projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: Add threshold workqueue based on kernel workqueue
[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;
8b712842 58static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 59static void free_fs_root(struct btrfs_root *root);
fcd1f065 60static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 61 int read_only);
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;
8b712842 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;
8b712842 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;
8b712842
CM		 681 end_io_wq->work.func = end_workqueue_fn;
682 end_io_wq->work.flags = 0;
d20f7043 683
7b6d91da 684 if (bio->bi_rw & REQ_WRITE) {
53b381b3 685 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA)
cad321ad
CM		 686 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
687 &end_io_wq->work);
53b381b3 688 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE)
0cb59c99
JB		 689 btrfs_queue_worker(&fs_info->endio_freespace_worker,
690 &end_io_wq->work);
53b381b3
DW		 691 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
692 btrfs_queue_worker(&fs_info->endio_raid56_workers,
693 &end_io_wq->work);
cad321ad
CM		 694 else
695 btrfs_queue_worker(&fs_info->endio_write_workers,
696 &end_io_wq->work);
d20f7043 697 } else {
53b381b3
DW		 698 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
699 btrfs_queue_worker(&fs_info->endio_raid56_workers,
700 &end_io_wq->work);
701 else if (end_io_wq->metadata)
d20f7043
CM		 702 btrfs_queue_worker(&fs_info->endio_meta_workers,
703 &end_io_wq->work);
704 else
705 btrfs_queue_worker(&fs_info->endio_workers,
706 &end_io_wq->work);
707 }
ce9adaa5
CM		 708}
709
0cb59c99
JB		 710/*
711 * For the metadata arg you want
712 *
713 * 0 - if data
714 * 1 - if normal metadta
715 * 2 - if writing to the free space cache area
53b381b3 716 * 3 - raid parity work
0cb59c99 717 */
22c59948
CM		 718int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
719 int metadata)
0b86a832 720{
ce9adaa5 721 struct end_io_wq *end_io_wq;
ce9adaa5
CM		 722 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
723 if (!end_io_wq)
724 return -ENOMEM;
725
726 end_io_wq->private = bio->bi_private;
727 end_io_wq->end_io = bio->bi_end_io;
22c59948 728 end_io_wq->info = info;
ce9adaa5
CM		 729 end_io_wq->error = 0;
730 end_io_wq->bio = bio;
22c59948 731 end_io_wq->metadata = metadata;
ce9adaa5
CM		 732
733 bio->bi_private = end_io_wq;
734 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 735 return 0;
736}
737
b64a2851 738unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 739{
4854ddd0
CM		 740 unsigned long limit = min_t(unsigned long,
741 info->workers.max_workers,
742 info->fs_devices->open_devices);
743 return 256 * limit;
744}
0986fe9e 745
4a69a410
CM		 746static void run_one_async_start(struct btrfs_work *work)
747{
4a69a410 748 struct async_submit_bio *async;
79787eaa 749 int ret;
4a69a410
CM		 750
751 async = container_of(work, struct async_submit_bio, work);
79787eaa
JM		 752 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
753 async->mirror_num, async->bio_flags,
754 async->bio_offset);
755 if (ret)
756 async->error = ret;
4a69a410
CM		 757}
758
759static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 760{
761 struct btrfs_fs_info *fs_info;
762 struct async_submit_bio *async;
4854ddd0 763 int limit;
8b712842
CM		 764
765 async = container_of(work, struct async_submit_bio, work);
766 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 767
b64a2851 768 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 769 limit = limit * 2 / 3;
770
66657b31 771 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
b64a2851 772 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 773 wake_up(&fs_info->async_submit_wait);
774
79787eaa
JM		 775 /* If an error occured we just want to clean up the bio and move on */
776 if (async->error) {
777 bio_endio(async->bio, async->error);
778 return;
779 }
780
4a69a410 781 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 782 async->mirror_num, async->bio_flags,
783 async->bio_offset);
4a69a410
CM		 784}
785
786static void run_one_async_free(struct btrfs_work *work)
787{
788 struct async_submit_bio *async;
789
790 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 791 kfree(async);
792}
793
44b8bd7e
CM		 794int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
795 int rw, struct bio *bio, int mirror_num,
c8b97818 796 unsigned long bio_flags,
eaf25d93 797 u64 bio_offset,
4a69a410
CM		 798 extent_submit_bio_hook_t *submit_bio_start,
799 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 800{
801 struct async_submit_bio *async;
802
803 async = kmalloc(sizeof(*async), GFP_NOFS);
804 if (!async)
805 return -ENOMEM;
806
807 async->inode = inode;
808 async->rw = rw;
809 async->bio = bio;
810 async->mirror_num = mirror_num;
4a69a410
CM		 811 async->submit_bio_start = submit_bio_start;
812 async->submit_bio_done = submit_bio_done;
813
814 async->work.func = run_one_async_start;
815 async->work.ordered_func = run_one_async_done;
816 async->work.ordered_free = run_one_async_free;
817
8b712842 818 async->work.flags = 0;
c8b97818 819 async->bio_flags = bio_flags;
eaf25d93 820 async->bio_offset = bio_offset;
8c8bee1d 821
79787eaa
JM		 822 async->error = 0;
823
cb03c743 824 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 825
7b6d91da 826 if (rw & REQ_SYNC)
d313d7a3
CM		 827 btrfs_set_work_high_prio(&async->work);
828
8b712842 829 btrfs_queue_worker(&fs_info->workers, &async->work);
9473f16c 830
d397712b 831 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 832 atomic_read(&fs_info->nr_async_submits)) {
833 wait_event(fs_info->async_submit_wait,
834 (atomic_read(&fs_info->nr_async_submits) == 0));
835 }
836
44b8bd7e
CM		 837 return 0;
838}
839
ce3ed71a
CM		 840static int btree_csum_one_bio(struct bio *bio)
841{
842 struct bio_vec *bvec = bio->bi_io_vec;
843 int bio_index = 0;
844 struct btrfs_root *root;
79787eaa 845 int ret = 0;
ce3ed71a
CM		 846
847 WARN_ON(bio->bi_vcnt <= 0);
d397712b 848 while (bio_index < bio->bi_vcnt) {
ce3ed71a 849 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
79787eaa
JM		 850 ret = csum_dirty_buffer(root, bvec->bv_page);
851 if (ret)
852 break;
ce3ed71a
CM		 853 bio_index++;
854 bvec++;
855 }
79787eaa 856 return ret;
ce3ed71a
CM		 857}
858
4a69a410
CM		 859static int __btree_submit_bio_start(struct inode *inode, int rw,
860 struct bio *bio, int mirror_num,
eaf25d93
CM		 861 unsigned long bio_flags,
862 u64 bio_offset)
22c59948 863{
8b712842
CM		 864 /*
865 * when we're called for a write, we're already in the async
5443be45 866 * submission context. Just jump into btrfs_map_bio
8b712842 867 */
79787eaa 868 return btree_csum_one_bio(bio);
4a69a410 869}
22c59948 870
4a69a410 871static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 872 int mirror_num, unsigned long bio_flags,
873 u64 bio_offset)
4a69a410 874{
61891923
SB		 875 int ret;
876
8b712842 877 /*
4a69a410
CM		 878 * when we're called for a write, we're already in the async
879 * submission context. Just jump into btrfs_map_bio
8b712842 880 */
61891923
SB		 881 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
882 if (ret)
883 bio_endio(bio, ret);
884 return ret;
0b86a832
CM		 885}
886
de0022b9
JB		 887static int check_async_write(struct inode *inode, unsigned long bio_flags)
888{
889 if (bio_flags & EXTENT_BIO_TREE_LOG)
890 return 0;
891#ifdef CONFIG_X86
892 if (cpu_has_xmm4_2)
893 return 0;
894#endif
895 return 1;
896}
897
44b8bd7e 898static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 899 int mirror_num, unsigned long bio_flags,
900 u64 bio_offset)
44b8bd7e 901{
de0022b9 902 int async = check_async_write(inode, bio_flags);
cad321ad
CM		 903 int ret;
904
7b6d91da 905 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 906 /*
907 * called for a read, do the setup so that checksum validation
908 * can happen in the async kernel threads
909 */
f3f266ab
CM		 910 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
911 bio, 1);
1d4284bd 912 if (ret)
61891923
SB		 913 goto out_w_error;
914 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
915 mirror_num, 0);
de0022b9
JB		 916 } else if (!async) {
917 ret = btree_csum_one_bio(bio);
918 if (ret)
61891923
SB		 919 goto out_w_error;
920 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
921 mirror_num, 0);
922 } else {
923 /*
924 * kthread helpers are used to submit writes so that
925 * checksumming can happen in parallel across all CPUs
926 */
927 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
928 inode, rw, bio, mirror_num, 0,
929 bio_offset,
930 __btree_submit_bio_start,
931 __btree_submit_bio_done);
44b8bd7e 932 }
d313d7a3 933
61891923
SB		 934 if (ret) {
935out_w_error:
936 bio_endio(bio, ret);
937 }
938 return ret;
44b8bd7e
CM		 939}
940
3dd1462e 941#ifdef CONFIG_MIGRATION
784b4e29 942static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG		 943 struct page *newpage, struct page *page,
944 enum migrate_mode mode)
784b4e29
CM		 945{
946 /*
947 * we can't safely write a btree page from here,
948 * we haven't done the locking hook
949 */
950 if (PageDirty(page))
951 return -EAGAIN;
952 /*
953 * Buffers may be managed in a filesystem specific way.
954 * We must have no buffers or drop them.
955 */
956 if (page_has_private(page) &&
957 !try_to_release_page(page, GFP_KERNEL))
958 return -EAGAIN;
a6bc32b8 959 return migrate_page(mapping, newpage, page, mode);
784b4e29 960}
3dd1462e 961#endif
784b4e29 962
0da5468f
CM		 963
964static int btree_writepages(struct address_space *mapping,
965 struct writeback_control *wbc)
966{
e2d84521
MX		 967 struct btrfs_fs_info *fs_info;
968 int ret;
969
d8d5f3e1 970 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM		 971
972 if (wbc->for_kupdate)
973 return 0;
974
e2d84521 975 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 976 /* this is a bit racy, but that's ok */
e2d84521
MX		 977 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
978 BTRFS_DIRTY_METADATA_THRESH);
979 if (ret < 0)
793955bc 980 return 0;
793955bc 981 }
0b32f4bb 982 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM		 983}
984
b2950863 985static int btree_readpage(struct file *file, struct page *page)
5f39d397 986{
d1310b2e
CM		 987 struct extent_io_tree *tree;
988 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 989 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 990}
22b0ebda 991
70dec807 992static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 993{
98509cfc 994 if (PageWriteback(page) || PageDirty(page))
d397712b 995 return 0;
0c4e538b 996
f7a52a40 997 return try_release_extent_buffer(page);
d98237b3
CM		 998}
999
d47992f8
LC		 1000static void btree_invalidatepage(struct page *page, unsigned int offset,
1001 unsigned int length)
d98237b3 1002{
d1310b2e
CM		 1003 struct extent_io_tree *tree;
1004 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 1005 extent_invalidatepage(tree, page, offset);
1006 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 1007 if (PagePrivate(page)) {
efe120a0
FH		 1008 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1009 "page private not zero on page %llu",
1010 (unsigned long long)page_offset(page));
9ad6b7bc
CM		 1011 ClearPagePrivate(page);
1012 set_page_private(page, 0);
1013 page_cache_release(page);
1014 }
d98237b3
CM		 1015}
1016
0b32f4bb
JB		 1017static int btree_set_page_dirty(struct page *page)
1018{
bb146eb2 1019#ifdef DEBUG
0b32f4bb
JB		 1020 struct extent_buffer *eb;
1021
1022 BUG_ON(!PagePrivate(page));
1023 eb = (struct extent_buffer *)page->private;
1024 BUG_ON(!eb);
1025 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1026 BUG_ON(!atomic_read(&eb->refs));
1027 btrfs_assert_tree_locked(eb);
bb146eb2 1028#endif
0b32f4bb
JB		 1029 return __set_page_dirty_nobuffers(page);
1030}
1031
7f09410b 1032static const struct address_space_operations btree_aops = {
d98237b3 1033 .readpage = btree_readpage,
0da5468f 1034 .writepages = btree_writepages,
5f39d397
CM		 1035 .releasepage = btree_releasepage,
1036 .invalidatepage = btree_invalidatepage,
5a92bc88 1037#ifdef CONFIG_MIGRATION
784b4e29 1038 .migratepage = btree_migratepage,
5a92bc88 1039#endif
0b32f4bb 1040 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM		 1041};
1042
ca7a79ad
CM		 1043int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1044 u64 parent_transid)
090d1875 1045{
5f39d397
CM		 1046 struct extent_buffer *buf = NULL;
1047 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 1048 int ret = 0;
090d1875 1049
db94535d 1050 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 1051 if (!buf)
090d1875 1052 return 0;
d1310b2e 1053 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
bb82ab88 1054 buf, 0, WAIT_NONE, btree_get_extent, 0);
5f39d397 1055 free_extent_buffer(buf);
de428b63 1056 return ret;
090d1875
CM		 1057}
1058
ab0fff03
AJ		 1059int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1060 int mirror_num, struct extent_buffer **eb)
1061{
1062 struct extent_buffer *buf = NULL;
1063 struct inode *btree_inode = root->fs_info->btree_inode;
1064 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1065 int ret;
1066
1067 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1068 if (!buf)
1069 return 0;
1070
1071 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1072
1073 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1074 btree_get_extent, mirror_num);
1075 if (ret) {
1076 free_extent_buffer(buf);
1077 return ret;
1078 }
1079
1080 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1081 free_extent_buffer(buf);
1082 return -EIO;
0b32f4bb 1083 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ		 1084 *eb = buf;
1085 } else {
1086 free_extent_buffer(buf);
1087 }
1088 return 0;
1089}
1090
0999df54
CM		 1091struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1092 u64 bytenr, u32 blocksize)
1093{
f28491e0 1094 return find_extent_buffer(root->fs_info, bytenr);
0999df54
CM		 1095}
1096
1097struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1098 u64 bytenr, u32 blocksize)
1099{
f28491e0 1100 return alloc_extent_buffer(root->fs_info, bytenr, blocksize);
0999df54
CM		 1101}
1102
1103
e02119d5
CM		 1104int btrfs_write_tree_block(struct extent_buffer *buf)
1105{
727011e0 1106 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1107 buf->start + buf->len - 1);
e02119d5
CM		 1108}
1109
1110int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1111{
727011e0 1112 return filemap_fdatawait_range(buf->pages[0]->mapping,
8aa38c31 1113 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1114}
1115
0999df54 1116struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 1117 u32 blocksize, u64 parent_transid)
0999df54
CM		 1118{
1119 struct extent_buffer *buf = NULL;
0999df54
CM		 1120 int ret;
1121
0999df54
CM		 1122 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1123 if (!buf)
1124 return NULL;
0999df54 1125
ca7a79ad 1126 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
0f0fe8f7
FDBM		 1127 if (ret) {
1128 free_extent_buffer(buf);
1129 return NULL;
1130 }
5f39d397 1131 return buf;
ce9adaa5 1132
eb60ceac
CM		 1133}
1134
d5c13f92
JM		 1135void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1136 struct extent_buffer *buf)
ed2ff2cb 1137{
e2d84521
MX		 1138 struct btrfs_fs_info *fs_info = root->fs_info;
1139
55c69072 1140 if (btrfs_header_generation(buf) ==
e2d84521 1141 fs_info->running_transaction->transid) {
b9447ef8 1142 btrfs_assert_tree_locked(buf);
b4ce94de 1143
b9473439 1144 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
e2d84521
MX		 1145 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1146 -buf->len,
1147 fs_info->dirty_metadata_batch);
ed7b63eb
JB		 1148 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1149 btrfs_set_lock_blocking(buf);
1150 clear_extent_buffer_dirty(buf);
1151 }
925baedd 1152 }
5f39d397
CM		 1153}
1154
143bede5
JM		 1155static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1156 u32 stripesize, struct btrfs_root *root,
1157 struct btrfs_fs_info *fs_info,
1158 u64 objectid)
d97e63b6 1159{
cfaa7295 1160 root->node = NULL;
a28ec197 1161 root->commit_root = NULL;
db94535d
CM		 1162 root->sectorsize = sectorsize;
1163 root->nodesize = nodesize;
1164 root->leafsize = leafsize;
87ee04eb 1165 root->stripesize = stripesize;
123abc88 1166 root->ref_cows = 0;
0b86a832 1167 root->track_dirty = 0;
c71bf099 1168 root->in_radix = 0;
d68fc57b
YZ		 1169 root->orphan_item_inserted = 0;
1170 root->orphan_cleanup_state = 0;
0b86a832 1171
0f7d52f4
CM		 1172 root->objectid = objectid;
1173 root->last_trans = 0;
13a8a7c8 1174 root->highest_objectid = 0;
eb73c1b7 1175 root->nr_delalloc_inodes = 0;
199c2a9c 1176 root->nr_ordered_extents = 0;
58176a96 1177 root->name = NULL;
6bef4d31 1178 root->inode_tree = RB_ROOT;
16cdcec7 1179 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1180 root->block_rsv = NULL;
d68fc57b 1181 root->orphan_block_rsv = NULL;
0b86a832
CM		 1182
1183 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1184 INIT_LIST_HEAD(&root->root_list);
eb73c1b7
MX		 1185 INIT_LIST_HEAD(&root->delalloc_inodes);
1186 INIT_LIST_HEAD(&root->delalloc_root);
199c2a9c
MX		 1187 INIT_LIST_HEAD(&root->ordered_extents);
1188 INIT_LIST_HEAD(&root->ordered_root);
2ab28f32
JB		 1189 INIT_LIST_HEAD(&root->logged_list[0]);
1190 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1191 spin_lock_init(&root->orphan_lock);
5d4f98a2 1192 spin_lock_init(&root->inode_lock);
eb73c1b7 1193 spin_lock_init(&root->delalloc_lock);
199c2a9c 1194 spin_lock_init(&root->ordered_extent_lock);
f0486c68 1195 spin_lock_init(&root->accounting_lock);
2ab28f32
JB		 1196 spin_lock_init(&root->log_extents_lock[0]);
1197 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1198 mutex_init(&root->objectid_mutex);
e02119d5 1199 mutex_init(&root->log_mutex);
7237f183
YZ		 1200 init_waitqueue_head(&root->log_writer_wait);
1201 init_waitqueue_head(&root->log_commit_wait[0]);
1202 init_waitqueue_head(&root->log_commit_wait[1]);
8b050d35
MX		 1203 INIT_LIST_HEAD(&root->log_ctxs[0]);
1204 INIT_LIST_HEAD(&root->log_ctxs[1]);
7237f183
YZ		 1205 atomic_set(&root->log_commit[0], 0);
1206 atomic_set(&root->log_commit[1], 0);
1207 atomic_set(&root->log_writers, 0);
2ecb7923 1208 atomic_set(&root->log_batch, 0);
8a35d95f 1209 atomic_set(&root->orphan_inodes, 0);
b0feb9d9 1210 atomic_set(&root->refs, 1);
7237f183 1211 root->log_transid = 0;
d1433deb 1212 root->log_transid_committed = -1;
257c62e1 1213 root->last_log_commit = 0;
06ea65a3
JB		 1214 if (fs_info)
1215 extent_io_tree_init(&root->dirty_log_pages,
1216 fs_info->btree_inode->i_mapping);
017e5369 1217
3768f368
CM		 1218 memset(&root->root_key, 0, sizeof(root->root_key));
1219 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1220 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1221 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
06ea65a3
JB		 1222 if (fs_info)
1223 root->defrag_trans_start = fs_info->generation;
1224 else
1225 root->defrag_trans_start = 0;
58176a96 1226 init_completion(&root->kobj_unregister);
6702ed49 1227 root->defrag_running = 0;
4d775673 1228 root->root_key.objectid = objectid;
0ee5dc67 1229 root->anon_dev = 0;
8ea05e3a 1230
5f3ab90a 1231 spin_lock_init(&root->root_item_lock);
3768f368
CM		 1232}
1233
f84a8bd6 1234static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
6f07e42e
AV		 1235{
1236 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1237 if (root)
1238 root->fs_info = fs_info;
1239 return root;
1240}
1241
06ea65a3
JB		 1242#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1243/* Should only be used by the testing infrastructure */
1244struct btrfs_root *btrfs_alloc_dummy_root(void)
1245{
1246 struct btrfs_root *root;
1247
1248 root = btrfs_alloc_root(NULL);
1249 if (!root)
1250 return ERR_PTR(-ENOMEM);
1251 __setup_root(4096, 4096, 4096, 4096, root, NULL, 1);
1252 root->dummy_root = 1;
1253
1254 return root;
1255}
1256#endif
1257
20897f5c
AJ		 1258struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1259 struct btrfs_fs_info *fs_info,
1260 u64 objectid)
1261{
1262 struct extent_buffer *leaf;
1263 struct btrfs_root *tree_root = fs_info->tree_root;
1264 struct btrfs_root *root;
1265 struct btrfs_key key;
1266 int ret = 0;
6463fe58 1267 uuid_le uuid;
20897f5c
AJ		 1268
1269 root = btrfs_alloc_root(fs_info);
1270 if (!root)
1271 return ERR_PTR(-ENOMEM);
1272
1273 __setup_root(tree_root->nodesize, tree_root->leafsize,
1274 tree_root->sectorsize, tree_root->stripesize,
1275 root, fs_info, objectid);
1276 root->root_key.objectid = objectid;
1277 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1278 root->root_key.offset = 0;
1279
1280 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1281 0, objectid, NULL, 0, 0, 0);
1282 if (IS_ERR(leaf)) {
1283 ret = PTR_ERR(leaf);
1dd05682 1284 leaf = NULL;
20897f5c
AJ		 1285 goto fail;
1286 }
1287
20897f5c
AJ		 1288 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1289 btrfs_set_header_bytenr(leaf, leaf->start);
1290 btrfs_set_header_generation(leaf, trans->transid);
1291 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1292 btrfs_set_header_owner(leaf, objectid);
1293 root->node = leaf;
1294
0a4e5586 1295 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
20897f5c
AJ		 1296 BTRFS_FSID_SIZE);
1297 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
b308bc2f 1298 btrfs_header_chunk_tree_uuid(leaf),
20897f5c
AJ		 1299 BTRFS_UUID_SIZE);
1300 btrfs_mark_buffer_dirty(leaf);
1301
1302 root->commit_root = btrfs_root_node(root);
1303 root->track_dirty = 1;
1304
1305
1306 root->root_item.flags = 0;
1307 root->root_item.byte_limit = 0;
1308 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1309 btrfs_set_root_generation(&root->root_item, trans->transid);
1310 btrfs_set_root_level(&root->root_item, 0);
1311 btrfs_set_root_refs(&root->root_item, 1);
1312 btrfs_set_root_used(&root->root_item, leaf->len);
1313 btrfs_set_root_last_snapshot(&root->root_item, 0);
1314 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB		 1315 uuid_le_gen(&uuid);
1316 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ		 1317 root->root_item.drop_level = 0;
1318
1319 key.objectid = objectid;
1320 key.type = BTRFS_ROOT_ITEM_KEY;
1321 key.offset = 0;
1322 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1323 if (ret)
1324 goto fail;
1325
1326 btrfs_tree_unlock(leaf);
1327
1dd05682
TI		 1328 return root;
1329
20897f5c 1330fail:
1dd05682
TI		 1331 if (leaf) {
1332 btrfs_tree_unlock(leaf);
1333 free_extent_buffer(leaf);
1334 }
1335 kfree(root);
20897f5c 1336
1dd05682 1337 return ERR_PTR(ret);
20897f5c
AJ		 1338}
1339
7237f183
YZ		 1340static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1341 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1342{
1343 struct btrfs_root *root;
1344 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1345 struct extent_buffer *leaf;
e02119d5 1346
6f07e42e 1347 root = btrfs_alloc_root(fs_info);
e02119d5 1348 if (!root)
7237f183 1349 return ERR_PTR(-ENOMEM);
e02119d5
CM		 1350
1351 __setup_root(tree_root->nodesize, tree_root->leafsize,
1352 tree_root->sectorsize, tree_root->stripesize,
1353 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1354
1355 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1356 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1357 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
7237f183
YZ		 1358 /*
1359 * log trees do not get reference counted because they go away
1360 * before a real commit is actually done. They do store pointers
1361 * to file data extents, and those reference counts still get
1362 * updated (along with back refs to the log tree).
1363 */
e02119d5
CM		 1364 root->ref_cows = 0;
1365
5d4f98a2 1366 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
66d7e7f0 1367 BTRFS_TREE_LOG_OBJECTID, NULL,
5581a51a 1368 0, 0, 0);
7237f183
YZ		 1369 if (IS_ERR(leaf)) {
1370 kfree(root);
1371 return ERR_CAST(leaf);
1372 }
e02119d5 1373
5d4f98a2
YZ		 1374 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1375 btrfs_set_header_bytenr(leaf, leaf->start);
1376 btrfs_set_header_generation(leaf, trans->transid);
1377 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1378 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1379 root->node = leaf;
e02119d5
CM		 1380
1381 write_extent_buffer(root->node, root->fs_info->fsid,
0a4e5586 1382 btrfs_header_fsid(), BTRFS_FSID_SIZE);
e02119d5
CM		 1383 btrfs_mark_buffer_dirty(root->node);
1384 btrfs_tree_unlock(root->node);
7237f183
YZ		 1385 return root;
1386}
1387
1388int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1389 struct btrfs_fs_info *fs_info)
1390{
1391 struct btrfs_root *log_root;
1392
1393 log_root = alloc_log_tree(trans, fs_info);
1394 if (IS_ERR(log_root))
1395 return PTR_ERR(log_root);
1396 WARN_ON(fs_info->log_root_tree);
1397 fs_info->log_root_tree = log_root;
1398 return 0;
1399}
1400
1401int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1402 struct btrfs_root *root)
1403{
1404 struct btrfs_root *log_root;
1405 struct btrfs_inode_item *inode_item;
1406
1407 log_root = alloc_log_tree(trans, root->fs_info);
1408 if (IS_ERR(log_root))
1409 return PTR_ERR(log_root);
1410
1411 log_root->last_trans = trans->transid;
1412 log_root->root_key.offset = root->root_key.objectid;
1413
1414 inode_item = &log_root->root_item.inode;
3cae210f
QW		 1415 btrfs_set_stack_inode_generation(inode_item, 1);
1416 btrfs_set_stack_inode_size(inode_item, 3);
1417 btrfs_set_stack_inode_nlink(inode_item, 1);
1418 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
1419 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
7237f183 1420
5d4f98a2 1421 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1422
1423 WARN_ON(root->log_root);
1424 root->log_root = log_root;
1425 root->log_transid = 0;
d1433deb 1426 root->log_transid_committed = -1;
257c62e1 1427 root->last_log_commit = 0;
e02119d5
CM		 1428 return 0;
1429}
1430
35a3621b
SB		 1431static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1432 struct btrfs_key *key)
e02119d5
CM		 1433{
1434 struct btrfs_root *root;
1435 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1436 struct btrfs_path *path;
84234f3a 1437 u64 generation;
db94535d 1438 u32 blocksize;
cb517eab 1439 int ret;
0f7d52f4 1440
cb517eab
MX		 1441 path = btrfs_alloc_path();
1442 if (!path)
0f7d52f4 1443 return ERR_PTR(-ENOMEM);
cb517eab
MX		 1444
1445 root = btrfs_alloc_root(fs_info);
1446 if (!root) {
1447 ret = -ENOMEM;
1448 goto alloc_fail;
0f7d52f4
CM		 1449 }
1450
db94535d 1451 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb 1452 tree_root->sectorsize, tree_root->stripesize,
cb517eab 1453 root, fs_info, key->objectid);
0f7d52f4 1454
cb517eab
MX		 1455 ret = btrfs_find_root(tree_root, key, path,
1456 &root->root_item, &root->root_key);
0f7d52f4 1457 if (ret) {
13a8a7c8
YZ		 1458 if (ret > 0)
1459 ret = -ENOENT;
cb517eab 1460 goto find_fail;
0f7d52f4 1461 }
13a8a7c8 1462
84234f3a 1463 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1464 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1465 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1466 blocksize, generation);
cb517eab
MX		 1467 if (!root->node) {
1468 ret = -ENOMEM;
1469 goto find_fail;
1470 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1471 ret = -EIO;
1472 goto read_fail;
416bc658 1473 }
5d4f98a2 1474 root->commit_root = btrfs_root_node(root);
13a8a7c8 1475out:
cb517eab
MX		 1476 btrfs_free_path(path);
1477 return root;
1478
1479read_fail:
1480 free_extent_buffer(root->node);
1481find_fail:
1482 kfree(root);
1483alloc_fail:
1484 root = ERR_PTR(ret);
1485 goto out;
1486}
1487
1488struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1489 struct btrfs_key *location)
1490{
1491 struct btrfs_root *root;
1492
1493 root = btrfs_read_tree_root(tree_root, location);
1494 if (IS_ERR(root))
1495 return root;
1496
1497 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
e02119d5 1498 root->ref_cows = 1;
08fe4db1
LZ		 1499 btrfs_check_and_init_root_item(&root->root_item);
1500 }
13a8a7c8 1501
5eda7b5e
CM		 1502 return root;
1503}
1504
cb517eab
MX		 1505int btrfs_init_fs_root(struct btrfs_root *root)
1506{
1507 int ret;
1508
1509 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1510 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1511 GFP_NOFS);
1512 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1513 ret = -ENOMEM;
1514 goto fail;
1515 }
1516
1517 btrfs_init_free_ino_ctl(root);
1518 mutex_init(&root->fs_commit_mutex);
1519 spin_lock_init(&root->cache_lock);
1520 init_waitqueue_head(&root->cache_wait);
1521
1522 ret = get_anon_bdev(&root->anon_dev);
1523 if (ret)
1524 goto fail;
1525 return 0;
1526fail:
1527 kfree(root->free_ino_ctl);
1528 kfree(root->free_ino_pinned);
1529 return ret;
1530}
1531
171170c1
ST		 1532static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1533 u64 root_id)
cb517eab
MX		 1534{
1535 struct btrfs_root *root;
1536
1537 spin_lock(&fs_info->fs_roots_radix_lock);
1538 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1539 (unsigned long)root_id);
1540 spin_unlock(&fs_info->fs_roots_radix_lock);
1541 return root;
1542}
1543
1544int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1545 struct btrfs_root *root)
1546{
1547 int ret;
1548
1549 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1550 if (ret)
1551 return ret;
1552
1553 spin_lock(&fs_info->fs_roots_radix_lock);
1554 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1555 (unsigned long)root->root_key.objectid,
1556 root);
1557 if (ret == 0)
1558 root->in_radix = 1;
1559 spin_unlock(&fs_info->fs_roots_radix_lock);
1560 radix_tree_preload_end();
1561
1562 return ret;
1563}
1564
c00869f1
MX		 1565struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1566 struct btrfs_key *location,
1567 bool check_ref)
5eda7b5e
CM		 1568{
1569 struct btrfs_root *root;
1570 int ret;
1571
edbd8d4e
CM		 1572 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1573 return fs_info->tree_root;
1574 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1575 return fs_info->extent_root;
8f18cf13
CM		 1576 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1577 return fs_info->chunk_root;
1578 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1579 return fs_info->dev_root;
0403e47e
YZ		 1580 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1581 return fs_info->csum_root;
bcef60f2
AJ		 1582 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1583 return fs_info->quota_root ? fs_info->quota_root :
1584 ERR_PTR(-ENOENT);
f7a81ea4
SB		 1585 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1586 return fs_info->uuid_root ? fs_info->uuid_root :
1587 ERR_PTR(-ENOENT);
4df27c4d 1588again:
cb517eab 1589 root = btrfs_lookup_fs_root(fs_info, location->objectid);
48475471 1590 if (root) {
c00869f1 1591 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
48475471 1592 return ERR_PTR(-ENOENT);
5eda7b5e 1593 return root;
48475471 1594 }
5eda7b5e 1595
cb517eab 1596 root = btrfs_read_fs_root(fs_info->tree_root, location);
5eda7b5e
CM		 1597 if (IS_ERR(root))
1598 return root;
3394e160 1599
c00869f1 1600 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
cb517eab 1601 ret = -ENOENT;
581bb050 1602 goto fail;
35a30d7c 1603 }
581bb050 1604
cb517eab 1605 ret = btrfs_init_fs_root(root);
ac08aedf
CM		 1606 if (ret)
1607 goto fail;
3394e160 1608
3f870c28
KN		 1609 ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID,
1610 location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL);
d68fc57b
YZ		 1611 if (ret < 0)
1612 goto fail;
1613 if (ret == 0)
1614 root->orphan_item_inserted = 1;
1615
cb517eab 1616 ret = btrfs_insert_fs_root(fs_info, root);
0f7d52f4 1617 if (ret) {
4df27c4d
YZ		 1618 if (ret == -EEXIST) {
1619 free_fs_root(root);
1620 goto again;
1621 }
1622 goto fail;
0f7d52f4 1623 }
edbd8d4e 1624 return root;
4df27c4d
YZ		 1625fail:
1626 free_fs_root(root);
1627 return ERR_PTR(ret);
edbd8d4e
CM		 1628}
1629
04160088
CM		 1630static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1631{
1632 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1633 int ret = 0;
04160088
CM		 1634 struct btrfs_device *device;
1635 struct backing_dev_info *bdi;
b7967db7 1636
1f78160c
XG		 1637 rcu_read_lock();
1638 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1639 if (!device->bdev)
1640 continue;
04160088
CM		 1641 bdi = blk_get_backing_dev_info(device->bdev);
1642 if (bdi && bdi_congested(bdi, bdi_bits)) {
1643 ret = 1;
1644 break;
1645 }
1646 }
1f78160c 1647 rcu_read_unlock();
04160088
CM		 1648 return ret;
1649}
1650
ad081f14
JA		 1651/*
1652 * If this fails, caller must call bdi_destroy() to get rid of the
1653 * bdi again.
1654 */
04160088
CM		 1655static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1656{
ad081f14
JA		 1657 int err;
1658
1659 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1660 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1661 if (err)
1662 return err;
1663
4575c9cc 1664 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1665 bdi->congested_fn = btrfs_congested_fn;
1666 bdi->congested_data = info;
1667 return 0;
1668}
1669
8b712842
CM		 1670/*
1671 * called by the kthread helper functions to finally call the bio end_io
1672 * functions. This is where read checksum verification actually happens
1673 */
1674static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1675{
ce9adaa5 1676 struct bio *bio;
8b712842 1677 struct end_io_wq *end_io_wq;
ce9adaa5 1678 int error;
ce9adaa5 1679
8b712842
CM		 1680 end_io_wq = container_of(work, struct end_io_wq, work);
1681 bio = end_io_wq->bio;
ce9adaa5 1682
8b712842
CM		 1683 error = end_io_wq->error;
1684 bio->bi_private = end_io_wq->private;
1685 bio->bi_end_io = end_io_wq->end_io;
1686 kfree(end_io_wq);
8b712842 1687 bio_endio(bio, error);
44b8bd7e
CM		 1688}
1689
a74a4b97
CM		 1690static int cleaner_kthread(void *arg)
1691{
1692 struct btrfs_root *root = arg;
d0278245 1693 int again;
a74a4b97
CM		 1694
1695 do {
d0278245 1696 again = 0;
a74a4b97 1697
d0278245 1698 /* Make the cleaner go to sleep early. */
babbf170 1699 if (btrfs_need_cleaner_sleep(root))
d0278245
MX		 1700 goto sleep;
1701
1702 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1703 goto sleep;
1704
dc7f370c
MX		 1705 /*
1706 * Avoid the problem that we change the status of the fs
1707 * during the above check and trylock.
1708 */
babbf170 1709 if (btrfs_need_cleaner_sleep(root)) {
dc7f370c
MX		 1710 mutex_unlock(&root->fs_info->cleaner_mutex);
1711 goto sleep;
76dda93c 1712 }
a74a4b97 1713
d0278245
MX		 1714 btrfs_run_delayed_iputs(root);
1715 again = btrfs_clean_one_deleted_snapshot(root);
1716 mutex_unlock(&root->fs_info->cleaner_mutex);
1717
1718 /*
05323cd1
MX		 1719 * The defragger has dealt with the R/O remount and umount,
1720 * needn't do anything special here.
d0278245
MX		 1721 */
1722 btrfs_run_defrag_inodes(root->fs_info);
1723sleep:
9d1a2a3a 1724 if (!try_to_freeze() && !again) {
a74a4b97 1725 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1726 if (!kthread_should_stop())
1727 schedule();
a74a4b97
CM		 1728 __set_current_state(TASK_RUNNING);
1729 }
1730 } while (!kthread_should_stop());
1731 return 0;
1732}
1733
1734static int transaction_kthread(void *arg)
1735{
1736 struct btrfs_root *root = arg;
1737 struct btrfs_trans_handle *trans;
1738 struct btrfs_transaction *cur;
8929ecfa 1739 u64 transid;
a74a4b97
CM		 1740 unsigned long now;
1741 unsigned long delay;
914b2007 1742 bool cannot_commit;
a74a4b97
CM		 1743
1744 do {
914b2007 1745 cannot_commit = false;
8b87dc17 1746 delay = HZ * root->fs_info->commit_interval;
a74a4b97
CM		 1747 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1748
a4abeea4 1749 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1750 cur = root->fs_info->running_transaction;
1751 if (!cur) {
a4abeea4 1752 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1753 goto sleep;
1754 }
31153d81 1755
a74a4b97 1756 now = get_seconds();
4a9d8bde 1757 if (cur->state < TRANS_STATE_BLOCKED &&
8b87dc17
DS		 1758 (now < cur->start_time ||
1759 now - cur->start_time < root->fs_info->commit_interval)) {
a4abeea4 1760 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1761 delay = HZ * 5;
1762 goto sleep;
1763 }
8929ecfa 1764 transid = cur->transid;
a4abeea4 1765 spin_unlock(&root->fs_info->trans_lock);
56bec294 1766
79787eaa 1767 /* If the file system is aborted, this will always fail. */
354aa0fb 1768 trans = btrfs_attach_transaction(root);
914b2007 1769 if (IS_ERR(trans)) {
354aa0fb
MX		 1770 if (PTR_ERR(trans) != -ENOENT)
1771 cannot_commit = true;
79787eaa 1772 goto sleep;
914b2007 1773 }
8929ecfa 1774 if (transid == trans->transid) {
79787eaa 1775 btrfs_commit_transaction(trans, root);
8929ecfa
YZ		 1776 } else {
1777 btrfs_end_transaction(trans, root);
1778 }
a74a4b97
CM		 1779sleep:
1780 wake_up_process(root->fs_info->cleaner_kthread);
1781 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1782
4e121c06
JB		 1783 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1784 &root->fs_info->fs_state)))
1785 btrfs_cleanup_transaction(root);
a0acae0e 1786 if (!try_to_freeze()) {
a74a4b97 1787 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa 1788 if (!kthread_should_stop() &&
914b2007
JK		 1789 (!btrfs_transaction_blocked(root->fs_info) ||
1790 cannot_commit))
8929ecfa 1791 schedule_timeout(delay);
a74a4b97
CM		 1792 __set_current_state(TASK_RUNNING);
1793 }
1794 } while (!kthread_should_stop());
1795 return 0;
1796}
1797
af31f5e5
CM		 1798/*
1799 * this will find the highest generation in the array of
1800 * root backups. The index of the highest array is returned,
1801 * or -1 if we can't find anything.
1802 *
1803 * We check to make sure the array is valid by comparing the
1804 * generation of the latest root in the array with the generation
1805 * in the super block. If they don't match we pitch it.
1806 */
1807static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1808{
1809 u64 cur;
1810 int newest_index = -1;
1811 struct btrfs_root_backup *root_backup;
1812 int i;
1813
1814 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1815 root_backup = info->super_copy->super_roots + i;
1816 cur = btrfs_backup_tree_root_gen(root_backup);
1817 if (cur == newest_gen)
1818 newest_index = i;
1819 }
1820
1821 /* check to see if we actually wrapped around */
1822 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1823 root_backup = info->super_copy->super_roots;
1824 cur = btrfs_backup_tree_root_gen(root_backup);
1825 if (cur == newest_gen)
1826 newest_index = 0;
1827 }
1828 return newest_index;
1829}
1830
1831
1832/*
1833 * find the oldest backup so we know where to store new entries
1834 * in the backup array. This will set the backup_root_index
1835 * field in the fs_info struct
1836 */
1837static void find_oldest_super_backup(struct btrfs_fs_info *info,
1838 u64 newest_gen)
1839{
1840 int newest_index = -1;
1841
1842 newest_index = find_newest_super_backup(info, newest_gen);
1843 /* if there was garbage in there, just move along */
1844 if (newest_index == -1) {
1845 info->backup_root_index = 0;
1846 } else {
1847 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1848 }
1849}
1850
1851/*
1852 * copy all the root pointers into the super backup array.
1853 * this will bump the backup pointer by one when it is
1854 * done
1855 */
1856static void backup_super_roots(struct btrfs_fs_info *info)
1857{
1858 int next_backup;
1859 struct btrfs_root_backup *root_backup;
1860 int last_backup;
1861
1862 next_backup = info->backup_root_index;
1863 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1864 BTRFS_NUM_BACKUP_ROOTS;
1865
1866 /*
1867 * just overwrite the last backup if we're at the same generation
1868 * this happens only at umount
1869 */
1870 root_backup = info->super_for_commit->super_roots + last_backup;
1871 if (btrfs_backup_tree_root_gen(root_backup) ==
1872 btrfs_header_generation(info->tree_root->node))
1873 next_backup = last_backup;
1874
1875 root_backup = info->super_for_commit->super_roots + next_backup;
1876
1877 /*
1878 * make sure all of our padding and empty slots get zero filled
1879 * regardless of which ones we use today
1880 */
1881 memset(root_backup, 0, sizeof(*root_backup));
1882
1883 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1884
1885 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1886 btrfs_set_backup_tree_root_gen(root_backup,
1887 btrfs_header_generation(info->tree_root->node));
1888
1889 btrfs_set_backup_tree_root_level(root_backup,
1890 btrfs_header_level(info->tree_root->node));
1891
1892 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1893 btrfs_set_backup_chunk_root_gen(root_backup,
1894 btrfs_header_generation(info->chunk_root->node));
1895 btrfs_set_backup_chunk_root_level(root_backup,
1896 btrfs_header_level(info->chunk_root->node));
1897
1898 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1899 btrfs_set_backup_extent_root_gen(root_backup,
1900 btrfs_header_generation(info->extent_root->node));
1901 btrfs_set_backup_extent_root_level(root_backup,
1902 btrfs_header_level(info->extent_root->node));
1903
7c7e82a7
CM		 1904 /*
1905 * we might commit during log recovery, which happens before we set
1906 * the fs_root. Make sure it is valid before we fill it in.
1907 */
1908 if (info->fs_root && info->fs_root->node) {
1909 btrfs_set_backup_fs_root(root_backup,
1910 info->fs_root->node->start);
1911 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 1912 btrfs_header_generation(info->fs_root->node));
7c7e82a7 1913 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 1914 btrfs_header_level(info->fs_root->node));
7c7e82a7 1915 }
af31f5e5
CM		 1916
1917 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1918 btrfs_set_backup_dev_root_gen(root_backup,
1919 btrfs_header_generation(info->dev_root->node));
1920 btrfs_set_backup_dev_root_level(root_backup,
1921 btrfs_header_level(info->dev_root->node));
1922
1923 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1924 btrfs_set_backup_csum_root_gen(root_backup,
1925 btrfs_header_generation(info->csum_root->node));
1926 btrfs_set_backup_csum_root_level(root_backup,
1927 btrfs_header_level(info->csum_root->node));
1928
1929 btrfs_set_backup_total_bytes(root_backup,
1930 btrfs_super_total_bytes(info->super_copy));
1931 btrfs_set_backup_bytes_used(root_backup,
1932 btrfs_super_bytes_used(info->super_copy));
1933 btrfs_set_backup_num_devices(root_backup,
1934 btrfs_super_num_devices(info->super_copy));
1935
1936 /*
1937 * if we don't copy this out to the super_copy, it won't get remembered
1938 * for the next commit
1939 */
1940 memcpy(&info->super_copy->super_roots,
1941 &info->super_for_commit->super_roots,
1942 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1943}
1944
1945/*
1946 * this copies info out of the root backup array and back into
1947 * the in-memory super block. It is meant to help iterate through
1948 * the array, so you send it the number of backups you've already
1949 * tried and the last backup index you used.
1950 *
1951 * this returns -1 when it has tried all the backups
1952 */
1953static noinline int next_root_backup(struct btrfs_fs_info *info,
1954 struct btrfs_super_block *super,
1955 int *num_backups_tried, int *backup_index)
1956{
1957 struct btrfs_root_backup *root_backup;
1958 int newest = *backup_index;
1959
1960 if (*num_backups_tried == 0) {
1961 u64 gen = btrfs_super_generation(super);
1962
1963 newest = find_newest_super_backup(info, gen);
1964 if (newest == -1)
1965 return -1;
1966
1967 *backup_index = newest;
1968 *num_backups_tried = 1;
1969 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
1970 /* we've tried all the backups, all done */
1971 return -1;
1972 } else {
1973 /* jump to the next oldest backup */
1974 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
1975 BTRFS_NUM_BACKUP_ROOTS;
1976 *backup_index = newest;
1977 *num_backups_tried += 1;
1978 }
1979 root_backup = super->super_roots + newest;
1980
1981 btrfs_set_super_generation(super,
1982 btrfs_backup_tree_root_gen(root_backup));
1983 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
1984 btrfs_set_super_root_level(super,
1985 btrfs_backup_tree_root_level(root_backup));
1986 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
1987
1988 /*
1989 * fixme: the total bytes and num_devices need to match or we should
1990 * need a fsck
1991 */
1992 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
1993 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
1994 return 0;
1995}
1996
7abadb64
LB		 1997/* helper to cleanup workers */
1998static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
1999{
2000 btrfs_stop_workers(&fs_info->generic_worker);
2001 btrfs_stop_workers(&fs_info->fixup_workers);
2002 btrfs_stop_workers(&fs_info->delalloc_workers);
2003 btrfs_stop_workers(&fs_info->workers);
2004 btrfs_stop_workers(&fs_info->endio_workers);
2005 btrfs_stop_workers(&fs_info->endio_meta_workers);
2006 btrfs_stop_workers(&fs_info->endio_raid56_workers);
2007 btrfs_stop_workers(&fs_info->rmw_workers);
2008 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
2009 btrfs_stop_workers(&fs_info->endio_write_workers);
2010 btrfs_stop_workers(&fs_info->endio_freespace_worker);
2011 btrfs_stop_workers(&fs_info->submit_workers);
2012 btrfs_stop_workers(&fs_info->delayed_workers);
2013 btrfs_stop_workers(&fs_info->caching_workers);
2014 btrfs_stop_workers(&fs_info->readahead_workers);
2015 btrfs_stop_workers(&fs_info->flush_workers);
2f232036 2016 btrfs_stop_workers(&fs_info->qgroup_rescan_workers);
7abadb64
LB		 2017}
2018
2e9f5954
R		 2019static void free_root_extent_buffers(struct btrfs_root *root)
2020{
2021 if (root) {
2022 free_extent_buffer(root->node);
2023 free_extent_buffer(root->commit_root);
2024 root->node = NULL;
2025 root->commit_root = NULL;
2026 }
2027}
2028
af31f5e5
CM		 2029/* helper to cleanup tree roots */
2030static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2031{
2e9f5954 2032 free_root_extent_buffers(info->tree_root);
655b09fe 2033
2e9f5954
R		 2034 free_root_extent_buffers(info->dev_root);
2035 free_root_extent_buffers(info->extent_root);
2036 free_root_extent_buffers(info->csum_root);
2037 free_root_extent_buffers(info->quota_root);
2038 free_root_extent_buffers(info->uuid_root);
2039 if (chunk_root)
2040 free_root_extent_buffers(info->chunk_root);
af31f5e5
CM		 2041}
2042
171f6537
JB		 2043static void del_fs_roots(struct btrfs_fs_info *fs_info)
2044{
2045 int ret;
2046 struct btrfs_root *gang[8];
2047 int i;
2048
2049 while (!list_empty(&fs_info->dead_roots)) {
2050 gang[0] = list_entry(fs_info->dead_roots.next,
2051 struct btrfs_root, root_list);
2052 list_del(&gang[0]->root_list);
2053
2054 if (gang[0]->in_radix) {
cb517eab 2055 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
171f6537
JB		 2056 } else {
2057 free_extent_buffer(gang[0]->node);
2058 free_extent_buffer(gang[0]->commit_root);
b0feb9d9 2059 btrfs_put_fs_root(gang[0]);
171f6537
JB		 2060 }
2061 }
2062
2063 while (1) {
2064 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2065 (void **)gang, 0,
2066 ARRAY_SIZE(gang));
2067 if (!ret)
2068 break;
2069 for (i = 0; i < ret; i++)
cb517eab 2070 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
171f6537 2071 }
1a4319cc
LB		 2072
2073 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2074 btrfs_free_log_root_tree(NULL, fs_info);
2075 btrfs_destroy_pinned_extent(fs_info->tree_root,
2076 fs_info->pinned_extents);
2077 }
171f6537 2078}
af31f5e5 2079
ad2b2c80
AV		 2080int open_ctree(struct super_block *sb,
2081 struct btrfs_fs_devices *fs_devices,
2082 char *options)
2e635a27 2083{
db94535d
CM		 2084 u32 sectorsize;
2085 u32 nodesize;
2086 u32 leafsize;
2087 u32 blocksize;
87ee04eb 2088 u32 stripesize;
84234f3a 2089 u64 generation;
f2b636e8 2090 u64 features;
3de4586c 2091 struct btrfs_key location;
a061fc8d 2092 struct buffer_head *bh;
4d34b278 2093 struct btrfs_super_block *disk_super;
815745cf 2094 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2095 struct btrfs_root *tree_root;
4d34b278
ID		 2096 struct btrfs_root *extent_root;
2097 struct btrfs_root *csum_root;
2098 struct btrfs_root *chunk_root;
2099 struct btrfs_root *dev_root;
bcef60f2 2100 struct btrfs_root *quota_root;
f7a81ea4 2101 struct btrfs_root *uuid_root;
e02119d5 2102 struct btrfs_root *log_tree_root;
eb60ceac 2103 int ret;
e58ca020 2104 int err = -EINVAL;
af31f5e5
CM		 2105 int num_backups_tried = 0;
2106 int backup_index = 0;
70f80175
SB		 2107 bool create_uuid_tree;
2108 bool check_uuid_tree;
4543df7e 2109
f84a8bd6 2110 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
6f07e42e 2111 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
cb517eab 2112 if (!tree_root || !chunk_root) {
39279cc3
CM		 2113 err = -ENOMEM;
2114 goto fail;
2115 }
76dda93c
YZ		 2116
2117 ret = init_srcu_struct(&fs_info->subvol_srcu);
2118 if (ret) {
2119 err = ret;
2120 goto fail;
2121 }
2122
2123 ret = setup_bdi(fs_info, &fs_info->bdi);
2124 if (ret) {
2125 err = ret;
2126 goto fail_srcu;
2127 }
2128
e2d84521
MX		 2129 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2130 if (ret) {
2131 err = ret;
2132 goto fail_bdi;
2133 }
2134 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2135 (1 + ilog2(nr_cpu_ids));
2136
963d678b
MX		 2137 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2138 if (ret) {
2139 err = ret;
2140 goto fail_dirty_metadata_bytes;
2141 }
2142
c404e0dc
MX		 2143 ret = percpu_counter_init(&fs_info->bio_counter, 0);
2144 if (ret) {
2145 err = ret;
2146 goto fail_delalloc_bytes;
2147 }
2148
76dda93c
YZ		 2149 fs_info->btree_inode = new_inode(sb);
2150 if (!fs_info->btree_inode) {
2151 err = -ENOMEM;
c404e0dc 2152 goto fail_bio_counter;
76dda93c
YZ		 2153 }
2154
a6591715 2155 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2156
76dda93c 2157 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
f28491e0 2158 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
8fd17795 2159 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2160 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2161 INIT_LIST_HEAD(&fs_info->delayed_iputs);
eb73c1b7 2162 INIT_LIST_HEAD(&fs_info->delalloc_roots);
11833d66 2163 INIT_LIST_HEAD(&fs_info->caching_block_groups);
eb73c1b7 2164 spin_lock_init(&fs_info->delalloc_root_lock);
a4abeea4 2165 spin_lock_init(&fs_info->trans_lock);
76dda93c 2166 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2167 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2168 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 2169 spin_lock_init(&fs_info->free_chunk_lock);
f29021b2 2170 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2171 spin_lock_init(&fs_info->super_lock);
f28491e0 2172 spin_lock_init(&fs_info->buffer_lock);
f29021b2 2173 rwlock_init(&fs_info->tree_mod_log_lock);
7585717f 2174 mutex_init(&fs_info->reloc_mutex);
de98ced9 2175 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2176
58176a96 2177 init_completion(&fs_info->kobj_unregister);
0b86a832 2178 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2179 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2180 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
0b86a832 2181 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX		 2182 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2183 BTRFS_BLOCK_RSV_GLOBAL);
2184 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2185 BTRFS_BLOCK_RSV_DELALLOC);
2186 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2187 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2188 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2189 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2190 BTRFS_BLOCK_RSV_DELOPS);
cb03c743 2191 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 2192 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 2193 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 2194 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 2195 atomic_set(&fs_info->defrag_running, 0);
fc36ed7e 2196 atomic64_set(&fs_info->tree_mod_seq, 0);
e20d96d6 2197 fs_info->sb = sb;
6f568d35 2198 fs_info->max_inline = 8192 * 1024;
9ed74f2d 2199 fs_info->metadata_ratio = 0;
4cb5300b 2200 fs_info->defrag_inodes = RB_ROOT;
2bf64758 2201 fs_info->free_chunk_space = 0;
f29021b2 2202 fs_info->tree_mod_log = RB_ROOT;
8b87dc17 2203 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
0a2b2a84 2204 fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
90519d66
AJ		 2205 /* readahead state */
2206 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2207 spin_lock_init(&fs_info->reada_lock);
c8b97818 2208
b34b086c
CM		 2209 fs_info->thread_pool_size = min_t(unsigned long,
2210 num_online_cpus() + 2, 8);
0afbaf8c 2211
199c2a9c
MX		 2212 INIT_LIST_HEAD(&fs_info->ordered_roots);
2213 spin_lock_init(&fs_info->ordered_root_lock);
16cdcec7
MX		 2214 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2215 GFP_NOFS);
2216 if (!fs_info->delayed_root) {
2217 err = -ENOMEM;
2218 goto fail_iput;
2219 }
2220 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2221
a2de733c
AJ		 2222 mutex_init(&fs_info->scrub_lock);
2223 atomic_set(&fs_info->scrubs_running, 0);
2224 atomic_set(&fs_info->scrub_pause_req, 0);
2225 atomic_set(&fs_info->scrubs_paused, 0);
2226 atomic_set(&fs_info->scrub_cancel_req, 0);
c404e0dc 2227 init_waitqueue_head(&fs_info->replace_wait);
a2de733c 2228 init_waitqueue_head(&fs_info->scrub_pause_wait);
a2de733c 2229 fs_info->scrub_workers_refcnt = 0;
21adbd5c
SB		 2230#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2231 fs_info->check_integrity_print_mask = 0;
2232#endif
a2de733c 2233
c9e9f97b
ID		 2234 spin_lock_init(&fs_info->balance_lock);
2235 mutex_init(&fs_info->balance_mutex);
837d5b6e
ID		 2236 atomic_set(&fs_info->balance_running, 0);
2237 atomic_set(&fs_info->balance_pause_req, 0);
a7e99c69 2238 atomic_set(&fs_info->balance_cancel_req, 0);
c9e9f97b 2239 fs_info->balance_ctl = NULL;
837d5b6e 2240 init_waitqueue_head(&fs_info->balance_wait_q);
a2de733c 2241
a061fc8d
CM		 2242 sb->s_blocksize = 4096;
2243 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 2244 sb->s_bdi = &fs_info->bdi;
a061fc8d 2245
76dda93c 2246 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
bfe86848 2247 set_nlink(fs_info->btree_inode, 1);
0afbaf8c
CM		 2248 /*
2249 * we set the i_size on the btree inode to the max possible int.
2250 * the real end of the address space is determined by all of
2251 * the devices in the system
2252 */
2253 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 2254 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 2255 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2256
5d4f98a2 2257 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 2258 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 2259 fs_info->btree_inode->i_mapping);
0b32f4bb 2260 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
a8067e02 2261 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM		 2262
2263 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 2264
76dda93c
YZ		 2265 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2266 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2267 sizeof(struct btrfs_key));
72ac3c0d
JB		 2268 set_bit(BTRFS_INODE_DUMMY,
2269 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
778ba82b 2270 btrfs_insert_inode_hash(fs_info->btree_inode);
76dda93c 2271
0f9dd46c 2272 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2273 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2274 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2275
11833d66 2276 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 2277 fs_info->btree_inode->i_mapping);
11833d66 2278 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 2279 fs_info->btree_inode->i_mapping);
11833d66 2280 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 2281 fs_info->do_barriers = 1;
e18e4809 2282
39279cc3 2283
5a3f23d5 2284 mutex_init(&fs_info->ordered_operations_mutex);
9ffba8cd 2285 mutex_init(&fs_info->ordered_extent_flush_mutex);
e02119d5 2286 mutex_init(&fs_info->tree_log_mutex);
925baedd 2287 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 2288 mutex_init(&fs_info->transaction_kthread_mutex);
2289 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2290 mutex_init(&fs_info->volume_mutex);
276e680d 2291 init_rwsem(&fs_info->extent_commit_sem);
c71bf099 2292 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2293 init_rwsem(&fs_info->subvol_sem);
803b2f54 2294 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
e922e087
SB		 2295 fs_info->dev_replace.lock_owner = 0;
2296 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2297 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2298 mutex_init(&fs_info->dev_replace.lock_management_lock);
2299 mutex_init(&fs_info->dev_replace.lock);
fa9c0d79 2300
416ac51d 2301 spin_lock_init(&fs_info->qgroup_lock);
f2f6ed3d 2302 mutex_init(&fs_info->qgroup_ioctl_lock);
416ac51d
AJ		 2303 fs_info->qgroup_tree = RB_ROOT;
2304 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2305 fs_info->qgroup_seq = 1;
2306 fs_info->quota_enabled = 0;
2307 fs_info->pending_quota_state = 0;
1e8f9158 2308 fs_info->qgroup_ulist = NULL;
2f232036 2309 mutex_init(&fs_info->qgroup_rescan_lock);
416ac51d 2310
fa9c0d79
CM		 2311 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2312 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2313
e6dcd2dc 2314 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2315 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2316 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2317 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2318
53b381b3
DW		 2319 ret = btrfs_alloc_stripe_hash_table(fs_info);
2320 if (ret) {
83c8266a 2321 err = ret;
53b381b3
DW		 2322 goto fail_alloc;
2323 }
2324
0b86a832 2325 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 2326 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2327
3c4bb26b 2328 invalidate_bdev(fs_devices->latest_bdev);
1104a885
DS		 2329
2330 /*
2331 * Read super block and check the signature bytes only
2332 */
a512bbf8 2333 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY		 2334 if (!bh) {
2335 err = -EINVAL;
16cdcec7 2336 goto fail_alloc;
20b45077 2337 }
39279cc3 2338
1104a885
DS		 2339 /*
2340 * We want to check superblock checksum, the type is stored inside.
2341 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2342 */
2343 if (btrfs_check_super_csum(bh->b_data)) {
efe120a0 2344 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
1104a885
DS		 2345 err = -EINVAL;
2346 goto fail_alloc;
2347 }
2348
2349 /*
2350 * super_copy is zeroed at allocation time and we never touch the
2351 * following bytes up to INFO_SIZE, the checksum is calculated from
2352 * the whole block of INFO_SIZE
2353 */
6c41761f
DS		 2354 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2355 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2356 sizeof(*fs_info->super_for_commit));
a061fc8d 2357 brelse(bh);
5f39d397 2358
6c41761f 2359 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2360
1104a885
DS		 2361 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2362 if (ret) {
efe120a0 2363 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
1104a885
DS		 2364 err = -EINVAL;
2365 goto fail_alloc;
2366 }
2367
6c41761f 2368 disk_super = fs_info->super_copy;
0f7d52f4 2369 if (!btrfs_super_root(disk_super))
16cdcec7 2370 goto fail_alloc;
0f7d52f4 2371
acce952b 2372 /* check FS state, whether FS is broken. */
87533c47
MX		 2373 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2374 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2375
af31f5e5
CM		 2376 /*
2377 * run through our array of backup supers and setup
2378 * our ring pointer to the oldest one
2379 */
2380 generation = btrfs_super_generation(disk_super);
2381 find_oldest_super_backup(fs_info, generation);
2382
75e7cb7f
LB		 2383 /*
2384 * In the long term, we'll store the compression type in the super
2385 * block, and it'll be used for per file compression control.
2386 */
2387 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2388
2b82032c
YZ		 2389 ret = btrfs_parse_options(tree_root, options);
2390 if (ret) {
2391 err = ret;
16cdcec7 2392 goto fail_alloc;
2b82032c 2393 }
dfe25020 2394
f2b636e8
JB		 2395 features = btrfs_super_incompat_flags(disk_super) &
2396 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2397 if (features) {
2398 printk(KERN_ERR "BTRFS: couldn't mount because of "
2399 "unsupported optional features (%Lx).\n",
c1c9ff7c 2400 features);
f2b636e8 2401 err = -EINVAL;
16cdcec7 2402 goto fail_alloc;
f2b636e8
JB		 2403 }
2404
727011e0
CM		 2405 if (btrfs_super_leafsize(disk_super) !=
2406 btrfs_super_nodesize(disk_super)) {
2407 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2408 "blocksizes don't match. node %d leaf %d\n",
2409 btrfs_super_nodesize(disk_super),
2410 btrfs_super_leafsize(disk_super));
2411 err = -EINVAL;
2412 goto fail_alloc;
2413 }
2414 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2415 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2416 "blocksize (%d) was too large\n",
2417 btrfs_super_leafsize(disk_super));
2418 err = -EINVAL;
2419 goto fail_alloc;
2420 }
2421
5d4f98a2 2422 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2423 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
69e380d1 2424 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2425 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
727011e0 2426
3173a18f 2427 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
efe120a0 2428 printk(KERN_ERR "BTRFS: has skinny extents\n");
3173a18f 2429
727011e0
CM		 2430 /*
2431 * flag our filesystem as having big metadata blocks if
2432 * they are bigger than the page size
2433 */
2434 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2435 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
efe120a0 2436 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
727011e0
CM		 2437 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2438 }
2439
bc3f116f
CM		 2440 nodesize = btrfs_super_nodesize(disk_super);
2441 leafsize = btrfs_super_leafsize(disk_super);
2442 sectorsize = btrfs_super_sectorsize(disk_super);
2443 stripesize = btrfs_super_stripesize(disk_super);
e2d84521 2444 fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids));
963d678b 2445 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f
CM		 2446
2447 /*
2448 * mixed block groups end up with duplicate but slightly offset
2449 * extent buffers for the same range. It leads to corruptions
2450 */
2451 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2452 (sectorsize != leafsize)) {
efe120a0 2453 printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes "
bc3f116f
CM		 2454 "are not allowed for mixed block groups on %s\n",
2455 sb->s_id);
2456 goto fail_alloc;
2457 }
2458
ceda0864
MX		 2459 /*
2460 * Needn't use the lock because there is no other task which will
2461 * update the flag.
2462 */
a6fa6fae 2463 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2464
f2b636e8
JB		 2465 features = btrfs_super_compat_ro_flags(disk_super) &
2466 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2467 if (!(sb->s_flags & MS_RDONLY) && features) {
2468 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2469 "unsupported option features (%Lx).\n",
c1c9ff7c 2470 features);
f2b636e8 2471 err = -EINVAL;
16cdcec7 2472 goto fail_alloc;
f2b636e8 2473 }
61d92c32
CM		 2474
2475 btrfs_init_workers(&fs_info->generic_worker,
2476 "genwork", 1, NULL);
2477
5443be45 2478 btrfs_init_workers(&fs_info->workers, "worker",
61d92c32
CM		 2479 fs_info->thread_pool_size,
2480 &fs_info->generic_worker);
c8b97818 2481
771ed689 2482 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
45d5fd14 2483 fs_info->thread_pool_size, NULL);
771ed689 2484
8ccf6f19 2485 btrfs_init_workers(&fs_info->flush_workers, "flush_delalloc",
45d5fd14 2486 fs_info->thread_pool_size, NULL);
8ccf6f19 2487
5443be45 2488 btrfs_init_workers(&fs_info->submit_workers, "submit",
b720d209 2489 min_t(u64, fs_devices->num_devices,
45d5fd14 2490 fs_info->thread_pool_size), NULL);
61b49440 2491
bab39bf9 2492 btrfs_init_workers(&fs_info->caching_workers, "cache",
45d5fd14 2493 fs_info->thread_pool_size, NULL);
bab39bf9 2494
61b49440
CM		 2495 /* a higher idle thresh on the submit workers makes it much more
2496 * likely that bios will be send down in a sane order to the
2497 * devices
2498 */
2499 fs_info->submit_workers.idle_thresh = 64;
53863232 2500
771ed689 2501 fs_info->workers.idle_thresh = 16;
4a69a410 2502 fs_info->workers.ordered = 1;
61b49440 2503
771ed689
CM		 2504 fs_info->delalloc_workers.idle_thresh = 2;
2505 fs_info->delalloc_workers.ordered = 1;
2506
61d92c32
CM		 2507 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
2508 &fs_info->generic_worker);
5443be45 2509 btrfs_init_workers(&fs_info->endio_workers, "endio",
61d92c32
CM		 2510 fs_info->thread_pool_size,
2511 &fs_info->generic_worker);
d20f7043 2512 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
61d92c32
CM		 2513 fs_info->thread_pool_size,
2514 &fs_info->generic_worker);
cad321ad 2515 btrfs_init_workers(&fs_info->endio_meta_write_workers,
61d92c32
CM		 2516 "endio-meta-write", fs_info->thread_pool_size,
2517 &fs_info->generic_worker);
53b381b3
DW		 2518 btrfs_init_workers(&fs_info->endio_raid56_workers,
2519 "endio-raid56", fs_info->thread_pool_size,
2520 &fs_info->generic_worker);
2521 btrfs_init_workers(&fs_info->rmw_workers,
2522 "rmw", fs_info->thread_pool_size,
2523 &fs_info->generic_worker);
5443be45 2524 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
61d92c32
CM		 2525 fs_info->thread_pool_size,
2526 &fs_info->generic_worker);
0cb59c99
JB		 2527 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
2528 1, &fs_info->generic_worker);
16cdcec7
MX		 2529 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
2530 fs_info->thread_pool_size,
2531 &fs_info->generic_worker);
90519d66
AJ		 2532 btrfs_init_workers(&fs_info->readahead_workers, "readahead",
2533 fs_info->thread_pool_size,
2534 &fs_info->generic_worker);
2f232036
JS		 2535 btrfs_init_workers(&fs_info->qgroup_rescan_workers, "qgroup-rescan", 1,
2536 &fs_info->generic_worker);
61b49440
CM		 2537
2538 /*
2539 * endios are largely parallel and should have a very
2540 * low idle thresh
2541 */
2542 fs_info->endio_workers.idle_thresh = 4;
b51912c9 2543 fs_info->endio_meta_workers.idle_thresh = 4;
53b381b3
DW		 2544 fs_info->endio_raid56_workers.idle_thresh = 4;
2545 fs_info->rmw_workers.idle_thresh = 2;
b51912c9 2546
9042846b
CM		 2547 fs_info->endio_write_workers.idle_thresh = 2;
2548 fs_info->endio_meta_write_workers.idle_thresh = 2;
90519d66 2549 fs_info->readahead_workers.idle_thresh = 2;
9042846b 2550
0dc3b84a
JB		 2551 /*
2552 * btrfs_start_workers can really only fail because of ENOMEM so just
2553 * return -ENOMEM if any of these fail.
2554 */
2555 ret = btrfs_start_workers(&fs_info->workers);
2556 ret |= btrfs_start_workers(&fs_info->generic_worker);
2557 ret |= btrfs_start_workers(&fs_info->submit_workers);
2558 ret |= btrfs_start_workers(&fs_info->delalloc_workers);
2559 ret |= btrfs_start_workers(&fs_info->fixup_workers);
2560 ret |= btrfs_start_workers(&fs_info->endio_workers);
2561 ret |= btrfs_start_workers(&fs_info->endio_meta_workers);
53b381b3
DW		 2562 ret |= btrfs_start_workers(&fs_info->rmw_workers);
2563 ret |= btrfs_start_workers(&fs_info->endio_raid56_workers);
0dc3b84a
JB		 2564 ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers);
2565 ret |= btrfs_start_workers(&fs_info->endio_write_workers);
2566 ret |= btrfs_start_workers(&fs_info->endio_freespace_worker);
2567 ret |= btrfs_start_workers(&fs_info->delayed_workers);
2568 ret |= btrfs_start_workers(&fs_info->caching_workers);
2569 ret |= btrfs_start_workers(&fs_info->readahead_workers);
8ccf6f19 2570 ret |= btrfs_start_workers(&fs_info->flush_workers);
2f232036 2571 ret |= btrfs_start_workers(&fs_info->qgroup_rescan_workers);
0dc3b84a 2572 if (ret) {
fed425c7 2573 err = -ENOMEM;
0dc3b84a
JB		 2574 goto fail_sb_buffer;
2575 }
4543df7e 2576
4575c9cc 2577 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 2578 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2579 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 2580
db94535d
CM		 2581 tree_root->nodesize = nodesize;
2582 tree_root->leafsize = leafsize;
2583 tree_root->sectorsize = sectorsize;
87ee04eb 2584 tree_root->stripesize = stripesize;
a061fc8d
CM		 2585
2586 sb->s_blocksize = sectorsize;
2587 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 2588
3cae210f 2589 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
efe120a0 2590 printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 2591 goto fail_sb_buffer;
2592 }
19c00ddc 2593
8d082fb7 2594 if (sectorsize != PAGE_SIZE) {
efe120a0 2595 printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) "
8d082fb7 2596 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
941b2ddf
KM		 2597 goto fail_sb_buffer;
2598 }
2599
925baedd 2600 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 2601 ret = btrfs_read_sys_array(tree_root);
925baedd 2602 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2603 if (ret) {
efe120a0 2604 printk(KERN_WARNING "BTRFS: failed to read the system "
d397712b 2605 "array on %s\n", sb->s_id);
5d4f98a2 2606 goto fail_sb_buffer;
84eed90f 2607 }
0b86a832
CM		 2608
2609 blocksize = btrfs_level_size(tree_root,
2610 btrfs_super_chunk_root_level(disk_super));
84234f3a 2611 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 2612
2613 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2614 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2615
2616 chunk_root->node = read_tree_block(chunk_root,
2617 btrfs_super_chunk_root(disk_super),
84234f3a 2618 blocksize, generation);
416bc658
JB		 2619 if (!chunk_root->node ||
2620 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
efe120a0 2621 printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n",
83121942 2622 sb->s_id);
af31f5e5 2623 goto fail_tree_roots;
83121942 2624 }
5d4f98a2
YZ		 2625 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2626 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2627
e17cade2 2628 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
b308bc2f 2629 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
e17cade2 2630
0b86a832 2631 ret = btrfs_read_chunk_tree(chunk_root);
2b82032c 2632 if (ret) {
efe120a0 2633 printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n",
d397712b 2634 sb->s_id);
af31f5e5 2635 goto fail_tree_roots;
2b82032c 2636 }
0b86a832 2637
8dabb742
SB		 2638 /*
2639 * keep the device that is marked to be the target device for the
2640 * dev_replace procedure
2641 */
2642 btrfs_close_extra_devices(fs_info, fs_devices, 0);
dfe25020 2643
a6b0d5c8 2644 if (!fs_devices->latest_bdev) {
efe120a0 2645 printk(KERN_CRIT "BTRFS: failed to read devices on %s\n",
a6b0d5c8
CM		 2646 sb->s_id);
2647 goto fail_tree_roots;
2648 }
2649
af31f5e5 2650retry_root_backup:
db94535d
CM		 2651 blocksize = btrfs_level_size(tree_root,
2652 btrfs_super_root_level(disk_super));
84234f3a 2653 generation = btrfs_super_generation(disk_super);
0b86a832 2654
e20d96d6 2655 tree_root->node = read_tree_block(tree_root,
db94535d 2656 btrfs_super_root(disk_super),
84234f3a 2657 blocksize, generation);
af31f5e5
CM		 2658 if (!tree_root->node ||
2659 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
efe120a0 2660 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
83121942 2661 sb->s_id);
af31f5e5
CM		 2662
2663 goto recovery_tree_root;
83121942 2664 }
af31f5e5 2665
5d4f98a2
YZ		 2666 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2667 tree_root->commit_root = btrfs_root_node(tree_root);
69e9c6c6 2668 btrfs_set_root_refs(&tree_root->root_item, 1);
db94535d 2669
cb517eab
MX		 2670 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2671 location.type = BTRFS_ROOT_ITEM_KEY;
2672 location.offset = 0;
2673
2674 extent_root = btrfs_read_tree_root(tree_root, &location);
2675 if (IS_ERR(extent_root)) {
2676 ret = PTR_ERR(extent_root);
af31f5e5 2677 goto recovery_tree_root;
cb517eab 2678 }
0b86a832 2679 extent_root->track_dirty = 1;
cb517eab 2680 fs_info->extent_root = extent_root;
0b86a832 2681
cb517eab
MX		 2682 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2683 dev_root = btrfs_read_tree_root(tree_root, &location);
2684 if (IS_ERR(dev_root)) {
2685 ret = PTR_ERR(dev_root);
af31f5e5 2686 goto recovery_tree_root;
cb517eab 2687 }
5d4f98a2 2688 dev_root->track_dirty = 1;
cb517eab
MX		 2689 fs_info->dev_root = dev_root;
2690 btrfs_init_devices_late(fs_info);
3768f368 2691
cb517eab
MX		 2692 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2693 csum_root = btrfs_read_tree_root(tree_root, &location);
2694 if (IS_ERR(csum_root)) {
2695 ret = PTR_ERR(csum_root);
af31f5e5 2696 goto recovery_tree_root;
cb517eab 2697 }
d20f7043 2698 csum_root->track_dirty = 1;
cb517eab 2699 fs_info->csum_root = csum_root;
d20f7043 2700
cb517eab
MX		 2701 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2702 quota_root = btrfs_read_tree_root(tree_root, &location);
2703 if (!IS_ERR(quota_root)) {
bcef60f2
AJ		 2704 quota_root->track_dirty = 1;
2705 fs_info->quota_enabled = 1;
2706 fs_info->pending_quota_state = 1;
cb517eab 2707 fs_info->quota_root = quota_root;
bcef60f2
AJ		 2708 }
2709
f7a81ea4
SB		 2710 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2711 uuid_root = btrfs_read_tree_root(tree_root, &location);
2712 if (IS_ERR(uuid_root)) {
2713 ret = PTR_ERR(uuid_root);
2714 if (ret != -ENOENT)
2715 goto recovery_tree_root;
2716 create_uuid_tree = true;
70f80175 2717 check_uuid_tree = false;
f7a81ea4
SB		 2718 } else {
2719 uuid_root->track_dirty = 1;
2720 fs_info->uuid_root = uuid_root;
70f80175
SB		 2721 create_uuid_tree = false;
2722 check_uuid_tree =
2723 generation != btrfs_super_uuid_tree_generation(disk_super);
f7a81ea4
SB		 2724 }
2725
8929ecfa
YZ		 2726 fs_info->generation = generation;
2727 fs_info->last_trans_committed = generation;
8929ecfa 2728
68310a5e
ID		 2729 ret = btrfs_recover_balance(fs_info);
2730 if (ret) {
efe120a0 2731 printk(KERN_WARNING "BTRFS: failed to recover balance\n");
68310a5e
ID		 2732 goto fail_block_groups;
2733 }
2734
733f4fbb
SB		 2735 ret = btrfs_init_dev_stats(fs_info);
2736 if (ret) {
efe120a0 2737 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
733f4fbb
SB		 2738 ret);
2739 goto fail_block_groups;
2740 }
2741
8dabb742
SB		 2742 ret = btrfs_init_dev_replace(fs_info);
2743 if (ret) {
efe120a0 2744 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
8dabb742
SB		 2745 goto fail_block_groups;
2746 }
2747
2748 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2749
5ac1d209
JM		 2750 ret = btrfs_sysfs_add_one(fs_info);
2751 if (ret) {
efe120a0 2752 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
5ac1d209
JM		 2753 goto fail_block_groups;
2754 }
2755
c59021f8 2756 ret = btrfs_init_space_info(fs_info);
2757 if (ret) {
efe120a0 2758 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
2365dd3c 2759 goto fail_sysfs;
c59021f8 2760 }
2761
1b1d1f66
JB		 2762 ret = btrfs_read_block_groups(extent_root);
2763 if (ret) {
efe120a0 2764 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
2365dd3c 2765 goto fail_sysfs;
1b1d1f66 2766 }
5af3e8cc
SB		 2767 fs_info->num_tolerated_disk_barrier_failures =
2768 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
292fd7fc
SB		 2769 if (fs_info->fs_devices->missing_devices >
2770 fs_info->num_tolerated_disk_barrier_failures &&
2771 !(sb->s_flags & MS_RDONLY)) {
efe120a0
FH		 2772 printk(KERN_WARNING "BTRFS: "
2773 "too many missing devices, writeable mount is not allowed\n");
2365dd3c 2774 goto fail_sysfs;
292fd7fc 2775 }
9078a3e1 2776
a74a4b97
CM		 2777 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2778 "btrfs-cleaner");
57506d50 2779 if (IS_ERR(fs_info->cleaner_kthread))
2365dd3c 2780 goto fail_sysfs;
a74a4b97
CM		 2781
2782 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2783 tree_root,
2784 "btrfs-transaction");
57506d50 2785 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2786 goto fail_cleaner;
a74a4b97 2787
c289811c
CM		 2788 if (!btrfs_test_opt(tree_root, SSD) &&
2789 !btrfs_test_opt(tree_root, NOSSD) &&
2790 !fs_info->fs_devices->rotating) {
efe120a0 2791 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
c289811c
CM		 2792 "mode\n");
2793 btrfs_set_opt(fs_info->mount_opt, SSD);
2794 }
2795
3818aea2
QW		 2796 /* Set the real inode map cache flag */
2797 if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE))
2798 btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE);
2799
21adbd5c
SB		 2800#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2801 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2802 ret = btrfsic_mount(tree_root, fs_devices,
2803 btrfs_test_opt(tree_root,
2804 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2805 1 : 0,
2806 fs_info->check_integrity_print_mask);
2807 if (ret)
efe120a0 2808 printk(KERN_WARNING "BTRFS: failed to initialize"
21adbd5c
SB		 2809 " integrity check module %s\n", sb->s_id);
2810 }
2811#endif
bcef60f2
AJ		 2812 ret = btrfs_read_qgroup_config(fs_info);
2813 if (ret)
2814 goto fail_trans_kthread;
21adbd5c 2815
acce952b 2816 /* do not make disk changes in broken FS */
68ce9682 2817 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM		 2818 u64 bytenr = btrfs_super_log_root(disk_super);
2819
7c2ca468 2820 if (fs_devices->rw_devices == 0) {
efe120a0 2821 printk(KERN_WARNING "BTRFS: log replay required "
d397712b 2822 "on RO media\n");
7c2ca468 2823 err = -EIO;
bcef60f2 2824 goto fail_qgroup;
7c2ca468 2825 }
e02119d5
CM		 2826 blocksize =
2827 btrfs_level_size(tree_root,
2828 btrfs_super_log_root_level(disk_super));
d18a2c44 2829
6f07e42e 2830 log_tree_root = btrfs_alloc_root(fs_info);
676e4c86
DC		 2831 if (!log_tree_root) {
2832 err = -ENOMEM;
bcef60f2 2833 goto fail_qgroup;
676e4c86 2834 }
e02119d5
CM		 2835
2836 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2837 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2838
2839 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 2840 blocksize,
2841 generation + 1);
416bc658
JB		 2842 if (!log_tree_root->node ||
2843 !extent_buffer_uptodate(log_tree_root->node)) {
efe120a0 2844 printk(KERN_ERR "BTRFS: failed to read log tree\n");
416bc658
JB		 2845 free_extent_buffer(log_tree_root->node);
2846 kfree(log_tree_root);
2847 goto fail_trans_kthread;
2848 }
79787eaa 2849 /* returns with log_tree_root freed on success */
e02119d5 2850 ret = btrfs_recover_log_trees(log_tree_root);
79787eaa
JM		 2851 if (ret) {
2852 btrfs_error(tree_root->fs_info, ret,
2853 "Failed to recover log tree");
2854 free_extent_buffer(log_tree_root->node);
2855 kfree(log_tree_root);
2856 goto fail_trans_kthread;
2857 }
e556ce2c
YZ		 2858
2859 if (sb->s_flags & MS_RDONLY) {
79787eaa
JM		 2860 ret = btrfs_commit_super(tree_root);
2861 if (ret)
2862 goto fail_trans_kthread;
e556ce2c 2863 }
e02119d5 2864 }
1a40e23b 2865
76dda93c 2866 ret = btrfs_find_orphan_roots(tree_root);
79787eaa
JM		 2867 if (ret)
2868 goto fail_trans_kthread;
76dda93c 2869
7c2ca468 2870 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b 2871 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2
ID		 2872 if (ret)
2873 goto fail_trans_kthread;
d68fc57b 2874
5d4f98a2 2875 ret = btrfs_recover_relocation(tree_root);
d7ce5843
MX		 2876 if (ret < 0) {
2877 printk(KERN_WARNING
efe120a0 2878 "BTRFS: failed to recover relocation\n");
d7ce5843 2879 err = -EINVAL;
bcef60f2 2880 goto fail_qgroup;
d7ce5843 2881 }
7c2ca468 2882 }
1a40e23b 2883
3de4586c
CM		 2884 location.objectid = BTRFS_FS_TREE_OBJECTID;
2885 location.type = BTRFS_ROOT_ITEM_KEY;
cb517eab 2886 location.offset = 0;
3de4586c 2887
3de4586c 2888 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
3140c9a3
DC		 2889 if (IS_ERR(fs_info->fs_root)) {
2890 err = PTR_ERR(fs_info->fs_root);
bcef60f2 2891 goto fail_qgroup;
3140c9a3 2892 }
c289811c 2893
2b6ba629
ID		 2894 if (sb->s_flags & MS_RDONLY)
2895 return 0;
59641015 2896
2b6ba629
ID		 2897 down_read(&fs_info->cleanup_work_sem);
2898 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2899 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 2900 up_read(&fs_info->cleanup_work_sem);
2b6ba629
ID		 2901 close_ctree(tree_root);
2902 return ret;
2903 }
2904 up_read(&fs_info->cleanup_work_sem);
59641015 2905
2b6ba629
ID		 2906 ret = btrfs_resume_balance_async(fs_info);
2907 if (ret) {
efe120a0 2908 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
2b6ba629
ID		 2909 close_ctree(tree_root);
2910 return ret;
e3acc2a6
JB		 2911 }
2912
8dabb742
SB		 2913 ret = btrfs_resume_dev_replace_async(fs_info);
2914 if (ret) {
efe120a0 2915 pr_warn("BTRFS: failed to resume dev_replace\n");
8dabb742
SB		 2916 close_ctree(tree_root);
2917 return ret;
2918 }
2919
b382a324
JS		 2920 btrfs_qgroup_rescan_resume(fs_info);
2921
f7a81ea4 2922 if (create_uuid_tree) {
efe120a0 2923 pr_info("BTRFS: creating UUID tree\n");
f7a81ea4
SB		 2924 ret = btrfs_create_uuid_tree(fs_info);
2925 if (ret) {
efe120a0 2926 pr_warn("BTRFS: failed to create the UUID tree %d\n",
f7a81ea4
SB		 2927 ret);
2928 close_ctree(tree_root);
2929 return ret;
2930 }
f420ee1e
SB		 2931 } else if (check_uuid_tree ||
2932 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
efe120a0 2933 pr_info("BTRFS: checking UUID tree\n");
70f80175
SB		 2934 ret = btrfs_check_uuid_tree(fs_info);
2935 if (ret) {
efe120a0 2936 pr_warn("BTRFS: failed to check the UUID tree %d\n",
70f80175
SB		 2937 ret);
2938 close_ctree(tree_root);
2939 return ret;
2940 }
2941 } else {
2942 fs_info->update_uuid_tree_gen = 1;
f7a81ea4
SB		 2943 }
2944
ad2b2c80 2945 return 0;
39279cc3 2946
bcef60f2
AJ		 2947fail_qgroup:
2948 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM		 2949fail_trans_kthread:
2950 kthread_stop(fs_info->transaction_kthread);
54067ae9 2951 btrfs_cleanup_transaction(fs_info->tree_root);
7b5ff90e 2952 del_fs_roots(fs_info);
3f157a2f 2953fail_cleaner:
a74a4b97 2954 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2955
2956 /*
2957 * make sure we're done with the btree inode before we stop our
2958 * kthreads
2959 */
2960 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 2961
2365dd3c
AJ		 2962fail_sysfs:
2963 btrfs_sysfs_remove_one(fs_info);
2964
1b1d1f66 2965fail_block_groups:
54067ae9 2966 btrfs_put_block_group_cache(fs_info);
1b1d1f66 2967 btrfs_free_block_groups(fs_info);
af31f5e5
CM		 2968
2969fail_tree_roots:
2970 free_root_pointers(fs_info, 1);
2b8195bb 2971 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 2972
39279cc3 2973fail_sb_buffer:
7abadb64 2974 btrfs_stop_all_workers(fs_info);
16cdcec7 2975fail_alloc:
4543df7e 2976fail_iput:
586e46e2
ID		 2977 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2978
4543df7e 2979 iput(fs_info->btree_inode);
c404e0dc
MX		 2980fail_bio_counter:
2981 percpu_counter_destroy(&fs_info->bio_counter);
963d678b
MX		 2982fail_delalloc_bytes:
2983 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX		 2984fail_dirty_metadata_bytes:
2985 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
ad081f14 2986fail_bdi:
7e662854 2987 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 2988fail_srcu:
2989 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 2990fail:
53b381b3 2991 btrfs_free_stripe_hash_table(fs_info);
586e46e2 2992 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 2993 return err;
af31f5e5
CM		 2994
2995recovery_tree_root:
af31f5e5
CM		 2996 if (!btrfs_test_opt(tree_root, RECOVERY))
2997 goto fail_tree_roots;
2998
2999 free_root_pointers(fs_info, 0);
3000
3001 /* don't use the log in recovery mode, it won't be valid */
3002 btrfs_set_super_log_root(disk_super, 0);
3003
3004 /* we can't trust the free space cache either */
3005 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3006
3007 ret = next_root_backup(fs_info, fs_info->super_copy,
3008 &num_backups_tried, &backup_index);
3009 if (ret == -1)
3010 goto fail_block_groups;
3011 goto retry_root_backup;
eb60ceac
CM		 3012}
3013
f2984462
CM		 3014static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3015{
f2984462
CM		 3016 if (uptodate) {
3017 set_buffer_uptodate(bh);
3018 } else {
442a4f63
SB		 3019 struct btrfs_device *device = (struct btrfs_device *)
3020 bh->b_private;
3021
efe120a0 3022 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
606686ee
JB		 3023 "I/O error on %s\n",
3024 rcu_str_deref(device->name));
1259ab75
CM		 3025 /* note, we dont' set_buffer_write_io_error because we have
3026 * our own ways of dealing with the IO errors
3027 */
f2984462 3028 clear_buffer_uptodate(bh);
442a4f63 3029 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM		 3030 }
3031 unlock_buffer(bh);
3032 put_bh(bh);
3033}
3034
a512bbf8
YZ		 3035struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3036{
3037 struct buffer_head *bh;
3038 struct buffer_head *latest = NULL;
3039 struct btrfs_super_block *super;
3040 int i;
3041 u64 transid = 0;
3042 u64 bytenr;
3043
3044 /* we would like to check all the supers, but that would make
3045 * a btrfs mount succeed after a mkfs from a different FS.
3046 * So, we need to add a special mount option to scan for
3047 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3048 */
3049 for (i = 0; i < 1; i++) {
3050 bytenr = btrfs_sb_offset(i);
8068a47e
AJ		 3051 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3052 i_size_read(bdev->bd_inode))
a512bbf8 3053 break;
8068a47e
AJ		 3054 bh = __bread(bdev, bytenr / 4096,
3055 BTRFS_SUPER_INFO_SIZE);
a512bbf8
YZ		 3056 if (!bh)
3057 continue;
3058
3059 super = (struct btrfs_super_block *)bh->b_data;
3060 if (btrfs_super_bytenr(super) != bytenr ||
3cae210f 3061 btrfs_super_magic(super) != BTRFS_MAGIC) {
a512bbf8
YZ		 3062 brelse(bh);
3063 continue;
3064 }
3065
3066 if (!latest || btrfs_super_generation(super) > transid) {
3067 brelse(latest);
3068 latest = bh;
3069 transid = btrfs_super_generation(super);
3070 } else {
3071 brelse(bh);
3072 }
3073 }
3074 return latest;
3075}
3076
4eedeb75
HH		 3077/*
3078 * this should be called twice, once with wait == 0 and
3079 * once with wait == 1. When wait == 0 is done, all the buffer heads
3080 * we write are pinned.
3081 *
3082 * They are released when wait == 1 is done.
3083 * max_mirrors must be the same for both runs, and it indicates how
3084 * many supers on this one device should be written.
3085 *
3086 * max_mirrors == 0 means to write them all.
3087 */
a512bbf8
YZ		 3088static int write_dev_supers(struct btrfs_device *device,
3089 struct btrfs_super_block *sb,
3090 int do_barriers, int wait, int max_mirrors)
3091{
3092 struct buffer_head *bh;
3093 int i;
3094 int ret;
3095 int errors = 0;
3096 u32 crc;
3097 u64 bytenr;
a512bbf8
YZ		 3098
3099 if (max_mirrors == 0)
3100 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3101
a512bbf8
YZ		 3102 for (i = 0; i < max_mirrors; i++) {
3103 bytenr = btrfs_sb_offset(i);
3104 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
3105 break;
3106
3107 if (wait) {
3108 bh = __find_get_block(device->bdev, bytenr / 4096,
3109 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3110 if (!bh) {
3111 errors++;
3112 continue;
3113 }
a512bbf8 3114 wait_on_buffer(bh);
4eedeb75
HH		 3115 if (!buffer_uptodate(bh))
3116 errors++;
3117
3118 /* drop our reference */
3119 brelse(bh);
3120
3121 /* drop the reference from the wait == 0 run */
3122 brelse(bh);
3123 continue;
a512bbf8
YZ		 3124 } else {
3125 btrfs_set_super_bytenr(sb, bytenr);
3126
3127 crc = ~(u32)0;
b0496686 3128 crc = btrfs_csum_data((char *)sb +
a512bbf8
YZ		 3129 BTRFS_CSUM_SIZE, crc,
3130 BTRFS_SUPER_INFO_SIZE -
3131 BTRFS_CSUM_SIZE);
3132 btrfs_csum_final(crc, sb->csum);
3133
4eedeb75
HH		 3134 /*
3135 * one reference for us, and we leave it for the
3136 * caller
3137 */
a512bbf8
YZ		 3138 bh = __getblk(device->bdev, bytenr / 4096,
3139 BTRFS_SUPER_INFO_SIZE);
634554dc 3140 if (!bh) {
efe120a0 3141 printk(KERN_ERR "BTRFS: couldn't get super "
634554dc
JB		 3142 "buffer head for bytenr %Lu\n", bytenr);
3143 errors++;
3144 continue;
3145 }
3146
a512bbf8
YZ		 3147 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3148
4eedeb75 3149 /* one reference for submit_bh */
a512bbf8 3150 get_bh(bh);
4eedeb75
HH		 3151
3152 set_buffer_uptodate(bh);
a512bbf8
YZ		 3153 lock_buffer(bh);
3154 bh->b_end_io = btrfs_end_buffer_write_sync;
442a4f63 3155 bh->b_private = device;
a512bbf8
YZ		 3156 }
3157
387125fc
CM		 3158 /*
3159 * we fua the first super. The others we allow
3160 * to go down lazy.
3161 */
e8117c26
WS		 3162 if (i == 0)
3163 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3164 else
3165 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
4eedeb75 3166 if (ret)
a512bbf8 3167 errors++;
a512bbf8
YZ		 3168 }
3169 return errors < i ? 0 : -1;
3170}
3171
387125fc
CM		 3172/*
3173 * endio for the write_dev_flush, this will wake anyone waiting
3174 * for the barrier when it is done
3175 */
3176static void btrfs_end_empty_barrier(struct bio *bio, int err)
3177{
3178 if (err) {
3179 if (err == -EOPNOTSUPP)
3180 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3181 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3182 }
3183 if (bio->bi_private)
3184 complete(bio->bi_private);
3185 bio_put(bio);
3186}
3187
3188/*
3189 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3190 * sent down. With wait == 1, it waits for the previous flush.
3191 *
3192 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3193 * capable
3194 */
3195static int write_dev_flush(struct btrfs_device *device, int wait)
3196{
3197 struct bio *bio;
3198 int ret = 0;
3199
3200 if (device->nobarriers)
3201 return 0;
3202
3203 if (wait) {
3204 bio = device->flush_bio;
3205 if (!bio)
3206 return 0;
3207
3208 wait_for_completion(&device->flush_wait);
3209
3210 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
efe120a0 3211 printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
606686ee 3212 rcu_str_deref(device->name));
387125fc 3213 device->nobarriers = 1;
5af3e8cc 3214 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
387125fc 3215 ret = -EIO;
5af3e8cc
SB		 3216 btrfs_dev_stat_inc_and_print(device,
3217 BTRFS_DEV_STAT_FLUSH_ERRS);
387125fc
CM		 3218 }
3219
3220 /* drop the reference from the wait == 0 run */
3221 bio_put(bio);
3222 device->flush_bio = NULL;
3223
3224 return ret;
3225 }
3226
3227 /*
3228 * one reference for us, and we leave it for the
3229 * caller
3230 */
9c017abc 3231 device->flush_bio = NULL;
9be3395b 3232 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
387125fc
CM		 3233 if (!bio)
3234 return -ENOMEM;
3235
3236 bio->bi_end_io = btrfs_end_empty_barrier;
3237 bio->bi_bdev = device->bdev;
3238 init_completion(&device->flush_wait);
3239 bio->bi_private = &device->flush_wait;
3240 device->flush_bio = bio;
3241
3242 bio_get(bio);
21adbd5c 3243 btrfsic_submit_bio(WRITE_FLUSH, bio);
387125fc
CM		 3244
3245 return 0;
3246}
3247
3248/*
3249 * send an empty flush down to each device in parallel,
3250 * then wait for them
3251 */
3252static int barrier_all_devices(struct btrfs_fs_info *info)
3253{
3254 struct list_head *head;
3255 struct btrfs_device *dev;
5af3e8cc
SB		 3256 int errors_send = 0;
3257 int errors_wait = 0;
387125fc
CM		 3258 int ret;
3259
3260 /* send down all the barriers */
3261 head = &info->fs_devices->devices;
3262 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3263 if (dev->missing)
3264 continue;
387125fc 3265 if (!dev->bdev) {
5af3e8cc 3266 errors_send++;
387125fc
CM		 3267 continue;
3268 }
3269 if (!dev->in_fs_metadata || !dev->writeable)
3270 continue;
3271
3272 ret = write_dev_flush(dev, 0);
3273 if (ret)
5af3e8cc 3274 errors_send++;
387125fc
CM		 3275 }
3276
3277 /* wait for all the barriers */
3278 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3279 if (dev->missing)
3280 continue;
387125fc 3281 if (!dev->bdev) {
5af3e8cc 3282 errors_wait++;
387125fc
CM		 3283 continue;
3284 }
3285 if (!dev->in_fs_metadata || !dev->writeable)
3286 continue;
3287
3288 ret = write_dev_flush(dev, 1);
3289 if (ret)
5af3e8cc 3290 errors_wait++;
387125fc 3291 }
5af3e8cc
SB		 3292 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3293 errors_wait > info->num_tolerated_disk_barrier_failures)
387125fc
CM		 3294 return -EIO;
3295 return 0;
3296}
3297
5af3e8cc
SB		 3298int btrfs_calc_num_tolerated_disk_barrier_failures(
3299 struct btrfs_fs_info *fs_info)
3300{
3301 struct btrfs_ioctl_space_info space;
3302 struct btrfs_space_info *sinfo;
3303 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3304 BTRFS_BLOCK_GROUP_SYSTEM,
3305 BTRFS_BLOCK_GROUP_METADATA,
3306 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3307 int num_types = 4;
3308 int i;
3309 int c;
3310 int num_tolerated_disk_barrier_failures =
3311 (int)fs_info->fs_devices->num_devices;
3312
3313 for (i = 0; i < num_types; i++) {
3314 struct btrfs_space_info *tmp;
3315
3316 sinfo = NULL;
3317 rcu_read_lock();
3318 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3319 if (tmp->flags == types[i]) {
3320 sinfo = tmp;
3321 break;
3322 }
3323 }
3324 rcu_read_unlock();
3325
3326 if (!sinfo)
3327 continue;
3328
3329 down_read(&sinfo->groups_sem);
3330 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3331 if (!list_empty(&sinfo->block_groups[c])) {
3332 u64 flags;
3333
3334 btrfs_get_block_group_info(
3335 &sinfo->block_groups[c], &space);
3336 if (space.total_bytes == 0 ||
3337 space.used_bytes == 0)
3338 continue;
3339 flags = space.flags;
3340 /*
3341 * return
3342 * 0: if dup, single or RAID0 is configured for
3343 * any of metadata, system or data, else
3344 * 1: if RAID5 is configured, or if RAID1 or
3345 * RAID10 is configured and only two mirrors
3346 * are used, else
3347 * 2: if RAID6 is configured, else
3348 * num_mirrors - 1: if RAID1 or RAID10 is
3349 * configured and more than
3350 * 2 mirrors are used.
3351 */
3352 if (num_tolerated_disk_barrier_failures > 0 &&
3353 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3354 BTRFS_BLOCK_GROUP_RAID0)) ||
3355 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3356 == 0)))
3357 num_tolerated_disk_barrier_failures = 0;
53b381b3
DW		 3358 else if (num_tolerated_disk_barrier_failures > 1) {
3359 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3360 BTRFS_BLOCK_GROUP_RAID5 |
3361 BTRFS_BLOCK_GROUP_RAID10)) {
3362 num_tolerated_disk_barrier_failures = 1;
3363 } else if (flags &
15b0a89d 3364 BTRFS_BLOCK_GROUP_RAID6) {
53b381b3
DW		 3365 num_tolerated_disk_barrier_failures = 2;
3366 }
3367 }
5af3e8cc
SB		 3368 }
3369 }
3370 up_read(&sinfo->groups_sem);
3371 }
3372
3373 return num_tolerated_disk_barrier_failures;
3374}
3375
48a3b636 3376static int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 3377{
e5e9a520 3378 struct list_head *head;
f2984462 3379 struct btrfs_device *dev;
a061fc8d 3380 struct btrfs_super_block *sb;
f2984462 3381 struct btrfs_dev_item *dev_item;
f2984462
CM		 3382 int ret;
3383 int do_barriers;
a236aed1
CM		 3384 int max_errors;
3385 int total_errors = 0;
a061fc8d 3386 u64 flags;
f2984462
CM		 3387
3388 do_barriers = !btrfs_test_opt(root, NOBARRIER);
af31f5e5 3389 backup_super_roots(root->fs_info);
f2984462 3390
6c41761f 3391 sb = root->fs_info->super_for_commit;
a061fc8d 3392 dev_item = &sb->dev_item;
e5e9a520 3393
174ba509 3394 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 3395 head = &root->fs_info->fs_devices->devices;
d7306801 3396 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
387125fc 3397
5af3e8cc
SB		 3398 if (do_barriers) {
3399 ret = barrier_all_devices(root->fs_info);
3400 if (ret) {
3401 mutex_unlock(
3402 &root->fs_info->fs_devices->device_list_mutex);
3403 btrfs_error(root->fs_info, ret,
3404 "errors while submitting device barriers.");
3405 return ret;
3406 }
3407 }
387125fc 3408
1f78160c 3409 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3410 if (!dev->bdev) {
3411 total_errors++;
3412 continue;
3413 }
2b82032c 3414 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3415 continue;
3416
2b82032c 3417 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 3418 btrfs_set_stack_device_type(dev_item, dev->type);
3419 btrfs_set_stack_device_id(dev_item, dev->devid);
3420 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
3421 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
3422 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3423 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3424 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3425 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 3426 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 3427
a061fc8d
CM		 3428 flags = btrfs_super_flags(sb);
3429 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3430
a512bbf8 3431 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 3432 if (ret)
3433 total_errors++;
f2984462 3434 }
a236aed1 3435 if (total_errors > max_errors) {
efe120a0 3436 btrfs_err(root->fs_info, "%d errors while writing supers",
d397712b 3437 total_errors);
a724b436 3438 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
79787eaa 3439
9d565ba4
SB		 3440 /* FUA is masked off if unsupported and can't be the reason */
3441 btrfs_error(root->fs_info, -EIO,
3442 "%d errors while writing supers", total_errors);
3443 return -EIO;
a236aed1 3444 }
f2984462 3445
a512bbf8 3446 total_errors = 0;
1f78160c 3447 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3448 if (!dev->bdev)
3449 continue;
2b82032c 3450 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3451 continue;
3452
a512bbf8
YZ		 3453 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3454 if (ret)
3455 total_errors++;
f2984462 3456 }
174ba509 3457 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 3458 if (total_errors > max_errors) {
79787eaa
JM		 3459 btrfs_error(root->fs_info, -EIO,
3460 "%d errors while writing supers", total_errors);
3461 return -EIO;
a236aed1 3462 }
f2984462
CM		 3463 return 0;
3464}
3465
a512bbf8
YZ		 3466int write_ctree_super(struct btrfs_trans_handle *trans,
3467 struct btrfs_root *root, int max_mirrors)
eb60ceac 3468{
f570e757 3469 return write_all_supers(root, max_mirrors);
cfaa7295
CM		 3470}
3471
cb517eab
MX		 3472/* Drop a fs root from the radix tree and free it. */
3473void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3474 struct btrfs_root *root)
2619ba1f 3475{
4df27c4d 3476 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 3477 radix_tree_delete(&fs_info->fs_roots_radix,
3478 (unsigned long)root->root_key.objectid);
4df27c4d 3479 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 3480
3481 if (btrfs_root_refs(&root->root_item) == 0)
3482 synchronize_srcu(&fs_info->subvol_srcu);
3483
1a4319cc 3484 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
3321719e 3485 btrfs_free_log(NULL, root);
3321719e 3486
581bb050
LZ		 3487 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3488 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3489 free_fs_root(root);
4df27c4d
YZ		 3490}
3491
3492static void free_fs_root(struct btrfs_root *root)
3493{
82d5902d 3494 iput(root->cache_inode);
4df27c4d 3495 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
1cb048f5
FDBM		 3496 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3497 root->orphan_block_rsv = NULL;
0ee5dc67
AV		 3498 if (root->anon_dev)
3499 free_anon_bdev(root->anon_dev);
4df27c4d
YZ		 3500 free_extent_buffer(root->node);
3501 free_extent_buffer(root->commit_root);
581bb050
LZ		 3502 kfree(root->free_ino_ctl);
3503 kfree(root->free_ino_pinned);
d397712b 3504 kfree(root->name);
b0feb9d9 3505 btrfs_put_fs_root(root);
2619ba1f
CM		 3506}
3507
cb517eab
MX		 3508void btrfs_free_fs_root(struct btrfs_root *root)
3509{
3510 free_fs_root(root);
2619ba1f
CM		 3511}
3512
c146afad 3513int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3514{
c146afad
YZ		 3515 u64 root_objectid = 0;
3516 struct btrfs_root *gang[8];
3517 int i;
3768f368 3518 int ret;
e089f05c 3519
c146afad
YZ		 3520 while (1) {
3521 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3522 (void **)gang, root_objectid,
3523 ARRAY_SIZE(gang));
3524 if (!ret)
3525 break;
5d4f98a2
YZ		 3526
3527 root_objectid = gang[ret - 1]->root_key.objectid + 1;
c146afad 3528 for (i = 0; i < ret; i++) {
66b4ffd1
JB		 3529 int err;
3530
c146afad 3531 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 3532 err = btrfs_orphan_cleanup(gang[i]);
3533 if (err)
3534 return err;
c146afad
YZ		 3535 }
3536 root_objectid++;
3537 }
3538 return 0;
3539}
a2135011 3540
c146afad
YZ		 3541int btrfs_commit_super(struct btrfs_root *root)
3542{
3543 struct btrfs_trans_handle *trans;
a74a4b97 3544
c146afad 3545 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 3546 btrfs_run_delayed_iputs(root);
c146afad 3547 mutex_unlock(&root->fs_info->cleaner_mutex);
9d1a2a3a 3548 wake_up_process(root->fs_info->cleaner_kthread);
c71bf099
YZ		 3549
3550 /* wait until ongoing cleanup work done */
3551 down_write(&root->fs_info->cleanup_work_sem);
3552 up_write(&root->fs_info->cleanup_work_sem);
3553
7a7eaa40 3554 trans = btrfs_join_transaction(root);
3612b495
TI		 3555 if (IS_ERR(trans))
3556 return PTR_ERR(trans);
d52c1bcc 3557 return btrfs_commit_transaction(trans, root);
c146afad
YZ		 3558}
3559
3560int close_ctree(struct btrfs_root *root)
3561{
3562 struct btrfs_fs_info *fs_info = root->fs_info;
3563 int ret;
3564
3565 fs_info->closing = 1;
3566 smp_mb();
3567
803b2f54
SB		 3568 /* wait for the uuid_scan task to finish */
3569 down(&fs_info->uuid_tree_rescan_sem);
3570 /* avoid complains from lockdep et al., set sem back to initial state */
3571 up(&fs_info->uuid_tree_rescan_sem);
3572
837d5b6e 3573 /* pause restriper - we want to resume on mount */
aa1b8cd4 3574 btrfs_pause_balance(fs_info);
837d5b6e 3575
8dabb742
SB		 3576 btrfs_dev_replace_suspend_for_unmount(fs_info);
3577
aa1b8cd4 3578 btrfs_scrub_cancel(fs_info);
4cb5300b
CM		 3579
3580 /* wait for any defraggers to finish */
3581 wait_event(fs_info->transaction_wait,
3582 (atomic_read(&fs_info->defrag_running) == 0));
3583
3584 /* clear out the rbtree of defraggable inodes */
26176e7c 3585 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3586
c146afad 3587 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 3588 ret = btrfs_commit_super(root);
3589 if (ret)
efe120a0 3590 btrfs_err(root->fs_info, "commit super ret %d", ret);
acce952b 3591 }
3592
87533c47 3593 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
68ce9682 3594 btrfs_error_commit_super(root);
0f7d52f4 3595
e3029d9f
AV		 3596 kthread_stop(fs_info->transaction_kthread);
3597 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3598
f25784b3
YZ		 3599 fs_info->closing = 2;
3600 smp_mb();
3601
bcef60f2
AJ		 3602 btrfs_free_qgroup_config(root->fs_info);
3603
963d678b 3604 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
efe120a0 3605 btrfs_info(root->fs_info, "at unmount delalloc count %lld",
963d678b 3606 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3607 }
bcc63abb 3608
5ac1d209
JM		 3609 btrfs_sysfs_remove_one(fs_info);
3610
c146afad 3611 del_fs_roots(fs_info);
d10c5f31 3612
1a4319cc
LB		 3613 btrfs_put_block_group_cache(fs_info);
3614
2b1360da
JB		 3615 btrfs_free_block_groups(fs_info);
3616
96192499
JB		 3617 btrfs_stop_all_workers(fs_info);
3618
13e6c37b 3619 free_root_pointers(fs_info, 1);
9ad6b7bc 3620
13e6c37b 3621 iput(fs_info->btree_inode);
d6bfde87 3622
21adbd5c
SB		 3623#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3624 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3625 btrfsic_unmount(root, fs_info->fs_devices);
3626#endif
3627
dfe25020 3628 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3629 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3630
e2d84521 3631 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 3632 percpu_counter_destroy(&fs_info->delalloc_bytes);
c404e0dc 3633 percpu_counter_destroy(&fs_info->bio_counter);
04160088 3634 bdi_destroy(&fs_info->bdi);
76dda93c 3635 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 3636
53b381b3
DW		 3637 btrfs_free_stripe_hash_table(fs_info);
3638
1cb048f5
FDBM		 3639 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3640 root->orphan_block_rsv = NULL;
3641
eb60ceac
CM		 3642 return 0;
3643}
3644
b9fab919
CM		 3645int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3646 int atomic)
5f39d397 3647{
1259ab75 3648 int ret;
727011e0 3649 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 3650
0b32f4bb 3651 ret = extent_buffer_uptodate(buf);
1259ab75
CM		 3652 if (!ret)
3653 return ret;
3654
3655 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM		 3656 parent_transid, atomic);
3657 if (ret == -EAGAIN)
3658 return ret;
1259ab75 3659 return !ret;
5f39d397
CM		 3660}
3661
3662int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 3663{
0b32f4bb 3664 return set_extent_buffer_uptodate(buf);
5f39d397 3665}
6702ed49 3666
5f39d397
CM		 3667void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3668{
06ea65a3 3669 struct btrfs_root *root;
5f39d397 3670 u64 transid = btrfs_header_generation(buf);
b9473439 3671 int was_dirty;
b4ce94de 3672
06ea65a3
JB		 3673#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3674 /*
3675 * This is a fast path so only do this check if we have sanity tests
3676 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3677 * outside of the sanity tests.
3678 */
3679 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3680 return;
3681#endif
3682 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
b9447ef8 3683 btrfs_assert_tree_locked(buf);
31b1a2bd
JL		 3684 if (transid != root->fs_info->generation)
3685 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
d397712b 3686 "found %llu running %llu\n",
c1c9ff7c 3687 buf->start, transid, root->fs_info->generation);
0b32f4bb 3688 was_dirty = set_extent_buffer_dirty(buf);
e2d84521
MX		 3689 if (!was_dirty)
3690 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3691 buf->len,
3692 root->fs_info->dirty_metadata_batch);
eb60ceac
CM		 3693}
3694
b53d3f5d
LB		 3695static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3696 int flush_delayed)
16cdcec7
MX		 3697{
3698 /*
3699 * looks as though older kernels can get into trouble with
3700 * this code, they end up stuck in balance_dirty_pages forever
3701 */
e2d84521 3702 int ret;
16cdcec7
MX		 3703
3704 if (current->flags & PF_MEMALLOC)
3705 return;
3706
b53d3f5d
LB		 3707 if (flush_delayed)
3708 btrfs_balance_delayed_items(root);
16cdcec7 3709
e2d84521
MX		 3710 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3711 BTRFS_DIRTY_METADATA_THRESH);
3712 if (ret > 0) {
d0e1d66b
NJ		 3713 balance_dirty_pages_ratelimited(
3714 root->fs_info->btree_inode->i_mapping);
16cdcec7
MX		 3715 }
3716 return;
3717}
3718
b53d3f5d 3719void btrfs_btree_balance_dirty(struct btrfs_root *root)
35b7e476 3720{
b53d3f5d
LB		 3721 __btrfs_btree_balance_dirty(root, 1);
3722}
585ad2c3 3723
b53d3f5d
LB		 3724void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3725{
3726 __btrfs_btree_balance_dirty(root, 0);
35b7e476 3727}
6b80053d 3728
ca7a79ad 3729int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 3730{
727011e0 3731 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
0b32f4bb 3732 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
6b80053d 3733}
0da5468f 3734
fcd1f065 3735static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 3736 int read_only)
3737{
1104a885
DS		 3738 /*
3739 * Placeholder for checks
3740 */
fcd1f065 3741 return 0;
acce952b 3742}
3743
48a3b636 3744static void btrfs_error_commit_super(struct btrfs_root *root)
acce952b 3745{
acce952b 3746 mutex_lock(&root->fs_info->cleaner_mutex);
3747 btrfs_run_delayed_iputs(root);
3748 mutex_unlock(&root->fs_info->cleaner_mutex);
3749
3750 down_write(&root->fs_info->cleanup_work_sem);
3751 up_write(&root->fs_info->cleanup_work_sem);
3752
3753 /* cleanup FS via transaction */
3754 btrfs_cleanup_transaction(root);
acce952b 3755}
3756
569e0f35
JB		 3757static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
3758 struct btrfs_root *root)
acce952b 3759{
3760 struct btrfs_inode *btrfs_inode;
3761 struct list_head splice;
3762
3763 INIT_LIST_HEAD(&splice);
3764
3765 mutex_lock(&root->fs_info->ordered_operations_mutex);
199c2a9c 3766 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3767
569e0f35 3768 list_splice_init(&t->ordered_operations, &splice);
acce952b 3769 while (!list_empty(&splice)) {
3770 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3771 ordered_operations);
3772
3773 list_del_init(&btrfs_inode->ordered_operations);
199c2a9c 3774 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3775
3776 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3777
199c2a9c 3778 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3779 }
3780
199c2a9c 3781 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3782 mutex_unlock(&root->fs_info->ordered_operations_mutex);
acce952b 3783}
3784
143bede5 3785static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 3786{
acce952b 3787 struct btrfs_ordered_extent *ordered;
acce952b 3788
199c2a9c 3789 spin_lock(&root->ordered_extent_lock);
779880ef
JB		 3790 /*
3791 * This will just short circuit the ordered completion stuff which will
3792 * make sure the ordered extent gets properly cleaned up.
3793 */
199c2a9c 3794 list_for_each_entry(ordered, &root->ordered_extents,
779880ef
JB		 3795 root_extent_list)
3796 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
199c2a9c
MX		 3797 spin_unlock(&root->ordered_extent_lock);
3798}
3799
3800static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3801{
3802 struct btrfs_root *root;
3803 struct list_head splice;
3804
3805 INIT_LIST_HEAD(&splice);
3806
3807 spin_lock(&fs_info->ordered_root_lock);
3808 list_splice_init(&fs_info->ordered_roots, &splice);
3809 while (!list_empty(&splice)) {
3810 root = list_first_entry(&splice, struct btrfs_root,
3811 ordered_root);
1de2cfde
JB		 3812 list_move_tail(&root->ordered_root,
3813 &fs_info->ordered_roots);
199c2a9c 3814
2a85d9ca 3815 spin_unlock(&fs_info->ordered_root_lock);
199c2a9c
MX		 3816 btrfs_destroy_ordered_extents(root);
3817
2a85d9ca
LB		 3818 cond_resched();
3819 spin_lock(&fs_info->ordered_root_lock);
199c2a9c
MX		 3820 }
3821 spin_unlock(&fs_info->ordered_root_lock);
acce952b 3822}
3823
35a3621b
SB		 3824static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3825 struct btrfs_root *root)
acce952b 3826{
3827 struct rb_node *node;
3828 struct btrfs_delayed_ref_root *delayed_refs;
3829 struct btrfs_delayed_ref_node *ref;
3830 int ret = 0;
3831
3832 delayed_refs = &trans->delayed_refs;
3833
3834 spin_lock(&delayed_refs->lock);
d7df2c79 3835 if (atomic_read(&delayed_refs->num_entries) == 0) {
cfece4db 3836 spin_unlock(&delayed_refs->lock);
efe120a0 3837 btrfs_info(root->fs_info, "delayed_refs has NO entry");
acce952b 3838 return ret;
3839 }
3840
d7df2c79
JB		 3841 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
3842 struct btrfs_delayed_ref_head *head;
e78417d1 3843 bool pin_bytes = false;
acce952b 3844
d7df2c79
JB		 3845 head = rb_entry(node, struct btrfs_delayed_ref_head,
3846 href_node);
3847 if (!mutex_trylock(&head->mutex)) {
3848 atomic_inc(&head->node.refs);
3849 spin_unlock(&delayed_refs->lock);
acce952b 3850
d7df2c79 3851 mutex_lock(&head->mutex);
e78417d1 3852 mutex_unlock(&head->mutex);
d7df2c79
JB		 3853 btrfs_put_delayed_ref(&head->node);
3854 spin_lock(&delayed_refs->lock);
3855 continue;
3856 }
3857 spin_lock(&head->lock);
3858 while ((node = rb_first(&head->ref_root)) != NULL) {
3859 ref = rb_entry(node, struct btrfs_delayed_ref_node,
3860 rb_node);
3861 ref->in_tree = 0;
3862 rb_erase(&ref->rb_node, &head->ref_root);
3863 atomic_dec(&delayed_refs->num_entries);
3864 btrfs_put_delayed_ref(ref);
e78417d1 3865 }
d7df2c79
JB		 3866 if (head->must_insert_reserved)
3867 pin_bytes = true;
3868 btrfs_free_delayed_extent_op(head->extent_op);
3869 delayed_refs->num_heads--;
3870 if (head->processing == 0)
3871 delayed_refs->num_heads_ready--;
3872 atomic_dec(&delayed_refs->num_entries);
3873 head->node.in_tree = 0;
3874 rb_erase(&head->href_node, &delayed_refs->href_root);
3875 spin_unlock(&head->lock);
3876 spin_unlock(&delayed_refs->lock);
3877 mutex_unlock(&head->mutex);
acce952b 3878
d7df2c79
JB		 3879 if (pin_bytes)
3880 btrfs_pin_extent(root, head->node.bytenr,
3881 head->node.num_bytes, 1);
3882 btrfs_put_delayed_ref(&head->node);
acce952b 3883 cond_resched();
3884 spin_lock(&delayed_refs->lock);
3885 }
3886
3887 spin_unlock(&delayed_refs->lock);
3888
3889 return ret;
3890}
3891
143bede5 3892static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 3893{
3894 struct btrfs_inode *btrfs_inode;
3895 struct list_head splice;
3896
3897 INIT_LIST_HEAD(&splice);
3898
eb73c1b7
MX		 3899 spin_lock(&root->delalloc_lock);
3900 list_splice_init(&root->delalloc_inodes, &splice);
acce952b 3901
3902 while (!list_empty(&splice)) {
eb73c1b7
MX		 3903 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3904 delalloc_inodes);
acce952b 3905
3906 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX		 3907 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3908 &btrfs_inode->runtime_flags);
eb73c1b7 3909 spin_unlock(&root->delalloc_lock);
acce952b 3910
3911 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3912
eb73c1b7 3913 spin_lock(&root->delalloc_lock);
acce952b 3914 }
3915
eb73c1b7
MX		 3916 spin_unlock(&root->delalloc_lock);
3917}
3918
3919static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3920{
3921 struct btrfs_root *root;
3922 struct list_head splice;
3923
3924 INIT_LIST_HEAD(&splice);
3925
3926 spin_lock(&fs_info->delalloc_root_lock);
3927 list_splice_init(&fs_info->delalloc_roots, &splice);
3928 while (!list_empty(&splice)) {
3929 root = list_first_entry(&splice, struct btrfs_root,
3930 delalloc_root);
3931 list_del_init(&root->delalloc_root);
3932 root = btrfs_grab_fs_root(root);
3933 BUG_ON(!root);
3934 spin_unlock(&fs_info->delalloc_root_lock);
3935
3936 btrfs_destroy_delalloc_inodes(root);
3937 btrfs_put_fs_root(root);
3938
3939 spin_lock(&fs_info->delalloc_root_lock);
3940 }
3941 spin_unlock(&fs_info->delalloc_root_lock);
acce952b 3942}
3943
3944static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3945 struct extent_io_tree *dirty_pages,
3946 int mark)
3947{
3948 int ret;
acce952b 3949 struct extent_buffer *eb;
3950 u64 start = 0;
3951 u64 end;
acce952b 3952
3953 while (1) {
3954 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 3955 mark, NULL);
acce952b 3956 if (ret)
3957 break;
3958
3959 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
3960 while (start <= end) {
fd8b2b61
JB		 3961 eb = btrfs_find_tree_block(root, start,
3962 root->leafsize);
69a85bd8 3963 start += root->leafsize;
fd8b2b61 3964 if (!eb)
acce952b 3965 continue;
fd8b2b61 3966 wait_on_extent_buffer_writeback(eb);
acce952b 3967
fd8b2b61
JB		 3968 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
3969 &eb->bflags))
3970 clear_extent_buffer_dirty(eb);
3971 free_extent_buffer_stale(eb);
acce952b 3972 }
3973 }
3974
3975 return ret;
3976}
3977
3978static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3979 struct extent_io_tree *pinned_extents)
3980{
3981 struct extent_io_tree *unpin;
3982 u64 start;
3983 u64 end;
3984 int ret;
ed0eaa14 3985 bool loop = true;
acce952b 3986
3987 unpin = pinned_extents;
ed0eaa14 3988again:
acce952b 3989 while (1) {
3990 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 3991 EXTENT_DIRTY, NULL);
acce952b 3992 if (ret)
3993 break;
3994
3995 /* opt_discard */
5378e607
LD		 3996 if (btrfs_test_opt(root, DISCARD))
3997 ret = btrfs_error_discard_extent(root, start,
3998 end + 1 - start,
3999 NULL);
acce952b 4000
4001 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4002 btrfs_error_unpin_extent_range(root, start, end);
4003 cond_resched();
4004 }
4005
ed0eaa14
LB		 4006 if (loop) {
4007 if (unpin == &root->fs_info->freed_extents[0])
4008 unpin = &root->fs_info->freed_extents[1];
4009 else
4010 unpin = &root->fs_info->freed_extents[0];
4011 loop = false;
4012 goto again;
4013 }
4014
acce952b 4015 return 0;
4016}
4017
49b25e05
JM		 4018void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4019 struct btrfs_root *root)
4020{
724e2315
JB		 4021 btrfs_destroy_ordered_operations(cur_trans, root);
4022
49b25e05 4023 btrfs_destroy_delayed_refs(cur_trans, root);
49b25e05 4024
4a9d8bde 4025 cur_trans->state = TRANS_STATE_COMMIT_START;
d7096fc3 4026 wake_up(&root->fs_info->transaction_blocked_wait);
49b25e05 4027
4a9d8bde 4028 cur_trans->state = TRANS_STATE_UNBLOCKED;
d7096fc3 4029 wake_up(&root->fs_info->transaction_wait);
49b25e05 4030
67cde344
MX		 4031 btrfs_destroy_delayed_inodes(root);
4032 btrfs_assert_delayed_root_empty(root);
49b25e05 4033
49b25e05
JM		 4034 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4035 EXTENT_DIRTY);
6e841e32
LB		 4036 btrfs_destroy_pinned_extent(root,
4037 root->fs_info->pinned_extents);
49b25e05 4038
4a9d8bde
MX		 4039 cur_trans->state =TRANS_STATE_COMPLETED;
4040 wake_up(&cur_trans->commit_wait);
4041
49b25e05
JM		 4042 /*
4043 memset(cur_trans, 0, sizeof(*cur_trans));
4044 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4045 */
4046}
4047
48a3b636 4048static int btrfs_cleanup_transaction(struct btrfs_root *root)
acce952b 4049{
4050 struct btrfs_transaction *t;
acce952b 4051
acce952b 4052 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4053
a4abeea4 4054 spin_lock(&root->fs_info->trans_lock);
724e2315
JB		 4055 while (!list_empty(&root->fs_info->trans_list)) {
4056 t = list_first_entry(&root->fs_info->trans_list,
4057 struct btrfs_transaction, list);
4058 if (t->state >= TRANS_STATE_COMMIT_START) {
4059 atomic_inc(&t->use_count);
4060 spin_unlock(&root->fs_info->trans_lock);
4061 btrfs_wait_for_commit(root, t->transid);
4062 btrfs_put_transaction(t);
4063 spin_lock(&root->fs_info->trans_lock);
4064 continue;
4065 }
4066 if (t == root->fs_info->running_transaction) {
4067 t->state = TRANS_STATE_COMMIT_DOING;
4068 spin_unlock(&root->fs_info->trans_lock);
4069 /*
4070 * We wait for 0 num_writers since we don't hold a trans
4071 * handle open currently for this transaction.
4072 */
4073 wait_event(t->writer_wait,
4074 atomic_read(&t->num_writers) == 0);
4075 } else {
4076 spin_unlock(&root->fs_info->trans_lock);
4077 }
4078 btrfs_cleanup_one_transaction(t, root);
4a9d8bde 4079
724e2315
JB		 4080 spin_lock(&root->fs_info->trans_lock);
4081 if (t == root->fs_info->running_transaction)
4082 root->fs_info->running_transaction = NULL;
acce952b 4083 list_del_init(&t->list);
724e2315 4084 spin_unlock(&root->fs_info->trans_lock);
acce952b 4085
724e2315
JB		 4086 btrfs_put_transaction(t);
4087 trace_btrfs_transaction_commit(root);
4088 spin_lock(&root->fs_info->trans_lock);
4089 }
4090 spin_unlock(&root->fs_info->trans_lock);
4091 btrfs_destroy_all_ordered_extents(root->fs_info);
4092 btrfs_destroy_delayed_inodes(root);
4093 btrfs_assert_delayed_root_empty(root);
4094 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4095 btrfs_destroy_all_delalloc_inodes(root->fs_info);
acce952b 4096 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 4097
4098 return 0;
4099}
4100
d1310b2e 4101static struct extent_io_ops btree_extent_io_ops = {
ce9adaa5 4102 .readpage_end_io_hook = btree_readpage_end_io_hook,
4bb31e92 4103 .readpage_io_failed_hook = btree_io_failed_hook,
0b86a832 4104 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 4105 /* note we're sharing with inode.c for the merge bio hook */
4106 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 4107};
Linux 6.x block layer and io_uring tree(s)