projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
block: Really silence spurious compiler warnings
[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>
163e783e 29#include <linux/crc32c.h>
5a0e3ad6 30#include <linux/slab.h>
784b4e29 31#include <linux/migrate.h>
7a36ddec 32#include <linux/ratelimit.h>
6463fe58 33#include <linux/uuid.h>
803b2f54 34#include <linux/semaphore.h>
7e75bf3f 35#include <asm/unaligned.h>
eb60ceac
CM		 36#include "ctree.h"
37#include "disk-io.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"
eb60ceac 51
de0022b9
JB		 52#ifdef CONFIG_X86
53#include <asm/cpufeature.h>
54#endif
55
d1310b2e 56static struct extent_io_ops btree_extent_io_ops;
8b712842 57static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 58static void free_fs_root(struct btrfs_root *root);
fcd1f065 59static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 60 int read_only);
569e0f35
JB		 61static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
62 struct btrfs_root *root);
143bede5 63static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
acce952b 64static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
65 struct btrfs_root *root);
143bede5 66static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
acce952b 67static int btrfs_destroy_marked_extents(struct btrfs_root *root,
68 struct extent_io_tree *dirty_pages,
69 int mark);
70static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
71 struct extent_io_tree *pinned_extents);
48a3b636
ES		 72static int btrfs_cleanup_transaction(struct btrfs_root *root);
73static void btrfs_error_commit_super(struct btrfs_root *root);
ce9adaa5 74
d352ac68
CM		 75/*
76 * end_io_wq structs are used to do processing in task context when an IO is
77 * complete. This is used during reads to verify checksums, and it is used
78 * by writes to insert metadata for new file extents after IO is complete.
79 */
ce9adaa5
CM		 80struct end_io_wq {
81 struct bio *bio;
82 bio_end_io_t *end_io;
83 void *private;
84 struct btrfs_fs_info *info;
85 int error;
22c59948 86 int metadata;
ce9adaa5 87 struct list_head list;
8b712842 88 struct btrfs_work work;
ce9adaa5 89};
0da5468f 90
d352ac68
CM		 91/*
92 * async submit bios are used to offload expensive checksumming
93 * onto the worker threads. They checksum file and metadata bios
94 * just before they are sent down the IO stack.
95 */
44b8bd7e
CM		 96struct async_submit_bio {
97 struct inode *inode;
98 struct bio *bio;
99 struct list_head list;
4a69a410
CM		 100 extent_submit_bio_hook_t *submit_bio_start;
101 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM		 102 int rw;
103 int mirror_num;
c8b97818 104 unsigned long bio_flags;
eaf25d93
CM		 105 /*
106 * bio_offset is optional, can be used if the pages in the bio
107 * can't tell us where in the file the bio should go
108 */
109 u64 bio_offset;
8b712842 110 struct btrfs_work work;
79787eaa 111 int error;
44b8bd7e
CM		 112};
113
85d4e461
CM		 114/*
115 * Lockdep class keys for extent_buffer->lock's in this root. For a given
116 * eb, the lockdep key is determined by the btrfs_root it belongs to and
117 * the level the eb occupies in the tree.
118 *
119 * Different roots are used for different purposes and may nest inside each
120 * other and they require separate keysets. As lockdep keys should be
121 * static, assign keysets according to the purpose of the root as indicated
122 * by btrfs_root->objectid. This ensures that all special purpose roots
123 * have separate keysets.
4008c04a 124 *
85d4e461
CM		 125 * Lock-nesting across peer nodes is always done with the immediate parent
126 * node locked thus preventing deadlock. As lockdep doesn't know this, use
127 * subclass to avoid triggering lockdep warning in such cases.
4008c04a 128 *
85d4e461
CM		 129 * The key is set by the readpage_end_io_hook after the buffer has passed
130 * csum validation but before the pages are unlocked. It is also set by
131 * btrfs_init_new_buffer on freshly allocated blocks.
4008c04a 132 *
85d4e461
CM		 133 * We also add a check to make sure the highest level of the tree is the
134 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
135 * needs update as well.
4008c04a
CM		 136 */
137#ifdef CONFIG_DEBUG_LOCK_ALLOC
138# if BTRFS_MAX_LEVEL != 8
139# error
140# endif
85d4e461
CM		 141
142static struct btrfs_lockdep_keyset {
143 u64 id; /* root objectid */
144 const char *name_stem; /* lock name stem */
145 char names[BTRFS_MAX_LEVEL + 1][20];
146 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
147} btrfs_lockdep_keysets[] = {
148 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
149 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
150 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
151 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
152 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
153 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
60b62978 154 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
85d4e461
CM		 155 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
156 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
157 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
13fd8da9 158 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
85d4e461 159 { .id = 0, .name_stem = "tree" },
4008c04a 160};
85d4e461
CM		 161
162void __init btrfs_init_lockdep(void)
163{
164 int i, j;
165
166 /* initialize lockdep class names */
167 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
168 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
169
170 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
171 snprintf(ks->names[j], sizeof(ks->names[j]),
172 "btrfs-%s-%02d", ks->name_stem, j);
173 }
174}
175
176void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
177 int level)
178{
179 struct btrfs_lockdep_keyset *ks;
180
181 BUG_ON(level >= ARRAY_SIZE(ks->keys));
182
183 /* find the matching keyset, id 0 is the default entry */
184 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
185 if (ks->id == objectid)
186 break;
187
188 lockdep_set_class_and_name(&eb->lock,
189 &ks->keys[level], ks->names[level]);
190}
191
4008c04a
CM		 192#endif
193
d352ac68
CM		 194/*
195 * extents on the btree inode are pretty simple, there's one extent
196 * that covers the entire device
197 */
b2950863 198static struct extent_map *btree_get_extent(struct inode *inode,
306e16ce 199 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 200 int create)
7eccb903 201{
5f39d397
CM		 202 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
203 struct extent_map *em;
204 int ret;
205
890871be 206 read_lock(&em_tree->lock);
d1310b2e 207 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 208 if (em) {
209 em->bdev =
210 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 211 read_unlock(&em_tree->lock);
5f39d397 212 goto out;
a061fc8d 213 }
890871be 214 read_unlock(&em_tree->lock);
7b13b7b1 215
172ddd60 216 em = alloc_extent_map();
5f39d397
CM		 217 if (!em) {
218 em = ERR_PTR(-ENOMEM);
219 goto out;
220 }
221 em->start = 0;
0afbaf8c 222 em->len = (u64)-1;
c8b97818 223 em->block_len = (u64)-1;
5f39d397 224 em->block_start = 0;
a061fc8d 225 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 226
890871be 227 write_lock(&em_tree->lock);
09a2a8f9 228 ret = add_extent_mapping(em_tree, em, 0);
5f39d397
CM		 229 if (ret == -EEXIST) {
230 free_extent_map(em);
7b13b7b1 231 em = lookup_extent_mapping(em_tree, start, len);
b4f359ab 232 if (!em)
0433f20d 233 em = ERR_PTR(-EIO);
5f39d397 234 } else if (ret) {
7b13b7b1 235 free_extent_map(em);
0433f20d 236 em = ERR_PTR(ret);
5f39d397 237 }
890871be 238 write_unlock(&em_tree->lock);
7b13b7b1 239
5f39d397
CM		 240out:
241 return em;
7eccb903
CM		 242}
243
b0496686 244u32 btrfs_csum_data(char *data, u32 seed, size_t len)
19c00ddc 245{
163e783e 246 return crc32c(seed, data, len);
19c00ddc
CM		 247}
248
249void btrfs_csum_final(u32 crc, char *result)
250{
7e75bf3f 251 put_unaligned_le32(~crc, result);
19c00ddc
CM		 252}
253
d352ac68
CM		 254/*
255 * compute the csum for a btree block, and either verify it or write it
256 * into the csum field of the block.
257 */
19c00ddc
CM		 258static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
259 int verify)
260{
6c41761f 261 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
607d432d 262 char *result = NULL;
19c00ddc
CM		 263 unsigned long len;
264 unsigned long cur_len;
265 unsigned long offset = BTRFS_CSUM_SIZE;
19c00ddc
CM		 266 char *kaddr;
267 unsigned long map_start;
268 unsigned long map_len;
269 int err;
270 u32 crc = ~(u32)0;
607d432d 271 unsigned long inline_result;
19c00ddc
CM		 272
273 len = buf->len - offset;
d397712b 274 while (len > 0) {
19c00ddc 275 err = map_private_extent_buffer(buf, offset, 32,
a6591715 276 &kaddr, &map_start, &map_len);
d397712b 277 if (err)
19c00ddc 278 return 1;
19c00ddc 279 cur_len = min(len, map_len - (offset - map_start));
b0496686 280 crc = btrfs_csum_data(kaddr + offset - map_start,
19c00ddc
CM		 281 crc, cur_len);
282 len -= cur_len;
283 offset += cur_len;
19c00ddc 284 }
607d432d
JB		 285 if (csum_size > sizeof(inline_result)) {
286 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
287 if (!result)
288 return 1;
289 } else {
290 result = (char *)&inline_result;
291 }
292
19c00ddc
CM		 293 btrfs_csum_final(crc, result);
294
295 if (verify) {
607d432d 296 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 297 u32 val;
298 u32 found = 0;
607d432d 299 memcpy(&found, result, csum_size);
e4204ded 300
607d432d 301 read_extent_buffer(buf, &val, 0, csum_size);
7a36ddec 302 printk_ratelimited(KERN_INFO "btrfs: %s checksum verify "
193f284d
CM		 303 "failed on %llu wanted %X found %X "
304 "level %d\n",
c1c9ff7c
GU		 305 root->fs_info->sb->s_id, buf->start,
306 val, found, btrfs_header_level(buf));
607d432d
JB		 307 if (result != (char *)&inline_result)
308 kfree(result);
19c00ddc
CM		 309 return 1;
310 }
311 } else {
607d432d 312 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 313 }
607d432d
JB		 314 if (result != (char *)&inline_result)
315 kfree(result);
19c00ddc
CM		 316 return 0;
317}
318
d352ac68
CM		 319/*
320 * we can't consider a given block up to date unless the transid of the
321 * block matches the transid in the parent node's pointer. This is how we
322 * detect blocks that either didn't get written at all or got written
323 * in the wrong place.
324 */
1259ab75 325static int verify_parent_transid(struct extent_io_tree *io_tree,
b9fab919
CM		 326 struct extent_buffer *eb, u64 parent_transid,
327 int atomic)
1259ab75 328{
2ac55d41 329 struct extent_state *cached_state = NULL;
1259ab75
CM		 330 int ret;
331
332 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
333 return 0;
334
b9fab919
CM		 335 if (atomic)
336 return -EAGAIN;
337
2ac55d41 338 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
d0082371 339 0, &cached_state);
0b32f4bb 340 if (extent_buffer_uptodate(eb) &&
1259ab75
CM		 341 btrfs_header_generation(eb) == parent_transid) {
342 ret = 0;
343 goto out;
344 }
7a36ddec 345 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
193f284d 346 "found %llu\n",
c1c9ff7c 347 eb->start, parent_transid, btrfs_header_generation(eb));
1259ab75 348 ret = 1;
0b32f4bb 349 clear_extent_buffer_uptodate(eb);
33958dc6 350out:
2ac55d41
JB		 351 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
352 &cached_state, GFP_NOFS);
1259ab75 353 return ret;
1259ab75
CM		 354}
355
1104a885
DS		 356/*
357 * Return 0 if the superblock checksum type matches the checksum value of that
358 * algorithm. Pass the raw disk superblock data.
359 */
360static int btrfs_check_super_csum(char *raw_disk_sb)
361{
362 struct btrfs_super_block *disk_sb =
363 (struct btrfs_super_block *)raw_disk_sb;
364 u16 csum_type = btrfs_super_csum_type(disk_sb);
365 int ret = 0;
366
367 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
368 u32 crc = ~(u32)0;
369 const int csum_size = sizeof(crc);
370 char result[csum_size];
371
372 /*
373 * The super_block structure does not span the whole
374 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
375 * is filled with zeros and is included in the checkum.
376 */
377 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
378 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
379 btrfs_csum_final(crc, result);
380
381 if (memcmp(raw_disk_sb, result, csum_size))
382 ret = 1;
667e7d94
CM		 383
384 if (ret && btrfs_super_generation(disk_sb) < 10) {
385 printk(KERN_WARNING "btrfs: super block crcs don't match, older mkfs detected\n");
386 ret = 0;
387 }
1104a885
DS		 388 }
389
390 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
391 printk(KERN_ERR "btrfs: unsupported checksum algorithm %u\n",
392 csum_type);
393 ret = 1;
394 }
395
396 return ret;
397}
398
d352ac68
CM		 399/*
400 * helper to read a given tree block, doing retries as required when
401 * the checksums don't match and we have alternate mirrors to try.
402 */
f188591e
CM		 403static int btree_read_extent_buffer_pages(struct btrfs_root *root,
404 struct extent_buffer *eb,
ca7a79ad 405 u64 start, u64 parent_transid)
f188591e
CM		 406{
407 struct extent_io_tree *io_tree;
ea466794 408 int failed = 0;
f188591e
CM		 409 int ret;
410 int num_copies = 0;
411 int mirror_num = 0;
ea466794 412 int failed_mirror = 0;
f188591e 413
a826d6dc 414 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
f188591e
CM		 415 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
416 while (1) {
bb82ab88
AJ		 417 ret = read_extent_buffer_pages(io_tree, eb, start,
418 WAIT_COMPLETE,
f188591e 419 btree_get_extent, mirror_num);
256dd1bb
SB		 420 if (!ret) {
421 if (!verify_parent_transid(io_tree, eb,
b9fab919 422 parent_transid, 0))
256dd1bb
SB		 423 break;
424 else
425 ret = -EIO;
426 }
d397712b 427
a826d6dc
JB		 428 /*
429 * This buffer's crc is fine, but its contents are corrupted, so
430 * there is no reason to read the other copies, they won't be
431 * any less wrong.
432 */
433 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
ea466794
JB		 434 break;
435
5d964051 436 num_copies = btrfs_num_copies(root->fs_info,
f188591e 437 eb->start, eb->len);
4235298e 438 if (num_copies == 1)
ea466794 439 break;
4235298e 440
5cf1ab56
JB		 441 if (!failed_mirror) {
442 failed = 1;
443 failed_mirror = eb->read_mirror;
444 }
445
f188591e 446 mirror_num++;
ea466794
JB		 447 if (mirror_num == failed_mirror)
448 mirror_num++;
449
4235298e 450 if (mirror_num > num_copies)
ea466794 451 break;
f188591e 452 }
ea466794 453
c0901581 454 if (failed && !ret && failed_mirror)
ea466794
JB		 455 repair_eb_io_failure(root, eb, failed_mirror);
456
457 return ret;
f188591e 458}
19c00ddc 459
d352ac68 460/*
d397712b
CM		 461 * checksum a dirty tree block before IO. This has extra checks to make sure
462 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 463 */
d397712b 464
b2950863 465static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
19c00ddc 466{
d1310b2e 467 struct extent_io_tree *tree;
4eee4fa4 468 u64 start = page_offset(page);
19c00ddc 469 u64 found_start;
19c00ddc 470 struct extent_buffer *eb;
f188591e 471
d1310b2e 472 tree = &BTRFS_I(page->mapping->host)->io_tree;
19c00ddc 473
4f2de97a
JB		 474 eb = (struct extent_buffer *)page->private;
475 if (page != eb->pages[0])
476 return 0;
19c00ddc 477 found_start = btrfs_header_bytenr(eb);
fae7f21c 478 if (WARN_ON(found_start != start || !PageUptodate(page)))
4f2de97a 479 return 0;
19c00ddc 480 csum_tree_block(root, eb, 0);
19c00ddc
CM		 481 return 0;
482}
483
2b82032c
YZ		 484static int check_tree_block_fsid(struct btrfs_root *root,
485 struct extent_buffer *eb)
486{
487 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
488 u8 fsid[BTRFS_UUID_SIZE];
489 int ret = 1;
490
0a4e5586 491 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
2b82032c
YZ		 492 while (fs_devices) {
493 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
494 ret = 0;
495 break;
496 }
497 fs_devices = fs_devices->seed;
498 }
499 return ret;
500}
501
a826d6dc
JB		 502#define CORRUPT(reason, eb, root, slot) \
503 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
504 "root=%llu, slot=%d\n", reason, \
c1c9ff7c 505 btrfs_header_bytenr(eb), root->objectid, slot)
a826d6dc
JB		 506
507static noinline int check_leaf(struct btrfs_root *root,
508 struct extent_buffer *leaf)
509{
510 struct btrfs_key key;
511 struct btrfs_key leaf_key;
512 u32 nritems = btrfs_header_nritems(leaf);
513 int slot;
514
515 if (nritems == 0)
516 return 0;
517
518 /* Check the 0 item */
519 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
520 BTRFS_LEAF_DATA_SIZE(root)) {
521 CORRUPT("invalid item offset size pair", leaf, root, 0);
522 return -EIO;
523 }
524
525 /*
526 * Check to make sure each items keys are in the correct order and their
527 * offsets make sense. We only have to loop through nritems-1 because
528 * we check the current slot against the next slot, which verifies the
529 * next slot's offset+size makes sense and that the current's slot
530 * offset is correct.
531 */
532 for (slot = 0; slot < nritems - 1; slot++) {
533 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
534 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
535
536 /* Make sure the keys are in the right order */
537 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
538 CORRUPT("bad key order", leaf, root, slot);
539 return -EIO;
540 }
541
542 /*
543 * Make sure the offset and ends are right, remember that the
544 * item data starts at the end of the leaf and grows towards the
545 * front.
546 */
547 if (btrfs_item_offset_nr(leaf, slot) !=
548 btrfs_item_end_nr(leaf, slot + 1)) {
549 CORRUPT("slot offset bad", leaf, root, slot);
550 return -EIO;
551 }
552
553 /*
554 * Check to make sure that we don't point outside of the leaf,
555 * just incase all the items are consistent to eachother, but
556 * all point outside of the leaf.
557 */
558 if (btrfs_item_end_nr(leaf, slot) >
559 BTRFS_LEAF_DATA_SIZE(root)) {
560 CORRUPT("slot end outside of leaf", leaf, root, slot);
561 return -EIO;
562 }
563 }
564
565 return 0;
566}
567
facc8a22
MX		 568static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
569 u64 phy_offset, struct page *page,
570 u64 start, u64 end, int mirror)
ce9adaa5
CM		 571{
572 struct extent_io_tree *tree;
573 u64 found_start;
574 int found_level;
ce9adaa5
CM		 575 struct extent_buffer *eb;
576 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 577 int ret = 0;
727011e0 578 int reads_done;
ce9adaa5 579
ce9adaa5
CM		 580 if (!page->private)
581 goto out;
d397712b 582
727011e0 583 tree = &BTRFS_I(page->mapping->host)->io_tree;
4f2de97a 584 eb = (struct extent_buffer *)page->private;
d397712b 585
0b32f4bb
JB		 586 /* the pending IO might have been the only thing that kept this buffer
587 * in memory. Make sure we have a ref for all this other checks
588 */
589 extent_buffer_get(eb);
590
591 reads_done = atomic_dec_and_test(&eb->io_pages);
727011e0
CM		 592 if (!reads_done)
593 goto err;
f188591e 594
5cf1ab56 595 eb->read_mirror = mirror;
ea466794
JB		 596 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
597 ret = -EIO;
598 goto err;
599 }
600
ce9adaa5 601 found_start = btrfs_header_bytenr(eb);
727011e0 602 if (found_start != eb->start) {
7a36ddec 603 printk_ratelimited(KERN_INFO "btrfs bad tree block start "
193f284d 604 "%llu %llu\n",
c1c9ff7c 605 found_start, eb->start);
f188591e 606 ret = -EIO;
ce9adaa5
CM		 607 goto err;
608 }
2b82032c 609 if (check_tree_block_fsid(root, eb)) {
7a36ddec 610 printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n",
c1c9ff7c 611 eb->start);
1259ab75
CM		 612 ret = -EIO;
613 goto err;
614 }
ce9adaa5 615 found_level = btrfs_header_level(eb);
1c24c3ce
JB		 616 if (found_level >= BTRFS_MAX_LEVEL) {
617 btrfs_info(root->fs_info, "bad tree block level %d\n",
618 (int)btrfs_header_level(eb));
619 ret = -EIO;
620 goto err;
621 }
ce9adaa5 622
85d4e461
CM		 623 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
624 eb, found_level);
4008c04a 625
ce9adaa5 626 ret = csum_tree_block(root, eb, 1);
a826d6dc 627 if (ret) {
f188591e 628 ret = -EIO;
a826d6dc
JB		 629 goto err;
630 }
631
632 /*
633 * If this is a leaf block and it is corrupt, set the corrupt bit so
634 * that we don't try and read the other copies of this block, just
635 * return -EIO.
636 */
637 if (found_level == 0 && check_leaf(root, eb)) {
638 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
639 ret = -EIO;
640 }
ce9adaa5 641
0b32f4bb
JB		 642 if (!ret)
643 set_extent_buffer_uptodate(eb);
ce9adaa5 644err:
79fb65a1
JB		 645 if (reads_done &&
646 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 647 btree_readahead_hook(root, eb, eb->start, ret);
4bb31e92 648
53b381b3
DW		 649 if (ret) {
650 /*
651 * our io error hook is going to dec the io pages
652 * again, we have to make sure it has something
653 * to decrement
654 */
655 atomic_inc(&eb->io_pages);
0b32f4bb 656 clear_extent_buffer_uptodate(eb);
53b381b3 657 }
0b32f4bb 658 free_extent_buffer(eb);
ce9adaa5 659out:
f188591e 660 return ret;
ce9adaa5
CM		 661}
662
ea466794 663static int btree_io_failed_hook(struct page *page, int failed_mirror)
4bb31e92 664{
4bb31e92
AJ		 665 struct extent_buffer *eb;
666 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
667
4f2de97a 668 eb = (struct extent_buffer *)page->private;
ea466794 669 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
5cf1ab56 670 eb->read_mirror = failed_mirror;
53b381b3 671 atomic_dec(&eb->io_pages);
ea466794 672 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 673 btree_readahead_hook(root, eb, eb->start, -EIO);
4bb31e92
AJ		 674 return -EIO; /* we fixed nothing */
675}
676
ce9adaa5 677static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM		 678{
679 struct end_io_wq *end_io_wq = bio->bi_private;
680 struct btrfs_fs_info *fs_info;
ce9adaa5 681
ce9adaa5 682 fs_info = end_io_wq->info;
ce9adaa5 683 end_io_wq->error = err;
8b712842
CM		 684 end_io_wq->work.func = end_workqueue_fn;
685 end_io_wq->work.flags = 0;
d20f7043 686
7b6d91da 687 if (bio->bi_rw & REQ_WRITE) {
53b381b3 688 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA)
cad321ad
CM		 689 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
690 &end_io_wq->work);
53b381b3 691 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE)
0cb59c99
JB		 692 btrfs_queue_worker(&fs_info->endio_freespace_worker,
693 &end_io_wq->work);
53b381b3
DW		 694 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
695 btrfs_queue_worker(&fs_info->endio_raid56_workers,
696 &end_io_wq->work);
cad321ad
CM		 697 else
698 btrfs_queue_worker(&fs_info->endio_write_workers,
699 &end_io_wq->work);
d20f7043 700 } else {
53b381b3
DW		 701 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
702 btrfs_queue_worker(&fs_info->endio_raid56_workers,
703 &end_io_wq->work);
704 else if (end_io_wq->metadata)
d20f7043
CM		 705 btrfs_queue_worker(&fs_info->endio_meta_workers,
706 &end_io_wq->work);
707 else
708 btrfs_queue_worker(&fs_info->endio_workers,
709 &end_io_wq->work);
710 }
ce9adaa5
CM		 711}
712
0cb59c99
JB		 713/*
714 * For the metadata arg you want
715 *
716 * 0 - if data
717 * 1 - if normal metadta
718 * 2 - if writing to the free space cache area
53b381b3 719 * 3 - raid parity work
0cb59c99 720 */
22c59948
CM		 721int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
722 int metadata)
0b86a832 723{
ce9adaa5 724 struct end_io_wq *end_io_wq;
ce9adaa5
CM		 725 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
726 if (!end_io_wq)
727 return -ENOMEM;
728
729 end_io_wq->private = bio->bi_private;
730 end_io_wq->end_io = bio->bi_end_io;
22c59948 731 end_io_wq->info = info;
ce9adaa5
CM		 732 end_io_wq->error = 0;
733 end_io_wq->bio = bio;
22c59948 734 end_io_wq->metadata = metadata;
ce9adaa5
CM		 735
736 bio->bi_private = end_io_wq;
737 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 738 return 0;
739}
740
b64a2851 741unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 742{
4854ddd0
CM		 743 unsigned long limit = min_t(unsigned long,
744 info->workers.max_workers,
745 info->fs_devices->open_devices);
746 return 256 * limit;
747}
0986fe9e 748
4a69a410
CM		 749static void run_one_async_start(struct btrfs_work *work)
750{
4a69a410 751 struct async_submit_bio *async;
79787eaa 752 int ret;
4a69a410
CM		 753
754 async = container_of(work, struct async_submit_bio, work);
79787eaa
JM		 755 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
756 async->mirror_num, async->bio_flags,
757 async->bio_offset);
758 if (ret)
759 async->error = ret;
4a69a410
CM		 760}
761
762static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 763{
764 struct btrfs_fs_info *fs_info;
765 struct async_submit_bio *async;
4854ddd0 766 int limit;
8b712842
CM		 767
768 async = container_of(work, struct async_submit_bio, work);
769 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 770
b64a2851 771 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 772 limit = limit * 2 / 3;
773
66657b31 774 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
b64a2851 775 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 776 wake_up(&fs_info->async_submit_wait);
777
79787eaa
JM		 778 /* If an error occured we just want to clean up the bio and move on */
779 if (async->error) {
780 bio_endio(async->bio, async->error);
781 return;
782 }
783
4a69a410 784 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 785 async->mirror_num, async->bio_flags,
786 async->bio_offset);
4a69a410
CM		 787}
788
789static void run_one_async_free(struct btrfs_work *work)
790{
791 struct async_submit_bio *async;
792
793 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 794 kfree(async);
795}
796
44b8bd7e
CM		 797int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
798 int rw, struct bio *bio, int mirror_num,
c8b97818 799 unsigned long bio_flags,
eaf25d93 800 u64 bio_offset,
4a69a410
CM		 801 extent_submit_bio_hook_t *submit_bio_start,
802 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 803{
804 struct async_submit_bio *async;
805
806 async = kmalloc(sizeof(*async), GFP_NOFS);
807 if (!async)
808 return -ENOMEM;
809
810 async->inode = inode;
811 async->rw = rw;
812 async->bio = bio;
813 async->mirror_num = mirror_num;
4a69a410
CM		 814 async->submit_bio_start = submit_bio_start;
815 async->submit_bio_done = submit_bio_done;
816
817 async->work.func = run_one_async_start;
818 async->work.ordered_func = run_one_async_done;
819 async->work.ordered_free = run_one_async_free;
820
8b712842 821 async->work.flags = 0;
c8b97818 822 async->bio_flags = bio_flags;
eaf25d93 823 async->bio_offset = bio_offset;
8c8bee1d 824
79787eaa
JM		 825 async->error = 0;
826
cb03c743 827 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 828
7b6d91da 829 if (rw & REQ_SYNC)
d313d7a3
CM		 830 btrfs_set_work_high_prio(&async->work);
831
8b712842 832 btrfs_queue_worker(&fs_info->workers, &async->work);
9473f16c 833
d397712b 834 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 835 atomic_read(&fs_info->nr_async_submits)) {
836 wait_event(fs_info->async_submit_wait,
837 (atomic_read(&fs_info->nr_async_submits) == 0));
838 }
839
44b8bd7e
CM		 840 return 0;
841}
842
ce3ed71a
CM		 843static int btree_csum_one_bio(struct bio *bio)
844{
2c30c71b 845 struct bio_vec *bvec;
ce3ed71a 846 struct btrfs_root *root;
2c30c71b 847 int i, ret = 0;
ce3ed71a 848
2c30c71b 849 bio_for_each_segment_all(bvec, bio, i) {
ce3ed71a 850 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
79787eaa
JM		 851 ret = csum_dirty_buffer(root, bvec->bv_page);
852 if (ret)
853 break;
ce3ed71a 854 }
2c30c71b 855
79787eaa 856 return ret;
ce3ed71a
CM		 857}
858
4a69a410
CM		 859static int __btree_submit_bio_start(struct inode *inode, int rw,
860 struct bio *bio, int mirror_num,
eaf25d93
CM		 861 unsigned long bio_flags,
862 u64 bio_offset)
22c59948 863{
8b712842
CM		 864 /*
865 * when we're called for a write, we're already in the async
5443be45 866 * submission context. Just jump into btrfs_map_bio
8b712842 867 */
79787eaa 868 return btree_csum_one_bio(bio);
4a69a410 869}
22c59948 870
4a69a410 871static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 872 int mirror_num, unsigned long bio_flags,
873 u64 bio_offset)
4a69a410 874{
61891923
SB		 875 int ret;
876
8b712842 877 /*
4a69a410
CM		 878 * when we're called for a write, we're already in the async
879 * submission context. Just jump into btrfs_map_bio
8b712842 880 */
61891923
SB		 881 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
882 if (ret)
883 bio_endio(bio, ret);
884 return ret;
0b86a832
CM		 885}
886
de0022b9
JB		 887static int check_async_write(struct inode *inode, unsigned long bio_flags)
888{
889 if (bio_flags & EXTENT_BIO_TREE_LOG)
890 return 0;
891#ifdef CONFIG_X86
892 if (cpu_has_xmm4_2)
893 return 0;
894#endif
895 return 1;
896}
897
44b8bd7e 898static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 899 int mirror_num, unsigned long bio_flags,
900 u64 bio_offset)
44b8bd7e 901{
de0022b9 902 int async = check_async_write(inode, bio_flags);
cad321ad
CM		 903 int ret;
904
7b6d91da 905 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 906 /*
907 * called for a read, do the setup so that checksum validation
908 * can happen in the async kernel threads
909 */
f3f266ab
CM		 910 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
911 bio, 1);
1d4284bd 912 if (ret)
61891923
SB		 913 goto out_w_error;
914 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
915 mirror_num, 0);
de0022b9
JB		 916 } else if (!async) {
917 ret = btree_csum_one_bio(bio);
918 if (ret)
61891923
SB		 919 goto out_w_error;
920 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
921 mirror_num, 0);
922 } else {
923 /*
924 * kthread helpers are used to submit writes so that
925 * checksumming can happen in parallel across all CPUs
926 */
927 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
928 inode, rw, bio, mirror_num, 0,
929 bio_offset,
930 __btree_submit_bio_start,
931 __btree_submit_bio_done);
44b8bd7e 932 }
d313d7a3 933
61891923
SB		 934 if (ret) {
935out_w_error:
936 bio_endio(bio, ret);
937 }
938 return ret;
44b8bd7e
CM		 939}
940
3dd1462e 941#ifdef CONFIG_MIGRATION
784b4e29 942static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG		 943 struct page *newpage, struct page *page,
944 enum migrate_mode mode)
784b4e29
CM		 945{
946 /*
947 * we can't safely write a btree page from here,
948 * we haven't done the locking hook
949 */
950 if (PageDirty(page))
951 return -EAGAIN;
952 /*
953 * Buffers may be managed in a filesystem specific way.
954 * We must have no buffers or drop them.
955 */
956 if (page_has_private(page) &&
957 !try_to_release_page(page, GFP_KERNEL))
958 return -EAGAIN;
a6bc32b8 959 return migrate_page(mapping, newpage, page, mode);
784b4e29 960}
3dd1462e 961#endif
784b4e29 962
0da5468f
CM		 963
964static int btree_writepages(struct address_space *mapping,
965 struct writeback_control *wbc)
966{
d1310b2e 967 struct extent_io_tree *tree;
e2d84521
MX		 968 struct btrfs_fs_info *fs_info;
969 int ret;
970
d1310b2e 971 tree = &BTRFS_I(mapping->host)->io_tree;
d8d5f3e1 972 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM		 973
974 if (wbc->for_kupdate)
975 return 0;
976
e2d84521 977 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 978 /* this is a bit racy, but that's ok */
e2d84521
MX		 979 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
980 BTRFS_DIRTY_METADATA_THRESH);
981 if (ret < 0)
793955bc 982 return 0;
793955bc 983 }
0b32f4bb 984 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM		 985}
986
b2950863 987static int btree_readpage(struct file *file, struct page *page)
5f39d397 988{
d1310b2e
CM		 989 struct extent_io_tree *tree;
990 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 991 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 992}
22b0ebda 993
70dec807 994static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 995{
98509cfc 996 if (PageWriteback(page) || PageDirty(page))
d397712b 997 return 0;
0c4e538b 998
f7a52a40 999 return try_release_extent_buffer(page);
d98237b3
CM		 1000}
1001
d47992f8
LC		 1002static void btree_invalidatepage(struct page *page, unsigned int offset,
1003 unsigned int length)
d98237b3 1004{
d1310b2e
CM		 1005 struct extent_io_tree *tree;
1006 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 1007 extent_invalidatepage(tree, page, offset);
1008 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 1009 if (PagePrivate(page)) {
d397712b
CM		 1010 printk(KERN_WARNING "btrfs warning page private not zero "
1011 "on page %llu\n", (unsigned long long)page_offset(page));
9ad6b7bc
CM		 1012 ClearPagePrivate(page);
1013 set_page_private(page, 0);
1014 page_cache_release(page);
1015 }
d98237b3
CM		 1016}
1017
0b32f4bb
JB		 1018static int btree_set_page_dirty(struct page *page)
1019{
bb146eb2 1020#ifdef DEBUG
0b32f4bb
JB		 1021 struct extent_buffer *eb;
1022
1023 BUG_ON(!PagePrivate(page));
1024 eb = (struct extent_buffer *)page->private;
1025 BUG_ON(!eb);
1026 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1027 BUG_ON(!atomic_read(&eb->refs));
1028 btrfs_assert_tree_locked(eb);
bb146eb2 1029#endif
0b32f4bb
JB		 1030 return __set_page_dirty_nobuffers(page);
1031}
1032
7f09410b 1033static const struct address_space_operations btree_aops = {
d98237b3 1034 .readpage = btree_readpage,
0da5468f 1035 .writepages = btree_writepages,
5f39d397
CM		 1036 .releasepage = btree_releasepage,
1037 .invalidatepage = btree_invalidatepage,
5a92bc88 1038#ifdef CONFIG_MIGRATION
784b4e29 1039 .migratepage = btree_migratepage,
5a92bc88 1040#endif
0b32f4bb 1041 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM		 1042};
1043
ca7a79ad
CM		 1044int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1045 u64 parent_transid)
090d1875 1046{
5f39d397
CM		 1047 struct extent_buffer *buf = NULL;
1048 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 1049 int ret = 0;
090d1875 1050
db94535d 1051 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 1052 if (!buf)
090d1875 1053 return 0;
d1310b2e 1054 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
bb82ab88 1055 buf, 0, WAIT_NONE, btree_get_extent, 0);
5f39d397 1056 free_extent_buffer(buf);
de428b63 1057 return ret;
090d1875
CM		 1058}
1059
ab0fff03
AJ		 1060int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1061 int mirror_num, struct extent_buffer **eb)
1062{
1063 struct extent_buffer *buf = NULL;
1064 struct inode *btree_inode = root->fs_info->btree_inode;
1065 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1066 int ret;
1067
1068 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1069 if (!buf)
1070 return 0;
1071
1072 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1073
1074 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1075 btree_get_extent, mirror_num);
1076 if (ret) {
1077 free_extent_buffer(buf);
1078 return ret;
1079 }
1080
1081 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1082 free_extent_buffer(buf);
1083 return -EIO;
0b32f4bb 1084 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ		 1085 *eb = buf;
1086 } else {
1087 free_extent_buffer(buf);
1088 }
1089 return 0;
1090}
1091
0999df54
CM		 1092struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1093 u64 bytenr, u32 blocksize)
1094{
1095 struct inode *btree_inode = root->fs_info->btree_inode;
1096 struct extent_buffer *eb;
452c75c3 1097 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, bytenr);
0999df54
CM		 1098 return eb;
1099}
1100
1101struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1102 u64 bytenr, u32 blocksize)
1103{
1104 struct inode *btree_inode = root->fs_info->btree_inode;
1105 struct extent_buffer *eb;
1106
1107 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
727011e0 1108 bytenr, blocksize);
0999df54
CM		 1109 return eb;
1110}
1111
1112
e02119d5
CM		 1113int btrfs_write_tree_block(struct extent_buffer *buf)
1114{
727011e0 1115 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1116 buf->start + buf->len - 1);
e02119d5
CM		 1117}
1118
1119int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1120{
727011e0 1121 return filemap_fdatawait_range(buf->pages[0]->mapping,
8aa38c31 1122 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1123}
1124
0999df54 1125struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 1126 u32 blocksize, u64 parent_transid)
0999df54
CM		 1127{
1128 struct extent_buffer *buf = NULL;
0999df54
CM		 1129 int ret;
1130
0999df54
CM		 1131 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1132 if (!buf)
1133 return NULL;
0999df54 1134
ca7a79ad 1135 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
0f0fe8f7
FDBM		 1136 if (ret) {
1137 free_extent_buffer(buf);
1138 return NULL;
1139 }
5f39d397 1140 return buf;
ce9adaa5 1141
eb60ceac
CM		 1142}
1143
d5c13f92
JM		 1144void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1145 struct extent_buffer *buf)
ed2ff2cb 1146{
e2d84521
MX		 1147 struct btrfs_fs_info *fs_info = root->fs_info;
1148
55c69072 1149 if (btrfs_header_generation(buf) ==
e2d84521 1150 fs_info->running_transaction->transid) {
b9447ef8 1151 btrfs_assert_tree_locked(buf);
b4ce94de 1152
b9473439 1153 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
e2d84521
MX		 1154 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1155 -buf->len,
1156 fs_info->dirty_metadata_batch);
ed7b63eb
JB		 1157 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1158 btrfs_set_lock_blocking(buf);
1159 clear_extent_buffer_dirty(buf);
1160 }
925baedd 1161 }
5f39d397
CM		 1162}
1163
143bede5
JM		 1164static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1165 u32 stripesize, struct btrfs_root *root,
1166 struct btrfs_fs_info *fs_info,
1167 u64 objectid)
d97e63b6 1168{
cfaa7295 1169 root->node = NULL;
a28ec197 1170 root->commit_root = NULL;
db94535d
CM		 1171 root->sectorsize = sectorsize;
1172 root->nodesize = nodesize;
1173 root->leafsize = leafsize;
87ee04eb 1174 root->stripesize = stripesize;
123abc88 1175 root->ref_cows = 0;
0b86a832 1176 root->track_dirty = 0;
c71bf099 1177 root->in_radix = 0;
d68fc57b
YZ		 1178 root->orphan_item_inserted = 0;
1179 root->orphan_cleanup_state = 0;
0b86a832 1180
0f7d52f4
CM		 1181 root->objectid = objectid;
1182 root->last_trans = 0;
13a8a7c8 1183 root->highest_objectid = 0;
eb73c1b7 1184 root->nr_delalloc_inodes = 0;
199c2a9c 1185 root->nr_ordered_extents = 0;
58176a96 1186 root->name = NULL;
6bef4d31 1187 root->inode_tree = RB_ROOT;
16cdcec7 1188 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1189 root->block_rsv = NULL;
d68fc57b 1190 root->orphan_block_rsv = NULL;
0b86a832
CM		 1191
1192 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1193 INIT_LIST_HEAD(&root->root_list);
eb73c1b7
MX		 1194 INIT_LIST_HEAD(&root->delalloc_inodes);
1195 INIT_LIST_HEAD(&root->delalloc_root);
199c2a9c
MX		 1196 INIT_LIST_HEAD(&root->ordered_extents);
1197 INIT_LIST_HEAD(&root->ordered_root);
2ab28f32
JB		 1198 INIT_LIST_HEAD(&root->logged_list[0]);
1199 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1200 spin_lock_init(&root->orphan_lock);
5d4f98a2 1201 spin_lock_init(&root->inode_lock);
eb73c1b7 1202 spin_lock_init(&root->delalloc_lock);
199c2a9c 1203 spin_lock_init(&root->ordered_extent_lock);
f0486c68 1204 spin_lock_init(&root->accounting_lock);
2ab28f32
JB		 1205 spin_lock_init(&root->log_extents_lock[0]);
1206 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1207 mutex_init(&root->objectid_mutex);
e02119d5 1208 mutex_init(&root->log_mutex);
7237f183
YZ		 1209 init_waitqueue_head(&root->log_writer_wait);
1210 init_waitqueue_head(&root->log_commit_wait[0]);
1211 init_waitqueue_head(&root->log_commit_wait[1]);
1212 atomic_set(&root->log_commit[0], 0);
1213 atomic_set(&root->log_commit[1], 0);
1214 atomic_set(&root->log_writers, 0);
2ecb7923 1215 atomic_set(&root->log_batch, 0);
8a35d95f 1216 atomic_set(&root->orphan_inodes, 0);
b0feb9d9 1217 atomic_set(&root->refs, 1);
7237f183 1218 root->log_transid = 0;
257c62e1 1219 root->last_log_commit = 0;
06ea65a3
JB		 1220 if (fs_info)
1221 extent_io_tree_init(&root->dirty_log_pages,
1222 fs_info->btree_inode->i_mapping);
017e5369 1223
3768f368
CM		 1224 memset(&root->root_key, 0, sizeof(root->root_key));
1225 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1226 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1227 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
06ea65a3
JB		 1228 if (fs_info)
1229 root->defrag_trans_start = fs_info->generation;
1230 else
1231 root->defrag_trans_start = 0;
58176a96 1232 init_completion(&root->kobj_unregister);
6702ed49 1233 root->defrag_running = 0;
4d775673 1234 root->root_key.objectid = objectid;
0ee5dc67 1235 root->anon_dev = 0;
8ea05e3a 1236
5f3ab90a 1237 spin_lock_init(&root->root_item_lock);
3768f368
CM		 1238}
1239
f84a8bd6 1240static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
6f07e42e
AV		 1241{
1242 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1243 if (root)
1244 root->fs_info = fs_info;
1245 return root;
1246}
1247
06ea65a3
JB		 1248#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1249/* Should only be used by the testing infrastructure */
1250struct btrfs_root *btrfs_alloc_dummy_root(void)
1251{
1252 struct btrfs_root *root;
1253
1254 root = btrfs_alloc_root(NULL);
1255 if (!root)
1256 return ERR_PTR(-ENOMEM);
1257 __setup_root(4096, 4096, 4096, 4096, root, NULL, 1);
1258 root->dummy_root = 1;
1259
1260 return root;
1261}
1262#endif
1263
20897f5c
AJ		 1264struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1265 struct btrfs_fs_info *fs_info,
1266 u64 objectid)
1267{
1268 struct extent_buffer *leaf;
1269 struct btrfs_root *tree_root = fs_info->tree_root;
1270 struct btrfs_root *root;
1271 struct btrfs_key key;
1272 int ret = 0;
1273 u64 bytenr;
6463fe58 1274 uuid_le uuid;
20897f5c
AJ		 1275
1276 root = btrfs_alloc_root(fs_info);
1277 if (!root)
1278 return ERR_PTR(-ENOMEM);
1279
1280 __setup_root(tree_root->nodesize, tree_root->leafsize,
1281 tree_root->sectorsize, tree_root->stripesize,
1282 root, fs_info, objectid);
1283 root->root_key.objectid = objectid;
1284 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1285 root->root_key.offset = 0;
1286
1287 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1288 0, objectid, NULL, 0, 0, 0);
1289 if (IS_ERR(leaf)) {
1290 ret = PTR_ERR(leaf);
1dd05682 1291 leaf = NULL;
20897f5c
AJ		 1292 goto fail;
1293 }
1294
1295 bytenr = leaf->start;
1296 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1297 btrfs_set_header_bytenr(leaf, leaf->start);
1298 btrfs_set_header_generation(leaf, trans->transid);
1299 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1300 btrfs_set_header_owner(leaf, objectid);
1301 root->node = leaf;
1302
0a4e5586 1303 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
20897f5c
AJ		 1304 BTRFS_FSID_SIZE);
1305 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
b308bc2f 1306 btrfs_header_chunk_tree_uuid(leaf),
20897f5c
AJ		 1307 BTRFS_UUID_SIZE);
1308 btrfs_mark_buffer_dirty(leaf);
1309
1310 root->commit_root = btrfs_root_node(root);
1311 root->track_dirty = 1;
1312
1313
1314 root->root_item.flags = 0;
1315 root->root_item.byte_limit = 0;
1316 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1317 btrfs_set_root_generation(&root->root_item, trans->transid);
1318 btrfs_set_root_level(&root->root_item, 0);
1319 btrfs_set_root_refs(&root->root_item, 1);
1320 btrfs_set_root_used(&root->root_item, leaf->len);
1321 btrfs_set_root_last_snapshot(&root->root_item, 0);
1322 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB		 1323 uuid_le_gen(&uuid);
1324 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ		 1325 root->root_item.drop_level = 0;
1326
1327 key.objectid = objectid;
1328 key.type = BTRFS_ROOT_ITEM_KEY;
1329 key.offset = 0;
1330 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1331 if (ret)
1332 goto fail;
1333
1334 btrfs_tree_unlock(leaf);
1335
1dd05682
TI		 1336 return root;
1337
20897f5c 1338fail:
1dd05682
TI		 1339 if (leaf) {
1340 btrfs_tree_unlock(leaf);
1341 free_extent_buffer(leaf);
1342 }
1343 kfree(root);
20897f5c 1344
1dd05682 1345 return ERR_PTR(ret);
20897f5c
AJ		 1346}
1347
7237f183
YZ		 1348static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1349 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1350{
1351 struct btrfs_root *root;
1352 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1353 struct extent_buffer *leaf;
e02119d5 1354
6f07e42e 1355 root = btrfs_alloc_root(fs_info);
e02119d5 1356 if (!root)
7237f183 1357 return ERR_PTR(-ENOMEM);
e02119d5
CM		 1358
1359 __setup_root(tree_root->nodesize, tree_root->leafsize,
1360 tree_root->sectorsize, tree_root->stripesize,
1361 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1362
1363 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1364 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1365 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
7237f183
YZ		 1366 /*
1367 * log trees do not get reference counted because they go away
1368 * before a real commit is actually done. They do store pointers
1369 * to file data extents, and those reference counts still get
1370 * updated (along with back refs to the log tree).
1371 */
e02119d5
CM		 1372 root->ref_cows = 0;
1373
5d4f98a2 1374 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
66d7e7f0 1375 BTRFS_TREE_LOG_OBJECTID, NULL,
5581a51a 1376 0, 0, 0);
7237f183
YZ		 1377 if (IS_ERR(leaf)) {
1378 kfree(root);
1379 return ERR_CAST(leaf);
1380 }
e02119d5 1381
5d4f98a2
YZ		 1382 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1383 btrfs_set_header_bytenr(leaf, leaf->start);
1384 btrfs_set_header_generation(leaf, trans->transid);
1385 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1386 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1387 root->node = leaf;
e02119d5
CM		 1388
1389 write_extent_buffer(root->node, root->fs_info->fsid,
0a4e5586 1390 btrfs_header_fsid(), BTRFS_FSID_SIZE);
e02119d5
CM		 1391 btrfs_mark_buffer_dirty(root->node);
1392 btrfs_tree_unlock(root->node);
7237f183
YZ		 1393 return root;
1394}
1395
1396int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1397 struct btrfs_fs_info *fs_info)
1398{
1399 struct btrfs_root *log_root;
1400
1401 log_root = alloc_log_tree(trans, fs_info);
1402 if (IS_ERR(log_root))
1403 return PTR_ERR(log_root);
1404 WARN_ON(fs_info->log_root_tree);
1405 fs_info->log_root_tree = log_root;
1406 return 0;
1407}
1408
1409int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1410 struct btrfs_root *root)
1411{
1412 struct btrfs_root *log_root;
1413 struct btrfs_inode_item *inode_item;
1414
1415 log_root = alloc_log_tree(trans, root->fs_info);
1416 if (IS_ERR(log_root))
1417 return PTR_ERR(log_root);
1418
1419 log_root->last_trans = trans->transid;
1420 log_root->root_key.offset = root->root_key.objectid;
1421
1422 inode_item = &log_root->root_item.inode;
3cae210f
QW		 1423 btrfs_set_stack_inode_generation(inode_item, 1);
1424 btrfs_set_stack_inode_size(inode_item, 3);
1425 btrfs_set_stack_inode_nlink(inode_item, 1);
1426 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
1427 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
7237f183 1428
5d4f98a2 1429 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1430
1431 WARN_ON(root->log_root);
1432 root->log_root = log_root;
1433 root->log_transid = 0;
257c62e1 1434 root->last_log_commit = 0;
e02119d5
CM		 1435 return 0;
1436}
1437
35a3621b
SB		 1438static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1439 struct btrfs_key *key)
e02119d5
CM		 1440{
1441 struct btrfs_root *root;
1442 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1443 struct btrfs_path *path;
84234f3a 1444 u64 generation;
db94535d 1445 u32 blocksize;
cb517eab 1446 int ret;
0f7d52f4 1447
cb517eab
MX		 1448 path = btrfs_alloc_path();
1449 if (!path)
0f7d52f4 1450 return ERR_PTR(-ENOMEM);
cb517eab
MX		 1451
1452 root = btrfs_alloc_root(fs_info);
1453 if (!root) {
1454 ret = -ENOMEM;
1455 goto alloc_fail;
0f7d52f4
CM		 1456 }
1457
db94535d 1458 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb 1459 tree_root->sectorsize, tree_root->stripesize,
cb517eab 1460 root, fs_info, key->objectid);
0f7d52f4 1461
cb517eab
MX		 1462 ret = btrfs_find_root(tree_root, key, path,
1463 &root->root_item, &root->root_key);
0f7d52f4 1464 if (ret) {
13a8a7c8
YZ		 1465 if (ret > 0)
1466 ret = -ENOENT;
cb517eab 1467 goto find_fail;
0f7d52f4 1468 }
13a8a7c8 1469
84234f3a 1470 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1471 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1472 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1473 blocksize, generation);
cb517eab
MX		 1474 if (!root->node) {
1475 ret = -ENOMEM;
1476 goto find_fail;
1477 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1478 ret = -EIO;
1479 goto read_fail;
416bc658 1480 }
5d4f98a2 1481 root->commit_root = btrfs_root_node(root);
13a8a7c8 1482out:
cb517eab
MX		 1483 btrfs_free_path(path);
1484 return root;
1485
1486read_fail:
1487 free_extent_buffer(root->node);
1488find_fail:
1489 kfree(root);
1490alloc_fail:
1491 root = ERR_PTR(ret);
1492 goto out;
1493}
1494
1495struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1496 struct btrfs_key *location)
1497{
1498 struct btrfs_root *root;
1499
1500 root = btrfs_read_tree_root(tree_root, location);
1501 if (IS_ERR(root))
1502 return root;
1503
1504 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
e02119d5 1505 root->ref_cows = 1;
08fe4db1
LZ		 1506 btrfs_check_and_init_root_item(&root->root_item);
1507 }
13a8a7c8 1508
5eda7b5e
CM		 1509 return root;
1510}
1511
cb517eab
MX		 1512int btrfs_init_fs_root(struct btrfs_root *root)
1513{
1514 int ret;
1515
1516 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1517 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1518 GFP_NOFS);
1519 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1520 ret = -ENOMEM;
1521 goto fail;
1522 }
1523
1524 btrfs_init_free_ino_ctl(root);
1525 mutex_init(&root->fs_commit_mutex);
1526 spin_lock_init(&root->cache_lock);
1527 init_waitqueue_head(&root->cache_wait);
1528
1529 ret = get_anon_bdev(&root->anon_dev);
1530 if (ret)
1531 goto fail;
1532 return 0;
1533fail:
1534 kfree(root->free_ino_ctl);
1535 kfree(root->free_ino_pinned);
1536 return ret;
1537}
1538
171170c1
ST		 1539static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1540 u64 root_id)
cb517eab
MX		 1541{
1542 struct btrfs_root *root;
1543
1544 spin_lock(&fs_info->fs_roots_radix_lock);
1545 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1546 (unsigned long)root_id);
1547 spin_unlock(&fs_info->fs_roots_radix_lock);
1548 return root;
1549}
1550
1551int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1552 struct btrfs_root *root)
1553{
1554 int ret;
1555
1556 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1557 if (ret)
1558 return ret;
1559
1560 spin_lock(&fs_info->fs_roots_radix_lock);
1561 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1562 (unsigned long)root->root_key.objectid,
1563 root);
1564 if (ret == 0)
1565 root->in_radix = 1;
1566 spin_unlock(&fs_info->fs_roots_radix_lock);
1567 radix_tree_preload_end();
1568
1569 return ret;
1570}
1571
c00869f1
MX		 1572struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1573 struct btrfs_key *location,
1574 bool check_ref)
5eda7b5e
CM		 1575{
1576 struct btrfs_root *root;
1577 int ret;
1578
edbd8d4e
CM		 1579 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1580 return fs_info->tree_root;
1581 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1582 return fs_info->extent_root;
8f18cf13
CM		 1583 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1584 return fs_info->chunk_root;
1585 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1586 return fs_info->dev_root;
0403e47e
YZ		 1587 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1588 return fs_info->csum_root;
bcef60f2
AJ		 1589 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1590 return fs_info->quota_root ? fs_info->quota_root :
1591 ERR_PTR(-ENOENT);
f7a81ea4
SB		 1592 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1593 return fs_info->uuid_root ? fs_info->uuid_root :
1594 ERR_PTR(-ENOENT);
4df27c4d 1595again:
cb517eab 1596 root = btrfs_lookup_fs_root(fs_info, location->objectid);
48475471 1597 if (root) {
c00869f1 1598 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
48475471 1599 return ERR_PTR(-ENOENT);
5eda7b5e 1600 return root;
48475471 1601 }
5eda7b5e 1602
cb517eab 1603 root = btrfs_read_fs_root(fs_info->tree_root, location);
5eda7b5e
CM		 1604 if (IS_ERR(root))
1605 return root;
3394e160 1606
c00869f1 1607 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
cb517eab 1608 ret = -ENOENT;
581bb050 1609 goto fail;
35a30d7c 1610 }
581bb050 1611
cb517eab 1612 ret = btrfs_init_fs_root(root);
ac08aedf
CM		 1613 if (ret)
1614 goto fail;
3394e160 1615
d68fc57b
YZ		 1616 ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
1617 if (ret < 0)
1618 goto fail;
1619 if (ret == 0)
1620 root->orphan_item_inserted = 1;
1621
cb517eab 1622 ret = btrfs_insert_fs_root(fs_info, root);
0f7d52f4 1623 if (ret) {
4df27c4d
YZ		 1624 if (ret == -EEXIST) {
1625 free_fs_root(root);
1626 goto again;
1627 }
1628 goto fail;
0f7d52f4 1629 }
edbd8d4e 1630 return root;
4df27c4d
YZ		 1631fail:
1632 free_fs_root(root);
1633 return ERR_PTR(ret);
edbd8d4e
CM		 1634}
1635
04160088
CM		 1636static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1637{
1638 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1639 int ret = 0;
04160088
CM		 1640 struct btrfs_device *device;
1641 struct backing_dev_info *bdi;
b7967db7 1642
1f78160c
XG		 1643 rcu_read_lock();
1644 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1645 if (!device->bdev)
1646 continue;
04160088
CM		 1647 bdi = blk_get_backing_dev_info(device->bdev);
1648 if (bdi && bdi_congested(bdi, bdi_bits)) {
1649 ret = 1;
1650 break;
1651 }
1652 }
1f78160c 1653 rcu_read_unlock();
04160088
CM		 1654 return ret;
1655}
1656
ad081f14
JA		 1657/*
1658 * If this fails, caller must call bdi_destroy() to get rid of the
1659 * bdi again.
1660 */
04160088
CM		 1661static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1662{
ad081f14
JA		 1663 int err;
1664
1665 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1666 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1667 if (err)
1668 return err;
1669
4575c9cc 1670 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1671 bdi->congested_fn = btrfs_congested_fn;
1672 bdi->congested_data = info;
1673 return 0;
1674}
1675
8b712842
CM		 1676/*
1677 * called by the kthread helper functions to finally call the bio end_io
1678 * functions. This is where read checksum verification actually happens
1679 */
1680static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1681{
ce9adaa5 1682 struct bio *bio;
8b712842
CM		 1683 struct end_io_wq *end_io_wq;
1684 struct btrfs_fs_info *fs_info;
ce9adaa5 1685 int error;
ce9adaa5 1686
8b712842
CM		 1687 end_io_wq = container_of(work, struct end_io_wq, work);
1688 bio = end_io_wq->bio;
1689 fs_info = end_io_wq->info;
ce9adaa5 1690
8b712842
CM		 1691 error = end_io_wq->error;
1692 bio->bi_private = end_io_wq->private;
1693 bio->bi_end_io = end_io_wq->end_io;
1694 kfree(end_io_wq);
8b712842 1695 bio_endio(bio, error);
44b8bd7e
CM		 1696}
1697
a74a4b97
CM		 1698static int cleaner_kthread(void *arg)
1699{
1700 struct btrfs_root *root = arg;
d0278245 1701 int again;
a74a4b97
CM		 1702
1703 do {
d0278245 1704 again = 0;
a74a4b97 1705
d0278245 1706 /* Make the cleaner go to sleep early. */
babbf170 1707 if (btrfs_need_cleaner_sleep(root))
d0278245
MX		 1708 goto sleep;
1709
1710 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1711 goto sleep;
1712
dc7f370c
MX		 1713 /*
1714 * Avoid the problem that we change the status of the fs
1715 * during the above check and trylock.
1716 */
babbf170 1717 if (btrfs_need_cleaner_sleep(root)) {
dc7f370c
MX		 1718 mutex_unlock(&root->fs_info->cleaner_mutex);
1719 goto sleep;
76dda93c 1720 }
a74a4b97 1721
d0278245
MX		 1722 btrfs_run_delayed_iputs(root);
1723 again = btrfs_clean_one_deleted_snapshot(root);
1724 mutex_unlock(&root->fs_info->cleaner_mutex);
1725
1726 /*
05323cd1
MX		 1727 * The defragger has dealt with the R/O remount and umount,
1728 * needn't do anything special here.
d0278245
MX		 1729 */
1730 btrfs_run_defrag_inodes(root->fs_info);
1731sleep:
9d1a2a3a 1732 if (!try_to_freeze() && !again) {
a74a4b97 1733 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1734 if (!kthread_should_stop())
1735 schedule();
a74a4b97
CM		 1736 __set_current_state(TASK_RUNNING);
1737 }
1738 } while (!kthread_should_stop());
1739 return 0;
1740}
1741
1742static int transaction_kthread(void *arg)
1743{
1744 struct btrfs_root *root = arg;
1745 struct btrfs_trans_handle *trans;
1746 struct btrfs_transaction *cur;
8929ecfa 1747 u64 transid;
a74a4b97
CM		 1748 unsigned long now;
1749 unsigned long delay;
914b2007 1750 bool cannot_commit;
a74a4b97
CM		 1751
1752 do {
914b2007 1753 cannot_commit = false;
8b87dc17 1754 delay = HZ * root->fs_info->commit_interval;
a74a4b97
CM		 1755 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1756
a4abeea4 1757 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1758 cur = root->fs_info->running_transaction;
1759 if (!cur) {
a4abeea4 1760 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1761 goto sleep;
1762 }
31153d81 1763
a74a4b97 1764 now = get_seconds();
4a9d8bde 1765 if (cur->state < TRANS_STATE_BLOCKED &&
8b87dc17
DS		 1766 (now < cur->start_time ||
1767 now - cur->start_time < root->fs_info->commit_interval)) {
a4abeea4 1768 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1769 delay = HZ * 5;
1770 goto sleep;
1771 }
8929ecfa 1772 transid = cur->transid;
a4abeea4 1773 spin_unlock(&root->fs_info->trans_lock);
56bec294 1774
79787eaa 1775 /* If the file system is aborted, this will always fail. */
354aa0fb 1776 trans = btrfs_attach_transaction(root);
914b2007 1777 if (IS_ERR(trans)) {
354aa0fb
MX		 1778 if (PTR_ERR(trans) != -ENOENT)
1779 cannot_commit = true;
79787eaa 1780 goto sleep;
914b2007 1781 }
8929ecfa 1782 if (transid == trans->transid) {
79787eaa 1783 btrfs_commit_transaction(trans, root);
8929ecfa
YZ		 1784 } else {
1785 btrfs_end_transaction(trans, root);
1786 }
a74a4b97
CM		 1787sleep:
1788 wake_up_process(root->fs_info->cleaner_kthread);
1789 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1790
4e121c06
JB		 1791 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1792 &root->fs_info->fs_state)))
1793 btrfs_cleanup_transaction(root);
a0acae0e 1794 if (!try_to_freeze()) {
a74a4b97 1795 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa 1796 if (!kthread_should_stop() &&
914b2007
JK		 1797 (!btrfs_transaction_blocked(root->fs_info) ||
1798 cannot_commit))
8929ecfa 1799 schedule_timeout(delay);
a74a4b97
CM		 1800 __set_current_state(TASK_RUNNING);
1801 }
1802 } while (!kthread_should_stop());
1803 return 0;
1804}
1805
af31f5e5
CM		 1806/*
1807 * this will find the highest generation in the array of
1808 * root backups. The index of the highest array is returned,
1809 * or -1 if we can't find anything.
1810 *
1811 * We check to make sure the array is valid by comparing the
1812 * generation of the latest root in the array with the generation
1813 * in the super block. If they don't match we pitch it.
1814 */
1815static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1816{
1817 u64 cur;
1818 int newest_index = -1;
1819 struct btrfs_root_backup *root_backup;
1820 int i;
1821
1822 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1823 root_backup = info->super_copy->super_roots + i;
1824 cur = btrfs_backup_tree_root_gen(root_backup);
1825 if (cur == newest_gen)
1826 newest_index = i;
1827 }
1828
1829 /* check to see if we actually wrapped around */
1830 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1831 root_backup = info->super_copy->super_roots;
1832 cur = btrfs_backup_tree_root_gen(root_backup);
1833 if (cur == newest_gen)
1834 newest_index = 0;
1835 }
1836 return newest_index;
1837}
1838
1839
1840/*
1841 * find the oldest backup so we know where to store new entries
1842 * in the backup array. This will set the backup_root_index
1843 * field in the fs_info struct
1844 */
1845static void find_oldest_super_backup(struct btrfs_fs_info *info,
1846 u64 newest_gen)
1847{
1848 int newest_index = -1;
1849
1850 newest_index = find_newest_super_backup(info, newest_gen);
1851 /* if there was garbage in there, just move along */
1852 if (newest_index == -1) {
1853 info->backup_root_index = 0;
1854 } else {
1855 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1856 }
1857}
1858
1859/*
1860 * copy all the root pointers into the super backup array.
1861 * this will bump the backup pointer by one when it is
1862 * done
1863 */
1864static void backup_super_roots(struct btrfs_fs_info *info)
1865{
1866 int next_backup;
1867 struct btrfs_root_backup *root_backup;
1868 int last_backup;
1869
1870 next_backup = info->backup_root_index;
1871 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1872 BTRFS_NUM_BACKUP_ROOTS;
1873
1874 /*
1875 * just overwrite the last backup if we're at the same generation
1876 * this happens only at umount
1877 */
1878 root_backup = info->super_for_commit->super_roots + last_backup;
1879 if (btrfs_backup_tree_root_gen(root_backup) ==
1880 btrfs_header_generation(info->tree_root->node))
1881 next_backup = last_backup;
1882
1883 root_backup = info->super_for_commit->super_roots + next_backup;
1884
1885 /*
1886 * make sure all of our padding and empty slots get zero filled
1887 * regardless of which ones we use today
1888 */
1889 memset(root_backup, 0, sizeof(*root_backup));
1890
1891 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1892
1893 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1894 btrfs_set_backup_tree_root_gen(root_backup,
1895 btrfs_header_generation(info->tree_root->node));
1896
1897 btrfs_set_backup_tree_root_level(root_backup,
1898 btrfs_header_level(info->tree_root->node));
1899
1900 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1901 btrfs_set_backup_chunk_root_gen(root_backup,
1902 btrfs_header_generation(info->chunk_root->node));
1903 btrfs_set_backup_chunk_root_level(root_backup,
1904 btrfs_header_level(info->chunk_root->node));
1905
1906 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1907 btrfs_set_backup_extent_root_gen(root_backup,
1908 btrfs_header_generation(info->extent_root->node));
1909 btrfs_set_backup_extent_root_level(root_backup,
1910 btrfs_header_level(info->extent_root->node));
1911
7c7e82a7
CM		 1912 /*
1913 * we might commit during log recovery, which happens before we set
1914 * the fs_root. Make sure it is valid before we fill it in.
1915 */
1916 if (info->fs_root && info->fs_root->node) {
1917 btrfs_set_backup_fs_root(root_backup,
1918 info->fs_root->node->start);
1919 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 1920 btrfs_header_generation(info->fs_root->node));
7c7e82a7 1921 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 1922 btrfs_header_level(info->fs_root->node));
7c7e82a7 1923 }
af31f5e5
CM		 1924
1925 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1926 btrfs_set_backup_dev_root_gen(root_backup,
1927 btrfs_header_generation(info->dev_root->node));
1928 btrfs_set_backup_dev_root_level(root_backup,
1929 btrfs_header_level(info->dev_root->node));
1930
1931 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1932 btrfs_set_backup_csum_root_gen(root_backup,
1933 btrfs_header_generation(info->csum_root->node));
1934 btrfs_set_backup_csum_root_level(root_backup,
1935 btrfs_header_level(info->csum_root->node));
1936
1937 btrfs_set_backup_total_bytes(root_backup,
1938 btrfs_super_total_bytes(info->super_copy));
1939 btrfs_set_backup_bytes_used(root_backup,
1940 btrfs_super_bytes_used(info->super_copy));
1941 btrfs_set_backup_num_devices(root_backup,
1942 btrfs_super_num_devices(info->super_copy));
1943
1944 /*
1945 * if we don't copy this out to the super_copy, it won't get remembered
1946 * for the next commit
1947 */
1948 memcpy(&info->super_copy->super_roots,
1949 &info->super_for_commit->super_roots,
1950 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1951}
1952
1953/*
1954 * this copies info out of the root backup array and back into
1955 * the in-memory super block. It is meant to help iterate through
1956 * the array, so you send it the number of backups you've already
1957 * tried and the last backup index you used.
1958 *
1959 * this returns -1 when it has tried all the backups
1960 */
1961static noinline int next_root_backup(struct btrfs_fs_info *info,
1962 struct btrfs_super_block *super,
1963 int *num_backups_tried, int *backup_index)
1964{
1965 struct btrfs_root_backup *root_backup;
1966 int newest = *backup_index;
1967
1968 if (*num_backups_tried == 0) {
1969 u64 gen = btrfs_super_generation(super);
1970
1971 newest = find_newest_super_backup(info, gen);
1972 if (newest == -1)
1973 return -1;
1974
1975 *backup_index = newest;
1976 *num_backups_tried = 1;
1977 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
1978 /* we've tried all the backups, all done */
1979 return -1;
1980 } else {
1981 /* jump to the next oldest backup */
1982 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
1983 BTRFS_NUM_BACKUP_ROOTS;
1984 *backup_index = newest;
1985 *num_backups_tried += 1;
1986 }
1987 root_backup = super->super_roots + newest;
1988
1989 btrfs_set_super_generation(super,
1990 btrfs_backup_tree_root_gen(root_backup));
1991 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
1992 btrfs_set_super_root_level(super,
1993 btrfs_backup_tree_root_level(root_backup));
1994 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
1995
1996 /*
1997 * fixme: the total bytes and num_devices need to match or we should
1998 * need a fsck
1999 */
2000 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2001 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2002 return 0;
2003}
2004
7abadb64
LB		 2005/* helper to cleanup workers */
2006static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2007{
2008 btrfs_stop_workers(&fs_info->generic_worker);
2009 btrfs_stop_workers(&fs_info->fixup_workers);
2010 btrfs_stop_workers(&fs_info->delalloc_workers);
2011 btrfs_stop_workers(&fs_info->workers);
2012 btrfs_stop_workers(&fs_info->endio_workers);
2013 btrfs_stop_workers(&fs_info->endio_meta_workers);
2014 btrfs_stop_workers(&fs_info->endio_raid56_workers);
2015 btrfs_stop_workers(&fs_info->rmw_workers);
2016 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
2017 btrfs_stop_workers(&fs_info->endio_write_workers);
2018 btrfs_stop_workers(&fs_info->endio_freespace_worker);
2019 btrfs_stop_workers(&fs_info->submit_workers);
2020 btrfs_stop_workers(&fs_info->delayed_workers);
2021 btrfs_stop_workers(&fs_info->caching_workers);
2022 btrfs_stop_workers(&fs_info->readahead_workers);
2023 btrfs_stop_workers(&fs_info->flush_workers);
2f232036 2024 btrfs_stop_workers(&fs_info->qgroup_rescan_workers);
7abadb64
LB		 2025}
2026
2e9f5954
R		 2027static void free_root_extent_buffers(struct btrfs_root *root)
2028{
2029 if (root) {
2030 free_extent_buffer(root->node);
2031 free_extent_buffer(root->commit_root);
2032 root->node = NULL;
2033 root->commit_root = NULL;
2034 }
2035}
2036
af31f5e5
CM		 2037/* helper to cleanup tree roots */
2038static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2039{
2e9f5954 2040 free_root_extent_buffers(info->tree_root);
655b09fe 2041
2e9f5954
R		 2042 free_root_extent_buffers(info->dev_root);
2043 free_root_extent_buffers(info->extent_root);
2044 free_root_extent_buffers(info->csum_root);
2045 free_root_extent_buffers(info->quota_root);
2046 free_root_extent_buffers(info->uuid_root);
2047 if (chunk_root)
2048 free_root_extent_buffers(info->chunk_root);
af31f5e5
CM		 2049}
2050
171f6537
JB		 2051static void del_fs_roots(struct btrfs_fs_info *fs_info)
2052{
2053 int ret;
2054 struct btrfs_root *gang[8];
2055 int i;
2056
2057 while (!list_empty(&fs_info->dead_roots)) {
2058 gang[0] = list_entry(fs_info->dead_roots.next,
2059 struct btrfs_root, root_list);
2060 list_del(&gang[0]->root_list);
2061
2062 if (gang[0]->in_radix) {
cb517eab 2063 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
171f6537
JB		 2064 } else {
2065 free_extent_buffer(gang[0]->node);
2066 free_extent_buffer(gang[0]->commit_root);
b0feb9d9 2067 btrfs_put_fs_root(gang[0]);
171f6537
JB		 2068 }
2069 }
2070
2071 while (1) {
2072 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2073 (void **)gang, 0,
2074 ARRAY_SIZE(gang));
2075 if (!ret)
2076 break;
2077 for (i = 0; i < ret; i++)
cb517eab 2078 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
171f6537
JB		 2079 }
2080}
af31f5e5 2081
ad2b2c80
AV		 2082int open_ctree(struct super_block *sb,
2083 struct btrfs_fs_devices *fs_devices,
2084 char *options)
2e635a27 2085{
db94535d
CM		 2086 u32 sectorsize;
2087 u32 nodesize;
2088 u32 leafsize;
2089 u32 blocksize;
87ee04eb 2090 u32 stripesize;
84234f3a 2091 u64 generation;
f2b636e8 2092 u64 features;
3de4586c 2093 struct btrfs_key location;
a061fc8d 2094 struct buffer_head *bh;
4d34b278 2095 struct btrfs_super_block *disk_super;
815745cf 2096 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2097 struct btrfs_root *tree_root;
4d34b278
ID		 2098 struct btrfs_root *extent_root;
2099 struct btrfs_root *csum_root;
2100 struct btrfs_root *chunk_root;
2101 struct btrfs_root *dev_root;
bcef60f2 2102 struct btrfs_root *quota_root;
f7a81ea4 2103 struct btrfs_root *uuid_root;
e02119d5 2104 struct btrfs_root *log_tree_root;
eb60ceac 2105 int ret;
e58ca020 2106 int err = -EINVAL;
af31f5e5
CM		 2107 int num_backups_tried = 0;
2108 int backup_index = 0;
70f80175
SB		 2109 bool create_uuid_tree;
2110 bool check_uuid_tree;
4543df7e 2111
f84a8bd6 2112 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
6f07e42e 2113 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
cb517eab 2114 if (!tree_root || !chunk_root) {
39279cc3
CM		 2115 err = -ENOMEM;
2116 goto fail;
2117 }
76dda93c
YZ		 2118
2119 ret = init_srcu_struct(&fs_info->subvol_srcu);
2120 if (ret) {
2121 err = ret;
2122 goto fail;
2123 }
2124
2125 ret = setup_bdi(fs_info, &fs_info->bdi);
2126 if (ret) {
2127 err = ret;
2128 goto fail_srcu;
2129 }
2130
e2d84521
MX		 2131 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2132 if (ret) {
2133 err = ret;
2134 goto fail_bdi;
2135 }
2136 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2137 (1 + ilog2(nr_cpu_ids));
2138
963d678b
MX		 2139 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2140 if (ret) {
2141 err = ret;
2142 goto fail_dirty_metadata_bytes;
2143 }
2144
76dda93c
YZ		 2145 fs_info->btree_inode = new_inode(sb);
2146 if (!fs_info->btree_inode) {
2147 err = -ENOMEM;
963d678b 2148 goto fail_delalloc_bytes;
76dda93c
YZ		 2149 }
2150
a6591715 2151 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2152
76dda93c 2153 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
8fd17795 2154 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2155 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2156 INIT_LIST_HEAD(&fs_info->delayed_iputs);
eb73c1b7 2157 INIT_LIST_HEAD(&fs_info->delalloc_roots);
11833d66 2158 INIT_LIST_HEAD(&fs_info->caching_block_groups);
eb73c1b7 2159 spin_lock_init(&fs_info->delalloc_root_lock);
a4abeea4 2160 spin_lock_init(&fs_info->trans_lock);
76dda93c 2161 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2162 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2163 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 2164 spin_lock_init(&fs_info->free_chunk_lock);
f29021b2 2165 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2166 spin_lock_init(&fs_info->super_lock);
f29021b2 2167 rwlock_init(&fs_info->tree_mod_log_lock);
7585717f 2168 mutex_init(&fs_info->reloc_mutex);
de98ced9 2169 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2170
58176a96 2171 init_completion(&fs_info->kobj_unregister);
0b86a832 2172 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2173 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2174 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
0b86a832 2175 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX		 2176 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2177 BTRFS_BLOCK_RSV_GLOBAL);
2178 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2179 BTRFS_BLOCK_RSV_DELALLOC);
2180 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2181 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2182 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2183 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2184 BTRFS_BLOCK_RSV_DELOPS);
cb03c743 2185 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 2186 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 2187 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 2188 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 2189 atomic_set(&fs_info->defrag_running, 0);
fc36ed7e 2190 atomic64_set(&fs_info->tree_mod_seq, 0);
e20d96d6 2191 fs_info->sb = sb;
6f568d35 2192 fs_info->max_inline = 8192 * 1024;
9ed74f2d 2193 fs_info->metadata_ratio = 0;
4cb5300b 2194 fs_info->defrag_inodes = RB_ROOT;
2bf64758 2195 fs_info->free_chunk_space = 0;
f29021b2 2196 fs_info->tree_mod_log = RB_ROOT;
8b87dc17 2197 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
c8b97818 2198
90519d66
AJ		 2199 /* readahead state */
2200 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2201 spin_lock_init(&fs_info->reada_lock);
c8b97818 2202
b34b086c
CM		 2203 fs_info->thread_pool_size = min_t(unsigned long,
2204 num_online_cpus() + 2, 8);
0afbaf8c 2205
199c2a9c
MX		 2206 INIT_LIST_HEAD(&fs_info->ordered_roots);
2207 spin_lock_init(&fs_info->ordered_root_lock);
16cdcec7
MX		 2208 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2209 GFP_NOFS);
2210 if (!fs_info->delayed_root) {
2211 err = -ENOMEM;
2212 goto fail_iput;
2213 }
2214 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2215
a2de733c
AJ		 2216 mutex_init(&fs_info->scrub_lock);
2217 atomic_set(&fs_info->scrubs_running, 0);
2218 atomic_set(&fs_info->scrub_pause_req, 0);
2219 atomic_set(&fs_info->scrubs_paused, 0);
2220 atomic_set(&fs_info->scrub_cancel_req, 0);
2221 init_waitqueue_head(&fs_info->scrub_pause_wait);
a2de733c 2222 fs_info->scrub_workers_refcnt = 0;
21adbd5c
SB		 2223#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2224 fs_info->check_integrity_print_mask = 0;
2225#endif
a2de733c 2226
c9e9f97b
ID		 2227 spin_lock_init(&fs_info->balance_lock);
2228 mutex_init(&fs_info->balance_mutex);
837d5b6e
ID		 2229 atomic_set(&fs_info->balance_running, 0);
2230 atomic_set(&fs_info->balance_pause_req, 0);
a7e99c69 2231 atomic_set(&fs_info->balance_cancel_req, 0);
c9e9f97b 2232 fs_info->balance_ctl = NULL;
837d5b6e 2233 init_waitqueue_head(&fs_info->balance_wait_q);
a2de733c 2234
a061fc8d
CM		 2235 sb->s_blocksize = 4096;
2236 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 2237 sb->s_bdi = &fs_info->bdi;
a061fc8d 2238
76dda93c 2239 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
bfe86848 2240 set_nlink(fs_info->btree_inode, 1);
0afbaf8c
CM		 2241 /*
2242 * we set the i_size on the btree inode to the max possible int.
2243 * the real end of the address space is determined by all of
2244 * the devices in the system
2245 */
2246 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 2247 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 2248 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2249
5d4f98a2 2250 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 2251 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 2252 fs_info->btree_inode->i_mapping);
0b32f4bb 2253 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
a8067e02 2254 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM		 2255
2256 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 2257
76dda93c
YZ		 2258 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2259 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2260 sizeof(struct btrfs_key));
72ac3c0d
JB		 2261 set_bit(BTRFS_INODE_DUMMY,
2262 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
778ba82b 2263 btrfs_insert_inode_hash(fs_info->btree_inode);
76dda93c 2264
0f9dd46c 2265 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2266 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2267 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2268
11833d66 2269 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 2270 fs_info->btree_inode->i_mapping);
11833d66 2271 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 2272 fs_info->btree_inode->i_mapping);
11833d66 2273 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 2274 fs_info->do_barriers = 1;
e18e4809 2275
39279cc3 2276
5a3f23d5 2277 mutex_init(&fs_info->ordered_operations_mutex);
9ffba8cd 2278 mutex_init(&fs_info->ordered_extent_flush_mutex);
e02119d5 2279 mutex_init(&fs_info->tree_log_mutex);
925baedd 2280 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 2281 mutex_init(&fs_info->transaction_kthread_mutex);
2282 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2283 mutex_init(&fs_info->volume_mutex);
276e680d 2284 init_rwsem(&fs_info->extent_commit_sem);
c71bf099 2285 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2286 init_rwsem(&fs_info->subvol_sem);
803b2f54 2287 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
e922e087
SB		 2288 fs_info->dev_replace.lock_owner = 0;
2289 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2290 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2291 mutex_init(&fs_info->dev_replace.lock_management_lock);
2292 mutex_init(&fs_info->dev_replace.lock);
fa9c0d79 2293
416ac51d 2294 spin_lock_init(&fs_info->qgroup_lock);
f2f6ed3d 2295 mutex_init(&fs_info->qgroup_ioctl_lock);
416ac51d
AJ		 2296 fs_info->qgroup_tree = RB_ROOT;
2297 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2298 fs_info->qgroup_seq = 1;
2299 fs_info->quota_enabled = 0;
2300 fs_info->pending_quota_state = 0;
1e8f9158 2301 fs_info->qgroup_ulist = NULL;
2f232036 2302 mutex_init(&fs_info->qgroup_rescan_lock);
416ac51d 2303
fa9c0d79
CM		 2304 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2305 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2306
e6dcd2dc 2307 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2308 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2309 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2310 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2311
53b381b3
DW		 2312 ret = btrfs_alloc_stripe_hash_table(fs_info);
2313 if (ret) {
83c8266a 2314 err = ret;
53b381b3
DW		 2315 goto fail_alloc;
2316 }
2317
0b86a832 2318 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 2319 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2320
3c4bb26b 2321 invalidate_bdev(fs_devices->latest_bdev);
1104a885
DS		 2322
2323 /*
2324 * Read super block and check the signature bytes only
2325 */
a512bbf8 2326 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY		 2327 if (!bh) {
2328 err = -EINVAL;
16cdcec7 2329 goto fail_alloc;
20b45077 2330 }
39279cc3 2331
1104a885
DS		 2332 /*
2333 * We want to check superblock checksum, the type is stored inside.
2334 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2335 */
2336 if (btrfs_check_super_csum(bh->b_data)) {
2337 printk(KERN_ERR "btrfs: superblock checksum mismatch\n");
2338 err = -EINVAL;
2339 goto fail_alloc;
2340 }
2341
2342 /*
2343 * super_copy is zeroed at allocation time and we never touch the
2344 * following bytes up to INFO_SIZE, the checksum is calculated from
2345 * the whole block of INFO_SIZE
2346 */
6c41761f
DS		 2347 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2348 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2349 sizeof(*fs_info->super_for_commit));
a061fc8d 2350 brelse(bh);
5f39d397 2351
6c41761f 2352 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2353
1104a885
DS		 2354 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2355 if (ret) {
2356 printk(KERN_ERR "btrfs: superblock contains fatal errors\n");
2357 err = -EINVAL;
2358 goto fail_alloc;
2359 }
2360
6c41761f 2361 disk_super = fs_info->super_copy;
0f7d52f4 2362 if (!btrfs_super_root(disk_super))
16cdcec7 2363 goto fail_alloc;
0f7d52f4 2364
acce952b 2365 /* check FS state, whether FS is broken. */
87533c47
MX		 2366 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2367 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2368
af31f5e5
CM		 2369 /*
2370 * run through our array of backup supers and setup
2371 * our ring pointer to the oldest one
2372 */
2373 generation = btrfs_super_generation(disk_super);
2374 find_oldest_super_backup(fs_info, generation);
2375
75e7cb7f
LB		 2376 /*
2377 * In the long term, we'll store the compression type in the super
2378 * block, and it'll be used for per file compression control.
2379 */
2380 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2381
2b82032c
YZ		 2382 ret = btrfs_parse_options(tree_root, options);
2383 if (ret) {
2384 err = ret;
16cdcec7 2385 goto fail_alloc;
2b82032c 2386 }
dfe25020 2387
f2b636e8
JB		 2388 features = btrfs_super_incompat_flags(disk_super) &
2389 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2390 if (features) {
2391 printk(KERN_ERR "BTRFS: couldn't mount because of "
2392 "unsupported optional features (%Lx).\n",
c1c9ff7c 2393 features);
f2b636e8 2394 err = -EINVAL;
16cdcec7 2395 goto fail_alloc;
f2b636e8
JB		 2396 }
2397
727011e0
CM		 2398 if (btrfs_super_leafsize(disk_super) !=
2399 btrfs_super_nodesize(disk_super)) {
2400 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2401 "blocksizes don't match. node %d leaf %d\n",
2402 btrfs_super_nodesize(disk_super),
2403 btrfs_super_leafsize(disk_super));
2404 err = -EINVAL;
2405 goto fail_alloc;
2406 }
2407 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2408 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2409 "blocksize (%d) was too large\n",
2410 btrfs_super_leafsize(disk_super));
2411 err = -EINVAL;
2412 goto fail_alloc;
2413 }
2414
5d4f98a2 2415 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2416 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
69e380d1 2417 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2418 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
727011e0 2419
3173a18f
JB		 2420 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
2421 printk(KERN_ERR "btrfs: has skinny extents\n");
2422
727011e0
CM		 2423 /*
2424 * flag our filesystem as having big metadata blocks if
2425 * they are bigger than the page size
2426 */
2427 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2428 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
2429 printk(KERN_INFO "btrfs flagging fs with big metadata feature\n");
2430 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2431 }
2432
bc3f116f
CM		 2433 nodesize = btrfs_super_nodesize(disk_super);
2434 leafsize = btrfs_super_leafsize(disk_super);
2435 sectorsize = btrfs_super_sectorsize(disk_super);
2436 stripesize = btrfs_super_stripesize(disk_super);
e2d84521 2437 fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids));
963d678b 2438 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f
CM		 2439
2440 /*
2441 * mixed block groups end up with duplicate but slightly offset
2442 * extent buffers for the same range. It leads to corruptions
2443 */
2444 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2445 (sectorsize != leafsize)) {
2446 printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes "
2447 "are not allowed for mixed block groups on %s\n",
2448 sb->s_id);
2449 goto fail_alloc;
2450 }
2451
ceda0864
MX		 2452 /*
2453 * Needn't use the lock because there is no other task which will
2454 * update the flag.
2455 */
a6fa6fae 2456 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2457
f2b636e8
JB		 2458 features = btrfs_super_compat_ro_flags(disk_super) &
2459 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2460 if (!(sb->s_flags & MS_RDONLY) && features) {
2461 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2462 "unsupported option features (%Lx).\n",
c1c9ff7c 2463 features);
f2b636e8 2464 err = -EINVAL;
16cdcec7 2465 goto fail_alloc;
f2b636e8 2466 }
61d92c32
CM		 2467
2468 btrfs_init_workers(&fs_info->generic_worker,
2469 "genwork", 1, NULL);
2470
5443be45 2471 btrfs_init_workers(&fs_info->workers, "worker",
61d92c32
CM		 2472 fs_info->thread_pool_size,
2473 &fs_info->generic_worker);
c8b97818 2474
771ed689 2475 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
45d5fd14 2476 fs_info->thread_pool_size, NULL);
771ed689 2477
8ccf6f19 2478 btrfs_init_workers(&fs_info->flush_workers, "flush_delalloc",
45d5fd14 2479 fs_info->thread_pool_size, NULL);
8ccf6f19 2480
5443be45 2481 btrfs_init_workers(&fs_info->submit_workers, "submit",
b720d209 2482 min_t(u64, fs_devices->num_devices,
45d5fd14 2483 fs_info->thread_pool_size), NULL);
61b49440 2484
bab39bf9 2485 btrfs_init_workers(&fs_info->caching_workers, "cache",
45d5fd14 2486 fs_info->thread_pool_size, NULL);
bab39bf9 2487
61b49440
CM		 2488 /* a higher idle thresh on the submit workers makes it much more
2489 * likely that bios will be send down in a sane order to the
2490 * devices
2491 */
2492 fs_info->submit_workers.idle_thresh = 64;
53863232 2493
771ed689 2494 fs_info->workers.idle_thresh = 16;
4a69a410 2495 fs_info->workers.ordered = 1;
61b49440 2496
771ed689
CM		 2497 fs_info->delalloc_workers.idle_thresh = 2;
2498 fs_info->delalloc_workers.ordered = 1;
2499
61d92c32
CM		 2500 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
2501 &fs_info->generic_worker);
5443be45 2502 btrfs_init_workers(&fs_info->endio_workers, "endio",
61d92c32
CM		 2503 fs_info->thread_pool_size,
2504 &fs_info->generic_worker);
d20f7043 2505 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
61d92c32
CM		 2506 fs_info->thread_pool_size,
2507 &fs_info->generic_worker);
cad321ad 2508 btrfs_init_workers(&fs_info->endio_meta_write_workers,
61d92c32
CM		 2509 "endio-meta-write", fs_info->thread_pool_size,
2510 &fs_info->generic_worker);
53b381b3
DW		 2511 btrfs_init_workers(&fs_info->endio_raid56_workers,
2512 "endio-raid56", fs_info->thread_pool_size,
2513 &fs_info->generic_worker);
2514 btrfs_init_workers(&fs_info->rmw_workers,
2515 "rmw", fs_info->thread_pool_size,
2516 &fs_info->generic_worker);
5443be45 2517 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
61d92c32
CM		 2518 fs_info->thread_pool_size,
2519 &fs_info->generic_worker);
0cb59c99
JB		 2520 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
2521 1, &fs_info->generic_worker);
16cdcec7
MX		 2522 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
2523 fs_info->thread_pool_size,
2524 &fs_info->generic_worker);
90519d66
AJ		 2525 btrfs_init_workers(&fs_info->readahead_workers, "readahead",
2526 fs_info->thread_pool_size,
2527 &fs_info->generic_worker);
2f232036
JS		 2528 btrfs_init_workers(&fs_info->qgroup_rescan_workers, "qgroup-rescan", 1,
2529 &fs_info->generic_worker);
61b49440
CM		 2530
2531 /*
2532 * endios are largely parallel and should have a very
2533 * low idle thresh
2534 */
2535 fs_info->endio_workers.idle_thresh = 4;
b51912c9 2536 fs_info->endio_meta_workers.idle_thresh = 4;
53b381b3
DW		 2537 fs_info->endio_raid56_workers.idle_thresh = 4;
2538 fs_info->rmw_workers.idle_thresh = 2;
b51912c9 2539
9042846b
CM		 2540 fs_info->endio_write_workers.idle_thresh = 2;
2541 fs_info->endio_meta_write_workers.idle_thresh = 2;
90519d66 2542 fs_info->readahead_workers.idle_thresh = 2;
9042846b 2543
0dc3b84a
JB		 2544 /*
2545 * btrfs_start_workers can really only fail because of ENOMEM so just
2546 * return -ENOMEM if any of these fail.
2547 */
2548 ret = btrfs_start_workers(&fs_info->workers);
2549 ret |= btrfs_start_workers(&fs_info->generic_worker);
2550 ret |= btrfs_start_workers(&fs_info->submit_workers);
2551 ret |= btrfs_start_workers(&fs_info->delalloc_workers);
2552 ret |= btrfs_start_workers(&fs_info->fixup_workers);
2553 ret |= btrfs_start_workers(&fs_info->endio_workers);
2554 ret |= btrfs_start_workers(&fs_info->endio_meta_workers);
53b381b3
DW		 2555 ret |= btrfs_start_workers(&fs_info->rmw_workers);
2556 ret |= btrfs_start_workers(&fs_info->endio_raid56_workers);
0dc3b84a
JB		 2557 ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers);
2558 ret |= btrfs_start_workers(&fs_info->endio_write_workers);
2559 ret |= btrfs_start_workers(&fs_info->endio_freespace_worker);
2560 ret |= btrfs_start_workers(&fs_info->delayed_workers);
2561 ret |= btrfs_start_workers(&fs_info->caching_workers);
2562 ret |= btrfs_start_workers(&fs_info->readahead_workers);
8ccf6f19 2563 ret |= btrfs_start_workers(&fs_info->flush_workers);
2f232036 2564 ret |= btrfs_start_workers(&fs_info->qgroup_rescan_workers);
0dc3b84a 2565 if (ret) {
fed425c7 2566 err = -ENOMEM;
0dc3b84a
JB		 2567 goto fail_sb_buffer;
2568 }
4543df7e 2569
4575c9cc 2570 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 2571 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2572 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 2573
db94535d
CM		 2574 tree_root->nodesize = nodesize;
2575 tree_root->leafsize = leafsize;
2576 tree_root->sectorsize = sectorsize;
87ee04eb 2577 tree_root->stripesize = stripesize;
a061fc8d
CM		 2578
2579 sb->s_blocksize = sectorsize;
2580 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 2581
3cae210f 2582 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
d397712b 2583 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 2584 goto fail_sb_buffer;
2585 }
19c00ddc 2586
8d082fb7
LB		 2587 if (sectorsize != PAGE_SIZE) {
2588 printk(KERN_WARNING "btrfs: Incompatible sector size(%lu) "
2589 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
941b2ddf
KM		 2590 goto fail_sb_buffer;
2591 }
2592
925baedd 2593 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 2594 ret = btrfs_read_sys_array(tree_root);
925baedd 2595 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2596 if (ret) {
d397712b
CM		 2597 printk(KERN_WARNING "btrfs: failed to read the system "
2598 "array on %s\n", sb->s_id);
5d4f98a2 2599 goto fail_sb_buffer;
84eed90f 2600 }
0b86a832
CM		 2601
2602 blocksize = btrfs_level_size(tree_root,
2603 btrfs_super_chunk_root_level(disk_super));
84234f3a 2604 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 2605
2606 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2607 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2608
2609 chunk_root->node = read_tree_block(chunk_root,
2610 btrfs_super_chunk_root(disk_super),
84234f3a 2611 blocksize, generation);
416bc658
JB		 2612 if (!chunk_root->node ||
2613 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
83121942
DW		 2614 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
2615 sb->s_id);
af31f5e5 2616 goto fail_tree_roots;
83121942 2617 }
5d4f98a2
YZ		 2618 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2619 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2620
e17cade2 2621 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
b308bc2f 2622 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
e17cade2 2623
0b86a832 2624 ret = btrfs_read_chunk_tree(chunk_root);
2b82032c 2625 if (ret) {
d397712b
CM		 2626 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
2627 sb->s_id);
af31f5e5 2628 goto fail_tree_roots;
2b82032c 2629 }
0b86a832 2630
8dabb742
SB		 2631 /*
2632 * keep the device that is marked to be the target device for the
2633 * dev_replace procedure
2634 */
2635 btrfs_close_extra_devices(fs_info, fs_devices, 0);
dfe25020 2636
a6b0d5c8
CM		 2637 if (!fs_devices->latest_bdev) {
2638 printk(KERN_CRIT "btrfs: failed to read devices on %s\n",
2639 sb->s_id);
2640 goto fail_tree_roots;
2641 }
2642
af31f5e5 2643retry_root_backup:
db94535d
CM		 2644 blocksize = btrfs_level_size(tree_root,
2645 btrfs_super_root_level(disk_super));
84234f3a 2646 generation = btrfs_super_generation(disk_super);
0b86a832 2647
e20d96d6 2648 tree_root->node = read_tree_block(tree_root,
db94535d 2649 btrfs_super_root(disk_super),
84234f3a 2650 blocksize, generation);
af31f5e5
CM		 2651 if (!tree_root->node ||
2652 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
83121942
DW		 2653 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
2654 sb->s_id);
af31f5e5
CM		 2655
2656 goto recovery_tree_root;
83121942 2657 }
af31f5e5 2658
5d4f98a2
YZ		 2659 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2660 tree_root->commit_root = btrfs_root_node(tree_root);
69e9c6c6 2661 btrfs_set_root_refs(&tree_root->root_item, 1);
db94535d 2662
cb517eab
MX		 2663 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2664 location.type = BTRFS_ROOT_ITEM_KEY;
2665 location.offset = 0;
2666
2667 extent_root = btrfs_read_tree_root(tree_root, &location);
2668 if (IS_ERR(extent_root)) {
2669 ret = PTR_ERR(extent_root);
af31f5e5 2670 goto recovery_tree_root;
cb517eab 2671 }
0b86a832 2672 extent_root->track_dirty = 1;
cb517eab 2673 fs_info->extent_root = extent_root;
0b86a832 2674
cb517eab
MX		 2675 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2676 dev_root = btrfs_read_tree_root(tree_root, &location);
2677 if (IS_ERR(dev_root)) {
2678 ret = PTR_ERR(dev_root);
af31f5e5 2679 goto recovery_tree_root;
cb517eab 2680 }
5d4f98a2 2681 dev_root->track_dirty = 1;
cb517eab
MX		 2682 fs_info->dev_root = dev_root;
2683 btrfs_init_devices_late(fs_info);
3768f368 2684
cb517eab
MX		 2685 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2686 csum_root = btrfs_read_tree_root(tree_root, &location);
2687 if (IS_ERR(csum_root)) {
2688 ret = PTR_ERR(csum_root);
af31f5e5 2689 goto recovery_tree_root;
cb517eab 2690 }
d20f7043 2691 csum_root->track_dirty = 1;
cb517eab 2692 fs_info->csum_root = csum_root;
d20f7043 2693
cb517eab
MX		 2694 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2695 quota_root = btrfs_read_tree_root(tree_root, &location);
2696 if (!IS_ERR(quota_root)) {
bcef60f2
AJ		 2697 quota_root->track_dirty = 1;
2698 fs_info->quota_enabled = 1;
2699 fs_info->pending_quota_state = 1;
cb517eab 2700 fs_info->quota_root = quota_root;
bcef60f2
AJ		 2701 }
2702
f7a81ea4
SB		 2703 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2704 uuid_root = btrfs_read_tree_root(tree_root, &location);
2705 if (IS_ERR(uuid_root)) {
2706 ret = PTR_ERR(uuid_root);
2707 if (ret != -ENOENT)
2708 goto recovery_tree_root;
2709 create_uuid_tree = true;
70f80175 2710 check_uuid_tree = false;
f7a81ea4
SB		 2711 } else {
2712 uuid_root->track_dirty = 1;
2713 fs_info->uuid_root = uuid_root;
70f80175
SB		 2714 create_uuid_tree = false;
2715 check_uuid_tree =
2716 generation != btrfs_super_uuid_tree_generation(disk_super);
f7a81ea4
SB		 2717 }
2718
8929ecfa
YZ		 2719 fs_info->generation = generation;
2720 fs_info->last_trans_committed = generation;
8929ecfa 2721
68310a5e
ID		 2722 ret = btrfs_recover_balance(fs_info);
2723 if (ret) {
2724 printk(KERN_WARNING "btrfs: failed to recover balance\n");
2725 goto fail_block_groups;
2726 }
2727
733f4fbb
SB		 2728 ret = btrfs_init_dev_stats(fs_info);
2729 if (ret) {
2730 printk(KERN_ERR "btrfs: failed to init dev_stats: %d\n",
2731 ret);
2732 goto fail_block_groups;
2733 }
2734
8dabb742
SB		 2735 ret = btrfs_init_dev_replace(fs_info);
2736 if (ret) {
2737 pr_err("btrfs: failed to init dev_replace: %d\n", ret);
2738 goto fail_block_groups;
2739 }
2740
2741 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2742
c59021f8 2743 ret = btrfs_init_space_info(fs_info);
2744 if (ret) {
2745 printk(KERN_ERR "Failed to initial space info: %d\n", ret);
2746 goto fail_block_groups;
2747 }
2748
1b1d1f66
JB		 2749 ret = btrfs_read_block_groups(extent_root);
2750 if (ret) {
2751 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
2752 goto fail_block_groups;
2753 }
5af3e8cc
SB		 2754 fs_info->num_tolerated_disk_barrier_failures =
2755 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
292fd7fc
SB		 2756 if (fs_info->fs_devices->missing_devices >
2757 fs_info->num_tolerated_disk_barrier_failures &&
2758 !(sb->s_flags & MS_RDONLY)) {
2759 printk(KERN_WARNING
2760 "Btrfs: too many missing devices, writeable mount is not allowed\n");
2761 goto fail_block_groups;
2762 }
9078a3e1 2763
a74a4b97
CM		 2764 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2765 "btrfs-cleaner");
57506d50 2766 if (IS_ERR(fs_info->cleaner_kthread))
1b1d1f66 2767 goto fail_block_groups;
a74a4b97
CM		 2768
2769 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2770 tree_root,
2771 "btrfs-transaction");
57506d50 2772 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2773 goto fail_cleaner;
a74a4b97 2774
c289811c
CM		 2775 if (!btrfs_test_opt(tree_root, SSD) &&
2776 !btrfs_test_opt(tree_root, NOSSD) &&
2777 !fs_info->fs_devices->rotating) {
2778 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
2779 "mode\n");
2780 btrfs_set_opt(fs_info->mount_opt, SSD);
2781 }
2782
21adbd5c
SB		 2783#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2784 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2785 ret = btrfsic_mount(tree_root, fs_devices,
2786 btrfs_test_opt(tree_root,
2787 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2788 1 : 0,
2789 fs_info->check_integrity_print_mask);
2790 if (ret)
2791 printk(KERN_WARNING "btrfs: failed to initialize"
2792 " integrity check module %s\n", sb->s_id);
2793 }
2794#endif
bcef60f2
AJ		 2795 ret = btrfs_read_qgroup_config(fs_info);
2796 if (ret)
2797 goto fail_trans_kthread;
21adbd5c 2798
acce952b 2799 /* do not make disk changes in broken FS */
68ce9682 2800 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM		 2801 u64 bytenr = btrfs_super_log_root(disk_super);
2802
7c2ca468 2803 if (fs_devices->rw_devices == 0) {
d397712b
CM		 2804 printk(KERN_WARNING "Btrfs log replay required "
2805 "on RO media\n");
7c2ca468 2806 err = -EIO;
bcef60f2 2807 goto fail_qgroup;
7c2ca468 2808 }
e02119d5
CM		 2809 blocksize =
2810 btrfs_level_size(tree_root,
2811 btrfs_super_log_root_level(disk_super));
d18a2c44 2812
6f07e42e 2813 log_tree_root = btrfs_alloc_root(fs_info);
676e4c86
DC		 2814 if (!log_tree_root) {
2815 err = -ENOMEM;
bcef60f2 2816 goto fail_qgroup;
676e4c86 2817 }
e02119d5
CM		 2818
2819 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2820 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2821
2822 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 2823 blocksize,
2824 generation + 1);
416bc658
JB		 2825 if (!log_tree_root->node ||
2826 !extent_buffer_uptodate(log_tree_root->node)) {
2827 printk(KERN_ERR "btrfs: failed to read log tree\n");
2828 free_extent_buffer(log_tree_root->node);
2829 kfree(log_tree_root);
2830 goto fail_trans_kthread;
2831 }
79787eaa 2832 /* returns with log_tree_root freed on success */
e02119d5 2833 ret = btrfs_recover_log_trees(log_tree_root);
79787eaa
JM		 2834 if (ret) {
2835 btrfs_error(tree_root->fs_info, ret,
2836 "Failed to recover log tree");
2837 free_extent_buffer(log_tree_root->node);
2838 kfree(log_tree_root);
2839 goto fail_trans_kthread;
2840 }
e556ce2c
YZ		 2841
2842 if (sb->s_flags & MS_RDONLY) {
79787eaa
JM		 2843 ret = btrfs_commit_super(tree_root);
2844 if (ret)
2845 goto fail_trans_kthread;
e556ce2c 2846 }
e02119d5 2847 }
1a40e23b 2848
76dda93c 2849 ret = btrfs_find_orphan_roots(tree_root);
79787eaa
JM		 2850 if (ret)
2851 goto fail_trans_kthread;
76dda93c 2852
7c2ca468 2853 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b 2854 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2
ID		 2855 if (ret)
2856 goto fail_trans_kthread;
d68fc57b 2857
5d4f98a2 2858 ret = btrfs_recover_relocation(tree_root);
d7ce5843
MX		 2859 if (ret < 0) {
2860 printk(KERN_WARNING
2861 "btrfs: failed to recover relocation\n");
2862 err = -EINVAL;
bcef60f2 2863 goto fail_qgroup;
d7ce5843 2864 }
7c2ca468 2865 }
1a40e23b 2866
3de4586c
CM		 2867 location.objectid = BTRFS_FS_TREE_OBJECTID;
2868 location.type = BTRFS_ROOT_ITEM_KEY;
cb517eab 2869 location.offset = 0;
3de4586c 2870
3de4586c 2871 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
3140c9a3
DC		 2872 if (IS_ERR(fs_info->fs_root)) {
2873 err = PTR_ERR(fs_info->fs_root);
bcef60f2 2874 goto fail_qgroup;
3140c9a3 2875 }
c289811c 2876
2b6ba629
ID		 2877 if (sb->s_flags & MS_RDONLY)
2878 return 0;
59641015 2879
2b6ba629
ID		 2880 down_read(&fs_info->cleanup_work_sem);
2881 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2882 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 2883 up_read(&fs_info->cleanup_work_sem);
2b6ba629
ID		 2884 close_ctree(tree_root);
2885 return ret;
2886 }
2887 up_read(&fs_info->cleanup_work_sem);
59641015 2888
2b6ba629
ID		 2889 ret = btrfs_resume_balance_async(fs_info);
2890 if (ret) {
2891 printk(KERN_WARNING "btrfs: failed to resume balance\n");
2892 close_ctree(tree_root);
2893 return ret;
e3acc2a6
JB		 2894 }
2895
8dabb742
SB		 2896 ret = btrfs_resume_dev_replace_async(fs_info);
2897 if (ret) {
2898 pr_warn("btrfs: failed to resume dev_replace\n");
2899 close_ctree(tree_root);
2900 return ret;
2901 }
2902
b382a324
JS		 2903 btrfs_qgroup_rescan_resume(fs_info);
2904
f7a81ea4
SB		 2905 if (create_uuid_tree) {
2906 pr_info("btrfs: creating UUID tree\n");
2907 ret = btrfs_create_uuid_tree(fs_info);
2908 if (ret) {
2909 pr_warn("btrfs: failed to create the UUID tree %d\n",
2910 ret);
2911 close_ctree(tree_root);
2912 return ret;
2913 }
f420ee1e
SB		 2914 } else if (check_uuid_tree ||
2915 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
70f80175
SB		 2916 pr_info("btrfs: checking UUID tree\n");
2917 ret = btrfs_check_uuid_tree(fs_info);
2918 if (ret) {
2919 pr_warn("btrfs: failed to check the UUID tree %d\n",
2920 ret);
2921 close_ctree(tree_root);
2922 return ret;
2923 }
2924 } else {
2925 fs_info->update_uuid_tree_gen = 1;
f7a81ea4
SB		 2926 }
2927
ad2b2c80 2928 return 0;
39279cc3 2929
bcef60f2
AJ		 2930fail_qgroup:
2931 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM		 2932fail_trans_kthread:
2933 kthread_stop(fs_info->transaction_kthread);
54067ae9 2934 btrfs_cleanup_transaction(fs_info->tree_root);
7b5ff90e 2935 del_fs_roots(fs_info);
3f157a2f 2936fail_cleaner:
a74a4b97 2937 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2938
2939 /*
2940 * make sure we're done with the btree inode before we stop our
2941 * kthreads
2942 */
2943 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 2944
1b1d1f66 2945fail_block_groups:
54067ae9 2946 btrfs_put_block_group_cache(fs_info);
1b1d1f66 2947 btrfs_free_block_groups(fs_info);
af31f5e5
CM		 2948
2949fail_tree_roots:
2950 free_root_pointers(fs_info, 1);
2b8195bb 2951 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 2952
39279cc3 2953fail_sb_buffer:
7abadb64 2954 btrfs_stop_all_workers(fs_info);
16cdcec7 2955fail_alloc:
4543df7e 2956fail_iput:
586e46e2
ID		 2957 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2958
4543df7e 2959 iput(fs_info->btree_inode);
963d678b
MX		 2960fail_delalloc_bytes:
2961 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX		 2962fail_dirty_metadata_bytes:
2963 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
ad081f14 2964fail_bdi:
7e662854 2965 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 2966fail_srcu:
2967 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 2968fail:
53b381b3 2969 btrfs_free_stripe_hash_table(fs_info);
586e46e2 2970 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 2971 return err;
af31f5e5
CM		 2972
2973recovery_tree_root:
af31f5e5
CM		 2974 if (!btrfs_test_opt(tree_root, RECOVERY))
2975 goto fail_tree_roots;
2976
2977 free_root_pointers(fs_info, 0);
2978
2979 /* don't use the log in recovery mode, it won't be valid */
2980 btrfs_set_super_log_root(disk_super, 0);
2981
2982 /* we can't trust the free space cache either */
2983 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
2984
2985 ret = next_root_backup(fs_info, fs_info->super_copy,
2986 &num_backups_tried, &backup_index);
2987 if (ret == -1)
2988 goto fail_block_groups;
2989 goto retry_root_backup;
eb60ceac
CM		 2990}
2991
f2984462
CM		 2992static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2993{
f2984462
CM		 2994 if (uptodate) {
2995 set_buffer_uptodate(bh);
2996 } else {
442a4f63
SB		 2997 struct btrfs_device *device = (struct btrfs_device *)
2998 bh->b_private;
2999
606686ee
JB		 3000 printk_ratelimited_in_rcu(KERN_WARNING "lost page write due to "
3001 "I/O error on %s\n",
3002 rcu_str_deref(device->name));
1259ab75
CM		 3003 /* note, we dont' set_buffer_write_io_error because we have
3004 * our own ways of dealing with the IO errors
3005 */
f2984462 3006 clear_buffer_uptodate(bh);
442a4f63 3007 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM		 3008 }
3009 unlock_buffer(bh);
3010 put_bh(bh);
3011}
3012
a512bbf8
YZ		 3013struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3014{
3015 struct buffer_head *bh;
3016 struct buffer_head *latest = NULL;
3017 struct btrfs_super_block *super;
3018 int i;
3019 u64 transid = 0;
3020 u64 bytenr;
3021
3022 /* we would like to check all the supers, but that would make
3023 * a btrfs mount succeed after a mkfs from a different FS.
3024 * So, we need to add a special mount option to scan for
3025 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3026 */
3027 for (i = 0; i < 1; i++) {
3028 bytenr = btrfs_sb_offset(i);
8068a47e
AJ		 3029 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3030 i_size_read(bdev->bd_inode))
a512bbf8 3031 break;
8068a47e
AJ		 3032 bh = __bread(bdev, bytenr / 4096,
3033 BTRFS_SUPER_INFO_SIZE);
a512bbf8
YZ		 3034 if (!bh)
3035 continue;
3036
3037 super = (struct btrfs_super_block *)bh->b_data;
3038 if (btrfs_super_bytenr(super) != bytenr ||
3cae210f 3039 btrfs_super_magic(super) != BTRFS_MAGIC) {
a512bbf8
YZ		 3040 brelse(bh);
3041 continue;
3042 }
3043
3044 if (!latest || btrfs_super_generation(super) > transid) {
3045 brelse(latest);
3046 latest = bh;
3047 transid = btrfs_super_generation(super);
3048 } else {
3049 brelse(bh);
3050 }
3051 }
3052 return latest;
3053}
3054
4eedeb75
HH		 3055/*
3056 * this should be called twice, once with wait == 0 and
3057 * once with wait == 1. When wait == 0 is done, all the buffer heads
3058 * we write are pinned.
3059 *
3060 * They are released when wait == 1 is done.
3061 * max_mirrors must be the same for both runs, and it indicates how
3062 * many supers on this one device should be written.
3063 *
3064 * max_mirrors == 0 means to write them all.
3065 */
a512bbf8
YZ		 3066static int write_dev_supers(struct btrfs_device *device,
3067 struct btrfs_super_block *sb,
3068 int do_barriers, int wait, int max_mirrors)
3069{
3070 struct buffer_head *bh;
3071 int i;
3072 int ret;
3073 int errors = 0;
3074 u32 crc;
3075 u64 bytenr;
a512bbf8
YZ		 3076
3077 if (max_mirrors == 0)
3078 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3079
a512bbf8
YZ		 3080 for (i = 0; i < max_mirrors; i++) {
3081 bytenr = btrfs_sb_offset(i);
3082 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
3083 break;
3084
3085 if (wait) {
3086 bh = __find_get_block(device->bdev, bytenr / 4096,
3087 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3088 if (!bh) {
3089 errors++;
3090 continue;
3091 }
a512bbf8 3092 wait_on_buffer(bh);
4eedeb75
HH		 3093 if (!buffer_uptodate(bh))
3094 errors++;
3095
3096 /* drop our reference */
3097 brelse(bh);
3098
3099 /* drop the reference from the wait == 0 run */
3100 brelse(bh);
3101 continue;
a512bbf8
YZ		 3102 } else {
3103 btrfs_set_super_bytenr(sb, bytenr);
3104
3105 crc = ~(u32)0;
b0496686 3106 crc = btrfs_csum_data((char *)sb +
a512bbf8
YZ		 3107 BTRFS_CSUM_SIZE, crc,
3108 BTRFS_SUPER_INFO_SIZE -
3109 BTRFS_CSUM_SIZE);
3110 btrfs_csum_final(crc, sb->csum);
3111
4eedeb75
HH		 3112 /*
3113 * one reference for us, and we leave it for the
3114 * caller
3115 */
a512bbf8
YZ		 3116 bh = __getblk(device->bdev, bytenr / 4096,
3117 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3118 if (!bh) {
3119 printk(KERN_ERR "btrfs: couldn't get super "
3120 "buffer head for bytenr %Lu\n", bytenr);
3121 errors++;
3122 continue;
3123 }
3124
a512bbf8
YZ		 3125 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3126
4eedeb75 3127 /* one reference for submit_bh */
a512bbf8 3128 get_bh(bh);
4eedeb75
HH		 3129
3130 set_buffer_uptodate(bh);
a512bbf8
YZ		 3131 lock_buffer(bh);
3132 bh->b_end_io = btrfs_end_buffer_write_sync;
442a4f63 3133 bh->b_private = device;
a512bbf8
YZ		 3134 }
3135
387125fc
CM		 3136 /*
3137 * we fua the first super. The others we allow
3138 * to go down lazy.
3139 */
21adbd5c 3140 ret = btrfsic_submit_bh(WRITE_FUA, bh);
4eedeb75 3141 if (ret)
a512bbf8 3142 errors++;
a512bbf8
YZ		 3143 }
3144 return errors < i ? 0 : -1;
3145}
3146
387125fc
CM		 3147/*
3148 * endio for the write_dev_flush, this will wake anyone waiting
3149 * for the barrier when it is done
3150 */
3151static void btrfs_end_empty_barrier(struct bio *bio, int err)
3152{
3153 if (err) {
3154 if (err == -EOPNOTSUPP)
3155 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3156 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3157 }
3158 if (bio->bi_private)
3159 complete(bio->bi_private);
3160 bio_put(bio);
3161}
3162
3163/*
3164 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3165 * sent down. With wait == 1, it waits for the previous flush.
3166 *
3167 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3168 * capable
3169 */
3170static int write_dev_flush(struct btrfs_device *device, int wait)
3171{
3172 struct bio *bio;
3173 int ret = 0;
3174
3175 if (device->nobarriers)
3176 return 0;
3177
3178 if (wait) {
3179 bio = device->flush_bio;
3180 if (!bio)
3181 return 0;
3182
3183 wait_for_completion(&device->flush_wait);
3184
3185 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
606686ee
JB		 3186 printk_in_rcu("btrfs: disabling barriers on dev %s\n",
3187 rcu_str_deref(device->name));
387125fc 3188 device->nobarriers = 1;
5af3e8cc 3189 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
387125fc 3190 ret = -EIO;
5af3e8cc
SB		 3191 btrfs_dev_stat_inc_and_print(device,
3192 BTRFS_DEV_STAT_FLUSH_ERRS);
387125fc
CM		 3193 }
3194
3195 /* drop the reference from the wait == 0 run */
3196 bio_put(bio);
3197 device->flush_bio = NULL;
3198
3199 return ret;
3200 }
3201
3202 /*
3203 * one reference for us, and we leave it for the
3204 * caller
3205 */
9c017abc 3206 device->flush_bio = NULL;
9be3395b 3207 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
387125fc
CM		 3208 if (!bio)
3209 return -ENOMEM;
3210
3211 bio->bi_end_io = btrfs_end_empty_barrier;
3212 bio->bi_bdev = device->bdev;
3213 init_completion(&device->flush_wait);
3214 bio->bi_private = &device->flush_wait;
3215 device->flush_bio = bio;
3216
3217 bio_get(bio);
21adbd5c 3218 btrfsic_submit_bio(WRITE_FLUSH, bio);
387125fc
CM		 3219
3220 return 0;
3221}
3222
3223/*
3224 * send an empty flush down to each device in parallel,
3225 * then wait for them
3226 */
3227static int barrier_all_devices(struct btrfs_fs_info *info)
3228{
3229 struct list_head *head;
3230 struct btrfs_device *dev;
5af3e8cc
SB		 3231 int errors_send = 0;
3232 int errors_wait = 0;
387125fc
CM		 3233 int ret;
3234
3235 /* send down all the barriers */
3236 head = &info->fs_devices->devices;
3237 list_for_each_entry_rcu(dev, head, dev_list) {
3238 if (!dev->bdev) {
5af3e8cc 3239 errors_send++;
387125fc
CM		 3240 continue;
3241 }
3242 if (!dev->in_fs_metadata || !dev->writeable)
3243 continue;
3244
3245 ret = write_dev_flush(dev, 0);
3246 if (ret)
5af3e8cc 3247 errors_send++;
387125fc
CM		 3248 }
3249
3250 /* wait for all the barriers */
3251 list_for_each_entry_rcu(dev, head, dev_list) {
3252 if (!dev->bdev) {
5af3e8cc 3253 errors_wait++;
387125fc
CM		 3254 continue;
3255 }
3256 if (!dev->in_fs_metadata || !dev->writeable)
3257 continue;
3258
3259 ret = write_dev_flush(dev, 1);
3260 if (ret)
5af3e8cc 3261 errors_wait++;
387125fc 3262 }
5af3e8cc
SB		 3263 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3264 errors_wait > info->num_tolerated_disk_barrier_failures)
387125fc
CM		 3265 return -EIO;
3266 return 0;
3267}
3268
5af3e8cc
SB		 3269int btrfs_calc_num_tolerated_disk_barrier_failures(
3270 struct btrfs_fs_info *fs_info)
3271{
3272 struct btrfs_ioctl_space_info space;
3273 struct btrfs_space_info *sinfo;
3274 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3275 BTRFS_BLOCK_GROUP_SYSTEM,
3276 BTRFS_BLOCK_GROUP_METADATA,
3277 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3278 int num_types = 4;
3279 int i;
3280 int c;
3281 int num_tolerated_disk_barrier_failures =
3282 (int)fs_info->fs_devices->num_devices;
3283
3284 for (i = 0; i < num_types; i++) {
3285 struct btrfs_space_info *tmp;
3286
3287 sinfo = NULL;
3288 rcu_read_lock();
3289 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3290 if (tmp->flags == types[i]) {
3291 sinfo = tmp;
3292 break;
3293 }
3294 }
3295 rcu_read_unlock();
3296
3297 if (!sinfo)
3298 continue;
3299
3300 down_read(&sinfo->groups_sem);
3301 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3302 if (!list_empty(&sinfo->block_groups[c])) {
3303 u64 flags;
3304
3305 btrfs_get_block_group_info(
3306 &sinfo->block_groups[c], &space);
3307 if (space.total_bytes == 0 ||
3308 space.used_bytes == 0)
3309 continue;
3310 flags = space.flags;
3311 /*
3312 * return
3313 * 0: if dup, single or RAID0 is configured for
3314 * any of metadata, system or data, else
3315 * 1: if RAID5 is configured, or if RAID1 or
3316 * RAID10 is configured and only two mirrors
3317 * are used, else
3318 * 2: if RAID6 is configured, else
3319 * num_mirrors - 1: if RAID1 or RAID10 is
3320 * configured and more than
3321 * 2 mirrors are used.
3322 */
3323 if (num_tolerated_disk_barrier_failures > 0 &&
3324 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3325 BTRFS_BLOCK_GROUP_RAID0)) ||
3326 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3327 == 0)))
3328 num_tolerated_disk_barrier_failures = 0;
53b381b3
DW		 3329 else if (num_tolerated_disk_barrier_failures > 1) {
3330 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3331 BTRFS_BLOCK_GROUP_RAID5 |
3332 BTRFS_BLOCK_GROUP_RAID10)) {
3333 num_tolerated_disk_barrier_failures = 1;
3334 } else if (flags &
15b0a89d 3335 BTRFS_BLOCK_GROUP_RAID6) {
53b381b3
DW		 3336 num_tolerated_disk_barrier_failures = 2;
3337 }
3338 }
5af3e8cc
SB		 3339 }
3340 }
3341 up_read(&sinfo->groups_sem);
3342 }
3343
3344 return num_tolerated_disk_barrier_failures;
3345}
3346
48a3b636 3347static int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 3348{
e5e9a520 3349 struct list_head *head;
f2984462 3350 struct btrfs_device *dev;
a061fc8d 3351 struct btrfs_super_block *sb;
f2984462 3352 struct btrfs_dev_item *dev_item;
f2984462
CM		 3353 int ret;
3354 int do_barriers;
a236aed1
CM		 3355 int max_errors;
3356 int total_errors = 0;
a061fc8d 3357 u64 flags;
f2984462
CM		 3358
3359 do_barriers = !btrfs_test_opt(root, NOBARRIER);
af31f5e5 3360 backup_super_roots(root->fs_info);
f2984462 3361
6c41761f 3362 sb = root->fs_info->super_for_commit;
a061fc8d 3363 dev_item = &sb->dev_item;
e5e9a520 3364
174ba509 3365 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 3366 head = &root->fs_info->fs_devices->devices;
d7306801 3367 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
387125fc 3368
5af3e8cc
SB		 3369 if (do_barriers) {
3370 ret = barrier_all_devices(root->fs_info);
3371 if (ret) {
3372 mutex_unlock(
3373 &root->fs_info->fs_devices->device_list_mutex);
3374 btrfs_error(root->fs_info, ret,
3375 "errors while submitting device barriers.");
3376 return ret;
3377 }
3378 }
387125fc 3379
1f78160c 3380 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3381 if (!dev->bdev) {
3382 total_errors++;
3383 continue;
3384 }
2b82032c 3385 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3386 continue;
3387
2b82032c 3388 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 3389 btrfs_set_stack_device_type(dev_item, dev->type);
3390 btrfs_set_stack_device_id(dev_item, dev->devid);
3391 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
3392 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
3393 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3394 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3395 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3396 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 3397 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 3398
a061fc8d
CM		 3399 flags = btrfs_super_flags(sb);
3400 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3401
a512bbf8 3402 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 3403 if (ret)
3404 total_errors++;
f2984462 3405 }
a236aed1 3406 if (total_errors > max_errors) {
d397712b
CM		 3407 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
3408 total_errors);
a724b436 3409 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
79787eaa 3410
9d565ba4
SB		 3411 /* FUA is masked off if unsupported and can't be the reason */
3412 btrfs_error(root->fs_info, -EIO,
3413 "%d errors while writing supers", total_errors);
3414 return -EIO;
a236aed1 3415 }
f2984462 3416
a512bbf8 3417 total_errors = 0;
1f78160c 3418 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3419 if (!dev->bdev)
3420 continue;
2b82032c 3421 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3422 continue;
3423
a512bbf8
YZ		 3424 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3425 if (ret)
3426 total_errors++;
f2984462 3427 }
174ba509 3428 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 3429 if (total_errors > max_errors) {
79787eaa
JM		 3430 btrfs_error(root->fs_info, -EIO,
3431 "%d errors while writing supers", total_errors);
3432 return -EIO;
a236aed1 3433 }
f2984462
CM		 3434 return 0;
3435}
3436
a512bbf8
YZ		 3437int write_ctree_super(struct btrfs_trans_handle *trans,
3438 struct btrfs_root *root, int max_mirrors)
eb60ceac 3439{
f570e757 3440 return write_all_supers(root, max_mirrors);
cfaa7295
CM		 3441}
3442
cb517eab
MX		 3443/* Drop a fs root from the radix tree and free it. */
3444void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3445 struct btrfs_root *root)
2619ba1f 3446{
4df27c4d 3447 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 3448 radix_tree_delete(&fs_info->fs_roots_radix,
3449 (unsigned long)root->root_key.objectid);
4df27c4d 3450 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 3451
3452 if (btrfs_root_refs(&root->root_item) == 0)
3453 synchronize_srcu(&fs_info->subvol_srcu);
3454
d7634482 3455 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
3321719e
LB		 3456 btrfs_free_log(NULL, root);
3457 btrfs_free_log_root_tree(NULL, fs_info);
3458 }
3459
581bb050
LZ		 3460 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3461 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3462 free_fs_root(root);
4df27c4d
YZ		 3463}
3464
3465static void free_fs_root(struct btrfs_root *root)
3466{
82d5902d 3467 iput(root->cache_inode);
4df27c4d 3468 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
1cb048f5
FDBM		 3469 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3470 root->orphan_block_rsv = NULL;
0ee5dc67
AV		 3471 if (root->anon_dev)
3472 free_anon_bdev(root->anon_dev);
4df27c4d
YZ		 3473 free_extent_buffer(root->node);
3474 free_extent_buffer(root->commit_root);
581bb050
LZ		 3475 kfree(root->free_ino_ctl);
3476 kfree(root->free_ino_pinned);
d397712b 3477 kfree(root->name);
b0feb9d9 3478 btrfs_put_fs_root(root);
2619ba1f
CM		 3479}
3480
cb517eab
MX		 3481void btrfs_free_fs_root(struct btrfs_root *root)
3482{
3483 free_fs_root(root);
2619ba1f
CM		 3484}
3485
c146afad 3486int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3487{
c146afad
YZ		 3488 u64 root_objectid = 0;
3489 struct btrfs_root *gang[8];
3490 int i;
3768f368 3491 int ret;
e089f05c 3492
c146afad
YZ		 3493 while (1) {
3494 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3495 (void **)gang, root_objectid,
3496 ARRAY_SIZE(gang));
3497 if (!ret)
3498 break;
5d4f98a2
YZ		 3499
3500 root_objectid = gang[ret - 1]->root_key.objectid + 1;
c146afad 3501 for (i = 0; i < ret; i++) {
66b4ffd1
JB		 3502 int err;
3503
c146afad 3504 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 3505 err = btrfs_orphan_cleanup(gang[i]);
3506 if (err)
3507 return err;
c146afad
YZ		 3508 }
3509 root_objectid++;
3510 }
3511 return 0;
3512}
a2135011 3513
c146afad
YZ		 3514int btrfs_commit_super(struct btrfs_root *root)
3515{
3516 struct btrfs_trans_handle *trans;
a74a4b97 3517
c146afad 3518 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 3519 btrfs_run_delayed_iputs(root);
c146afad 3520 mutex_unlock(&root->fs_info->cleaner_mutex);
9d1a2a3a 3521 wake_up_process(root->fs_info->cleaner_kthread);
c71bf099
YZ		 3522
3523 /* wait until ongoing cleanup work done */
3524 down_write(&root->fs_info->cleanup_work_sem);
3525 up_write(&root->fs_info->cleanup_work_sem);
3526
7a7eaa40 3527 trans = btrfs_join_transaction(root);
3612b495
TI		 3528 if (IS_ERR(trans))
3529 return PTR_ERR(trans);
d52c1bcc 3530 return btrfs_commit_transaction(trans, root);
c146afad
YZ		 3531}
3532
3533int close_ctree(struct btrfs_root *root)
3534{
3535 struct btrfs_fs_info *fs_info = root->fs_info;
3536 int ret;
3537
3538 fs_info->closing = 1;
3539 smp_mb();
3540
803b2f54
SB		 3541 /* wait for the uuid_scan task to finish */
3542 down(&fs_info->uuid_tree_rescan_sem);
3543 /* avoid complains from lockdep et al., set sem back to initial state */
3544 up(&fs_info->uuid_tree_rescan_sem);
3545
837d5b6e 3546 /* pause restriper - we want to resume on mount */
aa1b8cd4 3547 btrfs_pause_balance(fs_info);
837d5b6e 3548
8dabb742
SB		 3549 btrfs_dev_replace_suspend_for_unmount(fs_info);
3550
aa1b8cd4 3551 btrfs_scrub_cancel(fs_info);
4cb5300b
CM		 3552
3553 /* wait for any defraggers to finish */
3554 wait_event(fs_info->transaction_wait,
3555 (atomic_read(&fs_info->defrag_running) == 0));
3556
3557 /* clear out the rbtree of defraggable inodes */
26176e7c 3558 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3559
c146afad 3560 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 3561 ret = btrfs_commit_super(root);
3562 if (ret)
3563 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
3564 }
3565
87533c47 3566 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
68ce9682 3567 btrfs_error_commit_super(root);
0f7d52f4 3568
300e4f8a
JB		 3569 btrfs_put_block_group_cache(fs_info);
3570
e3029d9f
AV		 3571 kthread_stop(fs_info->transaction_kthread);
3572 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3573
f25784b3
YZ		 3574 fs_info->closing = 2;
3575 smp_mb();
3576
bcef60f2
AJ		 3577 btrfs_free_qgroup_config(root->fs_info);
3578
963d678b
MX		 3579 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
3580 printk(KERN_INFO "btrfs: at unmount delalloc count %lld\n",
3581 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3582 }
bcc63abb 3583
c146afad 3584 del_fs_roots(fs_info);
d10c5f31 3585
2b1360da
JB		 3586 btrfs_free_block_groups(fs_info);
3587
96192499
JB		 3588 btrfs_stop_all_workers(fs_info);
3589
13e6c37b 3590 free_root_pointers(fs_info, 1);
9ad6b7bc 3591
13e6c37b 3592 iput(fs_info->btree_inode);
d6bfde87 3593
21adbd5c
SB		 3594#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3595 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3596 btrfsic_unmount(root, fs_info->fs_devices);
3597#endif
3598
dfe25020 3599 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3600 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3601
e2d84521 3602 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 3603 percpu_counter_destroy(&fs_info->delalloc_bytes);
04160088 3604 bdi_destroy(&fs_info->bdi);
76dda93c 3605 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 3606
53b381b3
DW		 3607 btrfs_free_stripe_hash_table(fs_info);
3608
1cb048f5
FDBM		 3609 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3610 root->orphan_block_rsv = NULL;
3611
eb60ceac
CM		 3612 return 0;
3613}
3614
b9fab919
CM		 3615int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3616 int atomic)
5f39d397 3617{
1259ab75 3618 int ret;
727011e0 3619 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 3620
0b32f4bb 3621 ret = extent_buffer_uptodate(buf);
1259ab75
CM		 3622 if (!ret)
3623 return ret;
3624
3625 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM		 3626 parent_transid, atomic);
3627 if (ret == -EAGAIN)
3628 return ret;
1259ab75 3629 return !ret;
5f39d397
CM		 3630}
3631
3632int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 3633{
0b32f4bb 3634 return set_extent_buffer_uptodate(buf);
5f39d397 3635}
6702ed49 3636
5f39d397
CM		 3637void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3638{
06ea65a3 3639 struct btrfs_root *root;
5f39d397 3640 u64 transid = btrfs_header_generation(buf);
b9473439 3641 int was_dirty;
b4ce94de 3642
06ea65a3
JB		 3643#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3644 /*
3645 * This is a fast path so only do this check if we have sanity tests
3646 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3647 * outside of the sanity tests.
3648 */
3649 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3650 return;
3651#endif
3652 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
b9447ef8 3653 btrfs_assert_tree_locked(buf);
31b1a2bd
JL		 3654 if (transid != root->fs_info->generation)
3655 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
d397712b 3656 "found %llu running %llu\n",
c1c9ff7c 3657 buf->start, transid, root->fs_info->generation);
0b32f4bb 3658 was_dirty = set_extent_buffer_dirty(buf);
e2d84521
MX		 3659 if (!was_dirty)
3660 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3661 buf->len,
3662 root->fs_info->dirty_metadata_batch);
eb60ceac
CM		 3663}
3664
b53d3f5d
LB		 3665static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3666 int flush_delayed)
16cdcec7
MX		 3667{
3668 /*
3669 * looks as though older kernels can get into trouble with
3670 * this code, they end up stuck in balance_dirty_pages forever
3671 */
e2d84521 3672 int ret;
16cdcec7
MX		 3673
3674 if (current->flags & PF_MEMALLOC)
3675 return;
3676
b53d3f5d
LB		 3677 if (flush_delayed)
3678 btrfs_balance_delayed_items(root);
16cdcec7 3679
e2d84521
MX		 3680 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3681 BTRFS_DIRTY_METADATA_THRESH);
3682 if (ret > 0) {
d0e1d66b
NJ		 3683 balance_dirty_pages_ratelimited(
3684 root->fs_info->btree_inode->i_mapping);
16cdcec7
MX		 3685 }
3686 return;
3687}
3688
b53d3f5d 3689void btrfs_btree_balance_dirty(struct btrfs_root *root)
35b7e476 3690{
b53d3f5d
LB		 3691 __btrfs_btree_balance_dirty(root, 1);
3692}
585ad2c3 3693
b53d3f5d
LB		 3694void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3695{
3696 __btrfs_btree_balance_dirty(root, 0);
35b7e476 3697}
6b80053d 3698
ca7a79ad 3699int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 3700{
727011e0 3701 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
0b32f4bb 3702 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
6b80053d 3703}
0da5468f 3704
fcd1f065 3705static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 3706 int read_only)
3707{
1104a885
DS		 3708 /*
3709 * Placeholder for checks
3710 */
fcd1f065 3711 return 0;
acce952b 3712}
3713
48a3b636 3714static void btrfs_error_commit_super(struct btrfs_root *root)
acce952b 3715{
acce952b 3716 mutex_lock(&root->fs_info->cleaner_mutex);
3717 btrfs_run_delayed_iputs(root);
3718 mutex_unlock(&root->fs_info->cleaner_mutex);
3719
3720 down_write(&root->fs_info->cleanup_work_sem);
3721 up_write(&root->fs_info->cleanup_work_sem);
3722
3723 /* cleanup FS via transaction */
3724 btrfs_cleanup_transaction(root);
acce952b 3725}
3726
569e0f35
JB		 3727static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
3728 struct btrfs_root *root)
acce952b 3729{
3730 struct btrfs_inode *btrfs_inode;
3731 struct list_head splice;
3732
3733 INIT_LIST_HEAD(&splice);
3734
3735 mutex_lock(&root->fs_info->ordered_operations_mutex);
199c2a9c 3736 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3737
569e0f35 3738 list_splice_init(&t->ordered_operations, &splice);
acce952b 3739 while (!list_empty(&splice)) {
3740 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3741 ordered_operations);
3742
3743 list_del_init(&btrfs_inode->ordered_operations);
199c2a9c 3744 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3745
3746 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3747
199c2a9c 3748 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3749 }
3750
199c2a9c 3751 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3752 mutex_unlock(&root->fs_info->ordered_operations_mutex);
acce952b 3753}
3754
143bede5 3755static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 3756{
acce952b 3757 struct btrfs_ordered_extent *ordered;
acce952b 3758
199c2a9c 3759 spin_lock(&root->ordered_extent_lock);
779880ef
JB		 3760 /*
3761 * This will just short circuit the ordered completion stuff which will
3762 * make sure the ordered extent gets properly cleaned up.
3763 */
199c2a9c 3764 list_for_each_entry(ordered, &root->ordered_extents,
779880ef
JB		 3765 root_extent_list)
3766 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
199c2a9c
MX		 3767 spin_unlock(&root->ordered_extent_lock);
3768}
3769
3770static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3771{
3772 struct btrfs_root *root;
3773 struct list_head splice;
3774
3775 INIT_LIST_HEAD(&splice);
3776
3777 spin_lock(&fs_info->ordered_root_lock);
3778 list_splice_init(&fs_info->ordered_roots, &splice);
3779 while (!list_empty(&splice)) {
3780 root = list_first_entry(&splice, struct btrfs_root,
3781 ordered_root);
1de2cfde
JB		 3782 list_move_tail(&root->ordered_root,
3783 &fs_info->ordered_roots);
199c2a9c
MX		 3784
3785 btrfs_destroy_ordered_extents(root);
3786
3787 cond_resched_lock(&fs_info->ordered_root_lock);
3788 }
3789 spin_unlock(&fs_info->ordered_root_lock);
acce952b 3790}
3791
35a3621b
SB		 3792static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3793 struct btrfs_root *root)
acce952b 3794{
3795 struct rb_node *node;
3796 struct btrfs_delayed_ref_root *delayed_refs;
3797 struct btrfs_delayed_ref_node *ref;
3798 int ret = 0;
3799
3800 delayed_refs = &trans->delayed_refs;
3801
3802 spin_lock(&delayed_refs->lock);
3803 if (delayed_refs->num_entries == 0) {
cfece4db 3804 spin_unlock(&delayed_refs->lock);
acce952b 3805 printk(KERN_INFO "delayed_refs has NO entry\n");
3806 return ret;
3807 }
3808
b939d1ab 3809 while ((node = rb_first(&delayed_refs->root)) != NULL) {
eb12db69 3810 struct btrfs_delayed_ref_head *head = NULL;
e78417d1 3811 bool pin_bytes = false;
acce952b 3812
eb12db69 3813 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
acce952b 3814 atomic_set(&ref->refs, 1);
3815 if (btrfs_delayed_ref_is_head(ref)) {
acce952b 3816
3817 head = btrfs_delayed_node_to_head(ref);
b939d1ab
JB		 3818 if (!mutex_trylock(&head->mutex)) {
3819 atomic_inc(&ref->refs);
3820 spin_unlock(&delayed_refs->lock);
3821
3822 /* Need to wait for the delayed ref to run */
3823 mutex_lock(&head->mutex);
3824 mutex_unlock(&head->mutex);
3825 btrfs_put_delayed_ref(ref);
3826
e18fca73 3827 spin_lock(&delayed_refs->lock);
b939d1ab
JB		 3828 continue;
3829 }
3830
54067ae9 3831 if (head->must_insert_reserved)
e78417d1 3832 pin_bytes = true;
78a6184a 3833 btrfs_free_delayed_extent_op(head->extent_op);
acce952b 3834 delayed_refs->num_heads--;
3835 if (list_empty(&head->cluster))
3836 delayed_refs->num_heads_ready--;
3837 list_del_init(&head->cluster);
acce952b 3838 }
eb12db69 3839
b939d1ab
JB		 3840 ref->in_tree = 0;
3841 rb_erase(&ref->rb_node, &delayed_refs->root);
3842 delayed_refs->num_entries--;
acce952b 3843 spin_unlock(&delayed_refs->lock);
e78417d1
JB		 3844 if (head) {
3845 if (pin_bytes)
3846 btrfs_pin_extent(root, ref->bytenr,
3847 ref->num_bytes, 1);
3848 mutex_unlock(&head->mutex);
3849 }
acce952b 3850 btrfs_put_delayed_ref(ref);
3851
3852 cond_resched();
3853 spin_lock(&delayed_refs->lock);
3854 }
3855
3856 spin_unlock(&delayed_refs->lock);
3857
3858 return ret;
3859}
3860
143bede5 3861static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 3862{
3863 struct btrfs_inode *btrfs_inode;
3864 struct list_head splice;
3865
3866 INIT_LIST_HEAD(&splice);
3867
eb73c1b7
MX		 3868 spin_lock(&root->delalloc_lock);
3869 list_splice_init(&root->delalloc_inodes, &splice);
acce952b 3870
3871 while (!list_empty(&splice)) {
eb73c1b7
MX		 3872 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3873 delalloc_inodes);
acce952b 3874
3875 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX		 3876 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3877 &btrfs_inode->runtime_flags);
eb73c1b7 3878 spin_unlock(&root->delalloc_lock);
acce952b 3879
3880 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3881
eb73c1b7 3882 spin_lock(&root->delalloc_lock);
acce952b 3883 }
3884
eb73c1b7
MX		 3885 spin_unlock(&root->delalloc_lock);
3886}
3887
3888static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3889{
3890 struct btrfs_root *root;
3891 struct list_head splice;
3892
3893 INIT_LIST_HEAD(&splice);
3894
3895 spin_lock(&fs_info->delalloc_root_lock);
3896 list_splice_init(&fs_info->delalloc_roots, &splice);
3897 while (!list_empty(&splice)) {
3898 root = list_first_entry(&splice, struct btrfs_root,
3899 delalloc_root);
3900 list_del_init(&root->delalloc_root);
3901 root = btrfs_grab_fs_root(root);
3902 BUG_ON(!root);
3903 spin_unlock(&fs_info->delalloc_root_lock);
3904
3905 btrfs_destroy_delalloc_inodes(root);
3906 btrfs_put_fs_root(root);
3907
3908 spin_lock(&fs_info->delalloc_root_lock);
3909 }
3910 spin_unlock(&fs_info->delalloc_root_lock);
acce952b 3911}
3912
3913static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3914 struct extent_io_tree *dirty_pages,
3915 int mark)
3916{
3917 int ret;
acce952b 3918 struct extent_buffer *eb;
3919 u64 start = 0;
3920 u64 end;
acce952b 3921
3922 while (1) {
3923 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 3924 mark, NULL);
acce952b 3925 if (ret)
3926 break;
3927
3928 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
3929 while (start <= end) {
fd8b2b61
JB		 3930 eb = btrfs_find_tree_block(root, start,
3931 root->leafsize);
69a85bd8 3932 start += root->leafsize;
fd8b2b61 3933 if (!eb)
acce952b 3934 continue;
fd8b2b61 3935 wait_on_extent_buffer_writeback(eb);
acce952b 3936
fd8b2b61
JB		 3937 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
3938 &eb->bflags))
3939 clear_extent_buffer_dirty(eb);
3940 free_extent_buffer_stale(eb);
acce952b 3941 }
3942 }
3943
3944 return ret;
3945}
3946
3947static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3948 struct extent_io_tree *pinned_extents)
3949{
3950 struct extent_io_tree *unpin;
3951 u64 start;
3952 u64 end;
3953 int ret;
ed0eaa14 3954 bool loop = true;
acce952b 3955
3956 unpin = pinned_extents;
ed0eaa14 3957again:
acce952b 3958 while (1) {
3959 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 3960 EXTENT_DIRTY, NULL);
acce952b 3961 if (ret)
3962 break;
3963
3964 /* opt_discard */
5378e607
LD		 3965 if (btrfs_test_opt(root, DISCARD))
3966 ret = btrfs_error_discard_extent(root, start,
3967 end + 1 - start,
3968 NULL);
acce952b 3969
3970 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3971 btrfs_error_unpin_extent_range(root, start, end);
3972 cond_resched();
3973 }
3974
ed0eaa14
LB		 3975 if (loop) {
3976 if (unpin == &root->fs_info->freed_extents[0])
3977 unpin = &root->fs_info->freed_extents[1];
3978 else
3979 unpin = &root->fs_info->freed_extents[0];
3980 loop = false;
3981 goto again;
3982 }
3983
acce952b 3984 return 0;
3985}
3986
49b25e05
JM		 3987void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
3988 struct btrfs_root *root)
3989{
724e2315
JB		 3990 btrfs_destroy_ordered_operations(cur_trans, root);
3991
49b25e05 3992 btrfs_destroy_delayed_refs(cur_trans, root);
49b25e05 3993
4a9d8bde 3994 cur_trans->state = TRANS_STATE_COMMIT_START;
d7096fc3 3995 wake_up(&root->fs_info->transaction_blocked_wait);
49b25e05 3996
4a9d8bde 3997 cur_trans->state = TRANS_STATE_UNBLOCKED;
d7096fc3 3998 wake_up(&root->fs_info->transaction_wait);
49b25e05 3999
67cde344
MX		 4000 btrfs_destroy_delayed_inodes(root);
4001 btrfs_assert_delayed_root_empty(root);
49b25e05 4002
49b25e05
JM		 4003 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4004 EXTENT_DIRTY);
6e841e32
LB		 4005 btrfs_destroy_pinned_extent(root,
4006 root->fs_info->pinned_extents);
49b25e05 4007
4a9d8bde
MX		 4008 cur_trans->state =TRANS_STATE_COMPLETED;
4009 wake_up(&cur_trans->commit_wait);
4010
49b25e05
JM		 4011 /*
4012 memset(cur_trans, 0, sizeof(*cur_trans));
4013 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4014 */
4015}
4016
48a3b636 4017static int btrfs_cleanup_transaction(struct btrfs_root *root)
acce952b 4018{
4019 struct btrfs_transaction *t;
acce952b 4020
acce952b 4021 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4022
a4abeea4 4023 spin_lock(&root->fs_info->trans_lock);
724e2315
JB		 4024 while (!list_empty(&root->fs_info->trans_list)) {
4025 t = list_first_entry(&root->fs_info->trans_list,
4026 struct btrfs_transaction, list);
4027 if (t->state >= TRANS_STATE_COMMIT_START) {
4028 atomic_inc(&t->use_count);
4029 spin_unlock(&root->fs_info->trans_lock);
4030 btrfs_wait_for_commit(root, t->transid);
4031 btrfs_put_transaction(t);
4032 spin_lock(&root->fs_info->trans_lock);
4033 continue;
4034 }
4035 if (t == root->fs_info->running_transaction) {
4036 t->state = TRANS_STATE_COMMIT_DOING;
4037 spin_unlock(&root->fs_info->trans_lock);
4038 /*
4039 * We wait for 0 num_writers since we don't hold a trans
4040 * handle open currently for this transaction.
4041 */
4042 wait_event(t->writer_wait,
4043 atomic_read(&t->num_writers) == 0);
4044 } else {
4045 spin_unlock(&root->fs_info->trans_lock);
4046 }
4047 btrfs_cleanup_one_transaction(t, root);
4a9d8bde 4048
724e2315
JB		 4049 spin_lock(&root->fs_info->trans_lock);
4050 if (t == root->fs_info->running_transaction)
4051 root->fs_info->running_transaction = NULL;
acce952b 4052 list_del_init(&t->list);
724e2315 4053 spin_unlock(&root->fs_info->trans_lock);
acce952b 4054
724e2315
JB		 4055 btrfs_put_transaction(t);
4056 trace_btrfs_transaction_commit(root);
4057 spin_lock(&root->fs_info->trans_lock);
4058 }
4059 spin_unlock(&root->fs_info->trans_lock);
4060 btrfs_destroy_all_ordered_extents(root->fs_info);
4061 btrfs_destroy_delayed_inodes(root);
4062 btrfs_assert_delayed_root_empty(root);
4063 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4064 btrfs_destroy_all_delalloc_inodes(root->fs_info);
acce952b 4065 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 4066
4067 return 0;
4068}
4069
d1310b2e 4070static struct extent_io_ops btree_extent_io_ops = {
ce9adaa5 4071 .readpage_end_io_hook = btree_readpage_end_io_hook,
4bb31e92 4072 .readpage_io_failed_hook = btree_io_failed_hook,
0b86a832 4073 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 4074 /* note we're sharing with inode.c for the merge bio hook */
4075 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 4076};
Linux 6.x block layer and io_uring tree(s)