projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: rework qgroup accounting
[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"
fcebe456 52#include "qgroup.h"
eb60ceac 53
de0022b9
JB		 54#ifdef CONFIG_X86
55#include <asm/cpufeature.h>
56#endif
57
d1310b2e 58static struct extent_io_ops btree_extent_io_ops;
8b712842 59static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 60static void free_fs_root(struct btrfs_root *root);
fcd1f065 61static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 62 int read_only);
569e0f35
JB		 63static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
64 struct btrfs_root *root);
143bede5 65static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
acce952b 66static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
67 struct btrfs_root *root);
143bede5 68static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
acce952b 69static int btrfs_destroy_marked_extents(struct btrfs_root *root,
70 struct extent_io_tree *dirty_pages,
71 int mark);
72static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
73 struct extent_io_tree *pinned_extents);
48a3b636
ES		 74static int btrfs_cleanup_transaction(struct btrfs_root *root);
75static void btrfs_error_commit_super(struct btrfs_root *root);
ce9adaa5 76
d352ac68
CM		 77/*
78 * end_io_wq structs are used to do processing in task context when an IO is
79 * complete. This is used during reads to verify checksums, and it is used
80 * by writes to insert metadata for new file extents after IO is complete.
81 */
ce9adaa5
CM		 82struct end_io_wq {
83 struct bio *bio;
84 bio_end_io_t *end_io;
85 void *private;
86 struct btrfs_fs_info *info;
87 int error;
22c59948 88 int metadata;
ce9adaa5 89 struct list_head list;
8b712842 90 struct btrfs_work work;
ce9adaa5 91};
0da5468f 92
d352ac68
CM		 93/*
94 * async submit bios are used to offload expensive checksumming
95 * onto the worker threads. They checksum file and metadata bios
96 * just before they are sent down the IO stack.
97 */
44b8bd7e
CM		 98struct async_submit_bio {
99 struct inode *inode;
100 struct bio *bio;
101 struct list_head list;
4a69a410
CM		 102 extent_submit_bio_hook_t *submit_bio_start;
103 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM		 104 int rw;
105 int mirror_num;
c8b97818 106 unsigned long bio_flags;
eaf25d93
CM		 107 /*
108 * bio_offset is optional, can be used if the pages in the bio
109 * can't tell us where in the file the bio should go
110 */
111 u64 bio_offset;
8b712842 112 struct btrfs_work work;
79787eaa 113 int error;
44b8bd7e
CM		 114};
115
85d4e461
CM		 116/*
117 * Lockdep class keys for extent_buffer->lock's in this root. For a given
118 * eb, the lockdep key is determined by the btrfs_root it belongs to and
119 * the level the eb occupies in the tree.
120 *
121 * Different roots are used for different purposes and may nest inside each
122 * other and they require separate keysets. As lockdep keys should be
123 * static, assign keysets according to the purpose of the root as indicated
124 * by btrfs_root->objectid. This ensures that all special purpose roots
125 * have separate keysets.
4008c04a 126 *
85d4e461
CM		 127 * Lock-nesting across peer nodes is always done with the immediate parent
128 * node locked thus preventing deadlock. As lockdep doesn't know this, use
129 * subclass to avoid triggering lockdep warning in such cases.
4008c04a 130 *
85d4e461
CM		 131 * The key is set by the readpage_end_io_hook after the buffer has passed
132 * csum validation but before the pages are unlocked. It is also set by
133 * btrfs_init_new_buffer on freshly allocated blocks.
4008c04a 134 *
85d4e461
CM		 135 * We also add a check to make sure the highest level of the tree is the
136 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
137 * needs update as well.
4008c04a
CM		 138 */
139#ifdef CONFIG_DEBUG_LOCK_ALLOC
140# if BTRFS_MAX_LEVEL != 8
141# error
142# endif
85d4e461
CM		 143
144static struct btrfs_lockdep_keyset {
145 u64 id; /* root objectid */
146 const char *name_stem; /* lock name stem */
147 char names[BTRFS_MAX_LEVEL + 1][20];
148 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
149} btrfs_lockdep_keysets[] = {
150 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
151 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
152 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
153 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
154 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
155 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
60b62978 156 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
85d4e461
CM		 157 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
158 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
159 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
13fd8da9 160 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
85d4e461 161 { .id = 0, .name_stem = "tree" },
4008c04a 162};
85d4e461
CM		 163
164void __init btrfs_init_lockdep(void)
165{
166 int i, j;
167
168 /* initialize lockdep class names */
169 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
170 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
171
172 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
173 snprintf(ks->names[j], sizeof(ks->names[j]),
174 "btrfs-%s-%02d", ks->name_stem, j);
175 }
176}
177
178void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
179 int level)
180{
181 struct btrfs_lockdep_keyset *ks;
182
183 BUG_ON(level >= ARRAY_SIZE(ks->keys));
184
185 /* find the matching keyset, id 0 is the default entry */
186 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
187 if (ks->id == objectid)
188 break;
189
190 lockdep_set_class_and_name(&eb->lock,
191 &ks->keys[level], ks->names[level]);
192}
193
4008c04a
CM		 194#endif
195
d352ac68
CM		 196/*
197 * extents on the btree inode are pretty simple, there's one extent
198 * that covers the entire device
199 */
b2950863 200static struct extent_map *btree_get_extent(struct inode *inode,
306e16ce 201 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 202 int create)
7eccb903 203{
5f39d397
CM		 204 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
205 struct extent_map *em;
206 int ret;
207
890871be 208 read_lock(&em_tree->lock);
d1310b2e 209 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 210 if (em) {
211 em->bdev =
212 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 213 read_unlock(&em_tree->lock);
5f39d397 214 goto out;
a061fc8d 215 }
890871be 216 read_unlock(&em_tree->lock);
7b13b7b1 217
172ddd60 218 em = alloc_extent_map();
5f39d397
CM		 219 if (!em) {
220 em = ERR_PTR(-ENOMEM);
221 goto out;
222 }
223 em->start = 0;
0afbaf8c 224 em->len = (u64)-1;
c8b97818 225 em->block_len = (u64)-1;
5f39d397 226 em->block_start = 0;
a061fc8d 227 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 228
890871be 229 write_lock(&em_tree->lock);
09a2a8f9 230 ret = add_extent_mapping(em_tree, em, 0);
5f39d397
CM		 231 if (ret == -EEXIST) {
232 free_extent_map(em);
7b13b7b1 233 em = lookup_extent_mapping(em_tree, start, len);
b4f359ab 234 if (!em)
0433f20d 235 em = ERR_PTR(-EIO);
5f39d397 236 } else if (ret) {
7b13b7b1 237 free_extent_map(em);
0433f20d 238 em = ERR_PTR(ret);
5f39d397 239 }
890871be 240 write_unlock(&em_tree->lock);
7b13b7b1 241
5f39d397
CM		 242out:
243 return em;
7eccb903
CM		 244}
245
b0496686 246u32 btrfs_csum_data(char *data, u32 seed, size_t len)
19c00ddc 247{
0b947aff 248 return btrfs_crc32c(seed, data, len);
19c00ddc
CM		 249}
250
251void btrfs_csum_final(u32 crc, char *result)
252{
7e75bf3f 253 put_unaligned_le32(~crc, result);
19c00ddc
CM		 254}
255
d352ac68
CM		 256/*
257 * compute the csum for a btree block, and either verify it or write it
258 * into the csum field of the block.
259 */
19c00ddc
CM		 260static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
261 int verify)
262{
6c41761f 263 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
607d432d 264 char *result = NULL;
19c00ddc
CM		 265 unsigned long len;
266 unsigned long cur_len;
267 unsigned long offset = BTRFS_CSUM_SIZE;
19c00ddc
CM		 268 char *kaddr;
269 unsigned long map_start;
270 unsigned long map_len;
271 int err;
272 u32 crc = ~(u32)0;
607d432d 273 unsigned long inline_result;
19c00ddc
CM		 274
275 len = buf->len - offset;
d397712b 276 while (len > 0) {
19c00ddc 277 err = map_private_extent_buffer(buf, offset, 32,
a6591715 278 &kaddr, &map_start, &map_len);
d397712b 279 if (err)
19c00ddc 280 return 1;
19c00ddc 281 cur_len = min(len, map_len - (offset - map_start));
b0496686 282 crc = btrfs_csum_data(kaddr + offset - map_start,
19c00ddc
CM		 283 crc, cur_len);
284 len -= cur_len;
285 offset += cur_len;
19c00ddc 286 }
607d432d
JB		 287 if (csum_size > sizeof(inline_result)) {
288 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
289 if (!result)
290 return 1;
291 } else {
292 result = (char *)&inline_result;
293 }
294
19c00ddc
CM		 295 btrfs_csum_final(crc, result);
296
297 if (verify) {
607d432d 298 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 299 u32 val;
300 u32 found = 0;
607d432d 301 memcpy(&found, result, csum_size);
e4204ded 302
607d432d 303 read_extent_buffer(buf, &val, 0, csum_size);
efe120a0
FH		 304 printk_ratelimited(KERN_INFO
305 "BTRFS: %s checksum verify failed on %llu wanted %X found %X "
306 "level %d\n",
307 root->fs_info->sb->s_id, buf->start,
308 val, found, btrfs_header_level(buf));
607d432d
JB		 309 if (result != (char *)&inline_result)
310 kfree(result);
19c00ddc
CM		 311 return 1;
312 }
313 } else {
607d432d 314 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 315 }
607d432d
JB		 316 if (result != (char *)&inline_result)
317 kfree(result);
19c00ddc
CM		 318 return 0;
319}
320
d352ac68
CM		 321/*
322 * we can't consider a given block up to date unless the transid of the
323 * block matches the transid in the parent node's pointer. This is how we
324 * detect blocks that either didn't get written at all or got written
325 * in the wrong place.
326 */
1259ab75 327static int verify_parent_transid(struct extent_io_tree *io_tree,
b9fab919
CM		 328 struct extent_buffer *eb, u64 parent_transid,
329 int atomic)
1259ab75 330{
2ac55d41 331 struct extent_state *cached_state = NULL;
1259ab75 332 int ret;
a26e8c9f
JB		 333 bool need_lock = (current->journal_info ==
334 (void *)BTRFS_SEND_TRANS_STUB);
1259ab75
CM		 335
336 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
337 return 0;
338
b9fab919
CM		 339 if (atomic)
340 return -EAGAIN;
341
a26e8c9f
JB		 342 if (need_lock) {
343 btrfs_tree_read_lock(eb);
344 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
345 }
346
2ac55d41 347 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
d0082371 348 0, &cached_state);
0b32f4bb 349 if (extent_buffer_uptodate(eb) &&
1259ab75
CM		 350 btrfs_header_generation(eb) == parent_transid) {
351 ret = 0;
352 goto out;
353 }
7a36ddec 354 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
193f284d 355 "found %llu\n",
c1c9ff7c 356 eb->start, parent_transid, btrfs_header_generation(eb));
1259ab75 357 ret = 1;
a26e8c9f
JB		 358
359 /*
360 * Things reading via commit roots that don't have normal protection,
361 * like send, can have a really old block in cache that may point at a
362 * block that has been free'd and re-allocated. So don't clear uptodate
363 * if we find an eb that is under IO (dirty/writeback) because we could
364 * end up reading in the stale data and then writing it back out and
365 * making everybody very sad.
366 */
367 if (!extent_buffer_under_io(eb))
368 clear_extent_buffer_uptodate(eb);
33958dc6 369out:
2ac55d41
JB		 370 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
371 &cached_state, GFP_NOFS);
a26e8c9f 372 btrfs_tree_read_unlock_blocking(eb);
1259ab75 373 return ret;
1259ab75
CM		 374}
375
1104a885
DS		 376/*
377 * Return 0 if the superblock checksum type matches the checksum value of that
378 * algorithm. Pass the raw disk superblock data.
379 */
380static int btrfs_check_super_csum(char *raw_disk_sb)
381{
382 struct btrfs_super_block *disk_sb =
383 (struct btrfs_super_block *)raw_disk_sb;
384 u16 csum_type = btrfs_super_csum_type(disk_sb);
385 int ret = 0;
386
387 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
388 u32 crc = ~(u32)0;
389 const int csum_size = sizeof(crc);
390 char result[csum_size];
391
392 /*
393 * The super_block structure does not span the whole
394 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
395 * is filled with zeros and is included in the checkum.
396 */
397 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
398 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
399 btrfs_csum_final(crc, result);
400
401 if (memcmp(raw_disk_sb, result, csum_size))
402 ret = 1;
667e7d94
CM		 403
404 if (ret && btrfs_super_generation(disk_sb) < 10) {
efe120a0
FH		 405 printk(KERN_WARNING
406 "BTRFS: super block crcs don't match, older mkfs detected\n");
667e7d94
CM		 407 ret = 0;
408 }
1104a885
DS		 409 }
410
411 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
efe120a0 412 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
1104a885
DS		 413 csum_type);
414 ret = 1;
415 }
416
417 return ret;
418}
419
d352ac68
CM		 420/*
421 * helper to read a given tree block, doing retries as required when
422 * the checksums don't match and we have alternate mirrors to try.
423 */
f188591e
CM		 424static int btree_read_extent_buffer_pages(struct btrfs_root *root,
425 struct extent_buffer *eb,
ca7a79ad 426 u64 start, u64 parent_transid)
f188591e
CM		 427{
428 struct extent_io_tree *io_tree;
ea466794 429 int failed = 0;
f188591e
CM		 430 int ret;
431 int num_copies = 0;
432 int mirror_num = 0;
ea466794 433 int failed_mirror = 0;
f188591e 434
a826d6dc 435 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
f188591e
CM		 436 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
437 while (1) {
bb82ab88
AJ		 438 ret = read_extent_buffer_pages(io_tree, eb, start,
439 WAIT_COMPLETE,
f188591e 440 btree_get_extent, mirror_num);
256dd1bb
SB		 441 if (!ret) {
442 if (!verify_parent_transid(io_tree, eb,
b9fab919 443 parent_transid, 0))
256dd1bb
SB		 444 break;
445 else
446 ret = -EIO;
447 }
d397712b 448
a826d6dc
JB		 449 /*
450 * This buffer's crc is fine, but its contents are corrupted, so
451 * there is no reason to read the other copies, they won't be
452 * any less wrong.
453 */
454 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
ea466794
JB		 455 break;
456
5d964051 457 num_copies = btrfs_num_copies(root->fs_info,
f188591e 458 eb->start, eb->len);
4235298e 459 if (num_copies == 1)
ea466794 460 break;
4235298e 461
5cf1ab56
JB		 462 if (!failed_mirror) {
463 failed = 1;
464 failed_mirror = eb->read_mirror;
465 }
466
f188591e 467 mirror_num++;
ea466794
JB		 468 if (mirror_num == failed_mirror)
469 mirror_num++;
470
4235298e 471 if (mirror_num > num_copies)
ea466794 472 break;
f188591e 473 }
ea466794 474
c0901581 475 if (failed && !ret && failed_mirror)
ea466794
JB		 476 repair_eb_io_failure(root, eb, failed_mirror);
477
478 return ret;
f188591e 479}
19c00ddc 480
d352ac68 481/*
d397712b
CM		 482 * checksum a dirty tree block before IO. This has extra checks to make sure
483 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 484 */
d397712b 485
b2950863 486static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
19c00ddc 487{
4eee4fa4 488 u64 start = page_offset(page);
19c00ddc 489 u64 found_start;
19c00ddc 490 struct extent_buffer *eb;
f188591e 491
4f2de97a
JB		 492 eb = (struct extent_buffer *)page->private;
493 if (page != eb->pages[0])
494 return 0;
19c00ddc 495 found_start = btrfs_header_bytenr(eb);
fae7f21c 496 if (WARN_ON(found_start != start || !PageUptodate(page)))
4f2de97a 497 return 0;
19c00ddc 498 csum_tree_block(root, eb, 0);
19c00ddc
CM		 499 return 0;
500}
501
2b82032c
YZ		 502static int check_tree_block_fsid(struct btrfs_root *root,
503 struct extent_buffer *eb)
504{
505 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
506 u8 fsid[BTRFS_UUID_SIZE];
507 int ret = 1;
508
0a4e5586 509 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
2b82032c
YZ		 510 while (fs_devices) {
511 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
512 ret = 0;
513 break;
514 }
515 fs_devices = fs_devices->seed;
516 }
517 return ret;
518}
519
a826d6dc 520#define CORRUPT(reason, eb, root, slot) \
efe120a0
FH		 521 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
522 "root=%llu, slot=%d", reason, \
c1c9ff7c 523 btrfs_header_bytenr(eb), root->objectid, slot)
a826d6dc
JB		 524
525static noinline int check_leaf(struct btrfs_root *root,
526 struct extent_buffer *leaf)
527{
528 struct btrfs_key key;
529 struct btrfs_key leaf_key;
530 u32 nritems = btrfs_header_nritems(leaf);
531 int slot;
532
533 if (nritems == 0)
534 return 0;
535
536 /* Check the 0 item */
537 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
538 BTRFS_LEAF_DATA_SIZE(root)) {
539 CORRUPT("invalid item offset size pair", leaf, root, 0);
540 return -EIO;
541 }
542
543 /*
544 * Check to make sure each items keys are in the correct order and their
545 * offsets make sense. We only have to loop through nritems-1 because
546 * we check the current slot against the next slot, which verifies the
547 * next slot's offset+size makes sense and that the current's slot
548 * offset is correct.
549 */
550 for (slot = 0; slot < nritems - 1; slot++) {
551 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
552 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
553
554 /* Make sure the keys are in the right order */
555 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
556 CORRUPT("bad key order", leaf, root, slot);
557 return -EIO;
558 }
559
560 /*
561 * Make sure the offset and ends are right, remember that the
562 * item data starts at the end of the leaf and grows towards the
563 * front.
564 */
565 if (btrfs_item_offset_nr(leaf, slot) !=
566 btrfs_item_end_nr(leaf, slot + 1)) {
567 CORRUPT("slot offset bad", leaf, root, slot);
568 return -EIO;
569 }
570
571 /*
572 * Check to make sure that we don't point outside of the leaf,
573 * just incase all the items are consistent to eachother, but
574 * all point outside of the leaf.
575 */
576 if (btrfs_item_end_nr(leaf, slot) >
577 BTRFS_LEAF_DATA_SIZE(root)) {
578 CORRUPT("slot end outside of leaf", leaf, root, slot);
579 return -EIO;
580 }
581 }
582
583 return 0;
584}
585
facc8a22
MX		 586static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
587 u64 phy_offset, struct page *page,
588 u64 start, u64 end, int mirror)
ce9adaa5 589{
ce9adaa5
CM		 590 u64 found_start;
591 int found_level;
ce9adaa5
CM		 592 struct extent_buffer *eb;
593 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 594 int ret = 0;
727011e0 595 int reads_done;
ce9adaa5 596
ce9adaa5
CM		 597 if (!page->private)
598 goto out;
d397712b 599
4f2de97a 600 eb = (struct extent_buffer *)page->private;
d397712b 601
0b32f4bb
JB		 602 /* the pending IO might have been the only thing that kept this buffer
603 * in memory. Make sure we have a ref for all this other checks
604 */
605 extent_buffer_get(eb);
606
607 reads_done = atomic_dec_and_test(&eb->io_pages);
727011e0
CM		 608 if (!reads_done)
609 goto err;
f188591e 610
5cf1ab56 611 eb->read_mirror = mirror;
ea466794
JB		 612 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
613 ret = -EIO;
614 goto err;
615 }
616
ce9adaa5 617 found_start = btrfs_header_bytenr(eb);
727011e0 618 if (found_start != eb->start) {
efe120a0 619 printk_ratelimited(KERN_INFO "BTRFS: bad tree block start "
193f284d 620 "%llu %llu\n",
c1c9ff7c 621 found_start, eb->start);
f188591e 622 ret = -EIO;
ce9adaa5
CM		 623 goto err;
624 }
2b82032c 625 if (check_tree_block_fsid(root, eb)) {
efe120a0 626 printk_ratelimited(KERN_INFO "BTRFS: bad fsid on block %llu\n",
c1c9ff7c 627 eb->start);
1259ab75
CM		 628 ret = -EIO;
629 goto err;
630 }
ce9adaa5 631 found_level = btrfs_header_level(eb);
1c24c3ce 632 if (found_level >= BTRFS_MAX_LEVEL) {
efe120a0 633 btrfs_info(root->fs_info, "bad tree block level %d",
1c24c3ce
JB		 634 (int)btrfs_header_level(eb));
635 ret = -EIO;
636 goto err;
637 }
ce9adaa5 638
85d4e461
CM		 639 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
640 eb, found_level);
4008c04a 641
ce9adaa5 642 ret = csum_tree_block(root, eb, 1);
a826d6dc 643 if (ret) {
f188591e 644 ret = -EIO;
a826d6dc
JB		 645 goto err;
646 }
647
648 /*
649 * If this is a leaf block and it is corrupt, set the corrupt bit so
650 * that we don't try and read the other copies of this block, just
651 * return -EIO.
652 */
653 if (found_level == 0 && check_leaf(root, eb)) {
654 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
655 ret = -EIO;
656 }
ce9adaa5 657
0b32f4bb
JB		 658 if (!ret)
659 set_extent_buffer_uptodate(eb);
ce9adaa5 660err:
79fb65a1
JB		 661 if (reads_done &&
662 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 663 btree_readahead_hook(root, eb, eb->start, ret);
4bb31e92 664
53b381b3
DW		 665 if (ret) {
666 /*
667 * our io error hook is going to dec the io pages
668 * again, we have to make sure it has something
669 * to decrement
670 */
671 atomic_inc(&eb->io_pages);
0b32f4bb 672 clear_extent_buffer_uptodate(eb);
53b381b3 673 }
0b32f4bb 674 free_extent_buffer(eb);
ce9adaa5 675out:
f188591e 676 return ret;
ce9adaa5
CM		 677}
678
ea466794 679static int btree_io_failed_hook(struct page *page, int failed_mirror)
4bb31e92 680{
4bb31e92
AJ		 681 struct extent_buffer *eb;
682 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
683
4f2de97a 684 eb = (struct extent_buffer *)page->private;
ea466794 685 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
5cf1ab56 686 eb->read_mirror = failed_mirror;
53b381b3 687 atomic_dec(&eb->io_pages);
ea466794 688 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 689 btree_readahead_hook(root, eb, eb->start, -EIO);
4bb31e92
AJ		 690 return -EIO; /* we fixed nothing */
691}
692
ce9adaa5 693static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM		 694{
695 struct end_io_wq *end_io_wq = bio->bi_private;
696 struct btrfs_fs_info *fs_info;
ce9adaa5 697
ce9adaa5 698 fs_info = end_io_wq->info;
ce9adaa5 699 end_io_wq->error = err;
fccb5d86 700 btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL);
d20f7043 701
7b6d91da 702 if (bio->bi_rw & REQ_WRITE) {
53b381b3 703 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA)
fccb5d86
QW		 704 btrfs_queue_work(fs_info->endio_meta_write_workers,
705 &end_io_wq->work);
53b381b3 706 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE)
fccb5d86
QW		 707 btrfs_queue_work(fs_info->endio_freespace_worker,
708 &end_io_wq->work);
53b381b3 709 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
fccb5d86
QW		 710 btrfs_queue_work(fs_info->endio_raid56_workers,
711 &end_io_wq->work);
cad321ad 712 else
fccb5d86
QW		 713 btrfs_queue_work(fs_info->endio_write_workers,
714 &end_io_wq->work);
d20f7043 715 } else {
53b381b3 716 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
fccb5d86
QW		 717 btrfs_queue_work(fs_info->endio_raid56_workers,
718 &end_io_wq->work);
53b381b3 719 else if (end_io_wq->metadata)
fccb5d86
QW		 720 btrfs_queue_work(fs_info->endio_meta_workers,
721 &end_io_wq->work);
d20f7043 722 else
fccb5d86
QW		 723 btrfs_queue_work(fs_info->endio_workers,
724 &end_io_wq->work);
d20f7043 725 }
ce9adaa5
CM		 726}
727
0cb59c99
JB		 728/*
729 * For the metadata arg you want
730 *
731 * 0 - if data
732 * 1 - if normal metadta
733 * 2 - if writing to the free space cache area
53b381b3 734 * 3 - raid parity work
0cb59c99 735 */
22c59948
CM		 736int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
737 int metadata)
0b86a832 738{
ce9adaa5 739 struct end_io_wq *end_io_wq;
ce9adaa5
CM		 740 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
741 if (!end_io_wq)
742 return -ENOMEM;
743
744 end_io_wq->private = bio->bi_private;
745 end_io_wq->end_io = bio->bi_end_io;
22c59948 746 end_io_wq->info = info;
ce9adaa5
CM		 747 end_io_wq->error = 0;
748 end_io_wq->bio = bio;
22c59948 749 end_io_wq->metadata = metadata;
ce9adaa5
CM		 750
751 bio->bi_private = end_io_wq;
752 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 753 return 0;
754}
755
b64a2851 756unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 757{
4854ddd0 758 unsigned long limit = min_t(unsigned long,
5cdc7ad3 759 info->thread_pool_size,
4854ddd0
CM		 760 info->fs_devices->open_devices);
761 return 256 * limit;
762}
0986fe9e 763
4a69a410
CM		 764static void run_one_async_start(struct btrfs_work *work)
765{
4a69a410 766 struct async_submit_bio *async;
79787eaa 767 int ret;
4a69a410
CM		 768
769 async = container_of(work, struct async_submit_bio, work);
79787eaa
JM		 770 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
771 async->mirror_num, async->bio_flags,
772 async->bio_offset);
773 if (ret)
774 async->error = ret;
4a69a410
CM		 775}
776
777static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 778{
779 struct btrfs_fs_info *fs_info;
780 struct async_submit_bio *async;
4854ddd0 781 int limit;
8b712842
CM		 782
783 async = container_of(work, struct async_submit_bio, work);
784 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 785
b64a2851 786 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 787 limit = limit * 2 / 3;
788
66657b31 789 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
b64a2851 790 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 791 wake_up(&fs_info->async_submit_wait);
792
79787eaa
JM		 793 /* If an error occured we just want to clean up the bio and move on */
794 if (async->error) {
795 bio_endio(async->bio, async->error);
796 return;
797 }
798
4a69a410 799 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 800 async->mirror_num, async->bio_flags,
801 async->bio_offset);
4a69a410
CM		 802}
803
804static void run_one_async_free(struct btrfs_work *work)
805{
806 struct async_submit_bio *async;
807
808 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 809 kfree(async);
810}
811
44b8bd7e
CM		 812int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
813 int rw, struct bio *bio, int mirror_num,
c8b97818 814 unsigned long bio_flags,
eaf25d93 815 u64 bio_offset,
4a69a410
CM		 816 extent_submit_bio_hook_t *submit_bio_start,
817 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 818{
819 struct async_submit_bio *async;
820
821 async = kmalloc(sizeof(*async), GFP_NOFS);
822 if (!async)
823 return -ENOMEM;
824
825 async->inode = inode;
826 async->rw = rw;
827 async->bio = bio;
828 async->mirror_num = mirror_num;
4a69a410
CM		 829 async->submit_bio_start = submit_bio_start;
830 async->submit_bio_done = submit_bio_done;
831
5cdc7ad3
QW		 832 btrfs_init_work(&async->work, run_one_async_start,
833 run_one_async_done, run_one_async_free);
4a69a410 834
c8b97818 835 async->bio_flags = bio_flags;
eaf25d93 836 async->bio_offset = bio_offset;
8c8bee1d 837
79787eaa
JM		 838 async->error = 0;
839
cb03c743 840 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 841
7b6d91da 842 if (rw & REQ_SYNC)
5cdc7ad3 843 btrfs_set_work_high_priority(&async->work);
d313d7a3 844
5cdc7ad3 845 btrfs_queue_work(fs_info->workers, &async->work);
9473f16c 846
d397712b 847 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 848 atomic_read(&fs_info->nr_async_submits)) {
849 wait_event(fs_info->async_submit_wait,
850 (atomic_read(&fs_info->nr_async_submits) == 0));
851 }
852
44b8bd7e
CM		 853 return 0;
854}
855
ce3ed71a
CM		 856static int btree_csum_one_bio(struct bio *bio)
857{
2c30c71b 858 struct bio_vec *bvec;
ce3ed71a 859 struct btrfs_root *root;
2c30c71b 860 int i, ret = 0;
ce3ed71a 861
2c30c71b 862 bio_for_each_segment_all(bvec, bio, i) {
ce3ed71a 863 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
79787eaa
JM		 864 ret = csum_dirty_buffer(root, bvec->bv_page);
865 if (ret)
866 break;
ce3ed71a 867 }
2c30c71b 868
79787eaa 869 return ret;
ce3ed71a
CM		 870}
871
4a69a410
CM		 872static int __btree_submit_bio_start(struct inode *inode, int rw,
873 struct bio *bio, int mirror_num,
eaf25d93
CM		 874 unsigned long bio_flags,
875 u64 bio_offset)
22c59948 876{
8b712842
CM		 877 /*
878 * when we're called for a write, we're already in the async
5443be45 879 * submission context. Just jump into btrfs_map_bio
8b712842 880 */
79787eaa 881 return btree_csum_one_bio(bio);
4a69a410 882}
22c59948 883
4a69a410 884static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 885 int mirror_num, unsigned long bio_flags,
886 u64 bio_offset)
4a69a410 887{
61891923
SB		 888 int ret;
889
8b712842 890 /*
4a69a410
CM		 891 * when we're called for a write, we're already in the async
892 * submission context. Just jump into btrfs_map_bio
8b712842 893 */
61891923
SB		 894 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
895 if (ret)
896 bio_endio(bio, ret);
897 return ret;
0b86a832
CM		 898}
899
de0022b9
JB		 900static int check_async_write(struct inode *inode, unsigned long bio_flags)
901{
902 if (bio_flags & EXTENT_BIO_TREE_LOG)
903 return 0;
904#ifdef CONFIG_X86
905 if (cpu_has_xmm4_2)
906 return 0;
907#endif
908 return 1;
909}
910
44b8bd7e 911static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 912 int mirror_num, unsigned long bio_flags,
913 u64 bio_offset)
44b8bd7e 914{
de0022b9 915 int async = check_async_write(inode, bio_flags);
cad321ad
CM		 916 int ret;
917
7b6d91da 918 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 919 /*
920 * called for a read, do the setup so that checksum validation
921 * can happen in the async kernel threads
922 */
f3f266ab
CM		 923 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
924 bio, 1);
1d4284bd 925 if (ret)
61891923
SB		 926 goto out_w_error;
927 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
928 mirror_num, 0);
de0022b9
JB		 929 } else if (!async) {
930 ret = btree_csum_one_bio(bio);
931 if (ret)
61891923
SB		 932 goto out_w_error;
933 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
934 mirror_num, 0);
935 } else {
936 /*
937 * kthread helpers are used to submit writes so that
938 * checksumming can happen in parallel across all CPUs
939 */
940 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
941 inode, rw, bio, mirror_num, 0,
942 bio_offset,
943 __btree_submit_bio_start,
944 __btree_submit_bio_done);
44b8bd7e 945 }
d313d7a3 946
61891923
SB		 947 if (ret) {
948out_w_error:
949 bio_endio(bio, ret);
950 }
951 return ret;
44b8bd7e
CM		 952}
953
3dd1462e 954#ifdef CONFIG_MIGRATION
784b4e29 955static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG		 956 struct page *newpage, struct page *page,
957 enum migrate_mode mode)
784b4e29
CM		 958{
959 /*
960 * we can't safely write a btree page from here,
961 * we haven't done the locking hook
962 */
963 if (PageDirty(page))
964 return -EAGAIN;
965 /*
966 * Buffers may be managed in a filesystem specific way.
967 * We must have no buffers or drop them.
968 */
969 if (page_has_private(page) &&
970 !try_to_release_page(page, GFP_KERNEL))
971 return -EAGAIN;
a6bc32b8 972 return migrate_page(mapping, newpage, page, mode);
784b4e29 973}
3dd1462e 974#endif
784b4e29 975
0da5468f
CM		 976
977static int btree_writepages(struct address_space *mapping,
978 struct writeback_control *wbc)
979{
e2d84521
MX		 980 struct btrfs_fs_info *fs_info;
981 int ret;
982
d8d5f3e1 983 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM		 984
985 if (wbc->for_kupdate)
986 return 0;
987
e2d84521 988 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 989 /* this is a bit racy, but that's ok */
e2d84521
MX		 990 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
991 BTRFS_DIRTY_METADATA_THRESH);
992 if (ret < 0)
793955bc 993 return 0;
793955bc 994 }
0b32f4bb 995 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM		 996}
997
b2950863 998static int btree_readpage(struct file *file, struct page *page)
5f39d397 999{
d1310b2e
CM		 1000 struct extent_io_tree *tree;
1001 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 1002 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 1003}
22b0ebda 1004
70dec807 1005static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 1006{
98509cfc 1007 if (PageWriteback(page) || PageDirty(page))
d397712b 1008 return 0;
0c4e538b 1009
f7a52a40 1010 return try_release_extent_buffer(page);
d98237b3
CM		 1011}
1012
d47992f8
LC		 1013static void btree_invalidatepage(struct page *page, unsigned int offset,
1014 unsigned int length)
d98237b3 1015{
d1310b2e
CM		 1016 struct extent_io_tree *tree;
1017 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 1018 extent_invalidatepage(tree, page, offset);
1019 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 1020 if (PagePrivate(page)) {
efe120a0
FH		 1021 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1022 "page private not zero on page %llu",
1023 (unsigned long long)page_offset(page));
9ad6b7bc
CM		 1024 ClearPagePrivate(page);
1025 set_page_private(page, 0);
1026 page_cache_release(page);
1027 }
d98237b3
CM		 1028}
1029
0b32f4bb
JB		 1030static int btree_set_page_dirty(struct page *page)
1031{
bb146eb2 1032#ifdef DEBUG
0b32f4bb
JB		 1033 struct extent_buffer *eb;
1034
1035 BUG_ON(!PagePrivate(page));
1036 eb = (struct extent_buffer *)page->private;
1037 BUG_ON(!eb);
1038 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1039 BUG_ON(!atomic_read(&eb->refs));
1040 btrfs_assert_tree_locked(eb);
bb146eb2 1041#endif
0b32f4bb
JB		 1042 return __set_page_dirty_nobuffers(page);
1043}
1044
7f09410b 1045static const struct address_space_operations btree_aops = {
d98237b3 1046 .readpage = btree_readpage,
0da5468f 1047 .writepages = btree_writepages,
5f39d397
CM		 1048 .releasepage = btree_releasepage,
1049 .invalidatepage = btree_invalidatepage,
5a92bc88 1050#ifdef CONFIG_MIGRATION
784b4e29 1051 .migratepage = btree_migratepage,
5a92bc88 1052#endif
0b32f4bb 1053 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM		 1054};
1055
ca7a79ad
CM		 1056int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1057 u64 parent_transid)
090d1875 1058{
5f39d397
CM		 1059 struct extent_buffer *buf = NULL;
1060 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 1061 int ret = 0;
090d1875 1062
db94535d 1063 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 1064 if (!buf)
090d1875 1065 return 0;
d1310b2e 1066 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
bb82ab88 1067 buf, 0, WAIT_NONE, btree_get_extent, 0);
5f39d397 1068 free_extent_buffer(buf);
de428b63 1069 return ret;
090d1875
CM		 1070}
1071
ab0fff03
AJ		 1072int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1073 int mirror_num, struct extent_buffer **eb)
1074{
1075 struct extent_buffer *buf = NULL;
1076 struct inode *btree_inode = root->fs_info->btree_inode;
1077 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1078 int ret;
1079
1080 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1081 if (!buf)
1082 return 0;
1083
1084 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1085
1086 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1087 btree_get_extent, mirror_num);
1088 if (ret) {
1089 free_extent_buffer(buf);
1090 return ret;
1091 }
1092
1093 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1094 free_extent_buffer(buf);
1095 return -EIO;
0b32f4bb 1096 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ		 1097 *eb = buf;
1098 } else {
1099 free_extent_buffer(buf);
1100 }
1101 return 0;
1102}
1103
0999df54
CM		 1104struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1105 u64 bytenr, u32 blocksize)
1106{
f28491e0 1107 return find_extent_buffer(root->fs_info, bytenr);
0999df54
CM		 1108}
1109
1110struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1111 u64 bytenr, u32 blocksize)
1112{
f28491e0 1113 return alloc_extent_buffer(root->fs_info, bytenr, blocksize);
0999df54
CM		 1114}
1115
1116
e02119d5
CM		 1117int btrfs_write_tree_block(struct extent_buffer *buf)
1118{
727011e0 1119 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1120 buf->start + buf->len - 1);
e02119d5
CM		 1121}
1122
1123int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1124{
727011e0 1125 return filemap_fdatawait_range(buf->pages[0]->mapping,
8aa38c31 1126 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1127}
1128
0999df54 1129struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 1130 u32 blocksize, u64 parent_transid)
0999df54
CM		 1131{
1132 struct extent_buffer *buf = NULL;
0999df54
CM		 1133 int ret;
1134
0999df54
CM		 1135 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1136 if (!buf)
1137 return NULL;
0999df54 1138
ca7a79ad 1139 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
0f0fe8f7
FDBM		 1140 if (ret) {
1141 free_extent_buffer(buf);
1142 return NULL;
1143 }
5f39d397 1144 return buf;
ce9adaa5 1145
eb60ceac
CM		 1146}
1147
d5c13f92
JM		 1148void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1149 struct extent_buffer *buf)
ed2ff2cb 1150{
e2d84521
MX		 1151 struct btrfs_fs_info *fs_info = root->fs_info;
1152
55c69072 1153 if (btrfs_header_generation(buf) ==
e2d84521 1154 fs_info->running_transaction->transid) {
b9447ef8 1155 btrfs_assert_tree_locked(buf);
b4ce94de 1156
b9473439 1157 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
e2d84521
MX		 1158 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1159 -buf->len,
1160 fs_info->dirty_metadata_batch);
ed7b63eb
JB		 1161 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1162 btrfs_set_lock_blocking(buf);
1163 clear_extent_buffer_dirty(buf);
1164 }
925baedd 1165 }
5f39d397
CM		 1166}
1167
8257b2dc
MX		 1168static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1169{
1170 struct btrfs_subvolume_writers *writers;
1171 int ret;
1172
1173 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1174 if (!writers)
1175 return ERR_PTR(-ENOMEM);
1176
1177 ret = percpu_counter_init(&writers->counter, 0);
1178 if (ret < 0) {
1179 kfree(writers);
1180 return ERR_PTR(ret);
1181 }
1182
1183 init_waitqueue_head(&writers->wait);
1184 return writers;
1185}
1186
1187static void
1188btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1189{
1190 percpu_counter_destroy(&writers->counter);
1191 kfree(writers);
1192}
1193
143bede5
JM		 1194static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1195 u32 stripesize, struct btrfs_root *root,
1196 struct btrfs_fs_info *fs_info,
1197 u64 objectid)
d97e63b6 1198{
cfaa7295 1199 root->node = NULL;
a28ec197 1200 root->commit_root = NULL;
db94535d
CM		 1201 root->sectorsize = sectorsize;
1202 root->nodesize = nodesize;
1203 root->leafsize = leafsize;
87ee04eb 1204 root->stripesize = stripesize;
27cdeb70 1205 root->state = 0;
d68fc57b 1206 root->orphan_cleanup_state = 0;
0b86a832 1207
0f7d52f4
CM		 1208 root->objectid = objectid;
1209 root->last_trans = 0;
13a8a7c8 1210 root->highest_objectid = 0;
eb73c1b7 1211 root->nr_delalloc_inodes = 0;
199c2a9c 1212 root->nr_ordered_extents = 0;
58176a96 1213 root->name = NULL;
6bef4d31 1214 root->inode_tree = RB_ROOT;
16cdcec7 1215 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1216 root->block_rsv = NULL;
d68fc57b 1217 root->orphan_block_rsv = NULL;
0b86a832
CM		 1218
1219 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1220 INIT_LIST_HEAD(&root->root_list);
eb73c1b7
MX		 1221 INIT_LIST_HEAD(&root->delalloc_inodes);
1222 INIT_LIST_HEAD(&root->delalloc_root);
199c2a9c
MX		 1223 INIT_LIST_HEAD(&root->ordered_extents);
1224 INIT_LIST_HEAD(&root->ordered_root);
2ab28f32
JB		 1225 INIT_LIST_HEAD(&root->logged_list[0]);
1226 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1227 spin_lock_init(&root->orphan_lock);
5d4f98a2 1228 spin_lock_init(&root->inode_lock);
eb73c1b7 1229 spin_lock_init(&root->delalloc_lock);
199c2a9c 1230 spin_lock_init(&root->ordered_extent_lock);
f0486c68 1231 spin_lock_init(&root->accounting_lock);
2ab28f32
JB		 1232 spin_lock_init(&root->log_extents_lock[0]);
1233 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1234 mutex_init(&root->objectid_mutex);
e02119d5 1235 mutex_init(&root->log_mutex);
31f3d255 1236 mutex_init(&root->ordered_extent_mutex);
573bfb72 1237 mutex_init(&root->delalloc_mutex);
7237f183
YZ		 1238 init_waitqueue_head(&root->log_writer_wait);
1239 init_waitqueue_head(&root->log_commit_wait[0]);
1240 init_waitqueue_head(&root->log_commit_wait[1]);
8b050d35
MX		 1241 INIT_LIST_HEAD(&root->log_ctxs[0]);
1242 INIT_LIST_HEAD(&root->log_ctxs[1]);
7237f183
YZ		 1243 atomic_set(&root->log_commit[0], 0);
1244 atomic_set(&root->log_commit[1], 0);
1245 atomic_set(&root->log_writers, 0);
2ecb7923 1246 atomic_set(&root->log_batch, 0);
8a35d95f 1247 atomic_set(&root->orphan_inodes, 0);
b0feb9d9 1248 atomic_set(&root->refs, 1);
8257b2dc 1249 atomic_set(&root->will_be_snapshoted, 0);
7237f183 1250 root->log_transid = 0;
d1433deb 1251 root->log_transid_committed = -1;
257c62e1 1252 root->last_log_commit = 0;
06ea65a3
JB		 1253 if (fs_info)
1254 extent_io_tree_init(&root->dirty_log_pages,
1255 fs_info->btree_inode->i_mapping);
017e5369 1256
3768f368
CM		 1257 memset(&root->root_key, 0, sizeof(root->root_key));
1258 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1259 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1260 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
06ea65a3
JB		 1261 if (fs_info)
1262 root->defrag_trans_start = fs_info->generation;
1263 else
1264 root->defrag_trans_start = 0;
58176a96 1265 init_completion(&root->kobj_unregister);
4d775673 1266 root->root_key.objectid = objectid;
0ee5dc67 1267 root->anon_dev = 0;
8ea05e3a 1268
5f3ab90a 1269 spin_lock_init(&root->root_item_lock);
3768f368
CM		 1270}
1271
f84a8bd6 1272static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
6f07e42e
AV		 1273{
1274 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1275 if (root)
1276 root->fs_info = fs_info;
1277 return root;
1278}
1279
06ea65a3
JB		 1280#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1281/* Should only be used by the testing infrastructure */
1282struct btrfs_root *btrfs_alloc_dummy_root(void)
1283{
1284 struct btrfs_root *root;
1285
1286 root = btrfs_alloc_root(NULL);
1287 if (!root)
1288 return ERR_PTR(-ENOMEM);
1289 __setup_root(4096, 4096, 4096, 4096, root, NULL, 1);
27cdeb70 1290 set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state);
06ea65a3
JB		 1291
1292 return root;
1293}
1294#endif
1295
20897f5c
AJ		 1296struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1297 struct btrfs_fs_info *fs_info,
1298 u64 objectid)
1299{
1300 struct extent_buffer *leaf;
1301 struct btrfs_root *tree_root = fs_info->tree_root;
1302 struct btrfs_root *root;
1303 struct btrfs_key key;
1304 int ret = 0;
6463fe58 1305 uuid_le uuid;
20897f5c
AJ		 1306
1307 root = btrfs_alloc_root(fs_info);
1308 if (!root)
1309 return ERR_PTR(-ENOMEM);
1310
1311 __setup_root(tree_root->nodesize, tree_root->leafsize,
1312 tree_root->sectorsize, tree_root->stripesize,
1313 root, fs_info, objectid);
1314 root->root_key.objectid = objectid;
1315 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1316 root->root_key.offset = 0;
1317
1318 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1319 0, objectid, NULL, 0, 0, 0);
1320 if (IS_ERR(leaf)) {
1321 ret = PTR_ERR(leaf);
1dd05682 1322 leaf = NULL;
20897f5c
AJ		 1323 goto fail;
1324 }
1325
20897f5c
AJ		 1326 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1327 btrfs_set_header_bytenr(leaf, leaf->start);
1328 btrfs_set_header_generation(leaf, trans->transid);
1329 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1330 btrfs_set_header_owner(leaf, objectid);
1331 root->node = leaf;
1332
0a4e5586 1333 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
20897f5c
AJ		 1334 BTRFS_FSID_SIZE);
1335 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
b308bc2f 1336 btrfs_header_chunk_tree_uuid(leaf),
20897f5c
AJ		 1337 BTRFS_UUID_SIZE);
1338 btrfs_mark_buffer_dirty(leaf);
1339
1340 root->commit_root = btrfs_root_node(root);
27cdeb70 1341 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
20897f5c
AJ		 1342
1343 root->root_item.flags = 0;
1344 root->root_item.byte_limit = 0;
1345 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1346 btrfs_set_root_generation(&root->root_item, trans->transid);
1347 btrfs_set_root_level(&root->root_item, 0);
1348 btrfs_set_root_refs(&root->root_item, 1);
1349 btrfs_set_root_used(&root->root_item, leaf->len);
1350 btrfs_set_root_last_snapshot(&root->root_item, 0);
1351 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB		 1352 uuid_le_gen(&uuid);
1353 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ		 1354 root->root_item.drop_level = 0;
1355
1356 key.objectid = objectid;
1357 key.type = BTRFS_ROOT_ITEM_KEY;
1358 key.offset = 0;
1359 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1360 if (ret)
1361 goto fail;
1362
1363 btrfs_tree_unlock(leaf);
1364
1dd05682
TI		 1365 return root;
1366
20897f5c 1367fail:
1dd05682
TI		 1368 if (leaf) {
1369 btrfs_tree_unlock(leaf);
59885b39 1370 free_extent_buffer(root->commit_root);
1dd05682
TI		 1371 free_extent_buffer(leaf);
1372 }
1373 kfree(root);
20897f5c 1374
1dd05682 1375 return ERR_PTR(ret);
20897f5c
AJ		 1376}
1377
7237f183
YZ		 1378static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1379 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1380{
1381 struct btrfs_root *root;
1382 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1383 struct extent_buffer *leaf;
e02119d5 1384
6f07e42e 1385 root = btrfs_alloc_root(fs_info);
e02119d5 1386 if (!root)
7237f183 1387 return ERR_PTR(-ENOMEM);
e02119d5
CM		 1388
1389 __setup_root(tree_root->nodesize, tree_root->leafsize,
1390 tree_root->sectorsize, tree_root->stripesize,
1391 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1392
1393 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1394 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1395 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
27cdeb70 1396
7237f183 1397 /*
27cdeb70
MX		 1398 * DON'T set REF_COWS for log trees
1399 *
7237f183
YZ		 1400 * log trees do not get reference counted because they go away
1401 * before a real commit is actually done. They do store pointers
1402 * to file data extents, and those reference counts still get
1403 * updated (along with back refs to the log tree).
1404 */
e02119d5 1405
5d4f98a2 1406 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
66d7e7f0 1407 BTRFS_TREE_LOG_OBJECTID, NULL,
5581a51a 1408 0, 0, 0);
7237f183
YZ		 1409 if (IS_ERR(leaf)) {
1410 kfree(root);
1411 return ERR_CAST(leaf);
1412 }
e02119d5 1413
5d4f98a2
YZ		 1414 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1415 btrfs_set_header_bytenr(leaf, leaf->start);
1416 btrfs_set_header_generation(leaf, trans->transid);
1417 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1418 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1419 root->node = leaf;
e02119d5
CM		 1420
1421 write_extent_buffer(root->node, root->fs_info->fsid,
0a4e5586 1422 btrfs_header_fsid(), BTRFS_FSID_SIZE);
e02119d5
CM		 1423 btrfs_mark_buffer_dirty(root->node);
1424 btrfs_tree_unlock(root->node);
7237f183
YZ		 1425 return root;
1426}
1427
1428int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1429 struct btrfs_fs_info *fs_info)
1430{
1431 struct btrfs_root *log_root;
1432
1433 log_root = alloc_log_tree(trans, fs_info);
1434 if (IS_ERR(log_root))
1435 return PTR_ERR(log_root);
1436 WARN_ON(fs_info->log_root_tree);
1437 fs_info->log_root_tree = log_root;
1438 return 0;
1439}
1440
1441int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1442 struct btrfs_root *root)
1443{
1444 struct btrfs_root *log_root;
1445 struct btrfs_inode_item *inode_item;
1446
1447 log_root = alloc_log_tree(trans, root->fs_info);
1448 if (IS_ERR(log_root))
1449 return PTR_ERR(log_root);
1450
1451 log_root->last_trans = trans->transid;
1452 log_root->root_key.offset = root->root_key.objectid;
1453
1454 inode_item = &log_root->root_item.inode;
3cae210f
QW		 1455 btrfs_set_stack_inode_generation(inode_item, 1);
1456 btrfs_set_stack_inode_size(inode_item, 3);
1457 btrfs_set_stack_inode_nlink(inode_item, 1);
1458 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
1459 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
7237f183 1460
5d4f98a2 1461 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1462
1463 WARN_ON(root->log_root);
1464 root->log_root = log_root;
1465 root->log_transid = 0;
d1433deb 1466 root->log_transid_committed = -1;
257c62e1 1467 root->last_log_commit = 0;
e02119d5
CM		 1468 return 0;
1469}
1470
35a3621b
SB		 1471static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1472 struct btrfs_key *key)
e02119d5
CM		 1473{
1474 struct btrfs_root *root;
1475 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1476 struct btrfs_path *path;
84234f3a 1477 u64 generation;
db94535d 1478 u32 blocksize;
cb517eab 1479 int ret;
0f7d52f4 1480
cb517eab
MX		 1481 path = btrfs_alloc_path();
1482 if (!path)
0f7d52f4 1483 return ERR_PTR(-ENOMEM);
cb517eab
MX		 1484
1485 root = btrfs_alloc_root(fs_info);
1486 if (!root) {
1487 ret = -ENOMEM;
1488 goto alloc_fail;
0f7d52f4
CM		 1489 }
1490
db94535d 1491 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb 1492 tree_root->sectorsize, tree_root->stripesize,
cb517eab 1493 root, fs_info, key->objectid);
0f7d52f4 1494
cb517eab
MX		 1495 ret = btrfs_find_root(tree_root, key, path,
1496 &root->root_item, &root->root_key);
0f7d52f4 1497 if (ret) {
13a8a7c8
YZ		 1498 if (ret > 0)
1499 ret = -ENOENT;
cb517eab 1500 goto find_fail;
0f7d52f4 1501 }
13a8a7c8 1502
84234f3a 1503 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1504 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1505 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1506 blocksize, generation);
cb517eab
MX		 1507 if (!root->node) {
1508 ret = -ENOMEM;
1509 goto find_fail;
1510 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1511 ret = -EIO;
1512 goto read_fail;
416bc658 1513 }
5d4f98a2 1514 root->commit_root = btrfs_root_node(root);
13a8a7c8 1515out:
cb517eab
MX		 1516 btrfs_free_path(path);
1517 return root;
1518
1519read_fail:
1520 free_extent_buffer(root->node);
1521find_fail:
1522 kfree(root);
1523alloc_fail:
1524 root = ERR_PTR(ret);
1525 goto out;
1526}
1527
1528struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1529 struct btrfs_key *location)
1530{
1531 struct btrfs_root *root;
1532
1533 root = btrfs_read_tree_root(tree_root, location);
1534 if (IS_ERR(root))
1535 return root;
1536
1537 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
27cdeb70 1538 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
08fe4db1
LZ		 1539 btrfs_check_and_init_root_item(&root->root_item);
1540 }
13a8a7c8 1541
5eda7b5e
CM		 1542 return root;
1543}
1544
cb517eab
MX		 1545int btrfs_init_fs_root(struct btrfs_root *root)
1546{
1547 int ret;
8257b2dc 1548 struct btrfs_subvolume_writers *writers;
cb517eab
MX		 1549
1550 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1551 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1552 GFP_NOFS);
1553 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1554 ret = -ENOMEM;
1555 goto fail;
1556 }
1557
8257b2dc
MX		 1558 writers = btrfs_alloc_subvolume_writers();
1559 if (IS_ERR(writers)) {
1560 ret = PTR_ERR(writers);
1561 goto fail;
1562 }
1563 root->subv_writers = writers;
1564
cb517eab 1565 btrfs_init_free_ino_ctl(root);
cb517eab
MX		 1566 spin_lock_init(&root->cache_lock);
1567 init_waitqueue_head(&root->cache_wait);
1568
1569 ret = get_anon_bdev(&root->anon_dev);
1570 if (ret)
8257b2dc 1571 goto free_writers;
cb517eab 1572 return 0;
8257b2dc
MX		 1573
1574free_writers:
1575 btrfs_free_subvolume_writers(root->subv_writers);
cb517eab
MX		 1576fail:
1577 kfree(root->free_ino_ctl);
1578 kfree(root->free_ino_pinned);
1579 return ret;
1580}
1581
171170c1
ST		 1582static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1583 u64 root_id)
cb517eab
MX		 1584{
1585 struct btrfs_root *root;
1586
1587 spin_lock(&fs_info->fs_roots_radix_lock);
1588 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1589 (unsigned long)root_id);
1590 spin_unlock(&fs_info->fs_roots_radix_lock);
1591 return root;
1592}
1593
1594int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1595 struct btrfs_root *root)
1596{
1597 int ret;
1598
1599 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1600 if (ret)
1601 return ret;
1602
1603 spin_lock(&fs_info->fs_roots_radix_lock);
1604 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1605 (unsigned long)root->root_key.objectid,
1606 root);
1607 if (ret == 0)
27cdeb70 1608 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
cb517eab
MX		 1609 spin_unlock(&fs_info->fs_roots_radix_lock);
1610 radix_tree_preload_end();
1611
1612 return ret;
1613}
1614
c00869f1
MX		 1615struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1616 struct btrfs_key *location,
1617 bool check_ref)
5eda7b5e
CM		 1618{
1619 struct btrfs_root *root;
1620 int ret;
1621
edbd8d4e
CM		 1622 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1623 return fs_info->tree_root;
1624 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1625 return fs_info->extent_root;
8f18cf13
CM		 1626 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1627 return fs_info->chunk_root;
1628 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1629 return fs_info->dev_root;
0403e47e
YZ		 1630 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1631 return fs_info->csum_root;
bcef60f2
AJ		 1632 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1633 return fs_info->quota_root ? fs_info->quota_root :
1634 ERR_PTR(-ENOENT);
f7a81ea4
SB		 1635 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1636 return fs_info->uuid_root ? fs_info->uuid_root :
1637 ERR_PTR(-ENOENT);
4df27c4d 1638again:
cb517eab 1639 root = btrfs_lookup_fs_root(fs_info, location->objectid);
48475471 1640 if (root) {
c00869f1 1641 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
48475471 1642 return ERR_PTR(-ENOENT);
5eda7b5e 1643 return root;
48475471 1644 }
5eda7b5e 1645
cb517eab 1646 root = btrfs_read_fs_root(fs_info->tree_root, location);
5eda7b5e
CM		 1647 if (IS_ERR(root))
1648 return root;
3394e160 1649
c00869f1 1650 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
cb517eab 1651 ret = -ENOENT;
581bb050 1652 goto fail;
35a30d7c 1653 }
581bb050 1654
cb517eab 1655 ret = btrfs_init_fs_root(root);
ac08aedf
CM		 1656 if (ret)
1657 goto fail;
3394e160 1658
3f870c28
KN		 1659 ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID,
1660 location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL);
d68fc57b
YZ		 1661 if (ret < 0)
1662 goto fail;
1663 if (ret == 0)
27cdeb70 1664 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
d68fc57b 1665
cb517eab 1666 ret = btrfs_insert_fs_root(fs_info, root);
0f7d52f4 1667 if (ret) {
4df27c4d
YZ		 1668 if (ret == -EEXIST) {
1669 free_fs_root(root);
1670 goto again;
1671 }
1672 goto fail;
0f7d52f4 1673 }
edbd8d4e 1674 return root;
4df27c4d
YZ		 1675fail:
1676 free_fs_root(root);
1677 return ERR_PTR(ret);
edbd8d4e
CM		 1678}
1679
04160088
CM		 1680static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1681{
1682 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1683 int ret = 0;
04160088
CM		 1684 struct btrfs_device *device;
1685 struct backing_dev_info *bdi;
b7967db7 1686
1f78160c
XG		 1687 rcu_read_lock();
1688 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1689 if (!device->bdev)
1690 continue;
04160088
CM		 1691 bdi = blk_get_backing_dev_info(device->bdev);
1692 if (bdi && bdi_congested(bdi, bdi_bits)) {
1693 ret = 1;
1694 break;
1695 }
1696 }
1f78160c 1697 rcu_read_unlock();
04160088
CM		 1698 return ret;
1699}
1700
ad081f14
JA		 1701/*
1702 * If this fails, caller must call bdi_destroy() to get rid of the
1703 * bdi again.
1704 */
04160088
CM		 1705static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1706{
ad081f14
JA		 1707 int err;
1708
1709 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1710 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1711 if (err)
1712 return err;
1713
4575c9cc 1714 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1715 bdi->congested_fn = btrfs_congested_fn;
1716 bdi->congested_data = info;
1717 return 0;
1718}
1719
8b712842
CM		 1720/*
1721 * called by the kthread helper functions to finally call the bio end_io
1722 * functions. This is where read checksum verification actually happens
1723 */
1724static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1725{
ce9adaa5 1726 struct bio *bio;
8b712842 1727 struct end_io_wq *end_io_wq;
ce9adaa5 1728 int error;
ce9adaa5 1729
8b712842
CM		 1730 end_io_wq = container_of(work, struct end_io_wq, work);
1731 bio = end_io_wq->bio;
ce9adaa5 1732
8b712842
CM		 1733 error = end_io_wq->error;
1734 bio->bi_private = end_io_wq->private;
1735 bio->bi_end_io = end_io_wq->end_io;
1736 kfree(end_io_wq);
bc1e79ac 1737 bio_endio_nodec(bio, error);
44b8bd7e
CM		 1738}
1739
a74a4b97
CM		 1740static int cleaner_kthread(void *arg)
1741{
1742 struct btrfs_root *root = arg;
d0278245 1743 int again;
a74a4b97
CM		 1744
1745 do {
d0278245 1746 again = 0;
a74a4b97 1747
d0278245 1748 /* Make the cleaner go to sleep early. */
babbf170 1749 if (btrfs_need_cleaner_sleep(root))
d0278245
MX		 1750 goto sleep;
1751
1752 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1753 goto sleep;
1754
dc7f370c
MX		 1755 /*
1756 * Avoid the problem that we change the status of the fs
1757 * during the above check and trylock.
1758 */
babbf170 1759 if (btrfs_need_cleaner_sleep(root)) {
dc7f370c
MX		 1760 mutex_unlock(&root->fs_info->cleaner_mutex);
1761 goto sleep;
76dda93c 1762 }
a74a4b97 1763
d0278245
MX		 1764 btrfs_run_delayed_iputs(root);
1765 again = btrfs_clean_one_deleted_snapshot(root);
1766 mutex_unlock(&root->fs_info->cleaner_mutex);
1767
1768 /*
05323cd1
MX		 1769 * The defragger has dealt with the R/O remount and umount,
1770 * needn't do anything special here.
d0278245
MX		 1771 */
1772 btrfs_run_defrag_inodes(root->fs_info);
1773sleep:
9d1a2a3a 1774 if (!try_to_freeze() && !again) {
a74a4b97 1775 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1776 if (!kthread_should_stop())
1777 schedule();
a74a4b97
CM		 1778 __set_current_state(TASK_RUNNING);
1779 }
1780 } while (!kthread_should_stop());
1781 return 0;
1782}
1783
1784static int transaction_kthread(void *arg)
1785{
1786 struct btrfs_root *root = arg;
1787 struct btrfs_trans_handle *trans;
1788 struct btrfs_transaction *cur;
8929ecfa 1789 u64 transid;
a74a4b97
CM		 1790 unsigned long now;
1791 unsigned long delay;
914b2007 1792 bool cannot_commit;
a74a4b97
CM		 1793
1794 do {
914b2007 1795 cannot_commit = false;
8b87dc17 1796 delay = HZ * root->fs_info->commit_interval;
a74a4b97
CM		 1797 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1798
a4abeea4 1799 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1800 cur = root->fs_info->running_transaction;
1801 if (!cur) {
a4abeea4 1802 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1803 goto sleep;
1804 }
31153d81 1805
a74a4b97 1806 now = get_seconds();
4a9d8bde 1807 if (cur->state < TRANS_STATE_BLOCKED &&
8b87dc17
DS		 1808 (now < cur->start_time ||
1809 now - cur->start_time < root->fs_info->commit_interval)) {
a4abeea4 1810 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1811 delay = HZ * 5;
1812 goto sleep;
1813 }
8929ecfa 1814 transid = cur->transid;
a4abeea4 1815 spin_unlock(&root->fs_info->trans_lock);
56bec294 1816
79787eaa 1817 /* If the file system is aborted, this will always fail. */
354aa0fb 1818 trans = btrfs_attach_transaction(root);
914b2007 1819 if (IS_ERR(trans)) {
354aa0fb
MX		 1820 if (PTR_ERR(trans) != -ENOENT)
1821 cannot_commit = true;
79787eaa 1822 goto sleep;
914b2007 1823 }
8929ecfa 1824 if (transid == trans->transid) {
79787eaa 1825 btrfs_commit_transaction(trans, root);
8929ecfa
YZ		 1826 } else {
1827 btrfs_end_transaction(trans, root);
1828 }
a74a4b97
CM		 1829sleep:
1830 wake_up_process(root->fs_info->cleaner_kthread);
1831 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1832
4e121c06
JB		 1833 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1834 &root->fs_info->fs_state)))
1835 btrfs_cleanup_transaction(root);
a0acae0e 1836 if (!try_to_freeze()) {
a74a4b97 1837 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa 1838 if (!kthread_should_stop() &&
914b2007
JK		 1839 (!btrfs_transaction_blocked(root->fs_info) ||
1840 cannot_commit))
8929ecfa 1841 schedule_timeout(delay);
a74a4b97
CM		 1842 __set_current_state(TASK_RUNNING);
1843 }
1844 } while (!kthread_should_stop());
1845 return 0;
1846}
1847
af31f5e5
CM		 1848/*
1849 * this will find the highest generation in the array of
1850 * root backups. The index of the highest array is returned,
1851 * or -1 if we can't find anything.
1852 *
1853 * We check to make sure the array is valid by comparing the
1854 * generation of the latest root in the array with the generation
1855 * in the super block. If they don't match we pitch it.
1856 */
1857static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1858{
1859 u64 cur;
1860 int newest_index = -1;
1861 struct btrfs_root_backup *root_backup;
1862 int i;
1863
1864 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1865 root_backup = info->super_copy->super_roots + i;
1866 cur = btrfs_backup_tree_root_gen(root_backup);
1867 if (cur == newest_gen)
1868 newest_index = i;
1869 }
1870
1871 /* check to see if we actually wrapped around */
1872 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1873 root_backup = info->super_copy->super_roots;
1874 cur = btrfs_backup_tree_root_gen(root_backup);
1875 if (cur == newest_gen)
1876 newest_index = 0;
1877 }
1878 return newest_index;
1879}
1880
1881
1882/*
1883 * find the oldest backup so we know where to store new entries
1884 * in the backup array. This will set the backup_root_index
1885 * field in the fs_info struct
1886 */
1887static void find_oldest_super_backup(struct btrfs_fs_info *info,
1888 u64 newest_gen)
1889{
1890 int newest_index = -1;
1891
1892 newest_index = find_newest_super_backup(info, newest_gen);
1893 /* if there was garbage in there, just move along */
1894 if (newest_index == -1) {
1895 info->backup_root_index = 0;
1896 } else {
1897 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1898 }
1899}
1900
1901/*
1902 * copy all the root pointers into the super backup array.
1903 * this will bump the backup pointer by one when it is
1904 * done
1905 */
1906static void backup_super_roots(struct btrfs_fs_info *info)
1907{
1908 int next_backup;
1909 struct btrfs_root_backup *root_backup;
1910 int last_backup;
1911
1912 next_backup = info->backup_root_index;
1913 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1914 BTRFS_NUM_BACKUP_ROOTS;
1915
1916 /*
1917 * just overwrite the last backup if we're at the same generation
1918 * this happens only at umount
1919 */
1920 root_backup = info->super_for_commit->super_roots + last_backup;
1921 if (btrfs_backup_tree_root_gen(root_backup) ==
1922 btrfs_header_generation(info->tree_root->node))
1923 next_backup = last_backup;
1924
1925 root_backup = info->super_for_commit->super_roots + next_backup;
1926
1927 /*
1928 * make sure all of our padding and empty slots get zero filled
1929 * regardless of which ones we use today
1930 */
1931 memset(root_backup, 0, sizeof(*root_backup));
1932
1933 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1934
1935 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1936 btrfs_set_backup_tree_root_gen(root_backup,
1937 btrfs_header_generation(info->tree_root->node));
1938
1939 btrfs_set_backup_tree_root_level(root_backup,
1940 btrfs_header_level(info->tree_root->node));
1941
1942 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1943 btrfs_set_backup_chunk_root_gen(root_backup,
1944 btrfs_header_generation(info->chunk_root->node));
1945 btrfs_set_backup_chunk_root_level(root_backup,
1946 btrfs_header_level(info->chunk_root->node));
1947
1948 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1949 btrfs_set_backup_extent_root_gen(root_backup,
1950 btrfs_header_generation(info->extent_root->node));
1951 btrfs_set_backup_extent_root_level(root_backup,
1952 btrfs_header_level(info->extent_root->node));
1953
7c7e82a7
CM		 1954 /*
1955 * we might commit during log recovery, which happens before we set
1956 * the fs_root. Make sure it is valid before we fill it in.
1957 */
1958 if (info->fs_root && info->fs_root->node) {
1959 btrfs_set_backup_fs_root(root_backup,
1960 info->fs_root->node->start);
1961 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 1962 btrfs_header_generation(info->fs_root->node));
7c7e82a7 1963 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 1964 btrfs_header_level(info->fs_root->node));
7c7e82a7 1965 }
af31f5e5
CM		 1966
1967 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1968 btrfs_set_backup_dev_root_gen(root_backup,
1969 btrfs_header_generation(info->dev_root->node));
1970 btrfs_set_backup_dev_root_level(root_backup,
1971 btrfs_header_level(info->dev_root->node));
1972
1973 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1974 btrfs_set_backup_csum_root_gen(root_backup,
1975 btrfs_header_generation(info->csum_root->node));
1976 btrfs_set_backup_csum_root_level(root_backup,
1977 btrfs_header_level(info->csum_root->node));
1978
1979 btrfs_set_backup_total_bytes(root_backup,
1980 btrfs_super_total_bytes(info->super_copy));
1981 btrfs_set_backup_bytes_used(root_backup,
1982 btrfs_super_bytes_used(info->super_copy));
1983 btrfs_set_backup_num_devices(root_backup,
1984 btrfs_super_num_devices(info->super_copy));
1985
1986 /*
1987 * if we don't copy this out to the super_copy, it won't get remembered
1988 * for the next commit
1989 */
1990 memcpy(&info->super_copy->super_roots,
1991 &info->super_for_commit->super_roots,
1992 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1993}
1994
1995/*
1996 * this copies info out of the root backup array and back into
1997 * the in-memory super block. It is meant to help iterate through
1998 * the array, so you send it the number of backups you've already
1999 * tried and the last backup index you used.
2000 *
2001 * this returns -1 when it has tried all the backups
2002 */
2003static noinline int next_root_backup(struct btrfs_fs_info *info,
2004 struct btrfs_super_block *super,
2005 int *num_backups_tried, int *backup_index)
2006{
2007 struct btrfs_root_backup *root_backup;
2008 int newest = *backup_index;
2009
2010 if (*num_backups_tried == 0) {
2011 u64 gen = btrfs_super_generation(super);
2012
2013 newest = find_newest_super_backup(info, gen);
2014 if (newest == -1)
2015 return -1;
2016
2017 *backup_index = newest;
2018 *num_backups_tried = 1;
2019 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2020 /* we've tried all the backups, all done */
2021 return -1;
2022 } else {
2023 /* jump to the next oldest backup */
2024 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2025 BTRFS_NUM_BACKUP_ROOTS;
2026 *backup_index = newest;
2027 *num_backups_tried += 1;
2028 }
2029 root_backup = super->super_roots + newest;
2030
2031 btrfs_set_super_generation(super,
2032 btrfs_backup_tree_root_gen(root_backup));
2033 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2034 btrfs_set_super_root_level(super,
2035 btrfs_backup_tree_root_level(root_backup));
2036 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2037
2038 /*
2039 * fixme: the total bytes and num_devices need to match or we should
2040 * need a fsck
2041 */
2042 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2043 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2044 return 0;
2045}
2046
7abadb64
LB		 2047/* helper to cleanup workers */
2048static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2049{
dc6e3209 2050 btrfs_destroy_workqueue(fs_info->fixup_workers);
afe3d242 2051 btrfs_destroy_workqueue(fs_info->delalloc_workers);
5cdc7ad3 2052 btrfs_destroy_workqueue(fs_info->workers);
fccb5d86
QW		 2053 btrfs_destroy_workqueue(fs_info->endio_workers);
2054 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2055 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
d05a33ac 2056 btrfs_destroy_workqueue(fs_info->rmw_workers);
fccb5d86
QW		 2057 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2058 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2059 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
a8c93d4e 2060 btrfs_destroy_workqueue(fs_info->submit_workers);
5b3bc44e 2061 btrfs_destroy_workqueue(fs_info->delayed_workers);
e66f0bb1 2062 btrfs_destroy_workqueue(fs_info->caching_workers);
736cfa15 2063 btrfs_destroy_workqueue(fs_info->readahead_workers);
a44903ab 2064 btrfs_destroy_workqueue(fs_info->flush_workers);
fc97fab0 2065 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
7abadb64
LB		 2066}
2067
2e9f5954
R		 2068static void free_root_extent_buffers(struct btrfs_root *root)
2069{
2070 if (root) {
2071 free_extent_buffer(root->node);
2072 free_extent_buffer(root->commit_root);
2073 root->node = NULL;
2074 root->commit_root = NULL;
2075 }
2076}
2077
af31f5e5
CM		 2078/* helper to cleanup tree roots */
2079static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2080{
2e9f5954 2081 free_root_extent_buffers(info->tree_root);
655b09fe 2082
2e9f5954
R		 2083 free_root_extent_buffers(info->dev_root);
2084 free_root_extent_buffers(info->extent_root);
2085 free_root_extent_buffers(info->csum_root);
2086 free_root_extent_buffers(info->quota_root);
2087 free_root_extent_buffers(info->uuid_root);
2088 if (chunk_root)
2089 free_root_extent_buffers(info->chunk_root);
af31f5e5
CM		 2090}
2091
171f6537
JB		 2092static void del_fs_roots(struct btrfs_fs_info *fs_info)
2093{
2094 int ret;
2095 struct btrfs_root *gang[8];
2096 int i;
2097
2098 while (!list_empty(&fs_info->dead_roots)) {
2099 gang[0] = list_entry(fs_info->dead_roots.next,
2100 struct btrfs_root, root_list);
2101 list_del(&gang[0]->root_list);
2102
27cdeb70 2103 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
cb517eab 2104 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
171f6537
JB		 2105 } else {
2106 free_extent_buffer(gang[0]->node);
2107 free_extent_buffer(gang[0]->commit_root);
b0feb9d9 2108 btrfs_put_fs_root(gang[0]);
171f6537
JB		 2109 }
2110 }
2111
2112 while (1) {
2113 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2114 (void **)gang, 0,
2115 ARRAY_SIZE(gang));
2116 if (!ret)
2117 break;
2118 for (i = 0; i < ret; i++)
cb517eab 2119 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
171f6537 2120 }
1a4319cc
LB		 2121
2122 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2123 btrfs_free_log_root_tree(NULL, fs_info);
2124 btrfs_destroy_pinned_extent(fs_info->tree_root,
2125 fs_info->pinned_extents);
2126 }
171f6537 2127}
af31f5e5 2128
ad2b2c80
AV		 2129int open_ctree(struct super_block *sb,
2130 struct btrfs_fs_devices *fs_devices,
2131 char *options)
2e635a27 2132{
db94535d
CM		 2133 u32 sectorsize;
2134 u32 nodesize;
2135 u32 leafsize;
2136 u32 blocksize;
87ee04eb 2137 u32 stripesize;
84234f3a 2138 u64 generation;
f2b636e8 2139 u64 features;
3de4586c 2140 struct btrfs_key location;
a061fc8d 2141 struct buffer_head *bh;
4d34b278 2142 struct btrfs_super_block *disk_super;
815745cf 2143 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2144 struct btrfs_root *tree_root;
4d34b278
ID		 2145 struct btrfs_root *extent_root;
2146 struct btrfs_root *csum_root;
2147 struct btrfs_root *chunk_root;
2148 struct btrfs_root *dev_root;
bcef60f2 2149 struct btrfs_root *quota_root;
f7a81ea4 2150 struct btrfs_root *uuid_root;
e02119d5 2151 struct btrfs_root *log_tree_root;
eb60ceac 2152 int ret;
e58ca020 2153 int err = -EINVAL;
af31f5e5
CM		 2154 int num_backups_tried = 0;
2155 int backup_index = 0;
5cdc7ad3
QW		 2156 int max_active;
2157 int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
70f80175
SB		 2158 bool create_uuid_tree;
2159 bool check_uuid_tree;
4543df7e 2160
f84a8bd6 2161 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
6f07e42e 2162 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
cb517eab 2163 if (!tree_root || !chunk_root) {
39279cc3
CM		 2164 err = -ENOMEM;
2165 goto fail;
2166 }
76dda93c
YZ		 2167
2168 ret = init_srcu_struct(&fs_info->subvol_srcu);
2169 if (ret) {
2170 err = ret;
2171 goto fail;
2172 }
2173
2174 ret = setup_bdi(fs_info, &fs_info->bdi);
2175 if (ret) {
2176 err = ret;
2177 goto fail_srcu;
2178 }
2179
e2d84521
MX		 2180 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2181 if (ret) {
2182 err = ret;
2183 goto fail_bdi;
2184 }
2185 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2186 (1 + ilog2(nr_cpu_ids));
2187
963d678b
MX		 2188 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2189 if (ret) {
2190 err = ret;
2191 goto fail_dirty_metadata_bytes;
2192 }
2193
c404e0dc
MX		 2194 ret = percpu_counter_init(&fs_info->bio_counter, 0);
2195 if (ret) {
2196 err = ret;
2197 goto fail_delalloc_bytes;
2198 }
2199
76dda93c
YZ		 2200 fs_info->btree_inode = new_inode(sb);
2201 if (!fs_info->btree_inode) {
2202 err = -ENOMEM;
c404e0dc 2203 goto fail_bio_counter;
76dda93c
YZ		 2204 }
2205
a6591715 2206 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2207
76dda93c 2208 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
f28491e0 2209 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
8fd17795 2210 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2211 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2212 INIT_LIST_HEAD(&fs_info->delayed_iputs);
eb73c1b7 2213 INIT_LIST_HEAD(&fs_info->delalloc_roots);
11833d66 2214 INIT_LIST_HEAD(&fs_info->caching_block_groups);
eb73c1b7 2215 spin_lock_init(&fs_info->delalloc_root_lock);
a4abeea4 2216 spin_lock_init(&fs_info->trans_lock);
76dda93c 2217 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2218 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2219 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 2220 spin_lock_init(&fs_info->free_chunk_lock);
f29021b2 2221 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2222 spin_lock_init(&fs_info->super_lock);
fcebe456 2223 spin_lock_init(&fs_info->qgroup_op_lock);
f28491e0 2224 spin_lock_init(&fs_info->buffer_lock);
f29021b2 2225 rwlock_init(&fs_info->tree_mod_log_lock);
7585717f 2226 mutex_init(&fs_info->reloc_mutex);
573bfb72 2227 mutex_init(&fs_info->delalloc_root_mutex);
de98ced9 2228 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2229
58176a96 2230 init_completion(&fs_info->kobj_unregister);
0b86a832 2231 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2232 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2233 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
0b86a832 2234 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX		 2235 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2236 BTRFS_BLOCK_RSV_GLOBAL);
2237 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2238 BTRFS_BLOCK_RSV_DELALLOC);
2239 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2240 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2241 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2242 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2243 BTRFS_BLOCK_RSV_DELOPS);
cb03c743 2244 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 2245 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 2246 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 2247 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 2248 atomic_set(&fs_info->defrag_running, 0);
fcebe456 2249 atomic_set(&fs_info->qgroup_op_seq, 0);
fc36ed7e 2250 atomic64_set(&fs_info->tree_mod_seq, 0);
e20d96d6 2251 fs_info->sb = sb;
6f568d35 2252 fs_info->max_inline = 8192 * 1024;
9ed74f2d 2253 fs_info->metadata_ratio = 0;
4cb5300b 2254 fs_info->defrag_inodes = RB_ROOT;
2bf64758 2255 fs_info->free_chunk_space = 0;
f29021b2 2256 fs_info->tree_mod_log = RB_ROOT;
8b87dc17 2257 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
0a2b2a84 2258 fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
90519d66
AJ		 2259 /* readahead state */
2260 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2261 spin_lock_init(&fs_info->reada_lock);
c8b97818 2262
b34b086c
CM		 2263 fs_info->thread_pool_size = min_t(unsigned long,
2264 num_online_cpus() + 2, 8);
0afbaf8c 2265
199c2a9c
MX		 2266 INIT_LIST_HEAD(&fs_info->ordered_roots);
2267 spin_lock_init(&fs_info->ordered_root_lock);
16cdcec7
MX		 2268 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2269 GFP_NOFS);
2270 if (!fs_info->delayed_root) {
2271 err = -ENOMEM;
2272 goto fail_iput;
2273 }
2274 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2275
a2de733c
AJ		 2276 mutex_init(&fs_info->scrub_lock);
2277 atomic_set(&fs_info->scrubs_running, 0);
2278 atomic_set(&fs_info->scrub_pause_req, 0);
2279 atomic_set(&fs_info->scrubs_paused, 0);
2280 atomic_set(&fs_info->scrub_cancel_req, 0);
c404e0dc 2281 init_waitqueue_head(&fs_info->replace_wait);
a2de733c 2282 init_waitqueue_head(&fs_info->scrub_pause_wait);
a2de733c 2283 fs_info->scrub_workers_refcnt = 0;
21adbd5c
SB		 2284#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2285 fs_info->check_integrity_print_mask = 0;
2286#endif
a2de733c 2287
c9e9f97b
ID		 2288 spin_lock_init(&fs_info->balance_lock);
2289 mutex_init(&fs_info->balance_mutex);
837d5b6e
ID		 2290 atomic_set(&fs_info->balance_running, 0);
2291 atomic_set(&fs_info->balance_pause_req, 0);
a7e99c69 2292 atomic_set(&fs_info->balance_cancel_req, 0);
c9e9f97b 2293 fs_info->balance_ctl = NULL;
837d5b6e 2294 init_waitqueue_head(&fs_info->balance_wait_q);
21c7e756 2295 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
a2de733c 2296
a061fc8d
CM		 2297 sb->s_blocksize = 4096;
2298 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 2299 sb->s_bdi = &fs_info->bdi;
a061fc8d 2300
76dda93c 2301 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
bfe86848 2302 set_nlink(fs_info->btree_inode, 1);
0afbaf8c
CM		 2303 /*
2304 * we set the i_size on the btree inode to the max possible int.
2305 * the real end of the address space is determined by all of
2306 * the devices in the system
2307 */
2308 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 2309 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 2310 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2311
5d4f98a2 2312 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 2313 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 2314 fs_info->btree_inode->i_mapping);
0b32f4bb 2315 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
a8067e02 2316 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM		 2317
2318 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 2319
76dda93c
YZ		 2320 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2321 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2322 sizeof(struct btrfs_key));
72ac3c0d
JB		 2323 set_bit(BTRFS_INODE_DUMMY,
2324 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
778ba82b 2325 btrfs_insert_inode_hash(fs_info->btree_inode);
76dda93c 2326
0f9dd46c 2327 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2328 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2329 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2330
11833d66 2331 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 2332 fs_info->btree_inode->i_mapping);
11833d66 2333 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 2334 fs_info->btree_inode->i_mapping);
11833d66 2335 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 2336 fs_info->do_barriers = 1;
e18e4809 2337
39279cc3 2338
5a3f23d5 2339 mutex_init(&fs_info->ordered_operations_mutex);
9ffba8cd 2340 mutex_init(&fs_info->ordered_extent_flush_mutex);
e02119d5 2341 mutex_init(&fs_info->tree_log_mutex);
925baedd 2342 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 2343 mutex_init(&fs_info->transaction_kthread_mutex);
2344 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2345 mutex_init(&fs_info->volume_mutex);
9e351cc8 2346 init_rwsem(&fs_info->commit_root_sem);
c71bf099 2347 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2348 init_rwsem(&fs_info->subvol_sem);
803b2f54 2349 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
e922e087
SB		 2350 fs_info->dev_replace.lock_owner = 0;
2351 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2352 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2353 mutex_init(&fs_info->dev_replace.lock_management_lock);
2354 mutex_init(&fs_info->dev_replace.lock);
fa9c0d79 2355
416ac51d 2356 spin_lock_init(&fs_info->qgroup_lock);
f2f6ed3d 2357 mutex_init(&fs_info->qgroup_ioctl_lock);
416ac51d 2358 fs_info->qgroup_tree = RB_ROOT;
fcebe456 2359 fs_info->qgroup_op_tree = RB_ROOT;
416ac51d
AJ		 2360 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2361 fs_info->qgroup_seq = 1;
2362 fs_info->quota_enabled = 0;
2363 fs_info->pending_quota_state = 0;
1e8f9158 2364 fs_info->qgroup_ulist = NULL;
2f232036 2365 mutex_init(&fs_info->qgroup_rescan_lock);
416ac51d 2366
fa9c0d79
CM		 2367 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2368 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2369
e6dcd2dc 2370 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2371 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2372 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2373 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2374
53b381b3
DW		 2375 ret = btrfs_alloc_stripe_hash_table(fs_info);
2376 if (ret) {
83c8266a 2377 err = ret;
53b381b3
DW		 2378 goto fail_alloc;
2379 }
2380
0b86a832 2381 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 2382 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2383
3c4bb26b 2384 invalidate_bdev(fs_devices->latest_bdev);
1104a885
DS		 2385
2386 /*
2387 * Read super block and check the signature bytes only
2388 */
a512bbf8 2389 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY		 2390 if (!bh) {
2391 err = -EINVAL;
16cdcec7 2392 goto fail_alloc;
20b45077 2393 }
39279cc3 2394
1104a885
DS		 2395 /*
2396 * We want to check superblock checksum, the type is stored inside.
2397 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2398 */
2399 if (btrfs_check_super_csum(bh->b_data)) {
efe120a0 2400 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
1104a885
DS		 2401 err = -EINVAL;
2402 goto fail_alloc;
2403 }
2404
2405 /*
2406 * super_copy is zeroed at allocation time and we never touch the
2407 * following bytes up to INFO_SIZE, the checksum is calculated from
2408 * the whole block of INFO_SIZE
2409 */
6c41761f
DS		 2410 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2411 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2412 sizeof(*fs_info->super_for_commit));
a061fc8d 2413 brelse(bh);
5f39d397 2414
6c41761f 2415 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2416
1104a885
DS		 2417 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2418 if (ret) {
efe120a0 2419 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
1104a885
DS		 2420 err = -EINVAL;
2421 goto fail_alloc;
2422 }
2423
6c41761f 2424 disk_super = fs_info->super_copy;
0f7d52f4 2425 if (!btrfs_super_root(disk_super))
16cdcec7 2426 goto fail_alloc;
0f7d52f4 2427
acce952b 2428 /* check FS state, whether FS is broken. */
87533c47
MX		 2429 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2430 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2431
af31f5e5
CM		 2432 /*
2433 * run through our array of backup supers and setup
2434 * our ring pointer to the oldest one
2435 */
2436 generation = btrfs_super_generation(disk_super);
2437 find_oldest_super_backup(fs_info, generation);
2438
75e7cb7f
LB		 2439 /*
2440 * In the long term, we'll store the compression type in the super
2441 * block, and it'll be used for per file compression control.
2442 */
2443 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2444
2b82032c
YZ		 2445 ret = btrfs_parse_options(tree_root, options);
2446 if (ret) {
2447 err = ret;
16cdcec7 2448 goto fail_alloc;
2b82032c 2449 }
dfe25020 2450
f2b636e8
JB		 2451 features = btrfs_super_incompat_flags(disk_super) &
2452 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2453 if (features) {
2454 printk(KERN_ERR "BTRFS: couldn't mount because of "
2455 "unsupported optional features (%Lx).\n",
c1c9ff7c 2456 features);
f2b636e8 2457 err = -EINVAL;
16cdcec7 2458 goto fail_alloc;
f2b636e8
JB		 2459 }
2460
727011e0
CM		 2461 if (btrfs_super_leafsize(disk_super) !=
2462 btrfs_super_nodesize(disk_super)) {
2463 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2464 "blocksizes don't match. node %d leaf %d\n",
2465 btrfs_super_nodesize(disk_super),
2466 btrfs_super_leafsize(disk_super));
2467 err = -EINVAL;
2468 goto fail_alloc;
2469 }
2470 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2471 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2472 "blocksize (%d) was too large\n",
2473 btrfs_super_leafsize(disk_super));
2474 err = -EINVAL;
2475 goto fail_alloc;
2476 }
2477
5d4f98a2 2478 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2479 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
69e380d1 2480 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2481 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
727011e0 2482
3173a18f 2483 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
efe120a0 2484 printk(KERN_ERR "BTRFS: has skinny extents\n");
3173a18f 2485
727011e0
CM		 2486 /*
2487 * flag our filesystem as having big metadata blocks if
2488 * they are bigger than the page size
2489 */
2490 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2491 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
efe120a0 2492 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
727011e0
CM		 2493 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2494 }
2495
bc3f116f
CM		 2496 nodesize = btrfs_super_nodesize(disk_super);
2497 leafsize = btrfs_super_leafsize(disk_super);
2498 sectorsize = btrfs_super_sectorsize(disk_super);
2499 stripesize = btrfs_super_stripesize(disk_super);
e2d84521 2500 fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids));
963d678b 2501 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f
CM		 2502
2503 /*
2504 * mixed block groups end up with duplicate but slightly offset
2505 * extent buffers for the same range. It leads to corruptions
2506 */
2507 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2508 (sectorsize != leafsize)) {
efe120a0 2509 printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes "
bc3f116f
CM		 2510 "are not allowed for mixed block groups on %s\n",
2511 sb->s_id);
2512 goto fail_alloc;
2513 }
2514
ceda0864
MX		 2515 /*
2516 * Needn't use the lock because there is no other task which will
2517 * update the flag.
2518 */
a6fa6fae 2519 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2520
f2b636e8
JB		 2521 features = btrfs_super_compat_ro_flags(disk_super) &
2522 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2523 if (!(sb->s_flags & MS_RDONLY) && features) {
2524 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2525 "unsupported option features (%Lx).\n",
c1c9ff7c 2526 features);
f2b636e8 2527 err = -EINVAL;
16cdcec7 2528 goto fail_alloc;
f2b636e8 2529 }
61d92c32 2530
5cdc7ad3 2531 max_active = fs_info->thread_pool_size;
61d92c32 2532
5cdc7ad3
QW		 2533 fs_info->workers =
2534 btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2535 max_active, 16);
c8b97818 2536
afe3d242
QW		 2537 fs_info->delalloc_workers =
2538 btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
771ed689 2539
a44903ab
QW		 2540 fs_info->flush_workers =
2541 btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
8ccf6f19 2542
e66f0bb1
QW		 2543 fs_info->caching_workers =
2544 btrfs_alloc_workqueue("cache", flags, max_active, 0);
61b49440 2545
a8c93d4e
QW		 2546 /*
2547 * a higher idle thresh on the submit workers makes it much more
61b49440
CM		 2548 * likely that bios will be send down in a sane order to the
2549 * devices
2550 */
a8c93d4e
QW		 2551 fs_info->submit_workers =
2552 btrfs_alloc_workqueue("submit", flags,
2553 min_t(u64, fs_devices->num_devices,
2554 max_active), 64);
53863232 2555
dc6e3209
QW		 2556 fs_info->fixup_workers =
2557 btrfs_alloc_workqueue("fixup", flags, 1, 0);
61b49440
CM		 2558
2559 /*
2560 * endios are largely parallel and should have a very
2561 * low idle thresh
2562 */
fccb5d86
QW		 2563 fs_info->endio_workers =
2564 btrfs_alloc_workqueue("endio", flags, max_active, 4);
2565 fs_info->endio_meta_workers =
2566 btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2567 fs_info->endio_meta_write_workers =
2568 btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2569 fs_info->endio_raid56_workers =
2570 btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
d05a33ac
QW		 2571 fs_info->rmw_workers =
2572 btrfs_alloc_workqueue("rmw", flags, max_active, 2);
fccb5d86
QW		 2573 fs_info->endio_write_workers =
2574 btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2575 fs_info->endio_freespace_worker =
2576 btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
5b3bc44e
QW		 2577 fs_info->delayed_workers =
2578 btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
736cfa15
QW		 2579 fs_info->readahead_workers =
2580 btrfs_alloc_workqueue("readahead", flags, max_active, 2);
fc97fab0
QW		 2581 fs_info->qgroup_rescan_workers =
2582 btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
61b49440 2583
a8c93d4e 2584 if (!(fs_info->workers && fs_info->delalloc_workers &&
fccb5d86
QW		 2585 fs_info->submit_workers && fs_info->flush_workers &&
2586 fs_info->endio_workers && fs_info->endio_meta_workers &&
2587 fs_info->endio_meta_write_workers &&
2588 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
e66f0bb1 2589 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
dc6e3209 2590 fs_info->caching_workers && fs_info->readahead_workers &&
fc97fab0
QW		 2591 fs_info->fixup_workers && fs_info->delayed_workers &&
2592 fs_info->qgroup_rescan_workers)) {
fed425c7 2593 err = -ENOMEM;
0dc3b84a
JB		 2594 goto fail_sb_buffer;
2595 }
4543df7e 2596
4575c9cc 2597 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 2598 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2599 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 2600
db94535d
CM		 2601 tree_root->nodesize = nodesize;
2602 tree_root->leafsize = leafsize;
2603 tree_root->sectorsize = sectorsize;
87ee04eb 2604 tree_root->stripesize = stripesize;
a061fc8d
CM		 2605
2606 sb->s_blocksize = sectorsize;
2607 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 2608
3cae210f 2609 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
efe120a0 2610 printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 2611 goto fail_sb_buffer;
2612 }
19c00ddc 2613
8d082fb7 2614 if (sectorsize != PAGE_SIZE) {
efe120a0 2615 printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) "
8d082fb7 2616 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
941b2ddf
KM		 2617 goto fail_sb_buffer;
2618 }
2619
925baedd 2620 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 2621 ret = btrfs_read_sys_array(tree_root);
925baedd 2622 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2623 if (ret) {
efe120a0 2624 printk(KERN_WARNING "BTRFS: failed to read the system "
d397712b 2625 "array on %s\n", sb->s_id);
5d4f98a2 2626 goto fail_sb_buffer;
84eed90f 2627 }
0b86a832
CM		 2628
2629 blocksize = btrfs_level_size(tree_root,
2630 btrfs_super_chunk_root_level(disk_super));
84234f3a 2631 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 2632
2633 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2634 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2635
2636 chunk_root->node = read_tree_block(chunk_root,
2637 btrfs_super_chunk_root(disk_super),
84234f3a 2638 blocksize, generation);
416bc658
JB		 2639 if (!chunk_root->node ||
2640 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
efe120a0 2641 printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n",
83121942 2642 sb->s_id);
af31f5e5 2643 goto fail_tree_roots;
83121942 2644 }
5d4f98a2
YZ		 2645 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2646 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2647
e17cade2 2648 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
b308bc2f 2649 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
e17cade2 2650
0b86a832 2651 ret = btrfs_read_chunk_tree(chunk_root);
2b82032c 2652 if (ret) {
efe120a0 2653 printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n",
d397712b 2654 sb->s_id);
af31f5e5 2655 goto fail_tree_roots;
2b82032c 2656 }
0b86a832 2657
8dabb742
SB		 2658 /*
2659 * keep the device that is marked to be the target device for the
2660 * dev_replace procedure
2661 */
2662 btrfs_close_extra_devices(fs_info, fs_devices, 0);
dfe25020 2663
a6b0d5c8 2664 if (!fs_devices->latest_bdev) {
efe120a0 2665 printk(KERN_CRIT "BTRFS: failed to read devices on %s\n",
a6b0d5c8
CM		 2666 sb->s_id);
2667 goto fail_tree_roots;
2668 }
2669
af31f5e5 2670retry_root_backup:
db94535d
CM		 2671 blocksize = btrfs_level_size(tree_root,
2672 btrfs_super_root_level(disk_super));
84234f3a 2673 generation = btrfs_super_generation(disk_super);
0b86a832 2674
e20d96d6 2675 tree_root->node = read_tree_block(tree_root,
db94535d 2676 btrfs_super_root(disk_super),
84234f3a 2677 blocksize, generation);
af31f5e5
CM		 2678 if (!tree_root->node ||
2679 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
efe120a0 2680 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
83121942 2681 sb->s_id);
af31f5e5
CM		 2682
2683 goto recovery_tree_root;
83121942 2684 }
af31f5e5 2685
5d4f98a2
YZ		 2686 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2687 tree_root->commit_root = btrfs_root_node(tree_root);
69e9c6c6 2688 btrfs_set_root_refs(&tree_root->root_item, 1);
db94535d 2689
cb517eab
MX		 2690 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2691 location.type = BTRFS_ROOT_ITEM_KEY;
2692 location.offset = 0;
2693
2694 extent_root = btrfs_read_tree_root(tree_root, &location);
2695 if (IS_ERR(extent_root)) {
2696 ret = PTR_ERR(extent_root);
af31f5e5 2697 goto recovery_tree_root;
cb517eab 2698 }
27cdeb70 2699 set_bit(BTRFS_ROOT_TRACK_DIRTY, &extent_root->state);
cb517eab 2700 fs_info->extent_root = extent_root;
0b86a832 2701
cb517eab
MX		 2702 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2703 dev_root = btrfs_read_tree_root(tree_root, &location);
2704 if (IS_ERR(dev_root)) {
2705 ret = PTR_ERR(dev_root);
af31f5e5 2706 goto recovery_tree_root;
cb517eab 2707 }
27cdeb70 2708 set_bit(BTRFS_ROOT_TRACK_DIRTY, &dev_root->state);
cb517eab
MX		 2709 fs_info->dev_root = dev_root;
2710 btrfs_init_devices_late(fs_info);
3768f368 2711
cb517eab
MX		 2712 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2713 csum_root = btrfs_read_tree_root(tree_root, &location);
2714 if (IS_ERR(csum_root)) {
2715 ret = PTR_ERR(csum_root);
af31f5e5 2716 goto recovery_tree_root;
cb517eab 2717 }
27cdeb70 2718 set_bit(BTRFS_ROOT_TRACK_DIRTY, &csum_root->state);
cb517eab 2719 fs_info->csum_root = csum_root;
d20f7043 2720
cb517eab
MX		 2721 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2722 quota_root = btrfs_read_tree_root(tree_root, &location);
2723 if (!IS_ERR(quota_root)) {
27cdeb70 2724 set_bit(BTRFS_ROOT_TRACK_DIRTY, &quota_root->state);
bcef60f2
AJ		 2725 fs_info->quota_enabled = 1;
2726 fs_info->pending_quota_state = 1;
cb517eab 2727 fs_info->quota_root = quota_root;
bcef60f2
AJ		 2728 }
2729
f7a81ea4
SB		 2730 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2731 uuid_root = btrfs_read_tree_root(tree_root, &location);
2732 if (IS_ERR(uuid_root)) {
2733 ret = PTR_ERR(uuid_root);
2734 if (ret != -ENOENT)
2735 goto recovery_tree_root;
2736 create_uuid_tree = true;
70f80175 2737 check_uuid_tree = false;
f7a81ea4 2738 } else {
27cdeb70 2739 set_bit(BTRFS_ROOT_TRACK_DIRTY, &uuid_root->state);
f7a81ea4 2740 fs_info->uuid_root = uuid_root;
70f80175
SB		 2741 create_uuid_tree = false;
2742 check_uuid_tree =
2743 generation != btrfs_super_uuid_tree_generation(disk_super);
f7a81ea4
SB		 2744 }
2745
8929ecfa
YZ		 2746 fs_info->generation = generation;
2747 fs_info->last_trans_committed = generation;
8929ecfa 2748
68310a5e
ID		 2749 ret = btrfs_recover_balance(fs_info);
2750 if (ret) {
efe120a0 2751 printk(KERN_WARNING "BTRFS: failed to recover balance\n");
68310a5e
ID		 2752 goto fail_block_groups;
2753 }
2754
733f4fbb
SB		 2755 ret = btrfs_init_dev_stats(fs_info);
2756 if (ret) {
efe120a0 2757 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
733f4fbb
SB		 2758 ret);
2759 goto fail_block_groups;
2760 }
2761
8dabb742
SB		 2762 ret = btrfs_init_dev_replace(fs_info);
2763 if (ret) {
efe120a0 2764 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
8dabb742
SB		 2765 goto fail_block_groups;
2766 }
2767
2768 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2769
5ac1d209 2770 ret = btrfs_sysfs_add_one(fs_info);
c59021f8 2771 if (ret) {
efe120a0 2772 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
c59021f8 2773 goto fail_block_groups;
2774 }
2775
c59021f8 2776 ret = btrfs_init_space_info(fs_info);
2777 if (ret) {
efe120a0 2778 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
2365dd3c 2779 goto fail_sysfs;
c59021f8 2780 }
2781
1b1d1f66
JB		 2782 ret = btrfs_read_block_groups(extent_root);
2783 if (ret) {
efe120a0 2784 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
2365dd3c 2785 goto fail_sysfs;
1b1d1f66 2786 }
5af3e8cc
SB		 2787 fs_info->num_tolerated_disk_barrier_failures =
2788 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
292fd7fc
SB		 2789 if (fs_info->fs_devices->missing_devices >
2790 fs_info->num_tolerated_disk_barrier_failures &&
2791 !(sb->s_flags & MS_RDONLY)) {
efe120a0
FH		 2792 printk(KERN_WARNING "BTRFS: "
2793 "too many missing devices, writeable mount is not allowed\n");
2365dd3c 2794 goto fail_sysfs;
292fd7fc 2795 }
9078a3e1 2796
a74a4b97
CM		 2797 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2798 "btrfs-cleaner");
57506d50 2799 if (IS_ERR(fs_info->cleaner_kthread))
2365dd3c 2800 goto fail_sysfs;
a74a4b97
CM		 2801
2802 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2803 tree_root,
2804 "btrfs-transaction");
57506d50 2805 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2806 goto fail_cleaner;
a74a4b97 2807
c289811c
CM		 2808 if (!btrfs_test_opt(tree_root, SSD) &&
2809 !btrfs_test_opt(tree_root, NOSSD) &&
2810 !fs_info->fs_devices->rotating) {
efe120a0 2811 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
c289811c
CM		 2812 "mode\n");
2813 btrfs_set_opt(fs_info->mount_opt, SSD);
2814 }
2815
3818aea2
QW		 2816 /* Set the real inode map cache flag */
2817 if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE))
2818 btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE);
2819
21adbd5c
SB		 2820#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2821 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2822 ret = btrfsic_mount(tree_root, fs_devices,
2823 btrfs_test_opt(tree_root,
2824 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2825 1 : 0,
2826 fs_info->check_integrity_print_mask);
2827 if (ret)
efe120a0 2828 printk(KERN_WARNING "BTRFS: failed to initialize"
21adbd5c
SB		 2829 " integrity check module %s\n", sb->s_id);
2830 }
2831#endif
bcef60f2
AJ		 2832 ret = btrfs_read_qgroup_config(fs_info);
2833 if (ret)
2834 goto fail_trans_kthread;
21adbd5c 2835
acce952b 2836 /* do not make disk changes in broken FS */
68ce9682 2837 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM		 2838 u64 bytenr = btrfs_super_log_root(disk_super);
2839
7c2ca468 2840 if (fs_devices->rw_devices == 0) {
efe120a0 2841 printk(KERN_WARNING "BTRFS: log replay required "
d397712b 2842 "on RO media\n");
7c2ca468 2843 err = -EIO;
bcef60f2 2844 goto fail_qgroup;
7c2ca468 2845 }
e02119d5
CM		 2846 blocksize =
2847 btrfs_level_size(tree_root,
2848 btrfs_super_log_root_level(disk_super));
d18a2c44 2849
6f07e42e 2850 log_tree_root = btrfs_alloc_root(fs_info);
676e4c86
DC		 2851 if (!log_tree_root) {
2852 err = -ENOMEM;
bcef60f2 2853 goto fail_qgroup;
676e4c86 2854 }
e02119d5
CM		 2855
2856 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2857 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2858
2859 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 2860 blocksize,
2861 generation + 1);
416bc658
JB		 2862 if (!log_tree_root->node ||
2863 !extent_buffer_uptodate(log_tree_root->node)) {
efe120a0 2864 printk(KERN_ERR "BTRFS: failed to read log tree\n");
416bc658
JB		 2865 free_extent_buffer(log_tree_root->node);
2866 kfree(log_tree_root);
28c16cbb 2867 goto fail_qgroup;
416bc658 2868 }
79787eaa 2869 /* returns with log_tree_root freed on success */
e02119d5 2870 ret = btrfs_recover_log_trees(log_tree_root);
79787eaa
JM		 2871 if (ret) {
2872 btrfs_error(tree_root->fs_info, ret,
2873 "Failed to recover log tree");
2874 free_extent_buffer(log_tree_root->node);
2875 kfree(log_tree_root);
28c16cbb 2876 goto fail_qgroup;
79787eaa 2877 }
e556ce2c
YZ		 2878
2879 if (sb->s_flags & MS_RDONLY) {
79787eaa
JM		 2880 ret = btrfs_commit_super(tree_root);
2881 if (ret)
28c16cbb 2882 goto fail_qgroup;
e556ce2c 2883 }
e02119d5 2884 }
1a40e23b 2885
76dda93c 2886 ret = btrfs_find_orphan_roots(tree_root);
79787eaa 2887 if (ret)
28c16cbb 2888 goto fail_qgroup;
76dda93c 2889
7c2ca468 2890 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b 2891 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2 2892 if (ret)
28c16cbb 2893 goto fail_qgroup;
d68fc57b 2894
5d4f98a2 2895 ret = btrfs_recover_relocation(tree_root);
d7ce5843
MX		 2896 if (ret < 0) {
2897 printk(KERN_WARNING
efe120a0 2898 "BTRFS: failed to recover relocation\n");
d7ce5843 2899 err = -EINVAL;
bcef60f2 2900 goto fail_qgroup;
d7ce5843 2901 }
7c2ca468 2902 }
1a40e23b 2903
3de4586c
CM		 2904 location.objectid = BTRFS_FS_TREE_OBJECTID;
2905 location.type = BTRFS_ROOT_ITEM_KEY;
cb517eab 2906 location.offset = 0;
3de4586c 2907
3de4586c 2908 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
3140c9a3
DC		 2909 if (IS_ERR(fs_info->fs_root)) {
2910 err = PTR_ERR(fs_info->fs_root);
bcef60f2 2911 goto fail_qgroup;
3140c9a3 2912 }
c289811c 2913
2b6ba629
ID		 2914 if (sb->s_flags & MS_RDONLY)
2915 return 0;
59641015 2916
2b6ba629
ID		 2917 down_read(&fs_info->cleanup_work_sem);
2918 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2919 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 2920 up_read(&fs_info->cleanup_work_sem);
2b6ba629
ID		 2921 close_ctree(tree_root);
2922 return ret;
2923 }
2924 up_read(&fs_info->cleanup_work_sem);
59641015 2925
2b6ba629
ID		 2926 ret = btrfs_resume_balance_async(fs_info);
2927 if (ret) {
efe120a0 2928 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
2b6ba629
ID		 2929 close_ctree(tree_root);
2930 return ret;
e3acc2a6
JB		 2931 }
2932
8dabb742
SB		 2933 ret = btrfs_resume_dev_replace_async(fs_info);
2934 if (ret) {
efe120a0 2935 pr_warn("BTRFS: failed to resume dev_replace\n");
8dabb742
SB		 2936 close_ctree(tree_root);
2937 return ret;
2938 }
2939
b382a324
JS		 2940 btrfs_qgroup_rescan_resume(fs_info);
2941
f7a81ea4 2942 if (create_uuid_tree) {
efe120a0 2943 pr_info("BTRFS: creating UUID tree\n");
f7a81ea4
SB		 2944 ret = btrfs_create_uuid_tree(fs_info);
2945 if (ret) {
efe120a0 2946 pr_warn("BTRFS: failed to create the UUID tree %d\n",
f7a81ea4
SB		 2947 ret);
2948 close_ctree(tree_root);
2949 return ret;
2950 }
f420ee1e
SB		 2951 } else if (check_uuid_tree ||
2952 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
efe120a0 2953 pr_info("BTRFS: checking UUID tree\n");
70f80175
SB		 2954 ret = btrfs_check_uuid_tree(fs_info);
2955 if (ret) {
efe120a0 2956 pr_warn("BTRFS: failed to check the UUID tree %d\n",
70f80175
SB		 2957 ret);
2958 close_ctree(tree_root);
2959 return ret;
2960 }
2961 } else {
2962 fs_info->update_uuid_tree_gen = 1;
f7a81ea4
SB		 2963 }
2964
ad2b2c80 2965 return 0;
39279cc3 2966
bcef60f2
AJ		 2967fail_qgroup:
2968 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM		 2969fail_trans_kthread:
2970 kthread_stop(fs_info->transaction_kthread);
54067ae9 2971 btrfs_cleanup_transaction(fs_info->tree_root);
7b5ff90e 2972 del_fs_roots(fs_info);
3f157a2f 2973fail_cleaner:
a74a4b97 2974 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2975
2976 /*
2977 * make sure we're done with the btree inode before we stop our
2978 * kthreads
2979 */
2980 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 2981
2365dd3c
AJ		 2982fail_sysfs:
2983 btrfs_sysfs_remove_one(fs_info);
2984
1b1d1f66 2985fail_block_groups:
54067ae9 2986 btrfs_put_block_group_cache(fs_info);
1b1d1f66 2987 btrfs_free_block_groups(fs_info);
af31f5e5
CM		 2988
2989fail_tree_roots:
2990 free_root_pointers(fs_info, 1);
2b8195bb 2991 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 2992
39279cc3 2993fail_sb_buffer:
7abadb64 2994 btrfs_stop_all_workers(fs_info);
16cdcec7 2995fail_alloc:
4543df7e 2996fail_iput:
586e46e2
ID		 2997 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2998
4543df7e 2999 iput(fs_info->btree_inode);
c404e0dc
MX		 3000fail_bio_counter:
3001 percpu_counter_destroy(&fs_info->bio_counter);
963d678b
MX		 3002fail_delalloc_bytes:
3003 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX		 3004fail_dirty_metadata_bytes:
3005 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
ad081f14 3006fail_bdi:
7e662854 3007 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 3008fail_srcu:
3009 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 3010fail:
53b381b3 3011 btrfs_free_stripe_hash_table(fs_info);
586e46e2 3012 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 3013 return err;
af31f5e5
CM		 3014
3015recovery_tree_root:
af31f5e5
CM		 3016 if (!btrfs_test_opt(tree_root, RECOVERY))
3017 goto fail_tree_roots;
3018
3019 free_root_pointers(fs_info, 0);
3020
3021 /* don't use the log in recovery mode, it won't be valid */
3022 btrfs_set_super_log_root(disk_super, 0);
3023
3024 /* we can't trust the free space cache either */
3025 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3026
3027 ret = next_root_backup(fs_info, fs_info->super_copy,
3028 &num_backups_tried, &backup_index);
3029 if (ret == -1)
3030 goto fail_block_groups;
3031 goto retry_root_backup;
eb60ceac
CM		 3032}
3033
f2984462
CM		 3034static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3035{
f2984462
CM		 3036 if (uptodate) {
3037 set_buffer_uptodate(bh);
3038 } else {
442a4f63
SB		 3039 struct btrfs_device *device = (struct btrfs_device *)
3040 bh->b_private;
3041
efe120a0 3042 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
606686ee
JB		 3043 "I/O error on %s\n",
3044 rcu_str_deref(device->name));
1259ab75
CM		 3045 /* note, we dont' set_buffer_write_io_error because we have
3046 * our own ways of dealing with the IO errors
3047 */
f2984462 3048 clear_buffer_uptodate(bh);
442a4f63 3049 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM		 3050 }
3051 unlock_buffer(bh);
3052 put_bh(bh);
3053}
3054
a512bbf8
YZ		 3055struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3056{
3057 struct buffer_head *bh;
3058 struct buffer_head *latest = NULL;
3059 struct btrfs_super_block *super;
3060 int i;
3061 u64 transid = 0;
3062 u64 bytenr;
3063
3064 /* we would like to check all the supers, but that would make
3065 * a btrfs mount succeed after a mkfs from a different FS.
3066 * So, we need to add a special mount option to scan for
3067 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3068 */
3069 for (i = 0; i < 1; i++) {
3070 bytenr = btrfs_sb_offset(i);
8068a47e
AJ		 3071 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3072 i_size_read(bdev->bd_inode))
a512bbf8 3073 break;
8068a47e
AJ		 3074 bh = __bread(bdev, bytenr / 4096,
3075 BTRFS_SUPER_INFO_SIZE);
a512bbf8
YZ		 3076 if (!bh)
3077 continue;
3078
3079 super = (struct btrfs_super_block *)bh->b_data;
3080 if (btrfs_super_bytenr(super) != bytenr ||
3cae210f 3081 btrfs_super_magic(super) != BTRFS_MAGIC) {
a512bbf8
YZ		 3082 brelse(bh);
3083 continue;
3084 }
3085
3086 if (!latest || btrfs_super_generation(super) > transid) {
3087 brelse(latest);
3088 latest = bh;
3089 transid = btrfs_super_generation(super);
3090 } else {
3091 brelse(bh);
3092 }
3093 }
3094 return latest;
3095}
3096
4eedeb75
HH		 3097/*
3098 * this should be called twice, once with wait == 0 and
3099 * once with wait == 1. When wait == 0 is done, all the buffer heads
3100 * we write are pinned.
3101 *
3102 * They are released when wait == 1 is done.
3103 * max_mirrors must be the same for both runs, and it indicates how
3104 * many supers on this one device should be written.
3105 *
3106 * max_mirrors == 0 means to write them all.
3107 */
a512bbf8
YZ		 3108static int write_dev_supers(struct btrfs_device *device,
3109 struct btrfs_super_block *sb,
3110 int do_barriers, int wait, int max_mirrors)
3111{
3112 struct buffer_head *bh;
3113 int i;
3114 int ret;
3115 int errors = 0;
3116 u32 crc;
3117 u64 bytenr;
a512bbf8
YZ		 3118
3119 if (max_mirrors == 0)
3120 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3121
a512bbf8
YZ		 3122 for (i = 0; i < max_mirrors; i++) {
3123 bytenr = btrfs_sb_offset(i);
3124 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
3125 break;
3126
3127 if (wait) {
3128 bh = __find_get_block(device->bdev, bytenr / 4096,
3129 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3130 if (!bh) {
3131 errors++;
3132 continue;
3133 }
a512bbf8 3134 wait_on_buffer(bh);
4eedeb75
HH		 3135 if (!buffer_uptodate(bh))
3136 errors++;
3137
3138 /* drop our reference */
3139 brelse(bh);
3140
3141 /* drop the reference from the wait == 0 run */
3142 brelse(bh);
3143 continue;
a512bbf8
YZ		 3144 } else {
3145 btrfs_set_super_bytenr(sb, bytenr);
3146
3147 crc = ~(u32)0;
b0496686 3148 crc = btrfs_csum_data((char *)sb +
a512bbf8
YZ		 3149 BTRFS_CSUM_SIZE, crc,
3150 BTRFS_SUPER_INFO_SIZE -
3151 BTRFS_CSUM_SIZE);
3152 btrfs_csum_final(crc, sb->csum);
3153
4eedeb75
HH		 3154 /*
3155 * one reference for us, and we leave it for the
3156 * caller
3157 */
a512bbf8
YZ		 3158 bh = __getblk(device->bdev, bytenr / 4096,
3159 BTRFS_SUPER_INFO_SIZE);
634554dc 3160 if (!bh) {
efe120a0 3161 printk(KERN_ERR "BTRFS: couldn't get super "
634554dc
JB		 3162 "buffer head for bytenr %Lu\n", bytenr);
3163 errors++;
3164 continue;
3165 }
3166
a512bbf8
YZ		 3167 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3168
4eedeb75 3169 /* one reference for submit_bh */
a512bbf8 3170 get_bh(bh);
4eedeb75
HH		 3171
3172 set_buffer_uptodate(bh);
a512bbf8
YZ		 3173 lock_buffer(bh);
3174 bh->b_end_io = btrfs_end_buffer_write_sync;
442a4f63 3175 bh->b_private = device;
a512bbf8
YZ		 3176 }
3177
387125fc
CM		 3178 /*
3179 * we fua the first super. The others we allow
3180 * to go down lazy.
3181 */
e8117c26
WS		 3182 if (i == 0)
3183 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3184 else
3185 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
4eedeb75 3186 if (ret)
a512bbf8 3187 errors++;
a512bbf8
YZ		 3188 }
3189 return errors < i ? 0 : -1;
3190}
3191
387125fc
CM		 3192/*
3193 * endio for the write_dev_flush, this will wake anyone waiting
3194 * for the barrier when it is done
3195 */
3196static void btrfs_end_empty_barrier(struct bio *bio, int err)
3197{
3198 if (err) {
3199 if (err == -EOPNOTSUPP)
3200 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3201 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3202 }
3203 if (bio->bi_private)
3204 complete(bio->bi_private);
3205 bio_put(bio);
3206}
3207
3208/*
3209 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3210 * sent down. With wait == 1, it waits for the previous flush.
3211 *
3212 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3213 * capable
3214 */
3215static int write_dev_flush(struct btrfs_device *device, int wait)
3216{
3217 struct bio *bio;
3218 int ret = 0;
3219
3220 if (device->nobarriers)
3221 return 0;
3222
3223 if (wait) {
3224 bio = device->flush_bio;
3225 if (!bio)
3226 return 0;
3227
3228 wait_for_completion(&device->flush_wait);
3229
3230 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
efe120a0 3231 printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
606686ee 3232 rcu_str_deref(device->name));
387125fc 3233 device->nobarriers = 1;
5af3e8cc 3234 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
387125fc 3235 ret = -EIO;
5af3e8cc
SB		 3236 btrfs_dev_stat_inc_and_print(device,
3237 BTRFS_DEV_STAT_FLUSH_ERRS);
387125fc
CM		 3238 }
3239
3240 /* drop the reference from the wait == 0 run */
3241 bio_put(bio);
3242 device->flush_bio = NULL;
3243
3244 return ret;
3245 }
3246
3247 /*
3248 * one reference for us, and we leave it for the
3249 * caller
3250 */
9c017abc 3251 device->flush_bio = NULL;
9be3395b 3252 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
387125fc
CM		 3253 if (!bio)
3254 return -ENOMEM;
3255
3256 bio->bi_end_io = btrfs_end_empty_barrier;
3257 bio->bi_bdev = device->bdev;
3258 init_completion(&device->flush_wait);
3259 bio->bi_private = &device->flush_wait;
3260 device->flush_bio = bio;
3261
3262 bio_get(bio);
21adbd5c 3263 btrfsic_submit_bio(WRITE_FLUSH, bio);
387125fc
CM		 3264
3265 return 0;
3266}
3267
3268/*
3269 * send an empty flush down to each device in parallel,
3270 * then wait for them
3271 */
3272static int barrier_all_devices(struct btrfs_fs_info *info)
3273{
3274 struct list_head *head;
3275 struct btrfs_device *dev;
5af3e8cc
SB		 3276 int errors_send = 0;
3277 int errors_wait = 0;
387125fc
CM		 3278 int ret;
3279
3280 /* send down all the barriers */
3281 head = &info->fs_devices->devices;
3282 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3283 if (dev->missing)
3284 continue;
387125fc 3285 if (!dev->bdev) {
5af3e8cc 3286 errors_send++;
387125fc
CM		 3287 continue;
3288 }
3289 if (!dev->in_fs_metadata || !dev->writeable)
3290 continue;
3291
3292 ret = write_dev_flush(dev, 0);
3293 if (ret)
5af3e8cc 3294 errors_send++;
387125fc
CM		 3295 }
3296
3297 /* wait for all the barriers */
3298 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS		 3299 if (dev->missing)
3300 continue;
387125fc 3301 if (!dev->bdev) {
5af3e8cc 3302 errors_wait++;
387125fc
CM		 3303 continue;
3304 }
3305 if (!dev->in_fs_metadata || !dev->writeable)
3306 continue;
3307
3308 ret = write_dev_flush(dev, 1);
3309 if (ret)
5af3e8cc 3310 errors_wait++;
387125fc 3311 }
5af3e8cc
SB		 3312 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3313 errors_wait > info->num_tolerated_disk_barrier_failures)
387125fc
CM		 3314 return -EIO;
3315 return 0;
3316}
3317
5af3e8cc
SB		 3318int btrfs_calc_num_tolerated_disk_barrier_failures(
3319 struct btrfs_fs_info *fs_info)
3320{
3321 struct btrfs_ioctl_space_info space;
3322 struct btrfs_space_info *sinfo;
3323 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3324 BTRFS_BLOCK_GROUP_SYSTEM,
3325 BTRFS_BLOCK_GROUP_METADATA,
3326 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3327 int num_types = 4;
3328 int i;
3329 int c;
3330 int num_tolerated_disk_barrier_failures =
3331 (int)fs_info->fs_devices->num_devices;
3332
3333 for (i = 0; i < num_types; i++) {
3334 struct btrfs_space_info *tmp;
3335
3336 sinfo = NULL;
3337 rcu_read_lock();
3338 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3339 if (tmp->flags == types[i]) {
3340 sinfo = tmp;
3341 break;
3342 }
3343 }
3344 rcu_read_unlock();
3345
3346 if (!sinfo)
3347 continue;
3348
3349 down_read(&sinfo->groups_sem);
3350 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3351 if (!list_empty(&sinfo->block_groups[c])) {
3352 u64 flags;
3353
3354 btrfs_get_block_group_info(
3355 &sinfo->block_groups[c], &space);
3356 if (space.total_bytes == 0 ||
3357 space.used_bytes == 0)
3358 continue;
3359 flags = space.flags;
3360 /*
3361 * return
3362 * 0: if dup, single or RAID0 is configured for
3363 * any of metadata, system or data, else
3364 * 1: if RAID5 is configured, or if RAID1 or
3365 * RAID10 is configured and only two mirrors
3366 * are used, else
3367 * 2: if RAID6 is configured, else
3368 * num_mirrors - 1: if RAID1 or RAID10 is
3369 * configured and more than
3370 * 2 mirrors are used.
3371 */
3372 if (num_tolerated_disk_barrier_failures > 0 &&
3373 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3374 BTRFS_BLOCK_GROUP_RAID0)) ||
3375 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3376 == 0)))
3377 num_tolerated_disk_barrier_failures = 0;
53b381b3
DW		 3378 else if (num_tolerated_disk_barrier_failures > 1) {
3379 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3380 BTRFS_BLOCK_GROUP_RAID5 |
3381 BTRFS_BLOCK_GROUP_RAID10)) {
3382 num_tolerated_disk_barrier_failures = 1;
3383 } else if (flags &
15b0a89d 3384 BTRFS_BLOCK_GROUP_RAID6) {
53b381b3
DW		 3385 num_tolerated_disk_barrier_failures = 2;
3386 }
3387 }
5af3e8cc
SB		 3388 }
3389 }
3390 up_read(&sinfo->groups_sem);
3391 }
3392
3393 return num_tolerated_disk_barrier_failures;
3394}
3395
48a3b636 3396static int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 3397{
e5e9a520 3398 struct list_head *head;
f2984462 3399 struct btrfs_device *dev;
a061fc8d 3400 struct btrfs_super_block *sb;
f2984462 3401 struct btrfs_dev_item *dev_item;
f2984462
CM		 3402 int ret;
3403 int do_barriers;
a236aed1
CM		 3404 int max_errors;
3405 int total_errors = 0;
a061fc8d 3406 u64 flags;
f2984462
CM		 3407
3408 do_barriers = !btrfs_test_opt(root, NOBARRIER);
af31f5e5 3409 backup_super_roots(root->fs_info);
f2984462 3410
6c41761f 3411 sb = root->fs_info->super_for_commit;
a061fc8d 3412 dev_item = &sb->dev_item;
e5e9a520 3413
174ba509 3414 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 3415 head = &root->fs_info->fs_devices->devices;
d7306801 3416 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
387125fc 3417
5af3e8cc
SB		 3418 if (do_barriers) {
3419 ret = barrier_all_devices(root->fs_info);
3420 if (ret) {
3421 mutex_unlock(
3422 &root->fs_info->fs_devices->device_list_mutex);
3423 btrfs_error(root->fs_info, ret,
3424 "errors while submitting device barriers.");
3425 return ret;
3426 }
3427 }
387125fc 3428
1f78160c 3429 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3430 if (!dev->bdev) {
3431 total_errors++;
3432 continue;
3433 }
2b82032c 3434 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3435 continue;
3436
2b82032c 3437 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 3438 btrfs_set_stack_device_type(dev_item, dev->type);
3439 btrfs_set_stack_device_id(dev_item, dev->devid);
3440 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
3441 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
3442 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3443 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3444 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3445 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 3446 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 3447
a061fc8d
CM		 3448 flags = btrfs_super_flags(sb);
3449 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3450
a512bbf8 3451 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 3452 if (ret)
3453 total_errors++;
f2984462 3454 }
a236aed1 3455 if (total_errors > max_errors) {
efe120a0 3456 btrfs_err(root->fs_info, "%d errors while writing supers",
d397712b 3457 total_errors);
a724b436 3458 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
79787eaa 3459
9d565ba4
SB		 3460 /* FUA is masked off if unsupported and can't be the reason */
3461 btrfs_error(root->fs_info, -EIO,
3462 "%d errors while writing supers", total_errors);
3463 return -EIO;
a236aed1 3464 }
f2984462 3465
a512bbf8 3466 total_errors = 0;
1f78160c 3467 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3468 if (!dev->bdev)
3469 continue;
2b82032c 3470 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3471 continue;
3472
a512bbf8
YZ		 3473 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3474 if (ret)
3475 total_errors++;
f2984462 3476 }
174ba509 3477 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 3478 if (total_errors > max_errors) {
79787eaa
JM		 3479 btrfs_error(root->fs_info, -EIO,
3480 "%d errors while writing supers", total_errors);
3481 return -EIO;
a236aed1 3482 }
f2984462
CM		 3483 return 0;
3484}
3485
a512bbf8
YZ		 3486int write_ctree_super(struct btrfs_trans_handle *trans,
3487 struct btrfs_root *root, int max_mirrors)
eb60ceac 3488{
f570e757 3489 return write_all_supers(root, max_mirrors);
cfaa7295
CM		 3490}
3491
cb517eab
MX		 3492/* Drop a fs root from the radix tree and free it. */
3493void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3494 struct btrfs_root *root)
2619ba1f 3495{
4df27c4d 3496 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 3497 radix_tree_delete(&fs_info->fs_roots_radix,
3498 (unsigned long)root->root_key.objectid);
4df27c4d 3499 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 3500
3501 if (btrfs_root_refs(&root->root_item) == 0)
3502 synchronize_srcu(&fs_info->subvol_srcu);
3503
1a4319cc 3504 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
3321719e 3505 btrfs_free_log(NULL, root);
3321719e 3506
581bb050
LZ		 3507 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3508 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3509 free_fs_root(root);
4df27c4d
YZ		 3510}
3511
3512static void free_fs_root(struct btrfs_root *root)
3513{
82d5902d 3514 iput(root->cache_inode);
4df27c4d 3515 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
1cb048f5
FDBM		 3516 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3517 root->orphan_block_rsv = NULL;
0ee5dc67
AV		 3518 if (root->anon_dev)
3519 free_anon_bdev(root->anon_dev);
8257b2dc
MX		 3520 if (root->subv_writers)
3521 btrfs_free_subvolume_writers(root->subv_writers);
4df27c4d
YZ		 3522 free_extent_buffer(root->node);
3523 free_extent_buffer(root->commit_root);
581bb050
LZ		 3524 kfree(root->free_ino_ctl);
3525 kfree(root->free_ino_pinned);
d397712b 3526 kfree(root->name);
b0feb9d9 3527 btrfs_put_fs_root(root);
2619ba1f
CM		 3528}
3529
cb517eab
MX		 3530void btrfs_free_fs_root(struct btrfs_root *root)
3531{
3532 free_fs_root(root);
2619ba1f
CM		 3533}
3534
c146afad 3535int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3536{
c146afad
YZ		 3537 u64 root_objectid = 0;
3538 struct btrfs_root *gang[8];
65d33fd7
QW		 3539 int i = 0;
3540 int err = 0;
3541 unsigned int ret = 0;
3542 int index;
e089f05c 3543
c146afad 3544 while (1) {
65d33fd7 3545 index = srcu_read_lock(&fs_info->subvol_srcu);
c146afad
YZ		 3546 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3547 (void **)gang, root_objectid,
3548 ARRAY_SIZE(gang));
65d33fd7
QW		 3549 if (!ret) {
3550 srcu_read_unlock(&fs_info->subvol_srcu, index);
c146afad 3551 break;
65d33fd7 3552 }
5d4f98a2 3553 root_objectid = gang[ret - 1]->root_key.objectid + 1;
65d33fd7 3554
c146afad 3555 for (i = 0; i < ret; i++) {
65d33fd7
QW		 3556 /* Avoid to grab roots in dead_roots */
3557 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3558 gang[i] = NULL;
3559 continue;
3560 }
3561 /* grab all the search result for later use */
3562 gang[i] = btrfs_grab_fs_root(gang[i]);
3563 }
3564 srcu_read_unlock(&fs_info->subvol_srcu, index);
66b4ffd1 3565
65d33fd7
QW		 3566 for (i = 0; i < ret; i++) {
3567 if (!gang[i])
3568 continue;
c146afad 3569 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 3570 err = btrfs_orphan_cleanup(gang[i]);
3571 if (err)
65d33fd7
QW		 3572 break;
3573 btrfs_put_fs_root(gang[i]);
c146afad
YZ		 3574 }
3575 root_objectid++;
3576 }
65d33fd7
QW		 3577
3578 /* release the uncleaned roots due to error */
3579 for (; i < ret; i++) {
3580 if (gang[i])
3581 btrfs_put_fs_root(gang[i]);
3582 }
3583 return err;
c146afad 3584}
a2135011 3585
c146afad
YZ		 3586int btrfs_commit_super(struct btrfs_root *root)
3587{
3588 struct btrfs_trans_handle *trans;
a74a4b97 3589
c146afad 3590 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 3591 btrfs_run_delayed_iputs(root);
c146afad 3592 mutex_unlock(&root->fs_info->cleaner_mutex);
9d1a2a3a 3593 wake_up_process(root->fs_info->cleaner_kthread);
c71bf099
YZ		 3594
3595 /* wait until ongoing cleanup work done */
3596 down_write(&root->fs_info->cleanup_work_sem);
3597 up_write(&root->fs_info->cleanup_work_sem);
3598
7a7eaa40 3599 trans = btrfs_join_transaction(root);
3612b495
TI		 3600 if (IS_ERR(trans))
3601 return PTR_ERR(trans);
d52c1bcc 3602 return btrfs_commit_transaction(trans, root);
c146afad
YZ		 3603}
3604
3605int close_ctree(struct btrfs_root *root)
3606{
3607 struct btrfs_fs_info *fs_info = root->fs_info;
3608 int ret;
3609
3610 fs_info->closing = 1;
3611 smp_mb();
3612
803b2f54
SB		 3613 /* wait for the uuid_scan task to finish */
3614 down(&fs_info->uuid_tree_rescan_sem);
3615 /* avoid complains from lockdep et al., set sem back to initial state */
3616 up(&fs_info->uuid_tree_rescan_sem);
3617
837d5b6e 3618 /* pause restriper - we want to resume on mount */
aa1b8cd4 3619 btrfs_pause_balance(fs_info);
837d5b6e 3620
8dabb742
SB		 3621 btrfs_dev_replace_suspend_for_unmount(fs_info);
3622
aa1b8cd4 3623 btrfs_scrub_cancel(fs_info);
4cb5300b
CM		 3624
3625 /* wait for any defraggers to finish */
3626 wait_event(fs_info->transaction_wait,
3627 (atomic_read(&fs_info->defrag_running) == 0));
3628
3629 /* clear out the rbtree of defraggable inodes */
26176e7c 3630 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3631
21c7e756
MX		 3632 cancel_work_sync(&fs_info->async_reclaim_work);
3633
c146afad 3634 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 3635 ret = btrfs_commit_super(root);
3636 if (ret)
efe120a0 3637 btrfs_err(root->fs_info, "commit super ret %d", ret);
acce952b 3638 }
3639
87533c47 3640 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
68ce9682 3641 btrfs_error_commit_super(root);
0f7d52f4 3642
e3029d9f
AV		 3643 kthread_stop(fs_info->transaction_kthread);
3644 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3645
f25784b3
YZ		 3646 fs_info->closing = 2;
3647 smp_mb();
3648
bcef60f2
AJ		 3649 btrfs_free_qgroup_config(root->fs_info);
3650
963d678b 3651 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
efe120a0 3652 btrfs_info(root->fs_info, "at unmount delalloc count %lld",
963d678b 3653 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3654 }
bcc63abb 3655
5ac1d209
JM		 3656 btrfs_sysfs_remove_one(fs_info);
3657
c146afad 3658 del_fs_roots(fs_info);
d10c5f31 3659
1a4319cc
LB		 3660 btrfs_put_block_group_cache(fs_info);
3661
2b1360da
JB		 3662 btrfs_free_block_groups(fs_info);
3663
de348ee0
WS		 3664 /*
3665 * we must make sure there is not any read request to
3666 * submit after we stopping all workers.
3667 */
3668 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
96192499
JB		 3669 btrfs_stop_all_workers(fs_info);
3670
13e6c37b 3671 free_root_pointers(fs_info, 1);
9ad6b7bc 3672
13e6c37b 3673 iput(fs_info->btree_inode);
d6bfde87 3674
21adbd5c
SB		 3675#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3676 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3677 btrfsic_unmount(root, fs_info->fs_devices);
3678#endif
3679
dfe25020 3680 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3681 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3682
e2d84521 3683 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 3684 percpu_counter_destroy(&fs_info->delalloc_bytes);
c404e0dc 3685 percpu_counter_destroy(&fs_info->bio_counter);
04160088 3686 bdi_destroy(&fs_info->bdi);
76dda93c 3687 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 3688
53b381b3
DW		 3689 btrfs_free_stripe_hash_table(fs_info);
3690
1cb048f5
FDBM		 3691 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3692 root->orphan_block_rsv = NULL;
3693
eb60ceac
CM		 3694 return 0;
3695}
3696
b9fab919
CM		 3697int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3698 int atomic)
5f39d397 3699{
1259ab75 3700 int ret;
727011e0 3701 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 3702
0b32f4bb 3703 ret = extent_buffer_uptodate(buf);
1259ab75
CM		 3704 if (!ret)
3705 return ret;
3706
3707 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM		 3708 parent_transid, atomic);
3709 if (ret == -EAGAIN)
3710 return ret;
1259ab75 3711 return !ret;
5f39d397
CM		 3712}
3713
3714int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 3715{
0b32f4bb 3716 return set_extent_buffer_uptodate(buf);
5f39d397 3717}
6702ed49 3718
5f39d397
CM		 3719void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3720{
06ea65a3 3721 struct btrfs_root *root;
5f39d397 3722 u64 transid = btrfs_header_generation(buf);
b9473439 3723 int was_dirty;
b4ce94de 3724
06ea65a3
JB		 3725#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3726 /*
3727 * This is a fast path so only do this check if we have sanity tests
3728 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3729 * outside of the sanity tests.
3730 */
3731 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3732 return;
3733#endif
3734 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
b9447ef8 3735 btrfs_assert_tree_locked(buf);
31b1a2bd
JL		 3736 if (transid != root->fs_info->generation)
3737 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
d397712b 3738 "found %llu running %llu\n",
c1c9ff7c 3739 buf->start, transid, root->fs_info->generation);
0b32f4bb 3740 was_dirty = set_extent_buffer_dirty(buf);
e2d84521
MX		 3741 if (!was_dirty)
3742 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3743 buf->len,
3744 root->fs_info->dirty_metadata_batch);
1f21ef0a
FM		 3745#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3746 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
3747 btrfs_print_leaf(root, buf);
3748 ASSERT(0);
3749 }
3750#endif
eb60ceac
CM		 3751}
3752
b53d3f5d
LB		 3753static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3754 int flush_delayed)
16cdcec7
MX		 3755{
3756 /*
3757 * looks as though older kernels can get into trouble with
3758 * this code, they end up stuck in balance_dirty_pages forever
3759 */
e2d84521 3760 int ret;
16cdcec7
MX		 3761
3762 if (current->flags & PF_MEMALLOC)
3763 return;
3764
b53d3f5d
LB		 3765 if (flush_delayed)
3766 btrfs_balance_delayed_items(root);
16cdcec7 3767
e2d84521
MX		 3768 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3769 BTRFS_DIRTY_METADATA_THRESH);
3770 if (ret > 0) {
d0e1d66b
NJ		 3771 balance_dirty_pages_ratelimited(
3772 root->fs_info->btree_inode->i_mapping);
16cdcec7
MX		 3773 }
3774 return;
3775}
3776
b53d3f5d 3777void btrfs_btree_balance_dirty(struct btrfs_root *root)
35b7e476 3778{
b53d3f5d
LB		 3779 __btrfs_btree_balance_dirty(root, 1);
3780}
585ad2c3 3781
b53d3f5d
LB		 3782void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3783{
3784 __btrfs_btree_balance_dirty(root, 0);
35b7e476 3785}
6b80053d 3786
ca7a79ad 3787int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 3788{
727011e0 3789 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
0b32f4bb 3790 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
6b80053d 3791}
0da5468f 3792
fcd1f065 3793static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 3794 int read_only)
3795{
1104a885
DS		 3796 /*
3797 * Placeholder for checks
3798 */
fcd1f065 3799 return 0;
acce952b 3800}
3801
48a3b636 3802static void btrfs_error_commit_super(struct btrfs_root *root)
acce952b 3803{
acce952b 3804 mutex_lock(&root->fs_info->cleaner_mutex);
3805 btrfs_run_delayed_iputs(root);
3806 mutex_unlock(&root->fs_info->cleaner_mutex);
3807
3808 down_write(&root->fs_info->cleanup_work_sem);
3809 up_write(&root->fs_info->cleanup_work_sem);
3810
3811 /* cleanup FS via transaction */
3812 btrfs_cleanup_transaction(root);
acce952b 3813}
3814
569e0f35
JB		 3815static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
3816 struct btrfs_root *root)
acce952b 3817{
3818 struct btrfs_inode *btrfs_inode;
3819 struct list_head splice;
3820
3821 INIT_LIST_HEAD(&splice);
3822
3823 mutex_lock(&root->fs_info->ordered_operations_mutex);
199c2a9c 3824 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3825
569e0f35 3826 list_splice_init(&t->ordered_operations, &splice);
acce952b 3827 while (!list_empty(&splice)) {
3828 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3829 ordered_operations);
3830
3831 list_del_init(&btrfs_inode->ordered_operations);
199c2a9c 3832 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3833
3834 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3835
199c2a9c 3836 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3837 }
3838
199c2a9c 3839 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3840 mutex_unlock(&root->fs_info->ordered_operations_mutex);
acce952b 3841}
3842
143bede5 3843static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 3844{
acce952b 3845 struct btrfs_ordered_extent *ordered;
acce952b 3846
199c2a9c 3847 spin_lock(&root->ordered_extent_lock);
779880ef
JB		 3848 /*
3849 * This will just short circuit the ordered completion stuff which will
3850 * make sure the ordered extent gets properly cleaned up.
3851 */
199c2a9c 3852 list_for_each_entry(ordered, &root->ordered_extents,
779880ef
JB		 3853 root_extent_list)
3854 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
199c2a9c
MX		 3855 spin_unlock(&root->ordered_extent_lock);
3856}
3857
3858static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3859{
3860 struct btrfs_root *root;
3861 struct list_head splice;
3862
3863 INIT_LIST_HEAD(&splice);
3864
3865 spin_lock(&fs_info->ordered_root_lock);
3866 list_splice_init(&fs_info->ordered_roots, &splice);
3867 while (!list_empty(&splice)) {
3868 root = list_first_entry(&splice, struct btrfs_root,
3869 ordered_root);
1de2cfde
JB		 3870 list_move_tail(&root->ordered_root,
3871 &fs_info->ordered_roots);
199c2a9c 3872
2a85d9ca 3873 spin_unlock(&fs_info->ordered_root_lock);
199c2a9c
MX		 3874 btrfs_destroy_ordered_extents(root);
3875
2a85d9ca
LB		 3876 cond_resched();
3877 spin_lock(&fs_info->ordered_root_lock);
199c2a9c
MX		 3878 }
3879 spin_unlock(&fs_info->ordered_root_lock);
acce952b 3880}
3881
35a3621b
SB		 3882static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3883 struct btrfs_root *root)
acce952b 3884{
3885 struct rb_node *node;
3886 struct btrfs_delayed_ref_root *delayed_refs;
3887 struct btrfs_delayed_ref_node *ref;
3888 int ret = 0;
3889
3890 delayed_refs = &trans->delayed_refs;
3891
3892 spin_lock(&delayed_refs->lock);
d7df2c79 3893 if (atomic_read(&delayed_refs->num_entries) == 0) {
cfece4db 3894 spin_unlock(&delayed_refs->lock);
efe120a0 3895 btrfs_info(root->fs_info, "delayed_refs has NO entry");
acce952b 3896 return ret;
3897 }
3898
d7df2c79
JB		 3899 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
3900 struct btrfs_delayed_ref_head *head;
e78417d1 3901 bool pin_bytes = false;
acce952b 3902
d7df2c79
JB		 3903 head = rb_entry(node, struct btrfs_delayed_ref_head,
3904 href_node);
3905 if (!mutex_trylock(&head->mutex)) {
3906 atomic_inc(&head->node.refs);
3907 spin_unlock(&delayed_refs->lock);
eb12db69 3908
d7df2c79 3909 mutex_lock(&head->mutex);
e78417d1 3910 mutex_unlock(&head->mutex);
d7df2c79
JB		 3911 btrfs_put_delayed_ref(&head->node);
3912 spin_lock(&delayed_refs->lock);
3913 continue;
3914 }
3915 spin_lock(&head->lock);
3916 while ((node = rb_first(&head->ref_root)) != NULL) {
3917 ref = rb_entry(node, struct btrfs_delayed_ref_node,
3918 rb_node);
3919 ref->in_tree = 0;
3920 rb_erase(&ref->rb_node, &head->ref_root);
3921 atomic_dec(&delayed_refs->num_entries);
3922 btrfs_put_delayed_ref(ref);
e78417d1 3923 }
d7df2c79
JB		 3924 if (head->must_insert_reserved)
3925 pin_bytes = true;
3926 btrfs_free_delayed_extent_op(head->extent_op);
3927 delayed_refs->num_heads--;
3928 if (head->processing == 0)
3929 delayed_refs->num_heads_ready--;
3930 atomic_dec(&delayed_refs->num_entries);
3931 head->node.in_tree = 0;
3932 rb_erase(&head->href_node, &delayed_refs->href_root);
3933 spin_unlock(&head->lock);
3934 spin_unlock(&delayed_refs->lock);
3935 mutex_unlock(&head->mutex);
acce952b 3936
d7df2c79
JB		 3937 if (pin_bytes)
3938 btrfs_pin_extent(root, head->node.bytenr,
3939 head->node.num_bytes, 1);
3940 btrfs_put_delayed_ref(&head->node);
acce952b 3941 cond_resched();
3942 spin_lock(&delayed_refs->lock);
3943 }
3944
3945 spin_unlock(&delayed_refs->lock);
3946
3947 return ret;
3948}
3949
143bede5 3950static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 3951{
3952 struct btrfs_inode *btrfs_inode;
3953 struct list_head splice;
3954
3955 INIT_LIST_HEAD(&splice);
3956
eb73c1b7
MX		 3957 spin_lock(&root->delalloc_lock);
3958 list_splice_init(&root->delalloc_inodes, &splice);
acce952b 3959
3960 while (!list_empty(&splice)) {
eb73c1b7
MX		 3961 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3962 delalloc_inodes);
acce952b 3963
3964 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX		 3965 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3966 &btrfs_inode->runtime_flags);
eb73c1b7 3967 spin_unlock(&root->delalloc_lock);
acce952b 3968
3969 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3970
eb73c1b7 3971 spin_lock(&root->delalloc_lock);
acce952b 3972 }
3973
eb73c1b7
MX		 3974 spin_unlock(&root->delalloc_lock);
3975}
3976
3977static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3978{
3979 struct btrfs_root *root;
3980 struct list_head splice;
3981
3982 INIT_LIST_HEAD(&splice);
3983
3984 spin_lock(&fs_info->delalloc_root_lock);
3985 list_splice_init(&fs_info->delalloc_roots, &splice);
3986 while (!list_empty(&splice)) {
3987 root = list_first_entry(&splice, struct btrfs_root,
3988 delalloc_root);
3989 list_del_init(&root->delalloc_root);
3990 root = btrfs_grab_fs_root(root);
3991 BUG_ON(!root);
3992 spin_unlock(&fs_info->delalloc_root_lock);
3993
3994 btrfs_destroy_delalloc_inodes(root);
3995 btrfs_put_fs_root(root);
3996
3997 spin_lock(&fs_info->delalloc_root_lock);
3998 }
3999 spin_unlock(&fs_info->delalloc_root_lock);
acce952b 4000}
4001
4002static int btrfs_destroy_marked_extents(struct btrfs_root *root,
4003 struct extent_io_tree *dirty_pages,
4004 int mark)
4005{
4006 int ret;
acce952b 4007 struct extent_buffer *eb;
4008 u64 start = 0;
4009 u64 end;
acce952b 4010
4011 while (1) {
4012 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 4013 mark, NULL);
acce952b 4014 if (ret)
4015 break;
4016
4017 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
4018 while (start <= end) {
fd8b2b61
JB		 4019 eb = btrfs_find_tree_block(root, start,
4020 root->leafsize);
69a85bd8 4021 start += root->leafsize;
fd8b2b61 4022 if (!eb)
acce952b 4023 continue;
fd8b2b61 4024 wait_on_extent_buffer_writeback(eb);
acce952b 4025
fd8b2b61
JB		 4026 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4027 &eb->bflags))
4028 clear_extent_buffer_dirty(eb);
4029 free_extent_buffer_stale(eb);
acce952b 4030 }
4031 }
4032
4033 return ret;
4034}
4035
4036static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4037 struct extent_io_tree *pinned_extents)
4038{
4039 struct extent_io_tree *unpin;
4040 u64 start;
4041 u64 end;
4042 int ret;
ed0eaa14 4043 bool loop = true;
acce952b 4044
4045 unpin = pinned_extents;
ed0eaa14 4046again:
acce952b 4047 while (1) {
4048 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 4049 EXTENT_DIRTY, NULL);
acce952b 4050 if (ret)
4051 break;
4052
4053 /* opt_discard */
5378e607
LD		 4054 if (btrfs_test_opt(root, DISCARD))
4055 ret = btrfs_error_discard_extent(root, start,
4056 end + 1 - start,
4057 NULL);
acce952b 4058
4059 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4060 btrfs_error_unpin_extent_range(root, start, end);
4061 cond_resched();
4062 }
4063
ed0eaa14
LB		 4064 if (loop) {
4065 if (unpin == &root->fs_info->freed_extents[0])
4066 unpin = &root->fs_info->freed_extents[1];
4067 else
4068 unpin = &root->fs_info->freed_extents[0];
4069 loop = false;
4070 goto again;
4071 }
4072
acce952b 4073 return 0;
4074}
4075
49b25e05
JM		 4076void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4077 struct btrfs_root *root)
4078{
724e2315
JB		 4079 btrfs_destroy_ordered_operations(cur_trans, root);
4080
49b25e05 4081 btrfs_destroy_delayed_refs(cur_trans, root);
49b25e05 4082
4a9d8bde 4083 cur_trans->state = TRANS_STATE_COMMIT_START;
d7096fc3 4084 wake_up(&root->fs_info->transaction_blocked_wait);
49b25e05 4085
4a9d8bde 4086 cur_trans->state = TRANS_STATE_UNBLOCKED;
d7096fc3 4087 wake_up(&root->fs_info->transaction_wait);
49b25e05 4088
67cde344
MX		 4089 btrfs_destroy_delayed_inodes(root);
4090 btrfs_assert_delayed_root_empty(root);
49b25e05 4091
49b25e05
JM		 4092 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4093 EXTENT_DIRTY);
6e841e32
LB		 4094 btrfs_destroy_pinned_extent(root,
4095 root->fs_info->pinned_extents);
49b25e05 4096
4a9d8bde
MX		 4097 cur_trans->state =TRANS_STATE_COMPLETED;
4098 wake_up(&cur_trans->commit_wait);
4099
49b25e05
JM		 4100 /*
4101 memset(cur_trans, 0, sizeof(*cur_trans));
4102 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4103 */
4104}
4105
48a3b636 4106static int btrfs_cleanup_transaction(struct btrfs_root *root)
acce952b 4107{
4108 struct btrfs_transaction *t;
acce952b 4109
acce952b 4110 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4111
a4abeea4 4112 spin_lock(&root->fs_info->trans_lock);
724e2315
JB		 4113 while (!list_empty(&root->fs_info->trans_list)) {
4114 t = list_first_entry(&root->fs_info->trans_list,
4115 struct btrfs_transaction, list);
4116 if (t->state >= TRANS_STATE_COMMIT_START) {
4117 atomic_inc(&t->use_count);
4118 spin_unlock(&root->fs_info->trans_lock);
4119 btrfs_wait_for_commit(root, t->transid);
4120 btrfs_put_transaction(t);
4121 spin_lock(&root->fs_info->trans_lock);
4122 continue;
4123 }
4124 if (t == root->fs_info->running_transaction) {
4125 t->state = TRANS_STATE_COMMIT_DOING;
4126 spin_unlock(&root->fs_info->trans_lock);
4127 /*
4128 * We wait for 0 num_writers since we don't hold a trans
4129 * handle open currently for this transaction.
4130 */
4131 wait_event(t->writer_wait,
4132 atomic_read(&t->num_writers) == 0);
4133 } else {
4134 spin_unlock(&root->fs_info->trans_lock);
4135 }
4136 btrfs_cleanup_one_transaction(t, root);
4a9d8bde 4137
724e2315
JB		 4138 spin_lock(&root->fs_info->trans_lock);
4139 if (t == root->fs_info->running_transaction)
4140 root->fs_info->running_transaction = NULL;
acce952b 4141 list_del_init(&t->list);
724e2315 4142 spin_unlock(&root->fs_info->trans_lock);
acce952b 4143
724e2315
JB		 4144 btrfs_put_transaction(t);
4145 trace_btrfs_transaction_commit(root);
4146 spin_lock(&root->fs_info->trans_lock);
4147 }
4148 spin_unlock(&root->fs_info->trans_lock);
4149 btrfs_destroy_all_ordered_extents(root->fs_info);
4150 btrfs_destroy_delayed_inodes(root);
4151 btrfs_assert_delayed_root_empty(root);
4152 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4153 btrfs_destroy_all_delalloc_inodes(root->fs_info);
acce952b 4154 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 4155
4156 return 0;
4157}
4158
d1310b2e 4159static struct extent_io_ops btree_extent_io_ops = {
ce9adaa5 4160 .readpage_end_io_hook = btree_readpage_end_io_hook,
4bb31e92 4161 .readpage_io_failed_hook = btree_io_failed_hook,
0b86a832 4162 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 4163 /* note we're sharing with inode.c for the merge bio hook */
4164 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 4165};
Linux kernel - Deliverables
This page took 0.864463 seconds and 5 git commands to generate.