ce9adaa5 73
d352ac68
CM		 74/*
75 * end_io_wq structs are used to do processing in task context when an IO is
76 * complete. This is used during reads to verify checksums, and it is used
77 * by writes to insert metadata for new file extents after IO is complete.
78 */
ce9adaa5
CM		 79struct end_io_wq {
80 struct bio *bio;
81 bio_end_io_t *end_io;
82 void *private;
83 struct btrfs_fs_info *info;
84 int error;
22c59948 85 int metadata;
ce9adaa5 86 struct list_head list;
8b712842 87 struct btrfs_work work;
ce9adaa5 88};
0da5468f 89
d352ac68
CM		 90/*
91 * async submit bios are used to offload expensive checksumming
92 * onto the worker threads. They checksum file and metadata bios
93 * just before they are sent down the IO stack.
94 */
44b8bd7e
CM		 95struct async_submit_bio {
96 struct inode *inode;
97 struct bio *bio;
98 struct list_head list;
4a69a410
CM		 99 extent_submit_bio_hook_t *submit_bio_start;
100 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM		 101 int rw;
102 int mirror_num;
c8b97818 103 unsigned long bio_flags;
eaf25d93
CM		 104 /*
105 * bio_offset is optional, can be used if the pages in the bio
106 * can't tell us where in the file the bio should go
107 */
108 u64 bio_offset;
8b712842 109 struct btrfs_work work;
79787eaa 110 int error;
44b8bd7e
CM		 111};
112
85d4e461
CM		 113/*
114 * Lockdep class keys for extent_buffer->lock's in this root. For a given
115 * eb, the lockdep key is determined by the btrfs_root it belongs to and
116 * the level the eb occupies in the tree.
117 *
118 * Different roots are used for different purposes and may nest inside each
119 * other and they require separate keysets. As lockdep keys should be
120 * static, assign keysets according to the purpose of the root as indicated
121 * by btrfs_root->objectid. This ensures that all special purpose roots
122 * have separate keysets.
4008c04a 123 *
85d4e461
CM		 124 * Lock-nesting across peer nodes is always done with the immediate parent
125 * node locked thus preventing deadlock. As lockdep doesn't know this, use
126 * subclass to avoid triggering lockdep warning in such cases.
4008c04a 127 *
85d4e461
CM		 128 * The key is set by the readpage_end_io_hook after the buffer has passed
129 * csum validation but before the pages are unlocked. It is also set by
130 * btrfs_init_new_buffer on freshly allocated blocks.
4008c04a 131 *
85d4e461
CM		 132 * We also add a check to make sure the highest level of the tree is the
133 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
134 * needs update as well.
4008c04a
CM		 135 */
136#ifdef CONFIG_DEBUG_LOCK_ALLOC
137# if BTRFS_MAX_LEVEL != 8
138# error
139# endif
85d4e461
CM		 140
141static struct btrfs_lockdep_keyset {
142 u64 id; /* root objectid */
143 const char *name_stem; /* lock name stem */
144 char names[BTRFS_MAX_LEVEL + 1][20];
145 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
146} btrfs_lockdep_keysets[] = {
147 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
148 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
149 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
150 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
151 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
152 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
60b62978 153 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
85d4e461
CM		 154 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
155 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
156 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
13fd8da9 157 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
85d4e461 158 { .id = 0, .name_stem = "tree" },
4008c04a 159};
85d4e461
CM		 160
161void __init btrfs_init_lockdep(void)
162{
163 int i, j;
164
165 /* initialize lockdep class names */
166 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
167 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
168
169 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
170 snprintf(ks->names[j], sizeof(ks->names[j]),
171 "btrfs-%s-%02d", ks->name_stem, j);
172 }
173}
174
175void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
176 int level)
177{
178 struct btrfs_lockdep_keyset *ks;
179
180 BUG_ON(level >= ARRAY_SIZE(ks->keys));
181
182 /* find the matching keyset, id 0 is the default entry */
183 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
184 if (ks->id == objectid)
185 break;
186
187 lockdep_set_class_and_name(&eb->lock,
188 &ks->keys[level], ks->names[level]);
189}
190
4008c04a
CM		 191#endif
192
d352ac68
CM		 193/*
194 * extents on the btree inode are pretty simple, there's one extent
195 * that covers the entire device
196 */
b2950863 197static struct extent_map *btree_get_extent(struct inode *inode,
306e16ce 198 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 199 int create)
7eccb903 200{
5f39d397
CM		 201 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
202 struct extent_map *em;
203 int ret;
204
890871be 205 read_lock(&em_tree->lock);
d1310b2e 206 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 207 if (em) {
208 em->bdev =
209 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 210 read_unlock(&em_tree->lock);
5f39d397 211 goto out;
a061fc8d 212 }
890871be 213 read_unlock(&em_tree->lock);
7b13b7b1 214
172ddd60 215 em = alloc_extent_map();
5f39d397
CM		 216 if (!em) {
217 em = ERR_PTR(-ENOMEM);
218 goto out;
219 }
220 em->start = 0;
0afbaf8c 221 em->len = (u64)-1;
c8b97818 222 em->block_len = (u64)-1;
5f39d397 223 em->block_start = 0;
a061fc8d 224 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 225
890871be 226 write_lock(&em_tree->lock);
09a2a8f9 227 ret = add_extent_mapping(em_tree, em, 0);
5f39d397
CM		 228 if (ret == -EEXIST) {
229 free_extent_map(em);
7b13b7b1 230 em = lookup_extent_mapping(em_tree, start, len);
b4f359ab 231 if (!em)
0433f20d 232 em = ERR_PTR(-EIO);
5f39d397 233 } else if (ret) {
7b13b7b1 234 free_extent_map(em);
0433f20d 235 em = ERR_PTR(ret);
5f39d397 236 }
890871be 237 write_unlock(&em_tree->lock);
7b13b7b1 238
5f39d397
CM		 239out:
240 return em;
7eccb903
CM		 241}
242
b0496686 243u32 btrfs_csum_data(char *data, u32 seed, size_t len)
19c00ddc 244{
0b947aff 245 return btrfs_crc32c(seed, data, len);
19c00ddc
CM		 246}
247
248void btrfs_csum_final(u32 crc, char *result)
249{
7e75bf3f 250 put_unaligned_le32(~crc, result);
19c00ddc
CM		 251}
252
d352ac68
CM		 253/*
254 * compute the csum for a btree block, and either verify it or write it
255 * into the csum field of the block.
256 */
19c00ddc
CM		 257static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
258 int verify)
259{
6c41761f 260 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
607d432d 261 char *result = NULL;
19c00ddc
CM		 262 unsigned long len;
263 unsigned long cur_len;
264 unsigned long offset = BTRFS_CSUM_SIZE;
19c00ddc
CM		 265 char *kaddr;
266 unsigned long map_start;
267 unsigned long map_len;
268 int err;
269 u32 crc = ~(u32)0;
607d432d 270 unsigned long inline_result;
19c00ddc
CM		 271
272 len = buf->len - offset;
d397712b 273 while (len > 0) {
19c00ddc 274 err = map_private_extent_buffer(buf, offset, 32,
a6591715 275 &kaddr, &map_start, &map_len);
d397712b 276 if (err)
19c00ddc 277 return 1;
19c00ddc 278 cur_len = min(len, map_len - (offset - map_start));
b0496686 279 crc = btrfs_csum_data(kaddr + offset - map_start,
19c00ddc
CM		 280 crc, cur_len);
281 len -= cur_len;
282 offset += cur_len;
19c00ddc 283 }
607d432d
JB		 284 if (csum_size > sizeof(inline_result)) {
285 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
286 if (!result)
287 return 1;
288 } else {
289 result = (char *)&inline_result;
290 }
291
19c00ddc
CM		 292 btrfs_csum_final(crc, result);
293
294 if (verify) {
607d432d 295 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 296 u32 val;
297 u32 found = 0;
607d432d 298 memcpy(&found, result, csum_size);
e4204ded 299
607d432d 300 read_extent_buffer(buf, &val, 0, csum_size);
efe120a0
FH		 301 printk_ratelimited(KERN_INFO
302 "BTRFS: %s checksum verify failed on %llu wanted %X found %X "
303 "level %d\n",
304 root->fs_info->sb->s_id, buf->start,
305 val, found, btrfs_header_level(buf));
607d432d
JB		 306 if (result != (char *)&inline_result)
307 kfree(result);
19c00ddc
CM		 308 return 1;
309 }
310 } else {
607d432d 311 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 312 }
607d432d
JB		 313 if (result != (char *)&inline_result)
314 kfree(result);
19c00ddc
CM		 315 return 0;
316}
317
d352ac68
CM		 318/*
319 * we can't consider a given block up to date unless the transid of the
320 * block matches the transid in the parent node's pointer. This is how we
321 * detect blocks that either didn't get written at all or got written
322 * in the wrong place.
323 */
1259ab75 324static int verify_parent_transid(struct extent_io_tree *io_tree,
b9fab919
CM		 325 struct extent_buffer *eb, u64 parent_transid,
326 int atomic)
1259ab75 327{
2ac55d41 328 struct extent_state *cached_state = NULL;
1259ab75 329 int ret;
a26e8c9f
JB		 330 bool need_lock = (current->journal_info ==
331 (void *)BTRFS_SEND_TRANS_STUB);
1259ab75
CM		 332
333 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
334 return 0;
335
b9fab919
CM		 336 if (atomic)
337 return -EAGAIN;
338
a26e8c9f
JB		 339 if (need_lock) {
340 btrfs_tree_read_lock(eb);
341 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
342 }
343
2ac55d41 344 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
d0082371 345 0, &cached_state);
0b32f4bb 346 if (extent_buffer_uptodate(eb) &&
1259ab75
CM		 347 btrfs_header_generation(eb) == parent_transid) {
348 ret = 0;
349 goto out;
350 }
29549aec
WS		 351 printk_ratelimited(KERN_INFO "BTRFS (device %s): parent transid verify failed on %llu wanted %llu found %llu\n",
352 eb->fs_info->sb->s_id, eb->start,
353 parent_transid, btrfs_header_generation(eb));
1259ab75 354 ret = 1;
a26e8c9f
JB		 355
356 /*
357 * Things reading via commit roots that don't have normal protection,
358 * like send, can have a really old block in cache that may point at a
359 * block that has been free'd and re-allocated. So don't clear uptodate
360 * if we find an eb that is under IO (dirty/writeback) because we could
361 * end up reading in the stale data and then writing it back out and
362 * making everybody very sad.
363 */
364 if (!extent_buffer_under_io(eb))
365 clear_extent_buffer_uptodate(eb);
33958dc6 366out:
2ac55d41
JB		 367 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
368 &cached_state, GFP_NOFS);
472b909f
JB		 369 if (need_lock)
370 btrfs_tree_read_unlock_blocking(eb);
1259ab75 371 return ret;
1259ab75
CM		 372}
373
1104a885
DS		 374/*
375 * Return 0 if the superblock checksum type matches the checksum value of that
376 * algorithm. Pass the raw disk superblock data.
377 */
378static int btrfs_check_super_csum(char *raw_disk_sb)
379{
380 struct btrfs_super_block *disk_sb =
381 (struct btrfs_super_block *)raw_disk_sb;
382 u16 csum_type = btrfs_super_csum_type(disk_sb);
383 int ret = 0;
384
385 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
386 u32 crc = ~(u32)0;
387 const int csum_size = sizeof(crc);
388 char result[csum_size];
389
390 /*
391 * The super_block structure does not span the whole
392 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
393 * is filled with zeros and is included in the checkum.
394 */
395 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
396 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
397 btrfs_csum_final(crc, result);
398
399 if (memcmp(raw_disk_sb, result, csum_size))
400 ret = 1;
667e7d94
CM		 401
402 if (ret && btrfs_super_generation(disk_sb) < 10) {
efe120a0
FH		 403 printk(KERN_WARNING
404 "BTRFS: super block crcs don't match, older mkfs detected\n");
667e7d94
CM		 405 ret = 0;
406 }
1104a885
DS		 407 }
408
409 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
efe120a0 410 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
1104a885
DS		 411 csum_type);
412 ret = 1;
413 }
414
415 return ret;
416}
417
d352ac68
CM		 418/*
419 * helper to read a given tree block, doing retries as required when
420 * the checksums don't match and we have alternate mirrors to try.
421 */
f188591e
CM		 422static int btree_read_extent_buffer_pages(struct btrfs_root *root,
423 struct extent_buffer *eb,
ca7a79ad 424 u64 start, u64 parent_transid)
f188591e
CM		 425{
426 struct extent_io_tree *io_tree;
ea466794 427 int failed = 0;
f188591e
CM		 428 int ret;
429 int num_copies = 0;
430 int mirror_num = 0;
ea466794 431 int failed_mirror = 0;
f188591e 432
a826d6dc 433 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
f188591e
CM		 434 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
435 while (1) {
bb82ab88
AJ		 436 ret = read_extent_buffer_pages(io_tree, eb, start,
437 WAIT_COMPLETE,
f188591e 438 btree_get_extent, mirror_num);
256dd1bb
SB		 439 if (!ret) {
440 if (!verify_parent_transid(io_tree, eb,
b9fab919 441 parent_transid, 0))
256dd1bb
SB		 442 break;
443 else
444 ret = -EIO;
445 }
d397712b 446
a826d6dc
JB		 447 /*
448 * This buffer's crc is fine, but its contents are corrupted, so
449 * there is no reason to read the other copies, they won't be
450 * any less wrong.
451 */
452 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
ea466794
JB		 453 break;
454
5d964051 455 num_copies = btrfs_num_copies(root->fs_info,
f188591e 456 eb->start, eb->len);
4235298e 457 if (num_copies == 1)
ea466794 458 break;
4235298e 459
5cf1ab56
JB		 460 if (!failed_mirror) {
461 failed = 1;
462 failed_mirror = eb->read_mirror;
463 }
464
f188591e 465 mirror_num++;
ea466794
JB		 466 if (mirror_num == failed_mirror)
467 mirror_num++;
468
4235298e 469 if (mirror_num > num_copies)
ea466794 470 break;
f188591e 471 }
ea466794 472
c0901581 473 if (failed && !ret && failed_mirror)
ea466794
JB		 474 repair_eb_io_failure(root, eb, failed_mirror);
475
476 return ret;
f188591e 477}
19c00ddc 478
d352ac68 479/*
d397712b
CM		 480 * checksum a dirty tree block before IO. This has extra checks to make sure
481 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 482 */
d397712b 483
b2950863 484static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
19c00ddc 485{
4eee4fa4 486 u64 start = page_offset(page);
19c00ddc 487 u64 found_start;
19c00ddc 488 struct extent_buffer *eb;
f188591e 489
4f2de97a
JB		 490 eb = (struct extent_buffer *)page->private;
491 if (page != eb->pages[0])
492 return 0;
19c00ddc 493 found_start = btrfs_header_bytenr(eb);
fae7f21c 494 if (WARN_ON(found_start != start || !PageUptodate(page)))
4f2de97a 495 return 0;
19c00ddc 496 csum_tree_block(root, eb, 0);
19c00ddc
CM		 497 return 0;
498}
499
2b82032c
YZ		 500static int check_tree_block_fsid(struct btrfs_root *root,
501 struct extent_buffer *eb)
502{
503 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
504 u8 fsid[BTRFS_UUID_SIZE];
505 int ret = 1;
506
0a4e5586 507 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
2b82032c
YZ		 508 while (fs_devices) {
509 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
510 ret = 0;
511 break;
512 }
513 fs_devices = fs_devices->seed;
514 }
515 return ret;
516}
517
a826d6dc 518#define CORRUPT(reason, eb, root, slot) \
efe120a0
FH		 519 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
520 "root=%llu, slot=%d", reason, \
c1c9ff7c 521 btrfs_header_bytenr(eb), root->objectid, slot)
a826d6dc
JB		 522
523static noinline int check_leaf(struct btrfs_root *root,
524 struct extent_buffer *leaf)
525{
526 struct btrfs_key key;
527 struct btrfs_key leaf_key;
528 u32 nritems = btrfs_header_nritems(leaf);
529 int slot;
530
531 if (nritems == 0)
532 return 0;
533
534 /* Check the 0 item */
535 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
536 BTRFS_LEAF_DATA_SIZE(root)) {
537 CORRUPT("invalid item offset size pair", leaf, root, 0);
538 return -EIO;
539 }
540
541 /*
542 * Check to make sure each items keys are in the correct order and their
543 * offsets make sense. We only have to loop through nritems-1 because
544 * we check the current slot against the next slot, which verifies the
545 * next slot's offset+size makes sense and that the current's slot
546 * offset is correct.
547 */
548 for (slot = 0; slot < nritems - 1; slot++) {
549 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
550 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
551
552 /* Make sure the keys are in the right order */
553 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
554 CORRUPT("bad key order", leaf, root, slot);
555 return -EIO;
556 }
557
558 /*
559 * Make sure the offset and ends are right, remember that the
560 * item data starts at the end of the leaf and grows towards the
561 * front.
562 */
563 if (btrfs_item_offset_nr(leaf, slot) !=
564 btrfs_item_end_nr(leaf, slot + 1)) {
565 CORRUPT("slot offset bad", leaf, root, slot);
566 return -EIO;
567 }
568
569 /*
570 * Check to make sure that we don't point outside of the leaf,
571 * just incase all the items are consistent to eachother, but
572 * all point outside of the leaf.
573 */
574 if (btrfs_item_end_nr(leaf, slot) >
575 BTRFS_LEAF_DATA_SIZE(root)) {
576 CORRUPT("slot end outside of leaf", leaf, root, slot);
577 return -EIO;
578 }
579 }
580
581 return 0;
582}
583
facc8a22
MX		 584static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
585 u64 phy_offset, struct page *page,
586 u64 start, u64 end, int mirror)
ce9adaa5 587{
ce9adaa5
CM		 588 u64 found_start;
589 int found_level;
ce9adaa5
CM		 590 struct extent_buffer *eb;
591 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 592 int ret = 0;
727011e0 593 int reads_done;
ce9adaa5 594
ce9adaa5
CM		 595 if (!page->private)
596 goto out;
d397712b 597
4f2de97a 598 eb = (struct extent_buffer *)page->private;
d397712b 599
0b32f4bb
JB		 600 /* the pending IO might have been the only thing that kept this buffer
601 * in memory. Make sure we have a ref for all this other checks
602 */
603 extent_buffer_get(eb);
604
605 reads_done = atomic_dec_and_test(&eb->io_pages);
727011e0
CM		 606 if (!reads_done)
607 goto err;
f188591e 608
5cf1ab56 609 eb->read_mirror = mirror;
ea466794
JB		 610 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
611 ret = -EIO;
612 goto err;
613 }
614
ce9adaa5 615 found_start = btrfs_header_bytenr(eb);
727011e0 616 if (found_start != eb->start) {
29549aec 617 printk_ratelimited(KERN_INFO "BTRFS (device %s): bad tree block start "
193f284d 618 "%llu %llu\n",
29549aec 619 eb->fs_info->sb->s_id, found_start, eb->start);
f188591e 620 ret = -EIO;
ce9adaa5
CM		 621 goto err;
622 }
2b82032c 623 if (check_tree_block_fsid(root, eb)) {
29549aec
WS		 624 printk_ratelimited(KERN_INFO "BTRFS (device %s): bad fsid on block %llu\n",
625 eb->fs_info->sb->s_id, eb->start);
1259ab75
CM		 626 ret = -EIO;
627 goto err;
628 }
ce9adaa5 629 found_level = btrfs_header_level(eb);
1c24c3ce 630 if (found_level >= BTRFS_MAX_LEVEL) {
efe120a0 631 btrfs_info(root->fs_info, "bad tree block level %d",
1c24c3ce
JB		 632 (int)btrfs_header_level(eb));
633 ret = -EIO;
634 goto err;
635 }
ce9adaa5 636
85d4e461
CM		 637 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
638 eb, found_level);
4008c04a 639
ce9adaa5 640 ret = csum_tree_block(root, eb, 1);
a826d6dc 641 if (ret) {
f188591e 642 ret = -EIO;
a826d6dc
JB		 643 goto err;
644 }
645
646 /*
647 * If this is a leaf block and it is corrupt, set the corrupt bit so
648 * that we don't try and read the other copies of this block, just
649 * return -EIO.
650 */
651 if (found_level == 0 && check_leaf(root, eb)) {
652 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
653 ret = -EIO;
654 }
ce9adaa5 655
0b32f4bb
JB		 656 if (!ret)
657 set_extent_buffer_uptodate(eb);
ce9adaa5 658err:
79fb65a1
JB		 659 if (reads_done &&
660 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 661 btree_readahead_hook(root, eb, eb->start, ret);
4bb31e92 662
53b381b3
DW		 663 if (ret) {
664 /*
665 * our io error hook is going to dec the io pages
666 * again, we have to make sure it has something
667 * to decrement
668 */
669 atomic_inc(&eb->io_pages);
0b32f4bb 670 clear_extent_buffer_uptodate(eb);
53b381b3 671 }
0b32f4bb 672 free_extent_buffer(eb);
ce9adaa5 673out:
f188591e 674 return ret;
ce9adaa5
CM		 675}
676
ea466794 677static int btree_io_failed_hook(struct page *page, int failed_mirror)
4bb31e92 678{
4bb31e92
AJ		 679 struct extent_buffer *eb;
680 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
681
4f2de97a 682 eb = (struct extent_buffer *)page->private;
ea466794 683 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
5cf1ab56 684 eb->read_mirror = failed_mirror;
53b381b3 685 atomic_dec(&eb->io_pages);
ea466794 686 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 687 btree_readahead_hook(root, eb, eb->start, -EIO);
4bb31e92
AJ		 688 return -EIO; /* we fixed nothing */
689}
690
ce9adaa5 691static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM		 692{
693 struct end_io_wq *end_io_wq = bio->bi_private;
694 struct btrfs_fs_info *fs_info;
9e0af237
LB		 695 struct btrfs_workqueue *wq;
696 btrfs_work_func_t func;
ce9adaa5 697
ce9adaa5 698 fs_info = end_io_wq->info;
ce9adaa5 699 end_io_wq->error = err;
d20f7043 700
7b6d91da 701 if (bio->bi_rw & REQ_WRITE) {
9e0af237
LB		 702 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) {
703 wq = fs_info->endio_meta_write_workers;
704 func = btrfs_endio_meta_write_helper;
705 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) {
706 wq = fs_info->endio_freespace_worker;
707 func = btrfs_freespace_write_helper;
708 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
709 wq = fs_info->endio_raid56_workers;
710 func = btrfs_endio_raid56_helper;
711 } else {
712 wq = fs_info->endio_write_workers;
713 func = btrfs_endio_write_helper;
714 }
d20f7043 715 } else {
8b110e39
MX		 716 if (unlikely(end_io_wq->metadata ==
717 BTRFS_WQ_ENDIO_DIO_REPAIR)) {
718 wq = fs_info->endio_repair_workers;
719 func = btrfs_endio_repair_helper;
720 } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
9e0af237
LB		 721 wq = fs_info->endio_raid56_workers;
722 func = btrfs_endio_raid56_helper;
723 } else if (end_io_wq->metadata) {
724 wq = fs_info->endio_meta_workers;
725 func = btrfs_endio_meta_helper;
726 } else {
727 wq = fs_info->endio_workers;
728 func = btrfs_endio_helper;
729 }
d20f7043 730 }
9e0af237
LB		 731
732 btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL);
733 btrfs_queue_work(wq, &end_io_wq->work);
ce9adaa5
CM		 734}
735
0cb59c99
JB		 736/*
737 * For the metadata arg you want
738 *
739 * 0 - if data
740 * 1 - if normal metadta
741 * 2 - if writing to the free space cache area
53b381b3 742 * 3 - raid parity work
0cb59c99 743 */
22c59948
CM		 744int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
745 int metadata)
0b86a832 746{
ce9adaa5 747 struct end_io_wq *end_io_wq;
8b110e39 748
ce9adaa5
CM		 749 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
750 if (!end_io_wq)
751 return -ENOMEM;
752
753 end_io_wq->private = bio->bi_private;
754 end_io_wq->end_io = bio->bi_end_io;
22c59948 755 end_io_wq->info = info;
ce9adaa5
CM		 756 end_io_wq->error = 0;
757 end_io_wq->bio = bio;
22c59948 758 end_io_wq->metadata = metadata;
ce9adaa5
CM		 759
760 bio->bi_private = end_io_wq;
761 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 762 return 0;
763}
764
b64a2851 765unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 766{
4854ddd0 767 unsigned long limit = min_t(unsigned long,
5cdc7ad3 768 info->thread_pool_size,
4854ddd0
CM		 769 info->fs_devices->open_devices);
770 return 256 * limit;
771}
0986fe9e 772
4a69a410
CM		 773static void run_one_async_start(struct btrfs_work *work)
774{
4a69a410 775 struct async_submit_bio *async;
79787eaa 776 int ret;
4a69a410
CM		 777
778 async = container_of(work, struct async_submit_bio, work);
79787eaa
JM		 779 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
780 async->mirror_num, async->bio_flags,
781 async->bio_offset);
782 if (ret)
783 async->error = ret;
4a69a410
CM		 784}
785
786static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 787{
788 struct btrfs_fs_info *fs_info;
789 struct async_submit_bio *async;
4854ddd0 790 int limit;
8b712842
CM		 791
792 async = container_of(work, struct async_submit_bio, work);
793 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 794
b64a2851 795 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 796 limit = limit * 2 / 3;
797
66657b31 798 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
b64a2851 799 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 800 wake_up(&fs_info->async_submit_wait);
801
79787eaa
JM		 802 /* If an error occured we just want to clean up the bio and move on */
803 if (async->error) {
804 bio_endio(async->bio, async->error);
805 return;
806 }
807
4a69a410 808 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 809 async->mirror_num, async->bio_flags,
810 async->bio_offset);
4a69a410
CM		 811}
812
813static void run_one_async_free(struct btrfs_work *work)
814{
815 struct async_submit_bio *async;
816
817 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 818 kfree(async);
819}
820
44b8bd7e
CM		 821int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
822 int rw, struct bio *bio, int mirror_num,
c8b97818 823 unsigned long bio_flags,
eaf25d93 824 u64 bio_offset,
4a69a410
CM		 825 extent_submit_bio_hook_t *submit_bio_start,
826 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 827{
828 struct async_submit_bio *async;
829
830 async = kmalloc(sizeof(*async), GFP_NOFS);
831 if (!async)
832 return -ENOMEM;
833
834 async->inode = inode;
835 async->rw = rw;
836 async->bio = bio;
837 async->mirror_num = mirror_num;
4a69a410
CM		 838 async->submit_bio_start = submit_bio_start;
839 async->submit_bio_done = submit_bio_done;
840
9e0af237 841 btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start,
5cdc7ad3 842 run_one_async_done, run_one_async_free);
4a69a410 843
c8b97818 844 async->bio_flags = bio_flags;
eaf25d93 845 async->bio_offset = bio_offset;
8c8bee1d 846
79787eaa
JM		 847 async->error = 0;
848
cb03c743 849 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 850
7b6d91da 851 if (rw & REQ_SYNC)
5cdc7ad3 852 btrfs_set_work_high_priority(&async->work);
d313d7a3 853
5cdc7ad3 854 btrfs_queue_work(fs_info->workers, &async->work);
9473f16c 855
d397712b 856 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 857 atomic_read(&fs_info->nr_async_submits)) {
858 wait_event(fs_info->async_submit_wait,
859 (atomic_read(&fs_info->nr_async_submits) == 0));
860 }
861
44b8bd7e
CM		 862 return 0;
863}
864
ce3ed71a
CM		 865static int btree_csum_one_bio(struct bio *bio)
866{
2c30c71b 867 struct bio_vec *bvec;
ce3ed71a 868 struct btrfs_root *root;
2c30c71b 869 int i, ret = 0;
ce3ed71a 870
2c30c71b 871 bio_for_each_segment_all(bvec, bio, i) {
ce3ed71a 872 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
79787eaa
JM		 873 ret = csum_dirty_buffer(root, bvec->bv_page);
874 if (ret)
875 break;
ce3ed71a 876 }
2c30c71b 877
79787eaa 878 return ret;
ce3ed71a
CM		 879}
880
4a69a410
CM		 881static int __btree_submit_bio_start(struct inode *inode, int rw,
882 struct bio *bio, int mirror_num,
eaf25d93
CM		 883 unsigned long bio_flags,
884 u64 bio_offset)
22c59948 885{
8b712842
CM		 886 /*
887 * when we're called for a write, we're already in the async
5443be45 888 * submission context. Just jump into btrfs_map_bio
8b712842 889 */
79787eaa 890 return btree_csum_one_bio(bio);
4a69a410 891}
22c59948 892
4a69a410 893static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 894 int mirror_num, unsigned long bio_flags,
895 u64 bio_offset)
4a69a410 896{
61891923
SB		 897 int ret;
898
8b712842 899 /*
4a69a410
CM		 900 * when we're called for a write, we're already in the async
901 * submission context. Just jump into btrfs_map_bio
8b712842 902 */
61891923
SB		 903 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
904 if (ret)
905 bio_endio(bio, ret);
906 return ret;
0b86a832
CM		 907}
908
de0022b9
JB		 909static int check_async_write(struct inode *inode, unsigned long bio_flags)
910{
911 if (bio_flags & EXTENT_BIO_TREE_LOG)
912 return 0;
913#ifdef CONFIG_X86
914 if (cpu_has_xmm4_2)
915 return 0;
916#endif
917 return 1;
918}
919
44b8bd7e 920static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 921 int mirror_num, unsigned long bio_flags,
922 u64 bio_offset)
44b8bd7e 923{
de0022b9 924 int async = check_async_write(inode, bio_flags);
cad321ad
CM		 925 int ret;
926
7b6d91da 927 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 928 /*
929 * called for a read, do the setup so that checksum validation
930 * can happen in the async kernel threads
931 */
f3f266ab
CM		 932 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
933 bio, 1);
1d4284bd 934 if (ret)
61891923
SB		 935 goto out_w_error;
936 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
937 mirror_num, 0);
de0022b9
JB		 938 } else if (!async) {
939 ret = btree_csum_one_bio(bio);
940 if (ret)
61891923
SB		 941 goto out_w_error;
942 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
943 mirror_num, 0);
944 } else {
945 /*
946 * kthread helpers are used to submit writes so that
947 * checksumming can happen in parallel across all CPUs
948 */
949 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
950 inode, rw, bio, mirror_num, 0,
951 bio_offset,
952 __btree_submit_bio_start,
953 __btree_submit_bio_done);
44b8bd7e 954 }
d313d7a3 955
61891923
SB		 956 if (ret) {
957out_w_error:
958 bio_endio(bio, ret);
959 }
960 return ret;
44b8bd7e
CM		 961}
962
3dd1462e 963#ifdef CONFIG_MIGRATION
784b4e29 964static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG		 965 struct page *newpage, struct page *page,
966 enum migrate_mode mode)
784b4e29
CM		 967{
968 /*
969 * we can't safely write a btree page from here,
970 * we haven't done the locking hook
971 */
972 if (PageDirty(page))
973 return -EAGAIN;
974 /*
975 * Buffers may be managed in a filesystem specific way.
976 * We must have no buffers or drop them.
977 */
978 if (page_has_private(page) &&
979 !try_to_release_page(page, GFP_KERNEL))
980 return -EAGAIN;
a6bc32b8 981 return migrate_page(mapping, newpage, page, mode);
784b4e29 982}
3dd1462e 983#endif
784b4e29 984
0da5468f
CM		 985
986static int btree_writepages(struct address_space *mapping,
987 struct writeback_control *wbc)
988{
e2d84521
MX		 989 struct btrfs_fs_info *fs_info;
990 int ret;
991
d8d5f3e1 992 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM		 993
994 if (wbc->for_kupdate)
995 return 0;
996
e2d84521 997 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 998 /* this is a bit racy, but that's ok */
e2d84521
MX		 999 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
1000 BTRFS_DIRTY_METADATA_THRESH);
1001 if (ret < 0)
793955bc 1002 return 0;
793955bc 1003 }
0b32f4bb 1004 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM		 1005}
1006
b2950863 1007static int btree_readpage(struct file *file, struct page *page)
5f39d397 1008{
d1310b2e
CM		 1009 struct extent_io_tree *tree;
1010 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 1011 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 1012}
22b0ebda 1013
70dec807 1014static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 1015{
98509cfc 1016 if (PageWriteback(page) || PageDirty(page))
d397712b 1017 return 0;
0c4e538b 1018
f7a52a40 1019 return try_release_extent_buffer(page);
d98237b3
CM		 1020}
1021
d47992f8
LC		 1022static void btree_invalidatepage(struct page *page, unsigned int offset,
1023 unsigned int length)
d98237b3 1024{
d1310b2e
CM		 1025 struct extent_io_tree *tree;
1026 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 1027 extent_invalidatepage(tree, page, offset);
1028 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 1029 if (PagePrivate(page)) {
efe120a0
FH		 1030 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1031 "page private not zero on page %llu",
1032 (unsigned long long)page_offset(page));
9ad6b7bc
CM		 1033 ClearPagePrivate(page);
1034 set_page_private(page, 0);
1035 page_cache_release(page);
1036 }
d98237b3
CM		 1037}
1038
0b32f4bb
JB		 1039static int btree_set_page_dirty(struct page *page)
1040{
bb146eb2 1041#ifdef DEBUG
0b32f4bb
JB		 1042 struct extent_buffer *eb;
1043
1044 BUG_ON(!PagePrivate(page));
1045 eb = (struct extent_buffer *)page->private;
1046 BUG_ON(!eb);
1047 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1048 BUG_ON(!atomic_read(&eb->refs));
1049 btrfs_assert_tree_locked(eb);
bb146eb2 1050#endif
0b32f4bb
JB		 1051 return __set_page_dirty_nobuffers(page);
1052}
1053
7f09410b 1054static const struct address_space_operations btree_aops = {
d98237b3 1055 .readpage = btree_readpage,
0da5468f 1056 .writepages = btree_writepages,
5f39d397
CM		 1057 .releasepage = btree_releasepage,
1058 .invalidatepage = btree_invalidatepage,
5a92bc88 1059#ifdef CONFIG_MIGRATION
784b4e29 1060 .migratepage = btree_migratepage,
5a92bc88 1061#endif
0b32f4bb 1062 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM		 1063};
1064
6197d86e 1065void readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
090d1875 1066{
5f39d397
CM		 1067 struct extent_buffer *buf = NULL;
1068 struct inode *btree_inode = root->fs_info->btree_inode;
090d1875 1069
db94535d 1070 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 1071 if (!buf)
6197d86e 1072 return;
d1310b2e 1073 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
bb82ab88 1074 buf, 0, WAIT_NONE, btree_get_extent, 0);
5f39d397 1075 free_extent_buffer(buf);
090d1875
CM		 1076}
1077
ab0fff03
AJ		 1078int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1079 int mirror_num, struct extent_buffer **eb)
1080{
1081 struct extent_buffer *buf = NULL;
1082 struct inode *btree_inode = root->fs_info->btree_inode;
1083 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1084 int ret;
1085
1086 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1087 if (!buf)
1088 return 0;
1089
1090 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1091
1092 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1093 btree_get_extent, mirror_num);
1094 if (ret) {
1095 free_extent_buffer(buf);
1096 return ret;
1097 }
1098
1099 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1100 free_extent_buffer(buf);
1101 return -EIO;
0b32f4bb 1102 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ		 1103 *eb = buf;
1104 } else {
1105 free_extent_buffer(buf);
1106 }
1107 return 0;
1108}
1109
0999df54
CM		 1110struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1111 u64 bytenr, u32 blocksize)
1112{
f28491e0 1113 return find_extent_buffer(root->fs_info, bytenr);
0999df54
CM		 1114}
1115
1116struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1117 u64 bytenr, u32 blocksize)
1118{
faa2dbf0
JB		 1119#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1120 if (unlikely(test_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state)))
1121 return alloc_test_extent_buffer(root->fs_info, bytenr,
1122 blocksize);
1123#endif
f28491e0 1124 return alloc_extent_buffer(root->fs_info, bytenr, blocksize);
0999df54
CM		 1125}
1126
1127
e02119d5
CM		 1128int btrfs_write_tree_block(struct extent_buffer *buf)
1129{
727011e0 1130 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1131 buf->start + buf->len - 1);
e02119d5
CM		 1132}
1133
1134int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1135{
727011e0 1136 return filemap_fdatawait_range(buf->pages[0]->mapping,
8aa38c31 1137 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1138}
1139
0999df54 1140struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ce86cd59 1141 u64 parent_transid)
0999df54
CM		 1142{
1143 struct extent_buffer *buf = NULL;
0999df54
CM		 1144 int ret;
1145
ce86cd59 1146 buf = btrfs_find_create_tree_block(root, bytenr, root->nodesize);
0999df54
CM		 1147 if (!buf)
1148 return NULL;
0999df54 1149
ca7a79ad 1150 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
0f0fe8f7
FDBM		 1151 if (ret) {
1152 free_extent_buffer(buf);
1153 return NULL;
1154 }
5f39d397 1155 return buf;
ce9adaa5 1156
eb60ceac
CM		 1157}
1158
d5c13f92
JM		 1159void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1160 struct extent_buffer *buf)
ed2ff2cb 1161{
e2d84521
MX		 1162 struct btrfs_fs_info *fs_info = root->fs_info;
1163
55c69072 1164 if (btrfs_header_generation(buf) ==
e2d84521 1165 fs_info->running_transaction->transid) {
b9447ef8 1166 btrfs_assert_tree_locked(buf);
b4ce94de 1167
b9473439 1168 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
e2d84521
MX		 1169 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1170 -buf->len,
1171 fs_info->dirty_metadata_batch);
ed7b63eb
JB		 1172 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1173 btrfs_set_lock_blocking(buf);
1174 clear_extent_buffer_dirty(buf);
1175 }
925baedd 1176 }
5f39d397
CM		 1177}
1178
8257b2dc
MX		 1179static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1180{
1181 struct btrfs_subvolume_writers *writers;
1182 int ret;
1183
1184 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1185 if (!writers)
1186 return ERR_PTR(-ENOMEM);
1187
1188 ret = percpu_counter_init(&writers->counter, 0);
1189 if (ret < 0) {
1190 kfree(writers);
1191 return ERR_PTR(ret);
1192 }
1193
1194 init_waitqueue_head(&writers->wait);
1195 return writers;
1196}
1197
1198static void
1199btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1200{
1201 percpu_counter_destroy(&writers->counter);
1202 kfree(writers);
1203}
1204
707e8a07
DS		 1205static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize,
1206 struct btrfs_root *root, struct btrfs_fs_info *fs_info,
143bede5 1207 u64 objectid)
d97e63b6 1208{
cfaa7295 1209 root->node = NULL;
a28ec197 1210 root->commit_root = NULL;
db94535d
CM		 1211 root->sectorsize = sectorsize;
1212 root->nodesize = nodesize;
87ee04eb 1213 root->stripesize = stripesize;
27cdeb70 1214 root->state = 0;
d68fc57b 1215 root->orphan_cleanup_state = 0;
0b86a832 1216
0f7d52f4
CM		 1217 root->objectid = objectid;
1218 root->last_trans = 0;
13a8a7c8 1219 root->highest_objectid = 0;
eb73c1b7 1220 root->nr_delalloc_inodes = 0;
199c2a9c 1221 root->nr_ordered_extents = 0;
58176a96 1222 root->name = NULL;
6bef4d31 1223 root->inode_tree = RB_ROOT;
16cdcec7 1224 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1225 root->block_rsv = NULL;
d68fc57b 1226 root->orphan_block_rsv = NULL;
0b86a832
CM		 1227
1228 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1229 INIT_LIST_HEAD(&root->root_list);
eb73c1b7
MX		 1230 INIT_LIST_HEAD(&root->delalloc_inodes);
1231 INIT_LIST_HEAD(&root->delalloc_root);
199c2a9c
MX		 1232 INIT_LIST_HEAD(&root->ordered_extents);
1233 INIT_LIST_HEAD(&root->ordered_root);
2ab28f32
JB		 1234 INIT_LIST_HEAD(&root->logged_list[0]);
1235 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1236 spin_lock_init(&root->orphan_lock);
5d4f98a2 1237 spin_lock_init(&root->inode_lock);
eb73c1b7 1238 spin_lock_init(&root->delalloc_lock);
199c2a9c 1239 spin_lock_init(&root->ordered_extent_lock);
f0486c68 1240 spin_lock_init(&root->accounting_lock);
2ab28f32
JB		 1241 spin_lock_init(&root->log_extents_lock[0]);
1242 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1243 mutex_init(&root->objectid_mutex);
e02119d5 1244 mutex_init(&root->log_mutex);
31f3d255 1245 mutex_init(&root->ordered_extent_mutex);
573bfb72 1246 mutex_init(&root->delalloc_mutex);
7237f183
YZ		 1247 init_waitqueue_head(&root->log_writer_wait);
1248 init_waitqueue_head(&root->log_commit_wait[0]);
1249 init_waitqueue_head(&root->log_commit_wait[1]);
8b050d35
MX		 1250 INIT_LIST_HEAD(&root->log_ctxs[0]);
1251 INIT_LIST_HEAD(&root->log_ctxs[1]);
7237f183
YZ		 1252 atomic_set(&root->log_commit[0], 0);
1253 atomic_set(&root->log_commit[1], 0);
1254 atomic_set(&root->log_writers, 0);
2ecb7923 1255 atomic_set(&root->log_batch, 0);
8a35d95f 1256 atomic_set(&root->orphan_inodes, 0);
b0feb9d9 1257 atomic_set(&root->refs, 1);
8257b2dc 1258 atomic_set(&root->will_be_snapshoted, 0);
7237f183 1259 root->log_transid = 0;
d1433deb 1260 root->log_transid_committed = -1;
257c62e1 1261 root->last_log_commit = 0;
06ea65a3
JB		 1262 if (fs_info)
1263 extent_io_tree_init(&root->dirty_log_pages,
1264 fs_info->btree_inode->i_mapping);
017e5369 1265
3768f368
CM		 1266 memset(&root->root_key, 0, sizeof(root->root_key));
1267 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1268 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1269 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
06ea65a3
JB		 1270 if (fs_info)
1271 root->defrag_trans_start = fs_info->generation;
1272 else
1273 root->defrag_trans_start = 0;
58176a96 1274 init_completion(&root->kobj_unregister);
4d775673 1275 root->root_key.objectid = objectid;
0ee5dc67 1276 root->anon_dev = 0;
8ea05e3a 1277
5f3ab90a 1278 spin_lock_init(&root->root_item_lock);
3768f368
CM		 1279}
1280
f84a8bd6 1281static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
6f07e42e
AV		 1282{
1283 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1284 if (root)
1285 root->fs_info = fs_info;
1286 return root;
1287}
1288
06ea65a3
JB		 1289#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1290/* Should only be used by the testing infrastructure */
1291struct btrfs_root *btrfs_alloc_dummy_root(void)
1292{
1293 struct btrfs_root *root;
1294
1295 root = btrfs_alloc_root(NULL);
1296 if (!root)
1297 return ERR_PTR(-ENOMEM);
707e8a07 1298 __setup_root(4096, 4096, 4096, root, NULL, 1);
27cdeb70 1299 set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state);
faa2dbf0 1300 root->alloc_bytenr = 0;
06ea65a3
JB		 1301
1302 return root;
1303}
1304#endif
1305
20897f5c
AJ		 1306struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1307 struct btrfs_fs_info *fs_info,
1308 u64 objectid)
1309{
1310 struct extent_buffer *leaf;
1311 struct btrfs_root *tree_root = fs_info->tree_root;
1312 struct btrfs_root *root;
1313 struct btrfs_key key;
1314 int ret = 0;
6463fe58 1315 uuid_le uuid;
20897f5c
AJ		 1316
1317 root = btrfs_alloc_root(fs_info);
1318 if (!root)
1319 return ERR_PTR(-ENOMEM);
1320
707e8a07
DS		 1321 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1322 tree_root->stripesize, root, fs_info, objectid);
20897f5c
AJ		 1323 root->root_key.objectid = objectid;
1324 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1325 root->root_key.offset = 0;
1326
707e8a07 1327 leaf = btrfs_alloc_free_block(trans, root, root->nodesize,
20897f5c
AJ		 1328 0, objectid, NULL, 0, 0, 0);
1329 if (IS_ERR(leaf)) {
1330 ret = PTR_ERR(leaf);
1dd05682 1331 leaf = NULL;
20897f5c
AJ		 1332 goto fail;
1333 }
1334
20897f5c
AJ		 1335 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1336 btrfs_set_header_bytenr(leaf, leaf->start);
1337 btrfs_set_header_generation(leaf, trans->transid);
1338 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1339 btrfs_set_header_owner(leaf, objectid);
1340 root->node = leaf;
1341
0a4e5586 1342 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
20897f5c
AJ		 1343 BTRFS_FSID_SIZE);
1344 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
b308bc2f 1345 btrfs_header_chunk_tree_uuid(leaf),
20897f5c
AJ		 1346 BTRFS_UUID_SIZE);
1347 btrfs_mark_buffer_dirty(leaf);
1348
1349 root->commit_root = btrfs_root_node(root);
27cdeb70 1350 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
20897f5c
AJ		 1351
1352 root->root_item.flags = 0;
1353 root->root_item.byte_limit = 0;
1354 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1355 btrfs_set_root_generation(&root->root_item, trans->transid);
1356 btrfs_set_root_level(&root->root_item, 0);
1357 btrfs_set_root_refs(&root->root_item, 1);
1358 btrfs_set_root_used(&root->root_item, leaf->len);
1359 btrfs_set_root_last_snapshot(&root->root_item, 0);
1360 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB		 1361 uuid_le_gen(&uuid);
1362 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ		 1363 root->root_item.drop_level = 0;
1364
1365 key.objectid = objectid;
1366 key.type = BTRFS_ROOT_ITEM_KEY;
1367 key.offset = 0;
1368 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1369 if (ret)
1370 goto fail;
1371
1372 btrfs_tree_unlock(leaf);
1373
1dd05682
TI		 1374 return root;
1375
20897f5c 1376fail:
1dd05682
TI		 1377 if (leaf) {
1378 btrfs_tree_unlock(leaf);
59885b39 1379 free_extent_buffer(root->commit_root);
1dd05682
TI		 1380 free_extent_buffer(leaf);
1381 }
1382 kfree(root);
20897f5c 1383
1dd05682 1384 return ERR_PTR(ret);
20897f5c
AJ		 1385}
1386
7237f183
YZ		 1387static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1388 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1389{
1390 struct btrfs_root *root;
1391 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1392 struct extent_buffer *leaf;
e02119d5 1393
6f07e42e 1394 root = btrfs_alloc_root(fs_info);
e02119d5 1395 if (!root)
7237f183 1396 return ERR_PTR(-ENOMEM);
e02119d5 1397
707e8a07
DS		 1398 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1399 tree_root->stripesize, root, fs_info,
1400 BTRFS_TREE_LOG_OBJECTID);
e02119d5
CM		 1401
1402 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1403 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1404 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
27cdeb70 1405
7237f183 1406 /*
27cdeb70
MX		 1407 * DON'T set REF_COWS for log trees
1408 *
7237f183
YZ		 1409 * log trees do not get reference counted because they go away
1410 * before a real commit is actually done. They do store pointers
1411 * to file data extents, and those reference counts still get
1412 * updated (along with back refs to the log tree).
1413 */
e02119d5 1414
707e8a07 1415 leaf = btrfs_alloc_free_block(trans, root, root->nodesize, 0,
66d7e7f0 1416 BTRFS_TREE_LOG_OBJECTID, NULL,
5581a51a 1417 0, 0, 0);
7237f183
YZ		 1418 if (IS_ERR(leaf)) {
1419 kfree(root);
1420 return ERR_CAST(leaf);
1421 }
e02119d5 1422
5d4f98a2
YZ		 1423 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1424 btrfs_set_header_bytenr(leaf, leaf->start);
1425 btrfs_set_header_generation(leaf, trans->transid);
1426 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1427 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1428 root->node = leaf;
e02119d5
CM		 1429
1430 write_extent_buffer(root->node, root->fs_info->fsid,
0a4e5586 1431 btrfs_header_fsid(), BTRFS_FSID_SIZE);
e02119d5
CM		 1432 btrfs_mark_buffer_dirty(root->node);
1433 btrfs_tree_unlock(root->node);
7237f183
YZ		 1434 return root;
1435}
1436
1437int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1438 struct btrfs_fs_info *fs_info)
1439{
1440 struct btrfs_root *log_root;
1441
1442 log_root = alloc_log_tree(trans, fs_info);
1443 if (IS_ERR(log_root))
1444 return PTR_ERR(log_root);
1445 WARN_ON(fs_info->log_root_tree);
1446 fs_info->log_root_tree = log_root;
1447 return 0;
1448}
1449
1450int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1451 struct btrfs_root *root)
1452{
1453 struct btrfs_root *log_root;
1454 struct btrfs_inode_item *inode_item;
1455
1456 log_root = alloc_log_tree(trans, root->fs_info);
1457 if (IS_ERR(log_root))
1458 return PTR_ERR(log_root);
1459
1460 log_root->last_trans = trans->transid;
1461 log_root->root_key.offset = root->root_key.objectid;
1462
1463 inode_item = &log_root->root_item.inode;
3cae210f
QW		 1464 btrfs_set_stack_inode_generation(inode_item, 1);
1465 btrfs_set_stack_inode_size(inode_item, 3);
1466 btrfs_set_stack_inode_nlink(inode_item, 1);
707e8a07 1467 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
3cae210f 1468 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
7237f183 1469
5d4f98a2 1470 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1471
1472 WARN_ON(root->log_root);
1473 root->log_root = log_root;
1474 root->log_transid = 0;
d1433deb 1475 root->log_transid_committed = -1;
257c62e1 1476 root->last_log_commit = 0;
e02119d5
CM		 1477 return 0;
1478}
1479
35a3621b
SB		 1480static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1481 struct btrfs_key *key)
e02119d5
CM		 1482{
1483 struct btrfs_root *root;
1484 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1485 struct btrfs_path *path;
84234f3a 1486 u64 generation;
cb517eab 1487 int ret;
0f7d52f4 1488
cb517eab
MX		 1489 path = btrfs_alloc_path();
1490 if (!path)
0f7d52f4 1491 return ERR_PTR(-ENOMEM);
cb517eab
MX		 1492
1493 root = btrfs_alloc_root(fs_info);
1494 if (!root) {
1495 ret = -ENOMEM;
1496 goto alloc_fail;
0f7d52f4
CM		 1497 }
1498
707e8a07
DS		 1499 __setup_root(tree_root->nodesize, tree_root->sectorsize,
1500 tree_root->stripesize, root, fs_info, key->objectid);
0f7d52f4 1501
cb517eab
MX		 1502 ret = btrfs_find_root(tree_root, key, path,
1503 &root->root_item, &root->root_key);
0f7d52f4 1504 if (ret) {
13a8a7c8
YZ		 1505 if (ret > 0)
1506 ret = -ENOENT;
cb517eab 1507 goto find_fail;
0f7d52f4 1508 }
13a8a7c8 1509
84234f3a 1510 generation = btrfs_root_generation(&root->root_item);
db94535d 1511 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
ce86cd59 1512 generation);
cb517eab
MX		 1513 if (!root->node) {
1514 ret = -ENOMEM;
1515 goto find_fail;
1516 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1517 ret = -EIO;
1518 goto read_fail;
416bc658 1519 }
5d4f98a2 1520 root->commit_root = btrfs_root_node(root);
13a8a7c8 1521out:
cb517eab
MX		 1522 btrfs_free_path(path);
1523 return root;
1524
1525read_fail:
1526 free_extent_buffer(root->node);
1527find_fail:
1528 kfree(root);
1529alloc_fail:
1530 root = ERR_PTR(ret);
1531 goto out;
1532}
1533
1534struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1535 struct btrfs_key *location)
1536{
1537 struct btrfs_root *root;
1538
1539 root = btrfs_read_tree_root(tree_root, location);
1540 if (IS_ERR(root))
1541 return root;
1542
1543 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
27cdeb70 1544 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
08fe4db1
LZ		 1545 btrfs_check_and_init_root_item(&root->root_item);
1546 }
13a8a7c8 1547
5eda7b5e
CM		 1548 return root;
1549}
1550
cb517eab
MX		 1551int btrfs_init_fs_root(struct btrfs_root *root)
1552{
1553 int ret;
8257b2dc 1554 struct btrfs_subvolume_writers *writers;
cb517eab
MX		 1555
1556 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1557 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1558 GFP_NOFS);
1559 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1560 ret = -ENOMEM;
1561 goto fail;
1562 }
1563
8257b2dc
MX		 1564 writers = btrfs_alloc_subvolume_writers();
1565 if (IS_ERR(writers)) {
1566 ret = PTR_ERR(writers);
1567 goto fail;
1568 }
1569 root->subv_writers = writers;
1570
cb517eab 1571 btrfs_init_free_ino_ctl(root);
57cdc8db
DS		 1572 spin_lock_init(&root->ino_cache_lock);
1573 init_waitqueue_head(&root->ino_cache_wait);
cb517eab
MX		 1574
1575 ret = get_anon_bdev(&root->anon_dev);
1576 if (ret)
8257b2dc 1577 goto free_writers;
cb517eab 1578 return 0;
8257b2dc
MX		 1579
1580free_writers:
1581 btrfs_free_subvolume_writers(root->subv_writers);
cb517eab
MX		 1582fail:
1583 kfree(root->free_ino_ctl);
1584 kfree(root->free_ino_pinned);
1585 return ret;
1586}
1587
171170c1
ST		 1588static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1589 u64 root_id)
cb517eab
MX		 1590{
1591 struct btrfs_root *root;
1592
1593 spin_lock(&fs_info->fs_roots_radix_lock);
1594 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1595 (unsigned long)root_id);
1596 spin_unlock(&fs_info->fs_roots_radix_lock);
1597 return root;
1598}
1599
1600int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1601 struct btrfs_root *root)
1602{
1603 int ret;
1604
1605 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1606 if (ret)
1607 return ret;
1608
1609 spin_lock(&fs_info->fs_roots_radix_lock);
1610 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1611 (unsigned long)root->root_key.objectid,
1612 root);
1613 if (ret == 0)
27cdeb70 1614 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
cb517eab
MX		 1615 spin_unlock(&fs_info->fs_roots_radix_lock);
1616 radix_tree_preload_end();
1617
1618 return ret;
1619}
1620
c00869f1
MX		 1621struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1622 struct btrfs_key *location,
1623 bool check_ref)
5eda7b5e
CM		 1624{
1625 struct btrfs_root *root;
1626 int ret;
1627
edbd8d4e
CM		 1628 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1629 return fs_info->tree_root;
1630 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1631 return fs_info->extent_root;
8f18cf13
CM		 1632 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1633 return fs_info->chunk_root;
1634 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1635 return fs_info->dev_root;
0403e47e
YZ		 1636 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1637 return fs_info->csum_root;
bcef60f2
AJ		 1638 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1639 return fs_info->quota_root ? fs_info->quota_root :
1640 ERR_PTR(-ENOENT);
f7a81ea4
SB		 1641 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1642 return fs_info->uuid_root ? fs_info->uuid_root :
1643 ERR_PTR(-ENOENT);
4df27c4d 1644again:
cb517eab 1645 root = btrfs_lookup_fs_root(fs_info, location->objectid);
48475471 1646 if (root) {
c00869f1 1647 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
48475471 1648 return ERR_PTR(-ENOENT);
5eda7b5e 1649 return root;
48475471 1650 }
5eda7b5e 1651
cb517eab 1652 root = btrfs_read_fs_root(fs_info->tree_root, location);
5eda7b5e
CM		 1653 if (IS_ERR(root))
1654 return root;
3394e160 1655
c00869f1 1656 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
cb517eab 1657 ret = -ENOENT;
581bb050 1658 goto fail;
35a30d7c 1659 }
581bb050 1660
cb517eab 1661 ret = btrfs_init_fs_root(root);
ac08aedf
CM		 1662 if (ret)
1663 goto fail;
3394e160 1664
3f870c28
KN		 1665 ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID,
1666 location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL);
d68fc57b
YZ		 1667 if (ret < 0)
1668 goto fail;
1669 if (ret == 0)
27cdeb70 1670 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
d68fc57b 1671
cb517eab 1672 ret = btrfs_insert_fs_root(fs_info, root);
0f7d52f4 1673 if (ret) {
4df27c4d
YZ		 1674 if (ret == -EEXIST) {
1675 free_fs_root(root);
1676 goto again;
1677 }
1678 goto fail;
0f7d52f4 1679 }
edbd8d4e 1680 return root;
4df27c4d
YZ		 1681fail:
1682 free_fs_root(root);
1683 return ERR_PTR(ret);
edbd8d4e
CM		 1684}
1685
04160088
CM		 1686static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1687{
1688 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1689 int ret = 0;
04160088
CM		 1690 struct btrfs_device *device;
1691 struct backing_dev_info *bdi;
b7967db7 1692
1f78160c
XG		 1693 rcu_read_lock();
1694 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1695 if (!device->bdev)
1696 continue;
04160088
CM		 1697 bdi = blk_get_backing_dev_info(device->bdev);
1698 if (bdi && bdi_congested(bdi, bdi_bits)) {
1699 ret = 1;
1700 break;
1701 }
1702 }
1f78160c 1703 rcu_read_unlock();
04160088
CM		 1704 return ret;
1705}
1706
04160088
CM		 1707static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1708{
ad081f14
JA		 1709 int err;
1710
1711 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1712 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1713 if (err)
1714 return err;
1715
4575c9cc 1716 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1717 bdi->congested_fn = btrfs_congested_fn;
1718 bdi->congested_data = info;
1719 return 0;
1720}
1721
8b712842
CM		 1722/*
1723 * called by the kthread helper functions to finally call the bio end_io
1724 * functions. This is where read checksum verification actually happens
1725 */
1726static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1727{
ce9adaa5 1728 struct bio *bio;
8b712842 1729 struct end_io_wq *end_io_wq;
ce9adaa5 1730 int error;
ce9adaa5 1731
8b712842
CM		 1732 end_io_wq = container_of(work, struct end_io_wq, work);
1733 bio = end_io_wq->bio;
ce9adaa5 1734
8b712842
CM		 1735 error = end_io_wq->error;
1736 bio->bi_private = end_io_wq->private;
1737 bio->bi_end_io = end_io_wq->end_io;
1738 kfree(end_io_wq);
bc1e79ac 1739 bio_endio_nodec(bio, error);
44b8bd7e
CM		 1740}
1741
a74a4b97
CM		 1742static int cleaner_kthread(void *arg)
1743{
1744 struct btrfs_root *root = arg;
d0278245 1745 int again;
a74a4b97
CM		 1746
1747 do {
d0278245 1748 again = 0;
a74a4b97 1749
d0278245 1750 /* Make the cleaner go to sleep early. */
babbf170 1751 if (btrfs_need_cleaner_sleep(root))
d0278245
MX		 1752 goto sleep;
1753
1754 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1755 goto sleep;
1756
dc7f370c
MX		 1757 /*
1758 * Avoid the problem that we change the status of the fs
1759 * during the above check and trylock.
1760 */
babbf170 1761 if (btrfs_need_cleaner_sleep(root)) {
dc7f370c
MX		 1762 mutex_unlock(&root->fs_info->cleaner_mutex);
1763 goto sleep;
76dda93c 1764 }
a74a4b97 1765
d0278245 1766 btrfs_run_delayed_iputs(root);
47ab2a6c 1767 btrfs_delete_unused_bgs(root->fs_info);
d0278245
MX		 1768 again = btrfs_clean_one_deleted_snapshot(root);
1769 mutex_unlock(&root->fs_info->cleaner_mutex);
1770
1771 /*
05323cd1
MX		 1772 * The defragger has dealt with the R/O remount and umount,
1773 * needn't do anything special here.
d0278245
MX		 1774 */
1775 btrfs_run_defrag_inodes(root->fs_info);
1776sleep:
9d1a2a3a 1777 if (!try_to_freeze() && !again) {
a74a4b97 1778 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1779 if (!kthread_should_stop())
1780 schedule();
a74a4b97
CM		 1781 __set_current_state(TASK_RUNNING);
1782 }
1783 } while (!kthread_should_stop());
1784 return 0;
1785}
1786
1787static int transaction_kthread(void *arg)
1788{
1789 struct btrfs_root *root = arg;
1790 struct btrfs_trans_handle *trans;
1791 struct btrfs_transaction *cur;
8929ecfa 1792 u64 transid;
a74a4b97
CM		 1793 unsigned long now;
1794 unsigned long delay;
914b2007 1795 bool cannot_commit;
a74a4b97
CM		 1796
1797 do {
914b2007 1798 cannot_commit = false;
8b87dc17 1799 delay = HZ * root->fs_info->commit_interval;
a74a4b97
CM		 1800 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1801
a4abeea4 1802 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1803 cur = root->fs_info->running_transaction;
1804 if (!cur) {
a4abeea4 1805 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1806 goto sleep;
1807 }
31153d81 1808
a74a4b97 1809 now = get_seconds();
4a9d8bde 1810 if (cur->state < TRANS_STATE_BLOCKED &&
8b87dc17
DS		 1811 (now < cur->start_time ||
1812 now - cur->start_time < root->fs_info->commit_interval)) {
a4abeea4 1813 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1814 delay = HZ * 5;
1815 goto sleep;
1816 }
8929ecfa 1817 transid = cur->transid;
a4abeea4 1818 spin_unlock(&root->fs_info->trans_lock);
56bec294 1819
79787eaa 1820 /* If the file system is aborted, this will always fail. */
354aa0fb 1821 trans = btrfs_attach_transaction(root);
914b2007 1822 if (IS_ERR(trans)) {
354aa0fb
MX		 1823 if (PTR_ERR(trans) != -ENOENT)
1824 cannot_commit = true;
79787eaa 1825 goto sleep;
914b2007 1826 }
8929ecfa 1827 if (transid == trans->transid) {
79787eaa 1828 btrfs_commit_transaction(trans, root);
8929ecfa
YZ		 1829 } else {
1830 btrfs_end_transaction(trans, root);
1831 }
a74a4b97
CM		 1832sleep:
1833 wake_up_process(root->fs_info->cleaner_kthread);
1834 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1835
4e121c06
JB		 1836 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1837 &root->fs_info->fs_state)))
1838 btrfs_cleanup_transaction(root);
a0acae0e 1839 if (!try_to_freeze()) {
a74a4b97 1840 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa 1841 if (!kthread_should_stop() &&
914b2007
JK		 1842 (!btrfs_transaction_blocked(root->fs_info) ||
1843 cannot_commit))
8929ecfa 1844 schedule_timeout(delay);
a74a4b97
CM		 1845 __set_current_state(TASK_RUNNING);
1846 }
1847 } while (!kthread_should_stop());
1848 return 0;
1849}
1850
af31f5e5
CM		 1851/*
1852 * this will find the highest generation in the array of
1853 * root backups. The index of the highest array is returned,
1854 * or -1 if we can't find anything.
1855 *
1856 * We check to make sure the array is valid by comparing the
1857 * generation of the latest root in the array with the generation
1858 * in the super block. If they don't match we pitch it.
1859 */
1860static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1861{
1862 u64 cur;
1863 int newest_index = -1;
1864 struct btrfs_root_backup *root_backup;
1865 int i;
1866
1867 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1868 root_backup = info->super_copy->super_roots + i;
1869 cur = btrfs_backup_tree_root_gen(root_backup);
1870 if (cur == newest_gen)
1871 newest_index = i;
1872 }
1873
1874 /* check to see if we actually wrapped around */
1875 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1876 root_backup = info->super_copy->super_roots;
1877 cur = btrfs_backup_tree_root_gen(root_backup);
1878 if (cur == newest_gen)
1879 newest_index = 0;
1880 }
1881 return newest_index;
1882}
1883
1884
1885/*
1886 * find the oldest backup so we know where to store new entries
1887 * in the backup array. This will set the backup_root_index
1888 * field in the fs_info struct
1889 */
1890static void find_oldest_super_backup(struct btrfs_fs_info *info,
1891 u64 newest_gen)
1892{
1893 int newest_index = -1;
1894
1895 newest_index = find_newest_super_backup(info, newest_gen);
1896 /* if there was garbage in there, just move along */
1897 if (newest_index == -1) {
1898 info->backup_root_index = 0;
1899 } else {
1900 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1901 }
1902}
1903
1904/*
1905 * copy all the root pointers into the super backup array.
1906 * this will bump the backup pointer by one when it is
1907 * done
1908 */
1909static void backup_super_roots(struct btrfs_fs_info *info)
1910{
1911 int next_backup;
1912 struct btrfs_root_backup *root_backup;
1913 int last_backup;
1914
1915 next_backup = info->backup_root_index;
1916 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1917 BTRFS_NUM_BACKUP_ROOTS;
1918
1919 /*
1920 * just overwrite the last backup if we're at the same generation
1921 * this happens only at umount
1922 */
1923 root_backup = info->super_for_commit->super_roots + last_backup;
1924 if (btrfs_backup_tree_root_gen(root_backup) ==
1925 btrfs_header_generation(info->tree_root->node))
1926 next_backup = last_backup;
1927
1928 root_backup = info->super_for_commit->super_roots + next_backup;
1929
1930 /*
1931 * make sure all of our padding and empty slots get zero filled
1932 * regardless of which ones we use today
1933 */
1934 memset(root_backup, 0, sizeof(*root_backup));
1935
1936 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1937
1938 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1939 btrfs_set_backup_tree_root_gen(root_backup,
1940 btrfs_header_generation(info->tree_root->node));
1941
1942 btrfs_set_backup_tree_root_level(root_backup,
1943 btrfs_header_level(info->tree_root->node));
1944
1945 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1946 btrfs_set_backup_chunk_root_gen(root_backup,
1947 btrfs_header_generation(info->chunk_root->node));
1948 btrfs_set_backup_chunk_root_level(root_backup,
1949 btrfs_header_level(info->chunk_root->node));
1950
1951 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1952 btrfs_set_backup_extent_root_gen(root_backup,
1953 btrfs_header_generation(info->extent_root->node));
1954 btrfs_set_backup_extent_root_level(root_backup,
1955 btrfs_header_level(info->extent_root->node));
1956
7c7e82a7
CM		 1957 /*
1958 * we might commit during log recovery, which happens before we set
1959 * the fs_root. Make sure it is valid before we fill it in.
1960 */
1961 if (info->fs_root && info->fs_root->node) {
1962 btrfs_set_backup_fs_root(root_backup,
1963 info->fs_root->node->start);
1964 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 1965 btrfs_header_generation(info->fs_root->node));
7c7e82a7 1966 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 1967 btrfs_header_level(info->fs_root->node));
7c7e82a7 1968 }
af31f5e5
CM		 1969
1970 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1971 btrfs_set_backup_dev_root_gen(root_backup,
1972 btrfs_header_generation(info->dev_root->node));
1973 btrfs_set_backup_dev_root_level(root_backup,
1974 btrfs_header_level(info->dev_root->node));
1975
1976 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1977 btrfs_set_backup_csum_root_gen(root_backup,
1978 btrfs_header_generation(info->csum_root->node));
1979 btrfs_set_backup_csum_root_level(root_backup,
1980 btrfs_header_level(info->csum_root->node));
1981
1982 btrfs_set_backup_total_bytes(root_backup,
1983 btrfs_super_total_bytes(info->super_copy));
1984 btrfs_set_backup_bytes_used(root_backup,
1985 btrfs_super_bytes_used(info->super_copy));
1986 btrfs_set_backup_num_devices(root_backup,
1987 btrfs_super_num_devices(info->super_copy));
1988
1989 /*
1990 * if we don't copy this out to the super_copy, it won't get remembered
1991 * for the next commit
1992 */
1993 memcpy(&info->super_copy->super_roots,
1994 &info->super_for_commit->super_roots,
1995 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1996}
1997
1998/*
1999 * this copies info out of the root backup array and back into
2000 * the in-memory super block. It is meant to help iterate through
2001 * the array, so you send it the number of backups you've already
2002 * tried and the last backup index you used.
2003 *
2004 * this returns -1 when it has tried all the backups
2005 */
2006static noinline int next_root_backup(struct btrfs_fs_info *info,
2007 struct btrfs_super_block *super,
2008 int *num_backups_tried, int *backup_index)
2009{
2010 struct btrfs_root_backup *root_backup;
2011 int newest = *backup_index;
2012
2013 if (*num_backups_tried == 0) {
2014 u64 gen = btrfs_super_generation(super);
2015
2016 newest = find_newest_super_backup(info, gen);
2017 if (newest == -1)
2018 return -1;
2019
2020 *backup_index = newest;
2021 *num_backups_tried = 1;
2022 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2023 /* we've tried all the backups, all done */
2024 return -1;
2025 } else {
2026 /* jump to the next oldest backup */
2027 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2028 BTRFS_NUM_BACKUP_ROOTS;
2029 *backup_index = newest;
2030 *num_backups_tried += 1;
2031 }
2032 root_backup = super->super_roots + newest;
2033
2034 btrfs_set_super_generation(super,
2035 btrfs_backup_tree_root_gen(root_backup));
2036 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2037 btrfs_set_super_root_level(super,
2038 btrfs_backup_tree_root_level(root_backup));
2039 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2040
2041 /*
2042 * fixme: the total bytes and num_devices need to match or we should
2043 * need a fsck
2044 */
2045 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2046 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2047 return 0;
2048}
2049
7abadb64
LB		 2050/* helper to cleanup workers */
2051static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2052{
dc6e3209 2053 btrfs_destroy_workqueue(fs_info->fixup_workers);
afe3d242 2054 btrfs_destroy_workqueue(fs_info->delalloc_workers);
5cdc7ad3 2055 btrfs_destroy_workqueue(fs_info->workers);
fccb5d86
QW		 2056 btrfs_destroy_workqueue(fs_info->endio_workers);
2057 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2058 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
8b110e39 2059 btrfs_destroy_workqueue(fs_info->endio_repair_workers);
d05a33ac 2060 btrfs_destroy_workqueue(fs_info->rmw_workers);
fccb5d86
QW		 2061 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2062 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2063 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
a8c93d4e 2064 btrfs_destroy_workqueue(fs_info->submit_workers);
5b3bc44e 2065 btrfs_destroy_workqueue(fs_info->delayed_workers);
e66f0bb1 2066 btrfs_destroy_workqueue(fs_info->caching_workers);
736cfa15 2067 btrfs_destroy_workqueue(fs_info->readahead_workers);
a44903ab 2068 btrfs_destroy_workqueue(fs_info->flush_workers);
fc97fab0 2069 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
a79b7d4b 2070 btrfs_destroy_workqueue(fs_info->extent_workers);
7abadb64
LB		 2071}
2072
2e9f5954
R		 2073static void free_root_extent_buffers(struct btrfs_root *root)
2074{
2075 if (root) {
2076 free_extent_buffer(root->node);
2077 free_extent_buffer(root->commit_root);
2078 root->node = NULL;
2079 root->commit_root = NULL;
2080 }
2081}
2082
af31f5e5
CM		 2083/* helper to cleanup tree roots */
2084static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2085{
2e9f5954 2086 free_root_extent_buffers(info->tree_root);
655b09fe 2087
2e9f5954
R		 2088 free_root_extent_buffers(info->dev_root);
2089 free_root_extent_buffers(info->extent_root);
2090 free_root_extent_buffers(info->csum_root);
2091 free_root_extent_buffers(info->quota_root);
2092 free_root_extent_buffers(info->uuid_root);
2093 if (chunk_root)
2094 free_root_extent_buffers(info->chunk_root);
af31f5e5
CM		 2095}
2096
faa2dbf0 2097void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
171f6537
JB		 2098{
2099 int ret;
2100 struct btrfs_root *gang[8];
2101 int i;
2102
2103 while (!list_empty(&fs_info->dead_roots)) {
2104 gang[0] = list_entry(fs_info->dead_roots.next,
2105 struct btrfs_root, root_list);
2106 list_del(&gang[0]->root_list);
2107
27cdeb70 2108 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
cb517eab 2109 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
171f6537
JB		 2110 } else {
2111 free_extent_buffer(gang[0]->node);
2112 free_extent_buffer(gang[0]->commit_root);
b0feb9d9 2113 btrfs_put_fs_root(gang[0]);
171f6537
JB		 2114 }
2115 }
2116
2117 while (1) {
2118 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2119 (void **)gang, 0,
2120 ARRAY_SIZE(gang));
2121 if (!ret)
2122 break;
2123 for (i = 0; i < ret; i++)
cb517eab 2124 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
171f6537 2125 }
1a4319cc
LB		 2126
2127 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2128 btrfs_free_log_root_tree(NULL, fs_info);
2129 btrfs_destroy_pinned_extent(fs_info->tree_root,
2130 fs_info->pinned_extents);
2131 }
171f6537 2132}
af31f5e5 2133
ad2b2c80
AV		 2134int open_ctree(struct super_block *sb,
2135 struct btrfs_fs_devices *fs_devices,
2136 char *options)
2e635a27 2137{
db94535d
CM		 2138 u32 sectorsize;
2139 u32 nodesize;
87ee04eb 2140 u32 stripesize;
84234f3a 2141 u64 generation;
f2b636e8 2142 u64 features;
3de4586c 2143 struct btrfs_key location;
a061fc8d 2144 struct buffer_head *bh;
4d34b278 2145 struct btrfs_super_block *disk_super;
815745cf 2146 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2147 struct btrfs_root *tree_root;
4d34b278
ID		 2148 struct btrfs_root *extent_root;
2149 struct btrfs_root *csum_root;
2150 struct btrfs_root *chunk_root;
2151 struct btrfs_root *dev_root;
bcef60f2 2152 struct btrfs_root *quota_root;
f7a81ea4 2153 struct btrfs_root *uuid_root;
e02119d5 2154 struct btrfs_root *log_tree_root;
eb60ceac 2155 int ret;
e58ca020 2156 int err = -EINVAL;
af31f5e5
CM		 2157 int num_backups_tried = 0;
2158 int backup_index = 0;
5cdc7ad3
QW		 2159 int max_active;
2160 int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
70f80175
SB		 2161 bool create_uuid_tree;
2162 bool check_uuid_tree;
4543df7e 2163
f84a8bd6 2164 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
6f07e42e 2165 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
cb517eab 2166 if (!tree_root || !chunk_root) {
39279cc3
CM		 2167 err = -ENOMEM;
2168 goto fail;
2169 }
76dda93c
YZ		 2170
2171 ret = init_srcu_struct(&fs_info->subvol_srcu);
2172 if (ret) {
2173 err = ret;
2174 goto fail;
2175 }
2176
2177 ret = setup_bdi(fs_info, &fs_info->bdi);
2178 if (ret) {
2179 err = ret;
2180 goto fail_srcu;
2181 }
2182
e2d84521
MX		 2183 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2184 if (ret) {
2185 err = ret;
2186 goto fail_bdi;
2187 }
2188 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2189 (1 + ilog2(nr_cpu_ids));
2190
963d678b
MX		 2191 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2192 if (ret) {
2193 err = ret;
2194 goto fail_dirty_metadata_bytes;
2195 }
2196
c404e0dc
MX		 2197 ret = percpu_counter_init(&fs_info->bio_counter, 0);
2198 if (ret) {
2199 err = ret;
2200 goto fail_delalloc_bytes;
2201 }
2202
76dda93c
YZ		 2203 fs_info->btree_inode = new_inode(sb);
2204 if (!fs_info->btree_inode) {
2205 err = -ENOMEM;
c404e0dc 2206 goto fail_bio_counter;
76dda93c
YZ		 2207 }
2208
a6591715 2209 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2210
76dda93c 2211 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
f28491e0 2212 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
8fd17795 2213 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2214 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2215 INIT_LIST_HEAD(&fs_info->delayed_iputs);
eb73c1b7 2216 INIT_LIST_HEAD(&fs_info->delalloc_roots);
11833d66 2217 INIT_LIST_HEAD(&fs_info->caching_block_groups);
eb73c1b7 2218 spin_lock_init(&fs_info->delalloc_root_lock);
a4abeea4 2219 spin_lock_init(&fs_info->trans_lock);
76dda93c 2220 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2221 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2222 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 2223 spin_lock_init(&fs_info->free_chunk_lock);
f29021b2 2224 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2225 spin_lock_init(&fs_info->super_lock);
fcebe456 2226 spin_lock_init(&fs_info->qgroup_op_lock);
f28491e0 2227 spin_lock_init(&fs_info->buffer_lock);
47ab2a6c 2228 spin_lock_init(&fs_info->unused_bgs_lock);
f29021b2 2229 rwlock_init(&fs_info->tree_mod_log_lock);
7585717f 2230 mutex_init(&fs_info->reloc_mutex);
573bfb72 2231 mutex_init(&fs_info->delalloc_root_mutex);
de98ced9 2232 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2233
58176a96 2234 init_completion(&fs_info->kobj_unregister);
0b86a832 2235 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2236 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2237 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
47ab2a6c 2238 INIT_LIST_HEAD(&fs_info->unused_bgs);
0b86a832 2239 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX		 2240 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2241 BTRFS_BLOCK_RSV_GLOBAL);
2242 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2243 BTRFS_BLOCK_RSV_DELALLOC);
2244 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2245 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2246 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2247 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2248 BTRFS_BLOCK_RSV_DELOPS);
cb03c743 2249 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 2250 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 2251 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 2252 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 2253 atomic_set(&fs_info->defrag_running, 0);
fcebe456 2254 atomic_set(&fs_info->qgroup_op_seq, 0);
fc36ed7e 2255 atomic64_set(&fs_info->tree_mod_seq, 0);
e20d96d6 2256 fs_info->sb = sb;
6f568d35 2257 fs_info->max_inline = 8192 * 1024;
9ed74f2d 2258 fs_info->metadata_ratio = 0;
4cb5300b 2259 fs_info->defrag_inodes = RB_ROOT;
2bf64758 2260 fs_info->free_chunk_space = 0;
f29021b2 2261 fs_info->tree_mod_log = RB_ROOT;
8b87dc17 2262 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
0a2b2a84 2263 fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
90519d66
AJ		 2264 /* readahead state */
2265 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2266 spin_lock_init(&fs_info->reada_lock);
c8b97818 2267
b34b086c
CM		 2268 fs_info->thread_pool_size = min_t(unsigned long,
2269 num_online_cpus() + 2, 8);
0afbaf8c 2270
199c2a9c
MX		 2271 INIT_LIST_HEAD(&fs_info->ordered_roots);
2272 spin_lock_init(&fs_info->ordered_root_lock);
16cdcec7
MX		 2273 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2274 GFP_NOFS);
2275 if (!fs_info->delayed_root) {
2276 err = -ENOMEM;
2277 goto fail_iput;
2278 }
2279 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2280
a2de733c
AJ		 2281 mutex_init(&fs_info->scrub_lock);
2282 atomic_set(&fs_info->scrubs_running, 0);
2283 atomic_set(&fs_info->scrub_pause_req, 0);
2284 atomic_set(&fs_info->scrubs_paused, 0);
2285 atomic_set(&fs_info->scrub_cancel_req, 0);
c404e0dc 2286 init_waitqueue_head(&fs_info->replace_wait);
a2de733c 2287 init_waitqueue_head(&fs_info->scrub_pause_wait);
a2de733c 2288 fs_info->scrub_workers_refcnt = 0;
21adbd5c
SB		 2289#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2290 fs_info->check_integrity_print_mask = 0;
2291#endif
a2de733c 2292
c9e9f97b
ID		 2293 spin_lock_init(&fs_info->balance_lock);
2294 mutex_init(&fs_info->balance_mutex);
837d5b6e
ID		 2295 atomic_set(&fs_info->balance_running, 0);
2296 atomic_set(&fs_info->balance_pause_req, 0);
a7e99c69 2297 atomic_set(&fs_info->balance_cancel_req, 0);
c9e9f97b 2298 fs_info->balance_ctl = NULL;
837d5b6e 2299 init_waitqueue_head(&fs_info->balance_wait_q);
21c7e756 2300 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
a2de733c 2301
a061fc8d
CM		 2302 sb->s_blocksize = 4096;
2303 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 2304 sb->s_bdi = &fs_info->bdi;
a061fc8d 2305
76dda93c 2306 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
bfe86848 2307 set_nlink(fs_info->btree_inode, 1);
0afbaf8c
CM		 2308 /*
2309 * we set the i_size on the btree inode to the max possible int.
2310 * the real end of the address space is determined by all of
2311 * the devices in the system
2312 */
2313 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 2314 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 2315 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2316
5d4f98a2 2317 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 2318 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 2319 fs_info->btree_inode->i_mapping);
0b32f4bb 2320 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
a8067e02 2321 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM		 2322
2323 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 2324
76dda93c
YZ		 2325 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2326 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2327 sizeof(struct btrfs_key));
72ac3c0d
JB		 2328 set_bit(BTRFS_INODE_DUMMY,
2329 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
778ba82b 2330 btrfs_insert_inode_hash(fs_info->btree_inode);
76dda93c 2331
0f9dd46c 2332 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2333 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2334 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2335
11833d66 2336 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 2337 fs_info->btree_inode->i_mapping);
11833d66 2338 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 2339 fs_info->btree_inode->i_mapping);
11833d66 2340 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 2341 fs_info->do_barriers = 1;
e18e4809 2342
39279cc3 2343
5a3f23d5 2344 mutex_init(&fs_info->ordered_operations_mutex);
9ffba8cd 2345 mutex_init(&fs_info->ordered_extent_flush_mutex);
e02119d5 2346 mutex_init(&fs_info->tree_log_mutex);
925baedd 2347 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 2348 mutex_init(&fs_info->transaction_kthread_mutex);
2349 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2350 mutex_init(&fs_info->volume_mutex);
9e351cc8 2351 init_rwsem(&fs_info->commit_root_sem);
c71bf099 2352 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2353 init_rwsem(&fs_info->subvol_sem);
803b2f54 2354 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
e922e087
SB		 2355 fs_info->dev_replace.lock_owner = 0;
2356 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2357 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2358 mutex_init(&fs_info->dev_replace.lock_management_lock);
2359 mutex_init(&fs_info->dev_replace.lock);
fa9c0d79 2360
416ac51d 2361 spin_lock_init(&fs_info->qgroup_lock);
f2f6ed3d 2362 mutex_init(&fs_info->qgroup_ioctl_lock);
416ac51d 2363 fs_info->qgroup_tree = RB_ROOT;
fcebe456 2364 fs_info->qgroup_op_tree = RB_ROOT;
416ac51d
AJ		 2365 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2366 fs_info->qgroup_seq = 1;
2367 fs_info->quota_enabled = 0;
2368 fs_info->pending_quota_state = 0;
1e8f9158 2369 fs_info->qgroup_ulist = NULL;
2f232036 2370 mutex_init(&fs_info->qgroup_rescan_lock);
416ac51d 2371
fa9c0d79
CM		 2372 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2373 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2374
e6dcd2dc 2375 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2376 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2377 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2378 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2379
53b381b3
DW		 2380 ret = btrfs_alloc_stripe_hash_table(fs_info);
2381 if (ret) {
83c8266a 2382 err = ret;
53b381b3
DW		 2383 goto fail_alloc;
2384 }
2385
707e8a07 2386 __setup_root(4096, 4096, 4096, tree_root,
2c90e5d6 2387 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2388
3c4bb26b 2389 invalidate_bdev(fs_devices->latest_bdev);
1104a885
DS		 2390
2391 /*
2392 * Read super block and check the signature bytes only
2393 */
a512bbf8 2394 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY		 2395 if (!bh) {
2396 err = -EINVAL;
16cdcec7 2397 goto fail_alloc;
20b45077 2398 }
39279cc3 2399
1104a885
DS		 2400 /*
2401 * We want to check superblock checksum, the type is stored inside.
2402 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2403 */
2404 if (btrfs_check_super_csum(bh->b_data)) {
efe120a0 2405 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
1104a885
DS		 2406 err = -EINVAL;
2407 goto fail_alloc;
2408 }
2409
2410 /*
2411 * super_copy is zeroed at allocation time and we never touch the
2412 * following bytes up to INFO_SIZE, the checksum is calculated from
2413 * the whole block of INFO_SIZE
2414 */
6c41761f
DS		 2415 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2416 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2417 sizeof(*fs_info->super_for_commit));
a061fc8d 2418 brelse(bh);
5f39d397 2419
6c41761f 2420 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2421
1104a885
DS		 2422 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2423 if (ret) {
efe120a0 2424 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
1104a885
DS		 2425 err = -EINVAL;
2426 goto fail_alloc;
2427 }
2428
6c41761f 2429 disk_super = fs_info->super_copy;
0f7d52f4 2430 if (!btrfs_super_root(disk_super))
16cdcec7 2431 goto fail_alloc;
0f7d52f4 2432
acce952b 2433 /* check FS state, whether FS is broken. */
87533c47
MX		 2434 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2435 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2436
af31f5e5
CM		 2437 /*
2438 * run through our array of backup supers and setup
2439 * our ring pointer to the oldest one
2440 */
2441 generation = btrfs_super_generation(disk_super);
2442 find_oldest_super_backup(fs_info, generation);
2443
75e7cb7f
LB		 2444 /*
2445 * In the long term, we'll store the compression type in the super
2446 * block, and it'll be used for per file compression control.
2447 */
2448 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2449
2b82032c
YZ		 2450 ret = btrfs_parse_options(tree_root, options);
2451 if (ret) {
2452 err = ret;
16cdcec7 2453 goto fail_alloc;
2b82032c 2454 }
dfe25020 2455
f2b636e8
JB		 2456 features = btrfs_super_incompat_flags(disk_super) &
2457 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2458 if (features) {
2459 printk(KERN_ERR "BTRFS: couldn't mount because of "
2460 "unsupported optional features (%Lx).\n",
c1c9ff7c 2461 features);
f2b636e8 2462 err = -EINVAL;
16cdcec7 2463 goto fail_alloc;
f2b636e8
JB		 2464 }
2465
707e8a07
DS		 2466 /*
2467 * Leafsize and nodesize were always equal, this is only a sanity check.
2468 */
2469 if (le32_to_cpu(disk_super->__unused_leafsize) !=
727011e0
CM		 2470 btrfs_super_nodesize(disk_super)) {
2471 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2472 "blocksizes don't match. node %d leaf %d\n",
2473 btrfs_super_nodesize(disk_super),
707e8a07 2474 le32_to_cpu(disk_super->__unused_leafsize));
727011e0
CM		 2475 err = -EINVAL;
2476 goto fail_alloc;
2477 }
707e8a07 2478 if (btrfs_super_nodesize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
727011e0
CM		 2479 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2480 "blocksize (%d) was too large\n",
707e8a07 2481 btrfs_super_nodesize(disk_super));
727011e0
CM		 2482 err = -EINVAL;
2483 goto fail_alloc;
2484 }
2485
5d4f98a2 2486 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2487 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
69e380d1 2488 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2489 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
727011e0 2490
3173a18f 2491 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
efe120a0 2492 printk(KERN_ERR "BTRFS: has skinny extents\n");
3173a18f 2493
727011e0
CM		 2494 /*
2495 * flag our filesystem as having big metadata blocks if
2496 * they are bigger than the page size
2497 */
707e8a07 2498 if (btrfs_super_nodesize(disk_super) > PAGE_CACHE_SIZE) {
727011e0 2499 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
efe120a0 2500 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
727011e0
CM		 2501 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2502 }
2503
bc3f116f 2504 nodesize = btrfs_super_nodesize(disk_super);
bc3f116f
CM		 2505 sectorsize = btrfs_super_sectorsize(disk_super);
2506 stripesize = btrfs_super_stripesize(disk_super);
707e8a07 2507 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids));
963d678b 2508 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f
CM		 2509
2510 /*
2511 * mixed block groups end up with duplicate but slightly offset
2512 * extent buffers for the same range. It leads to corruptions
2513 */
2514 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
707e8a07 2515 (sectorsize != nodesize)) {
efe120a0 2516 printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes "
bc3f116f
CM		 2517 "are not allowed for mixed block groups on %s\n",
2518 sb->s_id);
2519 goto fail_alloc;
2520 }
2521
ceda0864
MX		 2522 /*
2523 * Needn't use the lock because there is no other task which will
2524 * update the flag.
2525 */
a6fa6fae 2526 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2527
f2b636e8
JB		 2528 features = btrfs_super_compat_ro_flags(disk_super) &
2529 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2530 if (!(sb->s_flags & MS_RDONLY) && features) {
2531 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2532 "unsupported option features (%Lx).\n",
c1c9ff7c 2533 features);
f2b636e8 2534 err = -EINVAL;
16cdcec7 2535 goto fail_alloc;
f2b636e8 2536 }
61d92c32 2537
5cdc7ad3 2538 max_active = fs_info->thread_pool_size;
61d92c32 2539
5cdc7ad3
QW		 2540 fs_info->workers =
2541 btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2542 max_active, 16);
c8b97818 2543
afe3d242
QW		 2544 fs_info->delalloc_workers =
2545 btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
771ed689 2546
a44903ab
QW		 2547 fs_info->flush_workers =
2548 btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
8ccf6f19 2549
e66f0bb1
QW		 2550 fs_info->caching_workers =
2551 btrfs_alloc_workqueue("cache", flags, max_active, 0);
61b49440 2552
a8c93d4e
QW		 2553 /*
2554 * a higher idle thresh on the submit workers makes it much more
61b49440
CM		 2555 * likely that bios will be send down in a sane order to the
2556 * devices
2557 */
a8c93d4e
QW		 2558 fs_info->submit_workers =
2559 btrfs_alloc_workqueue("submit", flags,
2560 min_t(u64, fs_devices->num_devices,
2561 max_active), 64);
53863232 2562
dc6e3209
QW		 2563 fs_info->fixup_workers =
2564 btrfs_alloc_workqueue("fixup", flags, 1, 0);
61b49440
CM		 2565
2566 /*
2567 * endios are largely parallel and should have a very
2568 * low idle thresh
2569 */
fccb5d86
QW		 2570 fs_info->endio_workers =
2571 btrfs_alloc_workqueue("endio", flags, max_active, 4);
2572 fs_info->endio_meta_workers =
2573 btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2574 fs_info->endio_meta_write_workers =
2575 btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2576 fs_info->endio_raid56_workers =
2577 btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
8b110e39
MX		 2578 fs_info->endio_repair_workers =
2579 btrfs_alloc_workqueue("endio-repair", flags, 1, 0);
d05a33ac
QW		 2580 fs_info->rmw_workers =
2581 btrfs_alloc_workqueue("rmw", flags, max_active, 2);
fccb5d86
QW		 2582 fs_info->endio_write_workers =
2583 btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2584 fs_info->endio_freespace_worker =
2585 btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
5b3bc44e
QW		 2586 fs_info->delayed_workers =
2587 btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
736cfa15
QW		 2588 fs_info->readahead_workers =
2589 btrfs_alloc_workqueue("readahead", flags, max_active, 2);
fc97fab0
QW		 2590 fs_info->qgroup_rescan_workers =
2591 btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
a79b7d4b
CM		 2592 fs_info->extent_workers =
2593 btrfs_alloc_workqueue("extent-refs", flags,
2594 min_t(u64, fs_devices->num_devices,
2595 max_active), 8);
61b49440 2596
a8c93d4e 2597 if (!(fs_info->workers && fs_info->delalloc_workers &&
fccb5d86
QW		 2598 fs_info->submit_workers && fs_info->flush_workers &&
2599 fs_info->endio_workers && fs_info->endio_meta_workers &&
2600 fs_info->endio_meta_write_workers &&
8b110e39 2601 fs_info->endio_repair_workers &&
fccb5d86 2602 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
e66f0bb1 2603 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
dc6e3209 2604 fs_info->caching_workers && fs_info->readahead_workers &&
fc97fab0 2605 fs_info->fixup_workers && fs_info->delayed_workers &&
56094eec 2606 fs_info->extent_workers &&
fc97fab0 2607 fs_info->qgroup_rescan_workers)) {
fed425c7 2608 err = -ENOMEM;
0dc3b84a
JB		 2609 goto fail_sb_buffer;
2610 }
4543df7e 2611
4575c9cc 2612 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 2613 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2614 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 2615
db94535d 2616 tree_root->nodesize = nodesize;
db94535d 2617 tree_root->sectorsize = sectorsize;
87ee04eb 2618 tree_root->stripesize = stripesize;
a061fc8d
CM		 2619
2620 sb->s_blocksize = sectorsize;
2621 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 2622
3cae210f 2623 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
efe120a0 2624 printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 2625 goto fail_sb_buffer;
2626 }
19c00ddc 2627
8d082fb7 2628 if (sectorsize != PAGE_SIZE) {
efe120a0 2629 printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) "
8d082fb7 2630 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
941b2ddf
KM		 2631 goto fail_sb_buffer;
2632 }
2633
925baedd 2634 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 2635 ret = btrfs_read_sys_array(tree_root);
925baedd 2636 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2637 if (ret) {
efe120a0 2638 printk(KERN_WARNING "BTRFS: failed to read the system "
d397712b 2639 "array on %s\n", sb->s_id);
5d4f98a2 2640 goto fail_sb_buffer;
84eed90f 2641 }
0b86a832 2642
84234f3a 2643 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832 2644
707e8a07
DS		 2645 __setup_root(nodesize, sectorsize, stripesize, chunk_root,
2646 fs_info, BTRFS_CHUNK_TREE_OBJECTID);
0b86a832
CM		 2647
2648 chunk_root->node = read_tree_block(chunk_root,
2649 btrfs_super_chunk_root(disk_super),
ce86cd59 2650 generation);
416bc658
JB		 2651 if (!chunk_root->node ||
2652 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
efe120a0 2653 printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n",
83121942 2654 sb->s_id);
af31f5e5 2655 goto fail_tree_roots;
83121942 2656 }
5d4f98a2
YZ		 2657 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2658 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2659
e17cade2 2660 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
b308bc2f 2661 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
e17cade2 2662
0b86a832 2663 ret = btrfs_read_chunk_tree(chunk_root);
2b82032c 2664 if (ret) {
efe120a0 2665 printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n",
d397712b 2666 sb->s_id);
af31f5e5 2667 goto fail_tree_roots;
2b82032c 2668 }
0b86a832 2669
8dabb742
SB		 2670 /*
2671 * keep the device that is marked to be the target device for the
2672 * dev_replace procedure
2673 */
2674 btrfs_close_extra_devices(fs_info, fs_devices, 0);
dfe25020 2675
a6b0d5c8 2676 if (!fs_devices->latest_bdev) {
efe120a0 2677 printk(KERN_CRIT "BTRFS: failed to read devices on %s\n",
a6b0d5c8
CM		 2678 sb->s_id);
2679 goto fail_tree_roots;
2680 }
2681
af31f5e5 2682retry_root_backup:
84234f3a 2683 generation = btrfs_super_generation(disk_super);
0b86a832 2684
e20d96d6 2685 tree_root->node = read_tree_block(tree_root,
db94535d 2686 btrfs_super_root(disk_super),
ce86cd59 2687 generation);
af31f5e5
CM		 2688 if (!tree_root->node ||
2689 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
efe120a0 2690 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
83121942 2691 sb->s_id);
af31f5e5
CM		 2692
2693 goto recovery_tree_root;
83121942 2694 }
af31f5e5 2695
5d4f98a2
YZ		 2696 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2697 tree_root->commit_root = btrfs_root_node(tree_root);
69e9c6c6 2698 btrfs_set_root_refs(&tree_root->root_item, 1);
db94535d 2699
cb517eab
MX		 2700 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2701 location.type = BTRFS_ROOT_ITEM_KEY;
2702 location.offset = 0;
2703
2704 extent_root = btrfs_read_tree_root(tree_root, &location);
2705 if (IS_ERR(extent_root)) {
2706 ret = PTR_ERR(extent_root);
af31f5e5 2707 goto recovery_tree_root;
cb517eab 2708 }
27cdeb70 2709 set_bit(BTRFS_ROOT_TRACK_DIRTY, &extent_root->state);
cb517eab 2710 fs_info->extent_root = extent_root;
0b86a832 2711
cb517eab
MX		 2712 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2713 dev_root = btrfs_read_tree_root(tree_root, &location);
2714 if (IS_ERR(dev_root)) {
2715 ret = PTR_ERR(dev_root);
af31f5e5 2716 goto recovery_tree_root;
cb517eab 2717 }
27cdeb70 2718 set_bit(BTRFS_ROOT_TRACK_DIRTY, &dev_root->state);
cb517eab
MX		 2719 fs_info->dev_root = dev_root;
2720 btrfs_init_devices_late(fs_info);
3768f368 2721
cb517eab
MX		 2722 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2723 csum_root = btrfs_read_tree_root(tree_root, &location);
2724 if (IS_ERR(csum_root)) {
2725 ret = PTR_ERR(csum_root);
af31f5e5 2726 goto recovery_tree_root;
cb517eab 2727 }
27cdeb70 2728 set_bit(BTRFS_ROOT_TRACK_DIRTY, &csum_root->state);
cb517eab 2729 fs_info->csum_root = csum_root;
d20f7043 2730
cb517eab
MX		 2731 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2732 quota_root = btrfs_read_tree_root(tree_root, &location);
2733 if (!IS_ERR(quota_root)) {
27cdeb70 2734 set_bit(BTRFS_ROOT_TRACK_DIRTY, &quota_root->state);
bcef60f2
AJ		 2735 fs_info->quota_enabled = 1;
2736 fs_info->pending_quota_state = 1;
cb517eab 2737 fs_info->quota_root = quota_root;
bcef60f2
AJ		 2738 }
2739
f7a81ea4
SB		 2740 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2741 uuid_root = btrfs_read_tree_root(tree_root, &location);
2742 if (IS_ERR(uuid_root)) {
2743 ret = PTR_ERR(uuid_root);
2744 if (ret != -ENOENT)
2745 goto recovery_tree_root;
2746 create_uuid_tree = true;
70f80175 2747 check_uuid_tree = false;
f7a81ea4 2748 } else {
27cdeb70 2749 set_bit(BTRFS_ROOT_TRACK_DIRTY, &uuid_root->state);
f7a81ea4 2750 fs_info->uuid_root = uuid_root;
70f80175
SB		 2751 create_uuid_tree = false;
2752 check_uuid_tree =
2753 generation != btrfs_super_uuid_tree_generation(disk_super);
f7a81ea4
SB		 2754 }
2755
8929ecfa
YZ		 2756 fs_info->generation = generation;
2757 fs_info->last_trans_committed = generation;
8929ecfa 2758
68310a5e
ID		 2759 ret = btrfs_recover_balance(fs_info);
2760 if (ret) {
efe120a0 2761 printk(KERN_WARNING "BTRFS: failed to recover balance\n");
68310a5e
ID		 2762 goto fail_block_groups;
2763 }
2764
733f4fbb
SB		 2765 ret = btrfs_init_dev_stats(fs_info);
2766 if (ret) {
efe120a0 2767 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
733f4fbb
SB		 2768 ret);
2769 goto fail_block_groups;
2770 }
2771
8dabb742
SB		 2772 ret = btrfs_init_dev_replace(fs_info);
2773 if (ret) {
efe120a0 2774 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
8dabb742
SB		 2775 goto fail_block_groups;
2776 }
2777
2778 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2779
5ac1d209 2780 ret = btrfs_sysfs_add_one(fs_info);
c59021f8 2781 if (ret) {
efe120a0 2782 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
c59021f8 2783 goto fail_block_groups;
2784 }
2785
c59021f8 2786 ret = btrfs_init_space_info(fs_info);
2787 if (ret) {
efe120a0 2788 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
2365dd3c 2789 goto fail_sysfs;
c59021f8 2790 }
2791
1b1d1f66
JB		 2792 ret = btrfs_read_block_groups(extent_root);
2793 if (ret) {
efe120a0 2794 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
2365dd3c 2795 goto fail_sysfs;
1b1d1f66 2796 }
5af3e8cc
SB		 2797 fs_info->num_tolerated_disk_barrier_failures =
2798 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
292fd7fc
SB		 2799 if (fs_info->fs_devices->missing_devices >
2800 fs_info->num_tolerated_disk_barrier_failures &&
2801 !(sb->s_flags & MS_RDONLY)) {
efe120a0
FH		 2802 printk(KERN_WARNING "BTRFS: "
2803 "too many missing devices, writeable mount is not allowed\n");
2365dd3c 2804 goto fail_sysfs;
292fd7fc 2805 }
9078a3e1 2806
a74a4b97
CM		 2807 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2808 "btrfs-cleaner");
57506d50 2809 if (IS_ERR(fs_info->cleaner_kthread))
2365dd3c 2810 goto fail_sysfs;
a74a4b97
CM		 2811
2812 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2813 tree_root,
2814 "btrfs-transaction");
57506d50 2815 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2816 goto fail_cleaner;
a74a4b97 2817
c289811c
CM		 2818 if (!btrfs_test_opt(tree_root, SSD) &&
2819 !btrfs_test_opt(tree_root, NOSSD) &&
2820 !fs_info->fs_devices->rotating) {
efe120a0 2821 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
c289811c
CM		 2822 "mode\n");
2823 btrfs_set_opt(fs_info->mount_opt, SSD);
2824 }
2825
3818aea2
QW		 2826 /* Set the real inode map cache flag */
2827 if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE))
2828 btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE);
2829
21adbd5c
SB		 2830#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2831 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2832 ret = btrfsic_mount(tree_root, fs_devices,
2833 btrfs_test_opt(tree_root,
2834 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2835 1 : 0,
2836 fs_info->check_integrity_print_mask);
2837 if (ret)
efe120a0 2838 printk(KERN_WARNING "BTRFS: failed to initialize"
21adbd5c
SB		 2839 " integrity check module %s\n", sb->s_id);
2840 }
2841#endif
bcef60f2
AJ		 2842 ret = btrfs_read_qgroup_config(fs_info);
2843 if (ret)
2844 goto fail_trans_kthread;
21adbd5c 2845
acce952b 2846 /* do not make disk changes in broken FS */
68ce9682 2847 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM		 2848 u64 bytenr = btrfs_super_log_root(disk_super);
2849
7c2ca468 2850 if (fs_devices->rw_devices == 0) {
efe120a0 2851 printk(KERN_WARNING "BTRFS: log replay required "
d397712b 2852 "on RO media\n");
7c2ca468 2853 err = -EIO;
bcef60f2 2854 goto fail_qgroup;
7c2ca468 2855 }
d18a2c44 2856
6f07e42e 2857 log_tree_root = btrfs_alloc_root(fs_info);
676e4c86
DC		 2858 if (!log_tree_root) {
2859 err = -ENOMEM;
bcef60f2 2860 goto fail_qgroup;
676e4c86 2861 }
e02119d5 2862
707e8a07 2863 __setup_root(nodesize, sectorsize, stripesize,
e02119d5
CM		 2864 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2865
2866 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a 2867 generation + 1);
416bc658
JB		 2868 if (!log_tree_root->node ||
2869 !extent_buffer_uptodate(log_tree_root->node)) {
efe120a0 2870 printk(KERN_ERR "BTRFS: failed to read log tree\n");
416bc658
JB		 2871 free_extent_buffer(log_tree_root->node);
2872 kfree(log_tree_root);
28c16cbb 2873 goto fail_qgroup;
416bc658 2874 }
79787eaa 2875 /* returns with log_tree_root freed on success */
e02119d5 2876 ret = btrfs_recover_log_trees(log_tree_root);
79787eaa
JM		 2877 if (ret) {
2878 btrfs_error(tree_root->fs_info, ret,
2879 "Failed to recover log tree");
2880 free_extent_buffer(log_tree_root->node);
2881 kfree(log_tree_root);
28c16cbb 2882 goto fail_qgroup;
79787eaa 2883 }
e556ce2c
YZ		 2884
2885 if (sb->s_flags & MS_RDONLY) {
79787eaa
JM		 2886 ret = btrfs_commit_super(tree_root);
2887 if (ret)
28c16cbb 2888 goto fail_qgroup;
e556ce2c 2889 }
e02119d5 2890 }
1a40e23b 2891
76dda93c 2892 ret = btrfs_find_orphan_roots(tree_root);
79787eaa 2893 if (ret)
28c16cbb 2894 goto fail_qgroup;
76dda93c 2895
7c2ca468 2896 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b 2897 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2 2898 if (ret)
28c16cbb 2899 goto fail_qgroup;
d68fc57b 2900
5f316481 2901 mutex_lock(&fs_info->cleaner_mutex);
5d4f98a2 2902 ret = btrfs_recover_relocation(tree_root);
5f316481 2903 mutex_unlock(&fs_info->cleaner_mutex);
d7ce5843
MX		 2904 if (ret < 0) {
2905 printk(KERN_WARNING
efe120a0 2906 "BTRFS: failed to recover relocation\n");
d7ce5843 2907 err = -EINVAL;
bcef60f2 2908 goto fail_qgroup;
d7ce5843 2909 }
7c2ca468 2910 }
1a40e23b 2911
3de4586c
CM		 2912 location.objectid = BTRFS_FS_TREE_OBJECTID;
2913 location.type = BTRFS_ROOT_ITEM_KEY;
cb517eab 2914 location.offset = 0;
3de4586c 2915
3de4586c 2916 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
3140c9a3
DC		 2917 if (IS_ERR(fs_info->fs_root)) {
2918 err = PTR_ERR(fs_info->fs_root);
bcef60f2 2919 goto fail_qgroup;
3140c9a3 2920 }
c289811c 2921
2b6ba629
ID		 2922 if (sb->s_flags & MS_RDONLY)
2923 return 0;
59641015 2924
2b6ba629
ID		 2925 down_read(&fs_info->cleanup_work_sem);
2926 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2927 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 2928 up_read(&fs_info->cleanup_work_sem);
2b6ba629
ID		 2929 close_ctree(tree_root);
2930 return ret;
2931 }
2932 up_read(&fs_info->cleanup_work_sem);
59641015 2933
2b6ba629
ID		 2934 ret = btrfs_resume_balance_async(fs_info);
2935 if (ret) {
efe120a0 2936 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
2b6ba629
ID		 2937 close_ctree(tree_root);
2938 return ret;
e3acc2a6
JB		 2939 }
2940
8dabb742
SB		 2941 ret = btrfs_resume_dev_replace_async(fs_info);
2942 if (ret) {
efe120a0 2943 pr_warn("BTRFS: failed to resume dev_replace\n");
8dabb742
SB		 2944 close_ctree(tree_root);
2945 return ret;
2946 }
2947
b382a324
JS		 2948 btrfs_qgroup_rescan_resume(fs_info);
2949
f7a81ea4 2950 if (create_uuid_tree) {
efe120a0 2951 pr_info("BTRFS: creating UUID tree\n");
f7a81ea4
SB		 2952 ret = btrfs_create_uuid_tree(fs_info);
2953 if (ret) {
efe120a0 2954 pr_warn("BTRFS: failed to create the UUID tree %d\n",
f7a81ea4
SB		 2955 ret);
2956 close_ctree(tree_root);
2957 return ret;
2958 }
f420ee1e
SB		 2959 } else if (check_uuid_tree ||
2960 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
efe120a0 2961 pr_info("BTRFS: checking UUID tree\n");
70f80175
SB		 2962 ret = btrfs_check_uuid_tree(fs_info);
2963 if (ret) {
efe120a0 2964 pr_warn("BTRFS: failed to check the UUID tree %d\n",
70f80175
SB		 2965 ret);
2966 close_ctree(tree_root);
2967 return ret;
2968 }
2969 } else {
2970 fs_info->update_uuid_tree_gen = 1;
f7a81ea4
SB		 2971 }
2972
47ab2a6c
JB		 2973 fs_info->open = 1;
2974
ad2b2c80 2975 return 0;
39279cc3 2976
bcef60f2
AJ		 2977fail_qgroup:
2978 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM		 2979fail_trans_kthread:
2980 kthread_stop(fs_info->transaction_kthread);
54067ae9 2981 btrfs_cleanup_transaction(fs_info->tree_root);
faa2dbf0 2982 btrfs_free_fs_roots(fs_info);
3f157a2f 2983fail_cleaner:
a74a4b97 2984 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2985
2986 /*
2987 * make sure we're done with the btree inode before we stop our
2988 * kthreads
2989 */
2990 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 2991
2365dd3c
AJ		 2992fail_sysfs:
2993 btrfs_sysfs_remove_one(fs_info);
2994
1b1d1f66 2995fail_block_groups:
54067ae9 2996 btrfs_put_block_group_cache(fs_info);
1b1d1f66 2997 btrfs_free_block_groups(fs_info);
af31f5e5
CM		 2998
2999fail_tree_roots:
3000 free_root_pointers(fs_info, 1);
2b8195bb 3001 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 3002
39279cc3 3003fail_sb_buffer:
7abadb64 3004 btrfs_stop_all_workers(fs_info);
16cdcec7 3005fail_alloc:
4543df7e 3006fail_iput:
586e46e2
ID		 3007 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3008
4543df7e 3009 iput(fs_info->btree_inode);
c404e0dc
MX		 3010fail_bio_counter:
3011 percpu_counter_destroy(&fs_info->bio_counter);
963d678b
MX		 3012fail_delalloc_bytes:
3013 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX		 3014fail_dirty_metadata_bytes:
3015 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
ad081f14 3016fail_bdi:
7e662854 3017 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 3018fail_srcu:
3019 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 3020fail:
53b381b3 3021 btrfs_free_stripe_hash_table(fs_info);
586e46e2 3022 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 3023 return err;
af31f5e5
CM		 3024
3025recovery_tree_root:
af31f5e5
CM		 3026 if (!btrfs_test_opt(tree_root, RECOVERY))
3027 goto fail_tree_roots;
3028
3029 free_root_pointers(fs_info, 0);
3030
3031 /* don't use the log in recovery mode, it won't be valid */
3032 btrfs_set_super_log_root(disk_super, 0);
3033
3034 /* we can't trust the free space cache either */
3035 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3036
3037 ret = next_root_backup(fs_info, fs_info->super_copy,
3038 &num_backups_tried, &backup_index);
3039 if (ret == -1)
3040 goto fail_block_groups;
3041 goto retry_root_backup;
eb60ceac
CM		 3042}
3043
f2984462
CM		 3044static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3045{
f2984462
CM		 3046 if (uptodate) {
3047 set_buffer_uptodate(bh);
3048 } else {
442a4f63
SB		 3049 struct btrfs_device *device = (struct btrfs_device *)
3050 bh->b_private;
3051
efe120a0 3052 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
606686ee
JB		 3053 "I/O error on %s\n",
3054 rcu_str_deref(device->name));
1259ab75
CM		 3055 /* note, we dont' set_buffer_write_io_error because we have
3056 * our own ways of dealing with the IO errors
3057 */
f2984462 3058 clear_buffer_uptodate(bh);
442a4f63 3059 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM		 3060 }
3061 unlock_buffer(bh);
3062 put_bh(bh);
3063}
3064
a512bbf8
YZ		 3065struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3066{
3067 struct buffer_head *bh;
3068 struct buffer_head *latest = NULL;
3069 struct btrfs_super_block *super;
3070 int i;
3071 u64 transid = 0;
3072 u64 bytenr;
3073
3074 /* we would like to check all the supers, but that would make
3075 * a btrfs mount succeed after a mkfs from a different FS.
3076 * So, we need to add a special mount option to scan for
3077 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3078 */
3079 for (i = 0; i < 1; i++) {
3080 bytenr = btrfs_sb_offset(i);
8068a47e
AJ		 3081 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3082 i_size_read(bdev->bd_inode))
a512bbf8 3083 break;
8068a47e
AJ		 3084 bh = __bread(bdev, bytenr / 4096,
3085 BTRFS_SUPER_INFO_SIZE);
a512bbf8
YZ		 3086 if (!bh)
3087 continue;
3088
3089 super = (struct btrfs_super_block *)bh->b_data;
3090 if (btrfs_super_bytenr(super) != bytenr ||
3cae210f 3091 btrfs_super_magic(super) != BTRFS_MAGIC) {
a512bbf8
YZ		 3092 brelse(bh);
3093 continue;
3094 }
3095
3096 if (!latest || btrfs_super_generation(super) > transid) {
3097 brelse(latest);
3098 latest = bh;
3099 transid = btrfs_super_generation(super);
3100 } else {
3101 brelse(bh);
3102 }
3103 }
3104 return latest;
3105}
3106
4eedeb75
HH		 3107/*
3108 * this should be called twice, once with wait == 0 and
3109 * once with wait == 1. When wait == 0 is done, all the buffer heads
3110 * we write are pinned.
3111 *
3112 * They are released when wait == 1 is done.
3113 * max_mirrors must be the same for both runs, and it indicates how
3114 * many supers on this one device should be written.
3115 *
3116 * max_mirrors == 0 means to write them all.
3117 */
a512bbf8
YZ		 3118static int write_dev_supers(struct btrfs_device *device,
3119 struct btrfs_super_block *sb,
3120 int do_barriers, int wait, int max_mirrors)
3121{
3122 struct buffer_head *bh;
3123 int i;
3124 int ret;
3125 int errors = 0;
3126 u32 crc;
3127 u64 bytenr;
a512bbf8
YZ		 3128
3129 if (max_mirrors == 0)
3130 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3131
a512bbf8
YZ		 3132 for (i = 0; i < max_mirrors; i++) {
3133 bytenr = btrfs_sb_offset(i);
935e5cc9
MX		 3134 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3135 device->commit_total_bytes)
a512bbf8
YZ		 3136 break;
3137
3138 if (wait) {
3139 bh = __find_get_block(device->bdev, bytenr / 4096,
3140 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3141 if (!bh) {
3142 errors++;
3143 continue;
3144 }
a512bbf8 3145 wait_on_buffer(bh);
4eedeb75
HH		 3146 if (!buffer_uptodate(bh))
3147 errors++;
3148
3149 /* drop our reference */
3150 brelse(bh);
3151
3152 /* drop the reference from the wait == 0 run */
3153 brelse(bh);
3154 continue;
a512bbf8
YZ		 3155 } else {
3156 btrfs_set_super_bytenr(sb, bytenr);
3157
3158 crc = ~(u32)0;
b0496686 3159 crc = btrfs_csum_data((char *)sb +
a512bbf8
YZ		 3160 BTRFS_CSUM_SIZE, crc,
3161 BTRFS_SUPER_INFO_SIZE -
3162 BTRFS_CSUM_SIZE);
3163 btrfs_csum_final(crc, sb->csum);
3164
4eedeb75
HH		 3165 /*
3166 * one reference for us, and we leave it for the
3167 * caller
3168 */
a512bbf8
YZ		 3169 bh = __getblk(device->bdev, bytenr / 4096,
3170 BTRFS_SUPER_INFO_SIZE);
634554dc 3171 if (!bh) {
efe120a0 3172 printk(KERN_ERR "BTRFS: couldn't get super "
634554dc
JB		 3173 "buffer head for bytenr %Lu\n", bytenr);
3174 errors++;
3175 continue;
3176 }
3177
a512bbf8
YZ		 3178 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3179
4eedeb75 3180 /* one reference for submit_bh */
a512bbf8 3181 get_bh(bh);
4eedeb75
HH		 3182
3183 set_buffer_uptodate(bh);
a512bbf8
YZ		 3184 lock_buffer(bh);
3185 bh->b_end_io = btrfs_end_buffer_write_sync;
442a4f63 3186 bh->b_private = device;
a512bbf8
YZ		 3187 }
3188
387125fc
CM		 3189 /*
3190 * we fua the first super. The others we allow
3191 * to go down lazy.
3192 */
e8117c26
WS		 3193 if (i == 0)
3194 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3195 else
3196 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
4eedeb75 3197 if (ret)
a512bbf8 3198 errors++;
a512bbf8
YZ		 3199 }
3200 return errors < i ? 0 : -1;
3201}
3202
387125fc
CM		 3203/*
3204 * endio for the write_dev_flush, this will wake anyone waiting
3205 * for the barrier when it is done
3206 */
3207static void btrfs_end_empty_barrier(struct bio *bio, int err)
3208{
3209 if (err) {
3210 if (err == -EOPNOTSUPP)
3211 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3212 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3213 }
3214 if (bio->bi_private)
3215 complete(bio->bi_private);
3216 bio_put(bio);
3217}
3218
3219/*
3220 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3221 * sent down. With wait == 1, it waits for the previous flush.
3222 *
3223 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3224 * capable
3225 */
3226static int write_dev_flush(struct btrfs_device *device, int wait)
3227{
3228 struct bio *bio;
3229 int ret = 0;
3230
3231 if (device->nobarriers)
3232 return 0;
3233
3234 if (wait) {
3235 bio = device->flush_bio;
3236 if (!bio)
3237 return 0;
3238
3239 wait_for_completion(&device->flush_wait);
3240
3241 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
efe120a0 3242 printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
606686ee 3243 rcu_str_deref(device->name));
387125fc 3244 device->nobarriers = 1;
5af3e8cc 3245 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
387125fc 3246 ret = -EIO;
5af3e8cc
SB		 3247 btrfs_dev_stat_inc_and_print(device,
3248 BTRFS_DEV_STAT_FLUSH_ERRS);
387125fc
CM		 3249 }
3250
3251 /* drop the reference from the wait == 0 run */
3252 bio_put(bio);
3253 device->flush_bio = NULL;
3254
3255 return ret;
3256 }
3257
3258 /*
3259 * one reference for us, and we leave it for the
3260 * caller
3261 */
9c017abc 3262 device->flush_bio = NULL;
9be3395b 3263 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
387125fc
CM		 3264 if (!bio)
3265 return -ENOMEM;
3266
3267 bio->bi_end_io = btrfs_end_empty_barrier;
3268 bio->bi_bdev = device->bdev;
3269 init_completion(&device->flush_wait);
3270 bio->bi_private = &device->flush_wait;
3271 device->flush_bio = bio;
3272
3273 bio_get(bio);
21adbd5c 3274 btrfsic_submit_bio(WRITE_FLUSH, bio);
387125fc
CM		 3275
3276 return 0;
3277}
3278
3279/*
3280 * send an empty flush down to each device in parallel,
3281 * then wait for them
3282 */
3283static int barrier_all_devices(struct btrfs_fs_info *info)
3284{
3285 struct list_head *head;
3286 struct btrfs_device *dev;
5af3e8cc
SB		 3287 int errors_send = 0;
3288 int errors_wait = 0;
387125fc
CM		 3289 int ret;
3290
3291 /* send down all the barriers */
3292 head = &info->fs_devices->devices;
3293 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3294 if (dev->missing)
3295 continue;
387125fc 3296 if (!dev->bdev) {
5af3e8cc 3297 errors_send++;
387125fc
CM		 3298 continue;
3299 }
3300 if (!dev->in_fs_metadata || !dev->writeable)
3301 continue;
3302
3303 ret = write_dev_flush(dev, 0);
3304 if (ret)
5af3e8cc 3305 errors_send++;
387125fc
CM		 3306 }
3307
3308 /* wait for all the barriers */
3309 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3310 if (dev->missing)
3311 continue;
387125fc 3312 if (!dev->bdev) {
5af3e8cc 3313 errors_wait++;
387125fc
CM		 3314 continue;
3315 }
3316 if (!dev->in_fs_metadata || !dev->writeable)
3317 continue;
3318
3319 ret = write_dev_flush(dev, 1);
3320 if (ret)
5af3e8cc 3321 errors_wait++;
387125fc 3322 }
5af3e8cc
SB		 3323 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3324 errors_wait > info->num_tolerated_disk_barrier_failures)
387125fc
CM		 3325 return -EIO;
3326 return 0;
3327}
3328
5af3e8cc
SB		 3329int btrfs_calc_num_tolerated_disk_barrier_failures(
3330 struct btrfs_fs_info *fs_info)
3331{
3332 struct btrfs_ioctl_space_info space;
3333 struct btrfs_space_info *sinfo;
3334 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3335 BTRFS_BLOCK_GROUP_SYSTEM,
3336 BTRFS_BLOCK_GROUP_METADATA,
3337 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3338 int num_types = 4;
3339 int i;
3340 int c;
3341 int num_tolerated_disk_barrier_failures =
3342 (int)fs_info->fs_devices->num_devices;
3343
3344 for (i = 0; i < num_types; i++) {
3345 struct btrfs_space_info *tmp;
3346
3347 sinfo = NULL;
3348 rcu_read_lock();
3349 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3350 if (tmp->flags == types[i]) {
3351 sinfo = tmp;
3352 break;
3353 }
3354 }
3355 rcu_read_unlock();
3356
3357 if (!sinfo)
3358 continue;
3359
3360 down_read(&sinfo->groups_sem);
3361 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3362 if (!list_empty(&sinfo->block_groups[c])) {
3363 u64 flags;
3364
3365 btrfs_get_block_group_info(
3366 &sinfo->block_groups[c], &space);
3367 if (space.total_bytes == 0 ||
3368 space.used_bytes == 0)
3369 continue;
3370 flags = space.flags;
3371 /*
3372 * return
3373 * 0: if dup, single or RAID0 is configured for
3374 * any of metadata, system or data, else
3375 * 1: if RAID5 is configured, or if RAID1 or
3376 * RAID10 is configured and only two mirrors
3377 * are used, else
3378 * 2: if RAID6 is configured, else
3379 * num_mirrors - 1: if RAID1 or RAID10 is
3380 * configured and more than
3381 * 2 mirrors are used.
3382 */
3383 if (num_tolerated_disk_barrier_failures > 0 &&
3384 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3385 BTRFS_BLOCK_GROUP_RAID0)) ||
3386 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3387 == 0)))
3388 num_tolerated_disk_barrier_failures = 0;
53b381b3
DW		 3389 else if (num_tolerated_disk_barrier_failures > 1) {
3390 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3391 BTRFS_BLOCK_GROUP_RAID5 |
3392 BTRFS_BLOCK_GROUP_RAID10)) {
3393 num_tolerated_disk_barrier_failures = 1;
3394 } else if (flags &
15b0a89d 3395 BTRFS_BLOCK_GROUP_RAID6) {
53b381b3
DW		 3396 num_tolerated_disk_barrier_failures = 2;
3397 }
3398 }
5af3e8cc
SB		 3399 }
3400 }
3401 up_read(&sinfo->groups_sem);
3402 }
3403
3404 return num_tolerated_disk_barrier_failures;
3405}
3406
48a3b636 3407static int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 3408{
e5e9a520 3409 struct list_head *head;
f2984462 3410 struct btrfs_device *dev;
a061fc8d 3411 struct btrfs_super_block *sb;
f2984462 3412 struct btrfs_dev_item *dev_item;
f2984462
CM		 3413 int ret;
3414 int do_barriers;
a236aed1
CM		 3415 int max_errors;
3416 int total_errors = 0;
a061fc8d 3417 u64 flags;
f2984462
CM		 3418
3419 do_barriers = !btrfs_test_opt(root, NOBARRIER);
af31f5e5 3420 backup_super_roots(root->fs_info);
f2984462 3421
6c41761f 3422 sb = root->fs_info->super_for_commit;
a061fc8d 3423 dev_item = &sb->dev_item;
e5e9a520 3424
174ba509 3425 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 3426 head = &root->fs_info->fs_devices->devices;
d7306801 3427 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
387125fc 3428
5af3e8cc
SB		 3429 if (do_barriers) {
3430 ret = barrier_all_devices(root->fs_info);
3431 if (ret) {
3432 mutex_unlock(
3433 &root->fs_info->fs_devices->device_list_mutex);
3434 btrfs_error(root->fs_info, ret,
3435 "errors while submitting device barriers.");
3436 return ret;
3437 }
3438 }
387125fc 3439
1f78160c 3440 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3441 if (!dev->bdev) {
3442 total_errors++;
3443 continue;
3444 }
2b82032c 3445 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3446 continue;
3447
2b82032c 3448 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 3449 btrfs_set_stack_device_type(dev_item, dev->type);
3450 btrfs_set_stack_device_id(dev_item, dev->devid);
7df69d3e 3451 btrfs_set_stack_device_total_bytes(dev_item,
935e5cc9 3452 dev->commit_total_bytes);
ce7213c7
MX		 3453 btrfs_set_stack_device_bytes_used(dev_item,
3454 dev->commit_bytes_used);
a061fc8d
CM		 3455 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3456 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3457 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3458 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 3459 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 3460
a061fc8d
CM		 3461 flags = btrfs_super_flags(sb);
3462 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3463
a512bbf8 3464 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 3465 if (ret)
3466 total_errors++;
f2984462 3467 }
a236aed1 3468 if (total_errors > max_errors) {
efe120a0 3469 btrfs_err(root->fs_info, "%d errors while writing supers",
d397712b 3470 total_errors);
a724b436 3471 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
79787eaa 3472
9d565ba4
SB		 3473 /* FUA is masked off if unsupported and can't be the reason */
3474 btrfs_error(root->fs_info, -EIO,
3475 "%d errors while writing supers", total_errors);
3476 return -EIO;
a236aed1 3477 }
f2984462 3478
a512bbf8 3479 total_errors = 0;
1f78160c 3480 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3481 if (!dev->bdev)
3482 continue;
2b82032c 3483 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3484 continue;
3485
a512bbf8
YZ		 3486 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3487 if (ret)
3488 total_errors++;
f2984462 3489 }
174ba509 3490 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 3491 if (total_errors > max_errors) {
79787eaa
JM		 3492 btrfs_error(root->fs_info, -EIO,
3493 "%d errors while writing supers", total_errors);
3494 return -EIO;
a236aed1 3495 }
f2984462
CM		 3496 return 0;
3497}
3498
a512bbf8
YZ		 3499int write_ctree_super(struct btrfs_trans_handle *trans,
3500 struct btrfs_root *root, int max_mirrors)
eb60ceac 3501{
f570e757 3502 return write_all_supers(root, max_mirrors);
cfaa7295
CM		 3503}
3504
cb517eab
MX		 3505/* Drop a fs root from the radix tree and free it. */
3506void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3507 struct btrfs_root *root)
2619ba1f 3508{
4df27c4d 3509 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 3510 radix_tree_delete(&fs_info->fs_roots_radix,
3511 (unsigned long)root->root_key.objectid);
4df27c4d 3512 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 3513
3514 if (btrfs_root_refs(&root->root_item) == 0)
3515 synchronize_srcu(&fs_info->subvol_srcu);
3516
1a4319cc 3517 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
3321719e 3518 btrfs_free_log(NULL, root);
3321719e 3519
faa2dbf0
JB		 3520 if (root->free_ino_pinned)
3521 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3522 if (root->free_ino_ctl)
3523 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3524 free_fs_root(root);
4df27c4d
YZ		 3525}
3526
3527static void free_fs_root(struct btrfs_root *root)
3528{
57cdc8db 3529 iput(root->ino_cache_inode);
4df27c4d 3530 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
1cb048f5
FDBM		 3531 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3532 root->orphan_block_rsv = NULL;
0ee5dc67
AV		 3533 if (root->anon_dev)
3534 free_anon_bdev(root->anon_dev);
8257b2dc
MX		 3535 if (root->subv_writers)
3536 btrfs_free_subvolume_writers(root->subv_writers);
4df27c4d
YZ		 3537 free_extent_buffer(root->node);
3538 free_extent_buffer(root->commit_root);
581bb050
LZ		 3539 kfree(root->free_ino_ctl);
3540 kfree(root->free_ino_pinned);
d397712b 3541 kfree(root->name);
b0feb9d9 3542 btrfs_put_fs_root(root);
2619ba1f
CM		 3543}
3544
cb517eab
MX		 3545void btrfs_free_fs_root(struct btrfs_root *root)
3546{
3547 free_fs_root(root);
2619ba1f
CM		 3548}
3549
c146afad 3550int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3551{
c146afad
YZ		 3552 u64 root_objectid = 0;
3553 struct btrfs_root *gang[8];
65d33fd7
QW		 3554 int i = 0;
3555 int err = 0;
3556 unsigned int ret = 0;
3557 int index;
e089f05c 3558
c146afad 3559 while (1) {
65d33fd7 3560 index = srcu_read_lock(&fs_info->subvol_srcu);
c146afad
YZ		 3561 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3562 (void **)gang, root_objectid,
3563 ARRAY_SIZE(gang));
65d33fd7
QW		 3564 if (!ret) {
3565 srcu_read_unlock(&fs_info->subvol_srcu, index);
c146afad 3566 break;
65d33fd7 3567 }
5d4f98a2 3568 root_objectid = gang[ret - 1]->root_key.objectid + 1;
65d33fd7 3569
c146afad 3570 for (i = 0; i < ret; i++) {
65d33fd7
QW		 3571 /* Avoid to grab roots in dead_roots */
3572 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3573 gang[i] = NULL;
3574 continue;
3575 }
3576 /* grab all the search result for later use */
3577 gang[i] = btrfs_grab_fs_root(gang[i]);
3578 }
3579 srcu_read_unlock(&fs_info->subvol_srcu, index);
66b4ffd1 3580
65d33fd7
QW		 3581 for (i = 0; i < ret; i++) {
3582 if (!gang[i])
3583 continue;
c146afad 3584 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 3585 err = btrfs_orphan_cleanup(gang[i]);
3586 if (err)
65d33fd7
QW		 3587 break;
3588 btrfs_put_fs_root(gang[i]);
c146afad
YZ		 3589 }
3590 root_objectid++;
3591 }
65d33fd7
QW		 3592
3593 /* release the uncleaned roots due to error */
3594 for (; i < ret; i++) {
3595 if (gang[i])
3596 btrfs_put_fs_root(gang[i]);
3597 }
3598 return err;
c146afad 3599}
a2135011 3600
c146afad
YZ		 3601int btrfs_commit_super(struct btrfs_root *root)
3602{
3603 struct btrfs_trans_handle *trans;
a74a4b97 3604
c146afad 3605 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 3606 btrfs_run_delayed_iputs(root);
c146afad 3607 mutex_unlock(&root->fs_info->cleaner_mutex);
9d1a2a3a 3608 wake_up_process(root->fs_info->cleaner_kthread);
c71bf099
YZ		 3609
3610 /* wait until ongoing cleanup work done */
3611 down_write(&root->fs_info->cleanup_work_sem);
3612 up_write(&root->fs_info->cleanup_work_sem);
3613
7a7eaa40 3614 trans = btrfs_join_transaction(root);
3612b495
TI		 3615 if (IS_ERR(trans))
3616 return PTR_ERR(trans);
d52c1bcc 3617 return btrfs_commit_transaction(trans, root);
c146afad
YZ		 3618}
3619
3abdbd78 3620void close_ctree(struct btrfs_root *root)
c146afad
YZ		 3621{
3622 struct btrfs_fs_info *fs_info = root->fs_info;
3623 int ret;
3624
3625 fs_info->closing = 1;
3626 smp_mb();
3627
803b2f54
SB		 3628 /* wait for the uuid_scan task to finish */
3629 down(&fs_info->uuid_tree_rescan_sem);
3630 /* avoid complains from lockdep et al., set sem back to initial state */
3631 up(&fs_info->uuid_tree_rescan_sem);
3632
837d5b6e 3633 /* pause restriper - we want to resume on mount */
aa1b8cd4 3634 btrfs_pause_balance(fs_info);
837d5b6e 3635
8dabb742
SB		 3636 btrfs_dev_replace_suspend_for_unmount(fs_info);
3637
aa1b8cd4 3638 btrfs_scrub_cancel(fs_info);
4cb5300b
CM		 3639
3640 /* wait for any defraggers to finish */
3641 wait_event(fs_info->transaction_wait,
3642 (atomic_read(&fs_info->defrag_running) == 0));
3643
3644 /* clear out the rbtree of defraggable inodes */
26176e7c 3645 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3646
21c7e756
MX		 3647 cancel_work_sync(&fs_info->async_reclaim_work);
3648
c146afad 3649 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 3650 ret = btrfs_commit_super(root);
3651 if (ret)
efe120a0 3652 btrfs_err(root->fs_info, "commit super ret %d", ret);
acce952b 3653 }
3654
87533c47 3655 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
68ce9682 3656 btrfs_error_commit_super(root);
0f7d52f4 3657
e3029d9f
AV		 3658 kthread_stop(fs_info->transaction_kthread);
3659 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3660
f25784b3
YZ		 3661 fs_info->closing = 2;
3662 smp_mb();
3663
bcef60f2
AJ		 3664 btrfs_free_qgroup_config(root->fs_info);
3665
963d678b 3666 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
efe120a0 3667 btrfs_info(root->fs_info, "at unmount delalloc count %lld",
963d678b 3668 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3669 }
bcc63abb 3670
5ac1d209
JM		 3671 btrfs_sysfs_remove_one(fs_info);
3672
faa2dbf0 3673 btrfs_free_fs_roots(fs_info);
d10c5f31 3674
1a4319cc
LB		 3675 btrfs_put_block_group_cache(fs_info);
3676
2b1360da
JB		 3677 btrfs_free_block_groups(fs_info);
3678
de348ee0
WS		 3679 /*
3680 * we must make sure there is not any read request to
3681 * submit after we stopping all workers.
3682 */
3683 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
96192499
JB		 3684 btrfs_stop_all_workers(fs_info);
3685
47ab2a6c 3686 fs_info->open = 0;
13e6c37b 3687 free_root_pointers(fs_info, 1);
9ad6b7bc 3688
13e6c37b 3689 iput(fs_info->btree_inode);
d6bfde87 3690
21adbd5c
SB		 3691#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3692 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3693 btrfsic_unmount(root, fs_info->fs_devices);
3694#endif
3695
dfe25020 3696 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3697 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3698
e2d84521 3699 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 3700 percpu_counter_destroy(&fs_info->delalloc_bytes);
c404e0dc 3701 percpu_counter_destroy(&fs_info->bio_counter);
04160088 3702 bdi_destroy(&fs_info->bdi);
76dda93c 3703 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 3704
53b381b3
DW		 3705 btrfs_free_stripe_hash_table(fs_info);
3706
1cb048f5
FDBM		 3707 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3708 root->orphan_block_rsv = NULL;
eb60ceac
CM		 3709}
3710
b9fab919
CM		 3711int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3712 int atomic)
5f39d397 3713{
1259ab75 3714 int ret;
727011e0 3715 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 3716
0b32f4bb 3717 ret = extent_buffer_uptodate(buf);
1259ab75
CM		 3718 if (!ret)
3719 return ret;
3720
3721 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM		 3722 parent_transid, atomic);
3723 if (ret == -EAGAIN)
3724 return ret;
1259ab75 3725 return !ret;
5f39d397
CM		 3726}
3727
3728int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 3729{
0b32f4bb 3730 return set_extent_buffer_uptodate(buf);
5f39d397 3731}
6702ed49 3732
5f39d397
CM		 3733void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3734{
06ea65a3 3735 struct btrfs_root *root;
5f39d397 3736 u64 transid = btrfs_header_generation(buf);
b9473439 3737 int was_dirty;
b4ce94de 3738
06ea65a3
JB		 3739#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3740 /*
3741 * This is a fast path so only do this check if we have sanity tests
3742 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3743 * outside of the sanity tests.
3744 */
3745 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3746 return;
3747#endif
3748 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
b9447ef8 3749 btrfs_assert_tree_locked(buf);
31b1a2bd
JL		 3750 if (transid != root->fs_info->generation)
3751 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
d397712b 3752 "found %llu running %llu\n",
c1c9ff7c 3753 buf->start, transid, root->fs_info->generation);
0b32f4bb 3754 was_dirty = set_extent_buffer_dirty(buf);
e2d84521
MX		 3755 if (!was_dirty)
3756 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3757 buf->len,
3758 root->fs_info->dirty_metadata_batch);
1f21ef0a
FM		 3759#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3760 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
3761 btrfs_print_leaf(root, buf);
3762 ASSERT(0);
3763 }
3764#endif
eb60ceac
CM		 3765}
3766
b53d3f5d
LB		 3767static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3768 int flush_delayed)
16cdcec7
MX		 3769{
3770 /*
3771 * looks as though older kernels can get into trouble with
3772 * this code, they end up stuck in balance_dirty_pages forever
3773 */
e2d84521 3774 int ret;
16cdcec7
MX		 3775
3776 if (current->flags & PF_MEMALLOC)
3777 return;
3778
b53d3f5d
LB		 3779 if (flush_delayed)
3780 btrfs_balance_delayed_items(root);
16cdcec7 3781
e2d84521
MX		 3782 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3783 BTRFS_DIRTY_METADATA_THRESH);
3784 if (ret > 0) {
d0e1d66b
NJ		 3785 balance_dirty_pages_ratelimited(
3786 root->fs_info->btree_inode->i_mapping);
16cdcec7
MX		 3787 }
3788 return;
3789}
3790
b53d3f5d 3791void btrfs_btree_balance_dirty(struct btrfs_root *root)
35b7e476 3792{
b53d3f5d
LB		 3793 __btrfs_btree_balance_dirty(root, 1);
3794}
585ad2c3 3795
b53d3f5d
LB		 3796void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3797{
3798 __btrfs_btree_balance_dirty(root, 0);
35b7e476 3799}
6b80053d 3800
ca7a79ad 3801int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 3802{
727011e0 3803 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
0b32f4bb 3804 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
6b80053d 3805}
0da5468f 3806
fcd1f065 3807static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 3808 int read_only)
3809{
1104a885
DS		 3810 /*
3811 * Placeholder for checks
3812 */
fcd1f065 3813 return 0;
acce952b 3814}
3815
48a3b636 3816static void btrfs_error_commit_super(struct btrfs_root *root)
acce952b 3817{
acce952b 3818 mutex_lock(&root->fs_info->cleaner_mutex);
3819 btrfs_run_delayed_iputs(root);
3820 mutex_unlock(&root->fs_info->cleaner_mutex);
3821
3822 down_write(&root->fs_info->cleanup_work_sem);
3823 up_write(&root->fs_info->cleanup_work_sem);
3824
3825 /* cleanup FS via transaction */
3826 btrfs_cleanup_transaction(root);
acce952b 3827}
3828
143bede5 3829static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 3830{
acce952b 3831 struct btrfs_ordered_extent *ordered;
acce952b 3832
199c2a9c 3833 spin_lock(&root->ordered_extent_lock);
779880ef
JB		 3834 /*
3835 * This will just short circuit the ordered completion stuff which will
3836 * make sure the ordered extent gets properly cleaned up.
3837 */
199c2a9c 3838 list_for_each_entry(ordered, &root->ordered_extents,
779880ef
JB		 3839 root_extent_list)
3840 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
199c2a9c
MX		 3841 spin_unlock(&root->ordered_extent_lock);
3842}
3843
3844static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3845{
3846 struct btrfs_root *root;
3847 struct list_head splice;
3848
3849 INIT_LIST_HEAD(&splice);
3850
3851 spin_lock(&fs_info->ordered_root_lock);
3852 list_splice_init(&fs_info->ordered_roots, &splice);
3853 while (!list_empty(&splice)) {
3854 root = list_first_entry(&splice, struct btrfs_root,
3855 ordered_root);
1de2cfde
JB		 3856 list_move_tail(&root->ordered_root,
3857 &fs_info->ordered_roots);
199c2a9c 3858
2a85d9ca 3859 spin_unlock(&fs_info->ordered_root_lock);
199c2a9c
MX		 3860 btrfs_destroy_ordered_extents(root);
3861
2a85d9ca
LB		 3862 cond_resched();
3863 spin_lock(&fs_info->ordered_root_lock);
199c2a9c
MX		 3864 }
3865 spin_unlock(&fs_info->ordered_root_lock);
acce952b 3866}
3867
35a3621b
SB		 3868static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3869 struct btrfs_root *root)
acce952b 3870{
3871 struct rb_node *node;
3872 struct btrfs_delayed_ref_root *delayed_refs;
3873 struct btrfs_delayed_ref_node *ref;
3874 int ret = 0;
3875
3876 delayed_refs = &trans->delayed_refs;
3877
3878 spin_lock(&delayed_refs->lock);
d7df2c79 3879 if (atomic_read(&delayed_refs->num_entries) == 0) {
cfece4db 3880 spin_unlock(&delayed_refs->lock);
efe120a0 3881 btrfs_info(root->fs_info, "delayed_refs has NO entry");
acce952b 3882 return ret;
3883 }
3884
d7df2c79
JB		 3885 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
3886 struct btrfs_delayed_ref_head *head;
e78417d1 3887 bool pin_bytes = false;
acce952b 3888
d7df2c79
JB		 3889 head = rb_entry(node, struct btrfs_delayed_ref_head,
3890 href_node);
3891 if (!mutex_trylock(&head->mutex)) {
3892 atomic_inc(&head->node.refs);
3893 spin_unlock(&delayed_refs->lock);
eb12db69 3894
d7df2c79 3895 mutex_lock(&head->mutex);
e78417d1 3896 mutex_unlock(&head->mutex);
d7df2c79
JB		 3897 btrfs_put_delayed_ref(&head->node);
3898 spin_lock(&delayed_refs->lock);
3899 continue;
3900 }
3901 spin_lock(&head->lock);
3902 while ((node = rb_first(&head->ref_root)) != NULL) {
3903 ref = rb_entry(node, struct btrfs_delayed_ref_node,
3904 rb_node);
3905 ref->in_tree = 0;
3906 rb_erase(&ref->rb_node, &head->ref_root);
3907 atomic_dec(&delayed_refs->num_entries);
3908 btrfs_put_delayed_ref(ref);
e78417d1 3909 }
d7df2c79
JB		 3910 if (head->must_insert_reserved)
3911 pin_bytes = true;
3912 btrfs_free_delayed_extent_op(head->extent_op);
3913 delayed_refs->num_heads--;
3914 if (head->processing == 0)
3915 delayed_refs->num_heads_ready--;
3916 atomic_dec(&delayed_refs->num_entries);
3917 head->node.in_tree = 0;
3918 rb_erase(&head->href_node, &delayed_refs->href_root);
3919 spin_unlock(&head->lock);
3920 spin_unlock(&delayed_refs->lock);
3921 mutex_unlock(&head->mutex);
acce952b 3922
d7df2c79
JB		 3923 if (pin_bytes)
3924 btrfs_pin_extent(root, head->node.bytenr,
3925 head->node.num_bytes, 1);
3926 btrfs_put_delayed_ref(&head->node);
acce952b 3927 cond_resched();
3928 spin_lock(&delayed_refs->lock);
3929 }
3930
3931 spin_unlock(&delayed_refs->lock);
3932
3933 return ret;
3934}
3935
143bede5 3936static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 3937{
3938 struct btrfs_inode *btrfs_inode;
3939 struct list_head splice;
3940
3941 INIT_LIST_HEAD(&splice);
3942
eb73c1b7
MX		 3943 spin_lock(&root->delalloc_lock);
3944 list_splice_init(&root->delalloc_inodes, &splice);
acce952b 3945
3946 while (!list_empty(&splice)) {
eb73c1b7
MX		 3947 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3948 delalloc_inodes);
acce952b 3949
3950 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX		 3951 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3952 &btrfs_inode->runtime_flags);
eb73c1b7 3953 spin_unlock(&root->delalloc_lock);
acce952b 3954
3955 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3956
eb73c1b7 3957 spin_lock(&root->delalloc_lock);
acce952b 3958 }
3959
eb73c1b7
MX		 3960 spin_unlock(&root->delalloc_lock);
3961}
3962
3963static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3964{
3965 struct btrfs_root *root;
3966 struct list_head splice;
3967
3968 INIT_LIST_HEAD(&splice);
3969
3970 spin_lock(&fs_info->delalloc_root_lock);
3971 list_splice_init(&fs_info->delalloc_roots, &splice);
3972 while (!list_empty(&splice)) {
3973 root = list_first_entry(&splice, struct btrfs_root,
3974 delalloc_root);
3975 list_del_init(&root->delalloc_root);
3976 root = btrfs_grab_fs_root(root);
3977 BUG_ON(!root);
3978 spin_unlock(&fs_info->delalloc_root_lock);
3979
3980 btrfs_destroy_delalloc_inodes(root);
3981 btrfs_put_fs_root(root);
3982
3983 spin_lock(&fs_info->delalloc_root_lock);
3984 }
3985 spin_unlock(&fs_info->delalloc_root_lock);
acce952b 3986}
3987
3988static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3989 struct extent_io_tree *dirty_pages,
3990 int mark)
3991{
3992 int ret;
acce952b 3993 struct extent_buffer *eb;
3994 u64 start = 0;
3995 u64 end;
acce952b 3996
3997 while (1) {
3998 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 3999 mark, NULL);
acce952b 4000 if (ret)
4001 break;
4002
4003 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
4004 while (start <= end) {
fd8b2b61 4005 eb = btrfs_find_tree_block(root, start,
707e8a07
DS		 4006 root->nodesize);
4007 start += root->nodesize;
fd8b2b61 4008 if (!eb)
acce952b 4009 continue;
fd8b2b61 4010 wait_on_extent_buffer_writeback(eb);
acce952b 4011
fd8b2b61
JB		 4012 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4013 &eb->bflags))
4014 clear_extent_buffer_dirty(eb);
4015 free_extent_buffer_stale(eb);
acce952b 4016 }
4017 }
4018
4019 return ret;
4020}
4021
4022static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4023 struct extent_io_tree *pinned_extents)
4024{
4025 struct extent_io_tree *unpin;
4026 u64 start;
4027 u64 end;
4028 int ret;
ed0eaa14 4029 bool loop = true;
acce952b 4030
4031 unpin = pinned_extents;
ed0eaa14 4032again:
acce952b 4033 while (1) {
4034 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 4035 EXTENT_DIRTY, NULL);
acce952b 4036 if (ret)
4037 break;
4038
4039 /* opt_discard */
5378e607
LD		 4040 if (btrfs_test_opt(root, DISCARD))
4041 ret = btrfs_error_discard_extent(root, start,
4042 end + 1 - start,
4043 NULL);
acce952b 4044
4045 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4046 btrfs_error_unpin_extent_range(root, start, end);
4047 cond_resched();
4048 }
4049
ed0eaa14
LB		 4050 if (loop) {
4051 if (unpin == &root->fs_info->freed_extents[0])
4052 unpin = &root->fs_info->freed_extents[1];
4053 else
4054 unpin = &root->fs_info->freed_extents[0];
4055 loop = false;
4056 goto again;
4057 }
4058
acce952b 4059 return 0;
4060}
4061
49b25e05
JM		 4062void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4063 struct btrfs_root *root)
4064{
4065 btrfs_destroy_delayed_refs(cur_trans, root);
49b25e05 4066
4a9d8bde 4067 cur_trans->state = TRANS_STATE_COMMIT_START;
d7096fc3 4068 wake_up(&root->fs_info->transaction_blocked_wait);
49b25e05 4069
4a9d8bde 4070 cur_trans->state = TRANS_STATE_UNBLOCKED;
d7096fc3 4071 wake_up(&root->fs_info->transaction_wait);
49b25e05 4072
67cde344
MX		 4073 btrfs_destroy_delayed_inodes(root);
4074 btrfs_assert_delayed_root_empty(root);
49b25e05 4075
49b25e05
JM		 4076 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4077 EXTENT_DIRTY);
6e841e32
LB		 4078 btrfs_destroy_pinned_extent(root,
4079 root->fs_info->pinned_extents);
49b25e05 4080
4a9d8bde
MX		 4081 cur_trans->state =TRANS_STATE_COMPLETED;
4082 wake_up(&cur_trans->commit_wait);
4083
49b25e05
JM		 4084 /*
4085 memset(cur_trans, 0, sizeof(*cur_trans));
4086 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4087 */
4088}
4089
48a3b636 4090static int btrfs_cleanup_transaction(struct btrfs_root *root)
acce952b 4091{
4092 struct btrfs_transaction *t;
acce952b 4093
acce952b 4094 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4095
a4abeea4 4096 spin_lock(&root->fs_info->trans_lock);
724e2315
JB		 4097 while (!list_empty(&root->fs_info->trans_list)) {
4098 t = list_first_entry(&root->fs_info->trans_list,
4099 struct btrfs_transaction, list);
4100 if (t->state >= TRANS_STATE_COMMIT_START) {
4101 atomic_inc(&t->use_count);
4102 spin_unlock(&root->fs_info->trans_lock);
4103 btrfs_wait_for_commit(root, t->transid);
4104 btrfs_put_transaction(t);
4105 spin_lock(&root->fs_info->trans_lock);
4106 continue;
4107 }
4108 if (t == root->fs_info->running_transaction) {
4109 t->state = TRANS_STATE_COMMIT_DOING;
4110 spin_unlock(&root->fs_info->trans_lock);
4111 /*
4112 * We wait for 0 num_writers since we don't hold a trans
4113 * handle open currently for this transaction.
4114 */
4115 wait_event(t->writer_wait,
4116 atomic_read(&t->num_writers) == 0);
4117 } else {
4118 spin_unlock(&root->fs_info->trans_lock);
4119 }
4120 btrfs_cleanup_one_transaction(t, root);
4a9d8bde 4121
724e2315
JB		 4122 spin_lock(&root->fs_info->trans_lock);
4123 if (t == root->fs_info->running_transaction)
4124 root->fs_info->running_transaction = NULL;
acce952b 4125 list_del_init(&t->list);
724e2315 4126 spin_unlock(&root->fs_info->trans_lock);
acce952b 4127
724e2315
JB		 4128 btrfs_put_transaction(t);
4129 trace_btrfs_transaction_commit(root);
4130 spin_lock(&root->fs_info->trans_lock);
4131 }
4132 spin_unlock(&root->fs_info->trans_lock);
4133 btrfs_destroy_all_ordered_extents(root->fs_info);
4134 btrfs_destroy_delayed_inodes(root);
4135 btrfs_assert_delayed_root_empty(root);
4136 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4137 btrfs_destroy_all_delalloc_inodes(root->fs_info);
acce952b 4138 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 4139
4140 return 0;
4141}
4142
d1310b2e 4143static struct extent_io_ops btree_extent_io_ops = {
ce9adaa5 4144 .readpage_end_io_hook = btree_readpage_end_io_hook,
4bb31e92 4145 .readpage_io_failed_hook = btree_io_failed_hook,
0b86a832 4146 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 4147 /* note we're sharing with inode.c for the merge bio hook */
4148 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 4149};
Linux 6.x block layer and io_uring tree(s)