projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: check for an extent_op on the locked ref
[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"
8b712842 42#include "async-thread.h"
925baedd 43#include "locking.h"
e02119d5 44#include "tree-log.h"
fa9c0d79 45#include "free-space-cache.h"
581bb050 46#include "inode-map.h"
21adbd5c 47#include "check-integrity.h"
606686ee 48#include "rcu-string.h"
8dabb742 49#include "dev-replace.h"
53b381b3 50#include "raid56.h"
5ac1d209 51#include "sysfs.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;
d458b054 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);
569e0f35
JB		 62static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
63 struct btrfs_root *root);
143bede5 64static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
acce952b 65static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
66 struct btrfs_root *root);
143bede5 67static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
acce952b 68static int btrfs_destroy_marked_extents(struct btrfs_root *root,
69 struct extent_io_tree *dirty_pages,
70 int mark);
71static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
72 struct extent_io_tree *pinned_extents);
48a3b636
ES		 73static int btrfs_cleanup_transaction(struct btrfs_root *root);
74static void btrfs_error_commit_super(struct btrfs_root *root);
ce9adaa5 75
d352ac68
CM		 76/*
77 * end_io_wq structs are used to do processing in task context when an IO is
78 * complete. This is used during reads to verify checksums, and it is used
79 * by writes to insert metadata for new file extents after IO is complete.
80 */
ce9adaa5
CM		 81struct end_io_wq {
82 struct bio *bio;
83 bio_end_io_t *end_io;
84 void *private;
85 struct btrfs_fs_info *info;
86 int error;
22c59948 87 int metadata;
ce9adaa5 88 struct list_head list;
d458b054 89 struct btrfs_work work;
ce9adaa5 90};
0da5468f 91
d352ac68
CM		 92/*
93 * async submit bios are used to offload expensive checksumming
94 * onto the worker threads. They checksum file and metadata bios
95 * just before they are sent down the IO stack.
96 */
44b8bd7e
CM		 97struct async_submit_bio {
98 struct inode *inode;
99 struct bio *bio;
100 struct list_head list;
4a69a410
CM		 101 extent_submit_bio_hook_t *submit_bio_start;
102 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM		 103 int rw;
104 int mirror_num;
c8b97818 105 unsigned long bio_flags;
eaf25d93
CM		 106 /*
107 * bio_offset is optional, can be used if the pages in the bio
108 * can't tell us where in the file the bio should go
109 */
110 u64 bio_offset;
d458b054 111 struct btrfs_work work;
79787eaa 112 int error;
44b8bd7e
CM		 113};
114
85d4e461
CM		 115/*
116 * Lockdep class keys for extent_buffer->lock's in this root. For a given
117 * eb, the lockdep key is determined by the btrfs_root it belongs to and
118 * the level the eb occupies in the tree.
119 *
120 * Different roots are used for different purposes and may nest inside each
121 * other and they require separate keysets. As lockdep keys should be
122 * static, assign keysets according to the purpose of the root as indicated
123 * by btrfs_root->objectid. This ensures that all special purpose roots
124 * have separate keysets.
4008c04a 125 *
85d4e461
CM		 126 * Lock-nesting across peer nodes is always done with the immediate parent
127 * node locked thus preventing deadlock. As lockdep doesn't know this, use
128 * subclass to avoid triggering lockdep warning in such cases.
4008c04a 129 *
85d4e461
CM		 130 * The key is set by the readpage_end_io_hook after the buffer has passed
131 * csum validation but before the pages are unlocked. It is also set by
132 * btrfs_init_new_buffer on freshly allocated blocks.
4008c04a 133 *
85d4e461
CM		 134 * We also add a check to make sure the highest level of the tree is the
135 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
136 * needs update as well.
4008c04a
CM		 137 */
138#ifdef CONFIG_DEBUG_LOCK_ALLOC
139# if BTRFS_MAX_LEVEL != 8
140# error
141# endif
85d4e461
CM		 142
143static struct btrfs_lockdep_keyset {
144 u64 id; /* root objectid */
145 const char *name_stem; /* lock name stem */
146 char names[BTRFS_MAX_LEVEL + 1][20];
147 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
148} btrfs_lockdep_keysets[] = {
149 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
150 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
151 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
152 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
153 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
154 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
60b62978 155 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
85d4e461
CM		 156 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
157 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
158 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
13fd8da9 159 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
85d4e461 160 { .id = 0, .name_stem = "tree" },
4008c04a 161};
85d4e461
CM		 162
163void __init btrfs_init_lockdep(void)
164{
165 int i, j;
166
167 /* initialize lockdep class names */
168 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
169 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
170
171 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
172 snprintf(ks->names[j], sizeof(ks->names[j]),
173 "btrfs-%s-%02d", ks->name_stem, j);
174 }
175}
176
177void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
178 int level)
179{
180 struct btrfs_lockdep_keyset *ks;
181
182 BUG_ON(level >= ARRAY_SIZE(ks->keys));
183
184 /* find the matching keyset, id 0 is the default entry */
185 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
186 if (ks->id == objectid)
187 break;
188
189 lockdep_set_class_and_name(&eb->lock,
190 &ks->keys[level], ks->names[level]);
191}
192
4008c04a
CM		 193#endif
194
d352ac68
CM		 195/*
196 * extents on the btree inode are pretty simple, there's one extent
197 * that covers the entire device
198 */
b2950863 199static struct extent_map *btree_get_extent(struct inode *inode,
306e16ce 200 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 201 int create)
7eccb903 202{
5f39d397
CM		 203 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
204 struct extent_map *em;
205 int ret;
206
890871be 207 read_lock(&em_tree->lock);
d1310b2e 208 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 209 if (em) {
210 em->bdev =
211 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 212 read_unlock(&em_tree->lock);
5f39d397 213 goto out;
a061fc8d 214 }
890871be 215 read_unlock(&em_tree->lock);
7b13b7b1 216
172ddd60 217 em = alloc_extent_map();
5f39d397
CM		 218 if (!em) {
219 em = ERR_PTR(-ENOMEM);
220 goto out;
221 }
222 em->start = 0;
0afbaf8c 223 em->len = (u64)-1;
c8b97818 224 em->block_len = (u64)-1;
5f39d397 225 em->block_start = 0;
a061fc8d 226 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 227
890871be 228 write_lock(&em_tree->lock);
09a2a8f9 229 ret = add_extent_mapping(em_tree, em, 0);
5f39d397
CM		 230 if (ret == -EEXIST) {
231 free_extent_map(em);
7b13b7b1 232 em = lookup_extent_mapping(em_tree, start, len);
b4f359ab 233 if (!em)
0433f20d 234 em = ERR_PTR(-EIO);
5f39d397 235 } else if (ret) {
7b13b7b1 236 free_extent_map(em);
0433f20d 237 em = ERR_PTR(ret);
5f39d397 238 }
890871be 239 write_unlock(&em_tree->lock);
7b13b7b1 240
5f39d397
CM		 241out:
242 return em;
7eccb903
CM		 243}
244
b0496686 245u32 btrfs_csum_data(char *data, u32 seed, size_t len)
19c00ddc 246{
0b947aff 247 return btrfs_crc32c(seed, data, len);
19c00ddc
CM		 248}
249
250void btrfs_csum_final(u32 crc, char *result)
251{
7e75bf3f 252 put_unaligned_le32(~crc, result);
19c00ddc
CM		 253}
254
d352ac68
CM		 255/*
256 * compute the csum for a btree block, and either verify it or write it
257 * into the csum field of the block.
258 */
19c00ddc
CM		 259static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
260 int verify)
261{
6c41761f 262 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
607d432d 263 char *result = NULL;
19c00ddc
CM		 264 unsigned long len;
265 unsigned long cur_len;
266 unsigned long offset = BTRFS_CSUM_SIZE;
19c00ddc
CM		 267 char *kaddr;
268 unsigned long map_start;
269 unsigned long map_len;
270 int err;
271 u32 crc = ~(u32)0;
607d432d 272 unsigned long inline_result;
19c00ddc
CM		 273
274 len = buf->len - offset;
d397712b 275 while (len > 0) {
19c00ddc 276 err = map_private_extent_buffer(buf, offset, 32,
a6591715 277 &kaddr, &map_start, &map_len);
d397712b 278 if (err)
19c00ddc 279 return 1;
19c00ddc 280 cur_len = min(len, map_len - (offset - map_start));
b0496686 281 crc = btrfs_csum_data(kaddr + offset - map_start,
19c00ddc
CM		 282 crc, cur_len);
283 len -= cur_len;
284 offset += cur_len;
19c00ddc 285 }
607d432d
JB		 286 if (csum_size > sizeof(inline_result)) {
287 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
288 if (!result)
289 return 1;
290 } else {
291 result = (char *)&inline_result;
292 }
293
19c00ddc
CM		 294 btrfs_csum_final(crc, result);
295
296 if (verify) {
607d432d 297 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 298 u32 val;
299 u32 found = 0;
607d432d 300 memcpy(&found, result, csum_size);
e4204ded 301
607d432d 302 read_extent_buffer(buf, &val, 0, csum_size);
efe120a0
FH		 303 printk_ratelimited(KERN_INFO
304 "BTRFS: %s checksum verify failed on %llu wanted %X found %X "
305 "level %d\n",
306 root->fs_info->sb->s_id, buf->start,
307 val, found, btrfs_header_level(buf));
607d432d
JB		 308 if (result != (char *)&inline_result)
309 kfree(result);
19c00ddc
CM		 310 return 1;
311 }
312 } else {
607d432d 313 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 314 }
607d432d
JB		 315 if (result != (char *)&inline_result)
316 kfree(result);
19c00ddc
CM		 317 return 0;
318}
319
d352ac68
CM		 320/*
321 * we can't consider a given block up to date unless the transid of the
322 * block matches the transid in the parent node's pointer. This is how we
323 * detect blocks that either didn't get written at all or got written
324 * in the wrong place.
325 */
1259ab75 326static int verify_parent_transid(struct extent_io_tree *io_tree,
b9fab919
CM		 327 struct extent_buffer *eb, u64 parent_transid,
328 int atomic)
1259ab75 329{
2ac55d41 330 struct extent_state *cached_state = NULL;
1259ab75
CM		 331 int ret;
332
333 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
334 return 0;
335
b9fab919
CM		 336 if (atomic)
337 return -EAGAIN;
338
2ac55d41 339 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
d0082371 340 0, &cached_state);
0b32f4bb 341 if (extent_buffer_uptodate(eb) &&
1259ab75
CM		 342 btrfs_header_generation(eb) == parent_transid) {
343 ret = 0;
344 goto out;
345 }
7a36ddec 346 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
193f284d 347 "found %llu\n",
c1c9ff7c 348 eb->start, parent_transid, btrfs_header_generation(eb));
1259ab75 349 ret = 1;
0b32f4bb 350 clear_extent_buffer_uptodate(eb);
33958dc6 351out:
2ac55d41
JB		 352 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
353 &cached_state, GFP_NOFS);
1259ab75 354 return ret;
1259ab75
CM		 355}
356
1104a885
DS		 357/*
358 * Return 0 if the superblock checksum type matches the checksum value of that
359 * algorithm. Pass the raw disk superblock data.
360 */
361static int btrfs_check_super_csum(char *raw_disk_sb)
362{
363 struct btrfs_super_block *disk_sb =
364 (struct btrfs_super_block *)raw_disk_sb;
365 u16 csum_type = btrfs_super_csum_type(disk_sb);
366 int ret = 0;
367
368 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
369 u32 crc = ~(u32)0;
370 const int csum_size = sizeof(crc);
371 char result[csum_size];
372
373 /*
374 * The super_block structure does not span the whole
375 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
376 * is filled with zeros and is included in the checkum.
377 */
378 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
379 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
380 btrfs_csum_final(crc, result);
381
382 if (memcmp(raw_disk_sb, result, csum_size))
383 ret = 1;
667e7d94
CM		 384
385 if (ret && btrfs_super_generation(disk_sb) < 10) {
efe120a0
FH		 386 printk(KERN_WARNING
387 "BTRFS: super block crcs don't match, older mkfs detected\n");
667e7d94
CM		 388 ret = 0;
389 }
1104a885
DS		 390 }
391
392 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
efe120a0 393 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
1104a885
DS		 394 csum_type);
395 ret = 1;
396 }
397
398 return ret;
399}
400
d352ac68
CM		 401/*
402 * helper to read a given tree block, doing retries as required when
403 * the checksums don't match and we have alternate mirrors to try.
404 */
f188591e
CM		 405static int btree_read_extent_buffer_pages(struct btrfs_root *root,
406 struct extent_buffer *eb,
ca7a79ad 407 u64 start, u64 parent_transid)
f188591e
CM		 408{
409 struct extent_io_tree *io_tree;
ea466794 410 int failed = 0;
f188591e
CM		 411 int ret;
412 int num_copies = 0;
413 int mirror_num = 0;
ea466794 414 int failed_mirror = 0;
f188591e 415
a826d6dc 416 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
f188591e
CM		 417 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
418 while (1) {
bb82ab88
AJ		 419 ret = read_extent_buffer_pages(io_tree, eb, start,
420 WAIT_COMPLETE,
f188591e 421 btree_get_extent, mirror_num);
256dd1bb
SB		 422 if (!ret) {
423 if (!verify_parent_transid(io_tree, eb,
b9fab919 424 parent_transid, 0))
256dd1bb
SB		 425 break;
426 else
427 ret = -EIO;
428 }
d397712b 429
a826d6dc
JB		 430 /*
431 * This buffer's crc is fine, but its contents are corrupted, so
432 * there is no reason to read the other copies, they won't be
433 * any less wrong.
434 */
435 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
ea466794
JB		 436 break;
437
5d964051 438 num_copies = btrfs_num_copies(root->fs_info,
f188591e 439 eb->start, eb->len);
4235298e 440 if (num_copies == 1)
ea466794 441 break;
4235298e 442
5cf1ab56
JB		 443 if (!failed_mirror) {
444 failed = 1;
445 failed_mirror = eb->read_mirror;
446 }
447
f188591e 448 mirror_num++;
ea466794
JB		 449 if (mirror_num == failed_mirror)
450 mirror_num++;
451
4235298e 452 if (mirror_num > num_copies)
ea466794 453 break;
f188591e 454 }
ea466794 455
c0901581 456 if (failed && !ret && failed_mirror)
ea466794
JB		 457 repair_eb_io_failure(root, eb, failed_mirror);
458
459 return ret;
f188591e 460}
19c00ddc 461
d352ac68 462/*
d397712b
CM		 463 * checksum a dirty tree block before IO. This has extra checks to make sure
464 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 465 */
d397712b 466
b2950863 467static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
19c00ddc 468{
4eee4fa4 469 u64 start = page_offset(page);
19c00ddc 470 u64 found_start;
19c00ddc 471 struct extent_buffer *eb;
f188591e 472
4f2de97a
JB		 473 eb = (struct extent_buffer *)page->private;
474 if (page != eb->pages[0])
475 return 0;
19c00ddc 476 found_start = btrfs_header_bytenr(eb);
fae7f21c 477 if (WARN_ON(found_start != start || !PageUptodate(page)))
4f2de97a 478 return 0;
19c00ddc 479 csum_tree_block(root, eb, 0);
19c00ddc
CM		 480 return 0;
481}
482
2b82032c
YZ		 483static int check_tree_block_fsid(struct btrfs_root *root,
484 struct extent_buffer *eb)
485{
486 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
487 u8 fsid[BTRFS_UUID_SIZE];
488 int ret = 1;
489
0a4e5586 490 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
2b82032c
YZ		 491 while (fs_devices) {
492 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
493 ret = 0;
494 break;
495 }
496 fs_devices = fs_devices->seed;
497 }
498 return ret;
499}
500
a826d6dc 501#define CORRUPT(reason, eb, root, slot) \
efe120a0
FH		 502 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
503 "root=%llu, slot=%d", reason, \
c1c9ff7c 504 btrfs_header_bytenr(eb), root->objectid, slot)
a826d6dc
JB		 505
506static noinline int check_leaf(struct btrfs_root *root,
507 struct extent_buffer *leaf)
508{
509 struct btrfs_key key;
510 struct btrfs_key leaf_key;
511 u32 nritems = btrfs_header_nritems(leaf);
512 int slot;
513
514 if (nritems == 0)
515 return 0;
516
517 /* Check the 0 item */
518 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
519 BTRFS_LEAF_DATA_SIZE(root)) {
520 CORRUPT("invalid item offset size pair", leaf, root, 0);
521 return -EIO;
522 }
523
524 /*
525 * Check to make sure each items keys are in the correct order and their
526 * offsets make sense. We only have to loop through nritems-1 because
527 * we check the current slot against the next slot, which verifies the
528 * next slot's offset+size makes sense and that the current's slot
529 * offset is correct.
530 */
531 for (slot = 0; slot < nritems - 1; slot++) {
532 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
533 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
534
535 /* Make sure the keys are in the right order */
536 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
537 CORRUPT("bad key order", leaf, root, slot);
538 return -EIO;
539 }
540
541 /*
542 * Make sure the offset and ends are right, remember that the
543 * item data starts at the end of the leaf and grows towards the
544 * front.
545 */
546 if (btrfs_item_offset_nr(leaf, slot) !=
547 btrfs_item_end_nr(leaf, slot + 1)) {
548 CORRUPT("slot offset bad", leaf, root, slot);
549 return -EIO;
550 }
551
552 /*
553 * Check to make sure that we don't point outside of the leaf,
554 * just incase all the items are consistent to eachother, but
555 * all point outside of the leaf.
556 */
557 if (btrfs_item_end_nr(leaf, slot) >
558 BTRFS_LEAF_DATA_SIZE(root)) {
559 CORRUPT("slot end outside of leaf", leaf, root, slot);
560 return -EIO;
561 }
562 }
563
564 return 0;
565}
566
facc8a22
MX		 567static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
568 u64 phy_offset, struct page *page,
569 u64 start, u64 end, int mirror)
ce9adaa5 570{
ce9adaa5
CM		 571 u64 found_start;
572 int found_level;
ce9adaa5
CM		 573 struct extent_buffer *eb;
574 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 575 int ret = 0;
727011e0 576 int reads_done;
ce9adaa5 577
ce9adaa5
CM		 578 if (!page->private)
579 goto out;
d397712b 580
4f2de97a 581 eb = (struct extent_buffer *)page->private;
d397712b 582
0b32f4bb
JB		 583 /* the pending IO might have been the only thing that kept this buffer
584 * in memory. Make sure we have a ref for all this other checks
585 */
586 extent_buffer_get(eb);
587
588 reads_done = atomic_dec_and_test(&eb->io_pages);
727011e0
CM		 589 if (!reads_done)
590 goto err;
f188591e 591
5cf1ab56 592 eb->read_mirror = mirror;
ea466794
JB		 593 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
594 ret = -EIO;
595 goto err;
596 }
597
ce9adaa5 598 found_start = btrfs_header_bytenr(eb);
727011e0 599 if (found_start != eb->start) {
efe120a0 600 printk_ratelimited(KERN_INFO "BTRFS: bad tree block start "
193f284d 601 "%llu %llu\n",
c1c9ff7c 602 found_start, eb->start);
f188591e 603 ret = -EIO;
ce9adaa5
CM		 604 goto err;
605 }
2b82032c 606 if (check_tree_block_fsid(root, eb)) {
efe120a0 607 printk_ratelimited(KERN_INFO "BTRFS: bad fsid on block %llu\n",
c1c9ff7c 608 eb->start);
1259ab75
CM		 609 ret = -EIO;
610 goto err;
611 }
ce9adaa5 612 found_level = btrfs_header_level(eb);
1c24c3ce 613 if (found_level >= BTRFS_MAX_LEVEL) {
efe120a0 614 btrfs_info(root->fs_info, "bad tree block level %d",
1c24c3ce
JB		 615 (int)btrfs_header_level(eb));
616 ret = -EIO;
617 goto err;
618 }
ce9adaa5 619
85d4e461
CM		 620 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
621 eb, found_level);
4008c04a 622
ce9adaa5 623 ret = csum_tree_block(root, eb, 1);
a826d6dc 624 if (ret) {
f188591e 625 ret = -EIO;
a826d6dc
JB		 626 goto err;
627 }
628
629 /*
630 * If this is a leaf block and it is corrupt, set the corrupt bit so
631 * that we don't try and read the other copies of this block, just
632 * return -EIO.
633 */
634 if (found_level == 0 && check_leaf(root, eb)) {
635 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
636 ret = -EIO;
637 }
ce9adaa5 638
0b32f4bb
JB		 639 if (!ret)
640 set_extent_buffer_uptodate(eb);
ce9adaa5 641err:
79fb65a1
JB		 642 if (reads_done &&
643 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 644 btree_readahead_hook(root, eb, eb->start, ret);
4bb31e92 645
53b381b3
DW		 646 if (ret) {
647 /*
648 * our io error hook is going to dec the io pages
649 * again, we have to make sure it has something
650 * to decrement
651 */
652 atomic_inc(&eb->io_pages);
0b32f4bb 653 clear_extent_buffer_uptodate(eb);
53b381b3 654 }
0b32f4bb 655 free_extent_buffer(eb);
ce9adaa5 656out:
f188591e 657 return ret;
ce9adaa5
CM		 658}
659
ea466794 660static int btree_io_failed_hook(struct page *page, int failed_mirror)
4bb31e92 661{
4bb31e92
AJ		 662 struct extent_buffer *eb;
663 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
664
4f2de97a 665 eb = (struct extent_buffer *)page->private;
ea466794 666 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
5cf1ab56 667 eb->read_mirror = failed_mirror;
53b381b3 668 atomic_dec(&eb->io_pages);
ea466794 669 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 670 btree_readahead_hook(root, eb, eb->start, -EIO);
4bb31e92
AJ		 671 return -EIO; /* we fixed nothing */
672}
673
ce9adaa5 674static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM		 675{
676 struct end_io_wq *end_io_wq = bio->bi_private;
677 struct btrfs_fs_info *fs_info;
ce9adaa5 678
ce9adaa5 679 fs_info = end_io_wq->info;
ce9adaa5 680 end_io_wq->error = err;
fccb5d86 681 btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL);
d20f7043 682
7b6d91da 683 if (bio->bi_rw & REQ_WRITE) {
53b381b3 684 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA)
fccb5d86
QW		 685 btrfs_queue_work(fs_info->endio_meta_write_workers,
686 &end_io_wq->work);
53b381b3 687 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE)
fccb5d86
QW		 688 btrfs_queue_work(fs_info->endio_freespace_worker,
689 &end_io_wq->work);
53b381b3 690 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
fccb5d86
QW		 691 btrfs_queue_work(fs_info->endio_raid56_workers,
692 &end_io_wq->work);
cad321ad 693 else
fccb5d86
QW		 694 btrfs_queue_work(fs_info->endio_write_workers,
695 &end_io_wq->work);
d20f7043 696 } else {
53b381b3 697 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
fccb5d86
QW		 698 btrfs_queue_work(fs_info->endio_raid56_workers,
699 &end_io_wq->work);
53b381b3 700 else if (end_io_wq->metadata)
fccb5d86
QW		 701 btrfs_queue_work(fs_info->endio_meta_workers,
702 &end_io_wq->work);
d20f7043 703 else
fccb5d86
QW		 704 btrfs_queue_work(fs_info->endio_workers,
705 &end_io_wq->work);
d20f7043 706 }
ce9adaa5
CM		 707}
708
0cb59c99
JB		 709/*
710 * For the metadata arg you want
711 *
712 * 0 - if data
713 * 1 - if normal metadta
714 * 2 - if writing to the free space cache area
53b381b3 715 * 3 - raid parity work
0cb59c99 716 */
22c59948
CM		 717int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
718 int metadata)
0b86a832 719{
ce9adaa5 720 struct end_io_wq *end_io_wq;
ce9adaa5
CM		 721 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
722 if (!end_io_wq)
723 return -ENOMEM;
724
725 end_io_wq->private = bio->bi_private;
726 end_io_wq->end_io = bio->bi_end_io;
22c59948 727 end_io_wq->info = info;
ce9adaa5
CM		 728 end_io_wq->error = 0;
729 end_io_wq->bio = bio;
22c59948 730 end_io_wq->metadata = metadata;
ce9adaa5
CM		 731
732 bio->bi_private = end_io_wq;
733 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 734 return 0;
735}
736
b64a2851 737unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 738{
4854ddd0 739 unsigned long limit = min_t(unsigned long,
5cdc7ad3 740 info->thread_pool_size,
4854ddd0
CM		 741 info->fs_devices->open_devices);
742 return 256 * limit;
743}
0986fe9e 744
d458b054 745static void run_one_async_start(struct btrfs_work *work)
4a69a410 746{
4a69a410 747 struct async_submit_bio *async;
79787eaa 748 int ret;
4a69a410
CM		 749
750 async = container_of(work, struct async_submit_bio, work);
79787eaa
JM		 751 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
752 async->mirror_num, async->bio_flags,
753 async->bio_offset);
754 if (ret)
755 async->error = ret;
4a69a410
CM		 756}
757
d458b054 758static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 759{
760 struct btrfs_fs_info *fs_info;
761 struct async_submit_bio *async;
4854ddd0 762 int limit;
8b712842
CM		 763
764 async = container_of(work, struct async_submit_bio, work);
765 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 766
b64a2851 767 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 768 limit = limit * 2 / 3;
769
66657b31 770 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
b64a2851 771 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 772 wake_up(&fs_info->async_submit_wait);
773
79787eaa
JM		 774 /* If an error occured we just want to clean up the bio and move on */
775 if (async->error) {
776 bio_endio(async->bio, async->error);
777 return;
778 }
779
4a69a410 780 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 781 async->mirror_num, async->bio_flags,
782 async->bio_offset);
4a69a410
CM		 783}
784
d458b054 785static void run_one_async_free(struct btrfs_work *work)
4a69a410
CM		 786{
787 struct async_submit_bio *async;
788
789 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 790 kfree(async);
791}
792
44b8bd7e
CM		 793int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
794 int rw, struct bio *bio, int mirror_num,
c8b97818 795 unsigned long bio_flags,
eaf25d93 796 u64 bio_offset,
4a69a410
CM		 797 extent_submit_bio_hook_t *submit_bio_start,
798 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 799{
800 struct async_submit_bio *async;
801
802 async = kmalloc(sizeof(*async), GFP_NOFS);
803 if (!async)
804 return -ENOMEM;
805
806 async->inode = inode;
807 async->rw = rw;
808 async->bio = bio;
809 async->mirror_num = mirror_num;
4a69a410
CM		 810 async->submit_bio_start = submit_bio_start;
811 async->submit_bio_done = submit_bio_done;
812
5cdc7ad3
QW		 813 btrfs_init_work(&async->work, run_one_async_start,
814 run_one_async_done, run_one_async_free);
4a69a410 815
c8b97818 816 async->bio_flags = bio_flags;
eaf25d93 817 async->bio_offset = bio_offset;
8c8bee1d 818
79787eaa
JM		 819 async->error = 0;
820
cb03c743 821 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 822
7b6d91da 823 if (rw & REQ_SYNC)
5cdc7ad3 824 btrfs_set_work_high_priority(&async->work);
d313d7a3 825
5cdc7ad3 826 btrfs_queue_work(fs_info->workers, &async->work);
9473f16c 827
d397712b 828 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 829 atomic_read(&fs_info->nr_async_submits)) {
830 wait_event(fs_info->async_submit_wait,
831 (atomic_read(&fs_info->nr_async_submits) == 0));
832 }
833
44b8bd7e
CM		 834 return 0;
835}
836
ce3ed71a
CM		 837static int btree_csum_one_bio(struct bio *bio)
838{
839 struct bio_vec *bvec = bio->bi_io_vec;
840 int bio_index = 0;
841 struct btrfs_root *root;
79787eaa 842 int ret = 0;
ce3ed71a
CM		 843
844 WARN_ON(bio->bi_vcnt <= 0);
d397712b 845 while (bio_index < bio->bi_vcnt) {
ce3ed71a 846 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
79787eaa
JM		 847 ret = csum_dirty_buffer(root, bvec->bv_page);
848 if (ret)
849 break;
ce3ed71a
CM		 850 bio_index++;
851 bvec++;
852 }
79787eaa 853 return ret;
ce3ed71a
CM		 854}
855
4a69a410
CM		 856static int __btree_submit_bio_start(struct inode *inode, int rw,
857 struct bio *bio, int mirror_num,
eaf25d93
CM		 858 unsigned long bio_flags,
859 u64 bio_offset)
22c59948 860{
8b712842
CM		 861 /*
862 * when we're called for a write, we're already in the async
5443be45 863 * submission context. Just jump into btrfs_map_bio
8b712842 864 */
79787eaa 865 return btree_csum_one_bio(bio);
4a69a410 866}
22c59948 867
4a69a410 868static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 869 int mirror_num, unsigned long bio_flags,
870 u64 bio_offset)
4a69a410 871{
61891923
SB		 872 int ret;
873
8b712842 874 /*
4a69a410
CM		 875 * when we're called for a write, we're already in the async
876 * submission context. Just jump into btrfs_map_bio
8b712842 877 */
61891923
SB		 878 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
879 if (ret)
880 bio_endio(bio, ret);
881 return ret;
0b86a832
CM		 882}
883
de0022b9
JB		 884static int check_async_write(struct inode *inode, unsigned long bio_flags)
885{
886 if (bio_flags & EXTENT_BIO_TREE_LOG)
887 return 0;
888#ifdef CONFIG_X86
889 if (cpu_has_xmm4_2)
890 return 0;
891#endif
892 return 1;
893}
894
44b8bd7e 895static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 896 int mirror_num, unsigned long bio_flags,
897 u64 bio_offset)
44b8bd7e 898{
de0022b9 899 int async = check_async_write(inode, bio_flags);
cad321ad
CM		 900 int ret;
901
7b6d91da 902 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 903 /*
904 * called for a read, do the setup so that checksum validation
905 * can happen in the async kernel threads
906 */
f3f266ab
CM		 907 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
908 bio, 1);
1d4284bd 909 if (ret)
61891923
SB		 910 goto out_w_error;
911 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
912 mirror_num, 0);
de0022b9
JB		 913 } else if (!async) {
914 ret = btree_csum_one_bio(bio);
915 if (ret)
61891923
SB		 916 goto out_w_error;
917 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
918 mirror_num, 0);
919 } else {
920 /*
921 * kthread helpers are used to submit writes so that
922 * checksumming can happen in parallel across all CPUs
923 */
924 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
925 inode, rw, bio, mirror_num, 0,
926 bio_offset,
927 __btree_submit_bio_start,
928 __btree_submit_bio_done);
44b8bd7e 929 }
d313d7a3 930
61891923
SB		 931 if (ret) {
932out_w_error:
933 bio_endio(bio, ret);
934 }
935 return ret;
44b8bd7e
CM		 936}
937
3dd1462e 938#ifdef CONFIG_MIGRATION
784b4e29 939static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG		 940 struct page *newpage, struct page *page,
941 enum migrate_mode mode)
784b4e29
CM		 942{
943 /*
944 * we can't safely write a btree page from here,
945 * we haven't done the locking hook
946 */
947 if (PageDirty(page))
948 return -EAGAIN;
949 /*
950 * Buffers may be managed in a filesystem specific way.
951 * We must have no buffers or drop them.
952 */
953 if (page_has_private(page) &&
954 !try_to_release_page(page, GFP_KERNEL))
955 return -EAGAIN;
a6bc32b8 956 return migrate_page(mapping, newpage, page, mode);
784b4e29 957}
3dd1462e 958#endif
784b4e29 959
0da5468f
CM		 960
961static int btree_writepages(struct address_space *mapping,
962 struct writeback_control *wbc)
963{
e2d84521
MX		 964 struct btrfs_fs_info *fs_info;
965 int ret;
966
d8d5f3e1 967 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM		 968
969 if (wbc->for_kupdate)
970 return 0;
971
e2d84521 972 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 973 /* this is a bit racy, but that's ok */
e2d84521
MX		 974 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
975 BTRFS_DIRTY_METADATA_THRESH);
976 if (ret < 0)
793955bc 977 return 0;
793955bc 978 }
0b32f4bb 979 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM		 980}
981
b2950863 982static int btree_readpage(struct file *file, struct page *page)
5f39d397 983{
d1310b2e
CM		 984 struct extent_io_tree *tree;
985 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 986 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 987}
22b0ebda 988
70dec807 989static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 990{
98509cfc 991 if (PageWriteback(page) || PageDirty(page))
d397712b 992 return 0;
0c4e538b 993
f7a52a40 994 return try_release_extent_buffer(page);
d98237b3
CM		 995}
996
d47992f8
LC		 997static void btree_invalidatepage(struct page *page, unsigned int offset,
998 unsigned int length)
d98237b3 999{
d1310b2e
CM		 1000 struct extent_io_tree *tree;
1001 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 1002 extent_invalidatepage(tree, page, offset);
1003 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 1004 if (PagePrivate(page)) {
efe120a0
FH		 1005 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1006 "page private not zero on page %llu",
1007 (unsigned long long)page_offset(page));
9ad6b7bc
CM		 1008 ClearPagePrivate(page);
1009 set_page_private(page, 0);
1010 page_cache_release(page);
1011 }
d98237b3
CM		 1012}
1013
0b32f4bb
JB		 1014static int btree_set_page_dirty(struct page *page)
1015{
bb146eb2 1016#ifdef DEBUG
0b32f4bb
JB		 1017 struct extent_buffer *eb;
1018
1019 BUG_ON(!PagePrivate(page));
1020 eb = (struct extent_buffer *)page->private;
1021 BUG_ON(!eb);
1022 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1023 BUG_ON(!atomic_read(&eb->refs));
1024 btrfs_assert_tree_locked(eb);
bb146eb2 1025#endif
0b32f4bb
JB		 1026 return __set_page_dirty_nobuffers(page);
1027}
1028
7f09410b 1029static const struct address_space_operations btree_aops = {
d98237b3 1030 .readpage = btree_readpage,
0da5468f 1031 .writepages = btree_writepages,
5f39d397
CM		 1032 .releasepage = btree_releasepage,
1033 .invalidatepage = btree_invalidatepage,
5a92bc88 1034#ifdef CONFIG_MIGRATION
784b4e29 1035 .migratepage = btree_migratepage,
5a92bc88 1036#endif
0b32f4bb 1037 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM		 1038};
1039
ca7a79ad
CM		 1040int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1041 u64 parent_transid)
090d1875 1042{
5f39d397
CM		 1043 struct extent_buffer *buf = NULL;
1044 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 1045 int ret = 0;
090d1875 1046
db94535d 1047 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 1048 if (!buf)
090d1875 1049 return 0;
d1310b2e 1050 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
bb82ab88 1051 buf, 0, WAIT_NONE, btree_get_extent, 0);
5f39d397 1052 free_extent_buffer(buf);
de428b63 1053 return ret;
090d1875
CM		 1054}
1055
ab0fff03
AJ		 1056int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1057 int mirror_num, struct extent_buffer **eb)
1058{
1059 struct extent_buffer *buf = NULL;
1060 struct inode *btree_inode = root->fs_info->btree_inode;
1061 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1062 int ret;
1063
1064 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1065 if (!buf)
1066 return 0;
1067
1068 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1069
1070 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1071 btree_get_extent, mirror_num);
1072 if (ret) {
1073 free_extent_buffer(buf);
1074 return ret;
1075 }
1076
1077 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1078 free_extent_buffer(buf);
1079 return -EIO;
0b32f4bb 1080 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ		 1081 *eb = buf;
1082 } else {
1083 free_extent_buffer(buf);
1084 }
1085 return 0;
1086}
1087
0999df54
CM		 1088struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1089 u64 bytenr, u32 blocksize)
1090{
f28491e0 1091 return find_extent_buffer(root->fs_info, bytenr);
0999df54
CM		 1092}
1093
1094struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1095 u64 bytenr, u32 blocksize)
1096{
f28491e0 1097 return alloc_extent_buffer(root->fs_info, bytenr, blocksize);
0999df54
CM		 1098}
1099
1100
e02119d5
CM		 1101int btrfs_write_tree_block(struct extent_buffer *buf)
1102{
727011e0 1103 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1104 buf->start + buf->len - 1);
e02119d5
CM		 1105}
1106
1107int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1108{
727011e0 1109 return filemap_fdatawait_range(buf->pages[0]->mapping,
8aa38c31 1110 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1111}
1112
0999df54 1113struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 1114 u32 blocksize, u64 parent_transid)
0999df54
CM		 1115{
1116 struct extent_buffer *buf = NULL;
0999df54
CM		 1117 int ret;
1118
0999df54
CM		 1119 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1120 if (!buf)
1121 return NULL;
0999df54 1122
ca7a79ad 1123 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
0f0fe8f7
FDBM		 1124 if (ret) {
1125 free_extent_buffer(buf);
1126 return NULL;
1127 }
5f39d397 1128 return buf;
ce9adaa5 1129
eb60ceac
CM		 1130}
1131
d5c13f92
JM		 1132void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1133 struct extent_buffer *buf)
ed2ff2cb 1134{
e2d84521
MX		 1135 struct btrfs_fs_info *fs_info = root->fs_info;
1136
55c69072 1137 if (btrfs_header_generation(buf) ==
e2d84521 1138 fs_info->running_transaction->transid) {
b9447ef8 1139 btrfs_assert_tree_locked(buf);
b4ce94de 1140
b9473439 1141 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
e2d84521
MX		 1142 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1143 -buf->len,
1144 fs_info->dirty_metadata_batch);
ed7b63eb
JB		 1145 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1146 btrfs_set_lock_blocking(buf);
1147 clear_extent_buffer_dirty(buf);
1148 }
925baedd 1149 }
5f39d397
CM		 1150}
1151
8257b2dc
MX		 1152static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1153{
1154 struct btrfs_subvolume_writers *writers;
1155 int ret;
1156
1157 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1158 if (!writers)
1159 return ERR_PTR(-ENOMEM);
1160
1161 ret = percpu_counter_init(&writers->counter, 0);
1162 if (ret < 0) {
1163 kfree(writers);
1164 return ERR_PTR(ret);
1165 }
1166
1167 init_waitqueue_head(&writers->wait);
1168 return writers;
1169}
1170
1171static void
1172btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1173{
1174 percpu_counter_destroy(&writers->counter);
1175 kfree(writers);
1176}
1177
143bede5
JM		 1178static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1179 u32 stripesize, struct btrfs_root *root,
1180 struct btrfs_fs_info *fs_info,
1181 u64 objectid)
d97e63b6 1182{
cfaa7295 1183 root->node = NULL;
a28ec197 1184 root->commit_root = NULL;
db94535d
CM		 1185 root->sectorsize = sectorsize;
1186 root->nodesize = nodesize;
1187 root->leafsize = leafsize;
87ee04eb 1188 root->stripesize = stripesize;
123abc88 1189 root->ref_cows = 0;
0b86a832 1190 root->track_dirty = 0;
c71bf099 1191 root->in_radix = 0;
d68fc57b
YZ		 1192 root->orphan_item_inserted = 0;
1193 root->orphan_cleanup_state = 0;
0b86a832 1194
0f7d52f4
CM		 1195 root->objectid = objectid;
1196 root->last_trans = 0;
13a8a7c8 1197 root->highest_objectid = 0;
eb73c1b7 1198 root->nr_delalloc_inodes = 0;
199c2a9c 1199 root->nr_ordered_extents = 0;
58176a96 1200 root->name = NULL;
6bef4d31 1201 root->inode_tree = RB_ROOT;
16cdcec7 1202 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1203 root->block_rsv = NULL;
d68fc57b 1204 root->orphan_block_rsv = NULL;
0b86a832
CM		 1205
1206 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1207 INIT_LIST_HEAD(&root->root_list);
eb73c1b7
MX		 1208 INIT_LIST_HEAD(&root->delalloc_inodes);
1209 INIT_LIST_HEAD(&root->delalloc_root);
199c2a9c
MX		 1210 INIT_LIST_HEAD(&root->ordered_extents);
1211 INIT_LIST_HEAD(&root->ordered_root);
2ab28f32
JB		 1212 INIT_LIST_HEAD(&root->logged_list[0]);
1213 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1214 spin_lock_init(&root->orphan_lock);
5d4f98a2 1215 spin_lock_init(&root->inode_lock);
eb73c1b7 1216 spin_lock_init(&root->delalloc_lock);
199c2a9c 1217 spin_lock_init(&root->ordered_extent_lock);
f0486c68 1218 spin_lock_init(&root->accounting_lock);
2ab28f32
JB		 1219 spin_lock_init(&root->log_extents_lock[0]);
1220 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1221 mutex_init(&root->objectid_mutex);
e02119d5 1222 mutex_init(&root->log_mutex);
31f3d255 1223 mutex_init(&root->ordered_extent_mutex);
573bfb72 1224 mutex_init(&root->delalloc_mutex);
7237f183
YZ		 1225 init_waitqueue_head(&root->log_writer_wait);
1226 init_waitqueue_head(&root->log_commit_wait[0]);
1227 init_waitqueue_head(&root->log_commit_wait[1]);
8b050d35
MX		 1228 INIT_LIST_HEAD(&root->log_ctxs[0]);
1229 INIT_LIST_HEAD(&root->log_ctxs[1]);
7237f183
YZ		 1230 atomic_set(&root->log_commit[0], 0);
1231 atomic_set(&root->log_commit[1], 0);
1232 atomic_set(&root->log_writers, 0);
2ecb7923 1233 atomic_set(&root->log_batch, 0);
8a35d95f 1234 atomic_set(&root->orphan_inodes, 0);
b0feb9d9 1235 atomic_set(&root->refs, 1);
8257b2dc 1236 atomic_set(&root->will_be_snapshoted, 0);
7237f183 1237 root->log_transid = 0;
d1433deb 1238 root->log_transid_committed = -1;
257c62e1 1239 root->last_log_commit = 0;
06ea65a3
JB		 1240 if (fs_info)
1241 extent_io_tree_init(&root->dirty_log_pages,
1242 fs_info->btree_inode->i_mapping);
017e5369 1243
3768f368
CM		 1244 memset(&root->root_key, 0, sizeof(root->root_key));
1245 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1246 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1247 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
06ea65a3
JB		 1248 if (fs_info)
1249 root->defrag_trans_start = fs_info->generation;
1250 else
1251 root->defrag_trans_start = 0;
58176a96 1252 init_completion(&root->kobj_unregister);
6702ed49 1253 root->defrag_running = 0;
4d775673 1254 root->root_key.objectid = objectid;
0ee5dc67 1255 root->anon_dev = 0;
8ea05e3a 1256
5f3ab90a 1257 spin_lock_init(&root->root_item_lock);
3768f368
CM		 1258}
1259
f84a8bd6 1260static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
6f07e42e
AV		 1261{
1262 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1263 if (root)
1264 root->fs_info = fs_info;
1265 return root;
1266}
1267
06ea65a3
JB		 1268#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1269/* Should only be used by the testing infrastructure */
1270struct btrfs_root *btrfs_alloc_dummy_root(void)
1271{
1272 struct btrfs_root *root;
1273
1274 root = btrfs_alloc_root(NULL);
1275 if (!root)
1276 return ERR_PTR(-ENOMEM);
1277 __setup_root(4096, 4096, 4096, 4096, root, NULL, 1);
1278 root->dummy_root = 1;
1279
1280 return root;
1281}
1282#endif
1283
20897f5c
AJ		 1284struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1285 struct btrfs_fs_info *fs_info,
1286 u64 objectid)
1287{
1288 struct extent_buffer *leaf;
1289 struct btrfs_root *tree_root = fs_info->tree_root;
1290 struct btrfs_root *root;
1291 struct btrfs_key key;
1292 int ret = 0;
6463fe58 1293 uuid_le uuid;
20897f5c
AJ		 1294
1295 root = btrfs_alloc_root(fs_info);
1296 if (!root)
1297 return ERR_PTR(-ENOMEM);
1298
1299 __setup_root(tree_root->nodesize, tree_root->leafsize,
1300 tree_root->sectorsize, tree_root->stripesize,
1301 root, fs_info, objectid);
1302 root->root_key.objectid = objectid;
1303 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1304 root->root_key.offset = 0;
1305
1306 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1307 0, objectid, NULL, 0, 0, 0);
1308 if (IS_ERR(leaf)) {
1309 ret = PTR_ERR(leaf);
1dd05682 1310 leaf = NULL;
20897f5c
AJ		 1311 goto fail;
1312 }
1313
20897f5c
AJ		 1314 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1315 btrfs_set_header_bytenr(leaf, leaf->start);
1316 btrfs_set_header_generation(leaf, trans->transid);
1317 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1318 btrfs_set_header_owner(leaf, objectid);
1319 root->node = leaf;
1320
0a4e5586 1321 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
20897f5c
AJ		 1322 BTRFS_FSID_SIZE);
1323 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
b308bc2f 1324 btrfs_header_chunk_tree_uuid(leaf),
20897f5c
AJ		 1325 BTRFS_UUID_SIZE);
1326 btrfs_mark_buffer_dirty(leaf);
1327
1328 root->commit_root = btrfs_root_node(root);
1329 root->track_dirty = 1;
1330
1331
1332 root->root_item.flags = 0;
1333 root->root_item.byte_limit = 0;
1334 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1335 btrfs_set_root_generation(&root->root_item, trans->transid);
1336 btrfs_set_root_level(&root->root_item, 0);
1337 btrfs_set_root_refs(&root->root_item, 1);
1338 btrfs_set_root_used(&root->root_item, leaf->len);
1339 btrfs_set_root_last_snapshot(&root->root_item, 0);
1340 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB		 1341 uuid_le_gen(&uuid);
1342 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ		 1343 root->root_item.drop_level = 0;
1344
1345 key.objectid = objectid;
1346 key.type = BTRFS_ROOT_ITEM_KEY;
1347 key.offset = 0;
1348 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1349 if (ret)
1350 goto fail;
1351
1352 btrfs_tree_unlock(leaf);
1353
1dd05682
TI		 1354 return root;
1355
20897f5c 1356fail:
1dd05682
TI		 1357 if (leaf) {
1358 btrfs_tree_unlock(leaf);
1359 free_extent_buffer(leaf);
1360 }
1361 kfree(root);
20897f5c 1362
1dd05682 1363 return ERR_PTR(ret);
20897f5c
AJ		 1364}
1365
7237f183
YZ		 1366static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1367 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1368{
1369 struct btrfs_root *root;
1370 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1371 struct extent_buffer *leaf;
e02119d5 1372
6f07e42e 1373 root = btrfs_alloc_root(fs_info);
e02119d5 1374 if (!root)
7237f183 1375 return ERR_PTR(-ENOMEM);
e02119d5
CM		 1376
1377 __setup_root(tree_root->nodesize, tree_root->leafsize,
1378 tree_root->sectorsize, tree_root->stripesize,
1379 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1380
1381 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1382 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1383 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
7237f183
YZ		 1384 /*
1385 * log trees do not get reference counted because they go away
1386 * before a real commit is actually done. They do store pointers
1387 * to file data extents, and those reference counts still get
1388 * updated (along with back refs to the log tree).
1389 */
e02119d5
CM		 1390 root->ref_cows = 0;
1391
5d4f98a2 1392 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
66d7e7f0 1393 BTRFS_TREE_LOG_OBJECTID, NULL,
5581a51a 1394 0, 0, 0);
7237f183
YZ		 1395 if (IS_ERR(leaf)) {
1396 kfree(root);
1397 return ERR_CAST(leaf);
1398 }
e02119d5 1399
5d4f98a2
YZ		 1400 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1401 btrfs_set_header_bytenr(leaf, leaf->start);
1402 btrfs_set_header_generation(leaf, trans->transid);
1403 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1404 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1405 root->node = leaf;
e02119d5
CM		 1406
1407 write_extent_buffer(root->node, root->fs_info->fsid,
0a4e5586 1408 btrfs_header_fsid(), BTRFS_FSID_SIZE);
e02119d5
CM		 1409 btrfs_mark_buffer_dirty(root->node);
1410 btrfs_tree_unlock(root->node);
7237f183
YZ		 1411 return root;
1412}
1413
1414int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1415 struct btrfs_fs_info *fs_info)
1416{
1417 struct btrfs_root *log_root;
1418
1419 log_root = alloc_log_tree(trans, fs_info);
1420 if (IS_ERR(log_root))
1421 return PTR_ERR(log_root);
1422 WARN_ON(fs_info->log_root_tree);
1423 fs_info->log_root_tree = log_root;
1424 return 0;
1425}
1426
1427int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1428 struct btrfs_root *root)
1429{
1430 struct btrfs_root *log_root;
1431 struct btrfs_inode_item *inode_item;
1432
1433 log_root = alloc_log_tree(trans, root->fs_info);
1434 if (IS_ERR(log_root))
1435 return PTR_ERR(log_root);
1436
1437 log_root->last_trans = trans->transid;
1438 log_root->root_key.offset = root->root_key.objectid;
1439
1440 inode_item = &log_root->root_item.inode;
3cae210f
QW		 1441 btrfs_set_stack_inode_generation(inode_item, 1);
1442 btrfs_set_stack_inode_size(inode_item, 3);
1443 btrfs_set_stack_inode_nlink(inode_item, 1);
1444 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
1445 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
7237f183 1446
5d4f98a2 1447 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1448
1449 WARN_ON(root->log_root);
1450 root->log_root = log_root;
1451 root->log_transid = 0;
d1433deb 1452 root->log_transid_committed = -1;
257c62e1 1453 root->last_log_commit = 0;
e02119d5
CM		 1454 return 0;
1455}
1456
35a3621b
SB		 1457static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1458 struct btrfs_key *key)
e02119d5
CM		 1459{
1460 struct btrfs_root *root;
1461 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1462 struct btrfs_path *path;
84234f3a 1463 u64 generation;
db94535d 1464 u32 blocksize;
cb517eab 1465 int ret;
0f7d52f4 1466
cb517eab
MX		 1467 path = btrfs_alloc_path();
1468 if (!path)
0f7d52f4 1469 return ERR_PTR(-ENOMEM);
cb517eab
MX		 1470
1471 root = btrfs_alloc_root(fs_info);
1472 if (!root) {
1473 ret = -ENOMEM;
1474 goto alloc_fail;
0f7d52f4
CM		 1475 }
1476
db94535d 1477 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb 1478 tree_root->sectorsize, tree_root->stripesize,
cb517eab 1479 root, fs_info, key->objectid);
0f7d52f4 1480
cb517eab
MX		 1481 ret = btrfs_find_root(tree_root, key, path,
1482 &root->root_item, &root->root_key);
0f7d52f4 1483 if (ret) {
13a8a7c8
YZ		 1484 if (ret > 0)
1485 ret = -ENOENT;
cb517eab 1486 goto find_fail;
0f7d52f4 1487 }
13a8a7c8 1488
84234f3a 1489 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1490 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1491 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1492 blocksize, generation);
cb517eab
MX		 1493 if (!root->node) {
1494 ret = -ENOMEM;
1495 goto find_fail;
1496 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1497 ret = -EIO;
1498 goto read_fail;
416bc658 1499 }
5d4f98a2 1500 root->commit_root = btrfs_root_node(root);
13a8a7c8 1501out:
cb517eab
MX		 1502 btrfs_free_path(path);
1503 return root;
1504
1505read_fail:
1506 free_extent_buffer(root->node);
1507find_fail:
1508 kfree(root);
1509alloc_fail:
1510 root = ERR_PTR(ret);
1511 goto out;
1512}
1513
1514struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1515 struct btrfs_key *location)
1516{
1517 struct btrfs_root *root;
1518
1519 root = btrfs_read_tree_root(tree_root, location);
1520 if (IS_ERR(root))
1521 return root;
1522
1523 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
e02119d5 1524 root->ref_cows = 1;
08fe4db1
LZ		 1525 btrfs_check_and_init_root_item(&root->root_item);
1526 }
13a8a7c8 1527
5eda7b5e
CM		 1528 return root;
1529}
1530
cb517eab
MX		 1531int btrfs_init_fs_root(struct btrfs_root *root)
1532{
1533 int ret;
8257b2dc 1534 struct btrfs_subvolume_writers *writers;
cb517eab
MX		 1535
1536 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1537 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1538 GFP_NOFS);
1539 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1540 ret = -ENOMEM;
1541 goto fail;
1542 }
1543
8257b2dc
MX		 1544 writers = btrfs_alloc_subvolume_writers();
1545 if (IS_ERR(writers)) {
1546 ret = PTR_ERR(writers);
1547 goto fail;
1548 }
1549 root->subv_writers = writers;
1550
cb517eab
MX		 1551 btrfs_init_free_ino_ctl(root);
1552 mutex_init(&root->fs_commit_mutex);
1553 spin_lock_init(&root->cache_lock);
1554 init_waitqueue_head(&root->cache_wait);
1555
1556 ret = get_anon_bdev(&root->anon_dev);
1557 if (ret)
8257b2dc 1558 goto free_writers;
cb517eab 1559 return 0;
8257b2dc
MX		 1560
1561free_writers:
1562 btrfs_free_subvolume_writers(root->subv_writers);
cb517eab
MX		 1563fail:
1564 kfree(root->free_ino_ctl);
1565 kfree(root->free_ino_pinned);
1566 return ret;
1567}
1568
171170c1
ST		 1569static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1570 u64 root_id)
cb517eab
MX		 1571{
1572 struct btrfs_root *root;
1573
1574 spin_lock(&fs_info->fs_roots_radix_lock);
1575 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1576 (unsigned long)root_id);
1577 spin_unlock(&fs_info->fs_roots_radix_lock);
1578 return root;
1579}
1580
1581int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1582 struct btrfs_root *root)
1583{
1584 int ret;
1585
1586 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1587 if (ret)
1588 return ret;
1589
1590 spin_lock(&fs_info->fs_roots_radix_lock);
1591 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1592 (unsigned long)root->root_key.objectid,
1593 root);
1594 if (ret == 0)
1595 root->in_radix = 1;
1596 spin_unlock(&fs_info->fs_roots_radix_lock);
1597 radix_tree_preload_end();
1598
1599 return ret;
1600}
1601
c00869f1
MX		 1602struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1603 struct btrfs_key *location,
1604 bool check_ref)
5eda7b5e
CM		 1605{
1606 struct btrfs_root *root;
1607 int ret;
1608
edbd8d4e
CM		 1609 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1610 return fs_info->tree_root;
1611 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1612 return fs_info->extent_root;
8f18cf13
CM		 1613 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1614 return fs_info->chunk_root;
1615 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1616 return fs_info->dev_root;
0403e47e
YZ		 1617 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1618 return fs_info->csum_root;
bcef60f2
AJ		 1619 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1620 return fs_info->quota_root ? fs_info->quota_root :
1621 ERR_PTR(-ENOENT);
f7a81ea4
SB		 1622 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1623 return fs_info->uuid_root ? fs_info->uuid_root :
1624 ERR_PTR(-ENOENT);
4df27c4d 1625again:
cb517eab 1626 root = btrfs_lookup_fs_root(fs_info, location->objectid);
48475471 1627 if (root) {
c00869f1 1628 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
48475471 1629 return ERR_PTR(-ENOENT);
5eda7b5e 1630 return root;
48475471 1631 }
5eda7b5e 1632
cb517eab 1633 root = btrfs_read_fs_root(fs_info->tree_root, location);
5eda7b5e
CM		 1634 if (IS_ERR(root))
1635 return root;
3394e160 1636
c00869f1 1637 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
cb517eab 1638 ret = -ENOENT;
581bb050 1639 goto fail;
35a30d7c 1640 }
581bb050 1641
cb517eab 1642 ret = btrfs_init_fs_root(root);
ac08aedf
CM		 1643 if (ret)
1644 goto fail;
3394e160 1645
3f870c28
KN		 1646 ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID,
1647 location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL);
d68fc57b
YZ		 1648 if (ret < 0)
1649 goto fail;
1650 if (ret == 0)
1651 root->orphan_item_inserted = 1;
1652
cb517eab 1653 ret = btrfs_insert_fs_root(fs_info, root);
0f7d52f4 1654 if (ret) {
4df27c4d
YZ		 1655 if (ret == -EEXIST) {
1656 free_fs_root(root);
1657 goto again;
1658 }
1659 goto fail;
0f7d52f4 1660 }
edbd8d4e 1661 return root;
4df27c4d
YZ		 1662fail:
1663 free_fs_root(root);
1664 return ERR_PTR(ret);
edbd8d4e
CM		 1665}
1666
04160088
CM		 1667static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1668{
1669 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1670 int ret = 0;
04160088
CM		 1671 struct btrfs_device *device;
1672 struct backing_dev_info *bdi;
b7967db7 1673
1f78160c
XG		 1674 rcu_read_lock();
1675 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1676 if (!device->bdev)
1677 continue;
04160088
CM		 1678 bdi = blk_get_backing_dev_info(device->bdev);
1679 if (bdi && bdi_congested(bdi, bdi_bits)) {
1680 ret = 1;
1681 break;
1682 }
1683 }
1f78160c 1684 rcu_read_unlock();
04160088
CM		 1685 return ret;
1686}
1687
ad081f14
JA		 1688/*
1689 * If this fails, caller must call bdi_destroy() to get rid of the
1690 * bdi again.
1691 */
04160088
CM		 1692static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1693{
ad081f14
JA		 1694 int err;
1695
1696 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1697 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1698 if (err)
1699 return err;
1700
4575c9cc 1701 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1702 bdi->congested_fn = btrfs_congested_fn;
1703 bdi->congested_data = info;
1704 return 0;
1705}
1706
8b712842
CM		 1707/*
1708 * called by the kthread helper functions to finally call the bio end_io
1709 * functions. This is where read checksum verification actually happens
1710 */
d458b054 1711static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1712{
ce9adaa5 1713 struct bio *bio;
8b712842 1714 struct end_io_wq *end_io_wq;
ce9adaa5 1715 int error;
ce9adaa5 1716
8b712842
CM		 1717 end_io_wq = container_of(work, struct end_io_wq, work);
1718 bio = end_io_wq->bio;
ce9adaa5 1719
8b712842
CM		 1720 error = end_io_wq->error;
1721 bio->bi_private = end_io_wq->private;
1722 bio->bi_end_io = end_io_wq->end_io;
1723 kfree(end_io_wq);
8b712842 1724 bio_endio(bio, error);
44b8bd7e
CM		 1725}
1726
a74a4b97
CM		 1727static int cleaner_kthread(void *arg)
1728{
1729 struct btrfs_root *root = arg;
d0278245 1730 int again;
a74a4b97
CM		 1731
1732 do {
d0278245 1733 again = 0;
a74a4b97 1734
d0278245 1735 /* Make the cleaner go to sleep early. */
babbf170 1736 if (btrfs_need_cleaner_sleep(root))
d0278245
MX		 1737 goto sleep;
1738
1739 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1740 goto sleep;
1741
dc7f370c
MX		 1742 /*
1743 * Avoid the problem that we change the status of the fs
1744 * during the above check and trylock.
1745 */
babbf170 1746 if (btrfs_need_cleaner_sleep(root)) {
dc7f370c
MX		 1747 mutex_unlock(&root->fs_info->cleaner_mutex);
1748 goto sleep;
76dda93c 1749 }
a74a4b97 1750
d0278245
MX		 1751 btrfs_run_delayed_iputs(root);
1752 again = btrfs_clean_one_deleted_snapshot(root);
1753 mutex_unlock(&root->fs_info->cleaner_mutex);
1754
1755 /*
05323cd1
MX		 1756 * The defragger has dealt with the R/O remount and umount,
1757 * needn't do anything special here.
d0278245
MX		 1758 */
1759 btrfs_run_defrag_inodes(root->fs_info);
1760sleep:
9d1a2a3a 1761 if (!try_to_freeze() && !again) {
a74a4b97 1762 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1763 if (!kthread_should_stop())
1764 schedule();
a74a4b97
CM		 1765 __set_current_state(TASK_RUNNING);
1766 }
1767 } while (!kthread_should_stop());
1768 return 0;
1769}
1770
1771static int transaction_kthread(void *arg)
1772{
1773 struct btrfs_root *root = arg;
1774 struct btrfs_trans_handle *trans;
1775 struct btrfs_transaction *cur;
8929ecfa 1776 u64 transid;
a74a4b97
CM		 1777 unsigned long now;
1778 unsigned long delay;
914b2007 1779 bool cannot_commit;
a74a4b97
CM		 1780
1781 do {
914b2007 1782 cannot_commit = false;
8b87dc17 1783 delay = HZ * root->fs_info->commit_interval;
a74a4b97
CM		 1784 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1785
a4abeea4 1786 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1787 cur = root->fs_info->running_transaction;
1788 if (!cur) {
a4abeea4 1789 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1790 goto sleep;
1791 }
31153d81 1792
a74a4b97 1793 now = get_seconds();
4a9d8bde 1794 if (cur->state < TRANS_STATE_BLOCKED &&
8b87dc17
DS		 1795 (now < cur->start_time ||
1796 now - cur->start_time < root->fs_info->commit_interval)) {
a4abeea4 1797 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1798 delay = HZ * 5;
1799 goto sleep;
1800 }
8929ecfa 1801 transid = cur->transid;
a4abeea4 1802 spin_unlock(&root->fs_info->trans_lock);
56bec294 1803
79787eaa 1804 /* If the file system is aborted, this will always fail. */
354aa0fb 1805 trans = btrfs_attach_transaction(root);
914b2007 1806 if (IS_ERR(trans)) {
354aa0fb
MX		 1807 if (PTR_ERR(trans) != -ENOENT)
1808 cannot_commit = true;
79787eaa 1809 goto sleep;
914b2007 1810 }
8929ecfa 1811 if (transid == trans->transid) {
79787eaa 1812 btrfs_commit_transaction(trans, root);
8929ecfa
YZ		 1813 } else {
1814 btrfs_end_transaction(trans, root);
1815 }
a74a4b97
CM		 1816sleep:
1817 wake_up_process(root->fs_info->cleaner_kthread);
1818 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1819
4e121c06
JB		 1820 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1821 &root->fs_info->fs_state)))
1822 btrfs_cleanup_transaction(root);
a0acae0e 1823 if (!try_to_freeze()) {
a74a4b97 1824 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa 1825 if (!kthread_should_stop() &&
914b2007
JK		 1826 (!btrfs_transaction_blocked(root->fs_info) ||
1827 cannot_commit))
8929ecfa 1828 schedule_timeout(delay);
a74a4b97
CM		 1829 __set_current_state(TASK_RUNNING);
1830 }
1831 } while (!kthread_should_stop());
1832 return 0;
1833}
1834
af31f5e5
CM		 1835/*
1836 * this will find the highest generation in the array of
1837 * root backups. The index of the highest array is returned,
1838 * or -1 if we can't find anything.
1839 *
1840 * We check to make sure the array is valid by comparing the
1841 * generation of the latest root in the array with the generation
1842 * in the super block. If they don't match we pitch it.
1843 */
1844static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1845{
1846 u64 cur;
1847 int newest_index = -1;
1848 struct btrfs_root_backup *root_backup;
1849 int i;
1850
1851 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1852 root_backup = info->super_copy->super_roots + i;
1853 cur = btrfs_backup_tree_root_gen(root_backup);
1854 if (cur == newest_gen)
1855 newest_index = i;
1856 }
1857
1858 /* check to see if we actually wrapped around */
1859 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1860 root_backup = info->super_copy->super_roots;
1861 cur = btrfs_backup_tree_root_gen(root_backup);
1862 if (cur == newest_gen)
1863 newest_index = 0;
1864 }
1865 return newest_index;
1866}
1867
1868
1869/*
1870 * find the oldest backup so we know where to store new entries
1871 * in the backup array. This will set the backup_root_index
1872 * field in the fs_info struct
1873 */
1874static void find_oldest_super_backup(struct btrfs_fs_info *info,
1875 u64 newest_gen)
1876{
1877 int newest_index = -1;
1878
1879 newest_index = find_newest_super_backup(info, newest_gen);
1880 /* if there was garbage in there, just move along */
1881 if (newest_index == -1) {
1882 info->backup_root_index = 0;
1883 } else {
1884 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1885 }
1886}
1887
1888/*
1889 * copy all the root pointers into the super backup array.
1890 * this will bump the backup pointer by one when it is
1891 * done
1892 */
1893static void backup_super_roots(struct btrfs_fs_info *info)
1894{
1895 int next_backup;
1896 struct btrfs_root_backup *root_backup;
1897 int last_backup;
1898
1899 next_backup = info->backup_root_index;
1900 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1901 BTRFS_NUM_BACKUP_ROOTS;
1902
1903 /*
1904 * just overwrite the last backup if we're at the same generation
1905 * this happens only at umount
1906 */
1907 root_backup = info->super_for_commit->super_roots + last_backup;
1908 if (btrfs_backup_tree_root_gen(root_backup) ==
1909 btrfs_header_generation(info->tree_root->node))
1910 next_backup = last_backup;
1911
1912 root_backup = info->super_for_commit->super_roots + next_backup;
1913
1914 /*
1915 * make sure all of our padding and empty slots get zero filled
1916 * regardless of which ones we use today
1917 */
1918 memset(root_backup, 0, sizeof(*root_backup));
1919
1920 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1921
1922 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1923 btrfs_set_backup_tree_root_gen(root_backup,
1924 btrfs_header_generation(info->tree_root->node));
1925
1926 btrfs_set_backup_tree_root_level(root_backup,
1927 btrfs_header_level(info->tree_root->node));
1928
1929 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1930 btrfs_set_backup_chunk_root_gen(root_backup,
1931 btrfs_header_generation(info->chunk_root->node));
1932 btrfs_set_backup_chunk_root_level(root_backup,
1933 btrfs_header_level(info->chunk_root->node));
1934
1935 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1936 btrfs_set_backup_extent_root_gen(root_backup,
1937 btrfs_header_generation(info->extent_root->node));
1938 btrfs_set_backup_extent_root_level(root_backup,
1939 btrfs_header_level(info->extent_root->node));
1940
7c7e82a7
CM		 1941 /*
1942 * we might commit during log recovery, which happens before we set
1943 * the fs_root. Make sure it is valid before we fill it in.
1944 */
1945 if (info->fs_root && info->fs_root->node) {
1946 btrfs_set_backup_fs_root(root_backup,
1947 info->fs_root->node->start);
1948 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 1949 btrfs_header_generation(info->fs_root->node));
7c7e82a7 1950 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 1951 btrfs_header_level(info->fs_root->node));
7c7e82a7 1952 }
af31f5e5
CM		 1953
1954 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1955 btrfs_set_backup_dev_root_gen(root_backup,
1956 btrfs_header_generation(info->dev_root->node));
1957 btrfs_set_backup_dev_root_level(root_backup,
1958 btrfs_header_level(info->dev_root->node));
1959
1960 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1961 btrfs_set_backup_csum_root_gen(root_backup,
1962 btrfs_header_generation(info->csum_root->node));
1963 btrfs_set_backup_csum_root_level(root_backup,
1964 btrfs_header_level(info->csum_root->node));
1965
1966 btrfs_set_backup_total_bytes(root_backup,
1967 btrfs_super_total_bytes(info->super_copy));
1968 btrfs_set_backup_bytes_used(root_backup,
1969 btrfs_super_bytes_used(info->super_copy));
1970 btrfs_set_backup_num_devices(root_backup,
1971 btrfs_super_num_devices(info->super_copy));
1972
1973 /*
1974 * if we don't copy this out to the super_copy, it won't get remembered
1975 * for the next commit
1976 */
1977 memcpy(&info->super_copy->super_roots,
1978 &info->super_for_commit->super_roots,
1979 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1980}
1981
1982/*
1983 * this copies info out of the root backup array and back into
1984 * the in-memory super block. It is meant to help iterate through
1985 * the array, so you send it the number of backups you've already
1986 * tried and the last backup index you used.
1987 *
1988 * this returns -1 when it has tried all the backups
1989 */
1990static noinline int next_root_backup(struct btrfs_fs_info *info,
1991 struct btrfs_super_block *super,
1992 int *num_backups_tried, int *backup_index)
1993{
1994 struct btrfs_root_backup *root_backup;
1995 int newest = *backup_index;
1996
1997 if (*num_backups_tried == 0) {
1998 u64 gen = btrfs_super_generation(super);
1999
2000 newest = find_newest_super_backup(info, gen);
2001 if (newest == -1)
2002 return -1;
2003
2004 *backup_index = newest;
2005 *num_backups_tried = 1;
2006 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2007 /* we've tried all the backups, all done */
2008 return -1;
2009 } else {
2010 /* jump to the next oldest backup */
2011 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2012 BTRFS_NUM_BACKUP_ROOTS;
2013 *backup_index = newest;
2014 *num_backups_tried += 1;
2015 }
2016 root_backup = super->super_roots + newest;
2017
2018 btrfs_set_super_generation(super,
2019 btrfs_backup_tree_root_gen(root_backup));
2020 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2021 btrfs_set_super_root_level(super,
2022 btrfs_backup_tree_root_level(root_backup));
2023 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2024
2025 /*
2026 * fixme: the total bytes and num_devices need to match or we should
2027 * need a fsck
2028 */
2029 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2030 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2031 return 0;
2032}
2033
7abadb64
LB		 2034/* helper to cleanup workers */
2035static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2036{
dc6e3209 2037 btrfs_destroy_workqueue(fs_info->fixup_workers);
afe3d242 2038 btrfs_destroy_workqueue(fs_info->delalloc_workers);
5cdc7ad3 2039 btrfs_destroy_workqueue(fs_info->workers);
fccb5d86
QW		 2040 btrfs_destroy_workqueue(fs_info->endio_workers);
2041 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2042 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
d05a33ac 2043 btrfs_destroy_workqueue(fs_info->rmw_workers);
fccb5d86
QW		 2044 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2045 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2046 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
a8c93d4e 2047 btrfs_destroy_workqueue(fs_info->submit_workers);
5b3bc44e 2048 btrfs_destroy_workqueue(fs_info->delayed_workers);
e66f0bb1 2049 btrfs_destroy_workqueue(fs_info->caching_workers);
736cfa15 2050 btrfs_destroy_workqueue(fs_info->readahead_workers);
a44903ab 2051 btrfs_destroy_workqueue(fs_info->flush_workers);
fc97fab0 2052 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
7abadb64
LB		 2053}
2054
2e9f5954
R		 2055static void free_root_extent_buffers(struct btrfs_root *root)
2056{
2057 if (root) {
2058 free_extent_buffer(root->node);
2059 free_extent_buffer(root->commit_root);
2060 root->node = NULL;
2061 root->commit_root = NULL;
2062 }
2063}
2064
af31f5e5
CM		 2065/* helper to cleanup tree roots */
2066static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2067{
2e9f5954 2068 free_root_extent_buffers(info->tree_root);
655b09fe 2069
2e9f5954
R		 2070 free_root_extent_buffers(info->dev_root);
2071 free_root_extent_buffers(info->extent_root);
2072 free_root_extent_buffers(info->csum_root);
2073 free_root_extent_buffers(info->quota_root);
2074 free_root_extent_buffers(info->uuid_root);
2075 if (chunk_root)
2076 free_root_extent_buffers(info->chunk_root);
af31f5e5
CM		 2077}
2078
171f6537
JB		 2079static void del_fs_roots(struct btrfs_fs_info *fs_info)
2080{
2081 int ret;
2082 struct btrfs_root *gang[8];
2083 int i;
2084
2085 while (!list_empty(&fs_info->dead_roots)) {
2086 gang[0] = list_entry(fs_info->dead_roots.next,
2087 struct btrfs_root, root_list);
2088 list_del(&gang[0]->root_list);
2089
2090 if (gang[0]->in_radix) {
cb517eab 2091 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
171f6537
JB		 2092 } else {
2093 free_extent_buffer(gang[0]->node);
2094 free_extent_buffer(gang[0]->commit_root);
b0feb9d9 2095 btrfs_put_fs_root(gang[0]);
171f6537
JB		 2096 }
2097 }
2098
2099 while (1) {
2100 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2101 (void **)gang, 0,
2102 ARRAY_SIZE(gang));
2103 if (!ret)
2104 break;
2105 for (i = 0; i < ret; i++)
cb517eab 2106 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
171f6537 2107 }
1a4319cc
LB		 2108
2109 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2110 btrfs_free_log_root_tree(NULL, fs_info);
2111 btrfs_destroy_pinned_extent(fs_info->tree_root,
2112 fs_info->pinned_extents);
2113 }
171f6537 2114}
af31f5e5 2115
ad2b2c80
AV		 2116int open_ctree(struct super_block *sb,
2117 struct btrfs_fs_devices *fs_devices,
2118 char *options)
2e635a27 2119{
db94535d
CM		 2120 u32 sectorsize;
2121 u32 nodesize;
2122 u32 leafsize;
2123 u32 blocksize;
87ee04eb 2124 u32 stripesize;
84234f3a 2125 u64 generation;
f2b636e8 2126 u64 features;
3de4586c 2127 struct btrfs_key location;
a061fc8d 2128 struct buffer_head *bh;
4d34b278 2129 struct btrfs_super_block *disk_super;
815745cf 2130 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2131 struct btrfs_root *tree_root;
4d34b278
ID		 2132 struct btrfs_root *extent_root;
2133 struct btrfs_root *csum_root;
2134 struct btrfs_root *chunk_root;
2135 struct btrfs_root *dev_root;
bcef60f2 2136 struct btrfs_root *quota_root;
f7a81ea4 2137 struct btrfs_root *uuid_root;
e02119d5 2138 struct btrfs_root *log_tree_root;
eb60ceac 2139 int ret;
e58ca020 2140 int err = -EINVAL;
af31f5e5
CM		 2141 int num_backups_tried = 0;
2142 int backup_index = 0;
5cdc7ad3
QW		 2143 int max_active;
2144 int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
70f80175
SB		 2145 bool create_uuid_tree;
2146 bool check_uuid_tree;
4543df7e 2147
f84a8bd6 2148 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
6f07e42e 2149 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
cb517eab 2150 if (!tree_root || !chunk_root) {
39279cc3
CM		 2151 err = -ENOMEM;
2152 goto fail;
2153 }
76dda93c
YZ		 2154
2155 ret = init_srcu_struct(&fs_info->subvol_srcu);
2156 if (ret) {
2157 err = ret;
2158 goto fail;
2159 }
2160
2161 ret = setup_bdi(fs_info, &fs_info->bdi);
2162 if (ret) {
2163 err = ret;
2164 goto fail_srcu;
2165 }
2166
e2d84521
MX		 2167 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2168 if (ret) {
2169 err = ret;
2170 goto fail_bdi;
2171 }
2172 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2173 (1 + ilog2(nr_cpu_ids));
2174
963d678b
MX		 2175 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2176 if (ret) {
2177 err = ret;
2178 goto fail_dirty_metadata_bytes;
2179 }
2180
c404e0dc
MX		 2181 ret = percpu_counter_init(&fs_info->bio_counter, 0);
2182 if (ret) {
2183 err = ret;
2184 goto fail_delalloc_bytes;
2185 }
2186
76dda93c
YZ		 2187 fs_info->btree_inode = new_inode(sb);
2188 if (!fs_info->btree_inode) {
2189 err = -ENOMEM;
c404e0dc 2190 goto fail_bio_counter;
76dda93c
YZ		 2191 }
2192
a6591715 2193 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2194
76dda93c 2195 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
f28491e0 2196 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
8fd17795 2197 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2198 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2199 INIT_LIST_HEAD(&fs_info->delayed_iputs);
eb73c1b7 2200 INIT_LIST_HEAD(&fs_info->delalloc_roots);
11833d66 2201 INIT_LIST_HEAD(&fs_info->caching_block_groups);
eb73c1b7 2202 spin_lock_init(&fs_info->delalloc_root_lock);
a4abeea4 2203 spin_lock_init(&fs_info->trans_lock);
76dda93c 2204 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2205 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2206 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 2207 spin_lock_init(&fs_info->free_chunk_lock);
f29021b2 2208 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2209 spin_lock_init(&fs_info->super_lock);
f28491e0 2210 spin_lock_init(&fs_info->buffer_lock);
f29021b2 2211 rwlock_init(&fs_info->tree_mod_log_lock);
7585717f 2212 mutex_init(&fs_info->reloc_mutex);
573bfb72 2213 mutex_init(&fs_info->delalloc_root_mutex);
de98ced9 2214 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2215
58176a96 2216 init_completion(&fs_info->kobj_unregister);
0b86a832 2217 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2218 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2219 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
0b86a832 2220 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX		 2221 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2222 BTRFS_BLOCK_RSV_GLOBAL);
2223 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2224 BTRFS_BLOCK_RSV_DELALLOC);
2225 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2226 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2227 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2228 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2229 BTRFS_BLOCK_RSV_DELOPS);
cb03c743 2230 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 2231 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 2232 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 2233 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 2234 atomic_set(&fs_info->defrag_running, 0);
fc36ed7e 2235 atomic64_set(&fs_info->tree_mod_seq, 0);
e20d96d6 2236 fs_info->sb = sb;
6f568d35 2237 fs_info->max_inline = 8192 * 1024;
9ed74f2d 2238 fs_info->metadata_ratio = 0;
4cb5300b 2239 fs_info->defrag_inodes = RB_ROOT;
2bf64758 2240 fs_info->free_chunk_space = 0;
f29021b2 2241 fs_info->tree_mod_log = RB_ROOT;
8b87dc17 2242 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
0a2b2a84 2243 fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
90519d66
AJ		 2244 /* readahead state */
2245 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2246 spin_lock_init(&fs_info->reada_lock);
c8b97818 2247
b34b086c
CM		 2248 fs_info->thread_pool_size = min_t(unsigned long,
2249 num_online_cpus() + 2, 8);
0afbaf8c 2250
199c2a9c
MX		 2251 INIT_LIST_HEAD(&fs_info->ordered_roots);
2252 spin_lock_init(&fs_info->ordered_root_lock);
16cdcec7
MX		 2253 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2254 GFP_NOFS);
2255 if (!fs_info->delayed_root) {
2256 err = -ENOMEM;
2257 goto fail_iput;
2258 }
2259 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2260
a2de733c
AJ		 2261 mutex_init(&fs_info->scrub_lock);
2262 atomic_set(&fs_info->scrubs_running, 0);
2263 atomic_set(&fs_info->scrub_pause_req, 0);
2264 atomic_set(&fs_info->scrubs_paused, 0);
2265 atomic_set(&fs_info->scrub_cancel_req, 0);
c404e0dc 2266 init_waitqueue_head(&fs_info->replace_wait);
a2de733c 2267 init_waitqueue_head(&fs_info->scrub_pause_wait);
a2de733c 2268 fs_info->scrub_workers_refcnt = 0;
21adbd5c
SB		 2269#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2270 fs_info->check_integrity_print_mask = 0;
2271#endif
a2de733c 2272
c9e9f97b
ID		 2273 spin_lock_init(&fs_info->balance_lock);
2274 mutex_init(&fs_info->balance_mutex);
837d5b6e
ID		 2275 atomic_set(&fs_info->balance_running, 0);
2276 atomic_set(&fs_info->balance_pause_req, 0);
a7e99c69 2277 atomic_set(&fs_info->balance_cancel_req, 0);
c9e9f97b 2278 fs_info->balance_ctl = NULL;
837d5b6e 2279 init_waitqueue_head(&fs_info->balance_wait_q);
a2de733c 2280
a061fc8d
CM		 2281 sb->s_blocksize = 4096;
2282 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 2283 sb->s_bdi = &fs_info->bdi;
a061fc8d 2284
76dda93c 2285 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
bfe86848 2286 set_nlink(fs_info->btree_inode, 1);
0afbaf8c
CM		 2287 /*
2288 * we set the i_size on the btree inode to the max possible int.
2289 * the real end of the address space is determined by all of
2290 * the devices in the system
2291 */
2292 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 2293 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 2294 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2295
5d4f98a2 2296 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 2297 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 2298 fs_info->btree_inode->i_mapping);
0b32f4bb 2299 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
a8067e02 2300 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM		 2301
2302 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 2303
76dda93c
YZ		 2304 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2305 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2306 sizeof(struct btrfs_key));
72ac3c0d
JB		 2307 set_bit(BTRFS_INODE_DUMMY,
2308 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
778ba82b 2309 btrfs_insert_inode_hash(fs_info->btree_inode);
76dda93c 2310
0f9dd46c 2311 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2312 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2313 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2314
11833d66 2315 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 2316 fs_info->btree_inode->i_mapping);
11833d66 2317 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 2318 fs_info->btree_inode->i_mapping);
11833d66 2319 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 2320 fs_info->do_barriers = 1;
e18e4809 2321
39279cc3 2322
5a3f23d5 2323 mutex_init(&fs_info->ordered_operations_mutex);
9ffba8cd 2324 mutex_init(&fs_info->ordered_extent_flush_mutex);
e02119d5 2325 mutex_init(&fs_info->tree_log_mutex);
925baedd 2326 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 2327 mutex_init(&fs_info->transaction_kthread_mutex);
2328 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2329 mutex_init(&fs_info->volume_mutex);
276e680d 2330 init_rwsem(&fs_info->extent_commit_sem);
c71bf099 2331 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2332 init_rwsem(&fs_info->subvol_sem);
803b2f54 2333 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
e922e087
SB		 2334 fs_info->dev_replace.lock_owner = 0;
2335 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2336 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2337 mutex_init(&fs_info->dev_replace.lock_management_lock);
2338 mutex_init(&fs_info->dev_replace.lock);
fa9c0d79 2339
416ac51d 2340 spin_lock_init(&fs_info->qgroup_lock);
f2f6ed3d 2341 mutex_init(&fs_info->qgroup_ioctl_lock);
416ac51d
AJ		 2342 fs_info->qgroup_tree = RB_ROOT;
2343 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2344 fs_info->qgroup_seq = 1;
2345 fs_info->quota_enabled = 0;
2346 fs_info->pending_quota_state = 0;
1e8f9158 2347 fs_info->qgroup_ulist = NULL;
2f232036 2348 mutex_init(&fs_info->qgroup_rescan_lock);
416ac51d 2349
fa9c0d79
CM		 2350 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2351 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2352
e6dcd2dc 2353 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2354 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2355 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2356 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2357
53b381b3
DW		 2358 ret = btrfs_alloc_stripe_hash_table(fs_info);
2359 if (ret) {
83c8266a 2360 err = ret;
53b381b3
DW		 2361 goto fail_alloc;
2362 }
2363
0b86a832 2364 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 2365 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2366
3c4bb26b 2367 invalidate_bdev(fs_devices->latest_bdev);
1104a885
DS		 2368
2369 /*
2370 * Read super block and check the signature bytes only
2371 */
a512bbf8 2372 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY		 2373 if (!bh) {
2374 err = -EINVAL;
16cdcec7 2375 goto fail_alloc;
20b45077 2376 }
39279cc3 2377
1104a885
DS		 2378 /*
2379 * We want to check superblock checksum, the type is stored inside.
2380 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2381 */
2382 if (btrfs_check_super_csum(bh->b_data)) {
efe120a0 2383 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
1104a885
DS		 2384 err = -EINVAL;
2385 goto fail_alloc;
2386 }
2387
2388 /*
2389 * super_copy is zeroed at allocation time and we never touch the
2390 * following bytes up to INFO_SIZE, the checksum is calculated from
2391 * the whole block of INFO_SIZE
2392 */
6c41761f
DS		 2393 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2394 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2395 sizeof(*fs_info->super_for_commit));
a061fc8d 2396 brelse(bh);
5f39d397 2397
6c41761f 2398 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2399
1104a885
DS		 2400 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2401 if (ret) {
efe120a0 2402 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
1104a885
DS		 2403 err = -EINVAL;
2404 goto fail_alloc;
2405 }
2406
6c41761f 2407 disk_super = fs_info->super_copy;
0f7d52f4 2408 if (!btrfs_super_root(disk_super))
16cdcec7 2409 goto fail_alloc;
0f7d52f4 2410
acce952b 2411 /* check FS state, whether FS is broken. */
87533c47
MX		 2412 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2413 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2414
af31f5e5
CM		 2415 /*
2416 * run through our array of backup supers and setup
2417 * our ring pointer to the oldest one
2418 */
2419 generation = btrfs_super_generation(disk_super);
2420 find_oldest_super_backup(fs_info, generation);
2421
75e7cb7f
LB		 2422 /*
2423 * In the long term, we'll store the compression type in the super
2424 * block, and it'll be used for per file compression control.
2425 */
2426 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2427
2b82032c
YZ		 2428 ret = btrfs_parse_options(tree_root, options);
2429 if (ret) {
2430 err = ret;
16cdcec7 2431 goto fail_alloc;
2b82032c 2432 }
dfe25020 2433
f2b636e8
JB		 2434 features = btrfs_super_incompat_flags(disk_super) &
2435 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2436 if (features) {
2437 printk(KERN_ERR "BTRFS: couldn't mount because of "
2438 "unsupported optional features (%Lx).\n",
c1c9ff7c 2439 features);
f2b636e8 2440 err = -EINVAL;
16cdcec7 2441 goto fail_alloc;
f2b636e8
JB		 2442 }
2443
727011e0
CM		 2444 if (btrfs_super_leafsize(disk_super) !=
2445 btrfs_super_nodesize(disk_super)) {
2446 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2447 "blocksizes don't match. node %d leaf %d\n",
2448 btrfs_super_nodesize(disk_super),
2449 btrfs_super_leafsize(disk_super));
2450 err = -EINVAL;
2451 goto fail_alloc;
2452 }
2453 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2454 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2455 "blocksize (%d) was too large\n",
2456 btrfs_super_leafsize(disk_super));
2457 err = -EINVAL;
2458 goto fail_alloc;
2459 }
2460
5d4f98a2 2461 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2462 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
69e380d1 2463 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2464 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
727011e0 2465
3173a18f 2466 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
efe120a0 2467 printk(KERN_ERR "BTRFS: has skinny extents\n");
3173a18f 2468
727011e0
CM		 2469 /*
2470 * flag our filesystem as having big metadata blocks if
2471 * they are bigger than the page size
2472 */
2473 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2474 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
efe120a0 2475 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
727011e0
CM		 2476 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2477 }
2478
bc3f116f
CM		 2479 nodesize = btrfs_super_nodesize(disk_super);
2480 leafsize = btrfs_super_leafsize(disk_super);
2481 sectorsize = btrfs_super_sectorsize(disk_super);
2482 stripesize = btrfs_super_stripesize(disk_super);
e2d84521 2483 fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids));
963d678b 2484 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f
CM		 2485
2486 /*
2487 * mixed block groups end up with duplicate but slightly offset
2488 * extent buffers for the same range. It leads to corruptions
2489 */
2490 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2491 (sectorsize != leafsize)) {
efe120a0 2492 printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes "
bc3f116f
CM		 2493 "are not allowed for mixed block groups on %s\n",
2494 sb->s_id);
2495 goto fail_alloc;
2496 }
2497
ceda0864
MX		 2498 /*
2499 * Needn't use the lock because there is no other task which will
2500 * update the flag.
2501 */
a6fa6fae 2502 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2503
f2b636e8
JB		 2504 features = btrfs_super_compat_ro_flags(disk_super) &
2505 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2506 if (!(sb->s_flags & MS_RDONLY) && features) {
2507 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2508 "unsupported option features (%Lx).\n",
c1c9ff7c 2509 features);
f2b636e8 2510 err = -EINVAL;
16cdcec7 2511 goto fail_alloc;
f2b636e8 2512 }
61d92c32 2513
5cdc7ad3 2514 max_active = fs_info->thread_pool_size;
61d92c32 2515
5cdc7ad3
QW		 2516 fs_info->workers =
2517 btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2518 max_active, 16);
c8b97818 2519
afe3d242
QW		 2520 fs_info->delalloc_workers =
2521 btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
771ed689 2522
a44903ab
QW		 2523 fs_info->flush_workers =
2524 btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
61b49440 2525
e66f0bb1
QW		 2526 fs_info->caching_workers =
2527 btrfs_alloc_workqueue("cache", flags, max_active, 0);
bab39bf9 2528
a8c93d4e
QW		 2529 /*
2530 * a higher idle thresh on the submit workers makes it much more
61b49440
CM		 2531 * likely that bios will be send down in a sane order to the
2532 * devices
2533 */
a8c93d4e
QW		 2534 fs_info->submit_workers =
2535 btrfs_alloc_workqueue("submit", flags,
2536 min_t(u64, fs_devices->num_devices,
2537 max_active), 64);
53863232 2538
dc6e3209
QW		 2539 fs_info->fixup_workers =
2540 btrfs_alloc_workqueue("fixup", flags, 1, 0);
fccb5d86
QW		 2541
2542 /*
2543 * endios are largely parallel and should have a very
2544 * low idle thresh
2545 */
2546 fs_info->endio_workers =
2547 btrfs_alloc_workqueue("endio", flags, max_active, 4);
2548 fs_info->endio_meta_workers =
2549 btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2550 fs_info->endio_meta_write_workers =
2551 btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2552 fs_info->endio_raid56_workers =
2553 btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
d05a33ac
QW		 2554 fs_info->rmw_workers =
2555 btrfs_alloc_workqueue("rmw", flags, max_active, 2);
fccb5d86
QW		 2556 fs_info->endio_write_workers =
2557 btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2558 fs_info->endio_freespace_worker =
2559 btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
5b3bc44e
QW		 2560 fs_info->delayed_workers =
2561 btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
736cfa15
QW		 2562 fs_info->readahead_workers =
2563 btrfs_alloc_workqueue("readahead", flags, max_active, 2);
fc97fab0
QW		 2564 fs_info->qgroup_rescan_workers =
2565 btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
61b49440 2566
a8c93d4e 2567 if (!(fs_info->workers && fs_info->delalloc_workers &&
fccb5d86
QW		 2568 fs_info->submit_workers && fs_info->flush_workers &&
2569 fs_info->endio_workers && fs_info->endio_meta_workers &&
2570 fs_info->endio_meta_write_workers &&
2571 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
e66f0bb1 2572 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
dc6e3209 2573 fs_info->caching_workers && fs_info->readahead_workers &&
fc97fab0
QW		 2574 fs_info->fixup_workers && fs_info->delayed_workers &&
2575 fs_info->qgroup_rescan_workers)) {
5cdc7ad3
QW		 2576 err = -ENOMEM;
2577 goto fail_sb_buffer;
2578 }
4543df7e 2579
4575c9cc 2580 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 2581 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2582 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 2583
db94535d
CM		 2584 tree_root->nodesize = nodesize;
2585 tree_root->leafsize = leafsize;
2586 tree_root->sectorsize = sectorsize;
87ee04eb 2587 tree_root->stripesize = stripesize;
a061fc8d
CM		 2588
2589 sb->s_blocksize = sectorsize;
2590 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 2591
3cae210f 2592 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
efe120a0 2593 printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 2594 goto fail_sb_buffer;
2595 }
19c00ddc 2596
8d082fb7 2597 if (sectorsize != PAGE_SIZE) {
efe120a0 2598 printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) "
8d082fb7 2599 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
941b2ddf
KM		 2600 goto fail_sb_buffer;
2601 }
2602
925baedd 2603 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 2604 ret = btrfs_read_sys_array(tree_root);
925baedd 2605 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2606 if (ret) {
efe120a0 2607 printk(KERN_WARNING "BTRFS: failed to read the system "
d397712b 2608 "array on %s\n", sb->s_id);
5d4f98a2 2609 goto fail_sb_buffer;
84eed90f 2610 }
0b86a832
CM		 2611
2612 blocksize = btrfs_level_size(tree_root,
2613 btrfs_super_chunk_root_level(disk_super));
84234f3a 2614 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 2615
2616 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2617 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2618
2619 chunk_root->node = read_tree_block(chunk_root,
2620 btrfs_super_chunk_root(disk_super),
84234f3a 2621 blocksize, generation);
416bc658
JB		 2622 if (!chunk_root->node ||
2623 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
efe120a0 2624 printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n",
83121942 2625 sb->s_id);
af31f5e5 2626 goto fail_tree_roots;
83121942 2627 }
5d4f98a2
YZ		 2628 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2629 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2630
e17cade2 2631 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
b308bc2f 2632 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
e17cade2 2633
0b86a832 2634 ret = btrfs_read_chunk_tree(chunk_root);
2b82032c 2635 if (ret) {
efe120a0 2636 printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n",
d397712b 2637 sb->s_id);
af31f5e5 2638 goto fail_tree_roots;
2b82032c 2639 }
0b86a832 2640
8dabb742
SB		 2641 /*
2642 * keep the device that is marked to be the target device for the
2643 * dev_replace procedure
2644 */
2645 btrfs_close_extra_devices(fs_info, fs_devices, 0);
dfe25020 2646
a6b0d5c8 2647 if (!fs_devices->latest_bdev) {
efe120a0 2648 printk(KERN_CRIT "BTRFS: failed to read devices on %s\n",
a6b0d5c8
CM		 2649 sb->s_id);
2650 goto fail_tree_roots;
2651 }
2652
af31f5e5 2653retry_root_backup:
db94535d
CM		 2654 blocksize = btrfs_level_size(tree_root,
2655 btrfs_super_root_level(disk_super));
84234f3a 2656 generation = btrfs_super_generation(disk_super);
0b86a832 2657
e20d96d6 2658 tree_root->node = read_tree_block(tree_root,
db94535d 2659 btrfs_super_root(disk_super),
84234f3a 2660 blocksize, generation);
af31f5e5
CM		 2661 if (!tree_root->node ||
2662 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
efe120a0 2663 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
83121942 2664 sb->s_id);
af31f5e5
CM		 2665
2666 goto recovery_tree_root;
83121942 2667 }
af31f5e5 2668
5d4f98a2
YZ		 2669 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2670 tree_root->commit_root = btrfs_root_node(tree_root);
69e9c6c6 2671 btrfs_set_root_refs(&tree_root->root_item, 1);
db94535d 2672
cb517eab
MX		 2673 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2674 location.type = BTRFS_ROOT_ITEM_KEY;
2675 location.offset = 0;
2676
2677 extent_root = btrfs_read_tree_root(tree_root, &location);
2678 if (IS_ERR(extent_root)) {
2679 ret = PTR_ERR(extent_root);
af31f5e5 2680 goto recovery_tree_root;
cb517eab 2681 }
0b86a832 2682 extent_root->track_dirty = 1;
cb517eab 2683 fs_info->extent_root = extent_root;
0b86a832 2684
cb517eab
MX		 2685 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2686 dev_root = btrfs_read_tree_root(tree_root, &location);
2687 if (IS_ERR(dev_root)) {
2688 ret = PTR_ERR(dev_root);
af31f5e5 2689 goto recovery_tree_root;
cb517eab 2690 }
5d4f98a2 2691 dev_root->track_dirty = 1;
cb517eab
MX		 2692 fs_info->dev_root = dev_root;
2693 btrfs_init_devices_late(fs_info);
3768f368 2694
cb517eab
MX		 2695 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2696 csum_root = btrfs_read_tree_root(tree_root, &location);
2697 if (IS_ERR(csum_root)) {
2698 ret = PTR_ERR(csum_root);
af31f5e5 2699 goto recovery_tree_root;
cb517eab 2700 }
d20f7043 2701 csum_root->track_dirty = 1;
cb517eab 2702 fs_info->csum_root = csum_root;
d20f7043 2703
cb517eab
MX		 2704 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2705 quota_root = btrfs_read_tree_root(tree_root, &location);
2706 if (!IS_ERR(quota_root)) {
bcef60f2
AJ		 2707 quota_root->track_dirty = 1;
2708 fs_info->quota_enabled = 1;
2709 fs_info->pending_quota_state = 1;
cb517eab 2710 fs_info->quota_root = quota_root;
bcef60f2
AJ		 2711 }
2712
f7a81ea4
SB		 2713 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2714 uuid_root = btrfs_read_tree_root(tree_root, &location);
2715 if (IS_ERR(uuid_root)) {
2716 ret = PTR_ERR(uuid_root);
2717 if (ret != -ENOENT)
2718 goto recovery_tree_root;
2719 create_uuid_tree = true;
70f80175 2720 check_uuid_tree = false;
f7a81ea4
SB		 2721 } else {
2722 uuid_root->track_dirty = 1;
2723 fs_info->uuid_root = uuid_root;
70f80175
SB		 2724 create_uuid_tree = false;
2725 check_uuid_tree =
2726 generation != btrfs_super_uuid_tree_generation(disk_super);
f7a81ea4
SB		 2727 }
2728
8929ecfa
YZ		 2729 fs_info->generation = generation;
2730 fs_info->last_trans_committed = generation;
8929ecfa 2731
68310a5e
ID		 2732 ret = btrfs_recover_balance(fs_info);
2733 if (ret) {
efe120a0 2734 printk(KERN_WARNING "BTRFS: failed to recover balance\n");
68310a5e
ID		 2735 goto fail_block_groups;
2736 }
2737
733f4fbb
SB		 2738 ret = btrfs_init_dev_stats(fs_info);
2739 if (ret) {
efe120a0 2740 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
733f4fbb
SB		 2741 ret);
2742 goto fail_block_groups;
2743 }
2744
8dabb742
SB		 2745 ret = btrfs_init_dev_replace(fs_info);
2746 if (ret) {
efe120a0 2747 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
8dabb742
SB		 2748 goto fail_block_groups;
2749 }
2750
2751 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2752
5ac1d209
JM		 2753 ret = btrfs_sysfs_add_one(fs_info);
2754 if (ret) {
efe120a0 2755 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
5ac1d209
JM		 2756 goto fail_block_groups;
2757 }
2758
c59021f8 2759 ret = btrfs_init_space_info(fs_info);
2760 if (ret) {
efe120a0 2761 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
2365dd3c 2762 goto fail_sysfs;
c59021f8 2763 }
2764
1b1d1f66
JB		 2765 ret = btrfs_read_block_groups(extent_root);
2766 if (ret) {
efe120a0 2767 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
2365dd3c 2768 goto fail_sysfs;
1b1d1f66 2769 }
5af3e8cc
SB		 2770 fs_info->num_tolerated_disk_barrier_failures =
2771 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
292fd7fc
SB		 2772 if (fs_info->fs_devices->missing_devices >
2773 fs_info->num_tolerated_disk_barrier_failures &&
2774 !(sb->s_flags & MS_RDONLY)) {
efe120a0
FH		 2775 printk(KERN_WARNING "BTRFS: "
2776 "too many missing devices, writeable mount is not allowed\n");
2365dd3c 2777 goto fail_sysfs;
292fd7fc 2778 }
9078a3e1 2779
a74a4b97
CM		 2780 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2781 "btrfs-cleaner");
57506d50 2782 if (IS_ERR(fs_info->cleaner_kthread))
2365dd3c 2783 goto fail_sysfs;
a74a4b97
CM		 2784
2785 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2786 tree_root,
2787 "btrfs-transaction");
57506d50 2788 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2789 goto fail_cleaner;
a74a4b97 2790
c289811c
CM		 2791 if (!btrfs_test_opt(tree_root, SSD) &&
2792 !btrfs_test_opt(tree_root, NOSSD) &&
2793 !fs_info->fs_devices->rotating) {
efe120a0 2794 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
c289811c
CM		 2795 "mode\n");
2796 btrfs_set_opt(fs_info->mount_opt, SSD);
2797 }
2798
3818aea2
QW		 2799 /* Set the real inode map cache flag */
2800 if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE))
2801 btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE);
2802
21adbd5c
SB		 2803#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2804 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2805 ret = btrfsic_mount(tree_root, fs_devices,
2806 btrfs_test_opt(tree_root,
2807 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2808 1 : 0,
2809 fs_info->check_integrity_print_mask);
2810 if (ret)
efe120a0 2811 printk(KERN_WARNING "BTRFS: failed to initialize"
21adbd5c
SB		 2812 " integrity check module %s\n", sb->s_id);
2813 }
2814#endif
bcef60f2
AJ		 2815 ret = btrfs_read_qgroup_config(fs_info);
2816 if (ret)
2817 goto fail_trans_kthread;
21adbd5c 2818
acce952b 2819 /* do not make disk changes in broken FS */
68ce9682 2820 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM		 2821 u64 bytenr = btrfs_super_log_root(disk_super);
2822
7c2ca468 2823 if (fs_devices->rw_devices == 0) {
efe120a0 2824 printk(KERN_WARNING "BTRFS: log replay required "
d397712b 2825 "on RO media\n");
7c2ca468 2826 err = -EIO;
bcef60f2 2827 goto fail_qgroup;
7c2ca468 2828 }
e02119d5
CM		 2829 blocksize =
2830 btrfs_level_size(tree_root,
2831 btrfs_super_log_root_level(disk_super));
d18a2c44 2832
6f07e42e 2833 log_tree_root = btrfs_alloc_root(fs_info);
676e4c86
DC		 2834 if (!log_tree_root) {
2835 err = -ENOMEM;
bcef60f2 2836 goto fail_qgroup;
676e4c86 2837 }
e02119d5
CM		 2838
2839 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2840 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2841
2842 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 2843 blocksize,
2844 generation + 1);
416bc658
JB		 2845 if (!log_tree_root->node ||
2846 !extent_buffer_uptodate(log_tree_root->node)) {
efe120a0 2847 printk(KERN_ERR "BTRFS: failed to read log tree\n");
416bc658
JB		 2848 free_extent_buffer(log_tree_root->node);
2849 kfree(log_tree_root);
2850 goto fail_trans_kthread;
2851 }
79787eaa 2852 /* returns with log_tree_root freed on success */
e02119d5 2853 ret = btrfs_recover_log_trees(log_tree_root);
79787eaa
JM		 2854 if (ret) {
2855 btrfs_error(tree_root->fs_info, ret,
2856 "Failed to recover log tree");
2857 free_extent_buffer(log_tree_root->node);
2858 kfree(log_tree_root);
2859 goto fail_trans_kthread;
2860 }
e556ce2c
YZ		 2861
2862 if (sb->s_flags & MS_RDONLY) {
79787eaa
JM		 2863 ret = btrfs_commit_super(tree_root);
2864 if (ret)
2865 goto fail_trans_kthread;
e556ce2c 2866 }
e02119d5 2867 }
1a40e23b 2868
76dda93c 2869 ret = btrfs_find_orphan_roots(tree_root);
79787eaa
JM		 2870 if (ret)
2871 goto fail_trans_kthread;
76dda93c 2872
7c2ca468 2873 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b 2874 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2
ID		 2875 if (ret)
2876 goto fail_trans_kthread;
d68fc57b 2877
5d4f98a2 2878 ret = btrfs_recover_relocation(tree_root);
d7ce5843
MX		 2879 if (ret < 0) {
2880 printk(KERN_WARNING
efe120a0 2881 "BTRFS: failed to recover relocation\n");
d7ce5843 2882 err = -EINVAL;
bcef60f2 2883 goto fail_qgroup;
d7ce5843 2884 }
7c2ca468 2885 }
1a40e23b 2886
3de4586c
CM		 2887 location.objectid = BTRFS_FS_TREE_OBJECTID;
2888 location.type = BTRFS_ROOT_ITEM_KEY;
cb517eab 2889 location.offset = 0;
3de4586c 2890
3de4586c 2891 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
3140c9a3
DC		 2892 if (IS_ERR(fs_info->fs_root)) {
2893 err = PTR_ERR(fs_info->fs_root);
bcef60f2 2894 goto fail_qgroup;
3140c9a3 2895 }
c289811c 2896
2b6ba629
ID		 2897 if (sb->s_flags & MS_RDONLY)
2898 return 0;
59641015 2899
2b6ba629
ID		 2900 down_read(&fs_info->cleanup_work_sem);
2901 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2902 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 2903 up_read(&fs_info->cleanup_work_sem);
2b6ba629
ID		 2904 close_ctree(tree_root);
2905 return ret;
2906 }
2907 up_read(&fs_info->cleanup_work_sem);
59641015 2908
2b6ba629
ID		 2909 ret = btrfs_resume_balance_async(fs_info);
2910 if (ret) {
efe120a0 2911 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
2b6ba629
ID		 2912 close_ctree(tree_root);
2913 return ret;
e3acc2a6
JB		 2914 }
2915
8dabb742
SB		 2916 ret = btrfs_resume_dev_replace_async(fs_info);
2917 if (ret) {
efe120a0 2918 pr_warn("BTRFS: failed to resume dev_replace\n");
8dabb742
SB		 2919 close_ctree(tree_root);
2920 return ret;
2921 }
2922
b382a324
JS		 2923 btrfs_qgroup_rescan_resume(fs_info);
2924
f7a81ea4 2925 if (create_uuid_tree) {
efe120a0 2926 pr_info("BTRFS: creating UUID tree\n");
f7a81ea4
SB		 2927 ret = btrfs_create_uuid_tree(fs_info);
2928 if (ret) {
efe120a0 2929 pr_warn("BTRFS: failed to create the UUID tree %d\n",
f7a81ea4
SB		 2930 ret);
2931 close_ctree(tree_root);
2932 return ret;
2933 }
f420ee1e
SB		 2934 } else if (check_uuid_tree ||
2935 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
efe120a0 2936 pr_info("BTRFS: checking UUID tree\n");
70f80175
SB		 2937 ret = btrfs_check_uuid_tree(fs_info);
2938 if (ret) {
efe120a0 2939 pr_warn("BTRFS: failed to check the UUID tree %d\n",
70f80175
SB		 2940 ret);
2941 close_ctree(tree_root);
2942 return ret;
2943 }
2944 } else {
2945 fs_info->update_uuid_tree_gen = 1;
f7a81ea4
SB		 2946 }
2947
ad2b2c80 2948 return 0;
39279cc3 2949
bcef60f2
AJ		 2950fail_qgroup:
2951 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM		 2952fail_trans_kthread:
2953 kthread_stop(fs_info->transaction_kthread);
54067ae9 2954 btrfs_cleanup_transaction(fs_info->tree_root);
7b5ff90e 2955 del_fs_roots(fs_info);
3f157a2f 2956fail_cleaner:
a74a4b97 2957 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2958
2959 /*
2960 * make sure we're done with the btree inode before we stop our
2961 * kthreads
2962 */
2963 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 2964
2365dd3c
AJ		 2965fail_sysfs:
2966 btrfs_sysfs_remove_one(fs_info);
2967
1b1d1f66 2968fail_block_groups:
54067ae9 2969 btrfs_put_block_group_cache(fs_info);
1b1d1f66 2970 btrfs_free_block_groups(fs_info);
af31f5e5
CM		 2971
2972fail_tree_roots:
2973 free_root_pointers(fs_info, 1);
2b8195bb 2974 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 2975
39279cc3 2976fail_sb_buffer:
7abadb64 2977 btrfs_stop_all_workers(fs_info);
16cdcec7 2978fail_alloc:
4543df7e 2979fail_iput:
586e46e2
ID		 2980 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2981
4543df7e 2982 iput(fs_info->btree_inode);
c404e0dc
MX		 2983fail_bio_counter:
2984 percpu_counter_destroy(&fs_info->bio_counter);
963d678b
MX		 2985fail_delalloc_bytes:
2986 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX		 2987fail_dirty_metadata_bytes:
2988 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
ad081f14 2989fail_bdi:
7e662854 2990 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 2991fail_srcu:
2992 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 2993fail:
53b381b3 2994 btrfs_free_stripe_hash_table(fs_info);
586e46e2 2995 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 2996 return err;
af31f5e5
CM		 2997
2998recovery_tree_root:
af31f5e5
CM		 2999 if (!btrfs_test_opt(tree_root, RECOVERY))
3000 goto fail_tree_roots;
3001
3002 free_root_pointers(fs_info, 0);
3003
3004 /* don't use the log in recovery mode, it won't be valid */
3005 btrfs_set_super_log_root(disk_super, 0);
3006
3007 /* we can't trust the free space cache either */
3008 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3009
3010 ret = next_root_backup(fs_info, fs_info->super_copy,
3011 &num_backups_tried, &backup_index);
3012 if (ret == -1)
3013 goto fail_block_groups;
3014 goto retry_root_backup;
eb60ceac
CM		 3015}
3016
f2984462
CM		 3017static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3018{
f2984462
CM		 3019 if (uptodate) {
3020 set_buffer_uptodate(bh);
3021 } else {
442a4f63
SB		 3022 struct btrfs_device *device = (struct btrfs_device *)
3023 bh->b_private;
3024
efe120a0 3025 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
606686ee
JB		 3026 "I/O error on %s\n",
3027 rcu_str_deref(device->name));
1259ab75
CM		 3028 /* note, we dont' set_buffer_write_io_error because we have
3029 * our own ways of dealing with the IO errors
3030 */
f2984462 3031 clear_buffer_uptodate(bh);
442a4f63 3032 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM		 3033 }
3034 unlock_buffer(bh);
3035 put_bh(bh);
3036}
3037
a512bbf8
YZ		 3038struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3039{
3040 struct buffer_head *bh;
3041 struct buffer_head *latest = NULL;
3042 struct btrfs_super_block *super;
3043 int i;
3044 u64 transid = 0;
3045 u64 bytenr;
3046
3047 /* we would like to check all the supers, but that would make
3048 * a btrfs mount succeed after a mkfs from a different FS.
3049 * So, we need to add a special mount option to scan for
3050 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3051 */
3052 for (i = 0; i < 1; i++) {
3053 bytenr = btrfs_sb_offset(i);
8068a47e
AJ		 3054 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3055 i_size_read(bdev->bd_inode))
a512bbf8 3056 break;
8068a47e
AJ		 3057 bh = __bread(bdev, bytenr / 4096,
3058 BTRFS_SUPER_INFO_SIZE);
a512bbf8
YZ		 3059 if (!bh)
3060 continue;
3061
3062 super = (struct btrfs_super_block *)bh->b_data;
3063 if (btrfs_super_bytenr(super) != bytenr ||
3cae210f 3064 btrfs_super_magic(super) != BTRFS_MAGIC) {
a512bbf8
YZ		 3065 brelse(bh);
3066 continue;
3067 }
3068
3069 if (!latest || btrfs_super_generation(super) > transid) {
3070 brelse(latest);
3071 latest = bh;
3072 transid = btrfs_super_generation(super);
3073 } else {
3074 brelse(bh);
3075 }
3076 }
3077 return latest;
3078}
3079
4eedeb75
HH		 3080/*
3081 * this should be called twice, once with wait == 0 and
3082 * once with wait == 1. When wait == 0 is done, all the buffer heads
3083 * we write are pinned.
3084 *
3085 * They are released when wait == 1 is done.
3086 * max_mirrors must be the same for both runs, and it indicates how
3087 * many supers on this one device should be written.
3088 *
3089 * max_mirrors == 0 means to write them all.
3090 */
a512bbf8
YZ		 3091static int write_dev_supers(struct btrfs_device *device,
3092 struct btrfs_super_block *sb,
3093 int do_barriers, int wait, int max_mirrors)
3094{
3095 struct buffer_head *bh;
3096 int i;
3097 int ret;
3098 int errors = 0;
3099 u32 crc;
3100 u64 bytenr;
a512bbf8
YZ		 3101
3102 if (max_mirrors == 0)
3103 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3104
a512bbf8
YZ		 3105 for (i = 0; i < max_mirrors; i++) {
3106 bytenr = btrfs_sb_offset(i);
3107 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
3108 break;
3109
3110 if (wait) {
3111 bh = __find_get_block(device->bdev, bytenr / 4096,
3112 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3113 if (!bh) {
3114 errors++;
3115 continue;
3116 }
a512bbf8 3117 wait_on_buffer(bh);
4eedeb75
HH		 3118 if (!buffer_uptodate(bh))
3119 errors++;
3120
3121 /* drop our reference */
3122 brelse(bh);
3123
3124 /* drop the reference from the wait == 0 run */
3125 brelse(bh);
3126 continue;
a512bbf8
YZ		 3127 } else {
3128 btrfs_set_super_bytenr(sb, bytenr);
3129
3130 crc = ~(u32)0;
b0496686 3131 crc = btrfs_csum_data((char *)sb +
a512bbf8
YZ		 3132 BTRFS_CSUM_SIZE, crc,
3133 BTRFS_SUPER_INFO_SIZE -
3134 BTRFS_CSUM_SIZE);
3135 btrfs_csum_final(crc, sb->csum);
3136
4eedeb75
HH		 3137 /*
3138 * one reference for us, and we leave it for the
3139 * caller
3140 */
a512bbf8
YZ		 3141 bh = __getblk(device->bdev, bytenr / 4096,
3142 BTRFS_SUPER_INFO_SIZE);
634554dc 3143 if (!bh) {
efe120a0 3144 printk(KERN_ERR "BTRFS: couldn't get super "
634554dc
JB		 3145 "buffer head for bytenr %Lu\n", bytenr);
3146 errors++;
3147 continue;
3148 }
3149
a512bbf8
YZ		 3150 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3151
4eedeb75 3152 /* one reference for submit_bh */
a512bbf8 3153 get_bh(bh);
4eedeb75
HH		 3154
3155 set_buffer_uptodate(bh);
a512bbf8
YZ		 3156 lock_buffer(bh);
3157 bh->b_end_io = btrfs_end_buffer_write_sync;
442a4f63 3158 bh->b_private = device;
a512bbf8
YZ		 3159 }
3160
387125fc
CM		 3161 /*
3162 * we fua the first super. The others we allow
3163 * to go down lazy.
3164 */
e8117c26
WS		 3165 if (i == 0)
3166 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3167 else
3168 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
4eedeb75 3169 if (ret)
a512bbf8 3170 errors++;
a512bbf8
YZ		 3171 }
3172 return errors < i ? 0 : -1;
3173}
3174
387125fc
CM		 3175/*
3176 * endio for the write_dev_flush, this will wake anyone waiting
3177 * for the barrier when it is done
3178 */
3179static void btrfs_end_empty_barrier(struct bio *bio, int err)
3180{
3181 if (err) {
3182 if (err == -EOPNOTSUPP)
3183 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3184 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3185 }
3186 if (bio->bi_private)
3187 complete(bio->bi_private);
3188 bio_put(bio);
3189}
3190
3191/*
3192 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3193 * sent down. With wait == 1, it waits for the previous flush.
3194 *
3195 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3196 * capable
3197 */
3198static int write_dev_flush(struct btrfs_device *device, int wait)
3199{
3200 struct bio *bio;
3201 int ret = 0;
3202
3203 if (device->nobarriers)
3204 return 0;
3205
3206 if (wait) {
3207 bio = device->flush_bio;
3208 if (!bio)
3209 return 0;
3210
3211 wait_for_completion(&device->flush_wait);
3212
3213 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
efe120a0 3214 printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
606686ee 3215 rcu_str_deref(device->name));
387125fc 3216 device->nobarriers = 1;
5af3e8cc 3217 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
387125fc 3218 ret = -EIO;
5af3e8cc
SB		 3219 btrfs_dev_stat_inc_and_print(device,
3220 BTRFS_DEV_STAT_FLUSH_ERRS);
387125fc
CM		 3221 }
3222
3223 /* drop the reference from the wait == 0 run */
3224 bio_put(bio);
3225 device->flush_bio = NULL;
3226
3227 return ret;
3228 }
3229
3230 /*
3231 * one reference for us, and we leave it for the
3232 * caller
3233 */
9c017abc 3234 device->flush_bio = NULL;
9be3395b 3235 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
387125fc
CM		 3236 if (!bio)
3237 return -ENOMEM;
3238
3239 bio->bi_end_io = btrfs_end_empty_barrier;
3240 bio->bi_bdev = device->bdev;
3241 init_completion(&device->flush_wait);
3242 bio->bi_private = &device->flush_wait;
3243 device->flush_bio = bio;
3244
3245 bio_get(bio);
21adbd5c 3246 btrfsic_submit_bio(WRITE_FLUSH, bio);
387125fc
CM		 3247
3248 return 0;
3249}
3250
3251/*
3252 * send an empty flush down to each device in parallel,
3253 * then wait for them
3254 */
3255static int barrier_all_devices(struct btrfs_fs_info *info)
3256{
3257 struct list_head *head;
3258 struct btrfs_device *dev;
5af3e8cc
SB		 3259 int errors_send = 0;
3260 int errors_wait = 0;
387125fc
CM		 3261 int ret;
3262
3263 /* send down all the barriers */
3264 head = &info->fs_devices->devices;
3265 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3266 if (dev->missing)
3267 continue;
387125fc 3268 if (!dev->bdev) {
5af3e8cc 3269 errors_send++;
387125fc
CM		 3270 continue;
3271 }
3272 if (!dev->in_fs_metadata || !dev->writeable)
3273 continue;
3274
3275 ret = write_dev_flush(dev, 0);
3276 if (ret)
5af3e8cc 3277 errors_send++;
387125fc
CM		 3278 }
3279
3280 /* wait for all the barriers */
3281 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3282 if (dev->missing)
3283 continue;
387125fc 3284 if (!dev->bdev) {
5af3e8cc 3285 errors_wait++;
387125fc
CM		 3286 continue;
3287 }
3288 if (!dev->in_fs_metadata || !dev->writeable)
3289 continue;
3290
3291 ret = write_dev_flush(dev, 1);
3292 if (ret)
5af3e8cc 3293 errors_wait++;
387125fc 3294 }
5af3e8cc
SB		 3295 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3296 errors_wait > info->num_tolerated_disk_barrier_failures)
387125fc
CM		 3297 return -EIO;
3298 return 0;
3299}
3300
5af3e8cc
SB		 3301int btrfs_calc_num_tolerated_disk_barrier_failures(
3302 struct btrfs_fs_info *fs_info)
3303{
3304 struct btrfs_ioctl_space_info space;
3305 struct btrfs_space_info *sinfo;
3306 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3307 BTRFS_BLOCK_GROUP_SYSTEM,
3308 BTRFS_BLOCK_GROUP_METADATA,
3309 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3310 int num_types = 4;
3311 int i;
3312 int c;
3313 int num_tolerated_disk_barrier_failures =
3314 (int)fs_info->fs_devices->num_devices;
3315
3316 for (i = 0; i < num_types; i++) {
3317 struct btrfs_space_info *tmp;
3318
3319 sinfo = NULL;
3320 rcu_read_lock();
3321 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3322 if (tmp->flags == types[i]) {
3323 sinfo = tmp;
3324 break;
3325 }
3326 }
3327 rcu_read_unlock();
3328
3329 if (!sinfo)
3330 continue;
3331
3332 down_read(&sinfo->groups_sem);
3333 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3334 if (!list_empty(&sinfo->block_groups[c])) {
3335 u64 flags;
3336
3337 btrfs_get_block_group_info(
3338 &sinfo->block_groups[c], &space);
3339 if (space.total_bytes == 0 ||
3340 space.used_bytes == 0)
3341 continue;
3342 flags = space.flags;
3343 /*
3344 * return
3345 * 0: if dup, single or RAID0 is configured for
3346 * any of metadata, system or data, else
3347 * 1: if RAID5 is configured, or if RAID1 or
3348 * RAID10 is configured and only two mirrors
3349 * are used, else
3350 * 2: if RAID6 is configured, else
3351 * num_mirrors - 1: if RAID1 or RAID10 is
3352 * configured and more than
3353 * 2 mirrors are used.
3354 */
3355 if (num_tolerated_disk_barrier_failures > 0 &&
3356 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3357 BTRFS_BLOCK_GROUP_RAID0)) ||
3358 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3359 == 0)))
3360 num_tolerated_disk_barrier_failures = 0;
53b381b3
DW		 3361 else if (num_tolerated_disk_barrier_failures > 1) {
3362 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3363 BTRFS_BLOCK_GROUP_RAID5 |
3364 BTRFS_BLOCK_GROUP_RAID10)) {
3365 num_tolerated_disk_barrier_failures = 1;
3366 } else if (flags &
15b0a89d 3367 BTRFS_BLOCK_GROUP_RAID6) {
53b381b3
DW		 3368 num_tolerated_disk_barrier_failures = 2;
3369 }
3370 }
5af3e8cc
SB		 3371 }
3372 }
3373 up_read(&sinfo->groups_sem);
3374 }
3375
3376 return num_tolerated_disk_barrier_failures;
3377}
3378
48a3b636 3379static int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 3380{
e5e9a520 3381 struct list_head *head;
f2984462 3382 struct btrfs_device *dev;
a061fc8d 3383 struct btrfs_super_block *sb;
f2984462 3384 struct btrfs_dev_item *dev_item;
f2984462
CM		 3385 int ret;
3386 int do_barriers;
a236aed1
CM		 3387 int max_errors;
3388 int total_errors = 0;
a061fc8d 3389 u64 flags;
f2984462
CM		 3390
3391 do_barriers = !btrfs_test_opt(root, NOBARRIER);
af31f5e5 3392 backup_super_roots(root->fs_info);
f2984462 3393
6c41761f 3394 sb = root->fs_info->super_for_commit;
a061fc8d 3395 dev_item = &sb->dev_item;
e5e9a520 3396
174ba509 3397 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 3398 head = &root->fs_info->fs_devices->devices;
d7306801 3399 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
387125fc 3400
5af3e8cc
SB		 3401 if (do_barriers) {
3402 ret = barrier_all_devices(root->fs_info);
3403 if (ret) {
3404 mutex_unlock(
3405 &root->fs_info->fs_devices->device_list_mutex);
3406 btrfs_error(root->fs_info, ret,
3407 "errors while submitting device barriers.");
3408 return ret;
3409 }
3410 }
387125fc 3411
1f78160c 3412 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3413 if (!dev->bdev) {
3414 total_errors++;
3415 continue;
3416 }
2b82032c 3417 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3418 continue;
3419
2b82032c 3420 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 3421 btrfs_set_stack_device_type(dev_item, dev->type);
3422 btrfs_set_stack_device_id(dev_item, dev->devid);
3423 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
3424 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
3425 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3426 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3427 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3428 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 3429 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 3430
a061fc8d
CM		 3431 flags = btrfs_super_flags(sb);
3432 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3433
a512bbf8 3434 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 3435 if (ret)
3436 total_errors++;
f2984462 3437 }
a236aed1 3438 if (total_errors > max_errors) {
efe120a0 3439 btrfs_err(root->fs_info, "%d errors while writing supers",
d397712b 3440 total_errors);
a724b436 3441 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
79787eaa 3442
9d565ba4
SB		 3443 /* FUA is masked off if unsupported and can't be the reason */
3444 btrfs_error(root->fs_info, -EIO,
3445 "%d errors while writing supers", total_errors);
3446 return -EIO;
a236aed1 3447 }
f2984462 3448
a512bbf8 3449 total_errors = 0;
1f78160c 3450 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3451 if (!dev->bdev)
3452 continue;
2b82032c 3453 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3454 continue;
3455
a512bbf8
YZ		 3456 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3457 if (ret)
3458 total_errors++;
f2984462 3459 }
174ba509 3460 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 3461 if (total_errors > max_errors) {
79787eaa
JM		 3462 btrfs_error(root->fs_info, -EIO,
3463 "%d errors while writing supers", total_errors);
3464 return -EIO;
a236aed1 3465 }
f2984462
CM		 3466 return 0;
3467}
3468
a512bbf8
YZ		 3469int write_ctree_super(struct btrfs_trans_handle *trans,
3470 struct btrfs_root *root, int max_mirrors)
eb60ceac 3471{
f570e757 3472 return write_all_supers(root, max_mirrors);
cfaa7295
CM		 3473}
3474
cb517eab
MX		 3475/* Drop a fs root from the radix tree and free it. */
3476void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3477 struct btrfs_root *root)
2619ba1f 3478{
4df27c4d 3479 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 3480 radix_tree_delete(&fs_info->fs_roots_radix,
3481 (unsigned long)root->root_key.objectid);
4df27c4d 3482 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 3483
3484 if (btrfs_root_refs(&root->root_item) == 0)
3485 synchronize_srcu(&fs_info->subvol_srcu);
3486
1a4319cc 3487 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
3321719e 3488 btrfs_free_log(NULL, root);
3321719e 3489
581bb050
LZ		 3490 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3491 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3492 free_fs_root(root);
4df27c4d
YZ		 3493}
3494
3495static void free_fs_root(struct btrfs_root *root)
3496{
82d5902d 3497 iput(root->cache_inode);
4df27c4d 3498 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
1cb048f5
FDBM		 3499 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3500 root->orphan_block_rsv = NULL;
0ee5dc67
AV		 3501 if (root->anon_dev)
3502 free_anon_bdev(root->anon_dev);
8257b2dc
MX		 3503 if (root->subv_writers)
3504 btrfs_free_subvolume_writers(root->subv_writers);
4df27c4d
YZ		 3505 free_extent_buffer(root->node);
3506 free_extent_buffer(root->commit_root);
581bb050
LZ		 3507 kfree(root->free_ino_ctl);
3508 kfree(root->free_ino_pinned);
d397712b 3509 kfree(root->name);
b0feb9d9 3510 btrfs_put_fs_root(root);
2619ba1f
CM		 3511}
3512
cb517eab
MX		 3513void btrfs_free_fs_root(struct btrfs_root *root)
3514{
3515 free_fs_root(root);
2619ba1f
CM		 3516}
3517
c146afad 3518int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3519{
c146afad
YZ		 3520 u64 root_objectid = 0;
3521 struct btrfs_root *gang[8];
3522 int i;
3768f368 3523 int ret;
e089f05c 3524
c146afad
YZ		 3525 while (1) {
3526 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3527 (void **)gang, root_objectid,
3528 ARRAY_SIZE(gang));
3529 if (!ret)
3530 break;
5d4f98a2
YZ		 3531
3532 root_objectid = gang[ret - 1]->root_key.objectid + 1;
c146afad 3533 for (i = 0; i < ret; i++) {
66b4ffd1
JB		 3534 int err;
3535
c146afad 3536 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 3537 err = btrfs_orphan_cleanup(gang[i]);
3538 if (err)
3539 return err;
c146afad
YZ		 3540 }
3541 root_objectid++;
3542 }
3543 return 0;
3544}
a2135011 3545
c146afad
YZ		 3546int btrfs_commit_super(struct btrfs_root *root)
3547{
3548 struct btrfs_trans_handle *trans;
a74a4b97 3549
c146afad 3550 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 3551 btrfs_run_delayed_iputs(root);
c146afad 3552 mutex_unlock(&root->fs_info->cleaner_mutex);
9d1a2a3a 3553 wake_up_process(root->fs_info->cleaner_kthread);
c71bf099
YZ		 3554
3555 /* wait until ongoing cleanup work done */
3556 down_write(&root->fs_info->cleanup_work_sem);
3557 up_write(&root->fs_info->cleanup_work_sem);
3558
7a7eaa40 3559 trans = btrfs_join_transaction(root);
3612b495
TI		 3560 if (IS_ERR(trans))
3561 return PTR_ERR(trans);
d52c1bcc 3562 return btrfs_commit_transaction(trans, root);
c146afad
YZ		 3563}
3564
3565int close_ctree(struct btrfs_root *root)
3566{
3567 struct btrfs_fs_info *fs_info = root->fs_info;
3568 int ret;
3569
3570 fs_info->closing = 1;
3571 smp_mb();
3572
803b2f54
SB		 3573 /* wait for the uuid_scan task to finish */
3574 down(&fs_info->uuid_tree_rescan_sem);
3575 /* avoid complains from lockdep et al., set sem back to initial state */
3576 up(&fs_info->uuid_tree_rescan_sem);
3577
837d5b6e 3578 /* pause restriper - we want to resume on mount */
aa1b8cd4 3579 btrfs_pause_balance(fs_info);
837d5b6e 3580
8dabb742
SB		 3581 btrfs_dev_replace_suspend_for_unmount(fs_info);
3582
aa1b8cd4 3583 btrfs_scrub_cancel(fs_info);
4cb5300b
CM		 3584
3585 /* wait for any defraggers to finish */
3586 wait_event(fs_info->transaction_wait,
3587 (atomic_read(&fs_info->defrag_running) == 0));
3588
3589 /* clear out the rbtree of defraggable inodes */
26176e7c 3590 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3591
c146afad 3592 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 3593 ret = btrfs_commit_super(root);
3594 if (ret)
efe120a0 3595 btrfs_err(root->fs_info, "commit super ret %d", ret);
acce952b 3596 }
3597
87533c47 3598 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
68ce9682 3599 btrfs_error_commit_super(root);
0f7d52f4 3600
e3029d9f
AV		 3601 kthread_stop(fs_info->transaction_kthread);
3602 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3603
f25784b3
YZ		 3604 fs_info->closing = 2;
3605 smp_mb();
3606
bcef60f2
AJ		 3607 btrfs_free_qgroup_config(root->fs_info);
3608
963d678b 3609 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
efe120a0 3610 btrfs_info(root->fs_info, "at unmount delalloc count %lld",
963d678b 3611 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3612 }
bcc63abb 3613
5ac1d209
JM		 3614 btrfs_sysfs_remove_one(fs_info);
3615
c146afad 3616 del_fs_roots(fs_info);
d10c5f31 3617
1a4319cc
LB		 3618 btrfs_put_block_group_cache(fs_info);
3619
2b1360da
JB		 3620 btrfs_free_block_groups(fs_info);
3621
96192499
JB		 3622 btrfs_stop_all_workers(fs_info);
3623
13e6c37b 3624 free_root_pointers(fs_info, 1);
9ad6b7bc 3625
13e6c37b 3626 iput(fs_info->btree_inode);
d6bfde87 3627
21adbd5c
SB		 3628#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3629 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3630 btrfsic_unmount(root, fs_info->fs_devices);
3631#endif
3632
dfe25020 3633 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3634 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3635
e2d84521 3636 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 3637 percpu_counter_destroy(&fs_info->delalloc_bytes);
c404e0dc 3638 percpu_counter_destroy(&fs_info->bio_counter);
04160088 3639 bdi_destroy(&fs_info->bdi);
76dda93c 3640 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 3641
53b381b3
DW		 3642 btrfs_free_stripe_hash_table(fs_info);
3643
1cb048f5
FDBM		 3644 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3645 root->orphan_block_rsv = NULL;
3646
eb60ceac
CM		 3647 return 0;
3648}
3649
b9fab919
CM		 3650int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3651 int atomic)
5f39d397 3652{
1259ab75 3653 int ret;
727011e0 3654 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 3655
0b32f4bb 3656 ret = extent_buffer_uptodate(buf);
1259ab75
CM		 3657 if (!ret)
3658 return ret;
3659
3660 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM		 3661 parent_transid, atomic);
3662 if (ret == -EAGAIN)
3663 return ret;
1259ab75 3664 return !ret;
5f39d397
CM		 3665}
3666
3667int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 3668{
0b32f4bb 3669 return set_extent_buffer_uptodate(buf);
5f39d397 3670}
6702ed49 3671
5f39d397
CM		 3672void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3673{
06ea65a3 3674 struct btrfs_root *root;
5f39d397 3675 u64 transid = btrfs_header_generation(buf);
b9473439 3676 int was_dirty;
b4ce94de 3677
06ea65a3
JB		 3678#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3679 /*
3680 * This is a fast path so only do this check if we have sanity tests
3681 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3682 * outside of the sanity tests.
3683 */
3684 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3685 return;
3686#endif
3687 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
b9447ef8 3688 btrfs_assert_tree_locked(buf);
31b1a2bd
JL		 3689 if (transid != root->fs_info->generation)
3690 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
d397712b 3691 "found %llu running %llu\n",
c1c9ff7c 3692 buf->start, transid, root->fs_info->generation);
0b32f4bb 3693 was_dirty = set_extent_buffer_dirty(buf);
e2d84521
MX		 3694 if (!was_dirty)
3695 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3696 buf->len,
3697 root->fs_info->dirty_metadata_batch);
eb60ceac
CM		 3698}
3699
b53d3f5d
LB		 3700static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3701 int flush_delayed)
16cdcec7
MX		 3702{
3703 /*
3704 * looks as though older kernels can get into trouble with
3705 * this code, they end up stuck in balance_dirty_pages forever
3706 */
e2d84521 3707 int ret;
16cdcec7
MX		 3708
3709 if (current->flags & PF_MEMALLOC)
3710 return;
3711
b53d3f5d
LB		 3712 if (flush_delayed)
3713 btrfs_balance_delayed_items(root);
16cdcec7 3714
e2d84521
MX		 3715 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3716 BTRFS_DIRTY_METADATA_THRESH);
3717 if (ret > 0) {
d0e1d66b
NJ		 3718 balance_dirty_pages_ratelimited(
3719 root->fs_info->btree_inode->i_mapping);
16cdcec7
MX		 3720 }
3721 return;
3722}
3723
b53d3f5d 3724void btrfs_btree_balance_dirty(struct btrfs_root *root)
35b7e476 3725{
b53d3f5d
LB		 3726 __btrfs_btree_balance_dirty(root, 1);
3727}
585ad2c3 3728
b53d3f5d
LB		 3729void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3730{
3731 __btrfs_btree_balance_dirty(root, 0);
35b7e476 3732}
6b80053d 3733
ca7a79ad 3734int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 3735{
727011e0 3736 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
0b32f4bb 3737 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
6b80053d 3738}
0da5468f 3739
fcd1f065 3740static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 3741 int read_only)
3742{
1104a885
DS		 3743 /*
3744 * Placeholder for checks
3745 */
fcd1f065 3746 return 0;
acce952b 3747}
3748
48a3b636 3749static void btrfs_error_commit_super(struct btrfs_root *root)
acce952b 3750{
acce952b 3751 mutex_lock(&root->fs_info->cleaner_mutex);
3752 btrfs_run_delayed_iputs(root);
3753 mutex_unlock(&root->fs_info->cleaner_mutex);
3754
3755 down_write(&root->fs_info->cleanup_work_sem);
3756 up_write(&root->fs_info->cleanup_work_sem);
3757
3758 /* cleanup FS via transaction */
3759 btrfs_cleanup_transaction(root);
acce952b 3760}
3761
569e0f35
JB		 3762static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
3763 struct btrfs_root *root)
acce952b 3764{
3765 struct btrfs_inode *btrfs_inode;
3766 struct list_head splice;
3767
3768 INIT_LIST_HEAD(&splice);
3769
3770 mutex_lock(&root->fs_info->ordered_operations_mutex);
199c2a9c 3771 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3772
569e0f35 3773 list_splice_init(&t->ordered_operations, &splice);
acce952b 3774 while (!list_empty(&splice)) {
3775 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3776 ordered_operations);
3777
3778 list_del_init(&btrfs_inode->ordered_operations);
199c2a9c 3779 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3780
3781 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3782
199c2a9c 3783 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3784 }
3785
199c2a9c 3786 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3787 mutex_unlock(&root->fs_info->ordered_operations_mutex);
acce952b 3788}
3789
143bede5 3790static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 3791{
acce952b 3792 struct btrfs_ordered_extent *ordered;
acce952b 3793
199c2a9c 3794 spin_lock(&root->ordered_extent_lock);
779880ef
JB		 3795 /*
3796 * This will just short circuit the ordered completion stuff which will
3797 * make sure the ordered extent gets properly cleaned up.
3798 */
199c2a9c 3799 list_for_each_entry(ordered, &root->ordered_extents,
779880ef
JB		 3800 root_extent_list)
3801 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
199c2a9c
MX		 3802 spin_unlock(&root->ordered_extent_lock);
3803}
3804
3805static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3806{
3807 struct btrfs_root *root;
3808 struct list_head splice;
3809
3810 INIT_LIST_HEAD(&splice);
3811
3812 spin_lock(&fs_info->ordered_root_lock);
3813 list_splice_init(&fs_info->ordered_roots, &splice);
3814 while (!list_empty(&splice)) {
3815 root = list_first_entry(&splice, struct btrfs_root,
3816 ordered_root);
1de2cfde
JB		 3817 list_move_tail(&root->ordered_root,
3818 &fs_info->ordered_roots);
199c2a9c 3819
2a85d9ca 3820 spin_unlock(&fs_info->ordered_root_lock);
199c2a9c
MX		 3821 btrfs_destroy_ordered_extents(root);
3822
2a85d9ca
LB		 3823 cond_resched();
3824 spin_lock(&fs_info->ordered_root_lock);
199c2a9c
MX		 3825 }
3826 spin_unlock(&fs_info->ordered_root_lock);
acce952b 3827}
3828
35a3621b
SB		 3829static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3830 struct btrfs_root *root)
acce952b 3831{
3832 struct rb_node *node;
3833 struct btrfs_delayed_ref_root *delayed_refs;
3834 struct btrfs_delayed_ref_node *ref;
3835 int ret = 0;
3836
3837 delayed_refs = &trans->delayed_refs;
3838
3839 spin_lock(&delayed_refs->lock);
d7df2c79 3840 if (atomic_read(&delayed_refs->num_entries) == 0) {
cfece4db 3841 spin_unlock(&delayed_refs->lock);
efe120a0 3842 btrfs_info(root->fs_info, "delayed_refs has NO entry");
acce952b 3843 return ret;
3844 }
3845
d7df2c79
JB		 3846 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
3847 struct btrfs_delayed_ref_head *head;
e78417d1 3848 bool pin_bytes = false;
acce952b 3849
d7df2c79
JB		 3850 head = rb_entry(node, struct btrfs_delayed_ref_head,
3851 href_node);
3852 if (!mutex_trylock(&head->mutex)) {
3853 atomic_inc(&head->node.refs);
3854 spin_unlock(&delayed_refs->lock);
acce952b 3855
d7df2c79 3856 mutex_lock(&head->mutex);
e78417d1 3857 mutex_unlock(&head->mutex);
d7df2c79
JB		 3858 btrfs_put_delayed_ref(&head->node);
3859 spin_lock(&delayed_refs->lock);
3860 continue;
3861 }
3862 spin_lock(&head->lock);
3863 while ((node = rb_first(&head->ref_root)) != NULL) {
3864 ref = rb_entry(node, struct btrfs_delayed_ref_node,
3865 rb_node);
3866 ref->in_tree = 0;
3867 rb_erase(&ref->rb_node, &head->ref_root);
3868 atomic_dec(&delayed_refs->num_entries);
3869 btrfs_put_delayed_ref(ref);
e78417d1 3870 }
d7df2c79
JB		 3871 if (head->must_insert_reserved)
3872 pin_bytes = true;
3873 btrfs_free_delayed_extent_op(head->extent_op);
3874 delayed_refs->num_heads--;
3875 if (head->processing == 0)
3876 delayed_refs->num_heads_ready--;
3877 atomic_dec(&delayed_refs->num_entries);
3878 head->node.in_tree = 0;
3879 rb_erase(&head->href_node, &delayed_refs->href_root);
3880 spin_unlock(&head->lock);
3881 spin_unlock(&delayed_refs->lock);
3882 mutex_unlock(&head->mutex);
acce952b 3883
d7df2c79
JB		 3884 if (pin_bytes)
3885 btrfs_pin_extent(root, head->node.bytenr,
3886 head->node.num_bytes, 1);
3887 btrfs_put_delayed_ref(&head->node);
acce952b 3888 cond_resched();
3889 spin_lock(&delayed_refs->lock);
3890 }
3891
3892 spin_unlock(&delayed_refs->lock);
3893
3894 return ret;
3895}
3896
143bede5 3897static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 3898{
3899 struct btrfs_inode *btrfs_inode;
3900 struct list_head splice;
3901
3902 INIT_LIST_HEAD(&splice);
3903
eb73c1b7
MX		 3904 spin_lock(&root->delalloc_lock);
3905 list_splice_init(&root->delalloc_inodes, &splice);
acce952b 3906
3907 while (!list_empty(&splice)) {
eb73c1b7
MX		 3908 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3909 delalloc_inodes);
acce952b 3910
3911 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX		 3912 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3913 &btrfs_inode->runtime_flags);
eb73c1b7 3914 spin_unlock(&root->delalloc_lock);
acce952b 3915
3916 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3917
eb73c1b7 3918 spin_lock(&root->delalloc_lock);
acce952b 3919 }
3920
eb73c1b7
MX		 3921 spin_unlock(&root->delalloc_lock);
3922}
3923
3924static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3925{
3926 struct btrfs_root *root;
3927 struct list_head splice;
3928
3929 INIT_LIST_HEAD(&splice);
3930
3931 spin_lock(&fs_info->delalloc_root_lock);
3932 list_splice_init(&fs_info->delalloc_roots, &splice);
3933 while (!list_empty(&splice)) {
3934 root = list_first_entry(&splice, struct btrfs_root,
3935 delalloc_root);
3936 list_del_init(&root->delalloc_root);
3937 root = btrfs_grab_fs_root(root);
3938 BUG_ON(!root);
3939 spin_unlock(&fs_info->delalloc_root_lock);
3940
3941 btrfs_destroy_delalloc_inodes(root);
3942 btrfs_put_fs_root(root);
3943
3944 spin_lock(&fs_info->delalloc_root_lock);
3945 }
3946 spin_unlock(&fs_info->delalloc_root_lock);
acce952b 3947}
3948
3949static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3950 struct extent_io_tree *dirty_pages,
3951 int mark)
3952{
3953 int ret;
acce952b 3954 struct extent_buffer *eb;
3955 u64 start = 0;
3956 u64 end;
acce952b 3957
3958 while (1) {
3959 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 3960 mark, NULL);
acce952b 3961 if (ret)
3962 break;
3963
3964 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
3965 while (start <= end) {
fd8b2b61
JB		 3966 eb = btrfs_find_tree_block(root, start,
3967 root->leafsize);
69a85bd8 3968 start += root->leafsize;
fd8b2b61 3969 if (!eb)
acce952b 3970 continue;
fd8b2b61 3971 wait_on_extent_buffer_writeback(eb);
acce952b 3972
fd8b2b61
JB		 3973 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
3974 &eb->bflags))
3975 clear_extent_buffer_dirty(eb);
3976 free_extent_buffer_stale(eb);
acce952b 3977 }
3978 }
3979
3980 return ret;
3981}
3982
3983static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3984 struct extent_io_tree *pinned_extents)
3985{
3986 struct extent_io_tree *unpin;
3987 u64 start;
3988 u64 end;
3989 int ret;
ed0eaa14 3990 bool loop = true;
acce952b 3991
3992 unpin = pinned_extents;
ed0eaa14 3993again:
acce952b 3994 while (1) {
3995 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 3996 EXTENT_DIRTY, NULL);
acce952b 3997 if (ret)
3998 break;
3999
4000 /* opt_discard */
5378e607
LD		 4001 if (btrfs_test_opt(root, DISCARD))
4002 ret = btrfs_error_discard_extent(root, start,
4003 end + 1 - start,
4004 NULL);
acce952b 4005
4006 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4007 btrfs_error_unpin_extent_range(root, start, end);
4008 cond_resched();
4009 }
4010
ed0eaa14
LB		 4011 if (loop) {
4012 if (unpin == &root->fs_info->freed_extents[0])
4013 unpin = &root->fs_info->freed_extents[1];
4014 else
4015 unpin = &root->fs_info->freed_extents[0];
4016 loop = false;
4017 goto again;
4018 }
4019
acce952b 4020 return 0;
4021}
4022
49b25e05
JM		 4023void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4024 struct btrfs_root *root)
4025{
724e2315
JB		 4026 btrfs_destroy_ordered_operations(cur_trans, root);
4027
49b25e05 4028 btrfs_destroy_delayed_refs(cur_trans, root);
49b25e05 4029
4a9d8bde 4030 cur_trans->state = TRANS_STATE_COMMIT_START;
d7096fc3 4031 wake_up(&root->fs_info->transaction_blocked_wait);
49b25e05 4032
4a9d8bde 4033 cur_trans->state = TRANS_STATE_UNBLOCKED;
d7096fc3 4034 wake_up(&root->fs_info->transaction_wait);
49b25e05 4035
67cde344
MX		 4036 btrfs_destroy_delayed_inodes(root);
4037 btrfs_assert_delayed_root_empty(root);
49b25e05 4038
49b25e05
JM		 4039 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4040 EXTENT_DIRTY);
6e841e32
LB		 4041 btrfs_destroy_pinned_extent(root,
4042 root->fs_info->pinned_extents);
49b25e05 4043
4a9d8bde
MX		 4044 cur_trans->state =TRANS_STATE_COMPLETED;
4045 wake_up(&cur_trans->commit_wait);
4046
49b25e05
JM		 4047 /*
4048 memset(cur_trans, 0, sizeof(*cur_trans));
4049 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4050 */
4051}
4052
48a3b636 4053static int btrfs_cleanup_transaction(struct btrfs_root *root)
acce952b 4054{
4055 struct btrfs_transaction *t;
acce952b 4056
acce952b 4057 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4058
a4abeea4 4059 spin_lock(&root->fs_info->trans_lock);
724e2315
JB		 4060 while (!list_empty(&root->fs_info->trans_list)) {
4061 t = list_first_entry(&root->fs_info->trans_list,
4062 struct btrfs_transaction, list);
4063 if (t->state >= TRANS_STATE_COMMIT_START) {
4064 atomic_inc(&t->use_count);
4065 spin_unlock(&root->fs_info->trans_lock);
4066 btrfs_wait_for_commit(root, t->transid);
4067 btrfs_put_transaction(t);
4068 spin_lock(&root->fs_info->trans_lock);
4069 continue;
4070 }
4071 if (t == root->fs_info->running_transaction) {
4072 t->state = TRANS_STATE_COMMIT_DOING;
4073 spin_unlock(&root->fs_info->trans_lock);
4074 /*
4075 * We wait for 0 num_writers since we don't hold a trans
4076 * handle open currently for this transaction.
4077 */
4078 wait_event(t->writer_wait,
4079 atomic_read(&t->num_writers) == 0);
4080 } else {
4081 spin_unlock(&root->fs_info->trans_lock);
4082 }
4083 btrfs_cleanup_one_transaction(t, root);
4a9d8bde 4084
724e2315
JB		 4085 spin_lock(&root->fs_info->trans_lock);
4086 if (t == root->fs_info->running_transaction)
4087 root->fs_info->running_transaction = NULL;
acce952b 4088 list_del_init(&t->list);
724e2315 4089 spin_unlock(&root->fs_info->trans_lock);
acce952b 4090
724e2315
JB		 4091 btrfs_put_transaction(t);
4092 trace_btrfs_transaction_commit(root);
4093 spin_lock(&root->fs_info->trans_lock);
4094 }
4095 spin_unlock(&root->fs_info->trans_lock);
4096 btrfs_destroy_all_ordered_extents(root->fs_info);
4097 btrfs_destroy_delayed_inodes(root);
4098 btrfs_assert_delayed_root_empty(root);
4099 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4100 btrfs_destroy_all_delalloc_inodes(root->fs_info);
acce952b 4101 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 4102
4103 return 0;
4104}
4105
d1310b2e 4106static struct extent_io_ops btree_extent_io_ops = {
ce9adaa5 4107 .readpage_end_io_hook = btree_readpage_end_io_hook,
4bb31e92 4108 .readpage_io_failed_hook = btree_io_failed_hook,
0b86a832 4109 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 4110 /* note we're sharing with inode.c for the merge bio hook */
4111 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 4112};
Linux 6.x block layer and io_uring tree(s)