projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ALSA: ice1712: add suspend support for ICE1712 chip
[deliverable/linux.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{
2c30c71b 842 struct bio_vec *bvec;
ce3ed71a 843 struct btrfs_root *root;
2c30c71b 844 int i, ret = 0;
ce3ed71a 845
2c30c71b 846 bio_for_each_segment_all(bvec, bio, i) {
ce3ed71a 847 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
79787eaa
JM		 848 ret = csum_dirty_buffer(root, bvec->bv_page);
849 if (ret)
850 break;
ce3ed71a 851 }
2c30c71b 852
79787eaa 853 return ret;
ce3ed71a
CM		 854}
855
4a69a410
CM		 856static int __btree_submit_bio_start(struct inode *inode, int rw,
857 struct bio *bio, int mirror_num,
eaf25d93
CM		 858 unsigned long bio_flags,
859 u64 bio_offset)
22c59948 860{
8b712842
CM		 861 /*
862 * when we're called for a write, we're already in the async
5443be45 863 * submission context. Just jump into btrfs_map_bio
8b712842 864 */
79787eaa 865 return btree_csum_one_bio(bio);
4a69a410 866}
22c59948 867
4a69a410 868static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 869 int mirror_num, unsigned long bio_flags,
870 u64 bio_offset)
4a69a410 871{
61891923
SB		 872 int ret;
873
8b712842 874 /*
4a69a410
CM		 875 * when we're called for a write, we're already in the async
876 * submission context. Just jump into btrfs_map_bio
8b712842 877 */
61891923
SB		 878 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
879 if (ret)
880 bio_endio(bio, ret);
881 return ret;
0b86a832
CM		 882}
883
de0022b9
JB		 884static int check_async_write(struct inode *inode, unsigned long bio_flags)
885{
886 if (bio_flags & EXTENT_BIO_TREE_LOG)
887 return 0;
888#ifdef CONFIG_X86
889 if (cpu_has_xmm4_2)
890 return 0;
891#endif
892 return 1;
893}
894
44b8bd7e 895static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 896 int mirror_num, unsigned long bio_flags,
897 u64 bio_offset)
44b8bd7e 898{
de0022b9 899 int async = check_async_write(inode, bio_flags);
cad321ad
CM		 900 int ret;
901
7b6d91da 902 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 903 /*
904 * called for a read, do the setup so that checksum validation
905 * can happen in the async kernel threads
906 */
f3f266ab
CM		 907 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
908 bio, 1);
1d4284bd 909 if (ret)
61891923
SB		 910 goto out_w_error;
911 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
912 mirror_num, 0);
de0022b9
JB		 913 } else if (!async) {
914 ret = btree_csum_one_bio(bio);
915 if (ret)
61891923
SB		 916 goto out_w_error;
917 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
918 mirror_num, 0);
919 } else {
920 /*
921 * kthread helpers are used to submit writes so that
922 * checksumming can happen in parallel across all CPUs
923 */
924 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
925 inode, rw, bio, mirror_num, 0,
926 bio_offset,
927 __btree_submit_bio_start,
928 __btree_submit_bio_done);
44b8bd7e 929 }
d313d7a3 930
61891923
SB		 931 if (ret) {
932out_w_error:
933 bio_endio(bio, ret);
934 }
935 return ret;
44b8bd7e
CM		 936}
937
3dd1462e 938#ifdef CONFIG_MIGRATION
784b4e29 939static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG		 940 struct page *newpage, struct page *page,
941 enum migrate_mode mode)
784b4e29
CM		 942{
943 /*
944 * we can't safely write a btree page from here,
945 * we haven't done the locking hook
946 */
947 if (PageDirty(page))
948 return -EAGAIN;
949 /*
950 * Buffers may be managed in a filesystem specific way.
951 * We must have no buffers or drop them.
952 */
953 if (page_has_private(page) &&
954 !try_to_release_page(page, GFP_KERNEL))
955 return -EAGAIN;
a6bc32b8 956 return migrate_page(mapping, newpage, page, mode);
784b4e29 957}
3dd1462e 958#endif
784b4e29 959
0da5468f
CM		 960
961static int btree_writepages(struct address_space *mapping,
962 struct writeback_control *wbc)
963{
e2d84521
MX		 964 struct btrfs_fs_info *fs_info;
965 int ret;
966
d8d5f3e1 967 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM		 968
969 if (wbc->for_kupdate)
970 return 0;
971
e2d84521 972 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 973 /* this is a bit racy, but that's ok */
e2d84521
MX		 974 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
975 BTRFS_DIRTY_METADATA_THRESH);
976 if (ret < 0)
793955bc 977 return 0;
793955bc 978 }
0b32f4bb 979 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM		 980}
981
b2950863 982static int btree_readpage(struct file *file, struct page *page)
5f39d397 983{
d1310b2e
CM		 984 struct extent_io_tree *tree;
985 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 986 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 987}
22b0ebda 988
70dec807 989static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 990{
98509cfc 991 if (PageWriteback(page) || PageDirty(page))
d397712b 992 return 0;
0c4e538b 993
f7a52a40 994 return try_release_extent_buffer(page);
d98237b3
CM		 995}
996
d47992f8
LC		 997static void btree_invalidatepage(struct page *page, unsigned int offset,
998 unsigned int length)
d98237b3 999{
d1310b2e
CM		 1000 struct extent_io_tree *tree;
1001 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 1002 extent_invalidatepage(tree, page, offset);
1003 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 1004 if (PagePrivate(page)) {
efe120a0
FH		 1005 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1006 "page private not zero on page %llu",
1007 (unsigned long long)page_offset(page));
9ad6b7bc
CM		 1008 ClearPagePrivate(page);
1009 set_page_private(page, 0);
1010 page_cache_release(page);
1011 }
d98237b3
CM		 1012}
1013
0b32f4bb
JB		 1014static int btree_set_page_dirty(struct page *page)
1015{
bb146eb2 1016#ifdef DEBUG
0b32f4bb
JB		 1017 struct extent_buffer *eb;
1018
1019 BUG_ON(!PagePrivate(page));
1020 eb = (struct extent_buffer *)page->private;
1021 BUG_ON(!eb);
1022 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1023 BUG_ON(!atomic_read(&eb->refs));
1024 btrfs_assert_tree_locked(eb);
bb146eb2 1025#endif
0b32f4bb
JB		 1026 return __set_page_dirty_nobuffers(page);
1027}
1028
7f09410b 1029static const struct address_space_operations btree_aops = {
d98237b3 1030 .readpage = btree_readpage,
0da5468f 1031 .writepages = btree_writepages,
5f39d397
CM		 1032 .releasepage = btree_releasepage,
1033 .invalidatepage = btree_invalidatepage,
5a92bc88 1034#ifdef CONFIG_MIGRATION
784b4e29 1035 .migratepage = btree_migratepage,
5a92bc88 1036#endif
0b32f4bb 1037 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM		 1038};
1039
ca7a79ad
CM		 1040int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1041 u64 parent_transid)
090d1875 1042{
5f39d397
CM		 1043 struct extent_buffer *buf = NULL;
1044 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 1045 int ret = 0;
090d1875 1046
db94535d 1047 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 1048 if (!buf)
090d1875 1049 return 0;
d1310b2e 1050 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
bb82ab88 1051 buf, 0, WAIT_NONE, btree_get_extent, 0);
5f39d397 1052 free_extent_buffer(buf);
de428b63 1053 return ret;
090d1875
CM		 1054}
1055
ab0fff03
AJ		 1056int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1057 int mirror_num, struct extent_buffer **eb)
1058{
1059 struct extent_buffer *buf = NULL;
1060 struct inode *btree_inode = root->fs_info->btree_inode;
1061 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1062 int ret;
1063
1064 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1065 if (!buf)
1066 return 0;
1067
1068 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1069
1070 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1071 btree_get_extent, mirror_num);
1072 if (ret) {
1073 free_extent_buffer(buf);
1074 return ret;
1075 }
1076
1077 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1078 free_extent_buffer(buf);
1079 return -EIO;
0b32f4bb 1080 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ		 1081 *eb = buf;
1082 } else {
1083 free_extent_buffer(buf);
1084 }
1085 return 0;
1086}
1087
0999df54
CM		 1088struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1089 u64 bytenr, u32 blocksize)
1090{
f28491e0 1091 return find_extent_buffer(root->fs_info, bytenr);
0999df54
CM		 1092}
1093
1094struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1095 u64 bytenr, u32 blocksize)
1096{
f28491e0 1097 return alloc_extent_buffer(root->fs_info, bytenr, blocksize);
0999df54
CM		 1098}
1099
1100
e02119d5
CM		 1101int btrfs_write_tree_block(struct extent_buffer *buf)
1102{
727011e0 1103 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1104 buf->start + buf->len - 1);
e02119d5
CM		 1105}
1106
1107int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1108{
727011e0 1109 return filemap_fdatawait_range(buf->pages[0]->mapping,
8aa38c31 1110 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1111}
1112
0999df54 1113struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 1114 u32 blocksize, u64 parent_transid)
0999df54
CM		 1115{
1116 struct extent_buffer *buf = NULL;
0999df54
CM		 1117 int ret;
1118
0999df54
CM		 1119 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1120 if (!buf)
1121 return NULL;
0999df54 1122
ca7a79ad 1123 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
0f0fe8f7
FDBM		 1124 if (ret) {
1125 free_extent_buffer(buf);
1126 return NULL;
1127 }
5f39d397 1128 return buf;
ce9adaa5 1129
eb60ceac
CM		 1130}
1131
d5c13f92
JM		 1132void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1133 struct extent_buffer *buf)
ed2ff2cb 1134{
e2d84521
MX		 1135 struct btrfs_fs_info *fs_info = root->fs_info;
1136
55c69072 1137 if (btrfs_header_generation(buf) ==
e2d84521 1138 fs_info->running_transaction->transid) {
b9447ef8 1139 btrfs_assert_tree_locked(buf);
b4ce94de 1140
b9473439 1141 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
e2d84521
MX		 1142 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1143 -buf->len,
1144 fs_info->dirty_metadata_batch);
ed7b63eb
JB		 1145 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1146 btrfs_set_lock_blocking(buf);
1147 clear_extent_buffer_dirty(buf);
1148 }
925baedd 1149 }
5f39d397
CM		 1150}
1151
143bede5
JM		 1152static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1153 u32 stripesize, struct btrfs_root *root,
1154 struct btrfs_fs_info *fs_info,
1155 u64 objectid)
d97e63b6 1156{
cfaa7295 1157 root->node = NULL;
a28ec197 1158 root->commit_root = NULL;
db94535d
CM		 1159 root->sectorsize = sectorsize;
1160 root->nodesize = nodesize;
1161 root->leafsize = leafsize;
87ee04eb 1162 root->stripesize = stripesize;
123abc88 1163 root->ref_cows = 0;
0b86a832 1164 root->track_dirty = 0;
c71bf099 1165 root->in_radix = 0;
d68fc57b
YZ		 1166 root->orphan_item_inserted = 0;
1167 root->orphan_cleanup_state = 0;
0b86a832 1168
0f7d52f4
CM		 1169 root->objectid = objectid;
1170 root->last_trans = 0;
13a8a7c8 1171 root->highest_objectid = 0;
eb73c1b7 1172 root->nr_delalloc_inodes = 0;
199c2a9c 1173 root->nr_ordered_extents = 0;
58176a96 1174 root->name = NULL;
6bef4d31 1175 root->inode_tree = RB_ROOT;
16cdcec7 1176 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1177 root->block_rsv = NULL;
d68fc57b 1178 root->orphan_block_rsv = NULL;
0b86a832
CM		 1179
1180 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1181 INIT_LIST_HEAD(&root->root_list);
eb73c1b7
MX		 1182 INIT_LIST_HEAD(&root->delalloc_inodes);
1183 INIT_LIST_HEAD(&root->delalloc_root);
199c2a9c
MX		 1184 INIT_LIST_HEAD(&root->ordered_extents);
1185 INIT_LIST_HEAD(&root->ordered_root);
2ab28f32
JB		 1186 INIT_LIST_HEAD(&root->logged_list[0]);
1187 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1188 spin_lock_init(&root->orphan_lock);
5d4f98a2 1189 spin_lock_init(&root->inode_lock);
eb73c1b7 1190 spin_lock_init(&root->delalloc_lock);
199c2a9c 1191 spin_lock_init(&root->ordered_extent_lock);
f0486c68 1192 spin_lock_init(&root->accounting_lock);
2ab28f32
JB		 1193 spin_lock_init(&root->log_extents_lock[0]);
1194 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1195 mutex_init(&root->objectid_mutex);
e02119d5 1196 mutex_init(&root->log_mutex);
7237f183
YZ		 1197 init_waitqueue_head(&root->log_writer_wait);
1198 init_waitqueue_head(&root->log_commit_wait[0]);
1199 init_waitqueue_head(&root->log_commit_wait[1]);
1200 atomic_set(&root->log_commit[0], 0);
1201 atomic_set(&root->log_commit[1], 0);
1202 atomic_set(&root->log_writers, 0);
2ecb7923 1203 atomic_set(&root->log_batch, 0);
8a35d95f 1204 atomic_set(&root->orphan_inodes, 0);
b0feb9d9 1205 atomic_set(&root->refs, 1);
7237f183 1206 root->log_transid = 0;
257c62e1 1207 root->last_log_commit = 0;
06ea65a3
JB		 1208 if (fs_info)
1209 extent_io_tree_init(&root->dirty_log_pages,
1210 fs_info->btree_inode->i_mapping);
017e5369 1211
3768f368
CM		 1212 memset(&root->root_key, 0, sizeof(root->root_key));
1213 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1214 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1215 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
06ea65a3
JB		 1216 if (fs_info)
1217 root->defrag_trans_start = fs_info->generation;
1218 else
1219 root->defrag_trans_start = 0;
58176a96 1220 init_completion(&root->kobj_unregister);
6702ed49 1221 root->defrag_running = 0;
4d775673 1222 root->root_key.objectid = objectid;
0ee5dc67 1223 root->anon_dev = 0;
8ea05e3a 1224
5f3ab90a 1225 spin_lock_init(&root->root_item_lock);
3768f368
CM		 1226}
1227
f84a8bd6 1228static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
6f07e42e
AV		 1229{
1230 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1231 if (root)
1232 root->fs_info = fs_info;
1233 return root;
1234}
1235
06ea65a3
JB		 1236#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1237/* Should only be used by the testing infrastructure */
1238struct btrfs_root *btrfs_alloc_dummy_root(void)
1239{
1240 struct btrfs_root *root;
1241
1242 root = btrfs_alloc_root(NULL);
1243 if (!root)
1244 return ERR_PTR(-ENOMEM);
1245 __setup_root(4096, 4096, 4096, 4096, root, NULL, 1);
1246 root->dummy_root = 1;
1247
1248 return root;
1249}
1250#endif
1251
20897f5c
AJ		 1252struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1253 struct btrfs_fs_info *fs_info,
1254 u64 objectid)
1255{
1256 struct extent_buffer *leaf;
1257 struct btrfs_root *tree_root = fs_info->tree_root;
1258 struct btrfs_root *root;
1259 struct btrfs_key key;
1260 int ret = 0;
6463fe58 1261 uuid_le uuid;
20897f5c
AJ		 1262
1263 root = btrfs_alloc_root(fs_info);
1264 if (!root)
1265 return ERR_PTR(-ENOMEM);
1266
1267 __setup_root(tree_root->nodesize, tree_root->leafsize,
1268 tree_root->sectorsize, tree_root->stripesize,
1269 root, fs_info, objectid);
1270 root->root_key.objectid = objectid;
1271 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1272 root->root_key.offset = 0;
1273
1274 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1275 0, objectid, NULL, 0, 0, 0);
1276 if (IS_ERR(leaf)) {
1277 ret = PTR_ERR(leaf);
1dd05682 1278 leaf = NULL;
20897f5c
AJ		 1279 goto fail;
1280 }
1281
20897f5c
AJ		 1282 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1283 btrfs_set_header_bytenr(leaf, leaf->start);
1284 btrfs_set_header_generation(leaf, trans->transid);
1285 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1286 btrfs_set_header_owner(leaf, objectid);
1287 root->node = leaf;
1288
0a4e5586 1289 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
20897f5c
AJ		 1290 BTRFS_FSID_SIZE);
1291 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
b308bc2f 1292 btrfs_header_chunk_tree_uuid(leaf),
20897f5c
AJ		 1293 BTRFS_UUID_SIZE);
1294 btrfs_mark_buffer_dirty(leaf);
1295
1296 root->commit_root = btrfs_root_node(root);
1297 root->track_dirty = 1;
1298
1299
1300 root->root_item.flags = 0;
1301 root->root_item.byte_limit = 0;
1302 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1303 btrfs_set_root_generation(&root->root_item, trans->transid);
1304 btrfs_set_root_level(&root->root_item, 0);
1305 btrfs_set_root_refs(&root->root_item, 1);
1306 btrfs_set_root_used(&root->root_item, leaf->len);
1307 btrfs_set_root_last_snapshot(&root->root_item, 0);
1308 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB		 1309 uuid_le_gen(&uuid);
1310 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ		 1311 root->root_item.drop_level = 0;
1312
1313 key.objectid = objectid;
1314 key.type = BTRFS_ROOT_ITEM_KEY;
1315 key.offset = 0;
1316 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1317 if (ret)
1318 goto fail;
1319
1320 btrfs_tree_unlock(leaf);
1321
1dd05682
TI		 1322 return root;
1323
20897f5c 1324fail:
1dd05682
TI		 1325 if (leaf) {
1326 btrfs_tree_unlock(leaf);
1327 free_extent_buffer(leaf);
1328 }
1329 kfree(root);
20897f5c 1330
1dd05682 1331 return ERR_PTR(ret);
20897f5c
AJ		 1332}
1333
7237f183
YZ		 1334static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1335 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1336{
1337 struct btrfs_root *root;
1338 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1339 struct extent_buffer *leaf;
e02119d5 1340
6f07e42e 1341 root = btrfs_alloc_root(fs_info);
e02119d5 1342 if (!root)
7237f183 1343 return ERR_PTR(-ENOMEM);
e02119d5
CM		 1344
1345 __setup_root(tree_root->nodesize, tree_root->leafsize,
1346 tree_root->sectorsize, tree_root->stripesize,
1347 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1348
1349 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1350 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1351 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
7237f183
YZ		 1352 /*
1353 * log trees do not get reference counted because they go away
1354 * before a real commit is actually done. They do store pointers
1355 * to file data extents, and those reference counts still get
1356 * updated (along with back refs to the log tree).
1357 */
e02119d5
CM		 1358 root->ref_cows = 0;
1359
5d4f98a2 1360 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
66d7e7f0 1361 BTRFS_TREE_LOG_OBJECTID, NULL,
5581a51a 1362 0, 0, 0);
7237f183
YZ		 1363 if (IS_ERR(leaf)) {
1364 kfree(root);
1365 return ERR_CAST(leaf);
1366 }
e02119d5 1367
5d4f98a2
YZ		 1368 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1369 btrfs_set_header_bytenr(leaf, leaf->start);
1370 btrfs_set_header_generation(leaf, trans->transid);
1371 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1372 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1373 root->node = leaf;
e02119d5
CM		 1374
1375 write_extent_buffer(root->node, root->fs_info->fsid,
0a4e5586 1376 btrfs_header_fsid(), BTRFS_FSID_SIZE);
e02119d5
CM		 1377 btrfs_mark_buffer_dirty(root->node);
1378 btrfs_tree_unlock(root->node);
7237f183
YZ		 1379 return root;
1380}
1381
1382int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1383 struct btrfs_fs_info *fs_info)
1384{
1385 struct btrfs_root *log_root;
1386
1387 log_root = alloc_log_tree(trans, fs_info);
1388 if (IS_ERR(log_root))
1389 return PTR_ERR(log_root);
1390 WARN_ON(fs_info->log_root_tree);
1391 fs_info->log_root_tree = log_root;
1392 return 0;
1393}
1394
1395int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1396 struct btrfs_root *root)
1397{
1398 struct btrfs_root *log_root;
1399 struct btrfs_inode_item *inode_item;
1400
1401 log_root = alloc_log_tree(trans, root->fs_info);
1402 if (IS_ERR(log_root))
1403 return PTR_ERR(log_root);
1404
1405 log_root->last_trans = trans->transid;
1406 log_root->root_key.offset = root->root_key.objectid;
1407
1408 inode_item = &log_root->root_item.inode;
3cae210f
QW		 1409 btrfs_set_stack_inode_generation(inode_item, 1);
1410 btrfs_set_stack_inode_size(inode_item, 3);
1411 btrfs_set_stack_inode_nlink(inode_item, 1);
1412 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
1413 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
7237f183 1414
5d4f98a2 1415 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1416
1417 WARN_ON(root->log_root);
1418 root->log_root = log_root;
1419 root->log_transid = 0;
257c62e1 1420 root->last_log_commit = 0;
e02119d5
CM		 1421 return 0;
1422}
1423
35a3621b
SB		 1424static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1425 struct btrfs_key *key)
e02119d5
CM		 1426{
1427 struct btrfs_root *root;
1428 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1429 struct btrfs_path *path;
84234f3a 1430 u64 generation;
db94535d 1431 u32 blocksize;
cb517eab 1432 int ret;
0f7d52f4 1433
cb517eab
MX		 1434 path = btrfs_alloc_path();
1435 if (!path)
0f7d52f4 1436 return ERR_PTR(-ENOMEM);
cb517eab
MX		 1437
1438 root = btrfs_alloc_root(fs_info);
1439 if (!root) {
1440 ret = -ENOMEM;
1441 goto alloc_fail;
0f7d52f4
CM		 1442 }
1443
db94535d 1444 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb 1445 tree_root->sectorsize, tree_root->stripesize,
cb517eab 1446 root, fs_info, key->objectid);
0f7d52f4 1447
cb517eab
MX		 1448 ret = btrfs_find_root(tree_root, key, path,
1449 &root->root_item, &root->root_key);
0f7d52f4 1450 if (ret) {
13a8a7c8
YZ		 1451 if (ret > 0)
1452 ret = -ENOENT;
cb517eab 1453 goto find_fail;
0f7d52f4 1454 }
13a8a7c8 1455
84234f3a 1456 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1457 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1458 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1459 blocksize, generation);
cb517eab
MX		 1460 if (!root->node) {
1461 ret = -ENOMEM;
1462 goto find_fail;
1463 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1464 ret = -EIO;
1465 goto read_fail;
416bc658 1466 }
5d4f98a2 1467 root->commit_root = btrfs_root_node(root);
13a8a7c8 1468out:
cb517eab
MX		 1469 btrfs_free_path(path);
1470 return root;
1471
1472read_fail:
1473 free_extent_buffer(root->node);
1474find_fail:
1475 kfree(root);
1476alloc_fail:
1477 root = ERR_PTR(ret);
1478 goto out;
1479}
1480
1481struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1482 struct btrfs_key *location)
1483{
1484 struct btrfs_root *root;
1485
1486 root = btrfs_read_tree_root(tree_root, location);
1487 if (IS_ERR(root))
1488 return root;
1489
1490 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
e02119d5 1491 root->ref_cows = 1;
08fe4db1
LZ		 1492 btrfs_check_and_init_root_item(&root->root_item);
1493 }
13a8a7c8 1494
5eda7b5e
CM		 1495 return root;
1496}
1497
cb517eab
MX		 1498int btrfs_init_fs_root(struct btrfs_root *root)
1499{
1500 int ret;
1501
1502 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1503 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1504 GFP_NOFS);
1505 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1506 ret = -ENOMEM;
1507 goto fail;
1508 }
1509
1510 btrfs_init_free_ino_ctl(root);
1511 mutex_init(&root->fs_commit_mutex);
1512 spin_lock_init(&root->cache_lock);
1513 init_waitqueue_head(&root->cache_wait);
1514
1515 ret = get_anon_bdev(&root->anon_dev);
1516 if (ret)
1517 goto fail;
1518 return 0;
1519fail:
1520 kfree(root->free_ino_ctl);
1521 kfree(root->free_ino_pinned);
1522 return ret;
1523}
1524
171170c1
ST		 1525static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1526 u64 root_id)
cb517eab
MX		 1527{
1528 struct btrfs_root *root;
1529
1530 spin_lock(&fs_info->fs_roots_radix_lock);
1531 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1532 (unsigned long)root_id);
1533 spin_unlock(&fs_info->fs_roots_radix_lock);
1534 return root;
1535}
1536
1537int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1538 struct btrfs_root *root)
1539{
1540 int ret;
1541
1542 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1543 if (ret)
1544 return ret;
1545
1546 spin_lock(&fs_info->fs_roots_radix_lock);
1547 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1548 (unsigned long)root->root_key.objectid,
1549 root);
1550 if (ret == 0)
1551 root->in_radix = 1;
1552 spin_unlock(&fs_info->fs_roots_radix_lock);
1553 radix_tree_preload_end();
1554
1555 return ret;
1556}
1557
c00869f1
MX		 1558struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1559 struct btrfs_key *location,
1560 bool check_ref)
5eda7b5e
CM		 1561{
1562 struct btrfs_root *root;
1563 int ret;
1564
edbd8d4e
CM		 1565 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1566 return fs_info->tree_root;
1567 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1568 return fs_info->extent_root;
8f18cf13
CM		 1569 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1570 return fs_info->chunk_root;
1571 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1572 return fs_info->dev_root;
0403e47e
YZ		 1573 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1574 return fs_info->csum_root;
bcef60f2
AJ		 1575 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1576 return fs_info->quota_root ? fs_info->quota_root :
1577 ERR_PTR(-ENOENT);
f7a81ea4
SB		 1578 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1579 return fs_info->uuid_root ? fs_info->uuid_root :
1580 ERR_PTR(-ENOENT);
4df27c4d 1581again:
cb517eab 1582 root = btrfs_lookup_fs_root(fs_info, location->objectid);
48475471 1583 if (root) {
c00869f1 1584 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
48475471 1585 return ERR_PTR(-ENOENT);
5eda7b5e 1586 return root;
48475471 1587 }
5eda7b5e 1588
cb517eab 1589 root = btrfs_read_fs_root(fs_info->tree_root, location);
5eda7b5e
CM		 1590 if (IS_ERR(root))
1591 return root;
3394e160 1592
c00869f1 1593 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
cb517eab 1594 ret = -ENOENT;
581bb050 1595 goto fail;
35a30d7c 1596 }
581bb050 1597
cb517eab 1598 ret = btrfs_init_fs_root(root);
ac08aedf
CM		 1599 if (ret)
1600 goto fail;
3394e160 1601
3f870c28
KN		 1602 ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID,
1603 location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL);
d68fc57b
YZ		 1604 if (ret < 0)
1605 goto fail;
1606 if (ret == 0)
1607 root->orphan_item_inserted = 1;
1608
cb517eab 1609 ret = btrfs_insert_fs_root(fs_info, root);
0f7d52f4 1610 if (ret) {
4df27c4d
YZ		 1611 if (ret == -EEXIST) {
1612 free_fs_root(root);
1613 goto again;
1614 }
1615 goto fail;
0f7d52f4 1616 }
edbd8d4e 1617 return root;
4df27c4d
YZ		 1618fail:
1619 free_fs_root(root);
1620 return ERR_PTR(ret);
edbd8d4e
CM		 1621}
1622
04160088
CM		 1623static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1624{
1625 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1626 int ret = 0;
04160088
CM		 1627 struct btrfs_device *device;
1628 struct backing_dev_info *bdi;
b7967db7 1629
1f78160c
XG		 1630 rcu_read_lock();
1631 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1632 if (!device->bdev)
1633 continue;
04160088
CM		 1634 bdi = blk_get_backing_dev_info(device->bdev);
1635 if (bdi && bdi_congested(bdi, bdi_bits)) {
1636 ret = 1;
1637 break;
1638 }
1639 }
1f78160c 1640 rcu_read_unlock();
04160088
CM		 1641 return ret;
1642}
1643
ad081f14
JA		 1644/*
1645 * If this fails, caller must call bdi_destroy() to get rid of the
1646 * bdi again.
1647 */
04160088
CM		 1648static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1649{
ad081f14
JA		 1650 int err;
1651
1652 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1653 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1654 if (err)
1655 return err;
1656
4575c9cc 1657 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1658 bdi->congested_fn = btrfs_congested_fn;
1659 bdi->congested_data = info;
1660 return 0;
1661}
1662
8b712842
CM		 1663/*
1664 * called by the kthread helper functions to finally call the bio end_io
1665 * functions. This is where read checksum verification actually happens
1666 */
1667static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1668{
ce9adaa5 1669 struct bio *bio;
8b712842 1670 struct end_io_wq *end_io_wq;
ce9adaa5 1671 int error;
ce9adaa5 1672
8b712842
CM		 1673 end_io_wq = container_of(work, struct end_io_wq, work);
1674 bio = end_io_wq->bio;
ce9adaa5 1675
8b712842
CM		 1676 error = end_io_wq->error;
1677 bio->bi_private = end_io_wq->private;
1678 bio->bi_end_io = end_io_wq->end_io;
1679 kfree(end_io_wq);
bc1e79ac 1680 bio_endio_nodec(bio, error);
44b8bd7e
CM		 1681}
1682
a74a4b97
CM		 1683static int cleaner_kthread(void *arg)
1684{
1685 struct btrfs_root *root = arg;
d0278245 1686 int again;
a74a4b97
CM		 1687
1688 do {
d0278245 1689 again = 0;
a74a4b97 1690
d0278245 1691 /* Make the cleaner go to sleep early. */
babbf170 1692 if (btrfs_need_cleaner_sleep(root))
d0278245
MX		 1693 goto sleep;
1694
1695 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1696 goto sleep;
1697
dc7f370c
MX		 1698 /*
1699 * Avoid the problem that we change the status of the fs
1700 * during the above check and trylock.
1701 */
babbf170 1702 if (btrfs_need_cleaner_sleep(root)) {
dc7f370c
MX		 1703 mutex_unlock(&root->fs_info->cleaner_mutex);
1704 goto sleep;
76dda93c 1705 }
a74a4b97 1706
d0278245
MX		 1707 btrfs_run_delayed_iputs(root);
1708 again = btrfs_clean_one_deleted_snapshot(root);
1709 mutex_unlock(&root->fs_info->cleaner_mutex);
1710
1711 /*
05323cd1
MX		 1712 * The defragger has dealt with the R/O remount and umount,
1713 * needn't do anything special here.
d0278245
MX		 1714 */
1715 btrfs_run_defrag_inodes(root->fs_info);
1716sleep:
9d1a2a3a 1717 if (!try_to_freeze() && !again) {
a74a4b97 1718 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1719 if (!kthread_should_stop())
1720 schedule();
a74a4b97
CM		 1721 __set_current_state(TASK_RUNNING);
1722 }
1723 } while (!kthread_should_stop());
1724 return 0;
1725}
1726
1727static int transaction_kthread(void *arg)
1728{
1729 struct btrfs_root *root = arg;
1730 struct btrfs_trans_handle *trans;
1731 struct btrfs_transaction *cur;
8929ecfa 1732 u64 transid;
a74a4b97
CM		 1733 unsigned long now;
1734 unsigned long delay;
914b2007 1735 bool cannot_commit;
a74a4b97
CM		 1736
1737 do {
914b2007 1738 cannot_commit = false;
8b87dc17 1739 delay = HZ * root->fs_info->commit_interval;
a74a4b97
CM		 1740 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1741
a4abeea4 1742 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1743 cur = root->fs_info->running_transaction;
1744 if (!cur) {
a4abeea4 1745 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1746 goto sleep;
1747 }
31153d81 1748
a74a4b97 1749 now = get_seconds();
4a9d8bde 1750 if (cur->state < TRANS_STATE_BLOCKED &&
8b87dc17
DS		 1751 (now < cur->start_time ||
1752 now - cur->start_time < root->fs_info->commit_interval)) {
a4abeea4 1753 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1754 delay = HZ * 5;
1755 goto sleep;
1756 }
8929ecfa 1757 transid = cur->transid;
a4abeea4 1758 spin_unlock(&root->fs_info->trans_lock);
56bec294 1759
79787eaa 1760 /* If the file system is aborted, this will always fail. */
354aa0fb 1761 trans = btrfs_attach_transaction(root);
914b2007 1762 if (IS_ERR(trans)) {
354aa0fb
MX		 1763 if (PTR_ERR(trans) != -ENOENT)
1764 cannot_commit = true;
79787eaa 1765 goto sleep;
914b2007 1766 }
8929ecfa 1767 if (transid == trans->transid) {
79787eaa 1768 btrfs_commit_transaction(trans, root);
8929ecfa
YZ		 1769 } else {
1770 btrfs_end_transaction(trans, root);
1771 }
a74a4b97
CM		 1772sleep:
1773 wake_up_process(root->fs_info->cleaner_kthread);
1774 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1775
4e121c06
JB		 1776 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1777 &root->fs_info->fs_state)))
1778 btrfs_cleanup_transaction(root);
a0acae0e 1779 if (!try_to_freeze()) {
a74a4b97 1780 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa 1781 if (!kthread_should_stop() &&
914b2007
JK		 1782 (!btrfs_transaction_blocked(root->fs_info) ||
1783 cannot_commit))
8929ecfa 1784 schedule_timeout(delay);
a74a4b97
CM		 1785 __set_current_state(TASK_RUNNING);
1786 }
1787 } while (!kthread_should_stop());
1788 return 0;
1789}
1790
af31f5e5
CM		 1791/*
1792 * this will find the highest generation in the array of
1793 * root backups. The index of the highest array is returned,
1794 * or -1 if we can't find anything.
1795 *
1796 * We check to make sure the array is valid by comparing the
1797 * generation of the latest root in the array with the generation
1798 * in the super block. If they don't match we pitch it.
1799 */
1800static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1801{
1802 u64 cur;
1803 int newest_index = -1;
1804 struct btrfs_root_backup *root_backup;
1805 int i;
1806
1807 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1808 root_backup = info->super_copy->super_roots + i;
1809 cur = btrfs_backup_tree_root_gen(root_backup);
1810 if (cur == newest_gen)
1811 newest_index = i;
1812 }
1813
1814 /* check to see if we actually wrapped around */
1815 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1816 root_backup = info->super_copy->super_roots;
1817 cur = btrfs_backup_tree_root_gen(root_backup);
1818 if (cur == newest_gen)
1819 newest_index = 0;
1820 }
1821 return newest_index;
1822}
1823
1824
1825/*
1826 * find the oldest backup so we know where to store new entries
1827 * in the backup array. This will set the backup_root_index
1828 * field in the fs_info struct
1829 */
1830static void find_oldest_super_backup(struct btrfs_fs_info *info,
1831 u64 newest_gen)
1832{
1833 int newest_index = -1;
1834
1835 newest_index = find_newest_super_backup(info, newest_gen);
1836 /* if there was garbage in there, just move along */
1837 if (newest_index == -1) {
1838 info->backup_root_index = 0;
1839 } else {
1840 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1841 }
1842}
1843
1844/*
1845 * copy all the root pointers into the super backup array.
1846 * this will bump the backup pointer by one when it is
1847 * done
1848 */
1849static void backup_super_roots(struct btrfs_fs_info *info)
1850{
1851 int next_backup;
1852 struct btrfs_root_backup *root_backup;
1853 int last_backup;
1854
1855 next_backup = info->backup_root_index;
1856 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1857 BTRFS_NUM_BACKUP_ROOTS;
1858
1859 /*
1860 * just overwrite the last backup if we're at the same generation
1861 * this happens only at umount
1862 */
1863 root_backup = info->super_for_commit->super_roots + last_backup;
1864 if (btrfs_backup_tree_root_gen(root_backup) ==
1865 btrfs_header_generation(info->tree_root->node))
1866 next_backup = last_backup;
1867
1868 root_backup = info->super_for_commit->super_roots + next_backup;
1869
1870 /*
1871 * make sure all of our padding and empty slots get zero filled
1872 * regardless of which ones we use today
1873 */
1874 memset(root_backup, 0, sizeof(*root_backup));
1875
1876 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1877
1878 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1879 btrfs_set_backup_tree_root_gen(root_backup,
1880 btrfs_header_generation(info->tree_root->node));
1881
1882 btrfs_set_backup_tree_root_level(root_backup,
1883 btrfs_header_level(info->tree_root->node));
1884
1885 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1886 btrfs_set_backup_chunk_root_gen(root_backup,
1887 btrfs_header_generation(info->chunk_root->node));
1888 btrfs_set_backup_chunk_root_level(root_backup,
1889 btrfs_header_level(info->chunk_root->node));
1890
1891 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1892 btrfs_set_backup_extent_root_gen(root_backup,
1893 btrfs_header_generation(info->extent_root->node));
1894 btrfs_set_backup_extent_root_level(root_backup,
1895 btrfs_header_level(info->extent_root->node));
1896
7c7e82a7
CM		 1897 /*
1898 * we might commit during log recovery, which happens before we set
1899 * the fs_root. Make sure it is valid before we fill it in.
1900 */
1901 if (info->fs_root && info->fs_root->node) {
1902 btrfs_set_backup_fs_root(root_backup,
1903 info->fs_root->node->start);
1904 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 1905 btrfs_header_generation(info->fs_root->node));
7c7e82a7 1906 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 1907 btrfs_header_level(info->fs_root->node));
7c7e82a7 1908 }
af31f5e5
CM		 1909
1910 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1911 btrfs_set_backup_dev_root_gen(root_backup,
1912 btrfs_header_generation(info->dev_root->node));
1913 btrfs_set_backup_dev_root_level(root_backup,
1914 btrfs_header_level(info->dev_root->node));
1915
1916 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1917 btrfs_set_backup_csum_root_gen(root_backup,
1918 btrfs_header_generation(info->csum_root->node));
1919 btrfs_set_backup_csum_root_level(root_backup,
1920 btrfs_header_level(info->csum_root->node));
1921
1922 btrfs_set_backup_total_bytes(root_backup,
1923 btrfs_super_total_bytes(info->super_copy));
1924 btrfs_set_backup_bytes_used(root_backup,
1925 btrfs_super_bytes_used(info->super_copy));
1926 btrfs_set_backup_num_devices(root_backup,
1927 btrfs_super_num_devices(info->super_copy));
1928
1929 /*
1930 * if we don't copy this out to the super_copy, it won't get remembered
1931 * for the next commit
1932 */
1933 memcpy(&info->super_copy->super_roots,
1934 &info->super_for_commit->super_roots,
1935 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1936}
1937
1938/*
1939 * this copies info out of the root backup array and back into
1940 * the in-memory super block. It is meant to help iterate through
1941 * the array, so you send it the number of backups you've already
1942 * tried and the last backup index you used.
1943 *
1944 * this returns -1 when it has tried all the backups
1945 */
1946static noinline int next_root_backup(struct btrfs_fs_info *info,
1947 struct btrfs_super_block *super,
1948 int *num_backups_tried, int *backup_index)
1949{
1950 struct btrfs_root_backup *root_backup;
1951 int newest = *backup_index;
1952
1953 if (*num_backups_tried == 0) {
1954 u64 gen = btrfs_super_generation(super);
1955
1956 newest = find_newest_super_backup(info, gen);
1957 if (newest == -1)
1958 return -1;
1959
1960 *backup_index = newest;
1961 *num_backups_tried = 1;
1962 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
1963 /* we've tried all the backups, all done */
1964 return -1;
1965 } else {
1966 /* jump to the next oldest backup */
1967 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
1968 BTRFS_NUM_BACKUP_ROOTS;
1969 *backup_index = newest;
1970 *num_backups_tried += 1;
1971 }
1972 root_backup = super->super_roots + newest;
1973
1974 btrfs_set_super_generation(super,
1975 btrfs_backup_tree_root_gen(root_backup));
1976 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
1977 btrfs_set_super_root_level(super,
1978 btrfs_backup_tree_root_level(root_backup));
1979 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
1980
1981 /*
1982 * fixme: the total bytes and num_devices need to match or we should
1983 * need a fsck
1984 */
1985 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
1986 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
1987 return 0;
1988}
1989
7abadb64
LB		 1990/* helper to cleanup workers */
1991static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
1992{
1993 btrfs_stop_workers(&fs_info->generic_worker);
1994 btrfs_stop_workers(&fs_info->fixup_workers);
1995 btrfs_stop_workers(&fs_info->delalloc_workers);
1996 btrfs_stop_workers(&fs_info->workers);
1997 btrfs_stop_workers(&fs_info->endio_workers);
1998 btrfs_stop_workers(&fs_info->endio_meta_workers);
1999 btrfs_stop_workers(&fs_info->endio_raid56_workers);
2000 btrfs_stop_workers(&fs_info->rmw_workers);
2001 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
2002 btrfs_stop_workers(&fs_info->endio_write_workers);
2003 btrfs_stop_workers(&fs_info->endio_freespace_worker);
2004 btrfs_stop_workers(&fs_info->submit_workers);
2005 btrfs_stop_workers(&fs_info->delayed_workers);
2006 btrfs_stop_workers(&fs_info->caching_workers);
2007 btrfs_stop_workers(&fs_info->readahead_workers);
2008 btrfs_stop_workers(&fs_info->flush_workers);
2f232036 2009 btrfs_stop_workers(&fs_info->qgroup_rescan_workers);
7abadb64
LB		 2010}
2011
2e9f5954
R		 2012static void free_root_extent_buffers(struct btrfs_root *root)
2013{
2014 if (root) {
2015 free_extent_buffer(root->node);
2016 free_extent_buffer(root->commit_root);
2017 root->node = NULL;
2018 root->commit_root = NULL;
2019 }
2020}
2021
af31f5e5
CM		 2022/* helper to cleanup tree roots */
2023static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2024{
2e9f5954 2025 free_root_extent_buffers(info->tree_root);
655b09fe 2026
2e9f5954
R		 2027 free_root_extent_buffers(info->dev_root);
2028 free_root_extent_buffers(info->extent_root);
2029 free_root_extent_buffers(info->csum_root);
2030 free_root_extent_buffers(info->quota_root);
2031 free_root_extent_buffers(info->uuid_root);
2032 if (chunk_root)
2033 free_root_extent_buffers(info->chunk_root);
af31f5e5
CM		 2034}
2035
171f6537
JB		 2036static void del_fs_roots(struct btrfs_fs_info *fs_info)
2037{
2038 int ret;
2039 struct btrfs_root *gang[8];
2040 int i;
2041
2042 while (!list_empty(&fs_info->dead_roots)) {
2043 gang[0] = list_entry(fs_info->dead_roots.next,
2044 struct btrfs_root, root_list);
2045 list_del(&gang[0]->root_list);
2046
2047 if (gang[0]->in_radix) {
cb517eab 2048 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
171f6537
JB		 2049 } else {
2050 free_extent_buffer(gang[0]->node);
2051 free_extent_buffer(gang[0]->commit_root);
b0feb9d9 2052 btrfs_put_fs_root(gang[0]);
171f6537
JB		 2053 }
2054 }
2055
2056 while (1) {
2057 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2058 (void **)gang, 0,
2059 ARRAY_SIZE(gang));
2060 if (!ret)
2061 break;
2062 for (i = 0; i < ret; i++)
cb517eab 2063 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
171f6537 2064 }
1a4319cc
LB		 2065
2066 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2067 btrfs_free_log_root_tree(NULL, fs_info);
2068 btrfs_destroy_pinned_extent(fs_info->tree_root,
2069 fs_info->pinned_extents);
2070 }
171f6537 2071}
af31f5e5 2072
ad2b2c80
AV		 2073int open_ctree(struct super_block *sb,
2074 struct btrfs_fs_devices *fs_devices,
2075 char *options)
2e635a27 2076{
db94535d
CM		 2077 u32 sectorsize;
2078 u32 nodesize;
2079 u32 leafsize;
2080 u32 blocksize;
87ee04eb 2081 u32 stripesize;
84234f3a 2082 u64 generation;
f2b636e8 2083 u64 features;
3de4586c 2084 struct btrfs_key location;
a061fc8d 2085 struct buffer_head *bh;
4d34b278 2086 struct btrfs_super_block *disk_super;
815745cf 2087 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2088 struct btrfs_root *tree_root;
4d34b278
ID		 2089 struct btrfs_root *extent_root;
2090 struct btrfs_root *csum_root;
2091 struct btrfs_root *chunk_root;
2092 struct btrfs_root *dev_root;
bcef60f2 2093 struct btrfs_root *quota_root;
f7a81ea4 2094 struct btrfs_root *uuid_root;
e02119d5 2095 struct btrfs_root *log_tree_root;
eb60ceac 2096 int ret;
e58ca020 2097 int err = -EINVAL;
af31f5e5
CM		 2098 int num_backups_tried = 0;
2099 int backup_index = 0;
70f80175
SB		 2100 bool create_uuid_tree;
2101 bool check_uuid_tree;
4543df7e 2102
f84a8bd6 2103 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
6f07e42e 2104 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
cb517eab 2105 if (!tree_root || !chunk_root) {
39279cc3
CM		 2106 err = -ENOMEM;
2107 goto fail;
2108 }
76dda93c
YZ		 2109
2110 ret = init_srcu_struct(&fs_info->subvol_srcu);
2111 if (ret) {
2112 err = ret;
2113 goto fail;
2114 }
2115
2116 ret = setup_bdi(fs_info, &fs_info->bdi);
2117 if (ret) {
2118 err = ret;
2119 goto fail_srcu;
2120 }
2121
e2d84521
MX		 2122 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2123 if (ret) {
2124 err = ret;
2125 goto fail_bdi;
2126 }
2127 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2128 (1 + ilog2(nr_cpu_ids));
2129
963d678b
MX		 2130 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2131 if (ret) {
2132 err = ret;
2133 goto fail_dirty_metadata_bytes;
2134 }
2135
76dda93c
YZ		 2136 fs_info->btree_inode = new_inode(sb);
2137 if (!fs_info->btree_inode) {
2138 err = -ENOMEM;
963d678b 2139 goto fail_delalloc_bytes;
76dda93c
YZ		 2140 }
2141
a6591715 2142 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2143
76dda93c 2144 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
f28491e0 2145 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
8fd17795 2146 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2147 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2148 INIT_LIST_HEAD(&fs_info->delayed_iputs);
eb73c1b7 2149 INIT_LIST_HEAD(&fs_info->delalloc_roots);
11833d66 2150 INIT_LIST_HEAD(&fs_info->caching_block_groups);
eb73c1b7 2151 spin_lock_init(&fs_info->delalloc_root_lock);
a4abeea4 2152 spin_lock_init(&fs_info->trans_lock);
76dda93c 2153 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2154 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2155 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 2156 spin_lock_init(&fs_info->free_chunk_lock);
f29021b2 2157 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2158 spin_lock_init(&fs_info->super_lock);
f28491e0 2159 spin_lock_init(&fs_info->buffer_lock);
f29021b2 2160 rwlock_init(&fs_info->tree_mod_log_lock);
7585717f 2161 mutex_init(&fs_info->reloc_mutex);
de98ced9 2162 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2163
58176a96 2164 init_completion(&fs_info->kobj_unregister);
0b86a832 2165 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2166 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2167 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
0b86a832 2168 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX		 2169 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2170 BTRFS_BLOCK_RSV_GLOBAL);
2171 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2172 BTRFS_BLOCK_RSV_DELALLOC);
2173 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2174 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2175 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2176 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2177 BTRFS_BLOCK_RSV_DELOPS);
cb03c743 2178 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 2179 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 2180 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 2181 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 2182 atomic_set(&fs_info->defrag_running, 0);
fc36ed7e 2183 atomic64_set(&fs_info->tree_mod_seq, 0);
e20d96d6 2184 fs_info->sb = sb;
6f568d35 2185 fs_info->max_inline = 8192 * 1024;
9ed74f2d 2186 fs_info->metadata_ratio = 0;
4cb5300b 2187 fs_info->defrag_inodes = RB_ROOT;
2bf64758 2188 fs_info->free_chunk_space = 0;
f29021b2 2189 fs_info->tree_mod_log = RB_ROOT;
8b87dc17 2190 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
0a2b2a84 2191 fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
90519d66
AJ		 2192 /* readahead state */
2193 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2194 spin_lock_init(&fs_info->reada_lock);
c8b97818 2195
b34b086c
CM		 2196 fs_info->thread_pool_size = min_t(unsigned long,
2197 num_online_cpus() + 2, 8);
0afbaf8c 2198
199c2a9c
MX		 2199 INIT_LIST_HEAD(&fs_info->ordered_roots);
2200 spin_lock_init(&fs_info->ordered_root_lock);
16cdcec7
MX		 2201 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2202 GFP_NOFS);
2203 if (!fs_info->delayed_root) {
2204 err = -ENOMEM;
2205 goto fail_iput;
2206 }
2207 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2208
a2de733c
AJ		 2209 mutex_init(&fs_info->scrub_lock);
2210 atomic_set(&fs_info->scrubs_running, 0);
2211 atomic_set(&fs_info->scrub_pause_req, 0);
2212 atomic_set(&fs_info->scrubs_paused, 0);
2213 atomic_set(&fs_info->scrub_cancel_req, 0);
2214 init_waitqueue_head(&fs_info->scrub_pause_wait);
a2de733c 2215 fs_info->scrub_workers_refcnt = 0;
21adbd5c
SB		 2216#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2217 fs_info->check_integrity_print_mask = 0;
2218#endif
a2de733c 2219
c9e9f97b
ID		 2220 spin_lock_init(&fs_info->balance_lock);
2221 mutex_init(&fs_info->balance_mutex);
837d5b6e
ID		 2222 atomic_set(&fs_info->balance_running, 0);
2223 atomic_set(&fs_info->balance_pause_req, 0);
a7e99c69 2224 atomic_set(&fs_info->balance_cancel_req, 0);
c9e9f97b 2225 fs_info->balance_ctl = NULL;
837d5b6e 2226 init_waitqueue_head(&fs_info->balance_wait_q);
a2de733c 2227
a061fc8d
CM		 2228 sb->s_blocksize = 4096;
2229 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 2230 sb->s_bdi = &fs_info->bdi;
a061fc8d 2231
76dda93c 2232 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
bfe86848 2233 set_nlink(fs_info->btree_inode, 1);
0afbaf8c
CM		 2234 /*
2235 * we set the i_size on the btree inode to the max possible int.
2236 * the real end of the address space is determined by all of
2237 * the devices in the system
2238 */
2239 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 2240 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 2241 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2242
5d4f98a2 2243 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 2244 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 2245 fs_info->btree_inode->i_mapping);
0b32f4bb 2246 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
a8067e02 2247 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM		 2248
2249 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 2250
76dda93c
YZ		 2251 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2252 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2253 sizeof(struct btrfs_key));
72ac3c0d
JB		 2254 set_bit(BTRFS_INODE_DUMMY,
2255 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
778ba82b 2256 btrfs_insert_inode_hash(fs_info->btree_inode);
76dda93c 2257
0f9dd46c 2258 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2259 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2260 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2261
11833d66 2262 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 2263 fs_info->btree_inode->i_mapping);
11833d66 2264 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 2265 fs_info->btree_inode->i_mapping);
11833d66 2266 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 2267 fs_info->do_barriers = 1;
e18e4809 2268
39279cc3 2269
5a3f23d5 2270 mutex_init(&fs_info->ordered_operations_mutex);
9ffba8cd 2271 mutex_init(&fs_info->ordered_extent_flush_mutex);
e02119d5 2272 mutex_init(&fs_info->tree_log_mutex);
925baedd 2273 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 2274 mutex_init(&fs_info->transaction_kthread_mutex);
2275 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2276 mutex_init(&fs_info->volume_mutex);
276e680d 2277 init_rwsem(&fs_info->extent_commit_sem);
c71bf099 2278 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2279 init_rwsem(&fs_info->subvol_sem);
803b2f54 2280 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
e922e087
SB		 2281 fs_info->dev_replace.lock_owner = 0;
2282 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2283 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2284 mutex_init(&fs_info->dev_replace.lock_management_lock);
2285 mutex_init(&fs_info->dev_replace.lock);
fa9c0d79 2286
416ac51d 2287 spin_lock_init(&fs_info->qgroup_lock);
f2f6ed3d 2288 mutex_init(&fs_info->qgroup_ioctl_lock);
416ac51d
AJ		 2289 fs_info->qgroup_tree = RB_ROOT;
2290 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2291 fs_info->qgroup_seq = 1;
2292 fs_info->quota_enabled = 0;
2293 fs_info->pending_quota_state = 0;
1e8f9158 2294 fs_info->qgroup_ulist = NULL;
2f232036 2295 mutex_init(&fs_info->qgroup_rescan_lock);
416ac51d 2296
fa9c0d79
CM		 2297 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2298 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2299
e6dcd2dc 2300 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2301 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2302 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2303 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2304
53b381b3
DW		 2305 ret = btrfs_alloc_stripe_hash_table(fs_info);
2306 if (ret) {
83c8266a 2307 err = ret;
53b381b3
DW		 2308 goto fail_alloc;
2309 }
2310
0b86a832 2311 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 2312 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2313
3c4bb26b 2314 invalidate_bdev(fs_devices->latest_bdev);
1104a885
DS		 2315
2316 /*
2317 * Read super block and check the signature bytes only
2318 */
a512bbf8 2319 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY		 2320 if (!bh) {
2321 err = -EINVAL;
16cdcec7 2322 goto fail_alloc;
20b45077 2323 }
39279cc3 2324
1104a885
DS		 2325 /*
2326 * We want to check superblock checksum, the type is stored inside.
2327 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2328 */
2329 if (btrfs_check_super_csum(bh->b_data)) {
efe120a0 2330 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
1104a885
DS		 2331 err = -EINVAL;
2332 goto fail_alloc;
2333 }
2334
2335 /*
2336 * super_copy is zeroed at allocation time and we never touch the
2337 * following bytes up to INFO_SIZE, the checksum is calculated from
2338 * the whole block of INFO_SIZE
2339 */
6c41761f
DS		 2340 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2341 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2342 sizeof(*fs_info->super_for_commit));
a061fc8d 2343 brelse(bh);
5f39d397 2344
6c41761f 2345 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2346
1104a885
DS		 2347 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2348 if (ret) {
efe120a0 2349 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
1104a885
DS		 2350 err = -EINVAL;
2351 goto fail_alloc;
2352 }
2353
6c41761f 2354 disk_super = fs_info->super_copy;
0f7d52f4 2355 if (!btrfs_super_root(disk_super))
16cdcec7 2356 goto fail_alloc;
0f7d52f4 2357
acce952b 2358 /* check FS state, whether FS is broken. */
87533c47
MX		 2359 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2360 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2361
af31f5e5
CM		 2362 /*
2363 * run through our array of backup supers and setup
2364 * our ring pointer to the oldest one
2365 */
2366 generation = btrfs_super_generation(disk_super);
2367 find_oldest_super_backup(fs_info, generation);
2368
75e7cb7f
LB		 2369 /*
2370 * In the long term, we'll store the compression type in the super
2371 * block, and it'll be used for per file compression control.
2372 */
2373 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2374
2b82032c
YZ		 2375 ret = btrfs_parse_options(tree_root, options);
2376 if (ret) {
2377 err = ret;
16cdcec7 2378 goto fail_alloc;
2b82032c 2379 }
dfe25020 2380
f2b636e8
JB		 2381 features = btrfs_super_incompat_flags(disk_super) &
2382 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2383 if (features) {
2384 printk(KERN_ERR "BTRFS: couldn't mount because of "
2385 "unsupported optional features (%Lx).\n",
c1c9ff7c 2386 features);
f2b636e8 2387 err = -EINVAL;
16cdcec7 2388 goto fail_alloc;
f2b636e8
JB		 2389 }
2390
727011e0
CM		 2391 if (btrfs_super_leafsize(disk_super) !=
2392 btrfs_super_nodesize(disk_super)) {
2393 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2394 "blocksizes don't match. node %d leaf %d\n",
2395 btrfs_super_nodesize(disk_super),
2396 btrfs_super_leafsize(disk_super));
2397 err = -EINVAL;
2398 goto fail_alloc;
2399 }
2400 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2401 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2402 "blocksize (%d) was too large\n",
2403 btrfs_super_leafsize(disk_super));
2404 err = -EINVAL;
2405 goto fail_alloc;
2406 }
2407
5d4f98a2 2408 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2409 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
69e380d1 2410 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2411 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
727011e0 2412
3173a18f 2413 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
efe120a0 2414 printk(KERN_ERR "BTRFS: has skinny extents\n");
3173a18f 2415
727011e0
CM		 2416 /*
2417 * flag our filesystem as having big metadata blocks if
2418 * they are bigger than the page size
2419 */
2420 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2421 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
efe120a0 2422 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
727011e0
CM		 2423 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2424 }
2425
bc3f116f
CM		 2426 nodesize = btrfs_super_nodesize(disk_super);
2427 leafsize = btrfs_super_leafsize(disk_super);
2428 sectorsize = btrfs_super_sectorsize(disk_super);
2429 stripesize = btrfs_super_stripesize(disk_super);
e2d84521 2430 fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids));
963d678b 2431 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f
CM		 2432
2433 /*
2434 * mixed block groups end up with duplicate but slightly offset
2435 * extent buffers for the same range. It leads to corruptions
2436 */
2437 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2438 (sectorsize != leafsize)) {
efe120a0 2439 printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes "
bc3f116f
CM		 2440 "are not allowed for mixed block groups on %s\n",
2441 sb->s_id);
2442 goto fail_alloc;
2443 }
2444
ceda0864
MX		 2445 /*
2446 * Needn't use the lock because there is no other task which will
2447 * update the flag.
2448 */
a6fa6fae 2449 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2450
f2b636e8
JB		 2451 features = btrfs_super_compat_ro_flags(disk_super) &
2452 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2453 if (!(sb->s_flags & MS_RDONLY) && features) {
2454 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2455 "unsupported option features (%Lx).\n",
c1c9ff7c 2456 features);
f2b636e8 2457 err = -EINVAL;
16cdcec7 2458 goto fail_alloc;
f2b636e8 2459 }
61d92c32
CM		 2460
2461 btrfs_init_workers(&fs_info->generic_worker,
2462 "genwork", 1, NULL);
2463
5443be45 2464 btrfs_init_workers(&fs_info->workers, "worker",
61d92c32
CM		 2465 fs_info->thread_pool_size,
2466 &fs_info->generic_worker);
c8b97818 2467
771ed689 2468 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
45d5fd14 2469 fs_info->thread_pool_size, NULL);
771ed689 2470
8ccf6f19 2471 btrfs_init_workers(&fs_info->flush_workers, "flush_delalloc",
45d5fd14 2472 fs_info->thread_pool_size, NULL);
8ccf6f19 2473
5443be45 2474 btrfs_init_workers(&fs_info->submit_workers, "submit",
b720d209 2475 min_t(u64, fs_devices->num_devices,
45d5fd14 2476 fs_info->thread_pool_size), NULL);
61b49440 2477
bab39bf9 2478 btrfs_init_workers(&fs_info->caching_workers, "cache",
45d5fd14 2479 fs_info->thread_pool_size, NULL);
bab39bf9 2480
61b49440
CM		 2481 /* a higher idle thresh on the submit workers makes it much more
2482 * likely that bios will be send down in a sane order to the
2483 * devices
2484 */
2485 fs_info->submit_workers.idle_thresh = 64;
53863232 2486
771ed689 2487 fs_info->workers.idle_thresh = 16;
4a69a410 2488 fs_info->workers.ordered = 1;
61b49440 2489
771ed689
CM		 2490 fs_info->delalloc_workers.idle_thresh = 2;
2491 fs_info->delalloc_workers.ordered = 1;
2492
61d92c32
CM		 2493 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
2494 &fs_info->generic_worker);
5443be45 2495 btrfs_init_workers(&fs_info->endio_workers, "endio",
61d92c32
CM		 2496 fs_info->thread_pool_size,
2497 &fs_info->generic_worker);
d20f7043 2498 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
61d92c32
CM		 2499 fs_info->thread_pool_size,
2500 &fs_info->generic_worker);
cad321ad 2501 btrfs_init_workers(&fs_info->endio_meta_write_workers,
61d92c32
CM		 2502 "endio-meta-write", fs_info->thread_pool_size,
2503 &fs_info->generic_worker);
53b381b3
DW		 2504 btrfs_init_workers(&fs_info->endio_raid56_workers,
2505 "endio-raid56", fs_info->thread_pool_size,
2506 &fs_info->generic_worker);
2507 btrfs_init_workers(&fs_info->rmw_workers,
2508 "rmw", fs_info->thread_pool_size,
2509 &fs_info->generic_worker);
5443be45 2510 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
61d92c32
CM		 2511 fs_info->thread_pool_size,
2512 &fs_info->generic_worker);
0cb59c99
JB		 2513 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
2514 1, &fs_info->generic_worker);
16cdcec7
MX		 2515 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
2516 fs_info->thread_pool_size,
2517 &fs_info->generic_worker);
90519d66
AJ		 2518 btrfs_init_workers(&fs_info->readahead_workers, "readahead",
2519 fs_info->thread_pool_size,
2520 &fs_info->generic_worker);
2f232036
JS		 2521 btrfs_init_workers(&fs_info->qgroup_rescan_workers, "qgroup-rescan", 1,
2522 &fs_info->generic_worker);
61b49440
CM		 2523
2524 /*
2525 * endios are largely parallel and should have a very
2526 * low idle thresh
2527 */
2528 fs_info->endio_workers.idle_thresh = 4;
b51912c9 2529 fs_info->endio_meta_workers.idle_thresh = 4;
53b381b3
DW		 2530 fs_info->endio_raid56_workers.idle_thresh = 4;
2531 fs_info->rmw_workers.idle_thresh = 2;
b51912c9 2532
9042846b
CM		 2533 fs_info->endio_write_workers.idle_thresh = 2;
2534 fs_info->endio_meta_write_workers.idle_thresh = 2;
90519d66 2535 fs_info->readahead_workers.idle_thresh = 2;
9042846b 2536
0dc3b84a
JB		 2537 /*
2538 * btrfs_start_workers can really only fail because of ENOMEM so just
2539 * return -ENOMEM if any of these fail.
2540 */
2541 ret = btrfs_start_workers(&fs_info->workers);
2542 ret |= btrfs_start_workers(&fs_info->generic_worker);
2543 ret |= btrfs_start_workers(&fs_info->submit_workers);
2544 ret |= btrfs_start_workers(&fs_info->delalloc_workers);
2545 ret |= btrfs_start_workers(&fs_info->fixup_workers);
2546 ret |= btrfs_start_workers(&fs_info->endio_workers);
2547 ret |= btrfs_start_workers(&fs_info->endio_meta_workers);
53b381b3
DW		 2548 ret |= btrfs_start_workers(&fs_info->rmw_workers);
2549 ret |= btrfs_start_workers(&fs_info->endio_raid56_workers);
0dc3b84a
JB		 2550 ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers);
2551 ret |= btrfs_start_workers(&fs_info->endio_write_workers);
2552 ret |= btrfs_start_workers(&fs_info->endio_freespace_worker);
2553 ret |= btrfs_start_workers(&fs_info->delayed_workers);
2554 ret |= btrfs_start_workers(&fs_info->caching_workers);
2555 ret |= btrfs_start_workers(&fs_info->readahead_workers);
8ccf6f19 2556 ret |= btrfs_start_workers(&fs_info->flush_workers);
2f232036 2557 ret |= btrfs_start_workers(&fs_info->qgroup_rescan_workers);
0dc3b84a 2558 if (ret) {
fed425c7 2559 err = -ENOMEM;
0dc3b84a
JB		 2560 goto fail_sb_buffer;
2561 }
4543df7e 2562
4575c9cc 2563 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 2564 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2565 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 2566
db94535d
CM		 2567 tree_root->nodesize = nodesize;
2568 tree_root->leafsize = leafsize;
2569 tree_root->sectorsize = sectorsize;
87ee04eb 2570 tree_root->stripesize = stripesize;
a061fc8d
CM		 2571
2572 sb->s_blocksize = sectorsize;
2573 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 2574
3cae210f 2575 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
efe120a0 2576 printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 2577 goto fail_sb_buffer;
2578 }
19c00ddc 2579
8d082fb7 2580 if (sectorsize != PAGE_SIZE) {
efe120a0 2581 printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) "
8d082fb7 2582 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
941b2ddf
KM		 2583 goto fail_sb_buffer;
2584 }
2585
925baedd 2586 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 2587 ret = btrfs_read_sys_array(tree_root);
925baedd 2588 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2589 if (ret) {
efe120a0 2590 printk(KERN_WARNING "BTRFS: failed to read the system "
d397712b 2591 "array on %s\n", sb->s_id);
5d4f98a2 2592 goto fail_sb_buffer;
84eed90f 2593 }
0b86a832
CM		 2594
2595 blocksize = btrfs_level_size(tree_root,
2596 btrfs_super_chunk_root_level(disk_super));
84234f3a 2597 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 2598
2599 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2600 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2601
2602 chunk_root->node = read_tree_block(chunk_root,
2603 btrfs_super_chunk_root(disk_super),
84234f3a 2604 blocksize, generation);
416bc658
JB		 2605 if (!chunk_root->node ||
2606 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
efe120a0 2607 printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n",
83121942 2608 sb->s_id);
af31f5e5 2609 goto fail_tree_roots;
83121942 2610 }
5d4f98a2
YZ		 2611 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2612 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2613
e17cade2 2614 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
b308bc2f 2615 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
e17cade2 2616
0b86a832 2617 ret = btrfs_read_chunk_tree(chunk_root);
2b82032c 2618 if (ret) {
efe120a0 2619 printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n",
d397712b 2620 sb->s_id);
af31f5e5 2621 goto fail_tree_roots;
2b82032c 2622 }
0b86a832 2623
8dabb742
SB		 2624 /*
2625 * keep the device that is marked to be the target device for the
2626 * dev_replace procedure
2627 */
2628 btrfs_close_extra_devices(fs_info, fs_devices, 0);
dfe25020 2629
a6b0d5c8 2630 if (!fs_devices->latest_bdev) {
efe120a0 2631 printk(KERN_CRIT "BTRFS: failed to read devices on %s\n",
a6b0d5c8
CM		 2632 sb->s_id);
2633 goto fail_tree_roots;
2634 }
2635
af31f5e5 2636retry_root_backup:
db94535d
CM		 2637 blocksize = btrfs_level_size(tree_root,
2638 btrfs_super_root_level(disk_super));
84234f3a 2639 generation = btrfs_super_generation(disk_super);
0b86a832 2640
e20d96d6 2641 tree_root->node = read_tree_block(tree_root,
db94535d 2642 btrfs_super_root(disk_super),
84234f3a 2643 blocksize, generation);
af31f5e5
CM		 2644 if (!tree_root->node ||
2645 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
efe120a0 2646 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
83121942 2647 sb->s_id);
af31f5e5
CM		 2648
2649 goto recovery_tree_root;
83121942 2650 }
af31f5e5 2651
5d4f98a2
YZ		 2652 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2653 tree_root->commit_root = btrfs_root_node(tree_root);
69e9c6c6 2654 btrfs_set_root_refs(&tree_root->root_item, 1);
db94535d 2655
cb517eab
MX		 2656 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2657 location.type = BTRFS_ROOT_ITEM_KEY;
2658 location.offset = 0;
2659
2660 extent_root = btrfs_read_tree_root(tree_root, &location);
2661 if (IS_ERR(extent_root)) {
2662 ret = PTR_ERR(extent_root);
af31f5e5 2663 goto recovery_tree_root;
cb517eab 2664 }
0b86a832 2665 extent_root->track_dirty = 1;
cb517eab 2666 fs_info->extent_root = extent_root;
0b86a832 2667
cb517eab
MX		 2668 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2669 dev_root = btrfs_read_tree_root(tree_root, &location);
2670 if (IS_ERR(dev_root)) {
2671 ret = PTR_ERR(dev_root);
af31f5e5 2672 goto recovery_tree_root;
cb517eab 2673 }
5d4f98a2 2674 dev_root->track_dirty = 1;
cb517eab
MX		 2675 fs_info->dev_root = dev_root;
2676 btrfs_init_devices_late(fs_info);
3768f368 2677
cb517eab
MX		 2678 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2679 csum_root = btrfs_read_tree_root(tree_root, &location);
2680 if (IS_ERR(csum_root)) {
2681 ret = PTR_ERR(csum_root);
af31f5e5 2682 goto recovery_tree_root;
cb517eab 2683 }
d20f7043 2684 csum_root->track_dirty = 1;
cb517eab 2685 fs_info->csum_root = csum_root;
d20f7043 2686
cb517eab
MX		 2687 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2688 quota_root = btrfs_read_tree_root(tree_root, &location);
2689 if (!IS_ERR(quota_root)) {
bcef60f2
AJ		 2690 quota_root->track_dirty = 1;
2691 fs_info->quota_enabled = 1;
2692 fs_info->pending_quota_state = 1;
cb517eab 2693 fs_info->quota_root = quota_root;
bcef60f2
AJ		 2694 }
2695
f7a81ea4
SB		 2696 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2697 uuid_root = btrfs_read_tree_root(tree_root, &location);
2698 if (IS_ERR(uuid_root)) {
2699 ret = PTR_ERR(uuid_root);
2700 if (ret != -ENOENT)
2701 goto recovery_tree_root;
2702 create_uuid_tree = true;
70f80175 2703 check_uuid_tree = false;
f7a81ea4
SB		 2704 } else {
2705 uuid_root->track_dirty = 1;
2706 fs_info->uuid_root = uuid_root;
70f80175
SB		 2707 create_uuid_tree = false;
2708 check_uuid_tree =
2709 generation != btrfs_super_uuid_tree_generation(disk_super);
f7a81ea4
SB		 2710 }
2711
8929ecfa
YZ		 2712 fs_info->generation = generation;
2713 fs_info->last_trans_committed = generation;
8929ecfa 2714
68310a5e
ID		 2715 ret = btrfs_recover_balance(fs_info);
2716 if (ret) {
efe120a0 2717 printk(KERN_WARNING "BTRFS: failed to recover balance\n");
68310a5e
ID		 2718 goto fail_block_groups;
2719 }
2720
733f4fbb
SB		 2721 ret = btrfs_init_dev_stats(fs_info);
2722 if (ret) {
efe120a0 2723 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
733f4fbb
SB		 2724 ret);
2725 goto fail_block_groups;
2726 }
2727
8dabb742
SB		 2728 ret = btrfs_init_dev_replace(fs_info);
2729 if (ret) {
efe120a0 2730 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
8dabb742
SB		 2731 goto fail_block_groups;
2732 }
2733
2734 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2735
5ac1d209 2736 ret = btrfs_sysfs_add_one(fs_info);
c59021f8 2737 if (ret) {
efe120a0 2738 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
c59021f8 2739 goto fail_block_groups;
2740 }
2741
c59021f8 2742 ret = btrfs_init_space_info(fs_info);
2743 if (ret) {
efe120a0 2744 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
2365dd3c 2745 goto fail_sysfs;
c59021f8 2746 }
2747
1b1d1f66
JB		 2748 ret = btrfs_read_block_groups(extent_root);
2749 if (ret) {
efe120a0 2750 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
2365dd3c 2751 goto fail_sysfs;
1b1d1f66 2752 }
5af3e8cc
SB		 2753 fs_info->num_tolerated_disk_barrier_failures =
2754 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
292fd7fc
SB		 2755 if (fs_info->fs_devices->missing_devices >
2756 fs_info->num_tolerated_disk_barrier_failures &&
2757 !(sb->s_flags & MS_RDONLY)) {
efe120a0
FH		 2758 printk(KERN_WARNING "BTRFS: "
2759 "too many missing devices, writeable mount is not allowed\n");
2365dd3c 2760 goto fail_sysfs;
292fd7fc 2761 }
9078a3e1 2762
a74a4b97
CM		 2763 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2764 "btrfs-cleaner");
57506d50 2765 if (IS_ERR(fs_info->cleaner_kthread))
2365dd3c 2766 goto fail_sysfs;
a74a4b97
CM		 2767
2768 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2769 tree_root,
2770 "btrfs-transaction");
57506d50 2771 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2772 goto fail_cleaner;
a74a4b97 2773
c289811c
CM		 2774 if (!btrfs_test_opt(tree_root, SSD) &&
2775 !btrfs_test_opt(tree_root, NOSSD) &&
2776 !fs_info->fs_devices->rotating) {
efe120a0 2777 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
c289811c
CM		 2778 "mode\n");
2779 btrfs_set_opt(fs_info->mount_opt, SSD);
2780 }
2781
3818aea2
QW		 2782 /* Set the real inode map cache flag */
2783 if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE))
2784 btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE);
2785
21adbd5c
SB		 2786#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2787 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2788 ret = btrfsic_mount(tree_root, fs_devices,
2789 btrfs_test_opt(tree_root,
2790 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2791 1 : 0,
2792 fs_info->check_integrity_print_mask);
2793 if (ret)
efe120a0 2794 printk(KERN_WARNING "BTRFS: failed to initialize"
21adbd5c
SB		 2795 " integrity check module %s\n", sb->s_id);
2796 }
2797#endif
bcef60f2
AJ		 2798 ret = btrfs_read_qgroup_config(fs_info);
2799 if (ret)
2800 goto fail_trans_kthread;
21adbd5c 2801
acce952b 2802 /* do not make disk changes in broken FS */
68ce9682 2803 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM		 2804 u64 bytenr = btrfs_super_log_root(disk_super);
2805
7c2ca468 2806 if (fs_devices->rw_devices == 0) {
efe120a0 2807 printk(KERN_WARNING "BTRFS: log replay required "
d397712b 2808 "on RO media\n");
7c2ca468 2809 err = -EIO;
bcef60f2 2810 goto fail_qgroup;
7c2ca468 2811 }
e02119d5
CM		 2812 blocksize =
2813 btrfs_level_size(tree_root,
2814 btrfs_super_log_root_level(disk_super));
d18a2c44 2815
6f07e42e 2816 log_tree_root = btrfs_alloc_root(fs_info);
676e4c86
DC		 2817 if (!log_tree_root) {
2818 err = -ENOMEM;
bcef60f2 2819 goto fail_qgroup;
676e4c86 2820 }
e02119d5
CM		 2821
2822 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2823 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2824
2825 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 2826 blocksize,
2827 generation + 1);
416bc658
JB		 2828 if (!log_tree_root->node ||
2829 !extent_buffer_uptodate(log_tree_root->node)) {
efe120a0 2830 printk(KERN_ERR "BTRFS: failed to read log tree\n");
416bc658
JB		 2831 free_extent_buffer(log_tree_root->node);
2832 kfree(log_tree_root);
2833 goto fail_trans_kthread;
2834 }
79787eaa 2835 /* returns with log_tree_root freed on success */
e02119d5 2836 ret = btrfs_recover_log_trees(log_tree_root);
79787eaa
JM		 2837 if (ret) {
2838 btrfs_error(tree_root->fs_info, ret,
2839 "Failed to recover log tree");
2840 free_extent_buffer(log_tree_root->node);
2841 kfree(log_tree_root);
2842 goto fail_trans_kthread;
2843 }
e556ce2c
YZ		 2844
2845 if (sb->s_flags & MS_RDONLY) {
79787eaa
JM		 2846 ret = btrfs_commit_super(tree_root);
2847 if (ret)
2848 goto fail_trans_kthread;
e556ce2c 2849 }
e02119d5 2850 }
1a40e23b 2851
76dda93c 2852 ret = btrfs_find_orphan_roots(tree_root);
79787eaa
JM		 2853 if (ret)
2854 goto fail_trans_kthread;
76dda93c 2855
7c2ca468 2856 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b 2857 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2
ID		 2858 if (ret)
2859 goto fail_trans_kthread;
d68fc57b 2860
5d4f98a2 2861 ret = btrfs_recover_relocation(tree_root);
d7ce5843
MX		 2862 if (ret < 0) {
2863 printk(KERN_WARNING
efe120a0 2864 "BTRFS: failed to recover relocation\n");
d7ce5843 2865 err = -EINVAL;
bcef60f2 2866 goto fail_qgroup;
d7ce5843 2867 }
7c2ca468 2868 }
1a40e23b 2869
3de4586c
CM		 2870 location.objectid = BTRFS_FS_TREE_OBJECTID;
2871 location.type = BTRFS_ROOT_ITEM_KEY;
cb517eab 2872 location.offset = 0;
3de4586c 2873
3de4586c 2874 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
3140c9a3
DC		 2875 if (IS_ERR(fs_info->fs_root)) {
2876 err = PTR_ERR(fs_info->fs_root);
bcef60f2 2877 goto fail_qgroup;
3140c9a3 2878 }
c289811c 2879
2b6ba629
ID		 2880 if (sb->s_flags & MS_RDONLY)
2881 return 0;
59641015 2882
2b6ba629
ID		 2883 down_read(&fs_info->cleanup_work_sem);
2884 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2885 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 2886 up_read(&fs_info->cleanup_work_sem);
2b6ba629
ID		 2887 close_ctree(tree_root);
2888 return ret;
2889 }
2890 up_read(&fs_info->cleanup_work_sem);
59641015 2891
2b6ba629
ID		 2892 ret = btrfs_resume_balance_async(fs_info);
2893 if (ret) {
efe120a0 2894 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
2b6ba629
ID		 2895 close_ctree(tree_root);
2896 return ret;
e3acc2a6
JB		 2897 }
2898
8dabb742
SB		 2899 ret = btrfs_resume_dev_replace_async(fs_info);
2900 if (ret) {
efe120a0 2901 pr_warn("BTRFS: failed to resume dev_replace\n");
8dabb742
SB		 2902 close_ctree(tree_root);
2903 return ret;
2904 }
2905
b382a324
JS		 2906 btrfs_qgroup_rescan_resume(fs_info);
2907
f7a81ea4 2908 if (create_uuid_tree) {
efe120a0 2909 pr_info("BTRFS: creating UUID tree\n");
f7a81ea4
SB		 2910 ret = btrfs_create_uuid_tree(fs_info);
2911 if (ret) {
efe120a0 2912 pr_warn("BTRFS: failed to create the UUID tree %d\n",
f7a81ea4
SB		 2913 ret);
2914 close_ctree(tree_root);
2915 return ret;
2916 }
f420ee1e
SB		 2917 } else if (check_uuid_tree ||
2918 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
efe120a0 2919 pr_info("BTRFS: checking UUID tree\n");
70f80175
SB		 2920 ret = btrfs_check_uuid_tree(fs_info);
2921 if (ret) {
efe120a0 2922 pr_warn("BTRFS: failed to check the UUID tree %d\n",
70f80175
SB		 2923 ret);
2924 close_ctree(tree_root);
2925 return ret;
2926 }
2927 } else {
2928 fs_info->update_uuid_tree_gen = 1;
f7a81ea4
SB		 2929 }
2930
ad2b2c80 2931 return 0;
39279cc3 2932
bcef60f2
AJ		 2933fail_qgroup:
2934 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM		 2935fail_trans_kthread:
2936 kthread_stop(fs_info->transaction_kthread);
54067ae9 2937 btrfs_cleanup_transaction(fs_info->tree_root);
7b5ff90e 2938 del_fs_roots(fs_info);
3f157a2f 2939fail_cleaner:
a74a4b97 2940 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2941
2942 /*
2943 * make sure we're done with the btree inode before we stop our
2944 * kthreads
2945 */
2946 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 2947
2365dd3c
AJ		 2948fail_sysfs:
2949 btrfs_sysfs_remove_one(fs_info);
2950
1b1d1f66 2951fail_block_groups:
54067ae9 2952 btrfs_put_block_group_cache(fs_info);
1b1d1f66 2953 btrfs_free_block_groups(fs_info);
af31f5e5
CM		 2954
2955fail_tree_roots:
2956 free_root_pointers(fs_info, 1);
2b8195bb 2957 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 2958
39279cc3 2959fail_sb_buffer:
7abadb64 2960 btrfs_stop_all_workers(fs_info);
16cdcec7 2961fail_alloc:
4543df7e 2962fail_iput:
586e46e2
ID		 2963 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2964
4543df7e 2965 iput(fs_info->btree_inode);
963d678b
MX		 2966fail_delalloc_bytes:
2967 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX		 2968fail_dirty_metadata_bytes:
2969 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
ad081f14 2970fail_bdi:
7e662854 2971 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 2972fail_srcu:
2973 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 2974fail:
53b381b3 2975 btrfs_free_stripe_hash_table(fs_info);
586e46e2 2976 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 2977 return err;
af31f5e5
CM		 2978
2979recovery_tree_root:
af31f5e5
CM		 2980 if (!btrfs_test_opt(tree_root, RECOVERY))
2981 goto fail_tree_roots;
2982
2983 free_root_pointers(fs_info, 0);
2984
2985 /* don't use the log in recovery mode, it won't be valid */
2986 btrfs_set_super_log_root(disk_super, 0);
2987
2988 /* we can't trust the free space cache either */
2989 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
2990
2991 ret = next_root_backup(fs_info, fs_info->super_copy,
2992 &num_backups_tried, &backup_index);
2993 if (ret == -1)
2994 goto fail_block_groups;
2995 goto retry_root_backup;
eb60ceac
CM		 2996}
2997
f2984462
CM		 2998static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2999{
f2984462
CM		 3000 if (uptodate) {
3001 set_buffer_uptodate(bh);
3002 } else {
442a4f63
SB		 3003 struct btrfs_device *device = (struct btrfs_device *)
3004 bh->b_private;
3005
efe120a0 3006 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
606686ee
JB		 3007 "I/O error on %s\n",
3008 rcu_str_deref(device->name));
1259ab75
CM		 3009 /* note, we dont' set_buffer_write_io_error because we have
3010 * our own ways of dealing with the IO errors
3011 */
f2984462 3012 clear_buffer_uptodate(bh);
442a4f63 3013 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM		 3014 }
3015 unlock_buffer(bh);
3016 put_bh(bh);
3017}
3018
a512bbf8
YZ		 3019struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3020{
3021 struct buffer_head *bh;
3022 struct buffer_head *latest = NULL;
3023 struct btrfs_super_block *super;
3024 int i;
3025 u64 transid = 0;
3026 u64 bytenr;
3027
3028 /* we would like to check all the supers, but that would make
3029 * a btrfs mount succeed after a mkfs from a different FS.
3030 * So, we need to add a special mount option to scan for
3031 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3032 */
3033 for (i = 0; i < 1; i++) {
3034 bytenr = btrfs_sb_offset(i);
8068a47e
AJ		 3035 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3036 i_size_read(bdev->bd_inode))
a512bbf8 3037 break;
8068a47e
AJ		 3038 bh = __bread(bdev, bytenr / 4096,
3039 BTRFS_SUPER_INFO_SIZE);
a512bbf8
YZ		 3040 if (!bh)
3041 continue;
3042
3043 super = (struct btrfs_super_block *)bh->b_data;
3044 if (btrfs_super_bytenr(super) != bytenr ||
3cae210f 3045 btrfs_super_magic(super) != BTRFS_MAGIC) {
a512bbf8
YZ		 3046 brelse(bh);
3047 continue;
3048 }
3049
3050 if (!latest || btrfs_super_generation(super) > transid) {
3051 brelse(latest);
3052 latest = bh;
3053 transid = btrfs_super_generation(super);
3054 } else {
3055 brelse(bh);
3056 }
3057 }
3058 return latest;
3059}
3060
4eedeb75
HH		 3061/*
3062 * this should be called twice, once with wait == 0 and
3063 * once with wait == 1. When wait == 0 is done, all the buffer heads
3064 * we write are pinned.
3065 *
3066 * They are released when wait == 1 is done.
3067 * max_mirrors must be the same for both runs, and it indicates how
3068 * many supers on this one device should be written.
3069 *
3070 * max_mirrors == 0 means to write them all.
3071 */
a512bbf8
YZ		 3072static int write_dev_supers(struct btrfs_device *device,
3073 struct btrfs_super_block *sb,
3074 int do_barriers, int wait, int max_mirrors)
3075{
3076 struct buffer_head *bh;
3077 int i;
3078 int ret;
3079 int errors = 0;
3080 u32 crc;
3081 u64 bytenr;
a512bbf8
YZ		 3082
3083 if (max_mirrors == 0)
3084 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3085
a512bbf8
YZ		 3086 for (i = 0; i < max_mirrors; i++) {
3087 bytenr = btrfs_sb_offset(i);
3088 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
3089 break;
3090
3091 if (wait) {
3092 bh = __find_get_block(device->bdev, bytenr / 4096,
3093 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3094 if (!bh) {
3095 errors++;
3096 continue;
3097 }
a512bbf8 3098 wait_on_buffer(bh);
4eedeb75
HH		 3099 if (!buffer_uptodate(bh))
3100 errors++;
3101
3102 /* drop our reference */
3103 brelse(bh);
3104
3105 /* drop the reference from the wait == 0 run */
3106 brelse(bh);
3107 continue;
a512bbf8
YZ		 3108 } else {
3109 btrfs_set_super_bytenr(sb, bytenr);
3110
3111 crc = ~(u32)0;
b0496686 3112 crc = btrfs_csum_data((char *)sb +
a512bbf8
YZ		 3113 BTRFS_CSUM_SIZE, crc,
3114 BTRFS_SUPER_INFO_SIZE -
3115 BTRFS_CSUM_SIZE);
3116 btrfs_csum_final(crc, sb->csum);
3117
4eedeb75
HH		 3118 /*
3119 * one reference for us, and we leave it for the
3120 * caller
3121 */
a512bbf8
YZ		 3122 bh = __getblk(device->bdev, bytenr / 4096,
3123 BTRFS_SUPER_INFO_SIZE);
634554dc 3124 if (!bh) {
efe120a0 3125 printk(KERN_ERR "BTRFS: couldn't get super "
634554dc
JB		 3126 "buffer head for bytenr %Lu\n", bytenr);
3127 errors++;
3128 continue;
3129 }
3130
a512bbf8
YZ		 3131 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3132
4eedeb75 3133 /* one reference for submit_bh */
a512bbf8 3134 get_bh(bh);
4eedeb75
HH		 3135
3136 set_buffer_uptodate(bh);
a512bbf8
YZ		 3137 lock_buffer(bh);
3138 bh->b_end_io = btrfs_end_buffer_write_sync;
442a4f63 3139 bh->b_private = device;
a512bbf8
YZ		 3140 }
3141
387125fc
CM		 3142 /*
3143 * we fua the first super. The others we allow
3144 * to go down lazy.
3145 */
e8117c26
WS		 3146 if (i == 0)
3147 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3148 else
3149 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
4eedeb75 3150 if (ret)
a512bbf8 3151 errors++;
a512bbf8
YZ		 3152 }
3153 return errors < i ? 0 : -1;
3154}
3155
387125fc
CM		 3156/*
3157 * endio for the write_dev_flush, this will wake anyone waiting
3158 * for the barrier when it is done
3159 */
3160static void btrfs_end_empty_barrier(struct bio *bio, int err)
3161{
3162 if (err) {
3163 if (err == -EOPNOTSUPP)
3164 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3165 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3166 }
3167 if (bio->bi_private)
3168 complete(bio->bi_private);
3169 bio_put(bio);
3170}
3171
3172/*
3173 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3174 * sent down. With wait == 1, it waits for the previous flush.
3175 *
3176 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3177 * capable
3178 */
3179static int write_dev_flush(struct btrfs_device *device, int wait)
3180{
3181 struct bio *bio;
3182 int ret = 0;
3183
3184 if (device->nobarriers)
3185 return 0;
3186
3187 if (wait) {
3188 bio = device->flush_bio;
3189 if (!bio)
3190 return 0;
3191
3192 wait_for_completion(&device->flush_wait);
3193
3194 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
efe120a0 3195 printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
606686ee 3196 rcu_str_deref(device->name));
387125fc 3197 device->nobarriers = 1;
5af3e8cc 3198 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
387125fc 3199 ret = -EIO;
5af3e8cc
SB		 3200 btrfs_dev_stat_inc_and_print(device,
3201 BTRFS_DEV_STAT_FLUSH_ERRS);
387125fc
CM		 3202 }
3203
3204 /* drop the reference from the wait == 0 run */
3205 bio_put(bio);
3206 device->flush_bio = NULL;
3207
3208 return ret;
3209 }
3210
3211 /*
3212 * one reference for us, and we leave it for the
3213 * caller
3214 */
9c017abc 3215 device->flush_bio = NULL;
9be3395b 3216 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
387125fc
CM		 3217 if (!bio)
3218 return -ENOMEM;
3219
3220 bio->bi_end_io = btrfs_end_empty_barrier;
3221 bio->bi_bdev = device->bdev;
3222 init_completion(&device->flush_wait);
3223 bio->bi_private = &device->flush_wait;
3224 device->flush_bio = bio;
3225
3226 bio_get(bio);
21adbd5c 3227 btrfsic_submit_bio(WRITE_FLUSH, bio);
387125fc
CM		 3228
3229 return 0;
3230}
3231
3232/*
3233 * send an empty flush down to each device in parallel,
3234 * then wait for them
3235 */
3236static int barrier_all_devices(struct btrfs_fs_info *info)
3237{
3238 struct list_head *head;
3239 struct btrfs_device *dev;
5af3e8cc
SB		 3240 int errors_send = 0;
3241 int errors_wait = 0;
387125fc
CM		 3242 int ret;
3243
3244 /* send down all the barriers */
3245 head = &info->fs_devices->devices;
3246 list_for_each_entry_rcu(dev, head, dev_list) {
3247 if (!dev->bdev) {
5af3e8cc 3248 errors_send++;
387125fc
CM		 3249 continue;
3250 }
3251 if (!dev->in_fs_metadata || !dev->writeable)
3252 continue;
3253
3254 ret = write_dev_flush(dev, 0);
3255 if (ret)
5af3e8cc 3256 errors_send++;
387125fc
CM		 3257 }
3258
3259 /* wait for all the barriers */
3260 list_for_each_entry_rcu(dev, head, dev_list) {
3261 if (!dev->bdev) {
5af3e8cc 3262 errors_wait++;
387125fc
CM		 3263 continue;
3264 }
3265 if (!dev->in_fs_metadata || !dev->writeable)
3266 continue;
3267
3268 ret = write_dev_flush(dev, 1);
3269 if (ret)
5af3e8cc 3270 errors_wait++;
387125fc 3271 }
5af3e8cc
SB		 3272 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3273 errors_wait > info->num_tolerated_disk_barrier_failures)
387125fc
CM		 3274 return -EIO;
3275 return 0;
3276}
3277
5af3e8cc
SB		 3278int btrfs_calc_num_tolerated_disk_barrier_failures(
3279 struct btrfs_fs_info *fs_info)
3280{
3281 struct btrfs_ioctl_space_info space;
3282 struct btrfs_space_info *sinfo;
3283 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3284 BTRFS_BLOCK_GROUP_SYSTEM,
3285 BTRFS_BLOCK_GROUP_METADATA,
3286 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3287 int num_types = 4;
3288 int i;
3289 int c;
3290 int num_tolerated_disk_barrier_failures =
3291 (int)fs_info->fs_devices->num_devices;
3292
3293 for (i = 0; i < num_types; i++) {
3294 struct btrfs_space_info *tmp;
3295
3296 sinfo = NULL;
3297 rcu_read_lock();
3298 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3299 if (tmp->flags == types[i]) {
3300 sinfo = tmp;
3301 break;
3302 }
3303 }
3304 rcu_read_unlock();
3305
3306 if (!sinfo)
3307 continue;
3308
3309 down_read(&sinfo->groups_sem);
3310 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3311 if (!list_empty(&sinfo->block_groups[c])) {
3312 u64 flags;
3313
3314 btrfs_get_block_group_info(
3315 &sinfo->block_groups[c], &space);
3316 if (space.total_bytes == 0 ||
3317 space.used_bytes == 0)
3318 continue;
3319 flags = space.flags;
3320 /*
3321 * return
3322 * 0: if dup, single or RAID0 is configured for
3323 * any of metadata, system or data, else
3324 * 1: if RAID5 is configured, or if RAID1 or
3325 * RAID10 is configured and only two mirrors
3326 * are used, else
3327 * 2: if RAID6 is configured, else
3328 * num_mirrors - 1: if RAID1 or RAID10 is
3329 * configured and more than
3330 * 2 mirrors are used.
3331 */
3332 if (num_tolerated_disk_barrier_failures > 0 &&
3333 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3334 BTRFS_BLOCK_GROUP_RAID0)) ||
3335 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3336 == 0)))
3337 num_tolerated_disk_barrier_failures = 0;
53b381b3
DW		 3338 else if (num_tolerated_disk_barrier_failures > 1) {
3339 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3340 BTRFS_BLOCK_GROUP_RAID5 |
3341 BTRFS_BLOCK_GROUP_RAID10)) {
3342 num_tolerated_disk_barrier_failures = 1;
3343 } else if (flags &
15b0a89d 3344 BTRFS_BLOCK_GROUP_RAID6) {
53b381b3
DW		 3345 num_tolerated_disk_barrier_failures = 2;
3346 }
3347 }
5af3e8cc
SB		 3348 }
3349 }
3350 up_read(&sinfo->groups_sem);
3351 }
3352
3353 return num_tolerated_disk_barrier_failures;
3354}
3355
48a3b636 3356static int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 3357{
e5e9a520 3358 struct list_head *head;
f2984462 3359 struct btrfs_device *dev;
a061fc8d 3360 struct btrfs_super_block *sb;
f2984462 3361 struct btrfs_dev_item *dev_item;
f2984462
CM		 3362 int ret;
3363 int do_barriers;
a236aed1
CM		 3364 int max_errors;
3365 int total_errors = 0;
a061fc8d 3366 u64 flags;
f2984462
CM		 3367
3368 do_barriers = !btrfs_test_opt(root, NOBARRIER);
af31f5e5 3369 backup_super_roots(root->fs_info);
f2984462 3370
6c41761f 3371 sb = root->fs_info->super_for_commit;
a061fc8d 3372 dev_item = &sb->dev_item;
e5e9a520 3373
174ba509 3374 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 3375 head = &root->fs_info->fs_devices->devices;
d7306801 3376 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
387125fc 3377
5af3e8cc
SB		 3378 if (do_barriers) {
3379 ret = barrier_all_devices(root->fs_info);
3380 if (ret) {
3381 mutex_unlock(
3382 &root->fs_info->fs_devices->device_list_mutex);
3383 btrfs_error(root->fs_info, ret,
3384 "errors while submitting device barriers.");
3385 return ret;
3386 }
3387 }
387125fc 3388
1f78160c 3389 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3390 if (!dev->bdev) {
3391 total_errors++;
3392 continue;
3393 }
2b82032c 3394 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3395 continue;
3396
2b82032c 3397 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 3398 btrfs_set_stack_device_type(dev_item, dev->type);
3399 btrfs_set_stack_device_id(dev_item, dev->devid);
3400 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
3401 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
3402 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3403 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3404 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3405 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 3406 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 3407
a061fc8d
CM		 3408 flags = btrfs_super_flags(sb);
3409 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3410
a512bbf8 3411 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 3412 if (ret)
3413 total_errors++;
f2984462 3414 }
a236aed1 3415 if (total_errors > max_errors) {
efe120a0 3416 btrfs_err(root->fs_info, "%d errors while writing supers",
d397712b 3417 total_errors);
a724b436 3418 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
79787eaa 3419
9d565ba4
SB		 3420 /* FUA is masked off if unsupported and can't be the reason */
3421 btrfs_error(root->fs_info, -EIO,
3422 "%d errors while writing supers", total_errors);
3423 return -EIO;
a236aed1 3424 }
f2984462 3425
a512bbf8 3426 total_errors = 0;
1f78160c 3427 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3428 if (!dev->bdev)
3429 continue;
2b82032c 3430 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3431 continue;
3432
a512bbf8
YZ		 3433 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3434 if (ret)
3435 total_errors++;
f2984462 3436 }
174ba509 3437 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 3438 if (total_errors > max_errors) {
79787eaa
JM		 3439 btrfs_error(root->fs_info, -EIO,
3440 "%d errors while writing supers", total_errors);
3441 return -EIO;
a236aed1 3442 }
f2984462
CM		 3443 return 0;
3444}
3445
a512bbf8
YZ		 3446int write_ctree_super(struct btrfs_trans_handle *trans,
3447 struct btrfs_root *root, int max_mirrors)
eb60ceac 3448{
f570e757 3449 return write_all_supers(root, max_mirrors);
cfaa7295
CM		 3450}
3451
cb517eab
MX		 3452/* Drop a fs root from the radix tree and free it. */
3453void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3454 struct btrfs_root *root)
2619ba1f 3455{
4df27c4d 3456 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 3457 radix_tree_delete(&fs_info->fs_roots_radix,
3458 (unsigned long)root->root_key.objectid);
4df27c4d 3459 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 3460
3461 if (btrfs_root_refs(&root->root_item) == 0)
3462 synchronize_srcu(&fs_info->subvol_srcu);
3463
1a4319cc 3464 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
3321719e 3465 btrfs_free_log(NULL, root);
3321719e 3466
581bb050
LZ		 3467 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3468 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3469 free_fs_root(root);
4df27c4d
YZ		 3470}
3471
3472static void free_fs_root(struct btrfs_root *root)
3473{
82d5902d 3474 iput(root->cache_inode);
4df27c4d 3475 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
1cb048f5
FDBM		 3476 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3477 root->orphan_block_rsv = NULL;
0ee5dc67
AV		 3478 if (root->anon_dev)
3479 free_anon_bdev(root->anon_dev);
4df27c4d
YZ		 3480 free_extent_buffer(root->node);
3481 free_extent_buffer(root->commit_root);
581bb050
LZ		 3482 kfree(root->free_ino_ctl);
3483 kfree(root->free_ino_pinned);
d397712b 3484 kfree(root->name);
b0feb9d9 3485 btrfs_put_fs_root(root);
2619ba1f
CM		 3486}
3487
cb517eab
MX		 3488void btrfs_free_fs_root(struct btrfs_root *root)
3489{
3490 free_fs_root(root);
2619ba1f
CM		 3491}
3492
c146afad 3493int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3494{
c146afad
YZ		 3495 u64 root_objectid = 0;
3496 struct btrfs_root *gang[8];
3497 int i;
3768f368 3498 int ret;
e089f05c 3499
c146afad
YZ		 3500 while (1) {
3501 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3502 (void **)gang, root_objectid,
3503 ARRAY_SIZE(gang));
3504 if (!ret)
3505 break;
5d4f98a2
YZ		 3506
3507 root_objectid = gang[ret - 1]->root_key.objectid + 1;
c146afad 3508 for (i = 0; i < ret; i++) {
66b4ffd1
JB		 3509 int err;
3510
c146afad 3511 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 3512 err = btrfs_orphan_cleanup(gang[i]);
3513 if (err)
3514 return err;
c146afad
YZ		 3515 }
3516 root_objectid++;
3517 }
3518 return 0;
3519}
a2135011 3520
c146afad
YZ		 3521int btrfs_commit_super(struct btrfs_root *root)
3522{
3523 struct btrfs_trans_handle *trans;
a74a4b97 3524
c146afad 3525 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 3526 btrfs_run_delayed_iputs(root);
c146afad 3527 mutex_unlock(&root->fs_info->cleaner_mutex);
9d1a2a3a 3528 wake_up_process(root->fs_info->cleaner_kthread);
c71bf099
YZ		 3529
3530 /* wait until ongoing cleanup work done */
3531 down_write(&root->fs_info->cleanup_work_sem);
3532 up_write(&root->fs_info->cleanup_work_sem);
3533
7a7eaa40 3534 trans = btrfs_join_transaction(root);
3612b495
TI		 3535 if (IS_ERR(trans))
3536 return PTR_ERR(trans);
d52c1bcc 3537 return btrfs_commit_transaction(trans, root);
c146afad
YZ		 3538}
3539
3540int close_ctree(struct btrfs_root *root)
3541{
3542 struct btrfs_fs_info *fs_info = root->fs_info;
3543 int ret;
3544
3545 fs_info->closing = 1;
3546 smp_mb();
3547
803b2f54
SB		 3548 /* wait for the uuid_scan task to finish */
3549 down(&fs_info->uuid_tree_rescan_sem);
3550 /* avoid complains from lockdep et al., set sem back to initial state */
3551 up(&fs_info->uuid_tree_rescan_sem);
3552
837d5b6e 3553 /* pause restriper - we want to resume on mount */
aa1b8cd4 3554 btrfs_pause_balance(fs_info);
837d5b6e 3555
8dabb742
SB		 3556 btrfs_dev_replace_suspend_for_unmount(fs_info);
3557
aa1b8cd4 3558 btrfs_scrub_cancel(fs_info);
4cb5300b
CM		 3559
3560 /* wait for any defraggers to finish */
3561 wait_event(fs_info->transaction_wait,
3562 (atomic_read(&fs_info->defrag_running) == 0));
3563
3564 /* clear out the rbtree of defraggable inodes */
26176e7c 3565 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3566
c146afad 3567 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 3568 ret = btrfs_commit_super(root);
3569 if (ret)
efe120a0 3570 btrfs_err(root->fs_info, "commit super ret %d", ret);
acce952b 3571 }
3572
87533c47 3573 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
68ce9682 3574 btrfs_error_commit_super(root);
0f7d52f4 3575
e3029d9f
AV		 3576 kthread_stop(fs_info->transaction_kthread);
3577 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3578
f25784b3
YZ		 3579 fs_info->closing = 2;
3580 smp_mb();
3581
bcef60f2
AJ		 3582 btrfs_free_qgroup_config(root->fs_info);
3583
963d678b 3584 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
efe120a0 3585 btrfs_info(root->fs_info, "at unmount delalloc count %lld",
963d678b 3586 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3587 }
bcc63abb 3588
5ac1d209
JM		 3589 btrfs_sysfs_remove_one(fs_info);
3590
c146afad 3591 del_fs_roots(fs_info);
d10c5f31 3592
1a4319cc
LB		 3593 btrfs_put_block_group_cache(fs_info);
3594
2b1360da
JB		 3595 btrfs_free_block_groups(fs_info);
3596
96192499
JB		 3597 btrfs_stop_all_workers(fs_info);
3598
13e6c37b 3599 free_root_pointers(fs_info, 1);
9ad6b7bc 3600
13e6c37b 3601 iput(fs_info->btree_inode);
d6bfde87 3602
21adbd5c
SB		 3603#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3604 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3605 btrfsic_unmount(root, fs_info->fs_devices);
3606#endif
3607
dfe25020 3608 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3609 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3610
e2d84521 3611 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 3612 percpu_counter_destroy(&fs_info->delalloc_bytes);
04160088 3613 bdi_destroy(&fs_info->bdi);
76dda93c 3614 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 3615
53b381b3
DW		 3616 btrfs_free_stripe_hash_table(fs_info);
3617
1cb048f5
FDBM		 3618 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3619 root->orphan_block_rsv = NULL;
3620
eb60ceac
CM		 3621 return 0;
3622}
3623
b9fab919
CM		 3624int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3625 int atomic)
5f39d397 3626{
1259ab75 3627 int ret;
727011e0 3628 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 3629
0b32f4bb 3630 ret = extent_buffer_uptodate(buf);
1259ab75
CM		 3631 if (!ret)
3632 return ret;
3633
3634 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM		 3635 parent_transid, atomic);
3636 if (ret == -EAGAIN)
3637 return ret;
1259ab75 3638 return !ret;
5f39d397
CM		 3639}
3640
3641int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 3642{
0b32f4bb 3643 return set_extent_buffer_uptodate(buf);
5f39d397 3644}
6702ed49 3645
5f39d397
CM		 3646void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3647{
06ea65a3 3648 struct btrfs_root *root;
5f39d397 3649 u64 transid = btrfs_header_generation(buf);
b9473439 3650 int was_dirty;
b4ce94de 3651
06ea65a3
JB		 3652#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3653 /*
3654 * This is a fast path so only do this check if we have sanity tests
3655 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3656 * outside of the sanity tests.
3657 */
3658 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3659 return;
3660#endif
3661 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
b9447ef8 3662 btrfs_assert_tree_locked(buf);
31b1a2bd
JL		 3663 if (transid != root->fs_info->generation)
3664 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
d397712b 3665 "found %llu running %llu\n",
c1c9ff7c 3666 buf->start, transid, root->fs_info->generation);
0b32f4bb 3667 was_dirty = set_extent_buffer_dirty(buf);
e2d84521
MX		 3668 if (!was_dirty)
3669 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3670 buf->len,
3671 root->fs_info->dirty_metadata_batch);
eb60ceac
CM		 3672}
3673
b53d3f5d
LB		 3674static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3675 int flush_delayed)
16cdcec7
MX		 3676{
3677 /*
3678 * looks as though older kernels can get into trouble with
3679 * this code, they end up stuck in balance_dirty_pages forever
3680 */
e2d84521 3681 int ret;
16cdcec7
MX		 3682
3683 if (current->flags & PF_MEMALLOC)
3684 return;
3685
b53d3f5d
LB		 3686 if (flush_delayed)
3687 btrfs_balance_delayed_items(root);
16cdcec7 3688
e2d84521
MX		 3689 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3690 BTRFS_DIRTY_METADATA_THRESH);
3691 if (ret > 0) {
d0e1d66b
NJ		 3692 balance_dirty_pages_ratelimited(
3693 root->fs_info->btree_inode->i_mapping);
16cdcec7
MX		 3694 }
3695 return;
3696}
3697
b53d3f5d 3698void btrfs_btree_balance_dirty(struct btrfs_root *root)
35b7e476 3699{
b53d3f5d
LB		 3700 __btrfs_btree_balance_dirty(root, 1);
3701}
585ad2c3 3702
b53d3f5d
LB		 3703void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3704{
3705 __btrfs_btree_balance_dirty(root, 0);
35b7e476 3706}
6b80053d 3707
ca7a79ad 3708int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 3709{
727011e0 3710 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
0b32f4bb 3711 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
6b80053d 3712}
0da5468f 3713
fcd1f065 3714static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 3715 int read_only)
3716{
1104a885
DS		 3717 /*
3718 * Placeholder for checks
3719 */
fcd1f065 3720 return 0;
acce952b 3721}
3722
48a3b636 3723static void btrfs_error_commit_super(struct btrfs_root *root)
acce952b 3724{
acce952b 3725 mutex_lock(&root->fs_info->cleaner_mutex);
3726 btrfs_run_delayed_iputs(root);
3727 mutex_unlock(&root->fs_info->cleaner_mutex);
3728
3729 down_write(&root->fs_info->cleanup_work_sem);
3730 up_write(&root->fs_info->cleanup_work_sem);
3731
3732 /* cleanup FS via transaction */
3733 btrfs_cleanup_transaction(root);
acce952b 3734}
3735
569e0f35
JB		 3736static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
3737 struct btrfs_root *root)
acce952b 3738{
3739 struct btrfs_inode *btrfs_inode;
3740 struct list_head splice;
3741
3742 INIT_LIST_HEAD(&splice);
3743
3744 mutex_lock(&root->fs_info->ordered_operations_mutex);
199c2a9c 3745 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3746
569e0f35 3747 list_splice_init(&t->ordered_operations, &splice);
acce952b 3748 while (!list_empty(&splice)) {
3749 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3750 ordered_operations);
3751
3752 list_del_init(&btrfs_inode->ordered_operations);
199c2a9c 3753 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3754
3755 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3756
199c2a9c 3757 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3758 }
3759
199c2a9c 3760 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3761 mutex_unlock(&root->fs_info->ordered_operations_mutex);
acce952b 3762}
3763
143bede5 3764static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 3765{
acce952b 3766 struct btrfs_ordered_extent *ordered;
acce952b 3767
199c2a9c 3768 spin_lock(&root->ordered_extent_lock);
779880ef
JB		 3769 /*
3770 * This will just short circuit the ordered completion stuff which will
3771 * make sure the ordered extent gets properly cleaned up.
3772 */
199c2a9c 3773 list_for_each_entry(ordered, &root->ordered_extents,
779880ef
JB		 3774 root_extent_list)
3775 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
199c2a9c
MX		 3776 spin_unlock(&root->ordered_extent_lock);
3777}
3778
3779static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3780{
3781 struct btrfs_root *root;
3782 struct list_head splice;
3783
3784 INIT_LIST_HEAD(&splice);
3785
3786 spin_lock(&fs_info->ordered_root_lock);
3787 list_splice_init(&fs_info->ordered_roots, &splice);
3788 while (!list_empty(&splice)) {
3789 root = list_first_entry(&splice, struct btrfs_root,
3790 ordered_root);
1de2cfde
JB		 3791 list_move_tail(&root->ordered_root,
3792 &fs_info->ordered_roots);
199c2a9c
MX		 3793
3794 btrfs_destroy_ordered_extents(root);
3795
3796 cond_resched_lock(&fs_info->ordered_root_lock);
3797 }
3798 spin_unlock(&fs_info->ordered_root_lock);
acce952b 3799}
3800
35a3621b
SB		 3801static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3802 struct btrfs_root *root)
acce952b 3803{
3804 struct rb_node *node;
3805 struct btrfs_delayed_ref_root *delayed_refs;
3806 struct btrfs_delayed_ref_node *ref;
3807 int ret = 0;
3808
3809 delayed_refs = &trans->delayed_refs;
3810
3811 spin_lock(&delayed_refs->lock);
d7df2c79 3812 if (atomic_read(&delayed_refs->num_entries) == 0) {
cfece4db 3813 spin_unlock(&delayed_refs->lock);
efe120a0 3814 btrfs_info(root->fs_info, "delayed_refs has NO entry");
acce952b 3815 return ret;
3816 }
3817
d7df2c79
JB		 3818 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
3819 struct btrfs_delayed_ref_head *head;
e78417d1 3820 bool pin_bytes = false;
acce952b 3821
d7df2c79
JB		 3822 head = rb_entry(node, struct btrfs_delayed_ref_head,
3823 href_node);
3824 if (!mutex_trylock(&head->mutex)) {
3825 atomic_inc(&head->node.refs);
3826 spin_unlock(&delayed_refs->lock);
eb12db69 3827
d7df2c79 3828 mutex_lock(&head->mutex);
e78417d1 3829 mutex_unlock(&head->mutex);
d7df2c79
JB		 3830 btrfs_put_delayed_ref(&head->node);
3831 spin_lock(&delayed_refs->lock);
3832 continue;
3833 }
3834 spin_lock(&head->lock);
3835 while ((node = rb_first(&head->ref_root)) != NULL) {
3836 ref = rb_entry(node, struct btrfs_delayed_ref_node,
3837 rb_node);
3838 ref->in_tree = 0;
3839 rb_erase(&ref->rb_node, &head->ref_root);
3840 atomic_dec(&delayed_refs->num_entries);
3841 btrfs_put_delayed_ref(ref);
e78417d1 3842 }
d7df2c79
JB		 3843 if (head->must_insert_reserved)
3844 pin_bytes = true;
3845 btrfs_free_delayed_extent_op(head->extent_op);
3846 delayed_refs->num_heads--;
3847 if (head->processing == 0)
3848 delayed_refs->num_heads_ready--;
3849 atomic_dec(&delayed_refs->num_entries);
3850 head->node.in_tree = 0;
3851 rb_erase(&head->href_node, &delayed_refs->href_root);
3852 spin_unlock(&head->lock);
3853 spin_unlock(&delayed_refs->lock);
3854 mutex_unlock(&head->mutex);
acce952b 3855
d7df2c79
JB		 3856 if (pin_bytes)
3857 btrfs_pin_extent(root, head->node.bytenr,
3858 head->node.num_bytes, 1);
3859 btrfs_put_delayed_ref(&head->node);
acce952b 3860 cond_resched();
3861 spin_lock(&delayed_refs->lock);
3862 }
3863
3864 spin_unlock(&delayed_refs->lock);
3865
3866 return ret;
3867}
3868
143bede5 3869static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 3870{
3871 struct btrfs_inode *btrfs_inode;
3872 struct list_head splice;
3873
3874 INIT_LIST_HEAD(&splice);
3875
eb73c1b7
MX		 3876 spin_lock(&root->delalloc_lock);
3877 list_splice_init(&root->delalloc_inodes, &splice);
acce952b 3878
3879 while (!list_empty(&splice)) {
eb73c1b7
MX		 3880 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3881 delalloc_inodes);
acce952b 3882
3883 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX		 3884 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3885 &btrfs_inode->runtime_flags);
eb73c1b7 3886 spin_unlock(&root->delalloc_lock);
acce952b 3887
3888 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3889
eb73c1b7 3890 spin_lock(&root->delalloc_lock);
acce952b 3891 }
3892
eb73c1b7
MX		 3893 spin_unlock(&root->delalloc_lock);
3894}
3895
3896static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3897{
3898 struct btrfs_root *root;
3899 struct list_head splice;
3900
3901 INIT_LIST_HEAD(&splice);
3902
3903 spin_lock(&fs_info->delalloc_root_lock);
3904 list_splice_init(&fs_info->delalloc_roots, &splice);
3905 while (!list_empty(&splice)) {
3906 root = list_first_entry(&splice, struct btrfs_root,
3907 delalloc_root);
3908 list_del_init(&root->delalloc_root);
3909 root = btrfs_grab_fs_root(root);
3910 BUG_ON(!root);
3911 spin_unlock(&fs_info->delalloc_root_lock);
3912
3913 btrfs_destroy_delalloc_inodes(root);
3914 btrfs_put_fs_root(root);
3915
3916 spin_lock(&fs_info->delalloc_root_lock);
3917 }
3918 spin_unlock(&fs_info->delalloc_root_lock);
acce952b 3919}
3920
3921static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3922 struct extent_io_tree *dirty_pages,
3923 int mark)
3924{
3925 int ret;
acce952b 3926 struct extent_buffer *eb;
3927 u64 start = 0;
3928 u64 end;
acce952b 3929
3930 while (1) {
3931 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 3932 mark, NULL);
acce952b 3933 if (ret)
3934 break;
3935
3936 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
3937 while (start <= end) {
fd8b2b61
JB		 3938 eb = btrfs_find_tree_block(root, start,
3939 root->leafsize);
69a85bd8 3940 start += root->leafsize;
fd8b2b61 3941 if (!eb)
acce952b 3942 continue;
fd8b2b61 3943 wait_on_extent_buffer_writeback(eb);
acce952b 3944
fd8b2b61
JB		 3945 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
3946 &eb->bflags))
3947 clear_extent_buffer_dirty(eb);
3948 free_extent_buffer_stale(eb);
acce952b 3949 }
3950 }
3951
3952 return ret;
3953}
3954
3955static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3956 struct extent_io_tree *pinned_extents)
3957{
3958 struct extent_io_tree *unpin;
3959 u64 start;
3960 u64 end;
3961 int ret;
ed0eaa14 3962 bool loop = true;
acce952b 3963
3964 unpin = pinned_extents;
ed0eaa14 3965again:
acce952b 3966 while (1) {
3967 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 3968 EXTENT_DIRTY, NULL);
acce952b 3969 if (ret)
3970 break;
3971
3972 /* opt_discard */
5378e607
LD		 3973 if (btrfs_test_opt(root, DISCARD))
3974 ret = btrfs_error_discard_extent(root, start,
3975 end + 1 - start,
3976 NULL);
acce952b 3977
3978 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3979 btrfs_error_unpin_extent_range(root, start, end);
3980 cond_resched();
3981 }
3982
ed0eaa14
LB		 3983 if (loop) {
3984 if (unpin == &root->fs_info->freed_extents[0])
3985 unpin = &root->fs_info->freed_extents[1];
3986 else
3987 unpin = &root->fs_info->freed_extents[0];
3988 loop = false;
3989 goto again;
3990 }
3991
acce952b 3992 return 0;
3993}
3994
49b25e05
JM		 3995void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
3996 struct btrfs_root *root)
3997{
724e2315
JB		 3998 btrfs_destroy_ordered_operations(cur_trans, root);
3999
49b25e05 4000 btrfs_destroy_delayed_refs(cur_trans, root);
49b25e05 4001
4a9d8bde 4002 cur_trans->state = TRANS_STATE_COMMIT_START;
d7096fc3 4003 wake_up(&root->fs_info->transaction_blocked_wait);
49b25e05 4004
4a9d8bde 4005 cur_trans->state = TRANS_STATE_UNBLOCKED;
d7096fc3 4006 wake_up(&root->fs_info->transaction_wait);
49b25e05 4007
67cde344
MX		 4008 btrfs_destroy_delayed_inodes(root);
4009 btrfs_assert_delayed_root_empty(root);
49b25e05 4010
49b25e05
JM		 4011 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4012 EXTENT_DIRTY);
6e841e32
LB		 4013 btrfs_destroy_pinned_extent(root,
4014 root->fs_info->pinned_extents);
49b25e05 4015
4a9d8bde
MX		 4016 cur_trans->state =TRANS_STATE_COMPLETED;
4017 wake_up(&cur_trans->commit_wait);
4018
49b25e05
JM		 4019 /*
4020 memset(cur_trans, 0, sizeof(*cur_trans));
4021 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4022 */
4023}
4024
48a3b636 4025static int btrfs_cleanup_transaction(struct btrfs_root *root)
acce952b 4026{
4027 struct btrfs_transaction *t;
acce952b 4028
acce952b 4029 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4030
a4abeea4 4031 spin_lock(&root->fs_info->trans_lock);
724e2315
JB		 4032 while (!list_empty(&root->fs_info->trans_list)) {
4033 t = list_first_entry(&root->fs_info->trans_list,
4034 struct btrfs_transaction, list);
4035 if (t->state >= TRANS_STATE_COMMIT_START) {
4036 atomic_inc(&t->use_count);
4037 spin_unlock(&root->fs_info->trans_lock);
4038 btrfs_wait_for_commit(root, t->transid);
4039 btrfs_put_transaction(t);
4040 spin_lock(&root->fs_info->trans_lock);
4041 continue;
4042 }
4043 if (t == root->fs_info->running_transaction) {
4044 t->state = TRANS_STATE_COMMIT_DOING;
4045 spin_unlock(&root->fs_info->trans_lock);
4046 /*
4047 * We wait for 0 num_writers since we don't hold a trans
4048 * handle open currently for this transaction.
4049 */
4050 wait_event(t->writer_wait,
4051 atomic_read(&t->num_writers) == 0);
4052 } else {
4053 spin_unlock(&root->fs_info->trans_lock);
4054 }
4055 btrfs_cleanup_one_transaction(t, root);
4a9d8bde 4056
724e2315
JB		 4057 spin_lock(&root->fs_info->trans_lock);
4058 if (t == root->fs_info->running_transaction)
4059 root->fs_info->running_transaction = NULL;
acce952b 4060 list_del_init(&t->list);
724e2315 4061 spin_unlock(&root->fs_info->trans_lock);
acce952b 4062
724e2315
JB		 4063 btrfs_put_transaction(t);
4064 trace_btrfs_transaction_commit(root);
4065 spin_lock(&root->fs_info->trans_lock);
4066 }
4067 spin_unlock(&root->fs_info->trans_lock);
4068 btrfs_destroy_all_ordered_extents(root->fs_info);
4069 btrfs_destroy_delayed_inodes(root);
4070 btrfs_assert_delayed_root_empty(root);
4071 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4072 btrfs_destroy_all_delalloc_inodes(root->fs_info);
acce952b 4073 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 4074
4075 return 0;
4076}
4077
d1310b2e 4078static struct extent_io_ops btree_extent_io_ops = {
ce9adaa5 4079 .readpage_end_io_hook = btree_readpage_end_io_hook,
4bb31e92 4080 .readpage_io_failed_hook = btree_io_failed_hook,
0b86a832 4081 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 4082 /* note we're sharing with inode.c for the merge bio hook */
4083 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 4084};
Linux kernel - Deliverables
This page took 0.703995 seconds and 5 git commands to generate.