btrfs: remove useless ACL check
[linux-2.6-block.git] / fs / btrfs / disk-io.c
CommitLineData
6cbd5570
CM
1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
e20d96d6 19#include <linux/fs.h>
d98237b3 20#include <linux/blkdev.h>
87cbda5c 21#include <linux/scatterlist.h>
22b0ebda 22#include <linux/swap.h>
0f7d52f4 23#include <linux/radix-tree.h>
35b7e476 24#include <linux/writeback.h>
d397712b 25#include <linux/buffer_head.h>
ce9adaa5 26#include <linux/workqueue.h>
a74a4b97 27#include <linux/kthread.h>
4b4e25f2 28#include <linux/freezer.h>
5a0e3ad6 29#include <linux/slab.h>
784b4e29 30#include <linux/migrate.h>
7a36ddec 31#include <linux/ratelimit.h>
6463fe58 32#include <linux/uuid.h>
803b2f54 33#include <linux/semaphore.h>
7e75bf3f 34#include <asm/unaligned.h>
eb60ceac
CM
35#include "ctree.h"
36#include "disk-io.h"
0b947aff 37#include "hash.h"
e089f05c 38#include "transaction.h"
0f7d52f4 39#include "btrfs_inode.h"
0b86a832 40#include "volumes.h"
db94535d 41#include "print-tree.h"
8b712842 42#include "async-thread.h"
925baedd 43#include "locking.h"
e02119d5 44#include "tree-log.h"
fa9c0d79 45#include "free-space-cache.h"
581bb050 46#include "inode-map.h"
21adbd5c 47#include "check-integrity.h"
606686ee 48#include "rcu-string.h"
8dabb742 49#include "dev-replace.h"
53b381b3 50#include "raid56.h"
5ac1d209 51#include "sysfs.h"
eb60ceac 52
de0022b9
JB
53#ifdef CONFIG_X86
54#include <asm/cpufeature.h>
55#endif
56
d1310b2e 57static struct extent_io_ops btree_extent_io_ops;
8b712842 58static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 59static void free_fs_root(struct btrfs_root *root);
fcd1f065 60static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 61 int read_only);
569e0f35
JB
62static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
63 struct btrfs_root *root);
143bede5 64static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
acce952b 65static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
66 struct btrfs_root *root);
143bede5 67static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
acce952b 68static int btrfs_destroy_marked_extents(struct btrfs_root *root,
69 struct extent_io_tree *dirty_pages,
70 int mark);
71static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
72 struct extent_io_tree *pinned_extents);
48a3b636
ES
73static int btrfs_cleanup_transaction(struct btrfs_root *root);
74static void btrfs_error_commit_super(struct btrfs_root *root);
ce9adaa5 75
d352ac68
CM
76/*
77 * end_io_wq structs are used to do processing in task context when an IO is
78 * complete. This is used during reads to verify checksums, and it is used
79 * by writes to insert metadata for new file extents after IO is complete.
80 */
ce9adaa5
CM
81struct end_io_wq {
82 struct bio *bio;
83 bio_end_io_t *end_io;
84 void *private;
85 struct btrfs_fs_info *info;
86 int error;
22c59948 87 int metadata;
ce9adaa5 88 struct list_head list;
8b712842 89 struct btrfs_work work;
ce9adaa5 90};
0da5468f 91
d352ac68
CM
92/*
93 * async submit bios are used to offload expensive checksumming
94 * onto the worker threads. They checksum file and metadata bios
95 * just before they are sent down the IO stack.
96 */
44b8bd7e
CM
97struct async_submit_bio {
98 struct inode *inode;
99 struct bio *bio;
100 struct list_head list;
4a69a410
CM
101 extent_submit_bio_hook_t *submit_bio_start;
102 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM
103 int rw;
104 int mirror_num;
c8b97818 105 unsigned long bio_flags;
eaf25d93
CM
106 /*
107 * bio_offset is optional, can be used if the pages in the bio
108 * can't tell us where in the file the bio should go
109 */
110 u64 bio_offset;
8b712842 111 struct btrfs_work work;
79787eaa 112 int error;
44b8bd7e
CM
113};
114
85d4e461
CM
115/*
116 * Lockdep class keys for extent_buffer->lock's in this root. For a given
117 * eb, the lockdep key is determined by the btrfs_root it belongs to and
118 * the level the eb occupies in the tree.
119 *
120 * Different roots are used for different purposes and may nest inside each
121 * other and they require separate keysets. As lockdep keys should be
122 * static, assign keysets according to the purpose of the root as indicated
123 * by btrfs_root->objectid. This ensures that all special purpose roots
124 * have separate keysets.
4008c04a 125 *
85d4e461
CM
126 * Lock-nesting across peer nodes is always done with the immediate parent
127 * node locked thus preventing deadlock. As lockdep doesn't know this, use
128 * subclass to avoid triggering lockdep warning in such cases.
4008c04a 129 *
85d4e461
CM
130 * The key is set by the readpage_end_io_hook after the buffer has passed
131 * csum validation but before the pages are unlocked. It is also set by
132 * btrfs_init_new_buffer on freshly allocated blocks.
4008c04a 133 *
85d4e461
CM
134 * We also add a check to make sure the highest level of the tree is the
135 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
136 * needs update as well.
4008c04a
CM
137 */
138#ifdef CONFIG_DEBUG_LOCK_ALLOC
139# if BTRFS_MAX_LEVEL != 8
140# error
141# endif
85d4e461
CM
142
143static struct btrfs_lockdep_keyset {
144 u64 id; /* root objectid */
145 const char *name_stem; /* lock name stem */
146 char names[BTRFS_MAX_LEVEL + 1][20];
147 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
148} btrfs_lockdep_keysets[] = {
149 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
150 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
151 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
152 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
153 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
154 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
60b62978 155 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
85d4e461
CM
156 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
157 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
158 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
13fd8da9 159 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
85d4e461 160 { .id = 0, .name_stem = "tree" },
4008c04a 161};
85d4e461
CM
162
163void __init btrfs_init_lockdep(void)
164{
165 int i, j;
166
167 /* initialize lockdep class names */
168 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
169 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
170
171 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
172 snprintf(ks->names[j], sizeof(ks->names[j]),
173 "btrfs-%s-%02d", ks->name_stem, j);
174 }
175}
176
177void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
178 int level)
179{
180 struct btrfs_lockdep_keyset *ks;
181
182 BUG_ON(level >= ARRAY_SIZE(ks->keys));
183
184 /* find the matching keyset, id 0 is the default entry */
185 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
186 if (ks->id == objectid)
187 break;
188
189 lockdep_set_class_and_name(&eb->lock,
190 &ks->keys[level], ks->names[level]);
191}
192
4008c04a
CM
193#endif
194
d352ac68
CM
195/*
196 * extents on the btree inode are pretty simple, there's one extent
197 * that covers the entire device
198 */
b2950863 199static struct extent_map *btree_get_extent(struct inode *inode,
306e16ce 200 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 201 int create)
7eccb903 202{
5f39d397
CM
203 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
204 struct extent_map *em;
205 int ret;
206
890871be 207 read_lock(&em_tree->lock);
d1310b2e 208 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM
209 if (em) {
210 em->bdev =
211 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 212 read_unlock(&em_tree->lock);
5f39d397 213 goto out;
a061fc8d 214 }
890871be 215 read_unlock(&em_tree->lock);
7b13b7b1 216
172ddd60 217 em = alloc_extent_map();
5f39d397
CM
218 if (!em) {
219 em = ERR_PTR(-ENOMEM);
220 goto out;
221 }
222 em->start = 0;
0afbaf8c 223 em->len = (u64)-1;
c8b97818 224 em->block_len = (u64)-1;
5f39d397 225 em->block_start = 0;
a061fc8d 226 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 227
890871be 228 write_lock(&em_tree->lock);
09a2a8f9 229 ret = add_extent_mapping(em_tree, em, 0);
5f39d397
CM
230 if (ret == -EEXIST) {
231 free_extent_map(em);
7b13b7b1 232 em = lookup_extent_mapping(em_tree, start, len);
b4f359ab 233 if (!em)
0433f20d 234 em = ERR_PTR(-EIO);
5f39d397 235 } else if (ret) {
7b13b7b1 236 free_extent_map(em);
0433f20d 237 em = ERR_PTR(ret);
5f39d397 238 }
890871be 239 write_unlock(&em_tree->lock);
7b13b7b1 240
5f39d397
CM
241out:
242 return em;
7eccb903
CM
243}
244
b0496686 245u32 btrfs_csum_data(char *data, u32 seed, size_t len)
19c00ddc 246{
0b947aff 247 return btrfs_crc32c(seed, data, len);
19c00ddc
CM
248}
249
250void btrfs_csum_final(u32 crc, char *result)
251{
7e75bf3f 252 put_unaligned_le32(~crc, result);
19c00ddc
CM
253}
254
d352ac68
CM
255/*
256 * compute the csum for a btree block, and either verify it or write it
257 * into the csum field of the block.
258 */
19c00ddc
CM
259static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
260 int verify)
261{
6c41761f 262 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
607d432d 263 char *result = NULL;
19c00ddc
CM
264 unsigned long len;
265 unsigned long cur_len;
266 unsigned long offset = BTRFS_CSUM_SIZE;
19c00ddc
CM
267 char *kaddr;
268 unsigned long map_start;
269 unsigned long map_len;
270 int err;
271 u32 crc = ~(u32)0;
607d432d 272 unsigned long inline_result;
19c00ddc
CM
273
274 len = buf->len - offset;
d397712b 275 while (len > 0) {
19c00ddc 276 err = map_private_extent_buffer(buf, offset, 32,
a6591715 277 &kaddr, &map_start, &map_len);
d397712b 278 if (err)
19c00ddc 279 return 1;
19c00ddc 280 cur_len = min(len, map_len - (offset - map_start));
b0496686 281 crc = btrfs_csum_data(kaddr + offset - map_start,
19c00ddc
CM
282 crc, cur_len);
283 len -= cur_len;
284 offset += cur_len;
19c00ddc 285 }
607d432d
JB
286 if (csum_size > sizeof(inline_result)) {
287 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
288 if (!result)
289 return 1;
290 } else {
291 result = (char *)&inline_result;
292 }
293
19c00ddc
CM
294 btrfs_csum_final(crc, result);
295
296 if (verify) {
607d432d 297 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM
298 u32 val;
299 u32 found = 0;
607d432d 300 memcpy(&found, result, csum_size);
e4204ded 301
607d432d 302 read_extent_buffer(buf, &val, 0, csum_size);
efe120a0
FH
303 printk_ratelimited(KERN_INFO
304 "BTRFS: %s checksum verify failed on %llu wanted %X found %X "
305 "level %d\n",
306 root->fs_info->sb->s_id, buf->start,
307 val, found, btrfs_header_level(buf));
607d432d
JB
308 if (result != (char *)&inline_result)
309 kfree(result);
19c00ddc
CM
310 return 1;
311 }
312 } else {
607d432d 313 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 314 }
607d432d
JB
315 if (result != (char *)&inline_result)
316 kfree(result);
19c00ddc
CM
317 return 0;
318}
319
d352ac68
CM
320/*
321 * we can't consider a given block up to date unless the transid of the
322 * block matches the transid in the parent node's pointer. This is how we
323 * detect blocks that either didn't get written at all or got written
324 * in the wrong place.
325 */
1259ab75 326static int verify_parent_transid(struct extent_io_tree *io_tree,
b9fab919
CM
327 struct extent_buffer *eb, u64 parent_transid,
328 int atomic)
1259ab75 329{
2ac55d41 330 struct extent_state *cached_state = NULL;
1259ab75 331 int ret;
a26e8c9f
JB
332 bool need_lock = (current->journal_info ==
333 (void *)BTRFS_SEND_TRANS_STUB);
1259ab75
CM
334
335 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
336 return 0;
337
b9fab919
CM
338 if (atomic)
339 return -EAGAIN;
340
a26e8c9f
JB
341 if (need_lock) {
342 btrfs_tree_read_lock(eb);
343 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
344 }
345
2ac55d41 346 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
d0082371 347 0, &cached_state);
0b32f4bb 348 if (extent_buffer_uptodate(eb) &&
1259ab75
CM
349 btrfs_header_generation(eb) == parent_transid) {
350 ret = 0;
351 goto out;
352 }
7a36ddec 353 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
193f284d 354 "found %llu\n",
c1c9ff7c 355 eb->start, parent_transid, btrfs_header_generation(eb));
1259ab75 356 ret = 1;
a26e8c9f
JB
357
358 /*
359 * Things reading via commit roots that don't have normal protection,
360 * like send, can have a really old block in cache that may point at a
361 * block that has been free'd and re-allocated. So don't clear uptodate
362 * if we find an eb that is under IO (dirty/writeback) because we could
363 * end up reading in the stale data and then writing it back out and
364 * making everybody very sad.
365 */
366 if (!extent_buffer_under_io(eb))
367 clear_extent_buffer_uptodate(eb);
33958dc6 368out:
2ac55d41
JB
369 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
370 &cached_state, GFP_NOFS);
a26e8c9f 371 btrfs_tree_read_unlock_blocking(eb);
1259ab75 372 return ret;
1259ab75
CM
373}
374
1104a885
DS
375/*
376 * Return 0 if the superblock checksum type matches the checksum value of that
377 * algorithm. Pass the raw disk superblock data.
378 */
379static int btrfs_check_super_csum(char *raw_disk_sb)
380{
381 struct btrfs_super_block *disk_sb =
382 (struct btrfs_super_block *)raw_disk_sb;
383 u16 csum_type = btrfs_super_csum_type(disk_sb);
384 int ret = 0;
385
386 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
387 u32 crc = ~(u32)0;
388 const int csum_size = sizeof(crc);
389 char result[csum_size];
390
391 /*
392 * The super_block structure does not span the whole
393 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
394 * is filled with zeros and is included in the checkum.
395 */
396 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
397 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
398 btrfs_csum_final(crc, result);
399
400 if (memcmp(raw_disk_sb, result, csum_size))
401 ret = 1;
667e7d94
CM
402
403 if (ret && btrfs_super_generation(disk_sb) < 10) {
efe120a0
FH
404 printk(KERN_WARNING
405 "BTRFS: super block crcs don't match, older mkfs detected\n");
667e7d94
CM
406 ret = 0;
407 }
1104a885
DS
408 }
409
410 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
efe120a0 411 printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
1104a885
DS
412 csum_type);
413 ret = 1;
414 }
415
416 return ret;
417}
418
d352ac68
CM
419/*
420 * helper to read a given tree block, doing retries as required when
421 * the checksums don't match and we have alternate mirrors to try.
422 */
f188591e
CM
423static int btree_read_extent_buffer_pages(struct btrfs_root *root,
424 struct extent_buffer *eb,
ca7a79ad 425 u64 start, u64 parent_transid)
f188591e
CM
426{
427 struct extent_io_tree *io_tree;
ea466794 428 int failed = 0;
f188591e
CM
429 int ret;
430 int num_copies = 0;
431 int mirror_num = 0;
ea466794 432 int failed_mirror = 0;
f188591e 433
a826d6dc 434 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
f188591e
CM
435 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
436 while (1) {
bb82ab88
AJ
437 ret = read_extent_buffer_pages(io_tree, eb, start,
438 WAIT_COMPLETE,
f188591e 439 btree_get_extent, mirror_num);
256dd1bb
SB
440 if (!ret) {
441 if (!verify_parent_transid(io_tree, eb,
b9fab919 442 parent_transid, 0))
256dd1bb
SB
443 break;
444 else
445 ret = -EIO;
446 }
d397712b 447
a826d6dc
JB
448 /*
449 * This buffer's crc is fine, but its contents are corrupted, so
450 * there is no reason to read the other copies, they won't be
451 * any less wrong.
452 */
453 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
ea466794
JB
454 break;
455
5d964051 456 num_copies = btrfs_num_copies(root->fs_info,
f188591e 457 eb->start, eb->len);
4235298e 458 if (num_copies == 1)
ea466794 459 break;
4235298e 460
5cf1ab56
JB
461 if (!failed_mirror) {
462 failed = 1;
463 failed_mirror = eb->read_mirror;
464 }
465
f188591e 466 mirror_num++;
ea466794
JB
467 if (mirror_num == failed_mirror)
468 mirror_num++;
469
4235298e 470 if (mirror_num > num_copies)
ea466794 471 break;
f188591e 472 }
ea466794 473
c0901581 474 if (failed && !ret && failed_mirror)
ea466794
JB
475 repair_eb_io_failure(root, eb, failed_mirror);
476
477 return ret;
f188591e 478}
19c00ddc 479
d352ac68 480/*
d397712b
CM
481 * checksum a dirty tree block before IO. This has extra checks to make sure
482 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 483 */
d397712b 484
b2950863 485static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
19c00ddc 486{
4eee4fa4 487 u64 start = page_offset(page);
19c00ddc 488 u64 found_start;
19c00ddc 489 struct extent_buffer *eb;
f188591e 490
4f2de97a
JB
491 eb = (struct extent_buffer *)page->private;
492 if (page != eb->pages[0])
493 return 0;
19c00ddc 494 found_start = btrfs_header_bytenr(eb);
fae7f21c 495 if (WARN_ON(found_start != start || !PageUptodate(page)))
4f2de97a 496 return 0;
19c00ddc 497 csum_tree_block(root, eb, 0);
19c00ddc
CM
498 return 0;
499}
500
2b82032c
YZ
501static int check_tree_block_fsid(struct btrfs_root *root,
502 struct extent_buffer *eb)
503{
504 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
505 u8 fsid[BTRFS_UUID_SIZE];
506 int ret = 1;
507
0a4e5586 508 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
2b82032c
YZ
509 while (fs_devices) {
510 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
511 ret = 0;
512 break;
513 }
514 fs_devices = fs_devices->seed;
515 }
516 return ret;
517}
518
a826d6dc 519#define CORRUPT(reason, eb, root, slot) \
efe120a0
FH
520 btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
521 "root=%llu, slot=%d", reason, \
c1c9ff7c 522 btrfs_header_bytenr(eb), root->objectid, slot)
a826d6dc
JB
523
524static noinline int check_leaf(struct btrfs_root *root,
525 struct extent_buffer *leaf)
526{
527 struct btrfs_key key;
528 struct btrfs_key leaf_key;
529 u32 nritems = btrfs_header_nritems(leaf);
530 int slot;
531
532 if (nritems == 0)
533 return 0;
534
535 /* Check the 0 item */
536 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
537 BTRFS_LEAF_DATA_SIZE(root)) {
538 CORRUPT("invalid item offset size pair", leaf, root, 0);
539 return -EIO;
540 }
541
542 /*
543 * Check to make sure each items keys are in the correct order and their
544 * offsets make sense. We only have to loop through nritems-1 because
545 * we check the current slot against the next slot, which verifies the
546 * next slot's offset+size makes sense and that the current's slot
547 * offset is correct.
548 */
549 for (slot = 0; slot < nritems - 1; slot++) {
550 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
551 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
552
553 /* Make sure the keys are in the right order */
554 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
555 CORRUPT("bad key order", leaf, root, slot);
556 return -EIO;
557 }
558
559 /*
560 * Make sure the offset and ends are right, remember that the
561 * item data starts at the end of the leaf and grows towards the
562 * front.
563 */
564 if (btrfs_item_offset_nr(leaf, slot) !=
565 btrfs_item_end_nr(leaf, slot + 1)) {
566 CORRUPT("slot offset bad", leaf, root, slot);
567 return -EIO;
568 }
569
570 /*
571 * Check to make sure that we don't point outside of the leaf,
572 * just incase all the items are consistent to eachother, but
573 * all point outside of the leaf.
574 */
575 if (btrfs_item_end_nr(leaf, slot) >
576 BTRFS_LEAF_DATA_SIZE(root)) {
577 CORRUPT("slot end outside of leaf", leaf, root, slot);
578 return -EIO;
579 }
580 }
581
582 return 0;
583}
584
facc8a22
MX
585static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
586 u64 phy_offset, struct page *page,
587 u64 start, u64 end, int mirror)
ce9adaa5 588{
ce9adaa5
CM
589 u64 found_start;
590 int found_level;
ce9adaa5
CM
591 struct extent_buffer *eb;
592 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 593 int ret = 0;
727011e0 594 int reads_done;
ce9adaa5 595
ce9adaa5
CM
596 if (!page->private)
597 goto out;
d397712b 598
4f2de97a 599 eb = (struct extent_buffer *)page->private;
d397712b 600
0b32f4bb
JB
601 /* the pending IO might have been the only thing that kept this buffer
602 * in memory. Make sure we have a ref for all this other checks
603 */
604 extent_buffer_get(eb);
605
606 reads_done = atomic_dec_and_test(&eb->io_pages);
727011e0
CM
607 if (!reads_done)
608 goto err;
f188591e 609
5cf1ab56 610 eb->read_mirror = mirror;
ea466794
JB
611 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
612 ret = -EIO;
613 goto err;
614 }
615
ce9adaa5 616 found_start = btrfs_header_bytenr(eb);
727011e0 617 if (found_start != eb->start) {
efe120a0 618 printk_ratelimited(KERN_INFO "BTRFS: bad tree block start "
193f284d 619 "%llu %llu\n",
c1c9ff7c 620 found_start, eb->start);
f188591e 621 ret = -EIO;
ce9adaa5
CM
622 goto err;
623 }
2b82032c 624 if (check_tree_block_fsid(root, eb)) {
efe120a0 625 printk_ratelimited(KERN_INFO "BTRFS: bad fsid on block %llu\n",
c1c9ff7c 626 eb->start);
1259ab75
CM
627 ret = -EIO;
628 goto err;
629 }
ce9adaa5 630 found_level = btrfs_header_level(eb);
1c24c3ce 631 if (found_level >= BTRFS_MAX_LEVEL) {
efe120a0 632 btrfs_info(root->fs_info, "bad tree block level %d",
1c24c3ce
JB
633 (int)btrfs_header_level(eb));
634 ret = -EIO;
635 goto err;
636 }
ce9adaa5 637
85d4e461
CM
638 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
639 eb, found_level);
4008c04a 640
ce9adaa5 641 ret = csum_tree_block(root, eb, 1);
a826d6dc 642 if (ret) {
f188591e 643 ret = -EIO;
a826d6dc
JB
644 goto err;
645 }
646
647 /*
648 * If this is a leaf block and it is corrupt, set the corrupt bit so
649 * that we don't try and read the other copies of this block, just
650 * return -EIO.
651 */
652 if (found_level == 0 && check_leaf(root, eb)) {
653 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
654 ret = -EIO;
655 }
ce9adaa5 656
0b32f4bb
JB
657 if (!ret)
658 set_extent_buffer_uptodate(eb);
ce9adaa5 659err:
79fb65a1
JB
660 if (reads_done &&
661 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 662 btree_readahead_hook(root, eb, eb->start, ret);
4bb31e92 663
53b381b3
DW
664 if (ret) {
665 /*
666 * our io error hook is going to dec the io pages
667 * again, we have to make sure it has something
668 * to decrement
669 */
670 atomic_inc(&eb->io_pages);
0b32f4bb 671 clear_extent_buffer_uptodate(eb);
53b381b3 672 }
0b32f4bb 673 free_extent_buffer(eb);
ce9adaa5 674out:
f188591e 675 return ret;
ce9adaa5
CM
676}
677
ea466794 678static int btree_io_failed_hook(struct page *page, int failed_mirror)
4bb31e92 679{
4bb31e92
AJ
680 struct extent_buffer *eb;
681 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
682
4f2de97a 683 eb = (struct extent_buffer *)page->private;
ea466794 684 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
5cf1ab56 685 eb->read_mirror = failed_mirror;
53b381b3 686 atomic_dec(&eb->io_pages);
ea466794 687 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 688 btree_readahead_hook(root, eb, eb->start, -EIO);
4bb31e92
AJ
689 return -EIO; /* we fixed nothing */
690}
691
ce9adaa5 692static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM
693{
694 struct end_io_wq *end_io_wq = bio->bi_private;
695 struct btrfs_fs_info *fs_info;
ce9adaa5 696
ce9adaa5 697 fs_info = end_io_wq->info;
ce9adaa5 698 end_io_wq->error = err;
fccb5d86 699 btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL);
d20f7043 700
7b6d91da 701 if (bio->bi_rw & REQ_WRITE) {
53b381b3 702 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA)
fccb5d86
QW
703 btrfs_queue_work(fs_info->endio_meta_write_workers,
704 &end_io_wq->work);
53b381b3 705 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE)
fccb5d86
QW
706 btrfs_queue_work(fs_info->endio_freespace_worker,
707 &end_io_wq->work);
53b381b3 708 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
fccb5d86
QW
709 btrfs_queue_work(fs_info->endio_raid56_workers,
710 &end_io_wq->work);
cad321ad 711 else
fccb5d86
QW
712 btrfs_queue_work(fs_info->endio_write_workers,
713 &end_io_wq->work);
d20f7043 714 } else {
53b381b3 715 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
fccb5d86
QW
716 btrfs_queue_work(fs_info->endio_raid56_workers,
717 &end_io_wq->work);
53b381b3 718 else if (end_io_wq->metadata)
fccb5d86
QW
719 btrfs_queue_work(fs_info->endio_meta_workers,
720 &end_io_wq->work);
d20f7043 721 else
fccb5d86
QW
722 btrfs_queue_work(fs_info->endio_workers,
723 &end_io_wq->work);
d20f7043 724 }
ce9adaa5
CM
725}
726
0cb59c99
JB
727/*
728 * For the metadata arg you want
729 *
730 * 0 - if data
731 * 1 - if normal metadta
732 * 2 - if writing to the free space cache area
53b381b3 733 * 3 - raid parity work
0cb59c99 734 */
22c59948
CM
735int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
736 int metadata)
0b86a832 737{
ce9adaa5 738 struct end_io_wq *end_io_wq;
ce9adaa5
CM
739 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
740 if (!end_io_wq)
741 return -ENOMEM;
742
743 end_io_wq->private = bio->bi_private;
744 end_io_wq->end_io = bio->bi_end_io;
22c59948 745 end_io_wq->info = info;
ce9adaa5
CM
746 end_io_wq->error = 0;
747 end_io_wq->bio = bio;
22c59948 748 end_io_wq->metadata = metadata;
ce9adaa5
CM
749
750 bio->bi_private = end_io_wq;
751 bio->bi_end_io = end_workqueue_bio;
22c59948
CM
752 return 0;
753}
754
b64a2851 755unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 756{
4854ddd0 757 unsigned long limit = min_t(unsigned long,
5cdc7ad3 758 info->thread_pool_size,
4854ddd0
CM
759 info->fs_devices->open_devices);
760 return 256 * limit;
761}
0986fe9e 762
4a69a410
CM
763static void run_one_async_start(struct btrfs_work *work)
764{
4a69a410 765 struct async_submit_bio *async;
79787eaa 766 int ret;
4a69a410
CM
767
768 async = container_of(work, struct async_submit_bio, work);
79787eaa
JM
769 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
770 async->mirror_num, async->bio_flags,
771 async->bio_offset);
772 if (ret)
773 async->error = ret;
4a69a410
CM
774}
775
776static void run_one_async_done(struct btrfs_work *work)
8b712842
CM
777{
778 struct btrfs_fs_info *fs_info;
779 struct async_submit_bio *async;
4854ddd0 780 int limit;
8b712842
CM
781
782 async = container_of(work, struct async_submit_bio, work);
783 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 784
b64a2851 785 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM
786 limit = limit * 2 / 3;
787
66657b31 788 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
b64a2851 789 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM
790 wake_up(&fs_info->async_submit_wait);
791
79787eaa
JM
792 /* If an error occured we just want to clean up the bio and move on */
793 if (async->error) {
794 bio_endio(async->bio, async->error);
795 return;
796 }
797
4a69a410 798 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM
799 async->mirror_num, async->bio_flags,
800 async->bio_offset);
4a69a410
CM
801}
802
803static void run_one_async_free(struct btrfs_work *work)
804{
805 struct async_submit_bio *async;
806
807 async = container_of(work, struct async_submit_bio, work);
8b712842
CM
808 kfree(async);
809}
810
44b8bd7e
CM
811int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
812 int rw, struct bio *bio, int mirror_num,
c8b97818 813 unsigned long bio_flags,
eaf25d93 814 u64 bio_offset,
4a69a410
CM
815 extent_submit_bio_hook_t *submit_bio_start,
816 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM
817{
818 struct async_submit_bio *async;
819
820 async = kmalloc(sizeof(*async), GFP_NOFS);
821 if (!async)
822 return -ENOMEM;
823
824 async->inode = inode;
825 async->rw = rw;
826 async->bio = bio;
827 async->mirror_num = mirror_num;
4a69a410
CM
828 async->submit_bio_start = submit_bio_start;
829 async->submit_bio_done = submit_bio_done;
830
5cdc7ad3
QW
831 btrfs_init_work(&async->work, run_one_async_start,
832 run_one_async_done, run_one_async_free);
4a69a410 833
c8b97818 834 async->bio_flags = bio_flags;
eaf25d93 835 async->bio_offset = bio_offset;
8c8bee1d 836
79787eaa
JM
837 async->error = 0;
838
cb03c743 839 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 840
7b6d91da 841 if (rw & REQ_SYNC)
5cdc7ad3 842 btrfs_set_work_high_priority(&async->work);
d313d7a3 843
5cdc7ad3 844 btrfs_queue_work(fs_info->workers, &async->work);
9473f16c 845
d397712b 846 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM
847 atomic_read(&fs_info->nr_async_submits)) {
848 wait_event(fs_info->async_submit_wait,
849 (atomic_read(&fs_info->nr_async_submits) == 0));
850 }
851
44b8bd7e
CM
852 return 0;
853}
854
ce3ed71a
CM
855static int btree_csum_one_bio(struct bio *bio)
856{
2c30c71b 857 struct bio_vec *bvec;
ce3ed71a 858 struct btrfs_root *root;
2c30c71b 859 int i, ret = 0;
ce3ed71a 860
2c30c71b 861 bio_for_each_segment_all(bvec, bio, i) {
ce3ed71a 862 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
79787eaa
JM
863 ret = csum_dirty_buffer(root, bvec->bv_page);
864 if (ret)
865 break;
ce3ed71a 866 }
2c30c71b 867
79787eaa 868 return ret;
ce3ed71a
CM
869}
870
4a69a410
CM
871static int __btree_submit_bio_start(struct inode *inode, int rw,
872 struct bio *bio, int mirror_num,
eaf25d93
CM
873 unsigned long bio_flags,
874 u64 bio_offset)
22c59948 875{
8b712842
CM
876 /*
877 * when we're called for a write, we're already in the async
5443be45 878 * submission context. Just jump into btrfs_map_bio
8b712842 879 */
79787eaa 880 return btree_csum_one_bio(bio);
4a69a410 881}
22c59948 882
4a69a410 883static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM
884 int mirror_num, unsigned long bio_flags,
885 u64 bio_offset)
4a69a410 886{
61891923
SB
887 int ret;
888
8b712842 889 /*
4a69a410
CM
890 * when we're called for a write, we're already in the async
891 * submission context. Just jump into btrfs_map_bio
8b712842 892 */
61891923
SB
893 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
894 if (ret)
895 bio_endio(bio, ret);
896 return ret;
0b86a832
CM
897}
898
de0022b9
JB
899static int check_async_write(struct inode *inode, unsigned long bio_flags)
900{
901 if (bio_flags & EXTENT_BIO_TREE_LOG)
902 return 0;
903#ifdef CONFIG_X86
904 if (cpu_has_xmm4_2)
905 return 0;
906#endif
907 return 1;
908}
909
44b8bd7e 910static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM
911 int mirror_num, unsigned long bio_flags,
912 u64 bio_offset)
44b8bd7e 913{
de0022b9 914 int async = check_async_write(inode, bio_flags);
cad321ad
CM
915 int ret;
916
7b6d91da 917 if (!(rw & REQ_WRITE)) {
4a69a410
CM
918 /*
919 * called for a read, do the setup so that checksum validation
920 * can happen in the async kernel threads
921 */
f3f266ab
CM
922 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
923 bio, 1);
1d4284bd 924 if (ret)
61891923
SB
925 goto out_w_error;
926 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
927 mirror_num, 0);
de0022b9
JB
928 } else if (!async) {
929 ret = btree_csum_one_bio(bio);
930 if (ret)
61891923
SB
931 goto out_w_error;
932 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
933 mirror_num, 0);
934 } else {
935 /*
936 * kthread helpers are used to submit writes so that
937 * checksumming can happen in parallel across all CPUs
938 */
939 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
940 inode, rw, bio, mirror_num, 0,
941 bio_offset,
942 __btree_submit_bio_start,
943 __btree_submit_bio_done);
44b8bd7e 944 }
d313d7a3 945
61891923
SB
946 if (ret) {
947out_w_error:
948 bio_endio(bio, ret);
949 }
950 return ret;
44b8bd7e
CM
951}
952
3dd1462e 953#ifdef CONFIG_MIGRATION
784b4e29 954static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG
955 struct page *newpage, struct page *page,
956 enum migrate_mode mode)
784b4e29
CM
957{
958 /*
959 * we can't safely write a btree page from here,
960 * we haven't done the locking hook
961 */
962 if (PageDirty(page))
963 return -EAGAIN;
964 /*
965 * Buffers may be managed in a filesystem specific way.
966 * We must have no buffers or drop them.
967 */
968 if (page_has_private(page) &&
969 !try_to_release_page(page, GFP_KERNEL))
970 return -EAGAIN;
a6bc32b8 971 return migrate_page(mapping, newpage, page, mode);
784b4e29 972}
3dd1462e 973#endif
784b4e29 974
0da5468f
CM
975
976static int btree_writepages(struct address_space *mapping,
977 struct writeback_control *wbc)
978{
e2d84521
MX
979 struct btrfs_fs_info *fs_info;
980 int ret;
981
d8d5f3e1 982 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM
983
984 if (wbc->for_kupdate)
985 return 0;
986
e2d84521 987 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 988 /* this is a bit racy, but that's ok */
e2d84521
MX
989 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
990 BTRFS_DIRTY_METADATA_THRESH);
991 if (ret < 0)
793955bc 992 return 0;
793955bc 993 }
0b32f4bb 994 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM
995}
996
b2950863 997static int btree_readpage(struct file *file, struct page *page)
5f39d397 998{
d1310b2e
CM
999 struct extent_io_tree *tree;
1000 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 1001 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 1002}
22b0ebda 1003
70dec807 1004static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 1005{
98509cfc 1006 if (PageWriteback(page) || PageDirty(page))
d397712b 1007 return 0;
0c4e538b 1008
f7a52a40 1009 return try_release_extent_buffer(page);
d98237b3
CM
1010}
1011
d47992f8
LC
1012static void btree_invalidatepage(struct page *page, unsigned int offset,
1013 unsigned int length)
d98237b3 1014{
d1310b2e
CM
1015 struct extent_io_tree *tree;
1016 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM
1017 extent_invalidatepage(tree, page, offset);
1018 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 1019 if (PagePrivate(page)) {
efe120a0
FH
1020 btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1021 "page private not zero on page %llu",
1022 (unsigned long long)page_offset(page));
9ad6b7bc
CM
1023 ClearPagePrivate(page);
1024 set_page_private(page, 0);
1025 page_cache_release(page);
1026 }
d98237b3
CM
1027}
1028
0b32f4bb
JB
1029static int btree_set_page_dirty(struct page *page)
1030{
bb146eb2 1031#ifdef DEBUG
0b32f4bb
JB
1032 struct extent_buffer *eb;
1033
1034 BUG_ON(!PagePrivate(page));
1035 eb = (struct extent_buffer *)page->private;
1036 BUG_ON(!eb);
1037 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1038 BUG_ON(!atomic_read(&eb->refs));
1039 btrfs_assert_tree_locked(eb);
bb146eb2 1040#endif
0b32f4bb
JB
1041 return __set_page_dirty_nobuffers(page);
1042}
1043
7f09410b 1044static const struct address_space_operations btree_aops = {
d98237b3 1045 .readpage = btree_readpage,
0da5468f 1046 .writepages = btree_writepages,
5f39d397
CM
1047 .releasepage = btree_releasepage,
1048 .invalidatepage = btree_invalidatepage,
5a92bc88 1049#ifdef CONFIG_MIGRATION
784b4e29 1050 .migratepage = btree_migratepage,
5a92bc88 1051#endif
0b32f4bb 1052 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM
1053};
1054
ca7a79ad
CM
1055int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1056 u64 parent_transid)
090d1875 1057{
5f39d397
CM
1058 struct extent_buffer *buf = NULL;
1059 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 1060 int ret = 0;
090d1875 1061
db94535d 1062 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 1063 if (!buf)
090d1875 1064 return 0;
d1310b2e 1065 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
bb82ab88 1066 buf, 0, WAIT_NONE, btree_get_extent, 0);
5f39d397 1067 free_extent_buffer(buf);
de428b63 1068 return ret;
090d1875
CM
1069}
1070
ab0fff03
AJ
1071int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1072 int mirror_num, struct extent_buffer **eb)
1073{
1074 struct extent_buffer *buf = NULL;
1075 struct inode *btree_inode = root->fs_info->btree_inode;
1076 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1077 int ret;
1078
1079 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1080 if (!buf)
1081 return 0;
1082
1083 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1084
1085 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1086 btree_get_extent, mirror_num);
1087 if (ret) {
1088 free_extent_buffer(buf);
1089 return ret;
1090 }
1091
1092 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1093 free_extent_buffer(buf);
1094 return -EIO;
0b32f4bb 1095 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ
1096 *eb = buf;
1097 } else {
1098 free_extent_buffer(buf);
1099 }
1100 return 0;
1101}
1102
0999df54
CM
1103struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1104 u64 bytenr, u32 blocksize)
1105{
f28491e0 1106 return find_extent_buffer(root->fs_info, bytenr);
0999df54
CM
1107}
1108
1109struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1110 u64 bytenr, u32 blocksize)
1111{
f28491e0 1112 return alloc_extent_buffer(root->fs_info, bytenr, blocksize);
0999df54
CM
1113}
1114
1115
e02119d5
CM
1116int btrfs_write_tree_block(struct extent_buffer *buf)
1117{
727011e0 1118 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1119 buf->start + buf->len - 1);
e02119d5
CM
1120}
1121
1122int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1123{
727011e0 1124 return filemap_fdatawait_range(buf->pages[0]->mapping,
8aa38c31 1125 buf->start, buf->start + buf->len - 1);
e02119d5
CM
1126}
1127
0999df54 1128struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 1129 u32 blocksize, u64 parent_transid)
0999df54
CM
1130{
1131 struct extent_buffer *buf = NULL;
0999df54
CM
1132 int ret;
1133
0999df54
CM
1134 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1135 if (!buf)
1136 return NULL;
0999df54 1137
ca7a79ad 1138 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
0f0fe8f7
FDBM
1139 if (ret) {
1140 free_extent_buffer(buf);
1141 return NULL;
1142 }
5f39d397 1143 return buf;
ce9adaa5 1144
eb60ceac
CM
1145}
1146
d5c13f92
JM
1147void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1148 struct extent_buffer *buf)
ed2ff2cb 1149{
e2d84521
MX
1150 struct btrfs_fs_info *fs_info = root->fs_info;
1151
55c69072 1152 if (btrfs_header_generation(buf) ==
e2d84521 1153 fs_info->running_transaction->transid) {
b9447ef8 1154 btrfs_assert_tree_locked(buf);
b4ce94de 1155
b9473439 1156 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
e2d84521
MX
1157 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1158 -buf->len,
1159 fs_info->dirty_metadata_batch);
ed7b63eb
JB
1160 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1161 btrfs_set_lock_blocking(buf);
1162 clear_extent_buffer_dirty(buf);
1163 }
925baedd 1164 }
5f39d397
CM
1165}
1166
8257b2dc
MX
1167static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1168{
1169 struct btrfs_subvolume_writers *writers;
1170 int ret;
1171
1172 writers = kmalloc(sizeof(*writers), GFP_NOFS);
1173 if (!writers)
1174 return ERR_PTR(-ENOMEM);
1175
1176 ret = percpu_counter_init(&writers->counter, 0);
1177 if (ret < 0) {
1178 kfree(writers);
1179 return ERR_PTR(ret);
1180 }
1181
1182 init_waitqueue_head(&writers->wait);
1183 return writers;
1184}
1185
1186static void
1187btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1188{
1189 percpu_counter_destroy(&writers->counter);
1190 kfree(writers);
1191}
1192
143bede5
JM
1193static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1194 u32 stripesize, struct btrfs_root *root,
1195 struct btrfs_fs_info *fs_info,
1196 u64 objectid)
d97e63b6 1197{
cfaa7295 1198 root->node = NULL;
a28ec197 1199 root->commit_root = NULL;
db94535d
CM
1200 root->sectorsize = sectorsize;
1201 root->nodesize = nodesize;
1202 root->leafsize = leafsize;
87ee04eb 1203 root->stripesize = stripesize;
27cdeb70 1204 root->state = 0;
d68fc57b 1205 root->orphan_cleanup_state = 0;
0b86a832 1206
0f7d52f4
CM
1207 root->objectid = objectid;
1208 root->last_trans = 0;
13a8a7c8 1209 root->highest_objectid = 0;
eb73c1b7 1210 root->nr_delalloc_inodes = 0;
199c2a9c 1211 root->nr_ordered_extents = 0;
58176a96 1212 root->name = NULL;
6bef4d31 1213 root->inode_tree = RB_ROOT;
16cdcec7 1214 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1215 root->block_rsv = NULL;
d68fc57b 1216 root->orphan_block_rsv = NULL;
0b86a832
CM
1217
1218 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1219 INIT_LIST_HEAD(&root->root_list);
eb73c1b7
MX
1220 INIT_LIST_HEAD(&root->delalloc_inodes);
1221 INIT_LIST_HEAD(&root->delalloc_root);
199c2a9c
MX
1222 INIT_LIST_HEAD(&root->ordered_extents);
1223 INIT_LIST_HEAD(&root->ordered_root);
2ab28f32
JB
1224 INIT_LIST_HEAD(&root->logged_list[0]);
1225 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1226 spin_lock_init(&root->orphan_lock);
5d4f98a2 1227 spin_lock_init(&root->inode_lock);
eb73c1b7 1228 spin_lock_init(&root->delalloc_lock);
199c2a9c 1229 spin_lock_init(&root->ordered_extent_lock);
f0486c68 1230 spin_lock_init(&root->accounting_lock);
2ab28f32
JB
1231 spin_lock_init(&root->log_extents_lock[0]);
1232 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1233 mutex_init(&root->objectid_mutex);
e02119d5 1234 mutex_init(&root->log_mutex);
31f3d255 1235 mutex_init(&root->ordered_extent_mutex);
573bfb72 1236 mutex_init(&root->delalloc_mutex);
7237f183
YZ
1237 init_waitqueue_head(&root->log_writer_wait);
1238 init_waitqueue_head(&root->log_commit_wait[0]);
1239 init_waitqueue_head(&root->log_commit_wait[1]);
8b050d35
MX
1240 INIT_LIST_HEAD(&root->log_ctxs[0]);
1241 INIT_LIST_HEAD(&root->log_ctxs[1]);
7237f183
YZ
1242 atomic_set(&root->log_commit[0], 0);
1243 atomic_set(&root->log_commit[1], 0);
1244 atomic_set(&root->log_writers, 0);
2ecb7923 1245 atomic_set(&root->log_batch, 0);
8a35d95f 1246 atomic_set(&root->orphan_inodes, 0);
b0feb9d9 1247 atomic_set(&root->refs, 1);
8257b2dc 1248 atomic_set(&root->will_be_snapshoted, 0);
7237f183 1249 root->log_transid = 0;
d1433deb 1250 root->log_transid_committed = -1;
257c62e1 1251 root->last_log_commit = 0;
06ea65a3
JB
1252 if (fs_info)
1253 extent_io_tree_init(&root->dirty_log_pages,
1254 fs_info->btree_inode->i_mapping);
017e5369 1255
3768f368
CM
1256 memset(&root->root_key, 0, sizeof(root->root_key));
1257 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1258 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1259 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
06ea65a3
JB
1260 if (fs_info)
1261 root->defrag_trans_start = fs_info->generation;
1262 else
1263 root->defrag_trans_start = 0;
58176a96 1264 init_completion(&root->kobj_unregister);
4d775673 1265 root->root_key.objectid = objectid;
0ee5dc67 1266 root->anon_dev = 0;
8ea05e3a 1267
5f3ab90a 1268 spin_lock_init(&root->root_item_lock);
3768f368
CM
1269}
1270
f84a8bd6 1271static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
6f07e42e
AV
1272{
1273 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1274 if (root)
1275 root->fs_info = fs_info;
1276 return root;
1277}
1278
06ea65a3
JB
1279#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1280/* Should only be used by the testing infrastructure */
1281struct btrfs_root *btrfs_alloc_dummy_root(void)
1282{
1283 struct btrfs_root *root;
1284
1285 root = btrfs_alloc_root(NULL);
1286 if (!root)
1287 return ERR_PTR(-ENOMEM);
1288 __setup_root(4096, 4096, 4096, 4096, root, NULL, 1);
27cdeb70 1289 set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state);
06ea65a3
JB
1290
1291 return root;
1292}
1293#endif
1294
20897f5c
AJ
1295struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1296 struct btrfs_fs_info *fs_info,
1297 u64 objectid)
1298{
1299 struct extent_buffer *leaf;
1300 struct btrfs_root *tree_root = fs_info->tree_root;
1301 struct btrfs_root *root;
1302 struct btrfs_key key;
1303 int ret = 0;
6463fe58 1304 uuid_le uuid;
20897f5c
AJ
1305
1306 root = btrfs_alloc_root(fs_info);
1307 if (!root)
1308 return ERR_PTR(-ENOMEM);
1309
1310 __setup_root(tree_root->nodesize, tree_root->leafsize,
1311 tree_root->sectorsize, tree_root->stripesize,
1312 root, fs_info, objectid);
1313 root->root_key.objectid = objectid;
1314 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1315 root->root_key.offset = 0;
1316
1317 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1318 0, objectid, NULL, 0, 0, 0);
1319 if (IS_ERR(leaf)) {
1320 ret = PTR_ERR(leaf);
1dd05682 1321 leaf = NULL;
20897f5c
AJ
1322 goto fail;
1323 }
1324
20897f5c
AJ
1325 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1326 btrfs_set_header_bytenr(leaf, leaf->start);
1327 btrfs_set_header_generation(leaf, trans->transid);
1328 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1329 btrfs_set_header_owner(leaf, objectid);
1330 root->node = leaf;
1331
0a4e5586 1332 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
20897f5c
AJ
1333 BTRFS_FSID_SIZE);
1334 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
b308bc2f 1335 btrfs_header_chunk_tree_uuid(leaf),
20897f5c
AJ
1336 BTRFS_UUID_SIZE);
1337 btrfs_mark_buffer_dirty(leaf);
1338
1339 root->commit_root = btrfs_root_node(root);
27cdeb70 1340 set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
20897f5c
AJ
1341
1342 root->root_item.flags = 0;
1343 root->root_item.byte_limit = 0;
1344 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1345 btrfs_set_root_generation(&root->root_item, trans->transid);
1346 btrfs_set_root_level(&root->root_item, 0);
1347 btrfs_set_root_refs(&root->root_item, 1);
1348 btrfs_set_root_used(&root->root_item, leaf->len);
1349 btrfs_set_root_last_snapshot(&root->root_item, 0);
1350 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB
1351 uuid_le_gen(&uuid);
1352 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ
1353 root->root_item.drop_level = 0;
1354
1355 key.objectid = objectid;
1356 key.type = BTRFS_ROOT_ITEM_KEY;
1357 key.offset = 0;
1358 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1359 if (ret)
1360 goto fail;
1361
1362 btrfs_tree_unlock(leaf);
1363
1dd05682
TI
1364 return root;
1365
20897f5c 1366fail:
1dd05682
TI
1367 if (leaf) {
1368 btrfs_tree_unlock(leaf);
1369 free_extent_buffer(leaf);
1370 }
1371 kfree(root);
20897f5c 1372
1dd05682 1373 return ERR_PTR(ret);
20897f5c
AJ
1374}
1375
7237f183
YZ
1376static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1377 struct btrfs_fs_info *fs_info)
0f7d52f4
CM
1378{
1379 struct btrfs_root *root;
1380 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1381 struct extent_buffer *leaf;
e02119d5 1382
6f07e42e 1383 root = btrfs_alloc_root(fs_info);
e02119d5 1384 if (!root)
7237f183 1385 return ERR_PTR(-ENOMEM);
e02119d5
CM
1386
1387 __setup_root(tree_root->nodesize, tree_root->leafsize,
1388 tree_root->sectorsize, tree_root->stripesize,
1389 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1390
1391 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1392 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1393 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
27cdeb70 1394
7237f183 1395 /*
27cdeb70
MX
1396 * DON'T set REF_COWS for log trees
1397 *
7237f183
YZ
1398 * log trees do not get reference counted because they go away
1399 * before a real commit is actually done. They do store pointers
1400 * to file data extents, and those reference counts still get
1401 * updated (along with back refs to the log tree).
1402 */
e02119d5 1403
5d4f98a2 1404 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
66d7e7f0 1405 BTRFS_TREE_LOG_OBJECTID, NULL,
5581a51a 1406 0, 0, 0);
7237f183
YZ
1407 if (IS_ERR(leaf)) {
1408 kfree(root);
1409 return ERR_CAST(leaf);
1410 }
e02119d5 1411
5d4f98a2
YZ
1412 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1413 btrfs_set_header_bytenr(leaf, leaf->start);
1414 btrfs_set_header_generation(leaf, trans->transid);
1415 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1416 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1417 root->node = leaf;
e02119d5
CM
1418
1419 write_extent_buffer(root->node, root->fs_info->fsid,
0a4e5586 1420 btrfs_header_fsid(), BTRFS_FSID_SIZE);
e02119d5
CM
1421 btrfs_mark_buffer_dirty(root->node);
1422 btrfs_tree_unlock(root->node);
7237f183
YZ
1423 return root;
1424}
1425
1426int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1427 struct btrfs_fs_info *fs_info)
1428{
1429 struct btrfs_root *log_root;
1430
1431 log_root = alloc_log_tree(trans, fs_info);
1432 if (IS_ERR(log_root))
1433 return PTR_ERR(log_root);
1434 WARN_ON(fs_info->log_root_tree);
1435 fs_info->log_root_tree = log_root;
1436 return 0;
1437}
1438
1439int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1440 struct btrfs_root *root)
1441{
1442 struct btrfs_root *log_root;
1443 struct btrfs_inode_item *inode_item;
1444
1445 log_root = alloc_log_tree(trans, root->fs_info);
1446 if (IS_ERR(log_root))
1447 return PTR_ERR(log_root);
1448
1449 log_root->last_trans = trans->transid;
1450 log_root->root_key.offset = root->root_key.objectid;
1451
1452 inode_item = &log_root->root_item.inode;
3cae210f
QW
1453 btrfs_set_stack_inode_generation(inode_item, 1);
1454 btrfs_set_stack_inode_size(inode_item, 3);
1455 btrfs_set_stack_inode_nlink(inode_item, 1);
1456 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
1457 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
7237f183 1458
5d4f98a2 1459 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ
1460
1461 WARN_ON(root->log_root);
1462 root->log_root = log_root;
1463 root->log_transid = 0;
d1433deb 1464 root->log_transid_committed = -1;
257c62e1 1465 root->last_log_commit = 0;
e02119d5
CM
1466 return 0;
1467}
1468
35a3621b
SB
1469static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1470 struct btrfs_key *key)
e02119d5
CM
1471{
1472 struct btrfs_root *root;
1473 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1474 struct btrfs_path *path;
84234f3a 1475 u64 generation;
db94535d 1476 u32 blocksize;
cb517eab 1477 int ret;
0f7d52f4 1478
cb517eab
MX
1479 path = btrfs_alloc_path();
1480 if (!path)
0f7d52f4 1481 return ERR_PTR(-ENOMEM);
cb517eab
MX
1482
1483 root = btrfs_alloc_root(fs_info);
1484 if (!root) {
1485 ret = -ENOMEM;
1486 goto alloc_fail;
0f7d52f4
CM
1487 }
1488
db94535d 1489 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb 1490 tree_root->sectorsize, tree_root->stripesize,
cb517eab 1491 root, fs_info, key->objectid);
0f7d52f4 1492
cb517eab
MX
1493 ret = btrfs_find_root(tree_root, key, path,
1494 &root->root_item, &root->root_key);
0f7d52f4 1495 if (ret) {
13a8a7c8
YZ
1496 if (ret > 0)
1497 ret = -ENOENT;
cb517eab 1498 goto find_fail;
0f7d52f4 1499 }
13a8a7c8 1500
84234f3a 1501 generation = btrfs_root_generation(&root->root_item);
db94535d
CM
1502 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1503 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1504 blocksize, generation);
cb517eab
MX
1505 if (!root->node) {
1506 ret = -ENOMEM;
1507 goto find_fail;
1508 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1509 ret = -EIO;
1510 goto read_fail;
416bc658 1511 }
5d4f98a2 1512 root->commit_root = btrfs_root_node(root);
13a8a7c8 1513out:
cb517eab
MX
1514 btrfs_free_path(path);
1515 return root;
1516
1517read_fail:
1518 free_extent_buffer(root->node);
1519find_fail:
1520 kfree(root);
1521alloc_fail:
1522 root = ERR_PTR(ret);
1523 goto out;
1524}
1525
1526struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1527 struct btrfs_key *location)
1528{
1529 struct btrfs_root *root;
1530
1531 root = btrfs_read_tree_root(tree_root, location);
1532 if (IS_ERR(root))
1533 return root;
1534
1535 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
27cdeb70 1536 set_bit(BTRFS_ROOT_REF_COWS, &root->state);
08fe4db1
LZ
1537 btrfs_check_and_init_root_item(&root->root_item);
1538 }
13a8a7c8 1539
5eda7b5e
CM
1540 return root;
1541}
1542
cb517eab
MX
1543int btrfs_init_fs_root(struct btrfs_root *root)
1544{
1545 int ret;
8257b2dc 1546 struct btrfs_subvolume_writers *writers;
cb517eab
MX
1547
1548 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1549 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1550 GFP_NOFS);
1551 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1552 ret = -ENOMEM;
1553 goto fail;
1554 }
1555
8257b2dc
MX
1556 writers = btrfs_alloc_subvolume_writers();
1557 if (IS_ERR(writers)) {
1558 ret = PTR_ERR(writers);
1559 goto fail;
1560 }
1561 root->subv_writers = writers;
1562
cb517eab 1563 btrfs_init_free_ino_ctl(root);
cb517eab
MX
1564 spin_lock_init(&root->cache_lock);
1565 init_waitqueue_head(&root->cache_wait);
1566
1567 ret = get_anon_bdev(&root->anon_dev);
1568 if (ret)
8257b2dc 1569 goto free_writers;
cb517eab 1570 return 0;
8257b2dc
MX
1571
1572free_writers:
1573 btrfs_free_subvolume_writers(root->subv_writers);
cb517eab
MX
1574fail:
1575 kfree(root->free_ino_ctl);
1576 kfree(root->free_ino_pinned);
1577 return ret;
1578}
1579
171170c1
ST
1580static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1581 u64 root_id)
cb517eab
MX
1582{
1583 struct btrfs_root *root;
1584
1585 spin_lock(&fs_info->fs_roots_radix_lock);
1586 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1587 (unsigned long)root_id);
1588 spin_unlock(&fs_info->fs_roots_radix_lock);
1589 return root;
1590}
1591
1592int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1593 struct btrfs_root *root)
1594{
1595 int ret;
1596
1597 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1598 if (ret)
1599 return ret;
1600
1601 spin_lock(&fs_info->fs_roots_radix_lock);
1602 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1603 (unsigned long)root->root_key.objectid,
1604 root);
1605 if (ret == 0)
27cdeb70 1606 set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
cb517eab
MX
1607 spin_unlock(&fs_info->fs_roots_radix_lock);
1608 radix_tree_preload_end();
1609
1610 return ret;
1611}
1612
c00869f1
MX
1613struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1614 struct btrfs_key *location,
1615 bool check_ref)
5eda7b5e
CM
1616{
1617 struct btrfs_root *root;
1618 int ret;
1619
edbd8d4e
CM
1620 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1621 return fs_info->tree_root;
1622 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1623 return fs_info->extent_root;
8f18cf13
CM
1624 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1625 return fs_info->chunk_root;
1626 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1627 return fs_info->dev_root;
0403e47e
YZ
1628 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1629 return fs_info->csum_root;
bcef60f2
AJ
1630 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1631 return fs_info->quota_root ? fs_info->quota_root :
1632 ERR_PTR(-ENOENT);
f7a81ea4
SB
1633 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1634 return fs_info->uuid_root ? fs_info->uuid_root :
1635 ERR_PTR(-ENOENT);
4df27c4d 1636again:
cb517eab 1637 root = btrfs_lookup_fs_root(fs_info, location->objectid);
48475471 1638 if (root) {
c00869f1 1639 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
48475471 1640 return ERR_PTR(-ENOENT);
5eda7b5e 1641 return root;
48475471 1642 }
5eda7b5e 1643
cb517eab 1644 root = btrfs_read_fs_root(fs_info->tree_root, location);
5eda7b5e
CM
1645 if (IS_ERR(root))
1646 return root;
3394e160 1647
c00869f1 1648 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
cb517eab 1649 ret = -ENOENT;
581bb050 1650 goto fail;
35a30d7c 1651 }
581bb050 1652
cb517eab 1653 ret = btrfs_init_fs_root(root);
ac08aedf
CM
1654 if (ret)
1655 goto fail;
3394e160 1656
3f870c28
KN
1657 ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID,
1658 location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL);
d68fc57b
YZ
1659 if (ret < 0)
1660 goto fail;
1661 if (ret == 0)
27cdeb70 1662 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
d68fc57b 1663
cb517eab 1664 ret = btrfs_insert_fs_root(fs_info, root);
0f7d52f4 1665 if (ret) {
4df27c4d
YZ
1666 if (ret == -EEXIST) {
1667 free_fs_root(root);
1668 goto again;
1669 }
1670 goto fail;
0f7d52f4 1671 }
edbd8d4e 1672 return root;
4df27c4d
YZ
1673fail:
1674 free_fs_root(root);
1675 return ERR_PTR(ret);
edbd8d4e
CM
1676}
1677
04160088
CM
1678static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1679{
1680 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1681 int ret = 0;
04160088
CM
1682 struct btrfs_device *device;
1683 struct backing_dev_info *bdi;
b7967db7 1684
1f78160c
XG
1685 rcu_read_lock();
1686 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM
1687 if (!device->bdev)
1688 continue;
04160088
CM
1689 bdi = blk_get_backing_dev_info(device->bdev);
1690 if (bdi && bdi_congested(bdi, bdi_bits)) {
1691 ret = 1;
1692 break;
1693 }
1694 }
1f78160c 1695 rcu_read_unlock();
04160088
CM
1696 return ret;
1697}
1698
ad081f14
JA
1699/*
1700 * If this fails, caller must call bdi_destroy() to get rid of the
1701 * bdi again.
1702 */
04160088
CM
1703static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1704{
ad081f14
JA
1705 int err;
1706
1707 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1708 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA
1709 if (err)
1710 return err;
1711
4575c9cc 1712 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM
1713 bdi->congested_fn = btrfs_congested_fn;
1714 bdi->congested_data = info;
1715 return 0;
1716}
1717
8b712842
CM
1718/*
1719 * called by the kthread helper functions to finally call the bio end_io
1720 * functions. This is where read checksum verification actually happens
1721 */
1722static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1723{
ce9adaa5 1724 struct bio *bio;
8b712842 1725 struct end_io_wq *end_io_wq;
ce9adaa5 1726 int error;
ce9adaa5 1727
8b712842
CM
1728 end_io_wq = container_of(work, struct end_io_wq, work);
1729 bio = end_io_wq->bio;
ce9adaa5 1730
8b712842
CM
1731 error = end_io_wq->error;
1732 bio->bi_private = end_io_wq->private;
1733 bio->bi_end_io = end_io_wq->end_io;
1734 kfree(end_io_wq);
bc1e79ac 1735 bio_endio_nodec(bio, error);
44b8bd7e
CM
1736}
1737
a74a4b97
CM
1738static int cleaner_kthread(void *arg)
1739{
1740 struct btrfs_root *root = arg;
d0278245 1741 int again;
a74a4b97
CM
1742
1743 do {
d0278245 1744 again = 0;
a74a4b97 1745
d0278245 1746 /* Make the cleaner go to sleep early. */
babbf170 1747 if (btrfs_need_cleaner_sleep(root))
d0278245
MX
1748 goto sleep;
1749
1750 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1751 goto sleep;
1752
dc7f370c
MX
1753 /*
1754 * Avoid the problem that we change the status of the fs
1755 * during the above check and trylock.
1756 */
babbf170 1757 if (btrfs_need_cleaner_sleep(root)) {
dc7f370c
MX
1758 mutex_unlock(&root->fs_info->cleaner_mutex);
1759 goto sleep;
76dda93c 1760 }
a74a4b97 1761
d0278245
MX
1762 btrfs_run_delayed_iputs(root);
1763 again = btrfs_clean_one_deleted_snapshot(root);
1764 mutex_unlock(&root->fs_info->cleaner_mutex);
1765
1766 /*
05323cd1
MX
1767 * The defragger has dealt with the R/O remount and umount,
1768 * needn't do anything special here.
d0278245
MX
1769 */
1770 btrfs_run_defrag_inodes(root->fs_info);
1771sleep:
9d1a2a3a 1772 if (!try_to_freeze() && !again) {
a74a4b97 1773 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ
1774 if (!kthread_should_stop())
1775 schedule();
a74a4b97
CM
1776 __set_current_state(TASK_RUNNING);
1777 }
1778 } while (!kthread_should_stop());
1779 return 0;
1780}
1781
1782static int transaction_kthread(void *arg)
1783{
1784 struct btrfs_root *root = arg;
1785 struct btrfs_trans_handle *trans;
1786 struct btrfs_transaction *cur;
8929ecfa 1787 u64 transid;
a74a4b97
CM
1788 unsigned long now;
1789 unsigned long delay;
914b2007 1790 bool cannot_commit;
a74a4b97
CM
1791
1792 do {
914b2007 1793 cannot_commit = false;
8b87dc17 1794 delay = HZ * root->fs_info->commit_interval;
a74a4b97
CM
1795 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1796
a4abeea4 1797 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM
1798 cur = root->fs_info->running_transaction;
1799 if (!cur) {
a4abeea4 1800 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM
1801 goto sleep;
1802 }
31153d81 1803
a74a4b97 1804 now = get_seconds();
4a9d8bde 1805 if (cur->state < TRANS_STATE_BLOCKED &&
8b87dc17
DS
1806 (now < cur->start_time ||
1807 now - cur->start_time < root->fs_info->commit_interval)) {
a4abeea4 1808 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM
1809 delay = HZ * 5;
1810 goto sleep;
1811 }
8929ecfa 1812 transid = cur->transid;
a4abeea4 1813 spin_unlock(&root->fs_info->trans_lock);
56bec294 1814
79787eaa 1815 /* If the file system is aborted, this will always fail. */
354aa0fb 1816 trans = btrfs_attach_transaction(root);
914b2007 1817 if (IS_ERR(trans)) {
354aa0fb
MX
1818 if (PTR_ERR(trans) != -ENOENT)
1819 cannot_commit = true;
79787eaa 1820 goto sleep;
914b2007 1821 }
8929ecfa 1822 if (transid == trans->transid) {
79787eaa 1823 btrfs_commit_transaction(trans, root);
8929ecfa
YZ
1824 } else {
1825 btrfs_end_transaction(trans, root);
1826 }
a74a4b97
CM
1827sleep:
1828 wake_up_process(root->fs_info->cleaner_kthread);
1829 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1830
4e121c06
JB
1831 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1832 &root->fs_info->fs_state)))
1833 btrfs_cleanup_transaction(root);
a0acae0e 1834 if (!try_to_freeze()) {
a74a4b97 1835 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa 1836 if (!kthread_should_stop() &&
914b2007
JK
1837 (!btrfs_transaction_blocked(root->fs_info) ||
1838 cannot_commit))
8929ecfa 1839 schedule_timeout(delay);
a74a4b97
CM
1840 __set_current_state(TASK_RUNNING);
1841 }
1842 } while (!kthread_should_stop());
1843 return 0;
1844}
1845
af31f5e5
CM
1846/*
1847 * this will find the highest generation in the array of
1848 * root backups. The index of the highest array is returned,
1849 * or -1 if we can't find anything.
1850 *
1851 * We check to make sure the array is valid by comparing the
1852 * generation of the latest root in the array with the generation
1853 * in the super block. If they don't match we pitch it.
1854 */
1855static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1856{
1857 u64 cur;
1858 int newest_index = -1;
1859 struct btrfs_root_backup *root_backup;
1860 int i;
1861
1862 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1863 root_backup = info->super_copy->super_roots + i;
1864 cur = btrfs_backup_tree_root_gen(root_backup);
1865 if (cur == newest_gen)
1866 newest_index = i;
1867 }
1868
1869 /* check to see if we actually wrapped around */
1870 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1871 root_backup = info->super_copy->super_roots;
1872 cur = btrfs_backup_tree_root_gen(root_backup);
1873 if (cur == newest_gen)
1874 newest_index = 0;
1875 }
1876 return newest_index;
1877}
1878
1879
1880/*
1881 * find the oldest backup so we know where to store new entries
1882 * in the backup array. This will set the backup_root_index
1883 * field in the fs_info struct
1884 */
1885static void find_oldest_super_backup(struct btrfs_fs_info *info,
1886 u64 newest_gen)
1887{
1888 int newest_index = -1;
1889
1890 newest_index = find_newest_super_backup(info, newest_gen);
1891 /* if there was garbage in there, just move along */
1892 if (newest_index == -1) {
1893 info->backup_root_index = 0;
1894 } else {
1895 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1896 }
1897}
1898
1899/*
1900 * copy all the root pointers into the super backup array.
1901 * this will bump the backup pointer by one when it is
1902 * done
1903 */
1904static void backup_super_roots(struct btrfs_fs_info *info)
1905{
1906 int next_backup;
1907 struct btrfs_root_backup *root_backup;
1908 int last_backup;
1909
1910 next_backup = info->backup_root_index;
1911 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1912 BTRFS_NUM_BACKUP_ROOTS;
1913
1914 /*
1915 * just overwrite the last backup if we're at the same generation
1916 * this happens only at umount
1917 */
1918 root_backup = info->super_for_commit->super_roots + last_backup;
1919 if (btrfs_backup_tree_root_gen(root_backup) ==
1920 btrfs_header_generation(info->tree_root->node))
1921 next_backup = last_backup;
1922
1923 root_backup = info->super_for_commit->super_roots + next_backup;
1924
1925 /*
1926 * make sure all of our padding and empty slots get zero filled
1927 * regardless of which ones we use today
1928 */
1929 memset(root_backup, 0, sizeof(*root_backup));
1930
1931 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1932
1933 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1934 btrfs_set_backup_tree_root_gen(root_backup,
1935 btrfs_header_generation(info->tree_root->node));
1936
1937 btrfs_set_backup_tree_root_level(root_backup,
1938 btrfs_header_level(info->tree_root->node));
1939
1940 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1941 btrfs_set_backup_chunk_root_gen(root_backup,
1942 btrfs_header_generation(info->chunk_root->node));
1943 btrfs_set_backup_chunk_root_level(root_backup,
1944 btrfs_header_level(info->chunk_root->node));
1945
1946 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1947 btrfs_set_backup_extent_root_gen(root_backup,
1948 btrfs_header_generation(info->extent_root->node));
1949 btrfs_set_backup_extent_root_level(root_backup,
1950 btrfs_header_level(info->extent_root->node));
1951
7c7e82a7
CM
1952 /*
1953 * we might commit during log recovery, which happens before we set
1954 * the fs_root. Make sure it is valid before we fill it in.
1955 */
1956 if (info->fs_root && info->fs_root->node) {
1957 btrfs_set_backup_fs_root(root_backup,
1958 info->fs_root->node->start);
1959 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 1960 btrfs_header_generation(info->fs_root->node));
7c7e82a7 1961 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 1962 btrfs_header_level(info->fs_root->node));
7c7e82a7 1963 }
af31f5e5
CM
1964
1965 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1966 btrfs_set_backup_dev_root_gen(root_backup,
1967 btrfs_header_generation(info->dev_root->node));
1968 btrfs_set_backup_dev_root_level(root_backup,
1969 btrfs_header_level(info->dev_root->node));
1970
1971 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1972 btrfs_set_backup_csum_root_gen(root_backup,
1973 btrfs_header_generation(info->csum_root->node));
1974 btrfs_set_backup_csum_root_level(root_backup,
1975 btrfs_header_level(info->csum_root->node));
1976
1977 btrfs_set_backup_total_bytes(root_backup,
1978 btrfs_super_total_bytes(info->super_copy));
1979 btrfs_set_backup_bytes_used(root_backup,
1980 btrfs_super_bytes_used(info->super_copy));
1981 btrfs_set_backup_num_devices(root_backup,
1982 btrfs_super_num_devices(info->super_copy));
1983
1984 /*
1985 * if we don't copy this out to the super_copy, it won't get remembered
1986 * for the next commit
1987 */
1988 memcpy(&info->super_copy->super_roots,
1989 &info->super_for_commit->super_roots,
1990 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1991}
1992
1993/*
1994 * this copies info out of the root backup array and back into
1995 * the in-memory super block. It is meant to help iterate through
1996 * the array, so you send it the number of backups you've already
1997 * tried and the last backup index you used.
1998 *
1999 * this returns -1 when it has tried all the backups
2000 */
2001static noinline int next_root_backup(struct btrfs_fs_info *info,
2002 struct btrfs_super_block *super,
2003 int *num_backups_tried, int *backup_index)
2004{
2005 struct btrfs_root_backup *root_backup;
2006 int newest = *backup_index;
2007
2008 if (*num_backups_tried == 0) {
2009 u64 gen = btrfs_super_generation(super);
2010
2011 newest = find_newest_super_backup(info, gen);
2012 if (newest == -1)
2013 return -1;
2014
2015 *backup_index = newest;
2016 *num_backups_tried = 1;
2017 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2018 /* we've tried all the backups, all done */
2019 return -1;
2020 } else {
2021 /* jump to the next oldest backup */
2022 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2023 BTRFS_NUM_BACKUP_ROOTS;
2024 *backup_index = newest;
2025 *num_backups_tried += 1;
2026 }
2027 root_backup = super->super_roots + newest;
2028
2029 btrfs_set_super_generation(super,
2030 btrfs_backup_tree_root_gen(root_backup));
2031 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2032 btrfs_set_super_root_level(super,
2033 btrfs_backup_tree_root_level(root_backup));
2034 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2035
2036 /*
2037 * fixme: the total bytes and num_devices need to match or we should
2038 * need a fsck
2039 */
2040 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2041 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2042 return 0;
2043}
2044
7abadb64
LB
2045/* helper to cleanup workers */
2046static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2047{
dc6e3209 2048 btrfs_destroy_workqueue(fs_info->fixup_workers);
afe3d242 2049 btrfs_destroy_workqueue(fs_info->delalloc_workers);
5cdc7ad3 2050 btrfs_destroy_workqueue(fs_info->workers);
fccb5d86
QW
2051 btrfs_destroy_workqueue(fs_info->endio_workers);
2052 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2053 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
d05a33ac 2054 btrfs_destroy_workqueue(fs_info->rmw_workers);
fccb5d86
QW
2055 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2056 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2057 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
a8c93d4e 2058 btrfs_destroy_workqueue(fs_info->submit_workers);
5b3bc44e 2059 btrfs_destroy_workqueue(fs_info->delayed_workers);
e66f0bb1 2060 btrfs_destroy_workqueue(fs_info->caching_workers);
736cfa15 2061 btrfs_destroy_workqueue(fs_info->readahead_workers);
a44903ab 2062 btrfs_destroy_workqueue(fs_info->flush_workers);
fc97fab0 2063 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
7abadb64
LB
2064}
2065
2e9f5954
R
2066static void free_root_extent_buffers(struct btrfs_root *root)
2067{
2068 if (root) {
2069 free_extent_buffer(root->node);
2070 free_extent_buffer(root->commit_root);
2071 root->node = NULL;
2072 root->commit_root = NULL;
2073 }
2074}
2075
af31f5e5
CM
2076/* helper to cleanup tree roots */
2077static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2078{
2e9f5954 2079 free_root_extent_buffers(info->tree_root);
655b09fe 2080
2e9f5954
R
2081 free_root_extent_buffers(info->dev_root);
2082 free_root_extent_buffers(info->extent_root);
2083 free_root_extent_buffers(info->csum_root);
2084 free_root_extent_buffers(info->quota_root);
2085 free_root_extent_buffers(info->uuid_root);
2086 if (chunk_root)
2087 free_root_extent_buffers(info->chunk_root);
af31f5e5
CM
2088}
2089
171f6537
JB
2090static void del_fs_roots(struct btrfs_fs_info *fs_info)
2091{
2092 int ret;
2093 struct btrfs_root *gang[8];
2094 int i;
2095
2096 while (!list_empty(&fs_info->dead_roots)) {
2097 gang[0] = list_entry(fs_info->dead_roots.next,
2098 struct btrfs_root, root_list);
2099 list_del(&gang[0]->root_list);
2100
27cdeb70 2101 if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
cb517eab 2102 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
171f6537
JB
2103 } else {
2104 free_extent_buffer(gang[0]->node);
2105 free_extent_buffer(gang[0]->commit_root);
b0feb9d9 2106 btrfs_put_fs_root(gang[0]);
171f6537
JB
2107 }
2108 }
2109
2110 while (1) {
2111 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2112 (void **)gang, 0,
2113 ARRAY_SIZE(gang));
2114 if (!ret)
2115 break;
2116 for (i = 0; i < ret; i++)
cb517eab 2117 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
171f6537 2118 }
1a4319cc
LB
2119
2120 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2121 btrfs_free_log_root_tree(NULL, fs_info);
2122 btrfs_destroy_pinned_extent(fs_info->tree_root,
2123 fs_info->pinned_extents);
2124 }
171f6537 2125}
af31f5e5 2126
ad2b2c80
AV
2127int open_ctree(struct super_block *sb,
2128 struct btrfs_fs_devices *fs_devices,
2129 char *options)
2e635a27 2130{
db94535d
CM
2131 u32 sectorsize;
2132 u32 nodesize;
2133 u32 leafsize;
2134 u32 blocksize;
87ee04eb 2135 u32 stripesize;
84234f3a 2136 u64 generation;
f2b636e8 2137 u64 features;
3de4586c 2138 struct btrfs_key location;
a061fc8d 2139 struct buffer_head *bh;
4d34b278 2140 struct btrfs_super_block *disk_super;
815745cf 2141 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2142 struct btrfs_root *tree_root;
4d34b278
ID
2143 struct btrfs_root *extent_root;
2144 struct btrfs_root *csum_root;
2145 struct btrfs_root *chunk_root;
2146 struct btrfs_root *dev_root;
bcef60f2 2147 struct btrfs_root *quota_root;
f7a81ea4 2148 struct btrfs_root *uuid_root;
e02119d5 2149 struct btrfs_root *log_tree_root;
eb60ceac 2150 int ret;
e58ca020 2151 int err = -EINVAL;
af31f5e5
CM
2152 int num_backups_tried = 0;
2153 int backup_index = 0;
5cdc7ad3
QW
2154 int max_active;
2155 int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
70f80175
SB
2156 bool create_uuid_tree;
2157 bool check_uuid_tree;
4543df7e 2158
f84a8bd6 2159 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
6f07e42e 2160 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
cb517eab 2161 if (!tree_root || !chunk_root) {
39279cc3
CM
2162 err = -ENOMEM;
2163 goto fail;
2164 }
76dda93c
YZ
2165
2166 ret = init_srcu_struct(&fs_info->subvol_srcu);
2167 if (ret) {
2168 err = ret;
2169 goto fail;
2170 }
2171
2172 ret = setup_bdi(fs_info, &fs_info->bdi);
2173 if (ret) {
2174 err = ret;
2175 goto fail_srcu;
2176 }
2177
e2d84521
MX
2178 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2179 if (ret) {
2180 err = ret;
2181 goto fail_bdi;
2182 }
2183 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2184 (1 + ilog2(nr_cpu_ids));
2185
963d678b
MX
2186 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2187 if (ret) {
2188 err = ret;
2189 goto fail_dirty_metadata_bytes;
2190 }
2191
c404e0dc
MX
2192 ret = percpu_counter_init(&fs_info->bio_counter, 0);
2193 if (ret) {
2194 err = ret;
2195 goto fail_delalloc_bytes;
2196 }
2197
76dda93c
YZ
2198 fs_info->btree_inode = new_inode(sb);
2199 if (!fs_info->btree_inode) {
2200 err = -ENOMEM;
c404e0dc 2201 goto fail_bio_counter;
76dda93c
YZ
2202 }
2203
a6591715 2204 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2205
76dda93c 2206 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
f28491e0 2207 INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
8fd17795 2208 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2209 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2210 INIT_LIST_HEAD(&fs_info->delayed_iputs);
eb73c1b7 2211 INIT_LIST_HEAD(&fs_info->delalloc_roots);
11833d66 2212 INIT_LIST_HEAD(&fs_info->caching_block_groups);
eb73c1b7 2213 spin_lock_init(&fs_info->delalloc_root_lock);
a4abeea4 2214 spin_lock_init(&fs_info->trans_lock);
76dda93c 2215 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2216 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2217 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 2218 spin_lock_init(&fs_info->free_chunk_lock);
f29021b2 2219 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2220 spin_lock_init(&fs_info->super_lock);
f28491e0 2221 spin_lock_init(&fs_info->buffer_lock);
f29021b2 2222 rwlock_init(&fs_info->tree_mod_log_lock);
7585717f 2223 mutex_init(&fs_info->reloc_mutex);
573bfb72 2224 mutex_init(&fs_info->delalloc_root_mutex);
de98ced9 2225 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2226
58176a96 2227 init_completion(&fs_info->kobj_unregister);
0b86a832 2228 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2229 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2230 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
0b86a832 2231 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX
2232 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2233 BTRFS_BLOCK_RSV_GLOBAL);
2234 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2235 BTRFS_BLOCK_RSV_DELALLOC);
2236 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2237 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2238 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2239 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2240 BTRFS_BLOCK_RSV_DELOPS);
cb03c743 2241 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 2242 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 2243 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 2244 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 2245 atomic_set(&fs_info->defrag_running, 0);
fc36ed7e 2246 atomic64_set(&fs_info->tree_mod_seq, 0);
e20d96d6 2247 fs_info->sb = sb;
6f568d35 2248 fs_info->max_inline = 8192 * 1024;
9ed74f2d 2249 fs_info->metadata_ratio = 0;
4cb5300b 2250 fs_info->defrag_inodes = RB_ROOT;
2bf64758 2251 fs_info->free_chunk_space = 0;
f29021b2 2252 fs_info->tree_mod_log = RB_ROOT;
8b87dc17 2253 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
0a2b2a84 2254 fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
90519d66
AJ
2255 /* readahead state */
2256 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2257 spin_lock_init(&fs_info->reada_lock);
c8b97818 2258
b34b086c
CM
2259 fs_info->thread_pool_size = min_t(unsigned long,
2260 num_online_cpus() + 2, 8);
0afbaf8c 2261
199c2a9c
MX
2262 INIT_LIST_HEAD(&fs_info->ordered_roots);
2263 spin_lock_init(&fs_info->ordered_root_lock);
16cdcec7
MX
2264 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2265 GFP_NOFS);
2266 if (!fs_info->delayed_root) {
2267 err = -ENOMEM;
2268 goto fail_iput;
2269 }
2270 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2271
a2de733c
AJ
2272 mutex_init(&fs_info->scrub_lock);
2273 atomic_set(&fs_info->scrubs_running, 0);
2274 atomic_set(&fs_info->scrub_pause_req, 0);
2275 atomic_set(&fs_info->scrubs_paused, 0);
2276 atomic_set(&fs_info->scrub_cancel_req, 0);
c404e0dc 2277 init_waitqueue_head(&fs_info->replace_wait);
a2de733c 2278 init_waitqueue_head(&fs_info->scrub_pause_wait);
a2de733c 2279 fs_info->scrub_workers_refcnt = 0;
21adbd5c
SB
2280#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2281 fs_info->check_integrity_print_mask = 0;
2282#endif
a2de733c 2283
c9e9f97b
ID
2284 spin_lock_init(&fs_info->balance_lock);
2285 mutex_init(&fs_info->balance_mutex);
837d5b6e
ID
2286 atomic_set(&fs_info->balance_running, 0);
2287 atomic_set(&fs_info->balance_pause_req, 0);
a7e99c69 2288 atomic_set(&fs_info->balance_cancel_req, 0);
c9e9f97b 2289 fs_info->balance_ctl = NULL;
837d5b6e 2290 init_waitqueue_head(&fs_info->balance_wait_q);
21c7e756 2291 btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
a2de733c 2292
a061fc8d
CM
2293 sb->s_blocksize = 4096;
2294 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 2295 sb->s_bdi = &fs_info->bdi;
a061fc8d 2296
76dda93c 2297 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
bfe86848 2298 set_nlink(fs_info->btree_inode, 1);
0afbaf8c
CM
2299 /*
2300 * we set the i_size on the btree inode to the max possible int.
2301 * the real end of the address space is determined by all of
2302 * the devices in the system
2303 */
2304 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 2305 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM
2306 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2307
5d4f98a2 2308 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 2309 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 2310 fs_info->btree_inode->i_mapping);
0b32f4bb 2311 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
a8067e02 2312 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM
2313
2314 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 2315
76dda93c
YZ
2316 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2317 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2318 sizeof(struct btrfs_key));
72ac3c0d
JB
2319 set_bit(BTRFS_INODE_DUMMY,
2320 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
778ba82b 2321 btrfs_insert_inode_hash(fs_info->btree_inode);
76dda93c 2322
0f9dd46c 2323 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2324 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2325 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2326
11833d66 2327 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 2328 fs_info->btree_inode->i_mapping);
11833d66 2329 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 2330 fs_info->btree_inode->i_mapping);
11833d66 2331 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 2332 fs_info->do_barriers = 1;
e18e4809 2333
39279cc3 2334
5a3f23d5 2335 mutex_init(&fs_info->ordered_operations_mutex);
9ffba8cd 2336 mutex_init(&fs_info->ordered_extent_flush_mutex);
e02119d5 2337 mutex_init(&fs_info->tree_log_mutex);
925baedd 2338 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM
2339 mutex_init(&fs_info->transaction_kthread_mutex);
2340 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2341 mutex_init(&fs_info->volume_mutex);
9e351cc8 2342 init_rwsem(&fs_info->commit_root_sem);
c71bf099 2343 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2344 init_rwsem(&fs_info->subvol_sem);
803b2f54 2345 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
e922e087
SB
2346 fs_info->dev_replace.lock_owner = 0;
2347 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2348 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2349 mutex_init(&fs_info->dev_replace.lock_management_lock);
2350 mutex_init(&fs_info->dev_replace.lock);
fa9c0d79 2351
416ac51d 2352 spin_lock_init(&fs_info->qgroup_lock);
f2f6ed3d 2353 mutex_init(&fs_info->qgroup_ioctl_lock);
416ac51d
AJ
2354 fs_info->qgroup_tree = RB_ROOT;
2355 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2356 fs_info->qgroup_seq = 1;
2357 fs_info->quota_enabled = 0;
2358 fs_info->pending_quota_state = 0;
1e8f9158 2359 fs_info->qgroup_ulist = NULL;
2f232036 2360 mutex_init(&fs_info->qgroup_rescan_lock);
416ac51d 2361
fa9c0d79
CM
2362 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2363 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2364
e6dcd2dc 2365 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2366 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2367 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2368 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2369
53b381b3
DW
2370 ret = btrfs_alloc_stripe_hash_table(fs_info);
2371 if (ret) {
83c8266a 2372 err = ret;
53b381b3
DW
2373 goto fail_alloc;
2374 }
2375
0b86a832 2376 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 2377 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2378
3c4bb26b 2379 invalidate_bdev(fs_devices->latest_bdev);
1104a885
DS
2380
2381 /*
2382 * Read super block and check the signature bytes only
2383 */
a512bbf8 2384 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY
2385 if (!bh) {
2386 err = -EINVAL;
16cdcec7 2387 goto fail_alloc;
20b45077 2388 }
39279cc3 2389
1104a885
DS
2390 /*
2391 * We want to check superblock checksum, the type is stored inside.
2392 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2393 */
2394 if (btrfs_check_super_csum(bh->b_data)) {
efe120a0 2395 printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
1104a885
DS
2396 err = -EINVAL;
2397 goto fail_alloc;
2398 }
2399
2400 /*
2401 * super_copy is zeroed at allocation time and we never touch the
2402 * following bytes up to INFO_SIZE, the checksum is calculated from
2403 * the whole block of INFO_SIZE
2404 */
6c41761f
DS
2405 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2406 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2407 sizeof(*fs_info->super_for_commit));
a061fc8d 2408 brelse(bh);
5f39d397 2409
6c41761f 2410 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2411
1104a885
DS
2412 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2413 if (ret) {
efe120a0 2414 printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
1104a885
DS
2415 err = -EINVAL;
2416 goto fail_alloc;
2417 }
2418
6c41761f 2419 disk_super = fs_info->super_copy;
0f7d52f4 2420 if (!btrfs_super_root(disk_super))
16cdcec7 2421 goto fail_alloc;
0f7d52f4 2422
acce952b 2423 /* check FS state, whether FS is broken. */
87533c47
MX
2424 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2425 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2426
af31f5e5
CM
2427 /*
2428 * run through our array of backup supers and setup
2429 * our ring pointer to the oldest one
2430 */
2431 generation = btrfs_super_generation(disk_super);
2432 find_oldest_super_backup(fs_info, generation);
2433
75e7cb7f
LB
2434 /*
2435 * In the long term, we'll store the compression type in the super
2436 * block, and it'll be used for per file compression control.
2437 */
2438 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2439
2b82032c
YZ
2440 ret = btrfs_parse_options(tree_root, options);
2441 if (ret) {
2442 err = ret;
16cdcec7 2443 goto fail_alloc;
2b82032c 2444 }
dfe25020 2445
f2b636e8
JB
2446 features = btrfs_super_incompat_flags(disk_super) &
2447 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2448 if (features) {
2449 printk(KERN_ERR "BTRFS: couldn't mount because of "
2450 "unsupported optional features (%Lx).\n",
c1c9ff7c 2451 features);
f2b636e8 2452 err = -EINVAL;
16cdcec7 2453 goto fail_alloc;
f2b636e8
JB
2454 }
2455
727011e0
CM
2456 if (btrfs_super_leafsize(disk_super) !=
2457 btrfs_super_nodesize(disk_super)) {
2458 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2459 "blocksizes don't match. node %d leaf %d\n",
2460 btrfs_super_nodesize(disk_super),
2461 btrfs_super_leafsize(disk_super));
2462 err = -EINVAL;
2463 goto fail_alloc;
2464 }
2465 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2466 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2467 "blocksize (%d) was too large\n",
2468 btrfs_super_leafsize(disk_super));
2469 err = -EINVAL;
2470 goto fail_alloc;
2471 }
2472
5d4f98a2 2473 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2474 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
69e380d1 2475 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2476 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
727011e0 2477
3173a18f 2478 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
efe120a0 2479 printk(KERN_ERR "BTRFS: has skinny extents\n");
3173a18f 2480
727011e0
CM
2481 /*
2482 * flag our filesystem as having big metadata blocks if
2483 * they are bigger than the page size
2484 */
2485 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2486 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
efe120a0 2487 printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
727011e0
CM
2488 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2489 }
2490
bc3f116f
CM
2491 nodesize = btrfs_super_nodesize(disk_super);
2492 leafsize = btrfs_super_leafsize(disk_super);
2493 sectorsize = btrfs_super_sectorsize(disk_super);
2494 stripesize = btrfs_super_stripesize(disk_super);
e2d84521 2495 fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids));
963d678b 2496 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f
CM
2497
2498 /*
2499 * mixed block groups end up with duplicate but slightly offset
2500 * extent buffers for the same range. It leads to corruptions
2501 */
2502 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2503 (sectorsize != leafsize)) {
efe120a0 2504 printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes "
bc3f116f
CM
2505 "are not allowed for mixed block groups on %s\n",
2506 sb->s_id);
2507 goto fail_alloc;
2508 }
2509
ceda0864
MX
2510 /*
2511 * Needn't use the lock because there is no other task which will
2512 * update the flag.
2513 */
a6fa6fae 2514 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2515
f2b636e8
JB
2516 features = btrfs_super_compat_ro_flags(disk_super) &
2517 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2518 if (!(sb->s_flags & MS_RDONLY) && features) {
2519 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2520 "unsupported option features (%Lx).\n",
c1c9ff7c 2521 features);
f2b636e8 2522 err = -EINVAL;
16cdcec7 2523 goto fail_alloc;
f2b636e8 2524 }
61d92c32 2525
5cdc7ad3 2526 max_active = fs_info->thread_pool_size;
61d92c32 2527
5cdc7ad3
QW
2528 fs_info->workers =
2529 btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2530 max_active, 16);
c8b97818 2531
afe3d242
QW
2532 fs_info->delalloc_workers =
2533 btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
771ed689 2534
a44903ab
QW
2535 fs_info->flush_workers =
2536 btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
8ccf6f19 2537
e66f0bb1
QW
2538 fs_info->caching_workers =
2539 btrfs_alloc_workqueue("cache", flags, max_active, 0);
61b49440 2540
a8c93d4e
QW
2541 /*
2542 * a higher idle thresh on the submit workers makes it much more
61b49440
CM
2543 * likely that bios will be send down in a sane order to the
2544 * devices
2545 */
a8c93d4e
QW
2546 fs_info->submit_workers =
2547 btrfs_alloc_workqueue("submit", flags,
2548 min_t(u64, fs_devices->num_devices,
2549 max_active), 64);
53863232 2550
dc6e3209
QW
2551 fs_info->fixup_workers =
2552 btrfs_alloc_workqueue("fixup", flags, 1, 0);
61b49440
CM
2553
2554 /*
2555 * endios are largely parallel and should have a very
2556 * low idle thresh
2557 */
fccb5d86
QW
2558 fs_info->endio_workers =
2559 btrfs_alloc_workqueue("endio", flags, max_active, 4);
2560 fs_info->endio_meta_workers =
2561 btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2562 fs_info->endio_meta_write_workers =
2563 btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2564 fs_info->endio_raid56_workers =
2565 btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
d05a33ac
QW
2566 fs_info->rmw_workers =
2567 btrfs_alloc_workqueue("rmw", flags, max_active, 2);
fccb5d86
QW
2568 fs_info->endio_write_workers =
2569 btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2570 fs_info->endio_freespace_worker =
2571 btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
5b3bc44e
QW
2572 fs_info->delayed_workers =
2573 btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
736cfa15
QW
2574 fs_info->readahead_workers =
2575 btrfs_alloc_workqueue("readahead", flags, max_active, 2);
fc97fab0
QW
2576 fs_info->qgroup_rescan_workers =
2577 btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
61b49440 2578
a8c93d4e 2579 if (!(fs_info->workers && fs_info->delalloc_workers &&
fccb5d86
QW
2580 fs_info->submit_workers && fs_info->flush_workers &&
2581 fs_info->endio_workers && fs_info->endio_meta_workers &&
2582 fs_info->endio_meta_write_workers &&
2583 fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
e66f0bb1 2584 fs_info->endio_freespace_worker && fs_info->rmw_workers &&
dc6e3209 2585 fs_info->caching_workers && fs_info->readahead_workers &&
fc97fab0
QW
2586 fs_info->fixup_workers && fs_info->delayed_workers &&
2587 fs_info->qgroup_rescan_workers)) {
fed425c7 2588 err = -ENOMEM;
0dc3b84a
JB
2589 goto fail_sb_buffer;
2590 }
4543df7e 2591
4575c9cc 2592 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM
2593 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2594 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 2595
db94535d
CM
2596 tree_root->nodesize = nodesize;
2597 tree_root->leafsize = leafsize;
2598 tree_root->sectorsize = sectorsize;
87ee04eb 2599 tree_root->stripesize = stripesize;
a061fc8d
CM
2600
2601 sb->s_blocksize = sectorsize;
2602 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 2603
3cae210f 2604 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
efe120a0 2605 printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id);
39279cc3
CM
2606 goto fail_sb_buffer;
2607 }
19c00ddc 2608
8d082fb7 2609 if (sectorsize != PAGE_SIZE) {
efe120a0 2610 printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) "
8d082fb7 2611 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
941b2ddf
KM
2612 goto fail_sb_buffer;
2613 }
2614
925baedd 2615 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 2616 ret = btrfs_read_sys_array(tree_root);
925baedd 2617 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2618 if (ret) {
efe120a0 2619 printk(KERN_WARNING "BTRFS: failed to read the system "
d397712b 2620 "array on %s\n", sb->s_id);
5d4f98a2 2621 goto fail_sb_buffer;
84eed90f 2622 }
0b86a832
CM
2623
2624 blocksize = btrfs_level_size(tree_root,
2625 btrfs_super_chunk_root_level(disk_super));
84234f3a 2626 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM
2627
2628 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2629 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2630
2631 chunk_root->node = read_tree_block(chunk_root,
2632 btrfs_super_chunk_root(disk_super),
84234f3a 2633 blocksize, generation);
416bc658
JB
2634 if (!chunk_root->node ||
2635 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
efe120a0 2636 printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n",
83121942 2637 sb->s_id);
af31f5e5 2638 goto fail_tree_roots;
83121942 2639 }
5d4f98a2
YZ
2640 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2641 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2642
e17cade2 2643 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
b308bc2f 2644 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
e17cade2 2645
0b86a832 2646 ret = btrfs_read_chunk_tree(chunk_root);
2b82032c 2647 if (ret) {
efe120a0 2648 printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n",
d397712b 2649 sb->s_id);
af31f5e5 2650 goto fail_tree_roots;
2b82032c 2651 }
0b86a832 2652
8dabb742
SB
2653 /*
2654 * keep the device that is marked to be the target device for the
2655 * dev_replace procedure
2656 */
2657 btrfs_close_extra_devices(fs_info, fs_devices, 0);
dfe25020 2658
a6b0d5c8 2659 if (!fs_devices->latest_bdev) {
efe120a0 2660 printk(KERN_CRIT "BTRFS: failed to read devices on %s\n",
a6b0d5c8
CM
2661 sb->s_id);
2662 goto fail_tree_roots;
2663 }
2664
af31f5e5 2665retry_root_backup:
db94535d
CM
2666 blocksize = btrfs_level_size(tree_root,
2667 btrfs_super_root_level(disk_super));
84234f3a 2668 generation = btrfs_super_generation(disk_super);
0b86a832 2669
e20d96d6 2670 tree_root->node = read_tree_block(tree_root,
db94535d 2671 btrfs_super_root(disk_super),
84234f3a 2672 blocksize, generation);
af31f5e5
CM
2673 if (!tree_root->node ||
2674 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
efe120a0 2675 printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
83121942 2676 sb->s_id);
af31f5e5
CM
2677
2678 goto recovery_tree_root;
83121942 2679 }
af31f5e5 2680
5d4f98a2
YZ
2681 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2682 tree_root->commit_root = btrfs_root_node(tree_root);
69e9c6c6 2683 btrfs_set_root_refs(&tree_root->root_item, 1);
db94535d 2684
cb517eab
MX
2685 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2686 location.type = BTRFS_ROOT_ITEM_KEY;
2687 location.offset = 0;
2688
2689 extent_root = btrfs_read_tree_root(tree_root, &location);
2690 if (IS_ERR(extent_root)) {
2691 ret = PTR_ERR(extent_root);
af31f5e5 2692 goto recovery_tree_root;
cb517eab 2693 }
27cdeb70 2694 set_bit(BTRFS_ROOT_TRACK_DIRTY, &extent_root->state);
cb517eab 2695 fs_info->extent_root = extent_root;
0b86a832 2696
cb517eab
MX
2697 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2698 dev_root = btrfs_read_tree_root(tree_root, &location);
2699 if (IS_ERR(dev_root)) {
2700 ret = PTR_ERR(dev_root);
af31f5e5 2701 goto recovery_tree_root;
cb517eab 2702 }
27cdeb70 2703 set_bit(BTRFS_ROOT_TRACK_DIRTY, &dev_root->state);
cb517eab
MX
2704 fs_info->dev_root = dev_root;
2705 btrfs_init_devices_late(fs_info);
3768f368 2706
cb517eab
MX
2707 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2708 csum_root = btrfs_read_tree_root(tree_root, &location);
2709 if (IS_ERR(csum_root)) {
2710 ret = PTR_ERR(csum_root);
af31f5e5 2711 goto recovery_tree_root;
cb517eab 2712 }
27cdeb70 2713 set_bit(BTRFS_ROOT_TRACK_DIRTY, &csum_root->state);
cb517eab 2714 fs_info->csum_root = csum_root;
d20f7043 2715
cb517eab
MX
2716 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2717 quota_root = btrfs_read_tree_root(tree_root, &location);
2718 if (!IS_ERR(quota_root)) {
27cdeb70 2719 set_bit(BTRFS_ROOT_TRACK_DIRTY, &quota_root->state);
bcef60f2
AJ
2720 fs_info->quota_enabled = 1;
2721 fs_info->pending_quota_state = 1;
cb517eab 2722 fs_info->quota_root = quota_root;
bcef60f2
AJ
2723 }
2724
f7a81ea4
SB
2725 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2726 uuid_root = btrfs_read_tree_root(tree_root, &location);
2727 if (IS_ERR(uuid_root)) {
2728 ret = PTR_ERR(uuid_root);
2729 if (ret != -ENOENT)
2730 goto recovery_tree_root;
2731 create_uuid_tree = true;
70f80175 2732 check_uuid_tree = false;
f7a81ea4 2733 } else {
27cdeb70 2734 set_bit(BTRFS_ROOT_TRACK_DIRTY, &uuid_root->state);
f7a81ea4 2735 fs_info->uuid_root = uuid_root;
70f80175
SB
2736 create_uuid_tree = false;
2737 check_uuid_tree =
2738 generation != btrfs_super_uuid_tree_generation(disk_super);
f7a81ea4
SB
2739 }
2740
8929ecfa
YZ
2741 fs_info->generation = generation;
2742 fs_info->last_trans_committed = generation;
8929ecfa 2743
68310a5e
ID
2744 ret = btrfs_recover_balance(fs_info);
2745 if (ret) {
efe120a0 2746 printk(KERN_WARNING "BTRFS: failed to recover balance\n");
68310a5e
ID
2747 goto fail_block_groups;
2748 }
2749
733f4fbb
SB
2750 ret = btrfs_init_dev_stats(fs_info);
2751 if (ret) {
efe120a0 2752 printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
733f4fbb
SB
2753 ret);
2754 goto fail_block_groups;
2755 }
2756
8dabb742
SB
2757 ret = btrfs_init_dev_replace(fs_info);
2758 if (ret) {
efe120a0 2759 pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
8dabb742
SB
2760 goto fail_block_groups;
2761 }
2762
2763 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2764
5ac1d209 2765 ret = btrfs_sysfs_add_one(fs_info);
c59021f8 2766 if (ret) {
efe120a0 2767 pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
c59021f8 2768 goto fail_block_groups;
2769 }
2770
c59021f8 2771 ret = btrfs_init_space_info(fs_info);
2772 if (ret) {
efe120a0 2773 printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
2365dd3c 2774 goto fail_sysfs;
c59021f8 2775 }
2776
1b1d1f66
JB
2777 ret = btrfs_read_block_groups(extent_root);
2778 if (ret) {
efe120a0 2779 printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
2365dd3c 2780 goto fail_sysfs;
1b1d1f66 2781 }
5af3e8cc
SB
2782 fs_info->num_tolerated_disk_barrier_failures =
2783 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
292fd7fc
SB
2784 if (fs_info->fs_devices->missing_devices >
2785 fs_info->num_tolerated_disk_barrier_failures &&
2786 !(sb->s_flags & MS_RDONLY)) {
efe120a0
FH
2787 printk(KERN_WARNING "BTRFS: "
2788 "too many missing devices, writeable mount is not allowed\n");
2365dd3c 2789 goto fail_sysfs;
292fd7fc 2790 }
9078a3e1 2791
a74a4b97
CM
2792 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2793 "btrfs-cleaner");
57506d50 2794 if (IS_ERR(fs_info->cleaner_kthread))
2365dd3c 2795 goto fail_sysfs;
a74a4b97
CM
2796
2797 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2798 tree_root,
2799 "btrfs-transaction");
57506d50 2800 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2801 goto fail_cleaner;
a74a4b97 2802
c289811c
CM
2803 if (!btrfs_test_opt(tree_root, SSD) &&
2804 !btrfs_test_opt(tree_root, NOSSD) &&
2805 !fs_info->fs_devices->rotating) {
efe120a0 2806 printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
c289811c
CM
2807 "mode\n");
2808 btrfs_set_opt(fs_info->mount_opt, SSD);
2809 }
2810
3818aea2
QW
2811 /* Set the real inode map cache flag */
2812 if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE))
2813 btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE);
2814
21adbd5c
SB
2815#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2816 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2817 ret = btrfsic_mount(tree_root, fs_devices,
2818 btrfs_test_opt(tree_root,
2819 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2820 1 : 0,
2821 fs_info->check_integrity_print_mask);
2822 if (ret)
efe120a0 2823 printk(KERN_WARNING "BTRFS: failed to initialize"
21adbd5c
SB
2824 " integrity check module %s\n", sb->s_id);
2825 }
2826#endif
bcef60f2
AJ
2827 ret = btrfs_read_qgroup_config(fs_info);
2828 if (ret)
2829 goto fail_trans_kthread;
21adbd5c 2830
acce952b 2831 /* do not make disk changes in broken FS */
68ce9682 2832 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM
2833 u64 bytenr = btrfs_super_log_root(disk_super);
2834
7c2ca468 2835 if (fs_devices->rw_devices == 0) {
efe120a0 2836 printk(KERN_WARNING "BTRFS: log replay required "
d397712b 2837 "on RO media\n");
7c2ca468 2838 err = -EIO;
bcef60f2 2839 goto fail_qgroup;
7c2ca468 2840 }
e02119d5
CM
2841 blocksize =
2842 btrfs_level_size(tree_root,
2843 btrfs_super_log_root_level(disk_super));
d18a2c44 2844
6f07e42e 2845 log_tree_root = btrfs_alloc_root(fs_info);
676e4c86
DC
2846 if (!log_tree_root) {
2847 err = -ENOMEM;
bcef60f2 2848 goto fail_qgroup;
676e4c86 2849 }
e02119d5
CM
2850
2851 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2852 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2853
2854 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ
2855 blocksize,
2856 generation + 1);
416bc658
JB
2857 if (!log_tree_root->node ||
2858 !extent_buffer_uptodate(log_tree_root->node)) {
efe120a0 2859 printk(KERN_ERR "BTRFS: failed to read log tree\n");
416bc658
JB
2860 free_extent_buffer(log_tree_root->node);
2861 kfree(log_tree_root);
28c16cbb 2862 goto fail_qgroup;
416bc658 2863 }
79787eaa 2864 /* returns with log_tree_root freed on success */
e02119d5 2865 ret = btrfs_recover_log_trees(log_tree_root);
79787eaa
JM
2866 if (ret) {
2867 btrfs_error(tree_root->fs_info, ret,
2868 "Failed to recover log tree");
2869 free_extent_buffer(log_tree_root->node);
2870 kfree(log_tree_root);
28c16cbb 2871 goto fail_qgroup;
79787eaa 2872 }
e556ce2c
YZ
2873
2874 if (sb->s_flags & MS_RDONLY) {
79787eaa
JM
2875 ret = btrfs_commit_super(tree_root);
2876 if (ret)
28c16cbb 2877 goto fail_qgroup;
e556ce2c 2878 }
e02119d5 2879 }
1a40e23b 2880
76dda93c 2881 ret = btrfs_find_orphan_roots(tree_root);
79787eaa 2882 if (ret)
28c16cbb 2883 goto fail_qgroup;
76dda93c 2884
7c2ca468 2885 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b 2886 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2 2887 if (ret)
28c16cbb 2888 goto fail_qgroup;
d68fc57b 2889
5d4f98a2 2890 ret = btrfs_recover_relocation(tree_root);
d7ce5843
MX
2891 if (ret < 0) {
2892 printk(KERN_WARNING
efe120a0 2893 "BTRFS: failed to recover relocation\n");
d7ce5843 2894 err = -EINVAL;
bcef60f2 2895 goto fail_qgroup;
d7ce5843 2896 }
7c2ca468 2897 }
1a40e23b 2898
3de4586c
CM
2899 location.objectid = BTRFS_FS_TREE_OBJECTID;
2900 location.type = BTRFS_ROOT_ITEM_KEY;
cb517eab 2901 location.offset = 0;
3de4586c 2902
3de4586c 2903 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
3140c9a3
DC
2904 if (IS_ERR(fs_info->fs_root)) {
2905 err = PTR_ERR(fs_info->fs_root);
bcef60f2 2906 goto fail_qgroup;
3140c9a3 2907 }
c289811c 2908
2b6ba629
ID
2909 if (sb->s_flags & MS_RDONLY)
2910 return 0;
59641015 2911
2b6ba629
ID
2912 down_read(&fs_info->cleanup_work_sem);
2913 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2914 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 2915 up_read(&fs_info->cleanup_work_sem);
2b6ba629
ID
2916 close_ctree(tree_root);
2917 return ret;
2918 }
2919 up_read(&fs_info->cleanup_work_sem);
59641015 2920
2b6ba629
ID
2921 ret = btrfs_resume_balance_async(fs_info);
2922 if (ret) {
efe120a0 2923 printk(KERN_WARNING "BTRFS: failed to resume balance\n");
2b6ba629
ID
2924 close_ctree(tree_root);
2925 return ret;
e3acc2a6
JB
2926 }
2927
8dabb742
SB
2928 ret = btrfs_resume_dev_replace_async(fs_info);
2929 if (ret) {
efe120a0 2930 pr_warn("BTRFS: failed to resume dev_replace\n");
8dabb742
SB
2931 close_ctree(tree_root);
2932 return ret;
2933 }
2934
b382a324
JS
2935 btrfs_qgroup_rescan_resume(fs_info);
2936
f7a81ea4 2937 if (create_uuid_tree) {
efe120a0 2938 pr_info("BTRFS: creating UUID tree\n");
f7a81ea4
SB
2939 ret = btrfs_create_uuid_tree(fs_info);
2940 if (ret) {
efe120a0 2941 pr_warn("BTRFS: failed to create the UUID tree %d\n",
f7a81ea4
SB
2942 ret);
2943 close_ctree(tree_root);
2944 return ret;
2945 }
f420ee1e
SB
2946 } else if (check_uuid_tree ||
2947 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
efe120a0 2948 pr_info("BTRFS: checking UUID tree\n");
70f80175
SB
2949 ret = btrfs_check_uuid_tree(fs_info);
2950 if (ret) {
efe120a0 2951 pr_warn("BTRFS: failed to check the UUID tree %d\n",
70f80175
SB
2952 ret);
2953 close_ctree(tree_root);
2954 return ret;
2955 }
2956 } else {
2957 fs_info->update_uuid_tree_gen = 1;
f7a81ea4
SB
2958 }
2959
ad2b2c80 2960 return 0;
39279cc3 2961
bcef60f2
AJ
2962fail_qgroup:
2963 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM
2964fail_trans_kthread:
2965 kthread_stop(fs_info->transaction_kthread);
54067ae9 2966 btrfs_cleanup_transaction(fs_info->tree_root);
7b5ff90e 2967 del_fs_roots(fs_info);
3f157a2f 2968fail_cleaner:
a74a4b97 2969 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM
2970
2971 /*
2972 * make sure we're done with the btree inode before we stop our
2973 * kthreads
2974 */
2975 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 2976
2365dd3c
AJ
2977fail_sysfs:
2978 btrfs_sysfs_remove_one(fs_info);
2979
1b1d1f66 2980fail_block_groups:
54067ae9 2981 btrfs_put_block_group_cache(fs_info);
1b1d1f66 2982 btrfs_free_block_groups(fs_info);
af31f5e5
CM
2983
2984fail_tree_roots:
2985 free_root_pointers(fs_info, 1);
2b8195bb 2986 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 2987
39279cc3 2988fail_sb_buffer:
7abadb64 2989 btrfs_stop_all_workers(fs_info);
16cdcec7 2990fail_alloc:
4543df7e 2991fail_iput:
586e46e2
ID
2992 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2993
4543df7e 2994 iput(fs_info->btree_inode);
c404e0dc
MX
2995fail_bio_counter:
2996 percpu_counter_destroy(&fs_info->bio_counter);
963d678b
MX
2997fail_delalloc_bytes:
2998 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX
2999fail_dirty_metadata_bytes:
3000 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
ad081f14 3001fail_bdi:
7e662854 3002 bdi_destroy(&fs_info->bdi);
76dda93c
YZ
3003fail_srcu:
3004 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 3005fail:
53b381b3 3006 btrfs_free_stripe_hash_table(fs_info);
586e46e2 3007 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 3008 return err;
af31f5e5
CM
3009
3010recovery_tree_root:
af31f5e5
CM
3011 if (!btrfs_test_opt(tree_root, RECOVERY))
3012 goto fail_tree_roots;
3013
3014 free_root_pointers(fs_info, 0);
3015
3016 /* don't use the log in recovery mode, it won't be valid */
3017 btrfs_set_super_log_root(disk_super, 0);
3018
3019 /* we can't trust the free space cache either */
3020 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3021
3022 ret = next_root_backup(fs_info, fs_info->super_copy,
3023 &num_backups_tried, &backup_index);
3024 if (ret == -1)
3025 goto fail_block_groups;
3026 goto retry_root_backup;
eb60ceac
CM
3027}
3028
f2984462
CM
3029static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3030{
f2984462
CM
3031 if (uptodate) {
3032 set_buffer_uptodate(bh);
3033 } else {
442a4f63
SB
3034 struct btrfs_device *device = (struct btrfs_device *)
3035 bh->b_private;
3036
efe120a0 3037 printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
606686ee
JB
3038 "I/O error on %s\n",
3039 rcu_str_deref(device->name));
1259ab75
CM
3040 /* note, we dont' set_buffer_write_io_error because we have
3041 * our own ways of dealing with the IO errors
3042 */
f2984462 3043 clear_buffer_uptodate(bh);
442a4f63 3044 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM
3045 }
3046 unlock_buffer(bh);
3047 put_bh(bh);
3048}
3049
a512bbf8
YZ
3050struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3051{
3052 struct buffer_head *bh;
3053 struct buffer_head *latest = NULL;
3054 struct btrfs_super_block *super;
3055 int i;
3056 u64 transid = 0;
3057 u64 bytenr;
3058
3059 /* we would like to check all the supers, but that would make
3060 * a btrfs mount succeed after a mkfs from a different FS.
3061 * So, we need to add a special mount option to scan for
3062 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3063 */
3064 for (i = 0; i < 1; i++) {
3065 bytenr = btrfs_sb_offset(i);
8068a47e
AJ
3066 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3067 i_size_read(bdev->bd_inode))
a512bbf8 3068 break;
8068a47e
AJ
3069 bh = __bread(bdev, bytenr / 4096,
3070 BTRFS_SUPER_INFO_SIZE);
a512bbf8
YZ
3071 if (!bh)
3072 continue;
3073
3074 super = (struct btrfs_super_block *)bh->b_data;
3075 if (btrfs_super_bytenr(super) != bytenr ||
3cae210f 3076 btrfs_super_magic(super) != BTRFS_MAGIC) {
a512bbf8
YZ
3077 brelse(bh);
3078 continue;
3079 }
3080
3081 if (!latest || btrfs_super_generation(super) > transid) {
3082 brelse(latest);
3083 latest = bh;
3084 transid = btrfs_super_generation(super);
3085 } else {
3086 brelse(bh);
3087 }
3088 }
3089 return latest;
3090}
3091
4eedeb75
HH
3092/*
3093 * this should be called twice, once with wait == 0 and
3094 * once with wait == 1. When wait == 0 is done, all the buffer heads
3095 * we write are pinned.
3096 *
3097 * They are released when wait == 1 is done.
3098 * max_mirrors must be the same for both runs, and it indicates how
3099 * many supers on this one device should be written.
3100 *
3101 * max_mirrors == 0 means to write them all.
3102 */
a512bbf8
YZ
3103static int write_dev_supers(struct btrfs_device *device,
3104 struct btrfs_super_block *sb,
3105 int do_barriers, int wait, int max_mirrors)
3106{
3107 struct buffer_head *bh;
3108 int i;
3109 int ret;
3110 int errors = 0;
3111 u32 crc;
3112 u64 bytenr;
a512bbf8
YZ
3113
3114 if (max_mirrors == 0)
3115 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3116
a512bbf8
YZ
3117 for (i = 0; i < max_mirrors; i++) {
3118 bytenr = btrfs_sb_offset(i);
3119 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
3120 break;
3121
3122 if (wait) {
3123 bh = __find_get_block(device->bdev, bytenr / 4096,
3124 BTRFS_SUPER_INFO_SIZE);
634554dc
JB
3125 if (!bh) {
3126 errors++;
3127 continue;
3128 }
a512bbf8 3129 wait_on_buffer(bh);
4eedeb75
HH
3130 if (!buffer_uptodate(bh))
3131 errors++;
3132
3133 /* drop our reference */
3134 brelse(bh);
3135
3136 /* drop the reference from the wait == 0 run */
3137 brelse(bh);
3138 continue;
a512bbf8
YZ
3139 } else {
3140 btrfs_set_super_bytenr(sb, bytenr);
3141
3142 crc = ~(u32)0;
b0496686 3143 crc = btrfs_csum_data((char *)sb +
a512bbf8
YZ
3144 BTRFS_CSUM_SIZE, crc,
3145 BTRFS_SUPER_INFO_SIZE -
3146 BTRFS_CSUM_SIZE);
3147 btrfs_csum_final(crc, sb->csum);
3148
4eedeb75
HH
3149 /*
3150 * one reference for us, and we leave it for the
3151 * caller
3152 */
a512bbf8
YZ
3153 bh = __getblk(device->bdev, bytenr / 4096,
3154 BTRFS_SUPER_INFO_SIZE);
634554dc 3155 if (!bh) {
efe120a0 3156 printk(KERN_ERR "BTRFS: couldn't get super "
634554dc
JB
3157 "buffer head for bytenr %Lu\n", bytenr);
3158 errors++;
3159 continue;
3160 }
3161
a512bbf8
YZ
3162 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3163
4eedeb75 3164 /* one reference for submit_bh */
a512bbf8 3165 get_bh(bh);
4eedeb75
HH
3166
3167 set_buffer_uptodate(bh);
a512bbf8
YZ
3168 lock_buffer(bh);
3169 bh->b_end_io = btrfs_end_buffer_write_sync;
442a4f63 3170 bh->b_private = device;
a512bbf8
YZ
3171 }
3172
387125fc
CM
3173 /*
3174 * we fua the first super. The others we allow
3175 * to go down lazy.
3176 */
e8117c26
WS
3177 if (i == 0)
3178 ret = btrfsic_submit_bh(WRITE_FUA, bh);
3179 else
3180 ret = btrfsic_submit_bh(WRITE_SYNC, bh);
4eedeb75 3181 if (ret)
a512bbf8 3182 errors++;
a512bbf8
YZ
3183 }
3184 return errors < i ? 0 : -1;
3185}
3186
387125fc
CM
3187/*
3188 * endio for the write_dev_flush, this will wake anyone waiting
3189 * for the barrier when it is done
3190 */
3191static void btrfs_end_empty_barrier(struct bio *bio, int err)
3192{
3193 if (err) {
3194 if (err == -EOPNOTSUPP)
3195 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3196 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3197 }
3198 if (bio->bi_private)
3199 complete(bio->bi_private);
3200 bio_put(bio);
3201}
3202
3203/*
3204 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3205 * sent down. With wait == 1, it waits for the previous flush.
3206 *
3207 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3208 * capable
3209 */
3210static int write_dev_flush(struct btrfs_device *device, int wait)
3211{
3212 struct bio *bio;
3213 int ret = 0;
3214
3215 if (device->nobarriers)
3216 return 0;
3217
3218 if (wait) {
3219 bio = device->flush_bio;
3220 if (!bio)
3221 return 0;
3222
3223 wait_for_completion(&device->flush_wait);
3224
3225 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
efe120a0 3226 printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
606686ee 3227 rcu_str_deref(device->name));
387125fc 3228 device->nobarriers = 1;
5af3e8cc 3229 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
387125fc 3230 ret = -EIO;
5af3e8cc
SB
3231 btrfs_dev_stat_inc_and_print(device,
3232 BTRFS_DEV_STAT_FLUSH_ERRS);
387125fc
CM
3233 }
3234
3235 /* drop the reference from the wait == 0 run */
3236 bio_put(bio);
3237 device->flush_bio = NULL;
3238
3239 return ret;
3240 }
3241
3242 /*
3243 * one reference for us, and we leave it for the
3244 * caller
3245 */
9c017abc 3246 device->flush_bio = NULL;
9be3395b 3247 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
387125fc
CM
3248 if (!bio)
3249 return -ENOMEM;
3250
3251 bio->bi_end_io = btrfs_end_empty_barrier;
3252 bio->bi_bdev = device->bdev;
3253 init_completion(&device->flush_wait);
3254 bio->bi_private = &device->flush_wait;
3255 device->flush_bio = bio;
3256
3257 bio_get(bio);
21adbd5c 3258 btrfsic_submit_bio(WRITE_FLUSH, bio);
387125fc
CM
3259
3260 return 0;
3261}
3262
3263/*
3264 * send an empty flush down to each device in parallel,
3265 * then wait for them
3266 */
3267static int barrier_all_devices(struct btrfs_fs_info *info)
3268{
3269 struct list_head *head;
3270 struct btrfs_device *dev;
5af3e8cc
SB
3271 int errors_send = 0;
3272 int errors_wait = 0;
387125fc
CM
3273 int ret;
3274
3275 /* send down all the barriers */
3276 head = &info->fs_devices->devices;
3277 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS
3278 if (dev->missing)
3279 continue;
387125fc 3280 if (!dev->bdev) {
5af3e8cc 3281 errors_send++;
387125fc
CM
3282 continue;
3283 }
3284 if (!dev->in_fs_metadata || !dev->writeable)
3285 continue;
3286
3287 ret = write_dev_flush(dev, 0);
3288 if (ret)
5af3e8cc 3289 errors_send++;
387125fc
CM
3290 }
3291
3292 /* wait for all the barriers */
3293 list_for_each_entry_rcu(dev, head, dev_list) {
f88ba6a2
HS
3294 if (dev->missing)
3295 continue;
387125fc 3296 if (!dev->bdev) {
5af3e8cc 3297 errors_wait++;
387125fc
CM
3298 continue;
3299 }
3300 if (!dev->in_fs_metadata || !dev->writeable)
3301 continue;
3302
3303 ret = write_dev_flush(dev, 1);
3304 if (ret)
5af3e8cc 3305 errors_wait++;
387125fc 3306 }
5af3e8cc
SB
3307 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3308 errors_wait > info->num_tolerated_disk_barrier_failures)
387125fc
CM
3309 return -EIO;
3310 return 0;
3311}
3312
5af3e8cc
SB
3313int btrfs_calc_num_tolerated_disk_barrier_failures(
3314 struct btrfs_fs_info *fs_info)
3315{
3316 struct btrfs_ioctl_space_info space;
3317 struct btrfs_space_info *sinfo;
3318 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3319 BTRFS_BLOCK_GROUP_SYSTEM,
3320 BTRFS_BLOCK_GROUP_METADATA,
3321 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3322 int num_types = 4;
3323 int i;
3324 int c;
3325 int num_tolerated_disk_barrier_failures =
3326 (int)fs_info->fs_devices->num_devices;
3327
3328 for (i = 0; i < num_types; i++) {
3329 struct btrfs_space_info *tmp;
3330
3331 sinfo = NULL;
3332 rcu_read_lock();
3333 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3334 if (tmp->flags == types[i]) {
3335 sinfo = tmp;
3336 break;
3337 }
3338 }
3339 rcu_read_unlock();
3340
3341 if (!sinfo)
3342 continue;
3343
3344 down_read(&sinfo->groups_sem);
3345 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3346 if (!list_empty(&sinfo->block_groups[c])) {
3347 u64 flags;
3348
3349 btrfs_get_block_group_info(
3350 &sinfo->block_groups[c], &space);
3351 if (space.total_bytes == 0 ||
3352 space.used_bytes == 0)
3353 continue;
3354 flags = space.flags;
3355 /*
3356 * return
3357 * 0: if dup, single or RAID0 is configured for
3358 * any of metadata, system or data, else
3359 * 1: if RAID5 is configured, or if RAID1 or
3360 * RAID10 is configured and only two mirrors
3361 * are used, else
3362 * 2: if RAID6 is configured, else
3363 * num_mirrors - 1: if RAID1 or RAID10 is
3364 * configured and more than
3365 * 2 mirrors are used.
3366 */
3367 if (num_tolerated_disk_barrier_failures > 0 &&
3368 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3369 BTRFS_BLOCK_GROUP_RAID0)) ||
3370 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3371 == 0)))
3372 num_tolerated_disk_barrier_failures = 0;
53b381b3
DW
3373 else if (num_tolerated_disk_barrier_failures > 1) {
3374 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3375 BTRFS_BLOCK_GROUP_RAID5 |
3376 BTRFS_BLOCK_GROUP_RAID10)) {
3377 num_tolerated_disk_barrier_failures = 1;
3378 } else if (flags &
15b0a89d 3379 BTRFS_BLOCK_GROUP_RAID6) {
53b381b3
DW
3380 num_tolerated_disk_barrier_failures = 2;
3381 }
3382 }
5af3e8cc
SB
3383 }
3384 }
3385 up_read(&sinfo->groups_sem);
3386 }
3387
3388 return num_tolerated_disk_barrier_failures;
3389}
3390
48a3b636 3391static int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 3392{
e5e9a520 3393 struct list_head *head;
f2984462 3394 struct btrfs_device *dev;
a061fc8d 3395 struct btrfs_super_block *sb;
f2984462 3396 struct btrfs_dev_item *dev_item;
f2984462
CM
3397 int ret;
3398 int do_barriers;
a236aed1
CM
3399 int max_errors;
3400 int total_errors = 0;
a061fc8d 3401 u64 flags;
f2984462
CM
3402
3403 do_barriers = !btrfs_test_opt(root, NOBARRIER);
af31f5e5 3404 backup_super_roots(root->fs_info);
f2984462 3405
6c41761f 3406 sb = root->fs_info->super_for_commit;
a061fc8d 3407 dev_item = &sb->dev_item;
e5e9a520 3408
174ba509 3409 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 3410 head = &root->fs_info->fs_devices->devices;
d7306801 3411 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
387125fc 3412
5af3e8cc
SB
3413 if (do_barriers) {
3414 ret = barrier_all_devices(root->fs_info);
3415 if (ret) {
3416 mutex_unlock(
3417 &root->fs_info->fs_devices->device_list_mutex);
3418 btrfs_error(root->fs_info, ret,
3419 "errors while submitting device barriers.");
3420 return ret;
3421 }
3422 }
387125fc 3423
1f78160c 3424 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM
3425 if (!dev->bdev) {
3426 total_errors++;
3427 continue;
3428 }
2b82032c 3429 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM
3430 continue;
3431
2b82032c 3432 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM
3433 btrfs_set_stack_device_type(dev_item, dev->type);
3434 btrfs_set_stack_device_id(dev_item, dev->devid);
3435 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
3436 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
3437 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3438 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3439 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3440 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 3441 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 3442
a061fc8d
CM
3443 flags = btrfs_super_flags(sb);
3444 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3445
a512bbf8 3446 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM
3447 if (ret)
3448 total_errors++;
f2984462 3449 }
a236aed1 3450 if (total_errors > max_errors) {
efe120a0 3451 btrfs_err(root->fs_info, "%d errors while writing supers",
d397712b 3452 total_errors);
a724b436 3453 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
79787eaa 3454
9d565ba4
SB
3455 /* FUA is masked off if unsupported and can't be the reason */
3456 btrfs_error(root->fs_info, -EIO,
3457 "%d errors while writing supers", total_errors);
3458 return -EIO;
a236aed1 3459 }
f2984462 3460
a512bbf8 3461 total_errors = 0;
1f78160c 3462 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM
3463 if (!dev->bdev)
3464 continue;
2b82032c 3465 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM
3466 continue;
3467
a512bbf8
YZ
3468 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3469 if (ret)
3470 total_errors++;
f2984462 3471 }
174ba509 3472 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 3473 if (total_errors > max_errors) {
79787eaa
JM
3474 btrfs_error(root->fs_info, -EIO,
3475 "%d errors while writing supers", total_errors);
3476 return -EIO;
a236aed1 3477 }
f2984462
CM
3478 return 0;
3479}
3480
a512bbf8
YZ
3481int write_ctree_super(struct btrfs_trans_handle *trans,
3482 struct btrfs_root *root, int max_mirrors)
eb60ceac 3483{
f570e757 3484 return write_all_supers(root, max_mirrors);
cfaa7295
CM
3485}
3486
cb517eab
MX
3487/* Drop a fs root from the radix tree and free it. */
3488void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3489 struct btrfs_root *root)
2619ba1f 3490{
4df27c4d 3491 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM
3492 radix_tree_delete(&fs_info->fs_roots_radix,
3493 (unsigned long)root->root_key.objectid);
4df27c4d 3494 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ
3495
3496 if (btrfs_root_refs(&root->root_item) == 0)
3497 synchronize_srcu(&fs_info->subvol_srcu);
3498
1a4319cc 3499 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
3321719e 3500 btrfs_free_log(NULL, root);
3321719e 3501
581bb050
LZ
3502 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3503 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3504 free_fs_root(root);
4df27c4d
YZ
3505}
3506
3507static void free_fs_root(struct btrfs_root *root)
3508{
82d5902d 3509 iput(root->cache_inode);
4df27c4d 3510 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
1cb048f5
FDBM
3511 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3512 root->orphan_block_rsv = NULL;
0ee5dc67
AV
3513 if (root->anon_dev)
3514 free_anon_bdev(root->anon_dev);
8257b2dc
MX
3515 if (root->subv_writers)
3516 btrfs_free_subvolume_writers(root->subv_writers);
4df27c4d
YZ
3517 free_extent_buffer(root->node);
3518 free_extent_buffer(root->commit_root);
581bb050
LZ
3519 kfree(root->free_ino_ctl);
3520 kfree(root->free_ino_pinned);
d397712b 3521 kfree(root->name);
b0feb9d9 3522 btrfs_put_fs_root(root);
2619ba1f
CM
3523}
3524
cb517eab
MX
3525void btrfs_free_fs_root(struct btrfs_root *root)
3526{
3527 free_fs_root(root);
2619ba1f
CM
3528}
3529
c146afad 3530int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3531{
c146afad
YZ
3532 u64 root_objectid = 0;
3533 struct btrfs_root *gang[8];
65d33fd7
QW
3534 int i = 0;
3535 int err = 0;
3536 unsigned int ret = 0;
3537 int index;
e089f05c 3538
c146afad 3539 while (1) {
65d33fd7 3540 index = srcu_read_lock(&fs_info->subvol_srcu);
c146afad
YZ
3541 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3542 (void **)gang, root_objectid,
3543 ARRAY_SIZE(gang));
65d33fd7
QW
3544 if (!ret) {
3545 srcu_read_unlock(&fs_info->subvol_srcu, index);
c146afad 3546 break;
65d33fd7 3547 }
5d4f98a2 3548 root_objectid = gang[ret - 1]->root_key.objectid + 1;
65d33fd7 3549
c146afad 3550 for (i = 0; i < ret; i++) {
65d33fd7
QW
3551 /* Avoid to grab roots in dead_roots */
3552 if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3553 gang[i] = NULL;
3554 continue;
3555 }
3556 /* grab all the search result for later use */
3557 gang[i] = btrfs_grab_fs_root(gang[i]);
3558 }
3559 srcu_read_unlock(&fs_info->subvol_srcu, index);
66b4ffd1 3560
65d33fd7
QW
3561 for (i = 0; i < ret; i++) {
3562 if (!gang[i])
3563 continue;
c146afad 3564 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB
3565 err = btrfs_orphan_cleanup(gang[i]);
3566 if (err)
65d33fd7
QW
3567 break;
3568 btrfs_put_fs_root(gang[i]);
c146afad
YZ
3569 }
3570 root_objectid++;
3571 }
65d33fd7
QW
3572
3573 /* release the uncleaned roots due to error */
3574 for (; i < ret; i++) {
3575 if (gang[i])
3576 btrfs_put_fs_root(gang[i]);
3577 }
3578 return err;
c146afad 3579}
a2135011 3580
c146afad
YZ
3581int btrfs_commit_super(struct btrfs_root *root)
3582{
3583 struct btrfs_trans_handle *trans;
a74a4b97 3584
c146afad 3585 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 3586 btrfs_run_delayed_iputs(root);
c146afad 3587 mutex_unlock(&root->fs_info->cleaner_mutex);
9d1a2a3a 3588 wake_up_process(root->fs_info->cleaner_kthread);
c71bf099
YZ
3589
3590 /* wait until ongoing cleanup work done */
3591 down_write(&root->fs_info->cleanup_work_sem);
3592 up_write(&root->fs_info->cleanup_work_sem);
3593
7a7eaa40 3594 trans = btrfs_join_transaction(root);
3612b495
TI
3595 if (IS_ERR(trans))
3596 return PTR_ERR(trans);
d52c1bcc 3597 return btrfs_commit_transaction(trans, root);
c146afad
YZ
3598}
3599
3600int close_ctree(struct btrfs_root *root)
3601{
3602 struct btrfs_fs_info *fs_info = root->fs_info;
3603 int ret;
3604
3605 fs_info->closing = 1;
3606 smp_mb();
3607
803b2f54
SB
3608 /* wait for the uuid_scan task to finish */
3609 down(&fs_info->uuid_tree_rescan_sem);
3610 /* avoid complains from lockdep et al., set sem back to initial state */
3611 up(&fs_info->uuid_tree_rescan_sem);
3612
837d5b6e 3613 /* pause restriper - we want to resume on mount */
aa1b8cd4 3614 btrfs_pause_balance(fs_info);
837d5b6e 3615
8dabb742
SB
3616 btrfs_dev_replace_suspend_for_unmount(fs_info);
3617
aa1b8cd4 3618 btrfs_scrub_cancel(fs_info);
4cb5300b
CM
3619
3620 /* wait for any defraggers to finish */
3621 wait_event(fs_info->transaction_wait,
3622 (atomic_read(&fs_info->defrag_running) == 0));
3623
3624 /* clear out the rbtree of defraggable inodes */
26176e7c 3625 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3626
21c7e756
MX
3627 cancel_work_sync(&fs_info->async_reclaim_work);
3628
c146afad 3629 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 3630 ret = btrfs_commit_super(root);
3631 if (ret)
efe120a0 3632 btrfs_err(root->fs_info, "commit super ret %d", ret);
acce952b 3633 }
3634
87533c47 3635 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
68ce9682 3636 btrfs_error_commit_super(root);
0f7d52f4 3637
e3029d9f
AV
3638 kthread_stop(fs_info->transaction_kthread);
3639 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3640
f25784b3
YZ
3641 fs_info->closing = 2;
3642 smp_mb();
3643
bcef60f2
AJ
3644 btrfs_free_qgroup_config(root->fs_info);
3645
963d678b 3646 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
efe120a0 3647 btrfs_info(root->fs_info, "at unmount delalloc count %lld",
963d678b 3648 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3649 }
bcc63abb 3650
5ac1d209
JM
3651 btrfs_sysfs_remove_one(fs_info);
3652
c146afad 3653 del_fs_roots(fs_info);
d10c5f31 3654
1a4319cc
LB
3655 btrfs_put_block_group_cache(fs_info);
3656
2b1360da
JB
3657 btrfs_free_block_groups(fs_info);
3658
96192499
JB
3659 btrfs_stop_all_workers(fs_info);
3660
13e6c37b 3661 free_root_pointers(fs_info, 1);
9ad6b7bc 3662
13e6c37b 3663 iput(fs_info->btree_inode);
d6bfde87 3664
21adbd5c
SB
3665#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3666 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3667 btrfsic_unmount(root, fs_info->fs_devices);
3668#endif
3669
dfe25020 3670 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3671 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3672
e2d84521 3673 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 3674 percpu_counter_destroy(&fs_info->delalloc_bytes);
c404e0dc 3675 percpu_counter_destroy(&fs_info->bio_counter);
04160088 3676 bdi_destroy(&fs_info->bdi);
76dda93c 3677 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 3678
53b381b3
DW
3679 btrfs_free_stripe_hash_table(fs_info);
3680
1cb048f5
FDBM
3681 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3682 root->orphan_block_rsv = NULL;
3683
eb60ceac
CM
3684 return 0;
3685}
3686
b9fab919
CM
3687int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3688 int atomic)
5f39d397 3689{
1259ab75 3690 int ret;
727011e0 3691 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 3692
0b32f4bb 3693 ret = extent_buffer_uptodate(buf);
1259ab75
CM
3694 if (!ret)
3695 return ret;
3696
3697 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM
3698 parent_transid, atomic);
3699 if (ret == -EAGAIN)
3700 return ret;
1259ab75 3701 return !ret;
5f39d397
CM
3702}
3703
3704int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 3705{
0b32f4bb 3706 return set_extent_buffer_uptodate(buf);
5f39d397 3707}
6702ed49 3708
5f39d397
CM
3709void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3710{
06ea65a3 3711 struct btrfs_root *root;
5f39d397 3712 u64 transid = btrfs_header_generation(buf);
b9473439 3713 int was_dirty;
b4ce94de 3714
06ea65a3
JB
3715#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3716 /*
3717 * This is a fast path so only do this check if we have sanity tests
3718 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3719 * outside of the sanity tests.
3720 */
3721 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3722 return;
3723#endif
3724 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
b9447ef8 3725 btrfs_assert_tree_locked(buf);
31b1a2bd
JL
3726 if (transid != root->fs_info->generation)
3727 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
d397712b 3728 "found %llu running %llu\n",
c1c9ff7c 3729 buf->start, transid, root->fs_info->generation);
0b32f4bb 3730 was_dirty = set_extent_buffer_dirty(buf);
e2d84521
MX
3731 if (!was_dirty)
3732 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3733 buf->len,
3734 root->fs_info->dirty_metadata_batch);
eb60ceac
CM
3735}
3736
b53d3f5d
LB
3737static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3738 int flush_delayed)
16cdcec7
MX
3739{
3740 /*
3741 * looks as though older kernels can get into trouble with
3742 * this code, they end up stuck in balance_dirty_pages forever
3743 */
e2d84521 3744 int ret;
16cdcec7
MX
3745
3746 if (current->flags & PF_MEMALLOC)
3747 return;
3748
b53d3f5d
LB
3749 if (flush_delayed)
3750 btrfs_balance_delayed_items(root);
16cdcec7 3751
e2d84521
MX
3752 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3753 BTRFS_DIRTY_METADATA_THRESH);
3754 if (ret > 0) {
d0e1d66b
NJ
3755 balance_dirty_pages_ratelimited(
3756 root->fs_info->btree_inode->i_mapping);
16cdcec7
MX
3757 }
3758 return;
3759}
3760
b53d3f5d 3761void btrfs_btree_balance_dirty(struct btrfs_root *root)
35b7e476 3762{
b53d3f5d
LB
3763 __btrfs_btree_balance_dirty(root, 1);
3764}
585ad2c3 3765
b53d3f5d
LB
3766void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3767{
3768 __btrfs_btree_balance_dirty(root, 0);
35b7e476 3769}
6b80053d 3770
ca7a79ad 3771int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 3772{
727011e0 3773 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
0b32f4bb 3774 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
6b80053d 3775}
0da5468f 3776
fcd1f065 3777static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 3778 int read_only)
3779{
1104a885
DS
3780 /*
3781 * Placeholder for checks
3782 */
fcd1f065 3783 return 0;
acce952b 3784}
3785
48a3b636 3786static void btrfs_error_commit_super(struct btrfs_root *root)
acce952b 3787{
acce952b 3788 mutex_lock(&root->fs_info->cleaner_mutex);
3789 btrfs_run_delayed_iputs(root);
3790 mutex_unlock(&root->fs_info->cleaner_mutex);
3791
3792 down_write(&root->fs_info->cleanup_work_sem);
3793 up_write(&root->fs_info->cleanup_work_sem);
3794
3795 /* cleanup FS via transaction */
3796 btrfs_cleanup_transaction(root);
acce952b 3797}
3798
569e0f35
JB
3799static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
3800 struct btrfs_root *root)
acce952b 3801{
3802 struct btrfs_inode *btrfs_inode;
3803 struct list_head splice;
3804
3805 INIT_LIST_HEAD(&splice);
3806
3807 mutex_lock(&root->fs_info->ordered_operations_mutex);
199c2a9c 3808 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3809
569e0f35 3810 list_splice_init(&t->ordered_operations, &splice);
acce952b 3811 while (!list_empty(&splice)) {
3812 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3813 ordered_operations);
3814
3815 list_del_init(&btrfs_inode->ordered_operations);
199c2a9c 3816 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3817
3818 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3819
199c2a9c 3820 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3821 }
3822
199c2a9c 3823 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3824 mutex_unlock(&root->fs_info->ordered_operations_mutex);
acce952b 3825}
3826
143bede5 3827static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 3828{
acce952b 3829 struct btrfs_ordered_extent *ordered;
acce952b 3830
199c2a9c 3831 spin_lock(&root->ordered_extent_lock);
779880ef
JB
3832 /*
3833 * This will just short circuit the ordered completion stuff which will
3834 * make sure the ordered extent gets properly cleaned up.
3835 */
199c2a9c 3836 list_for_each_entry(ordered, &root->ordered_extents,
779880ef
JB
3837 root_extent_list)
3838 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
199c2a9c
MX
3839 spin_unlock(&root->ordered_extent_lock);
3840}
3841
3842static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3843{
3844 struct btrfs_root *root;
3845 struct list_head splice;
3846
3847 INIT_LIST_HEAD(&splice);
3848
3849 spin_lock(&fs_info->ordered_root_lock);
3850 list_splice_init(&fs_info->ordered_roots, &splice);
3851 while (!list_empty(&splice)) {
3852 root = list_first_entry(&splice, struct btrfs_root,
3853 ordered_root);
1de2cfde
JB
3854 list_move_tail(&root->ordered_root,
3855 &fs_info->ordered_roots);
199c2a9c 3856
2a85d9ca 3857 spin_unlock(&fs_info->ordered_root_lock);
199c2a9c
MX
3858 btrfs_destroy_ordered_extents(root);
3859
2a85d9ca
LB
3860 cond_resched();
3861 spin_lock(&fs_info->ordered_root_lock);
199c2a9c
MX
3862 }
3863 spin_unlock(&fs_info->ordered_root_lock);
acce952b 3864}
3865
35a3621b
SB
3866static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3867 struct btrfs_root *root)
acce952b 3868{
3869 struct rb_node *node;
3870 struct btrfs_delayed_ref_root *delayed_refs;
3871 struct btrfs_delayed_ref_node *ref;
3872 int ret = 0;
3873
3874 delayed_refs = &trans->delayed_refs;
3875
3876 spin_lock(&delayed_refs->lock);
d7df2c79 3877 if (atomic_read(&delayed_refs->num_entries) == 0) {
cfece4db 3878 spin_unlock(&delayed_refs->lock);
efe120a0 3879 btrfs_info(root->fs_info, "delayed_refs has NO entry");
acce952b 3880 return ret;
3881 }
3882
d7df2c79
JB
3883 while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
3884 struct btrfs_delayed_ref_head *head;
e78417d1 3885 bool pin_bytes = false;
acce952b 3886
d7df2c79
JB
3887 head = rb_entry(node, struct btrfs_delayed_ref_head,
3888 href_node);
3889 if (!mutex_trylock(&head->mutex)) {
3890 atomic_inc(&head->node.refs);
3891 spin_unlock(&delayed_refs->lock);
eb12db69 3892
d7df2c79 3893 mutex_lock(&head->mutex);
e78417d1 3894 mutex_unlock(&head->mutex);
d7df2c79
JB
3895 btrfs_put_delayed_ref(&head->node);
3896 spin_lock(&delayed_refs->lock);
3897 continue;
3898 }
3899 spin_lock(&head->lock);
3900 while ((node = rb_first(&head->ref_root)) != NULL) {
3901 ref = rb_entry(node, struct btrfs_delayed_ref_node,
3902 rb_node);
3903 ref->in_tree = 0;
3904 rb_erase(&ref->rb_node, &head->ref_root);
3905 atomic_dec(&delayed_refs->num_entries);
3906 btrfs_put_delayed_ref(ref);
e78417d1 3907 }
d7df2c79
JB
3908 if (head->must_insert_reserved)
3909 pin_bytes = true;
3910 btrfs_free_delayed_extent_op(head->extent_op);
3911 delayed_refs->num_heads--;
3912 if (head->processing == 0)
3913 delayed_refs->num_heads_ready--;
3914 atomic_dec(&delayed_refs->num_entries);
3915 head->node.in_tree = 0;
3916 rb_erase(&head->href_node, &delayed_refs->href_root);
3917 spin_unlock(&head->lock);
3918 spin_unlock(&delayed_refs->lock);
3919 mutex_unlock(&head->mutex);
acce952b 3920
d7df2c79
JB
3921 if (pin_bytes)
3922 btrfs_pin_extent(root, head->node.bytenr,
3923 head->node.num_bytes, 1);
3924 btrfs_put_delayed_ref(&head->node);
acce952b 3925 cond_resched();
3926 spin_lock(&delayed_refs->lock);
3927 }
3928
3929 spin_unlock(&delayed_refs->lock);
3930
3931 return ret;
3932}
3933
143bede5 3934static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 3935{
3936 struct btrfs_inode *btrfs_inode;
3937 struct list_head splice;
3938
3939 INIT_LIST_HEAD(&splice);
3940
eb73c1b7
MX
3941 spin_lock(&root->delalloc_lock);
3942 list_splice_init(&root->delalloc_inodes, &splice);
acce952b 3943
3944 while (!list_empty(&splice)) {
eb73c1b7
MX
3945 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3946 delalloc_inodes);
acce952b 3947
3948 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX
3949 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3950 &btrfs_inode->runtime_flags);
eb73c1b7 3951 spin_unlock(&root->delalloc_lock);
acce952b 3952
3953 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3954
eb73c1b7 3955 spin_lock(&root->delalloc_lock);
acce952b 3956 }
3957
eb73c1b7
MX
3958 spin_unlock(&root->delalloc_lock);
3959}
3960
3961static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3962{
3963 struct btrfs_root *root;
3964 struct list_head splice;
3965
3966 INIT_LIST_HEAD(&splice);
3967
3968 spin_lock(&fs_info->delalloc_root_lock);
3969 list_splice_init(&fs_info->delalloc_roots, &splice);
3970 while (!list_empty(&splice)) {
3971 root = list_first_entry(&splice, struct btrfs_root,
3972 delalloc_root);
3973 list_del_init(&root->delalloc_root);
3974 root = btrfs_grab_fs_root(root);
3975 BUG_ON(!root);
3976 spin_unlock(&fs_info->delalloc_root_lock);
3977
3978 btrfs_destroy_delalloc_inodes(root);
3979 btrfs_put_fs_root(root);
3980
3981 spin_lock(&fs_info->delalloc_root_lock);
3982 }
3983 spin_unlock(&fs_info->delalloc_root_lock);
acce952b 3984}
3985
3986static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3987 struct extent_io_tree *dirty_pages,
3988 int mark)
3989{
3990 int ret;
acce952b 3991 struct extent_buffer *eb;
3992 u64 start = 0;
3993 u64 end;
acce952b 3994
3995 while (1) {
3996 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 3997 mark, NULL);
acce952b 3998 if (ret)
3999 break;
4000
4001 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
4002 while (start <= end) {
fd8b2b61
JB
4003 eb = btrfs_find_tree_block(root, start,
4004 root->leafsize);
69a85bd8 4005 start += root->leafsize;
fd8b2b61 4006 if (!eb)
acce952b 4007 continue;
fd8b2b61 4008 wait_on_extent_buffer_writeback(eb);
acce952b 4009
fd8b2b61
JB
4010 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4011 &eb->bflags))
4012 clear_extent_buffer_dirty(eb);
4013 free_extent_buffer_stale(eb);
acce952b 4014 }
4015 }
4016
4017 return ret;
4018}
4019
4020static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4021 struct extent_io_tree *pinned_extents)
4022{
4023 struct extent_io_tree *unpin;
4024 u64 start;
4025 u64 end;
4026 int ret;
ed0eaa14 4027 bool loop = true;
acce952b 4028
4029 unpin = pinned_extents;
ed0eaa14 4030again:
acce952b 4031 while (1) {
4032 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 4033 EXTENT_DIRTY, NULL);
acce952b 4034 if (ret)
4035 break;
4036
4037 /* opt_discard */
5378e607
LD
4038 if (btrfs_test_opt(root, DISCARD))
4039 ret = btrfs_error_discard_extent(root, start,
4040 end + 1 - start,
4041 NULL);
acce952b 4042
4043 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4044 btrfs_error_unpin_extent_range(root, start, end);
4045 cond_resched();
4046 }
4047
ed0eaa14
LB
4048 if (loop) {
4049 if (unpin == &root->fs_info->freed_extents[0])
4050 unpin = &root->fs_info->freed_extents[1];
4051 else
4052 unpin = &root->fs_info->freed_extents[0];
4053 loop = false;
4054 goto again;
4055 }
4056
acce952b 4057 return 0;
4058}
4059
49b25e05
JM
4060void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4061 struct btrfs_root *root)
4062{
724e2315
JB
4063 btrfs_destroy_ordered_operations(cur_trans, root);
4064
49b25e05 4065 btrfs_destroy_delayed_refs(cur_trans, root);
49b25e05 4066
4a9d8bde 4067 cur_trans->state = TRANS_STATE_COMMIT_START;
d7096fc3 4068 wake_up(&root->fs_info->transaction_blocked_wait);
49b25e05 4069
4a9d8bde 4070 cur_trans->state = TRANS_STATE_UNBLOCKED;
d7096fc3 4071 wake_up(&root->fs_info->transaction_wait);
49b25e05 4072
67cde344
MX
4073 btrfs_destroy_delayed_inodes(root);
4074 btrfs_assert_delayed_root_empty(root);
49b25e05 4075
49b25e05
JM
4076 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4077 EXTENT_DIRTY);
6e841e32
LB
4078 btrfs_destroy_pinned_extent(root,
4079 root->fs_info->pinned_extents);
49b25e05 4080
4a9d8bde
MX
4081 cur_trans->state =TRANS_STATE_COMPLETED;
4082 wake_up(&cur_trans->commit_wait);
4083
49b25e05
JM
4084 /*
4085 memset(cur_trans, 0, sizeof(*cur_trans));
4086 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4087 */
4088}
4089
48a3b636 4090static int btrfs_cleanup_transaction(struct btrfs_root *root)
acce952b 4091{
4092 struct btrfs_transaction *t;
acce952b 4093
acce952b 4094 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4095
a4abeea4 4096 spin_lock(&root->fs_info->trans_lock);
724e2315
JB
4097 while (!list_empty(&root->fs_info->trans_list)) {
4098 t = list_first_entry(&root->fs_info->trans_list,
4099 struct btrfs_transaction, list);
4100 if (t->state >= TRANS_STATE_COMMIT_START) {
4101 atomic_inc(&t->use_count);
4102 spin_unlock(&root->fs_info->trans_lock);
4103 btrfs_wait_for_commit(root, t->transid);
4104 btrfs_put_transaction(t);
4105 spin_lock(&root->fs_info->trans_lock);
4106 continue;
4107 }
4108 if (t == root->fs_info->running_transaction) {
4109 t->state = TRANS_STATE_COMMIT_DOING;
4110 spin_unlock(&root->fs_info->trans_lock);
4111 /*
4112 * We wait for 0 num_writers since we don't hold a trans
4113 * handle open currently for this transaction.
4114 */
4115 wait_event(t->writer_wait,
4116 atomic_read(&t->num_writers) == 0);
4117 } else {
4118 spin_unlock(&root->fs_info->trans_lock);
4119 }
4120 btrfs_cleanup_one_transaction(t, root);
4a9d8bde 4121
724e2315
JB
4122 spin_lock(&root->fs_info->trans_lock);
4123 if (t == root->fs_info->running_transaction)
4124 root->fs_info->running_transaction = NULL;
acce952b 4125 list_del_init(&t->list);
724e2315 4126 spin_unlock(&root->fs_info->trans_lock);
acce952b 4127
724e2315
JB
4128 btrfs_put_transaction(t);
4129 trace_btrfs_transaction_commit(root);
4130 spin_lock(&root->fs_info->trans_lock);
4131 }
4132 spin_unlock(&root->fs_info->trans_lock);
4133 btrfs_destroy_all_ordered_extents(root->fs_info);
4134 btrfs_destroy_delayed_inodes(root);
4135 btrfs_assert_delayed_root_empty(root);
4136 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4137 btrfs_destroy_all_delalloc_inodes(root->fs_info);
acce952b 4138 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 4139
4140 return 0;
4141}
4142
d1310b2e 4143static struct extent_io_ops btree_extent_io_ops = {
ce9adaa5 4144 .readpage_end_io_hook = btree_readpage_end_io_hook,
4bb31e92 4145 .readpage_io_failed_hook = btree_io_failed_hook,
0b86a832 4146 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM
4147 /* note we're sharing with inode.c for the merge bio hook */
4148 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 4149};