projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: Make btrfs_dev_extent_chunk_tree_uuid() return unsigned long
[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>
4b4e25f2 36#include "compat.h"
eb60ceac
CM		 37#include "ctree.h"
38#include "disk-io.h"
e089f05c 39#include "transaction.h"
0f7d52f4 40#include "btrfs_inode.h"
0b86a832 41#include "volumes.h"
db94535d 42#include "print-tree.h"
8b712842 43#include "async-thread.h"
925baedd 44#include "locking.h"
e02119d5 45#include "tree-log.h"
fa9c0d79 46#include "free-space-cache.h"
581bb050 47#include "inode-map.h"
21adbd5c 48#include "check-integrity.h"
606686ee 49#include "rcu-string.h"
8dabb742 50#include "dev-replace.h"
53b381b3 51#include "raid56.h"
eb60ceac 52
de0022b9
JB		 53#ifdef CONFIG_X86
54#include <asm/cpufeature.h>
55#endif
56
d1310b2e 57static struct extent_io_ops btree_extent_io_ops;
8b712842 58static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 59static void free_fs_root(struct btrfs_root *root);
fcd1f065 60static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 61 int read_only);
569e0f35
JB		 62static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
63 struct btrfs_root *root);
143bede5 64static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
acce952b 65static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
66 struct btrfs_root *root);
aec8030a 67static void btrfs_evict_pending_snapshots(struct btrfs_transaction *t);
143bede5 68static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
acce952b 69static int btrfs_destroy_marked_extents(struct btrfs_root *root,
70 struct extent_io_tree *dirty_pages,
71 int mark);
72static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
73 struct extent_io_tree *pinned_extents);
48a3b636
ES		 74static int btrfs_cleanup_transaction(struct btrfs_root *root);
75static void btrfs_error_commit_super(struct btrfs_root *root);
ce9adaa5 76
d352ac68
CM		 77/*
78 * end_io_wq structs are used to do processing in task context when an IO is
79 * complete. This is used during reads to verify checksums, and it is used
80 * by writes to insert metadata for new file extents after IO is complete.
81 */
ce9adaa5
CM		 82struct end_io_wq {
83 struct bio *bio;
84 bio_end_io_t *end_io;
85 void *private;
86 struct btrfs_fs_info *info;
87 int error;
22c59948 88 int metadata;
ce9adaa5 89 struct list_head list;
8b712842 90 struct btrfs_work work;
ce9adaa5 91};
0da5468f 92
d352ac68
CM		 93/*
94 * async submit bios are used to offload expensive checksumming
95 * onto the worker threads. They checksum file and metadata bios
96 * just before they are sent down the IO stack.
97 */
44b8bd7e
CM		 98struct async_submit_bio {
99 struct inode *inode;
100 struct bio *bio;
101 struct list_head list;
4a69a410
CM		 102 extent_submit_bio_hook_t *submit_bio_start;
103 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM		 104 int rw;
105 int mirror_num;
c8b97818 106 unsigned long bio_flags;
eaf25d93
CM		 107 /*
108 * bio_offset is optional, can be used if the pages in the bio
109 * can't tell us where in the file the bio should go
110 */
111 u64 bio_offset;
8b712842 112 struct btrfs_work work;
79787eaa 113 int error;
44b8bd7e
CM		 114};
115
85d4e461
CM		 116/*
117 * Lockdep class keys for extent_buffer->lock's in this root. For a given
118 * eb, the lockdep key is determined by the btrfs_root it belongs to and
119 * the level the eb occupies in the tree.
120 *
121 * Different roots are used for different purposes and may nest inside each
122 * other and they require separate keysets. As lockdep keys should be
123 * static, assign keysets according to the purpose of the root as indicated
124 * by btrfs_root->objectid. This ensures that all special purpose roots
125 * have separate keysets.
4008c04a 126 *
85d4e461
CM		 127 * Lock-nesting across peer nodes is always done with the immediate parent
128 * node locked thus preventing deadlock. As lockdep doesn't know this, use
129 * subclass to avoid triggering lockdep warning in such cases.
4008c04a 130 *
85d4e461
CM		 131 * The key is set by the readpage_end_io_hook after the buffer has passed
132 * csum validation but before the pages are unlocked. It is also set by
133 * btrfs_init_new_buffer on freshly allocated blocks.
4008c04a 134 *
85d4e461
CM		 135 * We also add a check to make sure the highest level of the tree is the
136 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
137 * needs update as well.
4008c04a
CM		 138 */
139#ifdef CONFIG_DEBUG_LOCK_ALLOC
140# if BTRFS_MAX_LEVEL != 8
141# error
142# endif
85d4e461
CM		 143
144static struct btrfs_lockdep_keyset {
145 u64 id; /* root objectid */
146 const char *name_stem; /* lock name stem */
147 char names[BTRFS_MAX_LEVEL + 1][20];
148 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
149} btrfs_lockdep_keysets[] = {
150 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
151 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
152 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
153 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
154 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
155 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
60b62978 156 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
85d4e461
CM		 157 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
158 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
159 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
160 { .id = 0, .name_stem = "tree" },
4008c04a 161};
85d4e461
CM		 162
163void __init btrfs_init_lockdep(void)
164{
165 int i, j;
166
167 /* initialize lockdep class names */
168 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
169 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
170
171 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
172 snprintf(ks->names[j], sizeof(ks->names[j]),
173 "btrfs-%s-%02d", ks->name_stem, j);
174 }
175}
176
177void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
178 int level)
179{
180 struct btrfs_lockdep_keyset *ks;
181
182 BUG_ON(level >= ARRAY_SIZE(ks->keys));
183
184 /* find the matching keyset, id 0 is the default entry */
185 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
186 if (ks->id == objectid)
187 break;
188
189 lockdep_set_class_and_name(&eb->lock,
190 &ks->keys[level], ks->names[level]);
191}
192
4008c04a
CM		 193#endif
194
d352ac68
CM		 195/*
196 * extents on the btree inode are pretty simple, there's one extent
197 * that covers the entire device
198 */
b2950863 199static struct extent_map *btree_get_extent(struct inode *inode,
306e16ce 200 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 201 int create)
7eccb903 202{
5f39d397
CM		 203 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
204 struct extent_map *em;
205 int ret;
206
890871be 207 read_lock(&em_tree->lock);
d1310b2e 208 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 209 if (em) {
210 em->bdev =
211 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 212 read_unlock(&em_tree->lock);
5f39d397 213 goto out;
a061fc8d 214 }
890871be 215 read_unlock(&em_tree->lock);
7b13b7b1 216
172ddd60 217 em = alloc_extent_map();
5f39d397
CM		 218 if (!em) {
219 em = ERR_PTR(-ENOMEM);
220 goto out;
221 }
222 em->start = 0;
0afbaf8c 223 em->len = (u64)-1;
c8b97818 224 em->block_len = (u64)-1;
5f39d397 225 em->block_start = 0;
a061fc8d 226 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 227
890871be 228 write_lock(&em_tree->lock);
09a2a8f9 229 ret = add_extent_mapping(em_tree, em, 0);
5f39d397
CM		 230 if (ret == -EEXIST) {
231 free_extent_map(em);
7b13b7b1 232 em = lookup_extent_mapping(em_tree, start, len);
b4f359ab 233 if (!em)
0433f20d 234 em = ERR_PTR(-EIO);
5f39d397 235 } else if (ret) {
7b13b7b1 236 free_extent_map(em);
0433f20d 237 em = ERR_PTR(ret);
5f39d397 238 }
890871be 239 write_unlock(&em_tree->lock);
7b13b7b1 240
5f39d397
CM		 241out:
242 return em;
7eccb903
CM		 243}
244
b0496686 245u32 btrfs_csum_data(char *data, u32 seed, size_t len)
19c00ddc 246{
163e783e 247 return crc32c(seed, data, len);
19c00ddc
CM		 248}
249
250void btrfs_csum_final(u32 crc, char *result)
251{
7e75bf3f 252 put_unaligned_le32(~crc, result);
19c00ddc
CM		 253}
254
d352ac68
CM		 255/*
256 * compute the csum for a btree block, and either verify it or write it
257 * into the csum field of the block.
258 */
19c00ddc
CM		 259static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
260 int verify)
261{
6c41761f 262 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
607d432d 263 char *result = NULL;
19c00ddc
CM		 264 unsigned long len;
265 unsigned long cur_len;
266 unsigned long offset = BTRFS_CSUM_SIZE;
19c00ddc
CM		 267 char *kaddr;
268 unsigned long map_start;
269 unsigned long map_len;
270 int err;
271 u32 crc = ~(u32)0;
607d432d 272 unsigned long inline_result;
19c00ddc
CM		 273
274 len = buf->len - offset;
d397712b 275 while (len > 0) {
19c00ddc 276 err = map_private_extent_buffer(buf, offset, 32,
a6591715 277 &kaddr, &map_start, &map_len);
d397712b 278 if (err)
19c00ddc 279 return 1;
19c00ddc 280 cur_len = min(len, map_len - (offset - map_start));
b0496686 281 crc = btrfs_csum_data(kaddr + offset - map_start,
19c00ddc
CM		 282 crc, cur_len);
283 len -= cur_len;
284 offset += cur_len;
19c00ddc 285 }
607d432d
JB		 286 if (csum_size > sizeof(inline_result)) {
287 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
288 if (!result)
289 return 1;
290 } else {
291 result = (char *)&inline_result;
292 }
293
19c00ddc
CM		 294 btrfs_csum_final(crc, result);
295
296 if (verify) {
607d432d 297 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 298 u32 val;
299 u32 found = 0;
607d432d 300 memcpy(&found, result, csum_size);
e4204ded 301
607d432d 302 read_extent_buffer(buf, &val, 0, csum_size);
7a36ddec 303 printk_ratelimited(KERN_INFO "btrfs: %s checksum verify "
193f284d
CM		 304 "failed on %llu wanted %X found %X "
305 "level %d\n",
c1c9ff7c
GU		 306 root->fs_info->sb->s_id, buf->start,
307 val, found, btrfs_header_level(buf));
607d432d
JB		 308 if (result != (char *)&inline_result)
309 kfree(result);
19c00ddc
CM		 310 return 1;
311 }
312 } else {
607d432d 313 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 314 }
607d432d
JB		 315 if (result != (char *)&inline_result)
316 kfree(result);
19c00ddc
CM		 317 return 0;
318}
319
d352ac68
CM		 320/*
321 * we can't consider a given block up to date unless the transid of the
322 * block matches the transid in the parent node's pointer. This is how we
323 * detect blocks that either didn't get written at all or got written
324 * in the wrong place.
325 */
1259ab75 326static int verify_parent_transid(struct extent_io_tree *io_tree,
b9fab919
CM		 327 struct extent_buffer *eb, u64 parent_transid,
328 int atomic)
1259ab75 329{
2ac55d41 330 struct extent_state *cached_state = NULL;
1259ab75
CM		 331 int ret;
332
333 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
334 return 0;
335
b9fab919
CM		 336 if (atomic)
337 return -EAGAIN;
338
2ac55d41 339 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
d0082371 340 0, &cached_state);
0b32f4bb 341 if (extent_buffer_uptodate(eb) &&
1259ab75
CM		 342 btrfs_header_generation(eb) == parent_transid) {
343 ret = 0;
344 goto out;
345 }
7a36ddec 346 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
193f284d 347 "found %llu\n",
c1c9ff7c 348 eb->start, parent_transid, btrfs_header_generation(eb));
1259ab75 349 ret = 1;
0b32f4bb 350 clear_extent_buffer_uptodate(eb);
33958dc6 351out:
2ac55d41
JB		 352 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
353 &cached_state, GFP_NOFS);
1259ab75 354 return ret;
1259ab75
CM		 355}
356
1104a885
DS		 357/*
358 * Return 0 if the superblock checksum type matches the checksum value of that
359 * algorithm. Pass the raw disk superblock data.
360 */
361static int btrfs_check_super_csum(char *raw_disk_sb)
362{
363 struct btrfs_super_block *disk_sb =
364 (struct btrfs_super_block *)raw_disk_sb;
365 u16 csum_type = btrfs_super_csum_type(disk_sb);
366 int ret = 0;
367
368 if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
369 u32 crc = ~(u32)0;
370 const int csum_size = sizeof(crc);
371 char result[csum_size];
372
373 /*
374 * The super_block structure does not span the whole
375 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
376 * is filled with zeros and is included in the checkum.
377 */
378 crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
379 crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
380 btrfs_csum_final(crc, result);
381
382 if (memcmp(raw_disk_sb, result, csum_size))
383 ret = 1;
667e7d94
CM		 384
385 if (ret && btrfs_super_generation(disk_sb) < 10) {
386 printk(KERN_WARNING "btrfs: super block crcs don't match, older mkfs detected\n");
387 ret = 0;
388 }
1104a885
DS		 389 }
390
391 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
392 printk(KERN_ERR "btrfs: unsupported checksum algorithm %u\n",
393 csum_type);
394 ret = 1;
395 }
396
397 return ret;
398}
399
d352ac68
CM		 400/*
401 * helper to read a given tree block, doing retries as required when
402 * the checksums don't match and we have alternate mirrors to try.
403 */
f188591e
CM		 404static int btree_read_extent_buffer_pages(struct btrfs_root *root,
405 struct extent_buffer *eb,
ca7a79ad 406 u64 start, u64 parent_transid)
f188591e
CM		 407{
408 struct extent_io_tree *io_tree;
ea466794 409 int failed = 0;
f188591e
CM		 410 int ret;
411 int num_copies = 0;
412 int mirror_num = 0;
ea466794 413 int failed_mirror = 0;
f188591e 414
a826d6dc 415 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
f188591e
CM		 416 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
417 while (1) {
bb82ab88
AJ		 418 ret = read_extent_buffer_pages(io_tree, eb, start,
419 WAIT_COMPLETE,
f188591e 420 btree_get_extent, mirror_num);
256dd1bb
SB		 421 if (!ret) {
422 if (!verify_parent_transid(io_tree, eb,
b9fab919 423 parent_transid, 0))
256dd1bb
SB		 424 break;
425 else
426 ret = -EIO;
427 }
d397712b 428
a826d6dc
JB		 429 /*
430 * This buffer's crc is fine, but its contents are corrupted, so
431 * there is no reason to read the other copies, they won't be
432 * any less wrong.
433 */
434 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
ea466794
JB		 435 break;
436
5d964051 437 num_copies = btrfs_num_copies(root->fs_info,
f188591e 438 eb->start, eb->len);
4235298e 439 if (num_copies == 1)
ea466794 440 break;
4235298e 441
5cf1ab56
JB		 442 if (!failed_mirror) {
443 failed = 1;
444 failed_mirror = eb->read_mirror;
445 }
446
f188591e 447 mirror_num++;
ea466794
JB		 448 if (mirror_num == failed_mirror)
449 mirror_num++;
450
4235298e 451 if (mirror_num > num_copies)
ea466794 452 break;
f188591e 453 }
ea466794 454
c0901581 455 if (failed && !ret && failed_mirror)
ea466794
JB		 456 repair_eb_io_failure(root, eb, failed_mirror);
457
458 return ret;
f188591e 459}
19c00ddc 460
d352ac68 461/*
d397712b
CM		 462 * checksum a dirty tree block before IO. This has extra checks to make sure
463 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 464 */
d397712b 465
b2950863 466static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
19c00ddc 467{
d1310b2e 468 struct extent_io_tree *tree;
4eee4fa4 469 u64 start = page_offset(page);
19c00ddc 470 u64 found_start;
19c00ddc 471 struct extent_buffer *eb;
f188591e 472
d1310b2e 473 tree = &BTRFS_I(page->mapping->host)->io_tree;
19c00ddc 474
4f2de97a
JB		 475 eb = (struct extent_buffer *)page->private;
476 if (page != eb->pages[0])
477 return 0;
19c00ddc
CM		 478 found_start = btrfs_header_bytenr(eb);
479 if (found_start != start) {
55c69072 480 WARN_ON(1);
4f2de97a 481 return 0;
55c69072 482 }
55c69072 483 if (!PageUptodate(page)) {
55c69072 484 WARN_ON(1);
4f2de97a 485 return 0;
19c00ddc 486 }
19c00ddc 487 csum_tree_block(root, eb, 0);
19c00ddc
CM		 488 return 0;
489}
490
2b82032c
YZ		 491static int check_tree_block_fsid(struct btrfs_root *root,
492 struct extent_buffer *eb)
493{
494 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
495 u8 fsid[BTRFS_UUID_SIZE];
496 int ret = 1;
497
498 read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
499 BTRFS_FSID_SIZE);
500 while (fs_devices) {
501 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
502 ret = 0;
503 break;
504 }
505 fs_devices = fs_devices->seed;
506 }
507 return ret;
508}
509
a826d6dc
JB		 510#define CORRUPT(reason, eb, root, slot) \
511 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
512 "root=%llu, slot=%d\n", reason, \
c1c9ff7c 513 btrfs_header_bytenr(eb), root->objectid, slot)
a826d6dc
JB		 514
515static noinline int check_leaf(struct btrfs_root *root,
516 struct extent_buffer *leaf)
517{
518 struct btrfs_key key;
519 struct btrfs_key leaf_key;
520 u32 nritems = btrfs_header_nritems(leaf);
521 int slot;
522
523 if (nritems == 0)
524 return 0;
525
526 /* Check the 0 item */
527 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
528 BTRFS_LEAF_DATA_SIZE(root)) {
529 CORRUPT("invalid item offset size pair", leaf, root, 0);
530 return -EIO;
531 }
532
533 /*
534 * Check to make sure each items keys are in the correct order and their
535 * offsets make sense. We only have to loop through nritems-1 because
536 * we check the current slot against the next slot, which verifies the
537 * next slot's offset+size makes sense and that the current's slot
538 * offset is correct.
539 */
540 for (slot = 0; slot < nritems - 1; slot++) {
541 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
542 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
543
544 /* Make sure the keys are in the right order */
545 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
546 CORRUPT("bad key order", leaf, root, slot);
547 return -EIO;
548 }
549
550 /*
551 * Make sure the offset and ends are right, remember that the
552 * item data starts at the end of the leaf and grows towards the
553 * front.
554 */
555 if (btrfs_item_offset_nr(leaf, slot) !=
556 btrfs_item_end_nr(leaf, slot + 1)) {
557 CORRUPT("slot offset bad", leaf, root, slot);
558 return -EIO;
559 }
560
561 /*
562 * Check to make sure that we don't point outside of the leaf,
563 * just incase all the items are consistent to eachother, but
564 * all point outside of the leaf.
565 */
566 if (btrfs_item_end_nr(leaf, slot) >
567 BTRFS_LEAF_DATA_SIZE(root)) {
568 CORRUPT("slot end outside of leaf", leaf, root, slot);
569 return -EIO;
570 }
571 }
572
573 return 0;
574}
575
facc8a22
MX		 576static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
577 u64 phy_offset, struct page *page,
578 u64 start, u64 end, int mirror)
ce9adaa5
CM		 579{
580 struct extent_io_tree *tree;
581 u64 found_start;
582 int found_level;
ce9adaa5
CM		 583 struct extent_buffer *eb;
584 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 585 int ret = 0;
727011e0 586 int reads_done;
ce9adaa5 587
ce9adaa5
CM		 588 if (!page->private)
589 goto out;
d397712b 590
727011e0 591 tree = &BTRFS_I(page->mapping->host)->io_tree;
4f2de97a 592 eb = (struct extent_buffer *)page->private;
d397712b 593
0b32f4bb
JB		 594 /* the pending IO might have been the only thing that kept this buffer
595 * in memory. Make sure we have a ref for all this other checks
596 */
597 extent_buffer_get(eb);
598
599 reads_done = atomic_dec_and_test(&eb->io_pages);
727011e0
CM		 600 if (!reads_done)
601 goto err;
f188591e 602
5cf1ab56 603 eb->read_mirror = mirror;
ea466794
JB		 604 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
605 ret = -EIO;
606 goto err;
607 }
608
ce9adaa5 609 found_start = btrfs_header_bytenr(eb);
727011e0 610 if (found_start != eb->start) {
7a36ddec 611 printk_ratelimited(KERN_INFO "btrfs bad tree block start "
193f284d 612 "%llu %llu\n",
c1c9ff7c 613 found_start, eb->start);
f188591e 614 ret = -EIO;
ce9adaa5
CM		 615 goto err;
616 }
2b82032c 617 if (check_tree_block_fsid(root, eb)) {
7a36ddec 618 printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n",
c1c9ff7c 619 eb->start);
1259ab75
CM		 620 ret = -EIO;
621 goto err;
622 }
ce9adaa5 623 found_level = btrfs_header_level(eb);
1c24c3ce
JB		 624 if (found_level >= BTRFS_MAX_LEVEL) {
625 btrfs_info(root->fs_info, "bad tree block level %d\n",
626 (int)btrfs_header_level(eb));
627 ret = -EIO;
628 goto err;
629 }
ce9adaa5 630
85d4e461
CM		 631 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
632 eb, found_level);
4008c04a 633
ce9adaa5 634 ret = csum_tree_block(root, eb, 1);
a826d6dc 635 if (ret) {
f188591e 636 ret = -EIO;
a826d6dc
JB		 637 goto err;
638 }
639
640 /*
641 * If this is a leaf block and it is corrupt, set the corrupt bit so
642 * that we don't try and read the other copies of this block, just
643 * return -EIO.
644 */
645 if (found_level == 0 && check_leaf(root, eb)) {
646 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
647 ret = -EIO;
648 }
ce9adaa5 649
0b32f4bb
JB		 650 if (!ret)
651 set_extent_buffer_uptodate(eb);
ce9adaa5 652err:
79fb65a1
JB		 653 if (reads_done &&
654 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 655 btree_readahead_hook(root, eb, eb->start, ret);
4bb31e92 656
53b381b3
DW		 657 if (ret) {
658 /*
659 * our io error hook is going to dec the io pages
660 * again, we have to make sure it has something
661 * to decrement
662 */
663 atomic_inc(&eb->io_pages);
0b32f4bb 664 clear_extent_buffer_uptodate(eb);
53b381b3 665 }
0b32f4bb 666 free_extent_buffer(eb);
ce9adaa5 667out:
f188591e 668 return ret;
ce9adaa5
CM		 669}
670
ea466794 671static int btree_io_failed_hook(struct page *page, int failed_mirror)
4bb31e92 672{
4bb31e92
AJ		 673 struct extent_buffer *eb;
674 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
675
4f2de97a 676 eb = (struct extent_buffer *)page->private;
ea466794 677 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
5cf1ab56 678 eb->read_mirror = failed_mirror;
53b381b3 679 atomic_dec(&eb->io_pages);
ea466794 680 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 681 btree_readahead_hook(root, eb, eb->start, -EIO);
4bb31e92
AJ		 682 return -EIO; /* we fixed nothing */
683}
684
ce9adaa5 685static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM		 686{
687 struct end_io_wq *end_io_wq = bio->bi_private;
688 struct btrfs_fs_info *fs_info;
ce9adaa5 689
ce9adaa5 690 fs_info = end_io_wq->info;
ce9adaa5 691 end_io_wq->error = err;
8b712842
CM		 692 end_io_wq->work.func = end_workqueue_fn;
693 end_io_wq->work.flags = 0;
d20f7043 694
7b6d91da 695 if (bio->bi_rw & REQ_WRITE) {
53b381b3 696 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA)
cad321ad
CM		 697 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
698 &end_io_wq->work);
53b381b3 699 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE)
0cb59c99
JB		 700 btrfs_queue_worker(&fs_info->endio_freespace_worker,
701 &end_io_wq->work);
53b381b3
DW		 702 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
703 btrfs_queue_worker(&fs_info->endio_raid56_workers,
704 &end_io_wq->work);
cad321ad
CM		 705 else
706 btrfs_queue_worker(&fs_info->endio_write_workers,
707 &end_io_wq->work);
d20f7043 708 } else {
53b381b3
DW		 709 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
710 btrfs_queue_worker(&fs_info->endio_raid56_workers,
711 &end_io_wq->work);
712 else if (end_io_wq->metadata)
d20f7043
CM		 713 btrfs_queue_worker(&fs_info->endio_meta_workers,
714 &end_io_wq->work);
715 else
716 btrfs_queue_worker(&fs_info->endio_workers,
717 &end_io_wq->work);
718 }
ce9adaa5
CM		 719}
720
0cb59c99
JB		 721/*
722 * For the metadata arg you want
723 *
724 * 0 - if data
725 * 1 - if normal metadta
726 * 2 - if writing to the free space cache area
53b381b3 727 * 3 - raid parity work
0cb59c99 728 */
22c59948
CM		 729int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
730 int metadata)
0b86a832 731{
ce9adaa5 732 struct end_io_wq *end_io_wq;
ce9adaa5
CM		 733 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
734 if (!end_io_wq)
735 return -ENOMEM;
736
737 end_io_wq->private = bio->bi_private;
738 end_io_wq->end_io = bio->bi_end_io;
22c59948 739 end_io_wq->info = info;
ce9adaa5
CM		 740 end_io_wq->error = 0;
741 end_io_wq->bio = bio;
22c59948 742 end_io_wq->metadata = metadata;
ce9adaa5
CM		 743
744 bio->bi_private = end_io_wq;
745 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 746 return 0;
747}
748
b64a2851 749unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 750{
4854ddd0
CM		 751 unsigned long limit = min_t(unsigned long,
752 info->workers.max_workers,
753 info->fs_devices->open_devices);
754 return 256 * limit;
755}
0986fe9e 756
4a69a410
CM		 757static void run_one_async_start(struct btrfs_work *work)
758{
4a69a410 759 struct async_submit_bio *async;
79787eaa 760 int ret;
4a69a410
CM		 761
762 async = container_of(work, struct async_submit_bio, work);
79787eaa
JM		 763 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
764 async->mirror_num, async->bio_flags,
765 async->bio_offset);
766 if (ret)
767 async->error = ret;
4a69a410
CM		 768}
769
770static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 771{
772 struct btrfs_fs_info *fs_info;
773 struct async_submit_bio *async;
4854ddd0 774 int limit;
8b712842
CM		 775
776 async = container_of(work, struct async_submit_bio, work);
777 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 778
b64a2851 779 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 780 limit = limit * 2 / 3;
781
66657b31 782 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
b64a2851 783 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 784 wake_up(&fs_info->async_submit_wait);
785
79787eaa
JM		 786 /* If an error occured we just want to clean up the bio and move on */
787 if (async->error) {
788 bio_endio(async->bio, async->error);
789 return;
790 }
791
4a69a410 792 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 793 async->mirror_num, async->bio_flags,
794 async->bio_offset);
4a69a410
CM		 795}
796
797static void run_one_async_free(struct btrfs_work *work)
798{
799 struct async_submit_bio *async;
800
801 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 802 kfree(async);
803}
804
44b8bd7e
CM		 805int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
806 int rw, struct bio *bio, int mirror_num,
c8b97818 807 unsigned long bio_flags,
eaf25d93 808 u64 bio_offset,
4a69a410
CM		 809 extent_submit_bio_hook_t *submit_bio_start,
810 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 811{
812 struct async_submit_bio *async;
813
814 async = kmalloc(sizeof(*async), GFP_NOFS);
815 if (!async)
816 return -ENOMEM;
817
818 async->inode = inode;
819 async->rw = rw;
820 async->bio = bio;
821 async->mirror_num = mirror_num;
4a69a410
CM		 822 async->submit_bio_start = submit_bio_start;
823 async->submit_bio_done = submit_bio_done;
824
825 async->work.func = run_one_async_start;
826 async->work.ordered_func = run_one_async_done;
827 async->work.ordered_free = run_one_async_free;
828
8b712842 829 async->work.flags = 0;
c8b97818 830 async->bio_flags = bio_flags;
eaf25d93 831 async->bio_offset = bio_offset;
8c8bee1d 832
79787eaa
JM		 833 async->error = 0;
834
cb03c743 835 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 836
7b6d91da 837 if (rw & REQ_SYNC)
d313d7a3
CM		 838 btrfs_set_work_high_prio(&async->work);
839
8b712842 840 btrfs_queue_worker(&fs_info->workers, &async->work);
9473f16c 841
d397712b 842 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 843 atomic_read(&fs_info->nr_async_submits)) {
844 wait_event(fs_info->async_submit_wait,
845 (atomic_read(&fs_info->nr_async_submits) == 0));
846 }
847
44b8bd7e
CM		 848 return 0;
849}
850
ce3ed71a
CM		 851static int btree_csum_one_bio(struct bio *bio)
852{
853 struct bio_vec *bvec = bio->bi_io_vec;
854 int bio_index = 0;
855 struct btrfs_root *root;
79787eaa 856 int ret = 0;
ce3ed71a
CM		 857
858 WARN_ON(bio->bi_vcnt <= 0);
d397712b 859 while (bio_index < bio->bi_vcnt) {
ce3ed71a 860 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
79787eaa
JM		 861 ret = csum_dirty_buffer(root, bvec->bv_page);
862 if (ret)
863 break;
ce3ed71a
CM		 864 bio_index++;
865 bvec++;
866 }
79787eaa 867 return ret;
ce3ed71a
CM		 868}
869
4a69a410
CM		 870static int __btree_submit_bio_start(struct inode *inode, int rw,
871 struct bio *bio, int mirror_num,
eaf25d93
CM		 872 unsigned long bio_flags,
873 u64 bio_offset)
22c59948 874{
8b712842
CM		 875 /*
876 * when we're called for a write, we're already in the async
5443be45 877 * submission context. Just jump into btrfs_map_bio
8b712842 878 */
79787eaa 879 return btree_csum_one_bio(bio);
4a69a410 880}
22c59948 881
4a69a410 882static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 883 int mirror_num, unsigned long bio_flags,
884 u64 bio_offset)
4a69a410 885{
61891923
SB		 886 int ret;
887
8b712842 888 /*
4a69a410
CM		 889 * when we're called for a write, we're already in the async
890 * submission context. Just jump into btrfs_map_bio
8b712842 891 */
61891923
SB		 892 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
893 if (ret)
894 bio_endio(bio, ret);
895 return ret;
0b86a832
CM		 896}
897
de0022b9
JB		 898static int check_async_write(struct inode *inode, unsigned long bio_flags)
899{
900 if (bio_flags & EXTENT_BIO_TREE_LOG)
901 return 0;
902#ifdef CONFIG_X86
903 if (cpu_has_xmm4_2)
904 return 0;
905#endif
906 return 1;
907}
908
44b8bd7e 909static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 910 int mirror_num, unsigned long bio_flags,
911 u64 bio_offset)
44b8bd7e 912{
de0022b9 913 int async = check_async_write(inode, bio_flags);
cad321ad
CM		 914 int ret;
915
7b6d91da 916 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 917 /*
918 * called for a read, do the setup so that checksum validation
919 * can happen in the async kernel threads
920 */
f3f266ab
CM		 921 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
922 bio, 1);
1d4284bd 923 if (ret)
61891923
SB		 924 goto out_w_error;
925 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
926 mirror_num, 0);
de0022b9
JB		 927 } else if (!async) {
928 ret = btree_csum_one_bio(bio);
929 if (ret)
61891923
SB		 930 goto out_w_error;
931 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
932 mirror_num, 0);
933 } else {
934 /*
935 * kthread helpers are used to submit writes so that
936 * checksumming can happen in parallel across all CPUs
937 */
938 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
939 inode, rw, bio, mirror_num, 0,
940 bio_offset,
941 __btree_submit_bio_start,
942 __btree_submit_bio_done);
44b8bd7e 943 }
d313d7a3 944
61891923
SB		 945 if (ret) {
946out_w_error:
947 bio_endio(bio, ret);
948 }
949 return ret;
44b8bd7e
CM		 950}
951
3dd1462e 952#ifdef CONFIG_MIGRATION
784b4e29 953static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG		 954 struct page *newpage, struct page *page,
955 enum migrate_mode mode)
784b4e29
CM		 956{
957 /*
958 * we can't safely write a btree page from here,
959 * we haven't done the locking hook
960 */
961 if (PageDirty(page))
962 return -EAGAIN;
963 /*
964 * Buffers may be managed in a filesystem specific way.
965 * We must have no buffers or drop them.
966 */
967 if (page_has_private(page) &&
968 !try_to_release_page(page, GFP_KERNEL))
969 return -EAGAIN;
a6bc32b8 970 return migrate_page(mapping, newpage, page, mode);
784b4e29 971}
3dd1462e 972#endif
784b4e29 973
0da5468f
CM		 974
975static int btree_writepages(struct address_space *mapping,
976 struct writeback_control *wbc)
977{
d1310b2e 978 struct extent_io_tree *tree;
e2d84521
MX		 979 struct btrfs_fs_info *fs_info;
980 int ret;
981
d1310b2e 982 tree = &BTRFS_I(mapping->host)->io_tree;
d8d5f3e1 983 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM		 984
985 if (wbc->for_kupdate)
986 return 0;
987
e2d84521 988 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 989 /* this is a bit racy, but that's ok */
e2d84521
MX		 990 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
991 BTRFS_DIRTY_METADATA_THRESH);
992 if (ret < 0)
793955bc 993 return 0;
793955bc 994 }
0b32f4bb 995 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM		 996}
997
b2950863 998static int btree_readpage(struct file *file, struct page *page)
5f39d397 999{
d1310b2e
CM		 1000 struct extent_io_tree *tree;
1001 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 1002 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 1003}
22b0ebda 1004
70dec807 1005static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 1006{
98509cfc 1007 if (PageWriteback(page) || PageDirty(page))
d397712b 1008 return 0;
0c4e538b 1009
f7a52a40 1010 return try_release_extent_buffer(page);
d98237b3
CM		 1011}
1012
d47992f8
LC		 1013static void btree_invalidatepage(struct page *page, unsigned int offset,
1014 unsigned int length)
d98237b3 1015{
d1310b2e
CM		 1016 struct extent_io_tree *tree;
1017 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 1018 extent_invalidatepage(tree, page, offset);
1019 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 1020 if (PagePrivate(page)) {
d397712b
CM		 1021 printk(KERN_WARNING "btrfs warning page private not zero "
1022 "on page %llu\n", (unsigned long long)page_offset(page));
9ad6b7bc
CM		 1023 ClearPagePrivate(page);
1024 set_page_private(page, 0);
1025 page_cache_release(page);
1026 }
d98237b3
CM		 1027}
1028
0b32f4bb
JB		 1029static int btree_set_page_dirty(struct page *page)
1030{
bb146eb2 1031#ifdef DEBUG
0b32f4bb
JB		 1032 struct extent_buffer *eb;
1033
1034 BUG_ON(!PagePrivate(page));
1035 eb = (struct extent_buffer *)page->private;
1036 BUG_ON(!eb);
1037 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1038 BUG_ON(!atomic_read(&eb->refs));
1039 btrfs_assert_tree_locked(eb);
bb146eb2 1040#endif
0b32f4bb
JB		 1041 return __set_page_dirty_nobuffers(page);
1042}
1043
7f09410b 1044static const struct address_space_operations btree_aops = {
d98237b3 1045 .readpage = btree_readpage,
0da5468f 1046 .writepages = btree_writepages,
5f39d397
CM		 1047 .releasepage = btree_releasepage,
1048 .invalidatepage = btree_invalidatepage,
5a92bc88 1049#ifdef CONFIG_MIGRATION
784b4e29 1050 .migratepage = btree_migratepage,
5a92bc88 1051#endif
0b32f4bb 1052 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM		 1053};
1054
ca7a79ad
CM		 1055int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1056 u64 parent_transid)
090d1875 1057{
5f39d397
CM		 1058 struct extent_buffer *buf = NULL;
1059 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 1060 int ret = 0;
090d1875 1061
db94535d 1062 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 1063 if (!buf)
090d1875 1064 return 0;
d1310b2e 1065 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
bb82ab88 1066 buf, 0, WAIT_NONE, btree_get_extent, 0);
5f39d397 1067 free_extent_buffer(buf);
de428b63 1068 return ret;
090d1875
CM		 1069}
1070
ab0fff03
AJ		 1071int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1072 int mirror_num, struct extent_buffer **eb)
1073{
1074 struct extent_buffer *buf = NULL;
1075 struct inode *btree_inode = root->fs_info->btree_inode;
1076 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1077 int ret;
1078
1079 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1080 if (!buf)
1081 return 0;
1082
1083 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1084
1085 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1086 btree_get_extent, mirror_num);
1087 if (ret) {
1088 free_extent_buffer(buf);
1089 return ret;
1090 }
1091
1092 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1093 free_extent_buffer(buf);
1094 return -EIO;
0b32f4bb 1095 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ		 1096 *eb = buf;
1097 } else {
1098 free_extent_buffer(buf);
1099 }
1100 return 0;
1101}
1102
0999df54
CM		 1103struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1104 u64 bytenr, u32 blocksize)
1105{
1106 struct inode *btree_inode = root->fs_info->btree_inode;
1107 struct extent_buffer *eb;
1108 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
f09d1f60 1109 bytenr, blocksize);
0999df54
CM		 1110 return eb;
1111}
1112
1113struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1114 u64 bytenr, u32 blocksize)
1115{
1116 struct inode *btree_inode = root->fs_info->btree_inode;
1117 struct extent_buffer *eb;
1118
1119 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
727011e0 1120 bytenr, blocksize);
0999df54
CM		 1121 return eb;
1122}
1123
1124
e02119d5
CM		 1125int btrfs_write_tree_block(struct extent_buffer *buf)
1126{
727011e0 1127 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1128 buf->start + buf->len - 1);
e02119d5
CM		 1129}
1130
1131int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1132{
727011e0 1133 return filemap_fdatawait_range(buf->pages[0]->mapping,
8aa38c31 1134 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1135}
1136
0999df54 1137struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 1138 u32 blocksize, u64 parent_transid)
0999df54
CM		 1139{
1140 struct extent_buffer *buf = NULL;
0999df54
CM		 1141 int ret;
1142
0999df54
CM		 1143 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1144 if (!buf)
1145 return NULL;
0999df54 1146
ca7a79ad 1147 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
0f0fe8f7
FDBM		 1148 if (ret) {
1149 free_extent_buffer(buf);
1150 return NULL;
1151 }
5f39d397 1152 return buf;
ce9adaa5 1153
eb60ceac
CM		 1154}
1155
d5c13f92
JM		 1156void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1157 struct extent_buffer *buf)
ed2ff2cb 1158{
e2d84521
MX		 1159 struct btrfs_fs_info *fs_info = root->fs_info;
1160
55c69072 1161 if (btrfs_header_generation(buf) ==
e2d84521 1162 fs_info->running_transaction->transid) {
b9447ef8 1163 btrfs_assert_tree_locked(buf);
b4ce94de 1164
b9473439 1165 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
e2d84521
MX		 1166 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1167 -buf->len,
1168 fs_info->dirty_metadata_batch);
ed7b63eb
JB		 1169 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1170 btrfs_set_lock_blocking(buf);
1171 clear_extent_buffer_dirty(buf);
1172 }
925baedd 1173 }
5f39d397
CM		 1174}
1175
143bede5
JM		 1176static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1177 u32 stripesize, struct btrfs_root *root,
1178 struct btrfs_fs_info *fs_info,
1179 u64 objectid)
d97e63b6 1180{
cfaa7295 1181 root->node = NULL;
a28ec197 1182 root->commit_root = NULL;
db94535d
CM		 1183 root->sectorsize = sectorsize;
1184 root->nodesize = nodesize;
1185 root->leafsize = leafsize;
87ee04eb 1186 root->stripesize = stripesize;
123abc88 1187 root->ref_cows = 0;
0b86a832 1188 root->track_dirty = 0;
c71bf099 1189 root->in_radix = 0;
d68fc57b
YZ		 1190 root->orphan_item_inserted = 0;
1191 root->orphan_cleanup_state = 0;
0b86a832 1192
0f7d52f4
CM		 1193 root->objectid = objectid;
1194 root->last_trans = 0;
13a8a7c8 1195 root->highest_objectid = 0;
eb73c1b7 1196 root->nr_delalloc_inodes = 0;
199c2a9c 1197 root->nr_ordered_extents = 0;
58176a96 1198 root->name = NULL;
6bef4d31 1199 root->inode_tree = RB_ROOT;
16cdcec7 1200 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1201 root->block_rsv = NULL;
d68fc57b 1202 root->orphan_block_rsv = NULL;
0b86a832
CM		 1203
1204 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1205 INIT_LIST_HEAD(&root->root_list);
eb73c1b7
MX		 1206 INIT_LIST_HEAD(&root->delalloc_inodes);
1207 INIT_LIST_HEAD(&root->delalloc_root);
199c2a9c
MX		 1208 INIT_LIST_HEAD(&root->ordered_extents);
1209 INIT_LIST_HEAD(&root->ordered_root);
2ab28f32
JB		 1210 INIT_LIST_HEAD(&root->logged_list[0]);
1211 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1212 spin_lock_init(&root->orphan_lock);
5d4f98a2 1213 spin_lock_init(&root->inode_lock);
eb73c1b7 1214 spin_lock_init(&root->delalloc_lock);
199c2a9c 1215 spin_lock_init(&root->ordered_extent_lock);
f0486c68 1216 spin_lock_init(&root->accounting_lock);
2ab28f32
JB		 1217 spin_lock_init(&root->log_extents_lock[0]);
1218 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1219 mutex_init(&root->objectid_mutex);
e02119d5 1220 mutex_init(&root->log_mutex);
7237f183
YZ		 1221 init_waitqueue_head(&root->log_writer_wait);
1222 init_waitqueue_head(&root->log_commit_wait[0]);
1223 init_waitqueue_head(&root->log_commit_wait[1]);
1224 atomic_set(&root->log_commit[0], 0);
1225 atomic_set(&root->log_commit[1], 0);
1226 atomic_set(&root->log_writers, 0);
2ecb7923 1227 atomic_set(&root->log_batch, 0);
8a35d95f 1228 atomic_set(&root->orphan_inodes, 0);
b0feb9d9 1229 atomic_set(&root->refs, 1);
7237f183 1230 root->log_transid = 0;
257c62e1 1231 root->last_log_commit = 0;
d0c803c4 1232 extent_io_tree_init(&root->dirty_log_pages,
f993c883 1233 fs_info->btree_inode->i_mapping);
017e5369 1234
3768f368
CM		 1235 memset(&root->root_key, 0, sizeof(root->root_key));
1236 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1237 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1238 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
3f157a2f 1239 root->defrag_trans_start = fs_info->generation;
58176a96 1240 init_completion(&root->kobj_unregister);
6702ed49 1241 root->defrag_running = 0;
4d775673 1242 root->root_key.objectid = objectid;
0ee5dc67 1243 root->anon_dev = 0;
8ea05e3a 1244
5f3ab90a 1245 spin_lock_init(&root->root_item_lock);
3768f368
CM		 1246}
1247
f84a8bd6 1248static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
6f07e42e
AV		 1249{
1250 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1251 if (root)
1252 root->fs_info = fs_info;
1253 return root;
1254}
1255
20897f5c
AJ		 1256struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1257 struct btrfs_fs_info *fs_info,
1258 u64 objectid)
1259{
1260 struct extent_buffer *leaf;
1261 struct btrfs_root *tree_root = fs_info->tree_root;
1262 struct btrfs_root *root;
1263 struct btrfs_key key;
1264 int ret = 0;
1265 u64 bytenr;
6463fe58 1266 uuid_le uuid;
20897f5c
AJ		 1267
1268 root = btrfs_alloc_root(fs_info);
1269 if (!root)
1270 return ERR_PTR(-ENOMEM);
1271
1272 __setup_root(tree_root->nodesize, tree_root->leafsize,
1273 tree_root->sectorsize, tree_root->stripesize,
1274 root, fs_info, objectid);
1275 root->root_key.objectid = objectid;
1276 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1277 root->root_key.offset = 0;
1278
1279 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1280 0, objectid, NULL, 0, 0, 0);
1281 if (IS_ERR(leaf)) {
1282 ret = PTR_ERR(leaf);
1dd05682 1283 leaf = NULL;
20897f5c
AJ		 1284 goto fail;
1285 }
1286
1287 bytenr = leaf->start;
1288 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1289 btrfs_set_header_bytenr(leaf, leaf->start);
1290 btrfs_set_header_generation(leaf, trans->transid);
1291 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1292 btrfs_set_header_owner(leaf, objectid);
1293 root->node = leaf;
1294
1295 write_extent_buffer(leaf, fs_info->fsid,
1296 (unsigned long)btrfs_header_fsid(leaf),
1297 BTRFS_FSID_SIZE);
1298 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
1299 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
1300 BTRFS_UUID_SIZE);
1301 btrfs_mark_buffer_dirty(leaf);
1302
1303 root->commit_root = btrfs_root_node(root);
1304 root->track_dirty = 1;
1305
1306
1307 root->root_item.flags = 0;
1308 root->root_item.byte_limit = 0;
1309 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1310 btrfs_set_root_generation(&root->root_item, trans->transid);
1311 btrfs_set_root_level(&root->root_item, 0);
1312 btrfs_set_root_refs(&root->root_item, 1);
1313 btrfs_set_root_used(&root->root_item, leaf->len);
1314 btrfs_set_root_last_snapshot(&root->root_item, 0);
1315 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB		 1316 uuid_le_gen(&uuid);
1317 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ		 1318 root->root_item.drop_level = 0;
1319
1320 key.objectid = objectid;
1321 key.type = BTRFS_ROOT_ITEM_KEY;
1322 key.offset = 0;
1323 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1324 if (ret)
1325 goto fail;
1326
1327 btrfs_tree_unlock(leaf);
1328
1dd05682
TI		 1329 return root;
1330
20897f5c 1331fail:
1dd05682
TI		 1332 if (leaf) {
1333 btrfs_tree_unlock(leaf);
1334 free_extent_buffer(leaf);
1335 }
1336 kfree(root);
20897f5c 1337
1dd05682 1338 return ERR_PTR(ret);
20897f5c
AJ		 1339}
1340
7237f183
YZ		 1341static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1342 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1343{
1344 struct btrfs_root *root;
1345 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1346 struct extent_buffer *leaf;
e02119d5 1347
6f07e42e 1348 root = btrfs_alloc_root(fs_info);
e02119d5 1349 if (!root)
7237f183 1350 return ERR_PTR(-ENOMEM);
e02119d5
CM		 1351
1352 __setup_root(tree_root->nodesize, tree_root->leafsize,
1353 tree_root->sectorsize, tree_root->stripesize,
1354 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1355
1356 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1357 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1358 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
7237f183
YZ		 1359 /*
1360 * log trees do not get reference counted because they go away
1361 * before a real commit is actually done. They do store pointers
1362 * to file data extents, and those reference counts still get
1363 * updated (along with back refs to the log tree).
1364 */
e02119d5
CM		 1365 root->ref_cows = 0;
1366
5d4f98a2 1367 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
66d7e7f0 1368 BTRFS_TREE_LOG_OBJECTID, NULL,
5581a51a 1369 0, 0, 0);
7237f183
YZ		 1370 if (IS_ERR(leaf)) {
1371 kfree(root);
1372 return ERR_CAST(leaf);
1373 }
e02119d5 1374
5d4f98a2
YZ		 1375 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1376 btrfs_set_header_bytenr(leaf, leaf->start);
1377 btrfs_set_header_generation(leaf, trans->transid);
1378 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1379 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1380 root->node = leaf;
e02119d5
CM		 1381
1382 write_extent_buffer(root->node, root->fs_info->fsid,
1383 (unsigned long)btrfs_header_fsid(root->node),
1384 BTRFS_FSID_SIZE);
1385 btrfs_mark_buffer_dirty(root->node);
1386 btrfs_tree_unlock(root->node);
7237f183
YZ		 1387 return root;
1388}
1389
1390int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1391 struct btrfs_fs_info *fs_info)
1392{
1393 struct btrfs_root *log_root;
1394
1395 log_root = alloc_log_tree(trans, fs_info);
1396 if (IS_ERR(log_root))
1397 return PTR_ERR(log_root);
1398 WARN_ON(fs_info->log_root_tree);
1399 fs_info->log_root_tree = log_root;
1400 return 0;
1401}
1402
1403int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1404 struct btrfs_root *root)
1405{
1406 struct btrfs_root *log_root;
1407 struct btrfs_inode_item *inode_item;
1408
1409 log_root = alloc_log_tree(trans, root->fs_info);
1410 if (IS_ERR(log_root))
1411 return PTR_ERR(log_root);
1412
1413 log_root->last_trans = trans->transid;
1414 log_root->root_key.offset = root->root_key.objectid;
1415
1416 inode_item = &log_root->root_item.inode;
3cae210f
QW		 1417 btrfs_set_stack_inode_generation(inode_item, 1);
1418 btrfs_set_stack_inode_size(inode_item, 3);
1419 btrfs_set_stack_inode_nlink(inode_item, 1);
1420 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
1421 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
7237f183 1422
5d4f98a2 1423 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1424
1425 WARN_ON(root->log_root);
1426 root->log_root = log_root;
1427 root->log_transid = 0;
257c62e1 1428 root->last_log_commit = 0;
e02119d5
CM		 1429 return 0;
1430}
1431
35a3621b
SB		 1432static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1433 struct btrfs_key *key)
e02119d5
CM		 1434{
1435 struct btrfs_root *root;
1436 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1437 struct btrfs_path *path;
84234f3a 1438 u64 generation;
db94535d 1439 u32 blocksize;
cb517eab 1440 int ret;
0f7d52f4 1441
cb517eab
MX		 1442 path = btrfs_alloc_path();
1443 if (!path)
0f7d52f4 1444 return ERR_PTR(-ENOMEM);
cb517eab
MX		 1445
1446 root = btrfs_alloc_root(fs_info);
1447 if (!root) {
1448 ret = -ENOMEM;
1449 goto alloc_fail;
0f7d52f4
CM		 1450 }
1451
db94535d 1452 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb 1453 tree_root->sectorsize, tree_root->stripesize,
cb517eab 1454 root, fs_info, key->objectid);
0f7d52f4 1455
cb517eab
MX		 1456 ret = btrfs_find_root(tree_root, key, path,
1457 &root->root_item, &root->root_key);
0f7d52f4 1458 if (ret) {
13a8a7c8
YZ		 1459 if (ret > 0)
1460 ret = -ENOENT;
cb517eab 1461 goto find_fail;
0f7d52f4 1462 }
13a8a7c8 1463
84234f3a 1464 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1465 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1466 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1467 blocksize, generation);
cb517eab
MX		 1468 if (!root->node) {
1469 ret = -ENOMEM;
1470 goto find_fail;
1471 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1472 ret = -EIO;
1473 goto read_fail;
416bc658 1474 }
5d4f98a2 1475 root->commit_root = btrfs_root_node(root);
13a8a7c8 1476out:
cb517eab
MX		 1477 btrfs_free_path(path);
1478 return root;
1479
1480read_fail:
1481 free_extent_buffer(root->node);
1482find_fail:
1483 kfree(root);
1484alloc_fail:
1485 root = ERR_PTR(ret);
1486 goto out;
1487}
1488
1489struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1490 struct btrfs_key *location)
1491{
1492 struct btrfs_root *root;
1493
1494 root = btrfs_read_tree_root(tree_root, location);
1495 if (IS_ERR(root))
1496 return root;
1497
1498 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
e02119d5 1499 root->ref_cows = 1;
08fe4db1
LZ		 1500 btrfs_check_and_init_root_item(&root->root_item);
1501 }
13a8a7c8 1502
5eda7b5e
CM		 1503 return root;
1504}
1505
cb517eab
MX		 1506int btrfs_init_fs_root(struct btrfs_root *root)
1507{
1508 int ret;
1509
1510 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1511 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1512 GFP_NOFS);
1513 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1514 ret = -ENOMEM;
1515 goto fail;
1516 }
1517
1518 btrfs_init_free_ino_ctl(root);
1519 mutex_init(&root->fs_commit_mutex);
1520 spin_lock_init(&root->cache_lock);
1521 init_waitqueue_head(&root->cache_wait);
1522
1523 ret = get_anon_bdev(&root->anon_dev);
1524 if (ret)
1525 goto fail;
1526 return 0;
1527fail:
1528 kfree(root->free_ino_ctl);
1529 kfree(root->free_ino_pinned);
1530 return ret;
1531}
1532
171170c1
ST		 1533static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1534 u64 root_id)
cb517eab
MX		 1535{
1536 struct btrfs_root *root;
1537
1538 spin_lock(&fs_info->fs_roots_radix_lock);
1539 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1540 (unsigned long)root_id);
1541 spin_unlock(&fs_info->fs_roots_radix_lock);
1542 return root;
1543}
1544
1545int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1546 struct btrfs_root *root)
1547{
1548 int ret;
1549
1550 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1551 if (ret)
1552 return ret;
1553
1554 spin_lock(&fs_info->fs_roots_radix_lock);
1555 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1556 (unsigned long)root->root_key.objectid,
1557 root);
1558 if (ret == 0)
1559 root->in_radix = 1;
1560 spin_unlock(&fs_info->fs_roots_radix_lock);
1561 radix_tree_preload_end();
1562
1563 return ret;
1564}
1565
edbd8d4e
CM		 1566struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
1567 struct btrfs_key *location)
5eda7b5e
CM		 1568{
1569 struct btrfs_root *root;
1570 int ret;
1571
edbd8d4e
CM		 1572 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1573 return fs_info->tree_root;
1574 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1575 return fs_info->extent_root;
8f18cf13
CM		 1576 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1577 return fs_info->chunk_root;
1578 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1579 return fs_info->dev_root;
0403e47e
YZ		 1580 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1581 return fs_info->csum_root;
bcef60f2
AJ		 1582 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1583 return fs_info->quota_root ? fs_info->quota_root :
1584 ERR_PTR(-ENOENT);
f7a81ea4
SB		 1585 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1586 return fs_info->uuid_root ? fs_info->uuid_root :
1587 ERR_PTR(-ENOENT);
4df27c4d 1588again:
cb517eab 1589 root = btrfs_lookup_fs_root(fs_info, location->objectid);
5eda7b5e
CM		 1590 if (root)
1591 return root;
1592
cb517eab 1593 root = btrfs_read_fs_root(fs_info->tree_root, location);
5eda7b5e
CM		 1594 if (IS_ERR(root))
1595 return root;
3394e160 1596
cb517eab
MX		 1597 if (btrfs_root_refs(&root->root_item) == 0) {
1598 ret = -ENOENT;
581bb050 1599 goto fail;
35a30d7c 1600 }
581bb050 1601
cb517eab 1602 ret = btrfs_init_fs_root(root);
ac08aedf
CM		 1603 if (ret)
1604 goto fail;
3394e160 1605
d68fc57b
YZ		 1606 ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
1607 if (ret < 0)
1608 goto fail;
1609 if (ret == 0)
1610 root->orphan_item_inserted = 1;
1611
cb517eab 1612 ret = btrfs_insert_fs_root(fs_info, root);
0f7d52f4 1613 if (ret) {
4df27c4d
YZ		 1614 if (ret == -EEXIST) {
1615 free_fs_root(root);
1616 goto again;
1617 }
1618 goto fail;
0f7d52f4 1619 }
edbd8d4e 1620 return root;
4df27c4d
YZ		 1621fail:
1622 free_fs_root(root);
1623 return ERR_PTR(ret);
edbd8d4e
CM		 1624}
1625
04160088
CM		 1626static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1627{
1628 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1629 int ret = 0;
04160088
CM		 1630 struct btrfs_device *device;
1631 struct backing_dev_info *bdi;
b7967db7 1632
1f78160c
XG		 1633 rcu_read_lock();
1634 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1635 if (!device->bdev)
1636 continue;
04160088
CM		 1637 bdi = blk_get_backing_dev_info(device->bdev);
1638 if (bdi && bdi_congested(bdi, bdi_bits)) {
1639 ret = 1;
1640 break;
1641 }
1642 }
1f78160c 1643 rcu_read_unlock();
04160088
CM		 1644 return ret;
1645}
1646
ad081f14
JA		 1647/*
1648 * If this fails, caller must call bdi_destroy() to get rid of the
1649 * bdi again.
1650 */
04160088
CM		 1651static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1652{
ad081f14
JA		 1653 int err;
1654
1655 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1656 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1657 if (err)
1658 return err;
1659
4575c9cc 1660 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1661 bdi->congested_fn = btrfs_congested_fn;
1662 bdi->congested_data = info;
1663 return 0;
1664}
1665
8b712842
CM		 1666/*
1667 * called by the kthread helper functions to finally call the bio end_io
1668 * functions. This is where read checksum verification actually happens
1669 */
1670static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1671{
ce9adaa5 1672 struct bio *bio;
8b712842
CM		 1673 struct end_io_wq *end_io_wq;
1674 struct btrfs_fs_info *fs_info;
ce9adaa5 1675 int error;
ce9adaa5 1676
8b712842
CM		 1677 end_io_wq = container_of(work, struct end_io_wq, work);
1678 bio = end_io_wq->bio;
1679 fs_info = end_io_wq->info;
ce9adaa5 1680
8b712842
CM		 1681 error = end_io_wq->error;
1682 bio->bi_private = end_io_wq->private;
1683 bio->bi_end_io = end_io_wq->end_io;
1684 kfree(end_io_wq);
8b712842 1685 bio_endio(bio, error);
44b8bd7e
CM		 1686}
1687
a74a4b97
CM		 1688static int cleaner_kthread(void *arg)
1689{
1690 struct btrfs_root *root = arg;
d0278245 1691 int again;
a74a4b97
CM		 1692
1693 do {
d0278245 1694 again = 0;
a74a4b97 1695
d0278245 1696 /* Make the cleaner go to sleep early. */
babbf170 1697 if (btrfs_need_cleaner_sleep(root))
d0278245
MX		 1698 goto sleep;
1699
1700 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1701 goto sleep;
1702
dc7f370c
MX		 1703 /*
1704 * Avoid the problem that we change the status of the fs
1705 * during the above check and trylock.
1706 */
babbf170 1707 if (btrfs_need_cleaner_sleep(root)) {
dc7f370c
MX		 1708 mutex_unlock(&root->fs_info->cleaner_mutex);
1709 goto sleep;
76dda93c 1710 }
a74a4b97 1711
d0278245
MX		 1712 btrfs_run_delayed_iputs(root);
1713 again = btrfs_clean_one_deleted_snapshot(root);
1714 mutex_unlock(&root->fs_info->cleaner_mutex);
1715
1716 /*
05323cd1
MX		 1717 * The defragger has dealt with the R/O remount and umount,
1718 * needn't do anything special here.
d0278245
MX		 1719 */
1720 btrfs_run_defrag_inodes(root->fs_info);
1721sleep:
9d1a2a3a 1722 if (!try_to_freeze() && !again) {
a74a4b97 1723 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1724 if (!kthread_should_stop())
1725 schedule();
a74a4b97
CM		 1726 __set_current_state(TASK_RUNNING);
1727 }
1728 } while (!kthread_should_stop());
1729 return 0;
1730}
1731
1732static int transaction_kthread(void *arg)
1733{
1734 struct btrfs_root *root = arg;
1735 struct btrfs_trans_handle *trans;
1736 struct btrfs_transaction *cur;
8929ecfa 1737 u64 transid;
a74a4b97
CM		 1738 unsigned long now;
1739 unsigned long delay;
914b2007 1740 bool cannot_commit;
a74a4b97
CM		 1741
1742 do {
914b2007 1743 cannot_commit = false;
8b87dc17 1744 delay = HZ * root->fs_info->commit_interval;
a74a4b97
CM		 1745 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1746
a4abeea4 1747 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1748 cur = root->fs_info->running_transaction;
1749 if (!cur) {
a4abeea4 1750 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1751 goto sleep;
1752 }
31153d81 1753
a74a4b97 1754 now = get_seconds();
4a9d8bde 1755 if (cur->state < TRANS_STATE_BLOCKED &&
8b87dc17
DS		 1756 (now < cur->start_time ||
1757 now - cur->start_time < root->fs_info->commit_interval)) {
a4abeea4 1758 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1759 delay = HZ * 5;
1760 goto sleep;
1761 }
8929ecfa 1762 transid = cur->transid;
a4abeea4 1763 spin_unlock(&root->fs_info->trans_lock);
56bec294 1764
79787eaa 1765 /* If the file system is aborted, this will always fail. */
354aa0fb 1766 trans = btrfs_attach_transaction(root);
914b2007 1767 if (IS_ERR(trans)) {
354aa0fb
MX		 1768 if (PTR_ERR(trans) != -ENOENT)
1769 cannot_commit = true;
79787eaa 1770 goto sleep;
914b2007 1771 }
8929ecfa 1772 if (transid == trans->transid) {
79787eaa 1773 btrfs_commit_transaction(trans, root);
8929ecfa
YZ		 1774 } else {
1775 btrfs_end_transaction(trans, root);
1776 }
a74a4b97
CM		 1777sleep:
1778 wake_up_process(root->fs_info->cleaner_kthread);
1779 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1780
a0acae0e 1781 if (!try_to_freeze()) {
a74a4b97 1782 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa 1783 if (!kthread_should_stop() &&
914b2007
JK		 1784 (!btrfs_transaction_blocked(root->fs_info) ||
1785 cannot_commit))
8929ecfa 1786 schedule_timeout(delay);
a74a4b97
CM		 1787 __set_current_state(TASK_RUNNING);
1788 }
1789 } while (!kthread_should_stop());
1790 return 0;
1791}
1792
af31f5e5
CM		 1793/*
1794 * this will find the highest generation in the array of
1795 * root backups. The index of the highest array is returned,
1796 * or -1 if we can't find anything.
1797 *
1798 * We check to make sure the array is valid by comparing the
1799 * generation of the latest root in the array with the generation
1800 * in the super block. If they don't match we pitch it.
1801 */
1802static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1803{
1804 u64 cur;
1805 int newest_index = -1;
1806 struct btrfs_root_backup *root_backup;
1807 int i;
1808
1809 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1810 root_backup = info->super_copy->super_roots + i;
1811 cur = btrfs_backup_tree_root_gen(root_backup);
1812 if (cur == newest_gen)
1813 newest_index = i;
1814 }
1815
1816 /* check to see if we actually wrapped around */
1817 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1818 root_backup = info->super_copy->super_roots;
1819 cur = btrfs_backup_tree_root_gen(root_backup);
1820 if (cur == newest_gen)
1821 newest_index = 0;
1822 }
1823 return newest_index;
1824}
1825
1826
1827/*
1828 * find the oldest backup so we know where to store new entries
1829 * in the backup array. This will set the backup_root_index
1830 * field in the fs_info struct
1831 */
1832static void find_oldest_super_backup(struct btrfs_fs_info *info,
1833 u64 newest_gen)
1834{
1835 int newest_index = -1;
1836
1837 newest_index = find_newest_super_backup(info, newest_gen);
1838 /* if there was garbage in there, just move along */
1839 if (newest_index == -1) {
1840 info->backup_root_index = 0;
1841 } else {
1842 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1843 }
1844}
1845
1846/*
1847 * copy all the root pointers into the super backup array.
1848 * this will bump the backup pointer by one when it is
1849 * done
1850 */
1851static void backup_super_roots(struct btrfs_fs_info *info)
1852{
1853 int next_backup;
1854 struct btrfs_root_backup *root_backup;
1855 int last_backup;
1856
1857 next_backup = info->backup_root_index;
1858 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1859 BTRFS_NUM_BACKUP_ROOTS;
1860
1861 /*
1862 * just overwrite the last backup if we're at the same generation
1863 * this happens only at umount
1864 */
1865 root_backup = info->super_for_commit->super_roots + last_backup;
1866 if (btrfs_backup_tree_root_gen(root_backup) ==
1867 btrfs_header_generation(info->tree_root->node))
1868 next_backup = last_backup;
1869
1870 root_backup = info->super_for_commit->super_roots + next_backup;
1871
1872 /*
1873 * make sure all of our padding and empty slots get zero filled
1874 * regardless of which ones we use today
1875 */
1876 memset(root_backup, 0, sizeof(*root_backup));
1877
1878 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1879
1880 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1881 btrfs_set_backup_tree_root_gen(root_backup,
1882 btrfs_header_generation(info->tree_root->node));
1883
1884 btrfs_set_backup_tree_root_level(root_backup,
1885 btrfs_header_level(info->tree_root->node));
1886
1887 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1888 btrfs_set_backup_chunk_root_gen(root_backup,
1889 btrfs_header_generation(info->chunk_root->node));
1890 btrfs_set_backup_chunk_root_level(root_backup,
1891 btrfs_header_level(info->chunk_root->node));
1892
1893 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1894 btrfs_set_backup_extent_root_gen(root_backup,
1895 btrfs_header_generation(info->extent_root->node));
1896 btrfs_set_backup_extent_root_level(root_backup,
1897 btrfs_header_level(info->extent_root->node));
1898
7c7e82a7
CM		 1899 /*
1900 * we might commit during log recovery, which happens before we set
1901 * the fs_root. Make sure it is valid before we fill it in.
1902 */
1903 if (info->fs_root && info->fs_root->node) {
1904 btrfs_set_backup_fs_root(root_backup,
1905 info->fs_root->node->start);
1906 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 1907 btrfs_header_generation(info->fs_root->node));
7c7e82a7 1908 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 1909 btrfs_header_level(info->fs_root->node));
7c7e82a7 1910 }
af31f5e5
CM		 1911
1912 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1913 btrfs_set_backup_dev_root_gen(root_backup,
1914 btrfs_header_generation(info->dev_root->node));
1915 btrfs_set_backup_dev_root_level(root_backup,
1916 btrfs_header_level(info->dev_root->node));
1917
1918 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1919 btrfs_set_backup_csum_root_gen(root_backup,
1920 btrfs_header_generation(info->csum_root->node));
1921 btrfs_set_backup_csum_root_level(root_backup,
1922 btrfs_header_level(info->csum_root->node));
1923
1924 btrfs_set_backup_total_bytes(root_backup,
1925 btrfs_super_total_bytes(info->super_copy));
1926 btrfs_set_backup_bytes_used(root_backup,
1927 btrfs_super_bytes_used(info->super_copy));
1928 btrfs_set_backup_num_devices(root_backup,
1929 btrfs_super_num_devices(info->super_copy));
1930
1931 /*
1932 * if we don't copy this out to the super_copy, it won't get remembered
1933 * for the next commit
1934 */
1935 memcpy(&info->super_copy->super_roots,
1936 &info->super_for_commit->super_roots,
1937 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1938}
1939
1940/*
1941 * this copies info out of the root backup array and back into
1942 * the in-memory super block. It is meant to help iterate through
1943 * the array, so you send it the number of backups you've already
1944 * tried and the last backup index you used.
1945 *
1946 * this returns -1 when it has tried all the backups
1947 */
1948static noinline int next_root_backup(struct btrfs_fs_info *info,
1949 struct btrfs_super_block *super,
1950 int *num_backups_tried, int *backup_index)
1951{
1952 struct btrfs_root_backup *root_backup;
1953 int newest = *backup_index;
1954
1955 if (*num_backups_tried == 0) {
1956 u64 gen = btrfs_super_generation(super);
1957
1958 newest = find_newest_super_backup(info, gen);
1959 if (newest == -1)
1960 return -1;
1961
1962 *backup_index = newest;
1963 *num_backups_tried = 1;
1964 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
1965 /* we've tried all the backups, all done */
1966 return -1;
1967 } else {
1968 /* jump to the next oldest backup */
1969 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
1970 BTRFS_NUM_BACKUP_ROOTS;
1971 *backup_index = newest;
1972 *num_backups_tried += 1;
1973 }
1974 root_backup = super->super_roots + newest;
1975
1976 btrfs_set_super_generation(super,
1977 btrfs_backup_tree_root_gen(root_backup));
1978 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
1979 btrfs_set_super_root_level(super,
1980 btrfs_backup_tree_root_level(root_backup));
1981 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
1982
1983 /*
1984 * fixme: the total bytes and num_devices need to match or we should
1985 * need a fsck
1986 */
1987 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
1988 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
1989 return 0;
1990}
1991
7abadb64
LB		 1992/* helper to cleanup workers */
1993static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
1994{
1995 btrfs_stop_workers(&fs_info->generic_worker);
1996 btrfs_stop_workers(&fs_info->fixup_workers);
1997 btrfs_stop_workers(&fs_info->delalloc_workers);
1998 btrfs_stop_workers(&fs_info->workers);
1999 btrfs_stop_workers(&fs_info->endio_workers);
2000 btrfs_stop_workers(&fs_info->endio_meta_workers);
2001 btrfs_stop_workers(&fs_info->endio_raid56_workers);
2002 btrfs_stop_workers(&fs_info->rmw_workers);
2003 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
2004 btrfs_stop_workers(&fs_info->endio_write_workers);
2005 btrfs_stop_workers(&fs_info->endio_freespace_worker);
2006 btrfs_stop_workers(&fs_info->submit_workers);
2007 btrfs_stop_workers(&fs_info->delayed_workers);
2008 btrfs_stop_workers(&fs_info->caching_workers);
2009 btrfs_stop_workers(&fs_info->readahead_workers);
2010 btrfs_stop_workers(&fs_info->flush_workers);
2f232036 2011 btrfs_stop_workers(&fs_info->qgroup_rescan_workers);
7abadb64
LB		 2012}
2013
af31f5e5
CM		 2014/* helper to cleanup tree roots */
2015static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2016{
2017 free_extent_buffer(info->tree_root->node);
2018 free_extent_buffer(info->tree_root->commit_root);
af31f5e5
CM		 2019 info->tree_root->node = NULL;
2020 info->tree_root->commit_root = NULL;
655b09fe
JB		 2021
2022 if (info->dev_root) {
2023 free_extent_buffer(info->dev_root->node);
2024 free_extent_buffer(info->dev_root->commit_root);
2025 info->dev_root->node = NULL;
2026 info->dev_root->commit_root = NULL;
2027 }
2028 if (info->extent_root) {
2029 free_extent_buffer(info->extent_root->node);
2030 free_extent_buffer(info->extent_root->commit_root);
2031 info->extent_root->node = NULL;
2032 info->extent_root->commit_root = NULL;
2033 }
2034 if (info->csum_root) {
2035 free_extent_buffer(info->csum_root->node);
2036 free_extent_buffer(info->csum_root->commit_root);
2037 info->csum_root->node = NULL;
2038 info->csum_root->commit_root = NULL;
2039 }
bcef60f2 2040 if (info->quota_root) {
655b09fe
JB		 2041 free_extent_buffer(info->quota_root->node);
2042 free_extent_buffer(info->quota_root->commit_root);
bcef60f2
AJ		 2043 info->quota_root->node = NULL;
2044 info->quota_root->commit_root = NULL;
2045 }
f7a81ea4
SB		 2046 if (info->uuid_root) {
2047 free_extent_buffer(info->uuid_root->node);
2048 free_extent_buffer(info->uuid_root->commit_root);
2049 info->uuid_root->node = NULL;
2050 info->uuid_root->commit_root = NULL;
2051 }
af31f5e5
CM		 2052 if (chunk_root) {
2053 free_extent_buffer(info->chunk_root->node);
2054 free_extent_buffer(info->chunk_root->commit_root);
2055 info->chunk_root->node = NULL;
2056 info->chunk_root->commit_root = NULL;
2057 }
2058}
2059
171f6537
JB		 2060static void del_fs_roots(struct btrfs_fs_info *fs_info)
2061{
2062 int ret;
2063 struct btrfs_root *gang[8];
2064 int i;
2065
2066 while (!list_empty(&fs_info->dead_roots)) {
2067 gang[0] = list_entry(fs_info->dead_roots.next,
2068 struct btrfs_root, root_list);
2069 list_del(&gang[0]->root_list);
2070
2071 if (gang[0]->in_radix) {
cb517eab 2072 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
171f6537
JB		 2073 } else {
2074 free_extent_buffer(gang[0]->node);
2075 free_extent_buffer(gang[0]->commit_root);
b0feb9d9 2076 btrfs_put_fs_root(gang[0]);
171f6537
JB		 2077 }
2078 }
2079
2080 while (1) {
2081 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2082 (void **)gang, 0,
2083 ARRAY_SIZE(gang));
2084 if (!ret)
2085 break;
2086 for (i = 0; i < ret; i++)
cb517eab 2087 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
171f6537
JB		 2088 }
2089}
af31f5e5 2090
ad2b2c80
AV		 2091int open_ctree(struct super_block *sb,
2092 struct btrfs_fs_devices *fs_devices,
2093 char *options)
2e635a27 2094{
db94535d
CM		 2095 u32 sectorsize;
2096 u32 nodesize;
2097 u32 leafsize;
2098 u32 blocksize;
87ee04eb 2099 u32 stripesize;
84234f3a 2100 u64 generation;
f2b636e8 2101 u64 features;
3de4586c 2102 struct btrfs_key location;
a061fc8d 2103 struct buffer_head *bh;
4d34b278 2104 struct btrfs_super_block *disk_super;
815745cf 2105 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2106 struct btrfs_root *tree_root;
4d34b278
ID		 2107 struct btrfs_root *extent_root;
2108 struct btrfs_root *csum_root;
2109 struct btrfs_root *chunk_root;
2110 struct btrfs_root *dev_root;
bcef60f2 2111 struct btrfs_root *quota_root;
f7a81ea4 2112 struct btrfs_root *uuid_root;
e02119d5 2113 struct btrfs_root *log_tree_root;
eb60ceac 2114 int ret;
e58ca020 2115 int err = -EINVAL;
af31f5e5
CM		 2116 int num_backups_tried = 0;
2117 int backup_index = 0;
70f80175
SB		 2118 bool create_uuid_tree;
2119 bool check_uuid_tree;
4543df7e 2120
f84a8bd6 2121 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
6f07e42e 2122 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
cb517eab 2123 if (!tree_root || !chunk_root) {
39279cc3
CM		 2124 err = -ENOMEM;
2125 goto fail;
2126 }
76dda93c
YZ		 2127
2128 ret = init_srcu_struct(&fs_info->subvol_srcu);
2129 if (ret) {
2130 err = ret;
2131 goto fail;
2132 }
2133
2134 ret = setup_bdi(fs_info, &fs_info->bdi);
2135 if (ret) {
2136 err = ret;
2137 goto fail_srcu;
2138 }
2139
e2d84521
MX		 2140 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2141 if (ret) {
2142 err = ret;
2143 goto fail_bdi;
2144 }
2145 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2146 (1 + ilog2(nr_cpu_ids));
2147
963d678b
MX		 2148 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2149 if (ret) {
2150 err = ret;
2151 goto fail_dirty_metadata_bytes;
2152 }
2153
76dda93c
YZ		 2154 fs_info->btree_inode = new_inode(sb);
2155 if (!fs_info->btree_inode) {
2156 err = -ENOMEM;
963d678b 2157 goto fail_delalloc_bytes;
76dda93c
YZ		 2158 }
2159
a6591715 2160 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2161
76dda93c 2162 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
8fd17795 2163 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2164 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2165 INIT_LIST_HEAD(&fs_info->delayed_iputs);
eb73c1b7 2166 INIT_LIST_HEAD(&fs_info->delalloc_roots);
11833d66 2167 INIT_LIST_HEAD(&fs_info->caching_block_groups);
eb73c1b7 2168 spin_lock_init(&fs_info->delalloc_root_lock);
a4abeea4 2169 spin_lock_init(&fs_info->trans_lock);
76dda93c 2170 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2171 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2172 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 2173 spin_lock_init(&fs_info->free_chunk_lock);
f29021b2 2174 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2175 spin_lock_init(&fs_info->super_lock);
f29021b2 2176 rwlock_init(&fs_info->tree_mod_log_lock);
7585717f 2177 mutex_init(&fs_info->reloc_mutex);
de98ced9 2178 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2179
58176a96 2180 init_completion(&fs_info->kobj_unregister);
0b86a832 2181 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2182 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2183 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
0b86a832 2184 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX		 2185 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2186 BTRFS_BLOCK_RSV_GLOBAL);
2187 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2188 BTRFS_BLOCK_RSV_DELALLOC);
2189 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2190 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2191 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2192 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2193 BTRFS_BLOCK_RSV_DELOPS);
cb03c743 2194 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 2195 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 2196 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 2197 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 2198 atomic_set(&fs_info->defrag_running, 0);
fc36ed7e 2199 atomic64_set(&fs_info->tree_mod_seq, 0);
e20d96d6 2200 fs_info->sb = sb;
6f568d35 2201 fs_info->max_inline = 8192 * 1024;
9ed74f2d 2202 fs_info->metadata_ratio = 0;
4cb5300b 2203 fs_info->defrag_inodes = RB_ROOT;
2bf64758 2204 fs_info->free_chunk_space = 0;
f29021b2 2205 fs_info->tree_mod_log = RB_ROOT;
8b87dc17 2206 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
c8b97818 2207
90519d66
AJ		 2208 /* readahead state */
2209 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2210 spin_lock_init(&fs_info->reada_lock);
c8b97818 2211
b34b086c
CM		 2212 fs_info->thread_pool_size = min_t(unsigned long,
2213 num_online_cpus() + 2, 8);
0afbaf8c 2214
199c2a9c
MX		 2215 INIT_LIST_HEAD(&fs_info->ordered_roots);
2216 spin_lock_init(&fs_info->ordered_root_lock);
16cdcec7
MX		 2217 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2218 GFP_NOFS);
2219 if (!fs_info->delayed_root) {
2220 err = -ENOMEM;
2221 goto fail_iput;
2222 }
2223 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2224
a2de733c
AJ		 2225 mutex_init(&fs_info->scrub_lock);
2226 atomic_set(&fs_info->scrubs_running, 0);
2227 atomic_set(&fs_info->scrub_pause_req, 0);
2228 atomic_set(&fs_info->scrubs_paused, 0);
2229 atomic_set(&fs_info->scrub_cancel_req, 0);
2230 init_waitqueue_head(&fs_info->scrub_pause_wait);
2231 init_rwsem(&fs_info->scrub_super_lock);
2232 fs_info->scrub_workers_refcnt = 0;
21adbd5c
SB		 2233#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2234 fs_info->check_integrity_print_mask = 0;
2235#endif
a2de733c 2236
c9e9f97b
ID		 2237 spin_lock_init(&fs_info->balance_lock);
2238 mutex_init(&fs_info->balance_mutex);
837d5b6e
ID		 2239 atomic_set(&fs_info->balance_running, 0);
2240 atomic_set(&fs_info->balance_pause_req, 0);
a7e99c69 2241 atomic_set(&fs_info->balance_cancel_req, 0);
c9e9f97b 2242 fs_info->balance_ctl = NULL;
837d5b6e 2243 init_waitqueue_head(&fs_info->balance_wait_q);
a2de733c 2244
a061fc8d
CM		 2245 sb->s_blocksize = 4096;
2246 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 2247 sb->s_bdi = &fs_info->bdi;
a061fc8d 2248
76dda93c 2249 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
bfe86848 2250 set_nlink(fs_info->btree_inode, 1);
0afbaf8c
CM		 2251 /*
2252 * we set the i_size on the btree inode to the max possible int.
2253 * the real end of the address space is determined by all of
2254 * the devices in the system
2255 */
2256 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 2257 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 2258 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2259
5d4f98a2 2260 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 2261 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 2262 fs_info->btree_inode->i_mapping);
0b32f4bb 2263 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
a8067e02 2264 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM		 2265
2266 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 2267
76dda93c
YZ		 2268 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2269 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2270 sizeof(struct btrfs_key));
72ac3c0d
JB		 2271 set_bit(BTRFS_INODE_DUMMY,
2272 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
c65ddb52 2273 insert_inode_hash(fs_info->btree_inode);
76dda93c 2274
0f9dd46c 2275 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2276 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2277 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2278
11833d66 2279 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 2280 fs_info->btree_inode->i_mapping);
11833d66 2281 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 2282 fs_info->btree_inode->i_mapping);
11833d66 2283 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 2284 fs_info->do_barriers = 1;
e18e4809 2285
39279cc3 2286
5a3f23d5 2287 mutex_init(&fs_info->ordered_operations_mutex);
9ffba8cd 2288 mutex_init(&fs_info->ordered_extent_flush_mutex);
e02119d5 2289 mutex_init(&fs_info->tree_log_mutex);
925baedd 2290 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 2291 mutex_init(&fs_info->transaction_kthread_mutex);
2292 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2293 mutex_init(&fs_info->volume_mutex);
276e680d 2294 init_rwsem(&fs_info->extent_commit_sem);
c71bf099 2295 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2296 init_rwsem(&fs_info->subvol_sem);
803b2f54 2297 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
e922e087
SB		 2298 fs_info->dev_replace.lock_owner = 0;
2299 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2300 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2301 mutex_init(&fs_info->dev_replace.lock_management_lock);
2302 mutex_init(&fs_info->dev_replace.lock);
fa9c0d79 2303
416ac51d 2304 spin_lock_init(&fs_info->qgroup_lock);
f2f6ed3d 2305 mutex_init(&fs_info->qgroup_ioctl_lock);
416ac51d
AJ		 2306 fs_info->qgroup_tree = RB_ROOT;
2307 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2308 fs_info->qgroup_seq = 1;
2309 fs_info->quota_enabled = 0;
2310 fs_info->pending_quota_state = 0;
1e8f9158 2311 fs_info->qgroup_ulist = NULL;
2f232036 2312 mutex_init(&fs_info->qgroup_rescan_lock);
416ac51d 2313
fa9c0d79
CM		 2314 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2315 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2316
e6dcd2dc 2317 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2318 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2319 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2320 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2321
53b381b3
DW		 2322 ret = btrfs_alloc_stripe_hash_table(fs_info);
2323 if (ret) {
83c8266a 2324 err = ret;
53b381b3
DW		 2325 goto fail_alloc;
2326 }
2327
0b86a832 2328 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 2329 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2330
3c4bb26b 2331 invalidate_bdev(fs_devices->latest_bdev);
1104a885
DS		 2332
2333 /*
2334 * Read super block and check the signature bytes only
2335 */
a512bbf8 2336 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY		 2337 if (!bh) {
2338 err = -EINVAL;
16cdcec7 2339 goto fail_alloc;
20b45077 2340 }
39279cc3 2341
1104a885
DS		 2342 /*
2343 * We want to check superblock checksum, the type is stored inside.
2344 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2345 */
2346 if (btrfs_check_super_csum(bh->b_data)) {
2347 printk(KERN_ERR "btrfs: superblock checksum mismatch\n");
2348 err = -EINVAL;
2349 goto fail_alloc;
2350 }
2351
2352 /*
2353 * super_copy is zeroed at allocation time and we never touch the
2354 * following bytes up to INFO_SIZE, the checksum is calculated from
2355 * the whole block of INFO_SIZE
2356 */
6c41761f
DS		 2357 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2358 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2359 sizeof(*fs_info->super_for_commit));
a061fc8d 2360 brelse(bh);
5f39d397 2361
6c41761f 2362 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2363
1104a885
DS		 2364 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2365 if (ret) {
2366 printk(KERN_ERR "btrfs: superblock contains fatal errors\n");
2367 err = -EINVAL;
2368 goto fail_alloc;
2369 }
2370
6c41761f 2371 disk_super = fs_info->super_copy;
0f7d52f4 2372 if (!btrfs_super_root(disk_super))
16cdcec7 2373 goto fail_alloc;
0f7d52f4 2374
acce952b 2375 /* check FS state, whether FS is broken. */
87533c47
MX		 2376 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2377 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2378
af31f5e5
CM		 2379 /*
2380 * run through our array of backup supers and setup
2381 * our ring pointer to the oldest one
2382 */
2383 generation = btrfs_super_generation(disk_super);
2384 find_oldest_super_backup(fs_info, generation);
2385
75e7cb7f
LB		 2386 /*
2387 * In the long term, we'll store the compression type in the super
2388 * block, and it'll be used for per file compression control.
2389 */
2390 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2391
2b82032c
YZ		 2392 ret = btrfs_parse_options(tree_root, options);
2393 if (ret) {
2394 err = ret;
16cdcec7 2395 goto fail_alloc;
2b82032c 2396 }
dfe25020 2397
f2b636e8
JB		 2398 features = btrfs_super_incompat_flags(disk_super) &
2399 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2400 if (features) {
2401 printk(KERN_ERR "BTRFS: couldn't mount because of "
2402 "unsupported optional features (%Lx).\n",
c1c9ff7c 2403 features);
f2b636e8 2404 err = -EINVAL;
16cdcec7 2405 goto fail_alloc;
f2b636e8
JB		 2406 }
2407
727011e0
CM		 2408 if (btrfs_super_leafsize(disk_super) !=
2409 btrfs_super_nodesize(disk_super)) {
2410 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2411 "blocksizes don't match. node %d leaf %d\n",
2412 btrfs_super_nodesize(disk_super),
2413 btrfs_super_leafsize(disk_super));
2414 err = -EINVAL;
2415 goto fail_alloc;
2416 }
2417 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2418 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2419 "blocksize (%d) was too large\n",
2420 btrfs_super_leafsize(disk_super));
2421 err = -EINVAL;
2422 goto fail_alloc;
2423 }
2424
5d4f98a2 2425 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2426 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
69e380d1 2427 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2428 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
727011e0 2429
3173a18f
JB		 2430 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
2431 printk(KERN_ERR "btrfs: has skinny extents\n");
2432
727011e0
CM		 2433 /*
2434 * flag our filesystem as having big metadata blocks if
2435 * they are bigger than the page size
2436 */
2437 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2438 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
2439 printk(KERN_INFO "btrfs flagging fs with big metadata feature\n");
2440 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2441 }
2442
bc3f116f
CM		 2443 nodesize = btrfs_super_nodesize(disk_super);
2444 leafsize = btrfs_super_leafsize(disk_super);
2445 sectorsize = btrfs_super_sectorsize(disk_super);
2446 stripesize = btrfs_super_stripesize(disk_super);
e2d84521 2447 fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids));
963d678b 2448 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f
CM		 2449
2450 /*
2451 * mixed block groups end up with duplicate but slightly offset
2452 * extent buffers for the same range. It leads to corruptions
2453 */
2454 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2455 (sectorsize != leafsize)) {
2456 printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes "
2457 "are not allowed for mixed block groups on %s\n",
2458 sb->s_id);
2459 goto fail_alloc;
2460 }
2461
ceda0864
MX		 2462 /*
2463 * Needn't use the lock because there is no other task which will
2464 * update the flag.
2465 */
a6fa6fae 2466 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2467
f2b636e8
JB		 2468 features = btrfs_super_compat_ro_flags(disk_super) &
2469 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2470 if (!(sb->s_flags & MS_RDONLY) && features) {
2471 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2472 "unsupported option features (%Lx).\n",
c1c9ff7c 2473 features);
f2b636e8 2474 err = -EINVAL;
16cdcec7 2475 goto fail_alloc;
f2b636e8 2476 }
61d92c32
CM		 2477
2478 btrfs_init_workers(&fs_info->generic_worker,
2479 "genwork", 1, NULL);
2480
5443be45 2481 btrfs_init_workers(&fs_info->workers, "worker",
61d92c32
CM		 2482 fs_info->thread_pool_size,
2483 &fs_info->generic_worker);
c8b97818 2484
771ed689 2485 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
61d92c32
CM		 2486 fs_info->thread_pool_size,
2487 &fs_info->generic_worker);
771ed689 2488
8ccf6f19
MX		 2489 btrfs_init_workers(&fs_info->flush_workers, "flush_delalloc",
2490 fs_info->thread_pool_size,
2491 &fs_info->generic_worker);
2492
5443be45 2493 btrfs_init_workers(&fs_info->submit_workers, "submit",
b720d209 2494 min_t(u64, fs_devices->num_devices,
61d92c32
CM		 2495 fs_info->thread_pool_size),
2496 &fs_info->generic_worker);
61b49440 2497
bab39bf9
JB		 2498 btrfs_init_workers(&fs_info->caching_workers, "cache",
2499 2, &fs_info->generic_worker);
2500
61b49440
CM		 2501 /* a higher idle thresh on the submit workers makes it much more
2502 * likely that bios will be send down in a sane order to the
2503 * devices
2504 */
2505 fs_info->submit_workers.idle_thresh = 64;
53863232 2506
771ed689 2507 fs_info->workers.idle_thresh = 16;
4a69a410 2508 fs_info->workers.ordered = 1;
61b49440 2509
771ed689
CM		 2510 fs_info->delalloc_workers.idle_thresh = 2;
2511 fs_info->delalloc_workers.ordered = 1;
2512
61d92c32
CM		 2513 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
2514 &fs_info->generic_worker);
5443be45 2515 btrfs_init_workers(&fs_info->endio_workers, "endio",
61d92c32
CM		 2516 fs_info->thread_pool_size,
2517 &fs_info->generic_worker);
d20f7043 2518 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
61d92c32
CM		 2519 fs_info->thread_pool_size,
2520 &fs_info->generic_worker);
cad321ad 2521 btrfs_init_workers(&fs_info->endio_meta_write_workers,
61d92c32
CM		 2522 "endio-meta-write", fs_info->thread_pool_size,
2523 &fs_info->generic_worker);
53b381b3
DW		 2524 btrfs_init_workers(&fs_info->endio_raid56_workers,
2525 "endio-raid56", fs_info->thread_pool_size,
2526 &fs_info->generic_worker);
2527 btrfs_init_workers(&fs_info->rmw_workers,
2528 "rmw", fs_info->thread_pool_size,
2529 &fs_info->generic_worker);
5443be45 2530 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
61d92c32
CM		 2531 fs_info->thread_pool_size,
2532 &fs_info->generic_worker);
0cb59c99
JB		 2533 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
2534 1, &fs_info->generic_worker);
16cdcec7
MX		 2535 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
2536 fs_info->thread_pool_size,
2537 &fs_info->generic_worker);
90519d66
AJ		 2538 btrfs_init_workers(&fs_info->readahead_workers, "readahead",
2539 fs_info->thread_pool_size,
2540 &fs_info->generic_worker);
2f232036
JS		 2541 btrfs_init_workers(&fs_info->qgroup_rescan_workers, "qgroup-rescan", 1,
2542 &fs_info->generic_worker);
61b49440
CM		 2543
2544 /*
2545 * endios are largely parallel and should have a very
2546 * low idle thresh
2547 */
2548 fs_info->endio_workers.idle_thresh = 4;
b51912c9 2549 fs_info->endio_meta_workers.idle_thresh = 4;
53b381b3
DW		 2550 fs_info->endio_raid56_workers.idle_thresh = 4;
2551 fs_info->rmw_workers.idle_thresh = 2;
b51912c9 2552
9042846b
CM		 2553 fs_info->endio_write_workers.idle_thresh = 2;
2554 fs_info->endio_meta_write_workers.idle_thresh = 2;
90519d66 2555 fs_info->readahead_workers.idle_thresh = 2;
9042846b 2556
0dc3b84a
JB		 2557 /*
2558 * btrfs_start_workers can really only fail because of ENOMEM so just
2559 * return -ENOMEM if any of these fail.
2560 */
2561 ret = btrfs_start_workers(&fs_info->workers);
2562 ret |= btrfs_start_workers(&fs_info->generic_worker);
2563 ret |= btrfs_start_workers(&fs_info->submit_workers);
2564 ret |= btrfs_start_workers(&fs_info->delalloc_workers);
2565 ret |= btrfs_start_workers(&fs_info->fixup_workers);
2566 ret |= btrfs_start_workers(&fs_info->endio_workers);
2567 ret |= btrfs_start_workers(&fs_info->endio_meta_workers);
53b381b3
DW		 2568 ret |= btrfs_start_workers(&fs_info->rmw_workers);
2569 ret |= btrfs_start_workers(&fs_info->endio_raid56_workers);
0dc3b84a
JB		 2570 ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers);
2571 ret |= btrfs_start_workers(&fs_info->endio_write_workers);
2572 ret |= btrfs_start_workers(&fs_info->endio_freespace_worker);
2573 ret |= btrfs_start_workers(&fs_info->delayed_workers);
2574 ret |= btrfs_start_workers(&fs_info->caching_workers);
2575 ret |= btrfs_start_workers(&fs_info->readahead_workers);
8ccf6f19 2576 ret |= btrfs_start_workers(&fs_info->flush_workers);
2f232036 2577 ret |= btrfs_start_workers(&fs_info->qgroup_rescan_workers);
0dc3b84a 2578 if (ret) {
fed425c7 2579 err = -ENOMEM;
0dc3b84a
JB		 2580 goto fail_sb_buffer;
2581 }
4543df7e 2582
4575c9cc 2583 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 2584 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2585 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 2586
db94535d
CM		 2587 tree_root->nodesize = nodesize;
2588 tree_root->leafsize = leafsize;
2589 tree_root->sectorsize = sectorsize;
87ee04eb 2590 tree_root->stripesize = stripesize;
a061fc8d
CM		 2591
2592 sb->s_blocksize = sectorsize;
2593 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 2594
3cae210f 2595 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
d397712b 2596 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 2597 goto fail_sb_buffer;
2598 }
19c00ddc 2599
8d082fb7
LB		 2600 if (sectorsize != PAGE_SIZE) {
2601 printk(KERN_WARNING "btrfs: Incompatible sector size(%lu) "
2602 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
941b2ddf
KM		 2603 goto fail_sb_buffer;
2604 }
2605
925baedd 2606 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 2607 ret = btrfs_read_sys_array(tree_root);
925baedd 2608 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2609 if (ret) {
d397712b
CM		 2610 printk(KERN_WARNING "btrfs: failed to read the system "
2611 "array on %s\n", sb->s_id);
5d4f98a2 2612 goto fail_sb_buffer;
84eed90f 2613 }
0b86a832
CM		 2614
2615 blocksize = btrfs_level_size(tree_root,
2616 btrfs_super_chunk_root_level(disk_super));
84234f3a 2617 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 2618
2619 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2620 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2621
2622 chunk_root->node = read_tree_block(chunk_root,
2623 btrfs_super_chunk_root(disk_super),
84234f3a 2624 blocksize, generation);
416bc658
JB		 2625 if (!chunk_root->node ||
2626 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
83121942
DW		 2627 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
2628 sb->s_id);
af31f5e5 2629 goto fail_tree_roots;
83121942 2630 }
5d4f98a2
YZ		 2631 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2632 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2633
e17cade2 2634 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
d397712b
CM		 2635 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
2636 BTRFS_UUID_SIZE);
e17cade2 2637
0b86a832 2638 ret = btrfs_read_chunk_tree(chunk_root);
2b82032c 2639 if (ret) {
d397712b
CM		 2640 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
2641 sb->s_id);
af31f5e5 2642 goto fail_tree_roots;
2b82032c 2643 }
0b86a832 2644
8dabb742
SB		 2645 /*
2646 * keep the device that is marked to be the target device for the
2647 * dev_replace procedure
2648 */
2649 btrfs_close_extra_devices(fs_info, fs_devices, 0);
dfe25020 2650
a6b0d5c8
CM		 2651 if (!fs_devices->latest_bdev) {
2652 printk(KERN_CRIT "btrfs: failed to read devices on %s\n",
2653 sb->s_id);
2654 goto fail_tree_roots;
2655 }
2656
af31f5e5 2657retry_root_backup:
db94535d
CM		 2658 blocksize = btrfs_level_size(tree_root,
2659 btrfs_super_root_level(disk_super));
84234f3a 2660 generation = btrfs_super_generation(disk_super);
0b86a832 2661
e20d96d6 2662 tree_root->node = read_tree_block(tree_root,
db94535d 2663 btrfs_super_root(disk_super),
84234f3a 2664 blocksize, generation);
af31f5e5
CM		 2665 if (!tree_root->node ||
2666 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
83121942
DW		 2667 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
2668 sb->s_id);
af31f5e5
CM		 2669
2670 goto recovery_tree_root;
83121942 2671 }
af31f5e5 2672
5d4f98a2
YZ		 2673 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2674 tree_root->commit_root = btrfs_root_node(tree_root);
db94535d 2675
cb517eab
MX		 2676 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2677 location.type = BTRFS_ROOT_ITEM_KEY;
2678 location.offset = 0;
2679
2680 extent_root = btrfs_read_tree_root(tree_root, &location);
2681 if (IS_ERR(extent_root)) {
2682 ret = PTR_ERR(extent_root);
af31f5e5 2683 goto recovery_tree_root;
cb517eab 2684 }
0b86a832 2685 extent_root->track_dirty = 1;
cb517eab 2686 fs_info->extent_root = extent_root;
0b86a832 2687
cb517eab
MX		 2688 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2689 dev_root = btrfs_read_tree_root(tree_root, &location);
2690 if (IS_ERR(dev_root)) {
2691 ret = PTR_ERR(dev_root);
af31f5e5 2692 goto recovery_tree_root;
cb517eab 2693 }
5d4f98a2 2694 dev_root->track_dirty = 1;
cb517eab
MX		 2695 fs_info->dev_root = dev_root;
2696 btrfs_init_devices_late(fs_info);
3768f368 2697
cb517eab
MX		 2698 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2699 csum_root = btrfs_read_tree_root(tree_root, &location);
2700 if (IS_ERR(csum_root)) {
2701 ret = PTR_ERR(csum_root);
af31f5e5 2702 goto recovery_tree_root;
cb517eab 2703 }
d20f7043 2704 csum_root->track_dirty = 1;
cb517eab 2705 fs_info->csum_root = csum_root;
d20f7043 2706
cb517eab
MX		 2707 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2708 quota_root = btrfs_read_tree_root(tree_root, &location);
2709 if (!IS_ERR(quota_root)) {
bcef60f2
AJ		 2710 quota_root->track_dirty = 1;
2711 fs_info->quota_enabled = 1;
2712 fs_info->pending_quota_state = 1;
cb517eab 2713 fs_info->quota_root = quota_root;
bcef60f2
AJ		 2714 }
2715
f7a81ea4
SB		 2716 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2717 uuid_root = btrfs_read_tree_root(tree_root, &location);
2718 if (IS_ERR(uuid_root)) {
2719 ret = PTR_ERR(uuid_root);
2720 if (ret != -ENOENT)
2721 goto recovery_tree_root;
2722 create_uuid_tree = true;
70f80175 2723 check_uuid_tree = false;
f7a81ea4
SB		 2724 } else {
2725 uuid_root->track_dirty = 1;
2726 fs_info->uuid_root = uuid_root;
70f80175
SB		 2727 create_uuid_tree = false;
2728 check_uuid_tree =
2729 generation != btrfs_super_uuid_tree_generation(disk_super);
f7a81ea4
SB		 2730 }
2731
8929ecfa
YZ		 2732 fs_info->generation = generation;
2733 fs_info->last_trans_committed = generation;
8929ecfa 2734
68310a5e
ID		 2735 ret = btrfs_recover_balance(fs_info);
2736 if (ret) {
2737 printk(KERN_WARNING "btrfs: failed to recover balance\n");
2738 goto fail_block_groups;
2739 }
2740
733f4fbb
SB		 2741 ret = btrfs_init_dev_stats(fs_info);
2742 if (ret) {
2743 printk(KERN_ERR "btrfs: failed to init dev_stats: %d\n",
2744 ret);
2745 goto fail_block_groups;
2746 }
2747
8dabb742
SB		 2748 ret = btrfs_init_dev_replace(fs_info);
2749 if (ret) {
2750 pr_err("btrfs: failed to init dev_replace: %d\n", ret);
2751 goto fail_block_groups;
2752 }
2753
2754 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2755
c59021f8 2756 ret = btrfs_init_space_info(fs_info);
2757 if (ret) {
2758 printk(KERN_ERR "Failed to initial space info: %d\n", ret);
2759 goto fail_block_groups;
2760 }
2761
1b1d1f66
JB		 2762 ret = btrfs_read_block_groups(extent_root);
2763 if (ret) {
2764 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
2765 goto fail_block_groups;
2766 }
5af3e8cc
SB		 2767 fs_info->num_tolerated_disk_barrier_failures =
2768 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
292fd7fc
SB		 2769 if (fs_info->fs_devices->missing_devices >
2770 fs_info->num_tolerated_disk_barrier_failures &&
2771 !(sb->s_flags & MS_RDONLY)) {
2772 printk(KERN_WARNING
2773 "Btrfs: too many missing devices, writeable mount is not allowed\n");
2774 goto fail_block_groups;
2775 }
9078a3e1 2776
a74a4b97
CM		 2777 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2778 "btrfs-cleaner");
57506d50 2779 if (IS_ERR(fs_info->cleaner_kthread))
1b1d1f66 2780 goto fail_block_groups;
a74a4b97
CM		 2781
2782 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2783 tree_root,
2784 "btrfs-transaction");
57506d50 2785 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2786 goto fail_cleaner;
a74a4b97 2787
c289811c
CM		 2788 if (!btrfs_test_opt(tree_root, SSD) &&
2789 !btrfs_test_opt(tree_root, NOSSD) &&
2790 !fs_info->fs_devices->rotating) {
2791 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
2792 "mode\n");
2793 btrfs_set_opt(fs_info->mount_opt, SSD);
2794 }
2795
21adbd5c
SB		 2796#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2797 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2798 ret = btrfsic_mount(tree_root, fs_devices,
2799 btrfs_test_opt(tree_root,
2800 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2801 1 : 0,
2802 fs_info->check_integrity_print_mask);
2803 if (ret)
2804 printk(KERN_WARNING "btrfs: failed to initialize"
2805 " integrity check module %s\n", sb->s_id);
2806 }
2807#endif
bcef60f2
AJ		 2808 ret = btrfs_read_qgroup_config(fs_info);
2809 if (ret)
2810 goto fail_trans_kthread;
21adbd5c 2811
acce952b 2812 /* do not make disk changes in broken FS */
68ce9682 2813 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM		 2814 u64 bytenr = btrfs_super_log_root(disk_super);
2815
7c2ca468 2816 if (fs_devices->rw_devices == 0) {
d397712b
CM		 2817 printk(KERN_WARNING "Btrfs log replay required "
2818 "on RO media\n");
7c2ca468 2819 err = -EIO;
bcef60f2 2820 goto fail_qgroup;
7c2ca468 2821 }
e02119d5
CM		 2822 blocksize =
2823 btrfs_level_size(tree_root,
2824 btrfs_super_log_root_level(disk_super));
d18a2c44 2825
6f07e42e 2826 log_tree_root = btrfs_alloc_root(fs_info);
676e4c86
DC		 2827 if (!log_tree_root) {
2828 err = -ENOMEM;
bcef60f2 2829 goto fail_qgroup;
676e4c86 2830 }
e02119d5
CM		 2831
2832 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2833 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2834
2835 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 2836 blocksize,
2837 generation + 1);
416bc658
JB		 2838 if (!log_tree_root->node ||
2839 !extent_buffer_uptodate(log_tree_root->node)) {
2840 printk(KERN_ERR "btrfs: failed to read log tree\n");
2841 free_extent_buffer(log_tree_root->node);
2842 kfree(log_tree_root);
2843 goto fail_trans_kthread;
2844 }
79787eaa 2845 /* returns with log_tree_root freed on success */
e02119d5 2846 ret = btrfs_recover_log_trees(log_tree_root);
79787eaa
JM		 2847 if (ret) {
2848 btrfs_error(tree_root->fs_info, ret,
2849 "Failed to recover log tree");
2850 free_extent_buffer(log_tree_root->node);
2851 kfree(log_tree_root);
2852 goto fail_trans_kthread;
2853 }
e556ce2c
YZ		 2854
2855 if (sb->s_flags & MS_RDONLY) {
79787eaa
JM		 2856 ret = btrfs_commit_super(tree_root);
2857 if (ret)
2858 goto fail_trans_kthread;
e556ce2c 2859 }
e02119d5 2860 }
1a40e23b 2861
76dda93c 2862 ret = btrfs_find_orphan_roots(tree_root);
79787eaa
JM		 2863 if (ret)
2864 goto fail_trans_kthread;
76dda93c 2865
7c2ca468 2866 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b 2867 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2
ID		 2868 if (ret)
2869 goto fail_trans_kthread;
d68fc57b 2870
5d4f98a2 2871 ret = btrfs_recover_relocation(tree_root);
d7ce5843
MX		 2872 if (ret < 0) {
2873 printk(KERN_WARNING
2874 "btrfs: failed to recover relocation\n");
2875 err = -EINVAL;
bcef60f2 2876 goto fail_qgroup;
d7ce5843 2877 }
7c2ca468 2878 }
1a40e23b 2879
3de4586c
CM		 2880 location.objectid = BTRFS_FS_TREE_OBJECTID;
2881 location.type = BTRFS_ROOT_ITEM_KEY;
cb517eab 2882 location.offset = 0;
3de4586c 2883
3de4586c 2884 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
3140c9a3
DC		 2885 if (IS_ERR(fs_info->fs_root)) {
2886 err = PTR_ERR(fs_info->fs_root);
bcef60f2 2887 goto fail_qgroup;
3140c9a3 2888 }
c289811c 2889
2b6ba629
ID		 2890 if (sb->s_flags & MS_RDONLY)
2891 return 0;
59641015 2892
2b6ba629
ID		 2893 down_read(&fs_info->cleanup_work_sem);
2894 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2895 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 2896 up_read(&fs_info->cleanup_work_sem);
2b6ba629
ID		 2897 close_ctree(tree_root);
2898 return ret;
2899 }
2900 up_read(&fs_info->cleanup_work_sem);
59641015 2901
2b6ba629
ID		 2902 ret = btrfs_resume_balance_async(fs_info);
2903 if (ret) {
2904 printk(KERN_WARNING "btrfs: failed to resume balance\n");
2905 close_ctree(tree_root);
2906 return ret;
e3acc2a6
JB		 2907 }
2908
8dabb742
SB		 2909 ret = btrfs_resume_dev_replace_async(fs_info);
2910 if (ret) {
2911 pr_warn("btrfs: failed to resume dev_replace\n");
2912 close_ctree(tree_root);
2913 return ret;
2914 }
2915
b382a324
JS		 2916 btrfs_qgroup_rescan_resume(fs_info);
2917
f7a81ea4
SB		 2918 if (create_uuid_tree) {
2919 pr_info("btrfs: creating UUID tree\n");
2920 ret = btrfs_create_uuid_tree(fs_info);
2921 if (ret) {
2922 pr_warn("btrfs: failed to create the UUID tree %d\n",
2923 ret);
2924 close_ctree(tree_root);
2925 return ret;
2926 }
f420ee1e
SB		 2927 } else if (check_uuid_tree ||
2928 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
70f80175
SB		 2929 pr_info("btrfs: checking UUID tree\n");
2930 ret = btrfs_check_uuid_tree(fs_info);
2931 if (ret) {
2932 pr_warn("btrfs: failed to check the UUID tree %d\n",
2933 ret);
2934 close_ctree(tree_root);
2935 return ret;
2936 }
2937 } else {
2938 fs_info->update_uuid_tree_gen = 1;
f7a81ea4
SB		 2939 }
2940
ad2b2c80 2941 return 0;
39279cc3 2942
bcef60f2
AJ		 2943fail_qgroup:
2944 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM		 2945fail_trans_kthread:
2946 kthread_stop(fs_info->transaction_kthread);
54067ae9 2947 btrfs_cleanup_transaction(fs_info->tree_root);
7b5ff90e 2948 del_fs_roots(fs_info);
3f157a2f 2949fail_cleaner:
a74a4b97 2950 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2951
2952 /*
2953 * make sure we're done with the btree inode before we stop our
2954 * kthreads
2955 */
2956 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 2957
1b1d1f66 2958fail_block_groups:
54067ae9 2959 btrfs_put_block_group_cache(fs_info);
1b1d1f66 2960 btrfs_free_block_groups(fs_info);
af31f5e5
CM		 2961
2962fail_tree_roots:
2963 free_root_pointers(fs_info, 1);
2b8195bb 2964 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 2965
39279cc3 2966fail_sb_buffer:
7abadb64 2967 btrfs_stop_all_workers(fs_info);
16cdcec7 2968fail_alloc:
4543df7e 2969fail_iput:
586e46e2
ID		 2970 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2971
4543df7e 2972 iput(fs_info->btree_inode);
963d678b
MX		 2973fail_delalloc_bytes:
2974 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX		 2975fail_dirty_metadata_bytes:
2976 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
ad081f14 2977fail_bdi:
7e662854 2978 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 2979fail_srcu:
2980 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 2981fail:
53b381b3 2982 btrfs_free_stripe_hash_table(fs_info);
586e46e2 2983 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 2984 return err;
af31f5e5
CM		 2985
2986recovery_tree_root:
af31f5e5
CM		 2987 if (!btrfs_test_opt(tree_root, RECOVERY))
2988 goto fail_tree_roots;
2989
2990 free_root_pointers(fs_info, 0);
2991
2992 /* don't use the log in recovery mode, it won't be valid */
2993 btrfs_set_super_log_root(disk_super, 0);
2994
2995 /* we can't trust the free space cache either */
2996 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
2997
2998 ret = next_root_backup(fs_info, fs_info->super_copy,
2999 &num_backups_tried, &backup_index);
3000 if (ret == -1)
3001 goto fail_block_groups;
3002 goto retry_root_backup;
eb60ceac
CM		 3003}
3004
f2984462
CM		 3005static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3006{
f2984462
CM		 3007 if (uptodate) {
3008 set_buffer_uptodate(bh);
3009 } else {
442a4f63
SB		 3010 struct btrfs_device *device = (struct btrfs_device *)
3011 bh->b_private;
3012
606686ee
JB		 3013 printk_ratelimited_in_rcu(KERN_WARNING "lost page write due to "
3014 "I/O error on %s\n",
3015 rcu_str_deref(device->name));
1259ab75
CM		 3016 /* note, we dont' set_buffer_write_io_error because we have
3017 * our own ways of dealing with the IO errors
3018 */
f2984462 3019 clear_buffer_uptodate(bh);
442a4f63 3020 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM		 3021 }
3022 unlock_buffer(bh);
3023 put_bh(bh);
3024}
3025
a512bbf8
YZ		 3026struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3027{
3028 struct buffer_head *bh;
3029 struct buffer_head *latest = NULL;
3030 struct btrfs_super_block *super;
3031 int i;
3032 u64 transid = 0;
3033 u64 bytenr;
3034
3035 /* we would like to check all the supers, but that would make
3036 * a btrfs mount succeed after a mkfs from a different FS.
3037 * So, we need to add a special mount option to scan for
3038 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3039 */
3040 for (i = 0; i < 1; i++) {
3041 bytenr = btrfs_sb_offset(i);
8068a47e
AJ		 3042 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3043 i_size_read(bdev->bd_inode))
a512bbf8 3044 break;
8068a47e
AJ		 3045 bh = __bread(bdev, bytenr / 4096,
3046 BTRFS_SUPER_INFO_SIZE);
a512bbf8
YZ		 3047 if (!bh)
3048 continue;
3049
3050 super = (struct btrfs_super_block *)bh->b_data;
3051 if (btrfs_super_bytenr(super) != bytenr ||
3cae210f 3052 btrfs_super_magic(super) != BTRFS_MAGIC) {
a512bbf8
YZ		 3053 brelse(bh);
3054 continue;
3055 }
3056
3057 if (!latest || btrfs_super_generation(super) > transid) {
3058 brelse(latest);
3059 latest = bh;
3060 transid = btrfs_super_generation(super);
3061 } else {
3062 brelse(bh);
3063 }
3064 }
3065 return latest;
3066}
3067
4eedeb75
HH		 3068/*
3069 * this should be called twice, once with wait == 0 and
3070 * once with wait == 1. When wait == 0 is done, all the buffer heads
3071 * we write are pinned.
3072 *
3073 * They are released when wait == 1 is done.
3074 * max_mirrors must be the same for both runs, and it indicates how
3075 * many supers on this one device should be written.
3076 *
3077 * max_mirrors == 0 means to write them all.
3078 */
a512bbf8
YZ		 3079static int write_dev_supers(struct btrfs_device *device,
3080 struct btrfs_super_block *sb,
3081 int do_barriers, int wait, int max_mirrors)
3082{
3083 struct buffer_head *bh;
3084 int i;
3085 int ret;
3086 int errors = 0;
3087 u32 crc;
3088 u64 bytenr;
a512bbf8
YZ		 3089
3090 if (max_mirrors == 0)
3091 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3092
a512bbf8
YZ		 3093 for (i = 0; i < max_mirrors; i++) {
3094 bytenr = btrfs_sb_offset(i);
3095 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
3096 break;
3097
3098 if (wait) {
3099 bh = __find_get_block(device->bdev, bytenr / 4096,
3100 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3101 if (!bh) {
3102 errors++;
3103 continue;
3104 }
a512bbf8 3105 wait_on_buffer(bh);
4eedeb75
HH		 3106 if (!buffer_uptodate(bh))
3107 errors++;
3108
3109 /* drop our reference */
3110 brelse(bh);
3111
3112 /* drop the reference from the wait == 0 run */
3113 brelse(bh);
3114 continue;
a512bbf8
YZ		 3115 } else {
3116 btrfs_set_super_bytenr(sb, bytenr);
3117
3118 crc = ~(u32)0;
b0496686 3119 crc = btrfs_csum_data((char *)sb +
a512bbf8
YZ		 3120 BTRFS_CSUM_SIZE, crc,
3121 BTRFS_SUPER_INFO_SIZE -
3122 BTRFS_CSUM_SIZE);
3123 btrfs_csum_final(crc, sb->csum);
3124
4eedeb75
HH		 3125 /*
3126 * one reference for us, and we leave it for the
3127 * caller
3128 */
a512bbf8
YZ		 3129 bh = __getblk(device->bdev, bytenr / 4096,
3130 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3131 if (!bh) {
3132 printk(KERN_ERR "btrfs: couldn't get super "
3133 "buffer head for bytenr %Lu\n", bytenr);
3134 errors++;
3135 continue;
3136 }
3137
a512bbf8
YZ		 3138 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3139
4eedeb75 3140 /* one reference for submit_bh */
a512bbf8 3141 get_bh(bh);
4eedeb75
HH		 3142
3143 set_buffer_uptodate(bh);
a512bbf8
YZ		 3144 lock_buffer(bh);
3145 bh->b_end_io = btrfs_end_buffer_write_sync;
442a4f63 3146 bh->b_private = device;
a512bbf8
YZ		 3147 }
3148
387125fc
CM		 3149 /*
3150 * we fua the first super. The others we allow
3151 * to go down lazy.
3152 */
21adbd5c 3153 ret = btrfsic_submit_bh(WRITE_FUA, bh);
4eedeb75 3154 if (ret)
a512bbf8 3155 errors++;
a512bbf8
YZ		 3156 }
3157 return errors < i ? 0 : -1;
3158}
3159
387125fc
CM		 3160/*
3161 * endio for the write_dev_flush, this will wake anyone waiting
3162 * for the barrier when it is done
3163 */
3164static void btrfs_end_empty_barrier(struct bio *bio, int err)
3165{
3166 if (err) {
3167 if (err == -EOPNOTSUPP)
3168 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3169 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3170 }
3171 if (bio->bi_private)
3172 complete(bio->bi_private);
3173 bio_put(bio);
3174}
3175
3176/*
3177 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3178 * sent down. With wait == 1, it waits for the previous flush.
3179 *
3180 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3181 * capable
3182 */
3183static int write_dev_flush(struct btrfs_device *device, int wait)
3184{
3185 struct bio *bio;
3186 int ret = 0;
3187
3188 if (device->nobarriers)
3189 return 0;
3190
3191 if (wait) {
3192 bio = device->flush_bio;
3193 if (!bio)
3194 return 0;
3195
3196 wait_for_completion(&device->flush_wait);
3197
3198 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
606686ee
JB		 3199 printk_in_rcu("btrfs: disabling barriers on dev %s\n",
3200 rcu_str_deref(device->name));
387125fc 3201 device->nobarriers = 1;
5af3e8cc 3202 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
387125fc 3203 ret = -EIO;
5af3e8cc
SB		 3204 btrfs_dev_stat_inc_and_print(device,
3205 BTRFS_DEV_STAT_FLUSH_ERRS);
387125fc
CM		 3206 }
3207
3208 /* drop the reference from the wait == 0 run */
3209 bio_put(bio);
3210 device->flush_bio = NULL;
3211
3212 return ret;
3213 }
3214
3215 /*
3216 * one reference for us, and we leave it for the
3217 * caller
3218 */
9c017abc 3219 device->flush_bio = NULL;
9be3395b 3220 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
387125fc
CM		 3221 if (!bio)
3222 return -ENOMEM;
3223
3224 bio->bi_end_io = btrfs_end_empty_barrier;
3225 bio->bi_bdev = device->bdev;
3226 init_completion(&device->flush_wait);
3227 bio->bi_private = &device->flush_wait;
3228 device->flush_bio = bio;
3229
3230 bio_get(bio);
21adbd5c 3231 btrfsic_submit_bio(WRITE_FLUSH, bio);
387125fc
CM		 3232
3233 return 0;
3234}
3235
3236/*
3237 * send an empty flush down to each device in parallel,
3238 * then wait for them
3239 */
3240static int barrier_all_devices(struct btrfs_fs_info *info)
3241{
3242 struct list_head *head;
3243 struct btrfs_device *dev;
5af3e8cc
SB		 3244 int errors_send = 0;
3245 int errors_wait = 0;
387125fc
CM		 3246 int ret;
3247
3248 /* send down all the barriers */
3249 head = &info->fs_devices->devices;
3250 list_for_each_entry_rcu(dev, head, dev_list) {
3251 if (!dev->bdev) {
5af3e8cc 3252 errors_send++;
387125fc
CM		 3253 continue;
3254 }
3255 if (!dev->in_fs_metadata || !dev->writeable)
3256 continue;
3257
3258 ret = write_dev_flush(dev, 0);
3259 if (ret)
5af3e8cc 3260 errors_send++;
387125fc
CM		 3261 }
3262
3263 /* wait for all the barriers */
3264 list_for_each_entry_rcu(dev, head, dev_list) {
3265 if (!dev->bdev) {
5af3e8cc 3266 errors_wait++;
387125fc
CM		 3267 continue;
3268 }
3269 if (!dev->in_fs_metadata || !dev->writeable)
3270 continue;
3271
3272 ret = write_dev_flush(dev, 1);
3273 if (ret)
5af3e8cc 3274 errors_wait++;
387125fc 3275 }
5af3e8cc
SB		 3276 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3277 errors_wait > info->num_tolerated_disk_barrier_failures)
387125fc
CM		 3278 return -EIO;
3279 return 0;
3280}
3281
5af3e8cc
SB		 3282int btrfs_calc_num_tolerated_disk_barrier_failures(
3283 struct btrfs_fs_info *fs_info)
3284{
3285 struct btrfs_ioctl_space_info space;
3286 struct btrfs_space_info *sinfo;
3287 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3288 BTRFS_BLOCK_GROUP_SYSTEM,
3289 BTRFS_BLOCK_GROUP_METADATA,
3290 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3291 int num_types = 4;
3292 int i;
3293 int c;
3294 int num_tolerated_disk_barrier_failures =
3295 (int)fs_info->fs_devices->num_devices;
3296
3297 for (i = 0; i < num_types; i++) {
3298 struct btrfs_space_info *tmp;
3299
3300 sinfo = NULL;
3301 rcu_read_lock();
3302 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3303 if (tmp->flags == types[i]) {
3304 sinfo = tmp;
3305 break;
3306 }
3307 }
3308 rcu_read_unlock();
3309
3310 if (!sinfo)
3311 continue;
3312
3313 down_read(&sinfo->groups_sem);
3314 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3315 if (!list_empty(&sinfo->block_groups[c])) {
3316 u64 flags;
3317
3318 btrfs_get_block_group_info(
3319 &sinfo->block_groups[c], &space);
3320 if (space.total_bytes == 0 ||
3321 space.used_bytes == 0)
3322 continue;
3323 flags = space.flags;
3324 /*
3325 * return
3326 * 0: if dup, single or RAID0 is configured for
3327 * any of metadata, system or data, else
3328 * 1: if RAID5 is configured, or if RAID1 or
3329 * RAID10 is configured and only two mirrors
3330 * are used, else
3331 * 2: if RAID6 is configured, else
3332 * num_mirrors - 1: if RAID1 or RAID10 is
3333 * configured and more than
3334 * 2 mirrors are used.
3335 */
3336 if (num_tolerated_disk_barrier_failures > 0 &&
3337 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3338 BTRFS_BLOCK_GROUP_RAID0)) ||
3339 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3340 == 0)))
3341 num_tolerated_disk_barrier_failures = 0;
53b381b3
DW		 3342 else if (num_tolerated_disk_barrier_failures > 1) {
3343 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3344 BTRFS_BLOCK_GROUP_RAID5 |
3345 BTRFS_BLOCK_GROUP_RAID10)) {
3346 num_tolerated_disk_barrier_failures = 1;
3347 } else if (flags &
15b0a89d 3348 BTRFS_BLOCK_GROUP_RAID6) {
53b381b3
DW		 3349 num_tolerated_disk_barrier_failures = 2;
3350 }
3351 }
5af3e8cc
SB		 3352 }
3353 }
3354 up_read(&sinfo->groups_sem);
3355 }
3356
3357 return num_tolerated_disk_barrier_failures;
3358}
3359
48a3b636 3360static int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 3361{
e5e9a520 3362 struct list_head *head;
f2984462 3363 struct btrfs_device *dev;
a061fc8d 3364 struct btrfs_super_block *sb;
f2984462 3365 struct btrfs_dev_item *dev_item;
f2984462
CM		 3366 int ret;
3367 int do_barriers;
a236aed1
CM		 3368 int max_errors;
3369 int total_errors = 0;
a061fc8d 3370 u64 flags;
f2984462 3371
6c41761f 3372 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
f2984462 3373 do_barriers = !btrfs_test_opt(root, NOBARRIER);
af31f5e5 3374 backup_super_roots(root->fs_info);
f2984462 3375
6c41761f 3376 sb = root->fs_info->super_for_commit;
a061fc8d 3377 dev_item = &sb->dev_item;
e5e9a520 3378
174ba509 3379 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 3380 head = &root->fs_info->fs_devices->devices;
387125fc 3381
5af3e8cc
SB		 3382 if (do_barriers) {
3383 ret = barrier_all_devices(root->fs_info);
3384 if (ret) {
3385 mutex_unlock(
3386 &root->fs_info->fs_devices->device_list_mutex);
3387 btrfs_error(root->fs_info, ret,
3388 "errors while submitting device barriers.");
3389 return ret;
3390 }
3391 }
387125fc 3392
1f78160c 3393 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3394 if (!dev->bdev) {
3395 total_errors++;
3396 continue;
3397 }
2b82032c 3398 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3399 continue;
3400
2b82032c 3401 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 3402 btrfs_set_stack_device_type(dev_item, dev->type);
3403 btrfs_set_stack_device_id(dev_item, dev->devid);
3404 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
3405 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
3406 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3407 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3408 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3409 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 3410 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 3411
a061fc8d
CM		 3412 flags = btrfs_super_flags(sb);
3413 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3414
a512bbf8 3415 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 3416 if (ret)
3417 total_errors++;
f2984462 3418 }
a236aed1 3419 if (total_errors > max_errors) {
d397712b
CM		 3420 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
3421 total_errors);
79787eaa
JM		 3422
3423 /* This shouldn't happen. FUA is masked off if unsupported */
a236aed1
CM		 3424 BUG();
3425 }
f2984462 3426
a512bbf8 3427 total_errors = 0;
1f78160c 3428 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3429 if (!dev->bdev)
3430 continue;
2b82032c 3431 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3432 continue;
3433
a512bbf8
YZ		 3434 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3435 if (ret)
3436 total_errors++;
f2984462 3437 }
174ba509 3438 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 3439 if (total_errors > max_errors) {
79787eaa
JM		 3440 btrfs_error(root->fs_info, -EIO,
3441 "%d errors while writing supers", total_errors);
3442 return -EIO;
a236aed1 3443 }
f2984462
CM		 3444 return 0;
3445}
3446
a512bbf8
YZ		 3447int write_ctree_super(struct btrfs_trans_handle *trans,
3448 struct btrfs_root *root, int max_mirrors)
eb60ceac 3449{
e66f709b 3450 int ret;
5f39d397 3451
a512bbf8 3452 ret = write_all_supers(root, max_mirrors);
5f39d397 3453 return ret;
cfaa7295
CM		 3454}
3455
cb517eab
MX		 3456/* Drop a fs root from the radix tree and free it. */
3457void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3458 struct btrfs_root *root)
2619ba1f 3459{
4df27c4d 3460 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 3461 radix_tree_delete(&fs_info->fs_roots_radix,
3462 (unsigned long)root->root_key.objectid);
4df27c4d 3463 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 3464
3465 if (btrfs_root_refs(&root->root_item) == 0)
3466 synchronize_srcu(&fs_info->subvol_srcu);
3467
d7634482 3468 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
3321719e
LB		 3469 btrfs_free_log(NULL, root);
3470 btrfs_free_log_root_tree(NULL, fs_info);
3471 }
3472
581bb050
LZ		 3473 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3474 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3475 free_fs_root(root);
4df27c4d
YZ		 3476}
3477
3478static void free_fs_root(struct btrfs_root *root)
3479{
82d5902d 3480 iput(root->cache_inode);
4df27c4d 3481 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
1cb048f5
FDBM		 3482 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3483 root->orphan_block_rsv = NULL;
0ee5dc67
AV		 3484 if (root->anon_dev)
3485 free_anon_bdev(root->anon_dev);
4df27c4d
YZ		 3486 free_extent_buffer(root->node);
3487 free_extent_buffer(root->commit_root);
581bb050
LZ		 3488 kfree(root->free_ino_ctl);
3489 kfree(root->free_ino_pinned);
d397712b 3490 kfree(root->name);
b0feb9d9 3491 btrfs_put_fs_root(root);
2619ba1f
CM		 3492}
3493
cb517eab
MX		 3494void btrfs_free_fs_root(struct btrfs_root *root)
3495{
3496 free_fs_root(root);
2619ba1f
CM		 3497}
3498
c146afad 3499int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3500{
c146afad
YZ		 3501 u64 root_objectid = 0;
3502 struct btrfs_root *gang[8];
3503 int i;
3768f368 3504 int ret;
e089f05c 3505
c146afad
YZ		 3506 while (1) {
3507 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3508 (void **)gang, root_objectid,
3509 ARRAY_SIZE(gang));
3510 if (!ret)
3511 break;
5d4f98a2
YZ		 3512
3513 root_objectid = gang[ret - 1]->root_key.objectid + 1;
c146afad 3514 for (i = 0; i < ret; i++) {
66b4ffd1
JB		 3515 int err;
3516
c146afad 3517 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 3518 err = btrfs_orphan_cleanup(gang[i]);
3519 if (err)
3520 return err;
c146afad
YZ		 3521 }
3522 root_objectid++;
3523 }
3524 return 0;
3525}
a2135011 3526
c146afad
YZ		 3527int btrfs_commit_super(struct btrfs_root *root)
3528{
3529 struct btrfs_trans_handle *trans;
3530 int ret;
a74a4b97 3531
c146afad 3532 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 3533 btrfs_run_delayed_iputs(root);
c146afad 3534 mutex_unlock(&root->fs_info->cleaner_mutex);
9d1a2a3a 3535 wake_up_process(root->fs_info->cleaner_kthread);
c71bf099
YZ		 3536
3537 /* wait until ongoing cleanup work done */
3538 down_write(&root->fs_info->cleanup_work_sem);
3539 up_write(&root->fs_info->cleanup_work_sem);
3540
7a7eaa40 3541 trans = btrfs_join_transaction(root);
3612b495
TI		 3542 if (IS_ERR(trans))
3543 return PTR_ERR(trans);
54aa1f4d 3544 ret = btrfs_commit_transaction(trans, root);
79787eaa
JM		 3545 if (ret)
3546 return ret;
c146afad 3547 /* run commit again to drop the original snapshot */
7a7eaa40 3548 trans = btrfs_join_transaction(root);
3612b495
TI		 3549 if (IS_ERR(trans))
3550 return PTR_ERR(trans);
79787eaa
JM		 3551 ret = btrfs_commit_transaction(trans, root);
3552 if (ret)
3553 return ret;
79154b1b 3554 ret = btrfs_write_and_wait_transaction(NULL, root);
79787eaa
JM		 3555 if (ret) {
3556 btrfs_error(root->fs_info, ret,
3557 "Failed to sync btree inode to disk.");
3558 return ret;
3559 }
d6bfde87 3560
a512bbf8 3561 ret = write_ctree_super(NULL, root, 0);
c146afad
YZ		 3562 return ret;
3563}
3564
3565int close_ctree(struct btrfs_root *root)
3566{
3567 struct btrfs_fs_info *fs_info = root->fs_info;
3568 int ret;
3569
3570 fs_info->closing = 1;
3571 smp_mb();
3572
803b2f54
SB		 3573 /* wait for the uuid_scan task to finish */
3574 down(&fs_info->uuid_tree_rescan_sem);
3575 /* avoid complains from lockdep et al., set sem back to initial state */
3576 up(&fs_info->uuid_tree_rescan_sem);
3577
837d5b6e 3578 /* pause restriper - we want to resume on mount */
aa1b8cd4 3579 btrfs_pause_balance(fs_info);
837d5b6e 3580
8dabb742
SB		 3581 btrfs_dev_replace_suspend_for_unmount(fs_info);
3582
aa1b8cd4 3583 btrfs_scrub_cancel(fs_info);
4cb5300b
CM		 3584
3585 /* wait for any defraggers to finish */
3586 wait_event(fs_info->transaction_wait,
3587 (atomic_read(&fs_info->defrag_running) == 0));
3588
3589 /* clear out the rbtree of defraggable inodes */
26176e7c 3590 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3591
c146afad 3592 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 3593 ret = btrfs_commit_super(root);
3594 if (ret)
3595 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
3596 }
3597
87533c47 3598 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
68ce9682 3599 btrfs_error_commit_super(root);
0f7d52f4 3600
300e4f8a
JB		 3601 btrfs_put_block_group_cache(fs_info);
3602
e3029d9f
AV		 3603 kthread_stop(fs_info->transaction_kthread);
3604 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3605
f25784b3
YZ		 3606 fs_info->closing = 2;
3607 smp_mb();
3608
bcef60f2
AJ		 3609 btrfs_free_qgroup_config(root->fs_info);
3610
963d678b
MX		 3611 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
3612 printk(KERN_INFO "btrfs: at unmount delalloc count %lld\n",
3613 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3614 }
bcc63abb 3615
e3029d9f 3616 btrfs_free_block_groups(fs_info);
d10c5f31 3617
13e6c37b 3618 btrfs_stop_all_workers(fs_info);
2932505a 3619
c146afad 3620 del_fs_roots(fs_info);
d10c5f31 3621
13e6c37b 3622 free_root_pointers(fs_info, 1);
9ad6b7bc 3623
13e6c37b 3624 iput(fs_info->btree_inode);
d6bfde87 3625
21adbd5c
SB		 3626#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3627 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3628 btrfsic_unmount(root, fs_info->fs_devices);
3629#endif
3630
dfe25020 3631 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3632 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3633
e2d84521 3634 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 3635 percpu_counter_destroy(&fs_info->delalloc_bytes);
04160088 3636 bdi_destroy(&fs_info->bdi);
76dda93c 3637 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 3638
53b381b3
DW		 3639 btrfs_free_stripe_hash_table(fs_info);
3640
1cb048f5
FDBM		 3641 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3642 root->orphan_block_rsv = NULL;
3643
eb60ceac
CM		 3644 return 0;
3645}
3646
b9fab919
CM		 3647int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3648 int atomic)
5f39d397 3649{
1259ab75 3650 int ret;
727011e0 3651 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 3652
0b32f4bb 3653 ret = extent_buffer_uptodate(buf);
1259ab75
CM		 3654 if (!ret)
3655 return ret;
3656
3657 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM		 3658 parent_transid, atomic);
3659 if (ret == -EAGAIN)
3660 return ret;
1259ab75 3661 return !ret;
5f39d397
CM		 3662}
3663
3664int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 3665{
0b32f4bb 3666 return set_extent_buffer_uptodate(buf);
5f39d397 3667}
6702ed49 3668
5f39d397
CM		 3669void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3670{
727011e0 3671 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
5f39d397 3672 u64 transid = btrfs_header_generation(buf);
b9473439 3673 int was_dirty;
b4ce94de 3674
b9447ef8 3675 btrfs_assert_tree_locked(buf);
31b1a2bd
JL		 3676 if (transid != root->fs_info->generation)
3677 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
d397712b 3678 "found %llu running %llu\n",
c1c9ff7c 3679 buf->start, transid, root->fs_info->generation);
0b32f4bb 3680 was_dirty = set_extent_buffer_dirty(buf);
e2d84521
MX		 3681 if (!was_dirty)
3682 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3683 buf->len,
3684 root->fs_info->dirty_metadata_batch);
eb60ceac
CM		 3685}
3686
b53d3f5d
LB		 3687static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3688 int flush_delayed)
16cdcec7
MX		 3689{
3690 /*
3691 * looks as though older kernels can get into trouble with
3692 * this code, they end up stuck in balance_dirty_pages forever
3693 */
e2d84521 3694 int ret;
16cdcec7
MX		 3695
3696 if (current->flags & PF_MEMALLOC)
3697 return;
3698
b53d3f5d
LB		 3699 if (flush_delayed)
3700 btrfs_balance_delayed_items(root);
16cdcec7 3701
e2d84521
MX		 3702 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3703 BTRFS_DIRTY_METADATA_THRESH);
3704 if (ret > 0) {
d0e1d66b
NJ		 3705 balance_dirty_pages_ratelimited(
3706 root->fs_info->btree_inode->i_mapping);
16cdcec7
MX		 3707 }
3708 return;
3709}
3710
b53d3f5d 3711void btrfs_btree_balance_dirty(struct btrfs_root *root)
35b7e476 3712{
b53d3f5d
LB		 3713 __btrfs_btree_balance_dirty(root, 1);
3714}
585ad2c3 3715
b53d3f5d
LB		 3716void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3717{
3718 __btrfs_btree_balance_dirty(root, 0);
35b7e476 3719}
6b80053d 3720
ca7a79ad 3721int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 3722{
727011e0 3723 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
0b32f4bb 3724 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
6b80053d 3725}
0da5468f 3726
fcd1f065 3727static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 3728 int read_only)
3729{
1104a885
DS		 3730 /*
3731 * Placeholder for checks
3732 */
fcd1f065 3733 return 0;
acce952b 3734}
3735
48a3b636 3736static void btrfs_error_commit_super(struct btrfs_root *root)
acce952b 3737{
acce952b 3738 mutex_lock(&root->fs_info->cleaner_mutex);
3739 btrfs_run_delayed_iputs(root);
3740 mutex_unlock(&root->fs_info->cleaner_mutex);
3741
3742 down_write(&root->fs_info->cleanup_work_sem);
3743 up_write(&root->fs_info->cleanup_work_sem);
3744
3745 /* cleanup FS via transaction */
3746 btrfs_cleanup_transaction(root);
acce952b 3747}
3748
569e0f35
JB		 3749static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
3750 struct btrfs_root *root)
acce952b 3751{
3752 struct btrfs_inode *btrfs_inode;
3753 struct list_head splice;
3754
3755 INIT_LIST_HEAD(&splice);
3756
3757 mutex_lock(&root->fs_info->ordered_operations_mutex);
199c2a9c 3758 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3759
569e0f35 3760 list_splice_init(&t->ordered_operations, &splice);
acce952b 3761 while (!list_empty(&splice)) {
3762 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3763 ordered_operations);
3764
3765 list_del_init(&btrfs_inode->ordered_operations);
199c2a9c 3766 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3767
3768 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3769
199c2a9c 3770 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3771 }
3772
199c2a9c 3773 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3774 mutex_unlock(&root->fs_info->ordered_operations_mutex);
acce952b 3775}
3776
143bede5 3777static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 3778{
acce952b 3779 struct btrfs_ordered_extent *ordered;
acce952b 3780
199c2a9c 3781 spin_lock(&root->ordered_extent_lock);
779880ef
JB		 3782 /*
3783 * This will just short circuit the ordered completion stuff which will
3784 * make sure the ordered extent gets properly cleaned up.
3785 */
199c2a9c 3786 list_for_each_entry(ordered, &root->ordered_extents,
779880ef
JB		 3787 root_extent_list)
3788 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
199c2a9c
MX		 3789 spin_unlock(&root->ordered_extent_lock);
3790}
3791
3792static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3793{
3794 struct btrfs_root *root;
3795 struct list_head splice;
3796
3797 INIT_LIST_HEAD(&splice);
3798
3799 spin_lock(&fs_info->ordered_root_lock);
3800 list_splice_init(&fs_info->ordered_roots, &splice);
3801 while (!list_empty(&splice)) {
3802 root = list_first_entry(&splice, struct btrfs_root,
3803 ordered_root);
3804 list_del_init(&root->ordered_root);
3805
3806 btrfs_destroy_ordered_extents(root);
3807
3808 cond_resched_lock(&fs_info->ordered_root_lock);
3809 }
3810 spin_unlock(&fs_info->ordered_root_lock);
acce952b 3811}
3812
35a3621b
SB		 3813static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3814 struct btrfs_root *root)
acce952b 3815{
3816 struct rb_node *node;
3817 struct btrfs_delayed_ref_root *delayed_refs;
3818 struct btrfs_delayed_ref_node *ref;
3819 int ret = 0;
3820
3821 delayed_refs = &trans->delayed_refs;
3822
3823 spin_lock(&delayed_refs->lock);
3824 if (delayed_refs->num_entries == 0) {
cfece4db 3825 spin_unlock(&delayed_refs->lock);
acce952b 3826 printk(KERN_INFO "delayed_refs has NO entry\n");
3827 return ret;
3828 }
3829
b939d1ab 3830 while ((node = rb_first(&delayed_refs->root)) != NULL) {
eb12db69 3831 struct btrfs_delayed_ref_head *head = NULL;
e78417d1 3832 bool pin_bytes = false;
acce952b 3833
eb12db69 3834 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
acce952b 3835 atomic_set(&ref->refs, 1);
3836 if (btrfs_delayed_ref_is_head(ref)) {
acce952b 3837
3838 head = btrfs_delayed_node_to_head(ref);
b939d1ab
JB		 3839 if (!mutex_trylock(&head->mutex)) {
3840 atomic_inc(&ref->refs);
3841 spin_unlock(&delayed_refs->lock);
3842
3843 /* Need to wait for the delayed ref to run */
3844 mutex_lock(&head->mutex);
3845 mutex_unlock(&head->mutex);
3846 btrfs_put_delayed_ref(ref);
3847
e18fca73 3848 spin_lock(&delayed_refs->lock);
b939d1ab
JB		 3849 continue;
3850 }
3851
54067ae9 3852 if (head->must_insert_reserved)
e78417d1 3853 pin_bytes = true;
78a6184a 3854 btrfs_free_delayed_extent_op(head->extent_op);
acce952b 3855 delayed_refs->num_heads--;
3856 if (list_empty(&head->cluster))
3857 delayed_refs->num_heads_ready--;
3858 list_del_init(&head->cluster);
acce952b 3859 }
eb12db69 3860
b939d1ab
JB		 3861 ref->in_tree = 0;
3862 rb_erase(&ref->rb_node, &delayed_refs->root);
3863 delayed_refs->num_entries--;
acce952b 3864 spin_unlock(&delayed_refs->lock);
e78417d1
JB		 3865 if (head) {
3866 if (pin_bytes)
3867 btrfs_pin_extent(root, ref->bytenr,
3868 ref->num_bytes, 1);
3869 mutex_unlock(&head->mutex);
3870 }
acce952b 3871 btrfs_put_delayed_ref(ref);
3872
3873 cond_resched();
3874 spin_lock(&delayed_refs->lock);
3875 }
3876
3877 spin_unlock(&delayed_refs->lock);
3878
3879 return ret;
3880}
3881
aec8030a 3882static void btrfs_evict_pending_snapshots(struct btrfs_transaction *t)
acce952b 3883{
3884 struct btrfs_pending_snapshot *snapshot;
3885 struct list_head splice;
3886
3887 INIT_LIST_HEAD(&splice);
3888
3889 list_splice_init(&t->pending_snapshots, &splice);
3890
3891 while (!list_empty(&splice)) {
3892 snapshot = list_entry(splice.next,
3893 struct btrfs_pending_snapshot,
3894 list);
aec8030a 3895 snapshot->error = -ECANCELED;
acce952b 3896 list_del_init(&snapshot->list);
acce952b 3897 }
acce952b 3898}
3899
143bede5 3900static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 3901{
3902 struct btrfs_inode *btrfs_inode;
3903 struct list_head splice;
3904
3905 INIT_LIST_HEAD(&splice);
3906
eb73c1b7
MX		 3907 spin_lock(&root->delalloc_lock);
3908 list_splice_init(&root->delalloc_inodes, &splice);
acce952b 3909
3910 while (!list_empty(&splice)) {
eb73c1b7
MX		 3911 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3912 delalloc_inodes);
acce952b 3913
3914 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX		 3915 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3916 &btrfs_inode->runtime_flags);
eb73c1b7 3917 spin_unlock(&root->delalloc_lock);
acce952b 3918
3919 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3920
eb73c1b7 3921 spin_lock(&root->delalloc_lock);
acce952b 3922 }
3923
eb73c1b7
MX		 3924 spin_unlock(&root->delalloc_lock);
3925}
3926
3927static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3928{
3929 struct btrfs_root *root;
3930 struct list_head splice;
3931
3932 INIT_LIST_HEAD(&splice);
3933
3934 spin_lock(&fs_info->delalloc_root_lock);
3935 list_splice_init(&fs_info->delalloc_roots, &splice);
3936 while (!list_empty(&splice)) {
3937 root = list_first_entry(&splice, struct btrfs_root,
3938 delalloc_root);
3939 list_del_init(&root->delalloc_root);
3940 root = btrfs_grab_fs_root(root);
3941 BUG_ON(!root);
3942 spin_unlock(&fs_info->delalloc_root_lock);
3943
3944 btrfs_destroy_delalloc_inodes(root);
3945 btrfs_put_fs_root(root);
3946
3947 spin_lock(&fs_info->delalloc_root_lock);
3948 }
3949 spin_unlock(&fs_info->delalloc_root_lock);
acce952b 3950}
3951
3952static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3953 struct extent_io_tree *dirty_pages,
3954 int mark)
3955{
3956 int ret;
acce952b 3957 struct extent_buffer *eb;
3958 u64 start = 0;
3959 u64 end;
acce952b 3960
3961 while (1) {
3962 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 3963 mark, NULL);
acce952b 3964 if (ret)
3965 break;
3966
3967 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
3968 while (start <= end) {
fd8b2b61
JB		 3969 eb = btrfs_find_tree_block(root, start,
3970 root->leafsize);
69a85bd8 3971 start += root->leafsize;
fd8b2b61 3972 if (!eb)
acce952b 3973 continue;
fd8b2b61 3974 wait_on_extent_buffer_writeback(eb);
acce952b 3975
fd8b2b61
JB		 3976 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
3977 &eb->bflags))
3978 clear_extent_buffer_dirty(eb);
3979 free_extent_buffer_stale(eb);
acce952b 3980 }
3981 }
3982
3983 return ret;
3984}
3985
3986static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3987 struct extent_io_tree *pinned_extents)
3988{
3989 struct extent_io_tree *unpin;
3990 u64 start;
3991 u64 end;
3992 int ret;
ed0eaa14 3993 bool loop = true;
acce952b 3994
3995 unpin = pinned_extents;
ed0eaa14 3996again:
acce952b 3997 while (1) {
3998 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 3999 EXTENT_DIRTY, NULL);
acce952b 4000 if (ret)
4001 break;
4002
4003 /* opt_discard */
5378e607
LD		 4004 if (btrfs_test_opt(root, DISCARD))
4005 ret = btrfs_error_discard_extent(root, start,
4006 end + 1 - start,
4007 NULL);
acce952b 4008
4009 clear_extent_dirty(unpin, start, end, GFP_NOFS);
4010 btrfs_error_unpin_extent_range(root, start, end);
4011 cond_resched();
4012 }
4013
ed0eaa14
LB		 4014 if (loop) {
4015 if (unpin == &root->fs_info->freed_extents[0])
4016 unpin = &root->fs_info->freed_extents[1];
4017 else
4018 unpin = &root->fs_info->freed_extents[0];
4019 loop = false;
4020 goto again;
4021 }
4022
acce952b 4023 return 0;
4024}
4025
49b25e05
JM		 4026void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4027 struct btrfs_root *root)
4028{
4029 btrfs_destroy_delayed_refs(cur_trans, root);
4030 btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv,
4031 cur_trans->dirty_pages.dirty_bytes);
4032
4a9d8bde 4033 cur_trans->state = TRANS_STATE_COMMIT_START;
d7096fc3 4034 wake_up(&root->fs_info->transaction_blocked_wait);
49b25e05 4035
aec8030a
MX		 4036 btrfs_evict_pending_snapshots(cur_trans);
4037
4a9d8bde 4038 cur_trans->state = TRANS_STATE_UNBLOCKED;
d7096fc3 4039 wake_up(&root->fs_info->transaction_wait);
49b25e05 4040
67cde344
MX		 4041 btrfs_destroy_delayed_inodes(root);
4042 btrfs_assert_delayed_root_empty(root);
49b25e05 4043
49b25e05
JM		 4044 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4045 EXTENT_DIRTY);
6e841e32
LB		 4046 btrfs_destroy_pinned_extent(root,
4047 root->fs_info->pinned_extents);
49b25e05 4048
4a9d8bde
MX		 4049 cur_trans->state =TRANS_STATE_COMPLETED;
4050 wake_up(&cur_trans->commit_wait);
4051
49b25e05
JM		 4052 /*
4053 memset(cur_trans, 0, sizeof(*cur_trans));
4054 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4055 */
4056}
4057
48a3b636 4058static int btrfs_cleanup_transaction(struct btrfs_root *root)
acce952b 4059{
4060 struct btrfs_transaction *t;
4061 LIST_HEAD(list);
4062
acce952b 4063 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4064
a4abeea4 4065 spin_lock(&root->fs_info->trans_lock);
acce952b 4066 list_splice_init(&root->fs_info->trans_list, &list);
ac673879 4067 root->fs_info->running_transaction = NULL;
a4abeea4
JB		 4068 spin_unlock(&root->fs_info->trans_lock);
4069
acce952b 4070 while (!list_empty(&list)) {
4071 t = list_entry(list.next, struct btrfs_transaction, list);
acce952b 4072
569e0f35 4073 btrfs_destroy_ordered_operations(t, root);
acce952b 4074
199c2a9c 4075 btrfs_destroy_all_ordered_extents(root->fs_info);
acce952b 4076
4077 btrfs_destroy_delayed_refs(t, root);
4078
4a9d8bde
MX		 4079 /*
4080 * FIXME: cleanup wait for commit
4081 * We needn't acquire the lock here, because we are during
4082 * the umount, there is no other task which will change it.
4083 */
4084 t->state = TRANS_STATE_COMMIT_START;
66657b31 4085 smp_mb();
acce952b 4086 if (waitqueue_active(&root->fs_info->transaction_blocked_wait))
4087 wake_up(&root->fs_info->transaction_blocked_wait);
4088
aec8030a
MX		 4089 btrfs_evict_pending_snapshots(t);
4090
4a9d8bde 4091 t->state = TRANS_STATE_UNBLOCKED;
66657b31 4092 smp_mb();
acce952b 4093 if (waitqueue_active(&root->fs_info->transaction_wait))
4094 wake_up(&root->fs_info->transaction_wait);
acce952b 4095
67cde344
MX		 4096 btrfs_destroy_delayed_inodes(root);
4097 btrfs_assert_delayed_root_empty(root);
4098
eb73c1b7 4099 btrfs_destroy_all_delalloc_inodes(root->fs_info);
acce952b 4100
4101 btrfs_destroy_marked_extents(root, &t->dirty_pages,
4102 EXTENT_DIRTY);
4103
4104 btrfs_destroy_pinned_extent(root,
4105 root->fs_info->pinned_extents);
4106
4a9d8bde
MX		 4107 t->state = TRANS_STATE_COMPLETED;
4108 smp_mb();
4109 if (waitqueue_active(&t->commit_wait))
4110 wake_up(&t->commit_wait);
4111
13c5a93e 4112 atomic_set(&t->use_count, 0);
acce952b 4113 list_del_init(&t->list);
4114 memset(t, 0, sizeof(*t));
4115 kmem_cache_free(btrfs_transaction_cachep, t);
4116 }
4117
4118 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 4119
4120 return 0;
4121}
4122
d1310b2e 4123static struct extent_io_ops btree_extent_io_ops = {
ce9adaa5 4124 .readpage_end_io_hook = btree_readpage_end_io_hook,
4bb31e92 4125 .readpage_io_failed_hook = btree_io_failed_hook,
0b86a832 4126 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 4127 /* note we're sharing with inode.c for the merge bio hook */
4128 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 4129};
Linux 6.x block layer and io_uring tree(s)