projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: publish per-super attributes in sysfs
[deliverable/linux.git] / fs / btrfs / disk-io.c
CommitLineData
6cbd5570
CM		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
e20d96d6 19#include <linux/fs.h>
d98237b3 20#include <linux/blkdev.h>
87cbda5c 21#include <linux/scatterlist.h>
22b0ebda 22#include <linux/swap.h>
0f7d52f4 23#include <linux/radix-tree.h>
35b7e476 24#include <linux/writeback.h>
d397712b 25#include <linux/buffer_head.h>
ce9adaa5 26#include <linux/workqueue.h>
a74a4b97 27#include <linux/kthread.h>
4b4e25f2 28#include <linux/freezer.h>
163e783e 29#include <linux/crc32c.h>
5a0e3ad6 30#include <linux/slab.h>
784b4e29 31#include <linux/migrate.h>
7a36ddec 32#include <linux/ratelimit.h>
6463fe58 33#include <linux/uuid.h>
803b2f54 34#include <linux/semaphore.h>
7e75bf3f 35#include <asm/unaligned.h>
eb60ceac
CM		 36#include "ctree.h"
37#include "disk-io.h"
e089f05c 38#include "transaction.h"
0f7d52f4 39#include "btrfs_inode.h"
0b86a832 40#include "volumes.h"
db94535d 41#include "print-tree.h"
8b712842 42#include "async-thread.h"
925baedd 43#include "locking.h"
e02119d5 44#include "tree-log.h"
fa9c0d79 45#include "free-space-cache.h"
581bb050 46#include "inode-map.h"
21adbd5c 47#include "check-integrity.h"
606686ee 48#include "rcu-string.h"
8dabb742 49#include "dev-replace.h"
53b381b3 50#include "raid56.h"
5ac1d209 51#include "sysfs.h"
eb60ceac 52
de0022b9
JB		 53#ifdef CONFIG_X86
54#include <asm/cpufeature.h>
55#endif
56
d1310b2e 57static struct extent_io_ops btree_extent_io_ops;
8b712842 58static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 59static void free_fs_root(struct btrfs_root *root);
fcd1f065 60static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 61 int read_only);
569e0f35
JB		 62static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
63 struct btrfs_root *root);
143bede5 64static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
acce952b 65static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
66 struct btrfs_root *root);
143bede5 67static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
acce952b 68static int btrfs_destroy_marked_extents(struct btrfs_root *root,
69 struct extent_io_tree *dirty_pages,
70 int mark);
71static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
72 struct extent_io_tree *pinned_extents);
48a3b636
ES		 73static int btrfs_cleanup_transaction(struct btrfs_root *root);
74static void btrfs_error_commit_super(struct btrfs_root *root);
ce9adaa5 75
d352ac68
CM		 76/*
77 * end_io_wq structs are used to do processing in task context when an IO is
78 * complete. This is used during reads to verify checksums, and it is used
79 * by writes to insert metadata for new file extents after IO is complete.
80 */
ce9adaa5
CM		 81struct end_io_wq {
82 struct bio *bio;
83 bio_end_io_t *end_io;
84 void *private;
85 struct btrfs_fs_info *info;
86 int error;
22c59948 87 int metadata;
ce9adaa5 88 struct list_head list;
8b712842 89 struct btrfs_work work;
ce9adaa5 90};
0da5468f 91
d352ac68
CM		 92/*
93 * async submit bios are used to offload expensive checksumming
94 * onto the worker threads. They checksum file and metadata bios
95 * just before they are sent down the IO stack.
96 */
44b8bd7e
CM		 97struct async_submit_bio {
98 struct inode *inode;
99 struct bio *bio;
100 struct list_head list;
4a69a410
CM		 101 extent_submit_bio_hook_t *submit_bio_start;
102 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM		 103 int rw;
104 int mirror_num;
c8b97818 105 unsigned long bio_flags;
eaf25d93
CM		 106 /*
107 * bio_offset is optional, can be used if the pages in the bio
108 * can't tell us where in the file the bio should go
109 */
110 u64 bio_offset;
8b712842 111 struct btrfs_work work;
79787eaa 112 int error;
44b8bd7e
CM		 113};
114
85d4e461
CM		 115/*
116 * Lockdep class keys for extent_buffer->lock's in this root. For a given
117 * eb, the lockdep key is determined by the btrfs_root it belongs to and
118 * the level the eb occupies in the tree.
119 *
120 * Different roots are used for different purposes and may nest inside each
121 * other and they require separate keysets. As lockdep keys should be
122 * static, assign keysets according to the purpose of the root as indicated
123 * by btrfs_root->objectid. This ensures that all special purpose roots
124 * have separate keysets.
4008c04a 125 *
85d4e461
CM		 126 * Lock-nesting across peer nodes is always done with the immediate parent
127 * node locked thus preventing deadlock. As lockdep doesn't know this, use
128 * subclass to avoid triggering lockdep warning in such cases.
4008c04a 129 *
85d4e461
CM		 130 * The key is set by the readpage_end_io_hook after the buffer has passed
131 * csum validation but before the pages are unlocked. It is also set by
132 * btrfs_init_new_buffer on freshly allocated blocks.
4008c04a 133 *
85d4e461
CM		 134 * We also add a check to make sure the highest level of the tree is the
135 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
136 * needs update as well.
4008c04a
CM		 137 */
138#ifdef CONFIG_DEBUG_LOCK_ALLOC
139# if BTRFS_MAX_LEVEL != 8
140# error
141# endif
85d4e461
CM		 142
143static struct btrfs_lockdep_keyset {
144 u64 id; /* root objectid */
145 const char *name_stem; /* lock name stem */
146 char names[BTRFS_MAX_LEVEL + 1][20];
147 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
148} btrfs_lockdep_keysets[] = {
149 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
150 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
151 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
152 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
153 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
154 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
60b62978 155 { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
85d4e461
CM		 156 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
157 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
158 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
13fd8da9 159 { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
85d4e461 160 { .id = 0, .name_stem = "tree" },
4008c04a 161};
85d4e461
CM		 162
163void __init btrfs_init_lockdep(void)
164{
165 int i, j;
166
167 /* initialize lockdep class names */
168 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
169 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
170
171 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
172 snprintf(ks->names[j], sizeof(ks->names[j]),
173 "btrfs-%s-%02d", ks->name_stem, j);
174 }
175}
176
177void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
178 int level)
179{
180 struct btrfs_lockdep_keyset *ks;
181
182 BUG_ON(level >= ARRAY_SIZE(ks->keys));
183
184 /* find the matching keyset, id 0 is the default entry */
185 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
186 if (ks->id == objectid)
187 break;
188
189 lockdep_set_class_and_name(&eb->lock,
190 &ks->keys[level], ks->names[level]);
191}
192
4008c04a
CM		 193#endif
194
d352ac68
CM		 195/*
196 * extents on the btree inode are pretty simple, there's one extent
197 * that covers the entire device
198 */
b2950863 199static struct extent_map *btree_get_extent(struct inode *inode,
306e16ce 200 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 201 int create)
7eccb903 202{
5f39d397
CM		 203 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
204 struct extent_map *em;
205 int ret;
206
890871be 207 read_lock(&em_tree->lock);
d1310b2e 208 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 209 if (em) {
210 em->bdev =
211 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 212 read_unlock(&em_tree->lock);
5f39d397 213 goto out;
a061fc8d 214 }
890871be 215 read_unlock(&em_tree->lock);
7b13b7b1 216
172ddd60 217 em = alloc_extent_map();
5f39d397
CM		 218 if (!em) {
219 em = ERR_PTR(-ENOMEM);
220 goto out;
221 }
222 em->start = 0;
0afbaf8c 223 em->len = (u64)-1;
c8b97818 224 em->block_len = (u64)-1;
5f39d397 225 em->block_start = 0;
a061fc8d 226 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 227
890871be 228 write_lock(&em_tree->lock);
09a2a8f9 229 ret = add_extent_mapping(em_tree, em, 0);
5f39d397
CM		 230 if (ret == -EEXIST) {
231 free_extent_map(em);
7b13b7b1 232 em = lookup_extent_mapping(em_tree, start, len);
b4f359ab 233 if (!em)
0433f20d 234 em = ERR_PTR(-EIO);
5f39d397 235 } else if (ret) {
7b13b7b1 236 free_extent_map(em);
0433f20d 237 em = ERR_PTR(ret);
5f39d397 238 }
890871be 239 write_unlock(&em_tree->lock);
7b13b7b1 240
5f39d397
CM		 241out:
242 return em;
7eccb903
CM		 243}
244
b0496686 245u32 btrfs_csum_data(char *data, u32 seed, size_t len)
19c00ddc 246{
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 478 found_start = btrfs_header_bytenr(eb);
fae7f21c 479 if (WARN_ON(found_start != start || !PageUptodate(page)))
4f2de97a 480 return 0;
19c00ddc 481 csum_tree_block(root, eb, 0);
19c00ddc
CM		 482 return 0;
483}
484
2b82032c
YZ		 485static int check_tree_block_fsid(struct btrfs_root *root,
486 struct extent_buffer *eb)
487{
488 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
489 u8 fsid[BTRFS_UUID_SIZE];
490 int ret = 1;
491
0a4e5586 492 read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
2b82032c
YZ		 493 while (fs_devices) {
494 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
495 ret = 0;
496 break;
497 }
498 fs_devices = fs_devices->seed;
499 }
500 return ret;
501}
502
a826d6dc
JB		 503#define CORRUPT(reason, eb, root, slot) \
504 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
505 "root=%llu, slot=%d\n", reason, \
c1c9ff7c 506 btrfs_header_bytenr(eb), root->objectid, slot)
a826d6dc
JB		 507
508static noinline int check_leaf(struct btrfs_root *root,
509 struct extent_buffer *leaf)
510{
511 struct btrfs_key key;
512 struct btrfs_key leaf_key;
513 u32 nritems = btrfs_header_nritems(leaf);
514 int slot;
515
516 if (nritems == 0)
517 return 0;
518
519 /* Check the 0 item */
520 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
521 BTRFS_LEAF_DATA_SIZE(root)) {
522 CORRUPT("invalid item offset size pair", leaf, root, 0);
523 return -EIO;
524 }
525
526 /*
527 * Check to make sure each items keys are in the correct order and their
528 * offsets make sense. We only have to loop through nritems-1 because
529 * we check the current slot against the next slot, which verifies the
530 * next slot's offset+size makes sense and that the current's slot
531 * offset is correct.
532 */
533 for (slot = 0; slot < nritems - 1; slot++) {
534 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
535 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
536
537 /* Make sure the keys are in the right order */
538 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
539 CORRUPT("bad key order", leaf, root, slot);
540 return -EIO;
541 }
542
543 /*
544 * Make sure the offset and ends are right, remember that the
545 * item data starts at the end of the leaf and grows towards the
546 * front.
547 */
548 if (btrfs_item_offset_nr(leaf, slot) !=
549 btrfs_item_end_nr(leaf, slot + 1)) {
550 CORRUPT("slot offset bad", leaf, root, slot);
551 return -EIO;
552 }
553
554 /*
555 * Check to make sure that we don't point outside of the leaf,
556 * just incase all the items are consistent to eachother, but
557 * all point outside of the leaf.
558 */
559 if (btrfs_item_end_nr(leaf, slot) >
560 BTRFS_LEAF_DATA_SIZE(root)) {
561 CORRUPT("slot end outside of leaf", leaf, root, slot);
562 return -EIO;
563 }
564 }
565
566 return 0;
567}
568
facc8a22
MX		 569static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
570 u64 phy_offset, struct page *page,
571 u64 start, u64 end, int mirror)
ce9adaa5
CM		 572{
573 struct extent_io_tree *tree;
574 u64 found_start;
575 int found_level;
ce9adaa5
CM		 576 struct extent_buffer *eb;
577 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 578 int ret = 0;
727011e0 579 int reads_done;
ce9adaa5 580
ce9adaa5
CM		 581 if (!page->private)
582 goto out;
d397712b 583
727011e0 584 tree = &BTRFS_I(page->mapping->host)->io_tree;
4f2de97a 585 eb = (struct extent_buffer *)page->private;
d397712b 586
0b32f4bb
JB		 587 /* the pending IO might have been the only thing that kept this buffer
588 * in memory. Make sure we have a ref for all this other checks
589 */
590 extent_buffer_get(eb);
591
592 reads_done = atomic_dec_and_test(&eb->io_pages);
727011e0
CM		 593 if (!reads_done)
594 goto err;
f188591e 595
5cf1ab56 596 eb->read_mirror = mirror;
ea466794
JB		 597 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
598 ret = -EIO;
599 goto err;
600 }
601
ce9adaa5 602 found_start = btrfs_header_bytenr(eb);
727011e0 603 if (found_start != eb->start) {
7a36ddec 604 printk_ratelimited(KERN_INFO "btrfs bad tree block start "
193f284d 605 "%llu %llu\n",
c1c9ff7c 606 found_start, eb->start);
f188591e 607 ret = -EIO;
ce9adaa5
CM		 608 goto err;
609 }
2b82032c 610 if (check_tree_block_fsid(root, eb)) {
7a36ddec 611 printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n",
c1c9ff7c 612 eb->start);
1259ab75
CM		 613 ret = -EIO;
614 goto err;
615 }
ce9adaa5 616 found_level = btrfs_header_level(eb);
1c24c3ce
JB		 617 if (found_level >= BTRFS_MAX_LEVEL) {
618 btrfs_info(root->fs_info, "bad tree block level %d\n",
619 (int)btrfs_header_level(eb));
620 ret = -EIO;
621 goto err;
622 }
ce9adaa5 623
85d4e461
CM		 624 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
625 eb, found_level);
4008c04a 626
ce9adaa5 627 ret = csum_tree_block(root, eb, 1);
a826d6dc 628 if (ret) {
f188591e 629 ret = -EIO;
a826d6dc
JB		 630 goto err;
631 }
632
633 /*
634 * If this is a leaf block and it is corrupt, set the corrupt bit so
635 * that we don't try and read the other copies of this block, just
636 * return -EIO.
637 */
638 if (found_level == 0 && check_leaf(root, eb)) {
639 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
640 ret = -EIO;
641 }
ce9adaa5 642
0b32f4bb
JB		 643 if (!ret)
644 set_extent_buffer_uptodate(eb);
ce9adaa5 645err:
79fb65a1
JB		 646 if (reads_done &&
647 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 648 btree_readahead_hook(root, eb, eb->start, ret);
4bb31e92 649
53b381b3
DW		 650 if (ret) {
651 /*
652 * our io error hook is going to dec the io pages
653 * again, we have to make sure it has something
654 * to decrement
655 */
656 atomic_inc(&eb->io_pages);
0b32f4bb 657 clear_extent_buffer_uptodate(eb);
53b381b3 658 }
0b32f4bb 659 free_extent_buffer(eb);
ce9adaa5 660out:
f188591e 661 return ret;
ce9adaa5
CM		 662}
663
ea466794 664static int btree_io_failed_hook(struct page *page, int failed_mirror)
4bb31e92 665{
4bb31e92
AJ		 666 struct extent_buffer *eb;
667 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
668
4f2de97a 669 eb = (struct extent_buffer *)page->private;
ea466794 670 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
5cf1ab56 671 eb->read_mirror = failed_mirror;
53b381b3 672 atomic_dec(&eb->io_pages);
ea466794 673 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 674 btree_readahead_hook(root, eb, eb->start, -EIO);
4bb31e92
AJ		 675 return -EIO; /* we fixed nothing */
676}
677
ce9adaa5 678static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM		 679{
680 struct end_io_wq *end_io_wq = bio->bi_private;
681 struct btrfs_fs_info *fs_info;
ce9adaa5 682
ce9adaa5 683 fs_info = end_io_wq->info;
ce9adaa5 684 end_io_wq->error = err;
8b712842
CM		 685 end_io_wq->work.func = end_workqueue_fn;
686 end_io_wq->work.flags = 0;
d20f7043 687
7b6d91da 688 if (bio->bi_rw & REQ_WRITE) {
53b381b3 689 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA)
cad321ad
CM		 690 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
691 &end_io_wq->work);
53b381b3 692 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE)
0cb59c99
JB		 693 btrfs_queue_worker(&fs_info->endio_freespace_worker,
694 &end_io_wq->work);
53b381b3
DW		 695 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
696 btrfs_queue_worker(&fs_info->endio_raid56_workers,
697 &end_io_wq->work);
cad321ad
CM		 698 else
699 btrfs_queue_worker(&fs_info->endio_write_workers,
700 &end_io_wq->work);
d20f7043 701 } else {
53b381b3
DW		 702 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
703 btrfs_queue_worker(&fs_info->endio_raid56_workers,
704 &end_io_wq->work);
705 else if (end_io_wq->metadata)
d20f7043
CM		 706 btrfs_queue_worker(&fs_info->endio_meta_workers,
707 &end_io_wq->work);
708 else
709 btrfs_queue_worker(&fs_info->endio_workers,
710 &end_io_wq->work);
711 }
ce9adaa5
CM		 712}
713
0cb59c99
JB		 714/*
715 * For the metadata arg you want
716 *
717 * 0 - if data
718 * 1 - if normal metadta
719 * 2 - if writing to the free space cache area
53b381b3 720 * 3 - raid parity work
0cb59c99 721 */
22c59948
CM		 722int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
723 int metadata)
0b86a832 724{
ce9adaa5 725 struct end_io_wq *end_io_wq;
ce9adaa5
CM		 726 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
727 if (!end_io_wq)
728 return -ENOMEM;
729
730 end_io_wq->private = bio->bi_private;
731 end_io_wq->end_io = bio->bi_end_io;
22c59948 732 end_io_wq->info = info;
ce9adaa5
CM		 733 end_io_wq->error = 0;
734 end_io_wq->bio = bio;
22c59948 735 end_io_wq->metadata = metadata;
ce9adaa5
CM		 736
737 bio->bi_private = end_io_wq;
738 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 739 return 0;
740}
741
b64a2851 742unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 743{
4854ddd0
CM		 744 unsigned long limit = min_t(unsigned long,
745 info->workers.max_workers,
746 info->fs_devices->open_devices);
747 return 256 * limit;
748}
0986fe9e 749
4a69a410
CM		 750static void run_one_async_start(struct btrfs_work *work)
751{
4a69a410 752 struct async_submit_bio *async;
79787eaa 753 int ret;
4a69a410
CM		 754
755 async = container_of(work, struct async_submit_bio, work);
79787eaa
JM		 756 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
757 async->mirror_num, async->bio_flags,
758 async->bio_offset);
759 if (ret)
760 async->error = ret;
4a69a410
CM		 761}
762
763static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 764{
765 struct btrfs_fs_info *fs_info;
766 struct async_submit_bio *async;
4854ddd0 767 int limit;
8b712842
CM		 768
769 async = container_of(work, struct async_submit_bio, work);
770 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 771
b64a2851 772 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 773 limit = limit * 2 / 3;
774
66657b31 775 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
b64a2851 776 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 777 wake_up(&fs_info->async_submit_wait);
778
79787eaa
JM		 779 /* If an error occured we just want to clean up the bio and move on */
780 if (async->error) {
781 bio_endio(async->bio, async->error);
782 return;
783 }
784
4a69a410 785 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 786 async->mirror_num, async->bio_flags,
787 async->bio_offset);
4a69a410
CM		 788}
789
790static void run_one_async_free(struct btrfs_work *work)
791{
792 struct async_submit_bio *async;
793
794 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 795 kfree(async);
796}
797
44b8bd7e
CM		 798int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
799 int rw, struct bio *bio, int mirror_num,
c8b97818 800 unsigned long bio_flags,
eaf25d93 801 u64 bio_offset,
4a69a410
CM		 802 extent_submit_bio_hook_t *submit_bio_start,
803 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 804{
805 struct async_submit_bio *async;
806
807 async = kmalloc(sizeof(*async), GFP_NOFS);
808 if (!async)
809 return -ENOMEM;
810
811 async->inode = inode;
812 async->rw = rw;
813 async->bio = bio;
814 async->mirror_num = mirror_num;
4a69a410
CM		 815 async->submit_bio_start = submit_bio_start;
816 async->submit_bio_done = submit_bio_done;
817
818 async->work.func = run_one_async_start;
819 async->work.ordered_func = run_one_async_done;
820 async->work.ordered_free = run_one_async_free;
821
8b712842 822 async->work.flags = 0;
c8b97818 823 async->bio_flags = bio_flags;
eaf25d93 824 async->bio_offset = bio_offset;
8c8bee1d 825
79787eaa
JM		 826 async->error = 0;
827
cb03c743 828 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 829
7b6d91da 830 if (rw & REQ_SYNC)
d313d7a3
CM		 831 btrfs_set_work_high_prio(&async->work);
832
8b712842 833 btrfs_queue_worker(&fs_info->workers, &async->work);
9473f16c 834
d397712b 835 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 836 atomic_read(&fs_info->nr_async_submits)) {
837 wait_event(fs_info->async_submit_wait,
838 (atomic_read(&fs_info->nr_async_submits) == 0));
839 }
840
44b8bd7e
CM		 841 return 0;
842}
843
ce3ed71a
CM		 844static int btree_csum_one_bio(struct bio *bio)
845{
846 struct bio_vec *bvec = bio->bi_io_vec;
847 int bio_index = 0;
848 struct btrfs_root *root;
79787eaa 849 int ret = 0;
ce3ed71a
CM		 850
851 WARN_ON(bio->bi_vcnt <= 0);
d397712b 852 while (bio_index < bio->bi_vcnt) {
ce3ed71a 853 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
79787eaa
JM		 854 ret = csum_dirty_buffer(root, bvec->bv_page);
855 if (ret)
856 break;
ce3ed71a
CM		 857 bio_index++;
858 bvec++;
859 }
79787eaa 860 return ret;
ce3ed71a
CM		 861}
862
4a69a410
CM		 863static int __btree_submit_bio_start(struct inode *inode, int rw,
864 struct bio *bio, int mirror_num,
eaf25d93
CM		 865 unsigned long bio_flags,
866 u64 bio_offset)
22c59948 867{
8b712842
CM		 868 /*
869 * when we're called for a write, we're already in the async
5443be45 870 * submission context. Just jump into btrfs_map_bio
8b712842 871 */
79787eaa 872 return btree_csum_one_bio(bio);
4a69a410 873}
22c59948 874
4a69a410 875static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 876 int mirror_num, unsigned long bio_flags,
877 u64 bio_offset)
4a69a410 878{
61891923
SB		 879 int ret;
880
8b712842 881 /*
4a69a410
CM		 882 * when we're called for a write, we're already in the async
883 * submission context. Just jump into btrfs_map_bio
8b712842 884 */
61891923
SB		 885 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
886 if (ret)
887 bio_endio(bio, ret);
888 return ret;
0b86a832
CM		 889}
890
de0022b9
JB		 891static int check_async_write(struct inode *inode, unsigned long bio_flags)
892{
893 if (bio_flags & EXTENT_BIO_TREE_LOG)
894 return 0;
895#ifdef CONFIG_X86
896 if (cpu_has_xmm4_2)
897 return 0;
898#endif
899 return 1;
900}
901
44b8bd7e 902static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 903 int mirror_num, unsigned long bio_flags,
904 u64 bio_offset)
44b8bd7e 905{
de0022b9 906 int async = check_async_write(inode, bio_flags);
cad321ad
CM		 907 int ret;
908
7b6d91da 909 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 910 /*
911 * called for a read, do the setup so that checksum validation
912 * can happen in the async kernel threads
913 */
f3f266ab
CM		 914 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
915 bio, 1);
1d4284bd 916 if (ret)
61891923
SB		 917 goto out_w_error;
918 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
919 mirror_num, 0);
de0022b9
JB		 920 } else if (!async) {
921 ret = btree_csum_one_bio(bio);
922 if (ret)
61891923
SB		 923 goto out_w_error;
924 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
925 mirror_num, 0);
926 } else {
927 /*
928 * kthread helpers are used to submit writes so that
929 * checksumming can happen in parallel across all CPUs
930 */
931 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
932 inode, rw, bio, mirror_num, 0,
933 bio_offset,
934 __btree_submit_bio_start,
935 __btree_submit_bio_done);
44b8bd7e 936 }
d313d7a3 937
61891923
SB		 938 if (ret) {
939out_w_error:
940 bio_endio(bio, ret);
941 }
942 return ret;
44b8bd7e
CM		 943}
944
3dd1462e 945#ifdef CONFIG_MIGRATION
784b4e29 946static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG		 947 struct page *newpage, struct page *page,
948 enum migrate_mode mode)
784b4e29
CM		 949{
950 /*
951 * we can't safely write a btree page from here,
952 * we haven't done the locking hook
953 */
954 if (PageDirty(page))
955 return -EAGAIN;
956 /*
957 * Buffers may be managed in a filesystem specific way.
958 * We must have no buffers or drop them.
959 */
960 if (page_has_private(page) &&
961 !try_to_release_page(page, GFP_KERNEL))
962 return -EAGAIN;
a6bc32b8 963 return migrate_page(mapping, newpage, page, mode);
784b4e29 964}
3dd1462e 965#endif
784b4e29 966
0da5468f
CM		 967
968static int btree_writepages(struct address_space *mapping,
969 struct writeback_control *wbc)
970{
d1310b2e 971 struct extent_io_tree *tree;
e2d84521
MX		 972 struct btrfs_fs_info *fs_info;
973 int ret;
974
d1310b2e 975 tree = &BTRFS_I(mapping->host)->io_tree;
d8d5f3e1 976 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM		 977
978 if (wbc->for_kupdate)
979 return 0;
980
e2d84521 981 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 982 /* this is a bit racy, but that's ok */
e2d84521
MX		 983 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
984 BTRFS_DIRTY_METADATA_THRESH);
985 if (ret < 0)
793955bc 986 return 0;
793955bc 987 }
0b32f4bb 988 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM		 989}
990
b2950863 991static int btree_readpage(struct file *file, struct page *page)
5f39d397 992{
d1310b2e
CM		 993 struct extent_io_tree *tree;
994 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 995 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 996}
22b0ebda 997
70dec807 998static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 999{
98509cfc 1000 if (PageWriteback(page) || PageDirty(page))
d397712b 1001 return 0;
0c4e538b 1002
f7a52a40 1003 return try_release_extent_buffer(page);
d98237b3
CM		 1004}
1005
d47992f8
LC		 1006static void btree_invalidatepage(struct page *page, unsigned int offset,
1007 unsigned int length)
d98237b3 1008{
d1310b2e
CM		 1009 struct extent_io_tree *tree;
1010 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 1011 extent_invalidatepage(tree, page, offset);
1012 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 1013 if (PagePrivate(page)) {
d397712b
CM		 1014 printk(KERN_WARNING "btrfs warning page private not zero "
1015 "on page %llu\n", (unsigned long long)page_offset(page));
9ad6b7bc
CM		 1016 ClearPagePrivate(page);
1017 set_page_private(page, 0);
1018 page_cache_release(page);
1019 }
d98237b3
CM		 1020}
1021
0b32f4bb
JB		 1022static int btree_set_page_dirty(struct page *page)
1023{
bb146eb2 1024#ifdef DEBUG
0b32f4bb
JB		 1025 struct extent_buffer *eb;
1026
1027 BUG_ON(!PagePrivate(page));
1028 eb = (struct extent_buffer *)page->private;
1029 BUG_ON(!eb);
1030 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1031 BUG_ON(!atomic_read(&eb->refs));
1032 btrfs_assert_tree_locked(eb);
bb146eb2 1033#endif
0b32f4bb
JB		 1034 return __set_page_dirty_nobuffers(page);
1035}
1036
7f09410b 1037static const struct address_space_operations btree_aops = {
d98237b3 1038 .readpage = btree_readpage,
0da5468f 1039 .writepages = btree_writepages,
5f39d397
CM		 1040 .releasepage = btree_releasepage,
1041 .invalidatepage = btree_invalidatepage,
5a92bc88 1042#ifdef CONFIG_MIGRATION
784b4e29 1043 .migratepage = btree_migratepage,
5a92bc88 1044#endif
0b32f4bb 1045 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM		 1046};
1047
ca7a79ad
CM		 1048int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1049 u64 parent_transid)
090d1875 1050{
5f39d397
CM		 1051 struct extent_buffer *buf = NULL;
1052 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 1053 int ret = 0;
090d1875 1054
db94535d 1055 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 1056 if (!buf)
090d1875 1057 return 0;
d1310b2e 1058 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
bb82ab88 1059 buf, 0, WAIT_NONE, btree_get_extent, 0);
5f39d397 1060 free_extent_buffer(buf);
de428b63 1061 return ret;
090d1875
CM		 1062}
1063
ab0fff03
AJ		 1064int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1065 int mirror_num, struct extent_buffer **eb)
1066{
1067 struct extent_buffer *buf = NULL;
1068 struct inode *btree_inode = root->fs_info->btree_inode;
1069 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1070 int ret;
1071
1072 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1073 if (!buf)
1074 return 0;
1075
1076 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1077
1078 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1079 btree_get_extent, mirror_num);
1080 if (ret) {
1081 free_extent_buffer(buf);
1082 return ret;
1083 }
1084
1085 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1086 free_extent_buffer(buf);
1087 return -EIO;
0b32f4bb 1088 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ		 1089 *eb = buf;
1090 } else {
1091 free_extent_buffer(buf);
1092 }
1093 return 0;
1094}
1095
0999df54
CM		 1096struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1097 u64 bytenr, u32 blocksize)
1098{
1099 struct inode *btree_inode = root->fs_info->btree_inode;
1100 struct extent_buffer *eb;
452c75c3 1101 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, bytenr);
0999df54
CM		 1102 return eb;
1103}
1104
1105struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1106 u64 bytenr, u32 blocksize)
1107{
1108 struct inode *btree_inode = root->fs_info->btree_inode;
1109 struct extent_buffer *eb;
1110
1111 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
727011e0 1112 bytenr, blocksize);
0999df54
CM		 1113 return eb;
1114}
1115
1116
e02119d5
CM		 1117int btrfs_write_tree_block(struct extent_buffer *buf)
1118{
727011e0 1119 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1120 buf->start + buf->len - 1);
e02119d5
CM		 1121}
1122
1123int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1124{
727011e0 1125 return filemap_fdatawait_range(buf->pages[0]->mapping,
8aa38c31 1126 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1127}
1128
0999df54 1129struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 1130 u32 blocksize, u64 parent_transid)
0999df54
CM		 1131{
1132 struct extent_buffer *buf = NULL;
0999df54
CM		 1133 int ret;
1134
0999df54
CM		 1135 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1136 if (!buf)
1137 return NULL;
0999df54 1138
ca7a79ad 1139 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
0f0fe8f7
FDBM		 1140 if (ret) {
1141 free_extent_buffer(buf);
1142 return NULL;
1143 }
5f39d397 1144 return buf;
ce9adaa5 1145
eb60ceac
CM		 1146}
1147
d5c13f92
JM		 1148void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1149 struct extent_buffer *buf)
ed2ff2cb 1150{
e2d84521
MX		 1151 struct btrfs_fs_info *fs_info = root->fs_info;
1152
55c69072 1153 if (btrfs_header_generation(buf) ==
e2d84521 1154 fs_info->running_transaction->transid) {
b9447ef8 1155 btrfs_assert_tree_locked(buf);
b4ce94de 1156
b9473439 1157 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
e2d84521
MX		 1158 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1159 -buf->len,
1160 fs_info->dirty_metadata_batch);
ed7b63eb
JB		 1161 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1162 btrfs_set_lock_blocking(buf);
1163 clear_extent_buffer_dirty(buf);
1164 }
925baedd 1165 }
5f39d397
CM		 1166}
1167
143bede5
JM		 1168static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1169 u32 stripesize, struct btrfs_root *root,
1170 struct btrfs_fs_info *fs_info,
1171 u64 objectid)
d97e63b6 1172{
cfaa7295 1173 root->node = NULL;
a28ec197 1174 root->commit_root = NULL;
db94535d
CM		 1175 root->sectorsize = sectorsize;
1176 root->nodesize = nodesize;
1177 root->leafsize = leafsize;
87ee04eb 1178 root->stripesize = stripesize;
123abc88 1179 root->ref_cows = 0;
0b86a832 1180 root->track_dirty = 0;
c71bf099 1181 root->in_radix = 0;
d68fc57b
YZ		 1182 root->orphan_item_inserted = 0;
1183 root->orphan_cleanup_state = 0;
0b86a832 1184
0f7d52f4
CM		 1185 root->objectid = objectid;
1186 root->last_trans = 0;
13a8a7c8 1187 root->highest_objectid = 0;
eb73c1b7 1188 root->nr_delalloc_inodes = 0;
199c2a9c 1189 root->nr_ordered_extents = 0;
58176a96 1190 root->name = NULL;
6bef4d31 1191 root->inode_tree = RB_ROOT;
16cdcec7 1192 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1193 root->block_rsv = NULL;
d68fc57b 1194 root->orphan_block_rsv = NULL;
0b86a832
CM		 1195
1196 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1197 INIT_LIST_HEAD(&root->root_list);
eb73c1b7
MX		 1198 INIT_LIST_HEAD(&root->delalloc_inodes);
1199 INIT_LIST_HEAD(&root->delalloc_root);
199c2a9c
MX		 1200 INIT_LIST_HEAD(&root->ordered_extents);
1201 INIT_LIST_HEAD(&root->ordered_root);
2ab28f32
JB		 1202 INIT_LIST_HEAD(&root->logged_list[0]);
1203 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1204 spin_lock_init(&root->orphan_lock);
5d4f98a2 1205 spin_lock_init(&root->inode_lock);
eb73c1b7 1206 spin_lock_init(&root->delalloc_lock);
199c2a9c 1207 spin_lock_init(&root->ordered_extent_lock);
f0486c68 1208 spin_lock_init(&root->accounting_lock);
2ab28f32
JB		 1209 spin_lock_init(&root->log_extents_lock[0]);
1210 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1211 mutex_init(&root->objectid_mutex);
e02119d5 1212 mutex_init(&root->log_mutex);
7237f183
YZ		 1213 init_waitqueue_head(&root->log_writer_wait);
1214 init_waitqueue_head(&root->log_commit_wait[0]);
1215 init_waitqueue_head(&root->log_commit_wait[1]);
1216 atomic_set(&root->log_commit[0], 0);
1217 atomic_set(&root->log_commit[1], 0);
1218 atomic_set(&root->log_writers, 0);
2ecb7923 1219 atomic_set(&root->log_batch, 0);
8a35d95f 1220 atomic_set(&root->orphan_inodes, 0);
b0feb9d9 1221 atomic_set(&root->refs, 1);
7237f183 1222 root->log_transid = 0;
257c62e1 1223 root->last_log_commit = 0;
06ea65a3
JB		 1224 if (fs_info)
1225 extent_io_tree_init(&root->dirty_log_pages,
1226 fs_info->btree_inode->i_mapping);
017e5369 1227
3768f368
CM		 1228 memset(&root->root_key, 0, sizeof(root->root_key));
1229 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1230 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1231 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
06ea65a3
JB		 1232 if (fs_info)
1233 root->defrag_trans_start = fs_info->generation;
1234 else
1235 root->defrag_trans_start = 0;
58176a96 1236 init_completion(&root->kobj_unregister);
6702ed49 1237 root->defrag_running = 0;
4d775673 1238 root->root_key.objectid = objectid;
0ee5dc67 1239 root->anon_dev = 0;
8ea05e3a 1240
5f3ab90a 1241 spin_lock_init(&root->root_item_lock);
3768f368
CM		 1242}
1243
f84a8bd6 1244static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
6f07e42e
AV		 1245{
1246 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1247 if (root)
1248 root->fs_info = fs_info;
1249 return root;
1250}
1251
06ea65a3
JB		 1252#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1253/* Should only be used by the testing infrastructure */
1254struct btrfs_root *btrfs_alloc_dummy_root(void)
1255{
1256 struct btrfs_root *root;
1257
1258 root = btrfs_alloc_root(NULL);
1259 if (!root)
1260 return ERR_PTR(-ENOMEM);
1261 __setup_root(4096, 4096, 4096, 4096, root, NULL, 1);
1262 root->dummy_root = 1;
1263
1264 return root;
1265}
1266#endif
1267
20897f5c
AJ		 1268struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1269 struct btrfs_fs_info *fs_info,
1270 u64 objectid)
1271{
1272 struct extent_buffer *leaf;
1273 struct btrfs_root *tree_root = fs_info->tree_root;
1274 struct btrfs_root *root;
1275 struct btrfs_key key;
1276 int ret = 0;
1277 u64 bytenr;
6463fe58 1278 uuid_le uuid;
20897f5c
AJ		 1279
1280 root = btrfs_alloc_root(fs_info);
1281 if (!root)
1282 return ERR_PTR(-ENOMEM);
1283
1284 __setup_root(tree_root->nodesize, tree_root->leafsize,
1285 tree_root->sectorsize, tree_root->stripesize,
1286 root, fs_info, objectid);
1287 root->root_key.objectid = objectid;
1288 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1289 root->root_key.offset = 0;
1290
1291 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1292 0, objectid, NULL, 0, 0, 0);
1293 if (IS_ERR(leaf)) {
1294 ret = PTR_ERR(leaf);
1dd05682 1295 leaf = NULL;
20897f5c
AJ		 1296 goto fail;
1297 }
1298
1299 bytenr = leaf->start;
1300 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1301 btrfs_set_header_bytenr(leaf, leaf->start);
1302 btrfs_set_header_generation(leaf, trans->transid);
1303 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1304 btrfs_set_header_owner(leaf, objectid);
1305 root->node = leaf;
1306
0a4e5586 1307 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
20897f5c
AJ		 1308 BTRFS_FSID_SIZE);
1309 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
b308bc2f 1310 btrfs_header_chunk_tree_uuid(leaf),
20897f5c
AJ		 1311 BTRFS_UUID_SIZE);
1312 btrfs_mark_buffer_dirty(leaf);
1313
1314 root->commit_root = btrfs_root_node(root);
1315 root->track_dirty = 1;
1316
1317
1318 root->root_item.flags = 0;
1319 root->root_item.byte_limit = 0;
1320 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1321 btrfs_set_root_generation(&root->root_item, trans->transid);
1322 btrfs_set_root_level(&root->root_item, 0);
1323 btrfs_set_root_refs(&root->root_item, 1);
1324 btrfs_set_root_used(&root->root_item, leaf->len);
1325 btrfs_set_root_last_snapshot(&root->root_item, 0);
1326 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB		 1327 uuid_le_gen(&uuid);
1328 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ		 1329 root->root_item.drop_level = 0;
1330
1331 key.objectid = objectid;
1332 key.type = BTRFS_ROOT_ITEM_KEY;
1333 key.offset = 0;
1334 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1335 if (ret)
1336 goto fail;
1337
1338 btrfs_tree_unlock(leaf);
1339
1dd05682
TI		 1340 return root;
1341
20897f5c 1342fail:
1dd05682
TI		 1343 if (leaf) {
1344 btrfs_tree_unlock(leaf);
1345 free_extent_buffer(leaf);
1346 }
1347 kfree(root);
20897f5c 1348
1dd05682 1349 return ERR_PTR(ret);
20897f5c
AJ		 1350}
1351
7237f183
YZ		 1352static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1353 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1354{
1355 struct btrfs_root *root;
1356 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1357 struct extent_buffer *leaf;
e02119d5 1358
6f07e42e 1359 root = btrfs_alloc_root(fs_info);
e02119d5 1360 if (!root)
7237f183 1361 return ERR_PTR(-ENOMEM);
e02119d5
CM		 1362
1363 __setup_root(tree_root->nodesize, tree_root->leafsize,
1364 tree_root->sectorsize, tree_root->stripesize,
1365 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1366
1367 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1368 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1369 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
7237f183
YZ		 1370 /*
1371 * log trees do not get reference counted because they go away
1372 * before a real commit is actually done. They do store pointers
1373 * to file data extents, and those reference counts still get
1374 * updated (along with back refs to the log tree).
1375 */
e02119d5
CM		 1376 root->ref_cows = 0;
1377
5d4f98a2 1378 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
66d7e7f0 1379 BTRFS_TREE_LOG_OBJECTID, NULL,
5581a51a 1380 0, 0, 0);
7237f183
YZ		 1381 if (IS_ERR(leaf)) {
1382 kfree(root);
1383 return ERR_CAST(leaf);
1384 }
e02119d5 1385
5d4f98a2
YZ		 1386 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1387 btrfs_set_header_bytenr(leaf, leaf->start);
1388 btrfs_set_header_generation(leaf, trans->transid);
1389 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1390 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1391 root->node = leaf;
e02119d5
CM		 1392
1393 write_extent_buffer(root->node, root->fs_info->fsid,
0a4e5586 1394 btrfs_header_fsid(), BTRFS_FSID_SIZE);
e02119d5
CM		 1395 btrfs_mark_buffer_dirty(root->node);
1396 btrfs_tree_unlock(root->node);
7237f183
YZ		 1397 return root;
1398}
1399
1400int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1401 struct btrfs_fs_info *fs_info)
1402{
1403 struct btrfs_root *log_root;
1404
1405 log_root = alloc_log_tree(trans, fs_info);
1406 if (IS_ERR(log_root))
1407 return PTR_ERR(log_root);
1408 WARN_ON(fs_info->log_root_tree);
1409 fs_info->log_root_tree = log_root;
1410 return 0;
1411}
1412
1413int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1414 struct btrfs_root *root)
1415{
1416 struct btrfs_root *log_root;
1417 struct btrfs_inode_item *inode_item;
1418
1419 log_root = alloc_log_tree(trans, root->fs_info);
1420 if (IS_ERR(log_root))
1421 return PTR_ERR(log_root);
1422
1423 log_root->last_trans = trans->transid;
1424 log_root->root_key.offset = root->root_key.objectid;
1425
1426 inode_item = &log_root->root_item.inode;
3cae210f
QW		 1427 btrfs_set_stack_inode_generation(inode_item, 1);
1428 btrfs_set_stack_inode_size(inode_item, 3);
1429 btrfs_set_stack_inode_nlink(inode_item, 1);
1430 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
1431 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
7237f183 1432
5d4f98a2 1433 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1434
1435 WARN_ON(root->log_root);
1436 root->log_root = log_root;
1437 root->log_transid = 0;
257c62e1 1438 root->last_log_commit = 0;
e02119d5
CM		 1439 return 0;
1440}
1441
35a3621b
SB		 1442static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1443 struct btrfs_key *key)
e02119d5
CM		 1444{
1445 struct btrfs_root *root;
1446 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1447 struct btrfs_path *path;
84234f3a 1448 u64 generation;
db94535d 1449 u32 blocksize;
cb517eab 1450 int ret;
0f7d52f4 1451
cb517eab
MX		 1452 path = btrfs_alloc_path();
1453 if (!path)
0f7d52f4 1454 return ERR_PTR(-ENOMEM);
cb517eab
MX		 1455
1456 root = btrfs_alloc_root(fs_info);
1457 if (!root) {
1458 ret = -ENOMEM;
1459 goto alloc_fail;
0f7d52f4
CM		 1460 }
1461
db94535d 1462 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb 1463 tree_root->sectorsize, tree_root->stripesize,
cb517eab 1464 root, fs_info, key->objectid);
0f7d52f4 1465
cb517eab
MX		 1466 ret = btrfs_find_root(tree_root, key, path,
1467 &root->root_item, &root->root_key);
0f7d52f4 1468 if (ret) {
13a8a7c8
YZ		 1469 if (ret > 0)
1470 ret = -ENOENT;
cb517eab 1471 goto find_fail;
0f7d52f4 1472 }
13a8a7c8 1473
84234f3a 1474 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1475 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1476 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1477 blocksize, generation);
cb517eab
MX		 1478 if (!root->node) {
1479 ret = -ENOMEM;
1480 goto find_fail;
1481 } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1482 ret = -EIO;
1483 goto read_fail;
416bc658 1484 }
5d4f98a2 1485 root->commit_root = btrfs_root_node(root);
13a8a7c8 1486out:
cb517eab
MX		 1487 btrfs_free_path(path);
1488 return root;
1489
1490read_fail:
1491 free_extent_buffer(root->node);
1492find_fail:
1493 kfree(root);
1494alloc_fail:
1495 root = ERR_PTR(ret);
1496 goto out;
1497}
1498
1499struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1500 struct btrfs_key *location)
1501{
1502 struct btrfs_root *root;
1503
1504 root = btrfs_read_tree_root(tree_root, location);
1505 if (IS_ERR(root))
1506 return root;
1507
1508 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
e02119d5 1509 root->ref_cows = 1;
08fe4db1
LZ		 1510 btrfs_check_and_init_root_item(&root->root_item);
1511 }
13a8a7c8 1512
5eda7b5e
CM		 1513 return root;
1514}
1515
cb517eab
MX		 1516int btrfs_init_fs_root(struct btrfs_root *root)
1517{
1518 int ret;
1519
1520 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1521 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1522 GFP_NOFS);
1523 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1524 ret = -ENOMEM;
1525 goto fail;
1526 }
1527
1528 btrfs_init_free_ino_ctl(root);
1529 mutex_init(&root->fs_commit_mutex);
1530 spin_lock_init(&root->cache_lock);
1531 init_waitqueue_head(&root->cache_wait);
1532
1533 ret = get_anon_bdev(&root->anon_dev);
1534 if (ret)
1535 goto fail;
1536 return 0;
1537fail:
1538 kfree(root->free_ino_ctl);
1539 kfree(root->free_ino_pinned);
1540 return ret;
1541}
1542
171170c1
ST		 1543static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1544 u64 root_id)
cb517eab
MX		 1545{
1546 struct btrfs_root *root;
1547
1548 spin_lock(&fs_info->fs_roots_radix_lock);
1549 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1550 (unsigned long)root_id);
1551 spin_unlock(&fs_info->fs_roots_radix_lock);
1552 return root;
1553}
1554
1555int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1556 struct btrfs_root *root)
1557{
1558 int ret;
1559
1560 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1561 if (ret)
1562 return ret;
1563
1564 spin_lock(&fs_info->fs_roots_radix_lock);
1565 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1566 (unsigned long)root->root_key.objectid,
1567 root);
1568 if (ret == 0)
1569 root->in_radix = 1;
1570 spin_unlock(&fs_info->fs_roots_radix_lock);
1571 radix_tree_preload_end();
1572
1573 return ret;
1574}
1575
c00869f1
MX		 1576struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1577 struct btrfs_key *location,
1578 bool check_ref)
5eda7b5e
CM		 1579{
1580 struct btrfs_root *root;
1581 int ret;
1582
edbd8d4e
CM		 1583 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1584 return fs_info->tree_root;
1585 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1586 return fs_info->extent_root;
8f18cf13
CM		 1587 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1588 return fs_info->chunk_root;
1589 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1590 return fs_info->dev_root;
0403e47e
YZ		 1591 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1592 return fs_info->csum_root;
bcef60f2
AJ		 1593 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1594 return fs_info->quota_root ? fs_info->quota_root :
1595 ERR_PTR(-ENOENT);
f7a81ea4
SB		 1596 if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1597 return fs_info->uuid_root ? fs_info->uuid_root :
1598 ERR_PTR(-ENOENT);
4df27c4d 1599again:
cb517eab 1600 root = btrfs_lookup_fs_root(fs_info, location->objectid);
48475471 1601 if (root) {
c00869f1 1602 if (check_ref && btrfs_root_refs(&root->root_item) == 0)
48475471 1603 return ERR_PTR(-ENOENT);
5eda7b5e 1604 return root;
48475471 1605 }
5eda7b5e 1606
cb517eab 1607 root = btrfs_read_fs_root(fs_info->tree_root, location);
5eda7b5e
CM		 1608 if (IS_ERR(root))
1609 return root;
3394e160 1610
c00869f1 1611 if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
cb517eab 1612 ret = -ENOENT;
581bb050 1613 goto fail;
35a30d7c 1614 }
581bb050 1615
cb517eab 1616 ret = btrfs_init_fs_root(root);
ac08aedf
CM		 1617 if (ret)
1618 goto fail;
3394e160 1619
d68fc57b
YZ		 1620 ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
1621 if (ret < 0)
1622 goto fail;
1623 if (ret == 0)
1624 root->orphan_item_inserted = 1;
1625
cb517eab 1626 ret = btrfs_insert_fs_root(fs_info, root);
0f7d52f4 1627 if (ret) {
4df27c4d
YZ		 1628 if (ret == -EEXIST) {
1629 free_fs_root(root);
1630 goto again;
1631 }
1632 goto fail;
0f7d52f4 1633 }
edbd8d4e 1634 return root;
4df27c4d
YZ		 1635fail:
1636 free_fs_root(root);
1637 return ERR_PTR(ret);
edbd8d4e
CM		 1638}
1639
04160088
CM		 1640static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1641{
1642 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1643 int ret = 0;
04160088
CM		 1644 struct btrfs_device *device;
1645 struct backing_dev_info *bdi;
b7967db7 1646
1f78160c
XG		 1647 rcu_read_lock();
1648 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1649 if (!device->bdev)
1650 continue;
04160088
CM		 1651 bdi = blk_get_backing_dev_info(device->bdev);
1652 if (bdi && bdi_congested(bdi, bdi_bits)) {
1653 ret = 1;
1654 break;
1655 }
1656 }
1f78160c 1657 rcu_read_unlock();
04160088
CM		 1658 return ret;
1659}
1660
ad081f14
JA		 1661/*
1662 * If this fails, caller must call bdi_destroy() to get rid of the
1663 * bdi again.
1664 */
04160088
CM		 1665static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1666{
ad081f14
JA		 1667 int err;
1668
1669 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1670 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1671 if (err)
1672 return err;
1673
4575c9cc 1674 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1675 bdi->congested_fn = btrfs_congested_fn;
1676 bdi->congested_data = info;
1677 return 0;
1678}
1679
8b712842
CM		 1680/*
1681 * called by the kthread helper functions to finally call the bio end_io
1682 * functions. This is where read checksum verification actually happens
1683 */
1684static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1685{
ce9adaa5 1686 struct bio *bio;
8b712842
CM		 1687 struct end_io_wq *end_io_wq;
1688 struct btrfs_fs_info *fs_info;
ce9adaa5 1689 int error;
ce9adaa5 1690
8b712842
CM		 1691 end_io_wq = container_of(work, struct end_io_wq, work);
1692 bio = end_io_wq->bio;
1693 fs_info = end_io_wq->info;
ce9adaa5 1694
8b712842
CM		 1695 error = end_io_wq->error;
1696 bio->bi_private = end_io_wq->private;
1697 bio->bi_end_io = end_io_wq->end_io;
1698 kfree(end_io_wq);
8b712842 1699 bio_endio(bio, error);
44b8bd7e
CM		 1700}
1701
a74a4b97
CM		 1702static int cleaner_kthread(void *arg)
1703{
1704 struct btrfs_root *root = arg;
d0278245 1705 int again;
a74a4b97
CM		 1706
1707 do {
d0278245 1708 again = 0;
a74a4b97 1709
d0278245 1710 /* Make the cleaner go to sleep early. */
babbf170 1711 if (btrfs_need_cleaner_sleep(root))
d0278245
MX		 1712 goto sleep;
1713
1714 if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1715 goto sleep;
1716
dc7f370c
MX		 1717 /*
1718 * Avoid the problem that we change the status of the fs
1719 * during the above check and trylock.
1720 */
babbf170 1721 if (btrfs_need_cleaner_sleep(root)) {
dc7f370c
MX		 1722 mutex_unlock(&root->fs_info->cleaner_mutex);
1723 goto sleep;
76dda93c 1724 }
a74a4b97 1725
d0278245
MX		 1726 btrfs_run_delayed_iputs(root);
1727 again = btrfs_clean_one_deleted_snapshot(root);
1728 mutex_unlock(&root->fs_info->cleaner_mutex);
1729
1730 /*
05323cd1
MX		 1731 * The defragger has dealt with the R/O remount and umount,
1732 * needn't do anything special here.
d0278245
MX		 1733 */
1734 btrfs_run_defrag_inodes(root->fs_info);
1735sleep:
9d1a2a3a 1736 if (!try_to_freeze() && !again) {
a74a4b97 1737 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1738 if (!kthread_should_stop())
1739 schedule();
a74a4b97
CM		 1740 __set_current_state(TASK_RUNNING);
1741 }
1742 } while (!kthread_should_stop());
1743 return 0;
1744}
1745
1746static int transaction_kthread(void *arg)
1747{
1748 struct btrfs_root *root = arg;
1749 struct btrfs_trans_handle *trans;
1750 struct btrfs_transaction *cur;
8929ecfa 1751 u64 transid;
a74a4b97
CM		 1752 unsigned long now;
1753 unsigned long delay;
914b2007 1754 bool cannot_commit;
a74a4b97
CM		 1755
1756 do {
914b2007 1757 cannot_commit = false;
8b87dc17 1758 delay = HZ * root->fs_info->commit_interval;
a74a4b97
CM		 1759 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1760
a4abeea4 1761 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1762 cur = root->fs_info->running_transaction;
1763 if (!cur) {
a4abeea4 1764 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1765 goto sleep;
1766 }
31153d81 1767
a74a4b97 1768 now = get_seconds();
4a9d8bde 1769 if (cur->state < TRANS_STATE_BLOCKED &&
8b87dc17
DS		 1770 (now < cur->start_time ||
1771 now - cur->start_time < root->fs_info->commit_interval)) {
a4abeea4 1772 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1773 delay = HZ * 5;
1774 goto sleep;
1775 }
8929ecfa 1776 transid = cur->transid;
a4abeea4 1777 spin_unlock(&root->fs_info->trans_lock);
56bec294 1778
79787eaa 1779 /* If the file system is aborted, this will always fail. */
354aa0fb 1780 trans = btrfs_attach_transaction(root);
914b2007 1781 if (IS_ERR(trans)) {
354aa0fb
MX		 1782 if (PTR_ERR(trans) != -ENOENT)
1783 cannot_commit = true;
79787eaa 1784 goto sleep;
914b2007 1785 }
8929ecfa 1786 if (transid == trans->transid) {
79787eaa 1787 btrfs_commit_transaction(trans, root);
8929ecfa
YZ		 1788 } else {
1789 btrfs_end_transaction(trans, root);
1790 }
a74a4b97
CM		 1791sleep:
1792 wake_up_process(root->fs_info->cleaner_kthread);
1793 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1794
4e121c06
JB		 1795 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1796 &root->fs_info->fs_state)))
1797 btrfs_cleanup_transaction(root);
a0acae0e 1798 if (!try_to_freeze()) {
a74a4b97 1799 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa 1800 if (!kthread_should_stop() &&
914b2007
JK		 1801 (!btrfs_transaction_blocked(root->fs_info) ||
1802 cannot_commit))
8929ecfa 1803 schedule_timeout(delay);
a74a4b97
CM		 1804 __set_current_state(TASK_RUNNING);
1805 }
1806 } while (!kthread_should_stop());
1807 return 0;
1808}
1809
af31f5e5
CM		 1810/*
1811 * this will find the highest generation in the array of
1812 * root backups. The index of the highest array is returned,
1813 * or -1 if we can't find anything.
1814 *
1815 * We check to make sure the array is valid by comparing the
1816 * generation of the latest root in the array with the generation
1817 * in the super block. If they don't match we pitch it.
1818 */
1819static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1820{
1821 u64 cur;
1822 int newest_index = -1;
1823 struct btrfs_root_backup *root_backup;
1824 int i;
1825
1826 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1827 root_backup = info->super_copy->super_roots + i;
1828 cur = btrfs_backup_tree_root_gen(root_backup);
1829 if (cur == newest_gen)
1830 newest_index = i;
1831 }
1832
1833 /* check to see if we actually wrapped around */
1834 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1835 root_backup = info->super_copy->super_roots;
1836 cur = btrfs_backup_tree_root_gen(root_backup);
1837 if (cur == newest_gen)
1838 newest_index = 0;
1839 }
1840 return newest_index;
1841}
1842
1843
1844/*
1845 * find the oldest backup so we know where to store new entries
1846 * in the backup array. This will set the backup_root_index
1847 * field in the fs_info struct
1848 */
1849static void find_oldest_super_backup(struct btrfs_fs_info *info,
1850 u64 newest_gen)
1851{
1852 int newest_index = -1;
1853
1854 newest_index = find_newest_super_backup(info, newest_gen);
1855 /* if there was garbage in there, just move along */
1856 if (newest_index == -1) {
1857 info->backup_root_index = 0;
1858 } else {
1859 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1860 }
1861}
1862
1863/*
1864 * copy all the root pointers into the super backup array.
1865 * this will bump the backup pointer by one when it is
1866 * done
1867 */
1868static void backup_super_roots(struct btrfs_fs_info *info)
1869{
1870 int next_backup;
1871 struct btrfs_root_backup *root_backup;
1872 int last_backup;
1873
1874 next_backup = info->backup_root_index;
1875 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1876 BTRFS_NUM_BACKUP_ROOTS;
1877
1878 /*
1879 * just overwrite the last backup if we're at the same generation
1880 * this happens only at umount
1881 */
1882 root_backup = info->super_for_commit->super_roots + last_backup;
1883 if (btrfs_backup_tree_root_gen(root_backup) ==
1884 btrfs_header_generation(info->tree_root->node))
1885 next_backup = last_backup;
1886
1887 root_backup = info->super_for_commit->super_roots + next_backup;
1888
1889 /*
1890 * make sure all of our padding and empty slots get zero filled
1891 * regardless of which ones we use today
1892 */
1893 memset(root_backup, 0, sizeof(*root_backup));
1894
1895 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1896
1897 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1898 btrfs_set_backup_tree_root_gen(root_backup,
1899 btrfs_header_generation(info->tree_root->node));
1900
1901 btrfs_set_backup_tree_root_level(root_backup,
1902 btrfs_header_level(info->tree_root->node));
1903
1904 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1905 btrfs_set_backup_chunk_root_gen(root_backup,
1906 btrfs_header_generation(info->chunk_root->node));
1907 btrfs_set_backup_chunk_root_level(root_backup,
1908 btrfs_header_level(info->chunk_root->node));
1909
1910 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1911 btrfs_set_backup_extent_root_gen(root_backup,
1912 btrfs_header_generation(info->extent_root->node));
1913 btrfs_set_backup_extent_root_level(root_backup,
1914 btrfs_header_level(info->extent_root->node));
1915
7c7e82a7
CM		 1916 /*
1917 * we might commit during log recovery, which happens before we set
1918 * the fs_root. Make sure it is valid before we fill it in.
1919 */
1920 if (info->fs_root && info->fs_root->node) {
1921 btrfs_set_backup_fs_root(root_backup,
1922 info->fs_root->node->start);
1923 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 1924 btrfs_header_generation(info->fs_root->node));
7c7e82a7 1925 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 1926 btrfs_header_level(info->fs_root->node));
7c7e82a7 1927 }
af31f5e5
CM		 1928
1929 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1930 btrfs_set_backup_dev_root_gen(root_backup,
1931 btrfs_header_generation(info->dev_root->node));
1932 btrfs_set_backup_dev_root_level(root_backup,
1933 btrfs_header_level(info->dev_root->node));
1934
1935 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1936 btrfs_set_backup_csum_root_gen(root_backup,
1937 btrfs_header_generation(info->csum_root->node));
1938 btrfs_set_backup_csum_root_level(root_backup,
1939 btrfs_header_level(info->csum_root->node));
1940
1941 btrfs_set_backup_total_bytes(root_backup,
1942 btrfs_super_total_bytes(info->super_copy));
1943 btrfs_set_backup_bytes_used(root_backup,
1944 btrfs_super_bytes_used(info->super_copy));
1945 btrfs_set_backup_num_devices(root_backup,
1946 btrfs_super_num_devices(info->super_copy));
1947
1948 /*
1949 * if we don't copy this out to the super_copy, it won't get remembered
1950 * for the next commit
1951 */
1952 memcpy(&info->super_copy->super_roots,
1953 &info->super_for_commit->super_roots,
1954 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1955}
1956
1957/*
1958 * this copies info out of the root backup array and back into
1959 * the in-memory super block. It is meant to help iterate through
1960 * the array, so you send it the number of backups you've already
1961 * tried and the last backup index you used.
1962 *
1963 * this returns -1 when it has tried all the backups
1964 */
1965static noinline int next_root_backup(struct btrfs_fs_info *info,
1966 struct btrfs_super_block *super,
1967 int *num_backups_tried, int *backup_index)
1968{
1969 struct btrfs_root_backup *root_backup;
1970 int newest = *backup_index;
1971
1972 if (*num_backups_tried == 0) {
1973 u64 gen = btrfs_super_generation(super);
1974
1975 newest = find_newest_super_backup(info, gen);
1976 if (newest == -1)
1977 return -1;
1978
1979 *backup_index = newest;
1980 *num_backups_tried = 1;
1981 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
1982 /* we've tried all the backups, all done */
1983 return -1;
1984 } else {
1985 /* jump to the next oldest backup */
1986 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
1987 BTRFS_NUM_BACKUP_ROOTS;
1988 *backup_index = newest;
1989 *num_backups_tried += 1;
1990 }
1991 root_backup = super->super_roots + newest;
1992
1993 btrfs_set_super_generation(super,
1994 btrfs_backup_tree_root_gen(root_backup));
1995 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
1996 btrfs_set_super_root_level(super,
1997 btrfs_backup_tree_root_level(root_backup));
1998 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
1999
2000 /*
2001 * fixme: the total bytes and num_devices need to match or we should
2002 * need a fsck
2003 */
2004 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2005 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2006 return 0;
2007}
2008
7abadb64
LB		 2009/* helper to cleanup workers */
2010static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2011{
2012 btrfs_stop_workers(&fs_info->generic_worker);
2013 btrfs_stop_workers(&fs_info->fixup_workers);
2014 btrfs_stop_workers(&fs_info->delalloc_workers);
2015 btrfs_stop_workers(&fs_info->workers);
2016 btrfs_stop_workers(&fs_info->endio_workers);
2017 btrfs_stop_workers(&fs_info->endio_meta_workers);
2018 btrfs_stop_workers(&fs_info->endio_raid56_workers);
2019 btrfs_stop_workers(&fs_info->rmw_workers);
2020 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
2021 btrfs_stop_workers(&fs_info->endio_write_workers);
2022 btrfs_stop_workers(&fs_info->endio_freespace_worker);
2023 btrfs_stop_workers(&fs_info->submit_workers);
2024 btrfs_stop_workers(&fs_info->delayed_workers);
2025 btrfs_stop_workers(&fs_info->caching_workers);
2026 btrfs_stop_workers(&fs_info->readahead_workers);
2027 btrfs_stop_workers(&fs_info->flush_workers);
2f232036 2028 btrfs_stop_workers(&fs_info->qgroup_rescan_workers);
7abadb64
LB		 2029}
2030
2e9f5954
R		 2031static void free_root_extent_buffers(struct btrfs_root *root)
2032{
2033 if (root) {
2034 free_extent_buffer(root->node);
2035 free_extent_buffer(root->commit_root);
2036 root->node = NULL;
2037 root->commit_root = NULL;
2038 }
2039}
2040
af31f5e5
CM		 2041/* helper to cleanup tree roots */
2042static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2043{
2e9f5954 2044 free_root_extent_buffers(info->tree_root);
655b09fe 2045
2e9f5954
R		 2046 free_root_extent_buffers(info->dev_root);
2047 free_root_extent_buffers(info->extent_root);
2048 free_root_extent_buffers(info->csum_root);
2049 free_root_extent_buffers(info->quota_root);
2050 free_root_extent_buffers(info->uuid_root);
2051 if (chunk_root)
2052 free_root_extent_buffers(info->chunk_root);
af31f5e5
CM		 2053}
2054
171f6537
JB		 2055static void del_fs_roots(struct btrfs_fs_info *fs_info)
2056{
2057 int ret;
2058 struct btrfs_root *gang[8];
2059 int i;
2060
2061 while (!list_empty(&fs_info->dead_roots)) {
2062 gang[0] = list_entry(fs_info->dead_roots.next,
2063 struct btrfs_root, root_list);
2064 list_del(&gang[0]->root_list);
2065
2066 if (gang[0]->in_radix) {
cb517eab 2067 btrfs_drop_and_free_fs_root(fs_info, gang[0]);
171f6537
JB		 2068 } else {
2069 free_extent_buffer(gang[0]->node);
2070 free_extent_buffer(gang[0]->commit_root);
b0feb9d9 2071 btrfs_put_fs_root(gang[0]);
171f6537
JB		 2072 }
2073 }
2074
2075 while (1) {
2076 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2077 (void **)gang, 0,
2078 ARRAY_SIZE(gang));
2079 if (!ret)
2080 break;
2081 for (i = 0; i < ret; i++)
cb517eab 2082 btrfs_drop_and_free_fs_root(fs_info, gang[i]);
171f6537
JB		 2083 }
2084}
af31f5e5 2085
ad2b2c80
AV		 2086int open_ctree(struct super_block *sb,
2087 struct btrfs_fs_devices *fs_devices,
2088 char *options)
2e635a27 2089{
db94535d
CM		 2090 u32 sectorsize;
2091 u32 nodesize;
2092 u32 leafsize;
2093 u32 blocksize;
87ee04eb 2094 u32 stripesize;
84234f3a 2095 u64 generation;
f2b636e8 2096 u64 features;
3de4586c 2097 struct btrfs_key location;
a061fc8d 2098 struct buffer_head *bh;
4d34b278 2099 struct btrfs_super_block *disk_super;
815745cf 2100 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2101 struct btrfs_root *tree_root;
4d34b278
ID		 2102 struct btrfs_root *extent_root;
2103 struct btrfs_root *csum_root;
2104 struct btrfs_root *chunk_root;
2105 struct btrfs_root *dev_root;
bcef60f2 2106 struct btrfs_root *quota_root;
f7a81ea4 2107 struct btrfs_root *uuid_root;
e02119d5 2108 struct btrfs_root *log_tree_root;
eb60ceac 2109 int ret;
e58ca020 2110 int err = -EINVAL;
af31f5e5
CM		 2111 int num_backups_tried = 0;
2112 int backup_index = 0;
70f80175
SB		 2113 bool create_uuid_tree;
2114 bool check_uuid_tree;
4543df7e 2115
f84a8bd6 2116 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
6f07e42e 2117 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
cb517eab 2118 if (!tree_root || !chunk_root) {
39279cc3
CM		 2119 err = -ENOMEM;
2120 goto fail;
2121 }
76dda93c
YZ		 2122
2123 ret = init_srcu_struct(&fs_info->subvol_srcu);
2124 if (ret) {
2125 err = ret;
2126 goto fail;
2127 }
2128
2129 ret = setup_bdi(fs_info, &fs_info->bdi);
2130 if (ret) {
2131 err = ret;
2132 goto fail_srcu;
2133 }
2134
e2d84521
MX		 2135 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2136 if (ret) {
2137 err = ret;
2138 goto fail_bdi;
2139 }
2140 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2141 (1 + ilog2(nr_cpu_ids));
2142
963d678b
MX		 2143 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2144 if (ret) {
2145 err = ret;
2146 goto fail_dirty_metadata_bytes;
2147 }
2148
76dda93c
YZ		 2149 fs_info->btree_inode = new_inode(sb);
2150 if (!fs_info->btree_inode) {
2151 err = -ENOMEM;
963d678b 2152 goto fail_delalloc_bytes;
76dda93c
YZ		 2153 }
2154
a6591715 2155 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2156
76dda93c 2157 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
8fd17795 2158 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2159 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2160 INIT_LIST_HEAD(&fs_info->delayed_iputs);
eb73c1b7 2161 INIT_LIST_HEAD(&fs_info->delalloc_roots);
11833d66 2162 INIT_LIST_HEAD(&fs_info->caching_block_groups);
eb73c1b7 2163 spin_lock_init(&fs_info->delalloc_root_lock);
a4abeea4 2164 spin_lock_init(&fs_info->trans_lock);
76dda93c 2165 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2166 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2167 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 2168 spin_lock_init(&fs_info->free_chunk_lock);
f29021b2 2169 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2170 spin_lock_init(&fs_info->super_lock);
f29021b2 2171 rwlock_init(&fs_info->tree_mod_log_lock);
7585717f 2172 mutex_init(&fs_info->reloc_mutex);
de98ced9 2173 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2174
58176a96 2175 init_completion(&fs_info->kobj_unregister);
0b86a832 2176 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2177 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2178 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
0b86a832 2179 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX		 2180 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2181 BTRFS_BLOCK_RSV_GLOBAL);
2182 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2183 BTRFS_BLOCK_RSV_DELALLOC);
2184 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2185 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2186 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2187 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2188 BTRFS_BLOCK_RSV_DELOPS);
cb03c743 2189 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 2190 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 2191 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 2192 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 2193 atomic_set(&fs_info->defrag_running, 0);
fc36ed7e 2194 atomic64_set(&fs_info->tree_mod_seq, 0);
e20d96d6 2195 fs_info->sb = sb;
6f568d35 2196 fs_info->max_inline = 8192 * 1024;
9ed74f2d 2197 fs_info->metadata_ratio = 0;
4cb5300b 2198 fs_info->defrag_inodes = RB_ROOT;
2bf64758 2199 fs_info->free_chunk_space = 0;
f29021b2 2200 fs_info->tree_mod_log = RB_ROOT;
8b87dc17 2201 fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
c8b97818 2202
90519d66
AJ		 2203 /* readahead state */
2204 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2205 spin_lock_init(&fs_info->reada_lock);
c8b97818 2206
b34b086c
CM		 2207 fs_info->thread_pool_size = min_t(unsigned long,
2208 num_online_cpus() + 2, 8);
0afbaf8c 2209
199c2a9c
MX		 2210 INIT_LIST_HEAD(&fs_info->ordered_roots);
2211 spin_lock_init(&fs_info->ordered_root_lock);
16cdcec7
MX		 2212 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2213 GFP_NOFS);
2214 if (!fs_info->delayed_root) {
2215 err = -ENOMEM;
2216 goto fail_iput;
2217 }
2218 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2219
a2de733c
AJ		 2220 mutex_init(&fs_info->scrub_lock);
2221 atomic_set(&fs_info->scrubs_running, 0);
2222 atomic_set(&fs_info->scrub_pause_req, 0);
2223 atomic_set(&fs_info->scrubs_paused, 0);
2224 atomic_set(&fs_info->scrub_cancel_req, 0);
2225 init_waitqueue_head(&fs_info->scrub_pause_wait);
a2de733c 2226 fs_info->scrub_workers_refcnt = 0;
21adbd5c
SB		 2227#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2228 fs_info->check_integrity_print_mask = 0;
2229#endif
a2de733c 2230
c9e9f97b
ID		 2231 spin_lock_init(&fs_info->balance_lock);
2232 mutex_init(&fs_info->balance_mutex);
837d5b6e
ID		 2233 atomic_set(&fs_info->balance_running, 0);
2234 atomic_set(&fs_info->balance_pause_req, 0);
a7e99c69 2235 atomic_set(&fs_info->balance_cancel_req, 0);
c9e9f97b 2236 fs_info->balance_ctl = NULL;
837d5b6e 2237 init_waitqueue_head(&fs_info->balance_wait_q);
a2de733c 2238
a061fc8d
CM		 2239 sb->s_blocksize = 4096;
2240 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 2241 sb->s_bdi = &fs_info->bdi;
a061fc8d 2242
76dda93c 2243 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
bfe86848 2244 set_nlink(fs_info->btree_inode, 1);
0afbaf8c
CM		 2245 /*
2246 * we set the i_size on the btree inode to the max possible int.
2247 * the real end of the address space is determined by all of
2248 * the devices in the system
2249 */
2250 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 2251 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 2252 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2253
5d4f98a2 2254 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 2255 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 2256 fs_info->btree_inode->i_mapping);
0b32f4bb 2257 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
a8067e02 2258 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM		 2259
2260 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 2261
76dda93c
YZ		 2262 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2263 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2264 sizeof(struct btrfs_key));
72ac3c0d
JB		 2265 set_bit(BTRFS_INODE_DUMMY,
2266 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
778ba82b 2267 btrfs_insert_inode_hash(fs_info->btree_inode);
76dda93c 2268
0f9dd46c 2269 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2270 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2271 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2272
11833d66 2273 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 2274 fs_info->btree_inode->i_mapping);
11833d66 2275 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 2276 fs_info->btree_inode->i_mapping);
11833d66 2277 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 2278 fs_info->do_barriers = 1;
e18e4809 2279
39279cc3 2280
5a3f23d5 2281 mutex_init(&fs_info->ordered_operations_mutex);
9ffba8cd 2282 mutex_init(&fs_info->ordered_extent_flush_mutex);
e02119d5 2283 mutex_init(&fs_info->tree_log_mutex);
925baedd 2284 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 2285 mutex_init(&fs_info->transaction_kthread_mutex);
2286 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2287 mutex_init(&fs_info->volume_mutex);
276e680d 2288 init_rwsem(&fs_info->extent_commit_sem);
c71bf099 2289 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2290 init_rwsem(&fs_info->subvol_sem);
803b2f54 2291 sema_init(&fs_info->uuid_tree_rescan_sem, 1);
e922e087
SB		 2292 fs_info->dev_replace.lock_owner = 0;
2293 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2294 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2295 mutex_init(&fs_info->dev_replace.lock_management_lock);
2296 mutex_init(&fs_info->dev_replace.lock);
fa9c0d79 2297
416ac51d 2298 spin_lock_init(&fs_info->qgroup_lock);
f2f6ed3d 2299 mutex_init(&fs_info->qgroup_ioctl_lock);
416ac51d
AJ		 2300 fs_info->qgroup_tree = RB_ROOT;
2301 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2302 fs_info->qgroup_seq = 1;
2303 fs_info->quota_enabled = 0;
2304 fs_info->pending_quota_state = 0;
1e8f9158 2305 fs_info->qgroup_ulist = NULL;
2f232036 2306 mutex_init(&fs_info->qgroup_rescan_lock);
416ac51d 2307
fa9c0d79
CM		 2308 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2309 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2310
e6dcd2dc 2311 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2312 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2313 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2314 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2315
53b381b3
DW		 2316 ret = btrfs_alloc_stripe_hash_table(fs_info);
2317 if (ret) {
83c8266a 2318 err = ret;
53b381b3
DW		 2319 goto fail_alloc;
2320 }
2321
0b86a832 2322 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 2323 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2324
3c4bb26b 2325 invalidate_bdev(fs_devices->latest_bdev);
1104a885
DS		 2326
2327 /*
2328 * Read super block and check the signature bytes only
2329 */
a512bbf8 2330 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY		 2331 if (!bh) {
2332 err = -EINVAL;
16cdcec7 2333 goto fail_alloc;
20b45077 2334 }
39279cc3 2335
1104a885
DS		 2336 /*
2337 * We want to check superblock checksum, the type is stored inside.
2338 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2339 */
2340 if (btrfs_check_super_csum(bh->b_data)) {
2341 printk(KERN_ERR "btrfs: superblock checksum mismatch\n");
2342 err = -EINVAL;
2343 goto fail_alloc;
2344 }
2345
2346 /*
2347 * super_copy is zeroed at allocation time and we never touch the
2348 * following bytes up to INFO_SIZE, the checksum is calculated from
2349 * the whole block of INFO_SIZE
2350 */
6c41761f
DS		 2351 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2352 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2353 sizeof(*fs_info->super_for_commit));
a061fc8d 2354 brelse(bh);
5f39d397 2355
6c41761f 2356 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2357
1104a885
DS		 2358 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2359 if (ret) {
2360 printk(KERN_ERR "btrfs: superblock contains fatal errors\n");
2361 err = -EINVAL;
2362 goto fail_alloc;
2363 }
2364
6c41761f 2365 disk_super = fs_info->super_copy;
0f7d52f4 2366 if (!btrfs_super_root(disk_super))
16cdcec7 2367 goto fail_alloc;
0f7d52f4 2368
acce952b 2369 /* check FS state, whether FS is broken. */
87533c47
MX		 2370 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2371 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2372
af31f5e5
CM		 2373 /*
2374 * run through our array of backup supers and setup
2375 * our ring pointer to the oldest one
2376 */
2377 generation = btrfs_super_generation(disk_super);
2378 find_oldest_super_backup(fs_info, generation);
2379
75e7cb7f
LB		 2380 /*
2381 * In the long term, we'll store the compression type in the super
2382 * block, and it'll be used for per file compression control.
2383 */
2384 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2385
2b82032c
YZ		 2386 ret = btrfs_parse_options(tree_root, options);
2387 if (ret) {
2388 err = ret;
16cdcec7 2389 goto fail_alloc;
2b82032c 2390 }
dfe25020 2391
f2b636e8
JB		 2392 features = btrfs_super_incompat_flags(disk_super) &
2393 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2394 if (features) {
2395 printk(KERN_ERR "BTRFS: couldn't mount because of "
2396 "unsupported optional features (%Lx).\n",
c1c9ff7c 2397 features);
f2b636e8 2398 err = -EINVAL;
16cdcec7 2399 goto fail_alloc;
f2b636e8
JB		 2400 }
2401
727011e0
CM		 2402 if (btrfs_super_leafsize(disk_super) !=
2403 btrfs_super_nodesize(disk_super)) {
2404 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2405 "blocksizes don't match. node %d leaf %d\n",
2406 btrfs_super_nodesize(disk_super),
2407 btrfs_super_leafsize(disk_super));
2408 err = -EINVAL;
2409 goto fail_alloc;
2410 }
2411 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2412 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2413 "blocksize (%d) was too large\n",
2414 btrfs_super_leafsize(disk_super));
2415 err = -EINVAL;
2416 goto fail_alloc;
2417 }
2418
5d4f98a2 2419 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2420 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
69e380d1 2421 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2422 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
727011e0 2423
3173a18f
JB		 2424 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
2425 printk(KERN_ERR "btrfs: has skinny extents\n");
2426
727011e0
CM		 2427 /*
2428 * flag our filesystem as having big metadata blocks if
2429 * they are bigger than the page size
2430 */
2431 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2432 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
2433 printk(KERN_INFO "btrfs flagging fs with big metadata feature\n");
2434 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2435 }
2436
bc3f116f
CM		 2437 nodesize = btrfs_super_nodesize(disk_super);
2438 leafsize = btrfs_super_leafsize(disk_super);
2439 sectorsize = btrfs_super_sectorsize(disk_super);
2440 stripesize = btrfs_super_stripesize(disk_super);
e2d84521 2441 fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids));
963d678b 2442 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f
CM		 2443
2444 /*
2445 * mixed block groups end up with duplicate but slightly offset
2446 * extent buffers for the same range. It leads to corruptions
2447 */
2448 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2449 (sectorsize != leafsize)) {
2450 printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes "
2451 "are not allowed for mixed block groups on %s\n",
2452 sb->s_id);
2453 goto fail_alloc;
2454 }
2455
ceda0864
MX		 2456 /*
2457 * Needn't use the lock because there is no other task which will
2458 * update the flag.
2459 */
a6fa6fae 2460 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2461
f2b636e8
JB		 2462 features = btrfs_super_compat_ro_flags(disk_super) &
2463 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2464 if (!(sb->s_flags & MS_RDONLY) && features) {
2465 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2466 "unsupported option features (%Lx).\n",
c1c9ff7c 2467 features);
f2b636e8 2468 err = -EINVAL;
16cdcec7 2469 goto fail_alloc;
f2b636e8 2470 }
61d92c32
CM		 2471
2472 btrfs_init_workers(&fs_info->generic_worker,
2473 "genwork", 1, NULL);
2474
5443be45 2475 btrfs_init_workers(&fs_info->workers, "worker",
61d92c32
CM		 2476 fs_info->thread_pool_size,
2477 &fs_info->generic_worker);
c8b97818 2478
771ed689 2479 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
45d5fd14 2480 fs_info->thread_pool_size, NULL);
771ed689 2481
8ccf6f19 2482 btrfs_init_workers(&fs_info->flush_workers, "flush_delalloc",
45d5fd14 2483 fs_info->thread_pool_size, NULL);
8ccf6f19 2484
5443be45 2485 btrfs_init_workers(&fs_info->submit_workers, "submit",
b720d209 2486 min_t(u64, fs_devices->num_devices,
45d5fd14 2487 fs_info->thread_pool_size), NULL);
61b49440 2488
bab39bf9 2489 btrfs_init_workers(&fs_info->caching_workers, "cache",
45d5fd14 2490 fs_info->thread_pool_size, NULL);
bab39bf9 2491
61b49440
CM		 2492 /* a higher idle thresh on the submit workers makes it much more
2493 * likely that bios will be send down in a sane order to the
2494 * devices
2495 */
2496 fs_info->submit_workers.idle_thresh = 64;
53863232 2497
771ed689 2498 fs_info->workers.idle_thresh = 16;
4a69a410 2499 fs_info->workers.ordered = 1;
61b49440 2500
771ed689
CM		 2501 fs_info->delalloc_workers.idle_thresh = 2;
2502 fs_info->delalloc_workers.ordered = 1;
2503
61d92c32
CM		 2504 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
2505 &fs_info->generic_worker);
5443be45 2506 btrfs_init_workers(&fs_info->endio_workers, "endio",
61d92c32
CM		 2507 fs_info->thread_pool_size,
2508 &fs_info->generic_worker);
d20f7043 2509 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
61d92c32
CM		 2510 fs_info->thread_pool_size,
2511 &fs_info->generic_worker);
cad321ad 2512 btrfs_init_workers(&fs_info->endio_meta_write_workers,
61d92c32
CM		 2513 "endio-meta-write", fs_info->thread_pool_size,
2514 &fs_info->generic_worker);
53b381b3
DW		 2515 btrfs_init_workers(&fs_info->endio_raid56_workers,
2516 "endio-raid56", fs_info->thread_pool_size,
2517 &fs_info->generic_worker);
2518 btrfs_init_workers(&fs_info->rmw_workers,
2519 "rmw", fs_info->thread_pool_size,
2520 &fs_info->generic_worker);
5443be45 2521 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
61d92c32
CM		 2522 fs_info->thread_pool_size,
2523 &fs_info->generic_worker);
0cb59c99
JB		 2524 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
2525 1, &fs_info->generic_worker);
16cdcec7
MX		 2526 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
2527 fs_info->thread_pool_size,
2528 &fs_info->generic_worker);
90519d66
AJ		 2529 btrfs_init_workers(&fs_info->readahead_workers, "readahead",
2530 fs_info->thread_pool_size,
2531 &fs_info->generic_worker);
2f232036
JS		 2532 btrfs_init_workers(&fs_info->qgroup_rescan_workers, "qgroup-rescan", 1,
2533 &fs_info->generic_worker);
61b49440
CM		 2534
2535 /*
2536 * endios are largely parallel and should have a very
2537 * low idle thresh
2538 */
2539 fs_info->endio_workers.idle_thresh = 4;
b51912c9 2540 fs_info->endio_meta_workers.idle_thresh = 4;
53b381b3
DW		 2541 fs_info->endio_raid56_workers.idle_thresh = 4;
2542 fs_info->rmw_workers.idle_thresh = 2;
b51912c9 2543
9042846b
CM		 2544 fs_info->endio_write_workers.idle_thresh = 2;
2545 fs_info->endio_meta_write_workers.idle_thresh = 2;
90519d66 2546 fs_info->readahead_workers.idle_thresh = 2;
9042846b 2547
0dc3b84a
JB		 2548 /*
2549 * btrfs_start_workers can really only fail because of ENOMEM so just
2550 * return -ENOMEM if any of these fail.
2551 */
2552 ret = btrfs_start_workers(&fs_info->workers);
2553 ret |= btrfs_start_workers(&fs_info->generic_worker);
2554 ret |= btrfs_start_workers(&fs_info->submit_workers);
2555 ret |= btrfs_start_workers(&fs_info->delalloc_workers);
2556 ret |= btrfs_start_workers(&fs_info->fixup_workers);
2557 ret |= btrfs_start_workers(&fs_info->endio_workers);
2558 ret |= btrfs_start_workers(&fs_info->endio_meta_workers);
53b381b3
DW		 2559 ret |= btrfs_start_workers(&fs_info->rmw_workers);
2560 ret |= btrfs_start_workers(&fs_info->endio_raid56_workers);
0dc3b84a
JB		 2561 ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers);
2562 ret |= btrfs_start_workers(&fs_info->endio_write_workers);
2563 ret |= btrfs_start_workers(&fs_info->endio_freespace_worker);
2564 ret |= btrfs_start_workers(&fs_info->delayed_workers);
2565 ret |= btrfs_start_workers(&fs_info->caching_workers);
2566 ret |= btrfs_start_workers(&fs_info->readahead_workers);
8ccf6f19 2567 ret |= btrfs_start_workers(&fs_info->flush_workers);
2f232036 2568 ret |= btrfs_start_workers(&fs_info->qgroup_rescan_workers);
0dc3b84a 2569 if (ret) {
fed425c7 2570 err = -ENOMEM;
0dc3b84a
JB		 2571 goto fail_sb_buffer;
2572 }
4543df7e 2573
4575c9cc 2574 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 2575 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2576 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 2577
db94535d
CM		 2578 tree_root->nodesize = nodesize;
2579 tree_root->leafsize = leafsize;
2580 tree_root->sectorsize = sectorsize;
87ee04eb 2581 tree_root->stripesize = stripesize;
a061fc8d
CM		 2582
2583 sb->s_blocksize = sectorsize;
2584 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 2585
3cae210f 2586 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
d397712b 2587 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 2588 goto fail_sb_buffer;
2589 }
19c00ddc 2590
8d082fb7
LB		 2591 if (sectorsize != PAGE_SIZE) {
2592 printk(KERN_WARNING "btrfs: Incompatible sector size(%lu) "
2593 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
941b2ddf
KM		 2594 goto fail_sb_buffer;
2595 }
2596
925baedd 2597 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 2598 ret = btrfs_read_sys_array(tree_root);
925baedd 2599 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2600 if (ret) {
d397712b
CM		 2601 printk(KERN_WARNING "btrfs: failed to read the system "
2602 "array on %s\n", sb->s_id);
5d4f98a2 2603 goto fail_sb_buffer;
84eed90f 2604 }
0b86a832
CM		 2605
2606 blocksize = btrfs_level_size(tree_root,
2607 btrfs_super_chunk_root_level(disk_super));
84234f3a 2608 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 2609
2610 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2611 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2612
2613 chunk_root->node = read_tree_block(chunk_root,
2614 btrfs_super_chunk_root(disk_super),
84234f3a 2615 blocksize, generation);
416bc658
JB		 2616 if (!chunk_root->node ||
2617 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
83121942
DW		 2618 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
2619 sb->s_id);
af31f5e5 2620 goto fail_tree_roots;
83121942 2621 }
5d4f98a2
YZ		 2622 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2623 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2624
e17cade2 2625 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
b308bc2f 2626 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
e17cade2 2627
0b86a832 2628 ret = btrfs_read_chunk_tree(chunk_root);
2b82032c 2629 if (ret) {
d397712b
CM		 2630 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
2631 sb->s_id);
af31f5e5 2632 goto fail_tree_roots;
2b82032c 2633 }
0b86a832 2634
8dabb742
SB		 2635 /*
2636 * keep the device that is marked to be the target device for the
2637 * dev_replace procedure
2638 */
2639 btrfs_close_extra_devices(fs_info, fs_devices, 0);
dfe25020 2640
a6b0d5c8
CM		 2641 if (!fs_devices->latest_bdev) {
2642 printk(KERN_CRIT "btrfs: failed to read devices on %s\n",
2643 sb->s_id);
2644 goto fail_tree_roots;
2645 }
2646
af31f5e5 2647retry_root_backup:
db94535d
CM		 2648 blocksize = btrfs_level_size(tree_root,
2649 btrfs_super_root_level(disk_super));
84234f3a 2650 generation = btrfs_super_generation(disk_super);
0b86a832 2651
e20d96d6 2652 tree_root->node = read_tree_block(tree_root,
db94535d 2653 btrfs_super_root(disk_super),
84234f3a 2654 blocksize, generation);
af31f5e5
CM		 2655 if (!tree_root->node ||
2656 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
83121942
DW		 2657 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
2658 sb->s_id);
af31f5e5
CM		 2659
2660 goto recovery_tree_root;
83121942 2661 }
af31f5e5 2662
5d4f98a2
YZ		 2663 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2664 tree_root->commit_root = btrfs_root_node(tree_root);
69e9c6c6 2665 btrfs_set_root_refs(&tree_root->root_item, 1);
db94535d 2666
cb517eab
MX		 2667 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2668 location.type = BTRFS_ROOT_ITEM_KEY;
2669 location.offset = 0;
2670
2671 extent_root = btrfs_read_tree_root(tree_root, &location);
2672 if (IS_ERR(extent_root)) {
2673 ret = PTR_ERR(extent_root);
af31f5e5 2674 goto recovery_tree_root;
cb517eab 2675 }
0b86a832 2676 extent_root->track_dirty = 1;
cb517eab 2677 fs_info->extent_root = extent_root;
0b86a832 2678
cb517eab
MX		 2679 location.objectid = BTRFS_DEV_TREE_OBJECTID;
2680 dev_root = btrfs_read_tree_root(tree_root, &location);
2681 if (IS_ERR(dev_root)) {
2682 ret = PTR_ERR(dev_root);
af31f5e5 2683 goto recovery_tree_root;
cb517eab 2684 }
5d4f98a2 2685 dev_root->track_dirty = 1;
cb517eab
MX		 2686 fs_info->dev_root = dev_root;
2687 btrfs_init_devices_late(fs_info);
3768f368 2688
cb517eab
MX		 2689 location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2690 csum_root = btrfs_read_tree_root(tree_root, &location);
2691 if (IS_ERR(csum_root)) {
2692 ret = PTR_ERR(csum_root);
af31f5e5 2693 goto recovery_tree_root;
cb517eab 2694 }
d20f7043 2695 csum_root->track_dirty = 1;
cb517eab 2696 fs_info->csum_root = csum_root;
d20f7043 2697
cb517eab
MX		 2698 location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2699 quota_root = btrfs_read_tree_root(tree_root, &location);
2700 if (!IS_ERR(quota_root)) {
bcef60f2
AJ		 2701 quota_root->track_dirty = 1;
2702 fs_info->quota_enabled = 1;
2703 fs_info->pending_quota_state = 1;
cb517eab 2704 fs_info->quota_root = quota_root;
bcef60f2
AJ		 2705 }
2706
f7a81ea4
SB		 2707 location.objectid = BTRFS_UUID_TREE_OBJECTID;
2708 uuid_root = btrfs_read_tree_root(tree_root, &location);
2709 if (IS_ERR(uuid_root)) {
2710 ret = PTR_ERR(uuid_root);
2711 if (ret != -ENOENT)
2712 goto recovery_tree_root;
2713 create_uuid_tree = true;
70f80175 2714 check_uuid_tree = false;
f7a81ea4
SB		 2715 } else {
2716 uuid_root->track_dirty = 1;
2717 fs_info->uuid_root = uuid_root;
70f80175
SB		 2718 create_uuid_tree = false;
2719 check_uuid_tree =
2720 generation != btrfs_super_uuid_tree_generation(disk_super);
f7a81ea4
SB		 2721 }
2722
8929ecfa
YZ		 2723 fs_info->generation = generation;
2724 fs_info->last_trans_committed = generation;
8929ecfa 2725
68310a5e
ID		 2726 ret = btrfs_recover_balance(fs_info);
2727 if (ret) {
2728 printk(KERN_WARNING "btrfs: failed to recover balance\n");
2729 goto fail_block_groups;
2730 }
2731
733f4fbb
SB		 2732 ret = btrfs_init_dev_stats(fs_info);
2733 if (ret) {
2734 printk(KERN_ERR "btrfs: failed to init dev_stats: %d\n",
2735 ret);
2736 goto fail_block_groups;
2737 }
2738
8dabb742
SB		 2739 ret = btrfs_init_dev_replace(fs_info);
2740 if (ret) {
2741 pr_err("btrfs: failed to init dev_replace: %d\n", ret);
2742 goto fail_block_groups;
2743 }
2744
2745 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2746
5ac1d209
JM		 2747 ret = btrfs_sysfs_add_one(fs_info);
2748 if (ret) {
2749 pr_err("btrfs: failed to init sysfs interface: %d\n", ret);
2750 goto fail_block_groups;
2751 }
2752
c59021f8 2753 ret = btrfs_init_space_info(fs_info);
2754 if (ret) {
2755 printk(KERN_ERR "Failed to initial space info: %d\n", ret);
2756 goto fail_block_groups;
2757 }
2758
1b1d1f66
JB		 2759 ret = btrfs_read_block_groups(extent_root);
2760 if (ret) {
2761 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
2762 goto fail_block_groups;
2763 }
5af3e8cc
SB		 2764 fs_info->num_tolerated_disk_barrier_failures =
2765 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
292fd7fc
SB		 2766 if (fs_info->fs_devices->missing_devices >
2767 fs_info->num_tolerated_disk_barrier_failures &&
2768 !(sb->s_flags & MS_RDONLY)) {
2769 printk(KERN_WARNING
2770 "Btrfs: too many missing devices, writeable mount is not allowed\n");
2771 goto fail_block_groups;
2772 }
9078a3e1 2773
a74a4b97
CM		 2774 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2775 "btrfs-cleaner");
57506d50 2776 if (IS_ERR(fs_info->cleaner_kthread))
1b1d1f66 2777 goto fail_block_groups;
a74a4b97
CM		 2778
2779 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2780 tree_root,
2781 "btrfs-transaction");
57506d50 2782 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2783 goto fail_cleaner;
a74a4b97 2784
c289811c
CM		 2785 if (!btrfs_test_opt(tree_root, SSD) &&
2786 !btrfs_test_opt(tree_root, NOSSD) &&
2787 !fs_info->fs_devices->rotating) {
2788 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
2789 "mode\n");
2790 btrfs_set_opt(fs_info->mount_opt, SSD);
2791 }
2792
21adbd5c
SB		 2793#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2794 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2795 ret = btrfsic_mount(tree_root, fs_devices,
2796 btrfs_test_opt(tree_root,
2797 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2798 1 : 0,
2799 fs_info->check_integrity_print_mask);
2800 if (ret)
2801 printk(KERN_WARNING "btrfs: failed to initialize"
2802 " integrity check module %s\n", sb->s_id);
2803 }
2804#endif
bcef60f2
AJ		 2805 ret = btrfs_read_qgroup_config(fs_info);
2806 if (ret)
2807 goto fail_trans_kthread;
21adbd5c 2808
acce952b 2809 /* do not make disk changes in broken FS */
68ce9682 2810 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM		 2811 u64 bytenr = btrfs_super_log_root(disk_super);
2812
7c2ca468 2813 if (fs_devices->rw_devices == 0) {
d397712b
CM		 2814 printk(KERN_WARNING "Btrfs log replay required "
2815 "on RO media\n");
7c2ca468 2816 err = -EIO;
bcef60f2 2817 goto fail_qgroup;
7c2ca468 2818 }
e02119d5
CM		 2819 blocksize =
2820 btrfs_level_size(tree_root,
2821 btrfs_super_log_root_level(disk_super));
d18a2c44 2822
6f07e42e 2823 log_tree_root = btrfs_alloc_root(fs_info);
676e4c86
DC		 2824 if (!log_tree_root) {
2825 err = -ENOMEM;
bcef60f2 2826 goto fail_qgroup;
676e4c86 2827 }
e02119d5
CM		 2828
2829 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2830 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2831
2832 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 2833 blocksize,
2834 generation + 1);
416bc658
JB		 2835 if (!log_tree_root->node ||
2836 !extent_buffer_uptodate(log_tree_root->node)) {
2837 printk(KERN_ERR "btrfs: failed to read log tree\n");
2838 free_extent_buffer(log_tree_root->node);
2839 kfree(log_tree_root);
2840 goto fail_trans_kthread;
2841 }
79787eaa 2842 /* returns with log_tree_root freed on success */
e02119d5 2843 ret = btrfs_recover_log_trees(log_tree_root);
79787eaa
JM		 2844 if (ret) {
2845 btrfs_error(tree_root->fs_info, ret,
2846 "Failed to recover log tree");
2847 free_extent_buffer(log_tree_root->node);
2848 kfree(log_tree_root);
2849 goto fail_trans_kthread;
2850 }
e556ce2c
YZ		 2851
2852 if (sb->s_flags & MS_RDONLY) {
79787eaa
JM		 2853 ret = btrfs_commit_super(tree_root);
2854 if (ret)
2855 goto fail_trans_kthread;
e556ce2c 2856 }
e02119d5 2857 }
1a40e23b 2858
76dda93c 2859 ret = btrfs_find_orphan_roots(tree_root);
79787eaa
JM		 2860 if (ret)
2861 goto fail_trans_kthread;
76dda93c 2862
7c2ca468 2863 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b 2864 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2
ID		 2865 if (ret)
2866 goto fail_trans_kthread;
d68fc57b 2867
5d4f98a2 2868 ret = btrfs_recover_relocation(tree_root);
d7ce5843
MX		 2869 if (ret < 0) {
2870 printk(KERN_WARNING
2871 "btrfs: failed to recover relocation\n");
2872 err = -EINVAL;
bcef60f2 2873 goto fail_qgroup;
d7ce5843 2874 }
7c2ca468 2875 }
1a40e23b 2876
3de4586c
CM		 2877 location.objectid = BTRFS_FS_TREE_OBJECTID;
2878 location.type = BTRFS_ROOT_ITEM_KEY;
cb517eab 2879 location.offset = 0;
3de4586c 2880
3de4586c 2881 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
3140c9a3
DC		 2882 if (IS_ERR(fs_info->fs_root)) {
2883 err = PTR_ERR(fs_info->fs_root);
bcef60f2 2884 goto fail_qgroup;
3140c9a3 2885 }
c289811c 2886
2b6ba629
ID		 2887 if (sb->s_flags & MS_RDONLY)
2888 return 0;
59641015 2889
2b6ba629
ID		 2890 down_read(&fs_info->cleanup_work_sem);
2891 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2892 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 2893 up_read(&fs_info->cleanup_work_sem);
2b6ba629
ID		 2894 close_ctree(tree_root);
2895 return ret;
2896 }
2897 up_read(&fs_info->cleanup_work_sem);
59641015 2898
2b6ba629
ID		 2899 ret = btrfs_resume_balance_async(fs_info);
2900 if (ret) {
2901 printk(KERN_WARNING "btrfs: failed to resume balance\n");
2902 close_ctree(tree_root);
2903 return ret;
e3acc2a6
JB		 2904 }
2905
8dabb742
SB		 2906 ret = btrfs_resume_dev_replace_async(fs_info);
2907 if (ret) {
2908 pr_warn("btrfs: failed to resume dev_replace\n");
2909 close_ctree(tree_root);
2910 return ret;
2911 }
2912
b382a324
JS		 2913 btrfs_qgroup_rescan_resume(fs_info);
2914
f7a81ea4
SB		 2915 if (create_uuid_tree) {
2916 pr_info("btrfs: creating UUID tree\n");
2917 ret = btrfs_create_uuid_tree(fs_info);
2918 if (ret) {
2919 pr_warn("btrfs: failed to create the UUID tree %d\n",
2920 ret);
2921 close_ctree(tree_root);
2922 return ret;
2923 }
f420ee1e
SB		 2924 } else if (check_uuid_tree ||
2925 btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
70f80175
SB		 2926 pr_info("btrfs: checking UUID tree\n");
2927 ret = btrfs_check_uuid_tree(fs_info);
2928 if (ret) {
2929 pr_warn("btrfs: failed to check the UUID tree %d\n",
2930 ret);
2931 close_ctree(tree_root);
2932 return ret;
2933 }
2934 } else {
2935 fs_info->update_uuid_tree_gen = 1;
f7a81ea4
SB		 2936 }
2937
ad2b2c80 2938 return 0;
39279cc3 2939
bcef60f2
AJ		 2940fail_qgroup:
2941 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM		 2942fail_trans_kthread:
2943 kthread_stop(fs_info->transaction_kthread);
54067ae9 2944 btrfs_cleanup_transaction(fs_info->tree_root);
7b5ff90e 2945 del_fs_roots(fs_info);
3f157a2f 2946fail_cleaner:
a74a4b97 2947 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2948
2949 /*
2950 * make sure we're done with the btree inode before we stop our
2951 * kthreads
2952 */
2953 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 2954
1b1d1f66 2955fail_block_groups:
54067ae9 2956 btrfs_put_block_group_cache(fs_info);
1b1d1f66 2957 btrfs_free_block_groups(fs_info);
af31f5e5
CM		 2958
2959fail_tree_roots:
2960 free_root_pointers(fs_info, 1);
2b8195bb 2961 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 2962
39279cc3 2963fail_sb_buffer:
7abadb64 2964 btrfs_stop_all_workers(fs_info);
16cdcec7 2965fail_alloc:
4543df7e 2966fail_iput:
586e46e2
ID		 2967 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2968
4543df7e 2969 iput(fs_info->btree_inode);
963d678b
MX		 2970fail_delalloc_bytes:
2971 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX		 2972fail_dirty_metadata_bytes:
2973 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
ad081f14 2974fail_bdi:
7e662854 2975 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 2976fail_srcu:
2977 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 2978fail:
53b381b3 2979 btrfs_free_stripe_hash_table(fs_info);
586e46e2 2980 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 2981 return err;
af31f5e5
CM		 2982
2983recovery_tree_root:
af31f5e5
CM		 2984 if (!btrfs_test_opt(tree_root, RECOVERY))
2985 goto fail_tree_roots;
2986
2987 free_root_pointers(fs_info, 0);
2988
2989 /* don't use the log in recovery mode, it won't be valid */
2990 btrfs_set_super_log_root(disk_super, 0);
2991
2992 /* we can't trust the free space cache either */
2993 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
2994
2995 ret = next_root_backup(fs_info, fs_info->super_copy,
2996 &num_backups_tried, &backup_index);
2997 if (ret == -1)
2998 goto fail_block_groups;
2999 goto retry_root_backup;
eb60ceac
CM		 3000}
3001
f2984462
CM		 3002static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3003{
f2984462
CM		 3004 if (uptodate) {
3005 set_buffer_uptodate(bh);
3006 } else {
442a4f63
SB		 3007 struct btrfs_device *device = (struct btrfs_device *)
3008 bh->b_private;
3009
606686ee
JB		 3010 printk_ratelimited_in_rcu(KERN_WARNING "lost page write due to "
3011 "I/O error on %s\n",
3012 rcu_str_deref(device->name));
1259ab75
CM		 3013 /* note, we dont' set_buffer_write_io_error because we have
3014 * our own ways of dealing with the IO errors
3015 */
f2984462 3016 clear_buffer_uptodate(bh);
442a4f63 3017 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM		 3018 }
3019 unlock_buffer(bh);
3020 put_bh(bh);
3021}
3022
a512bbf8
YZ		 3023struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3024{
3025 struct buffer_head *bh;
3026 struct buffer_head *latest = NULL;
3027 struct btrfs_super_block *super;
3028 int i;
3029 u64 transid = 0;
3030 u64 bytenr;
3031
3032 /* we would like to check all the supers, but that would make
3033 * a btrfs mount succeed after a mkfs from a different FS.
3034 * So, we need to add a special mount option to scan for
3035 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3036 */
3037 for (i = 0; i < 1; i++) {
3038 bytenr = btrfs_sb_offset(i);
8068a47e
AJ		 3039 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3040 i_size_read(bdev->bd_inode))
a512bbf8 3041 break;
8068a47e
AJ		 3042 bh = __bread(bdev, bytenr / 4096,
3043 BTRFS_SUPER_INFO_SIZE);
a512bbf8
YZ		 3044 if (!bh)
3045 continue;
3046
3047 super = (struct btrfs_super_block *)bh->b_data;
3048 if (btrfs_super_bytenr(super) != bytenr ||
3cae210f 3049 btrfs_super_magic(super) != BTRFS_MAGIC) {
a512bbf8
YZ		 3050 brelse(bh);
3051 continue;
3052 }
3053
3054 if (!latest || btrfs_super_generation(super) > transid) {
3055 brelse(latest);
3056 latest = bh;
3057 transid = btrfs_super_generation(super);
3058 } else {
3059 brelse(bh);
3060 }
3061 }
3062 return latest;
3063}
3064
4eedeb75
HH		 3065/*
3066 * this should be called twice, once with wait == 0 and
3067 * once with wait == 1. When wait == 0 is done, all the buffer heads
3068 * we write are pinned.
3069 *
3070 * They are released when wait == 1 is done.
3071 * max_mirrors must be the same for both runs, and it indicates how
3072 * many supers on this one device should be written.
3073 *
3074 * max_mirrors == 0 means to write them all.
3075 */
a512bbf8
YZ		 3076static int write_dev_supers(struct btrfs_device *device,
3077 struct btrfs_super_block *sb,
3078 int do_barriers, int wait, int max_mirrors)
3079{
3080 struct buffer_head *bh;
3081 int i;
3082 int ret;
3083 int errors = 0;
3084 u32 crc;
3085 u64 bytenr;
a512bbf8
YZ		 3086
3087 if (max_mirrors == 0)
3088 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3089
a512bbf8
YZ		 3090 for (i = 0; i < max_mirrors; i++) {
3091 bytenr = btrfs_sb_offset(i);
3092 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
3093 break;
3094
3095 if (wait) {
3096 bh = __find_get_block(device->bdev, bytenr / 4096,
3097 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3098 if (!bh) {
3099 errors++;
3100 continue;
3101 }
a512bbf8 3102 wait_on_buffer(bh);
4eedeb75
HH		 3103 if (!buffer_uptodate(bh))
3104 errors++;
3105
3106 /* drop our reference */
3107 brelse(bh);
3108
3109 /* drop the reference from the wait == 0 run */
3110 brelse(bh);
3111 continue;
a512bbf8
YZ		 3112 } else {
3113 btrfs_set_super_bytenr(sb, bytenr);
3114
3115 crc = ~(u32)0;
b0496686 3116 crc = btrfs_csum_data((char *)sb +
a512bbf8
YZ		 3117 BTRFS_CSUM_SIZE, crc,
3118 BTRFS_SUPER_INFO_SIZE -
3119 BTRFS_CSUM_SIZE);
3120 btrfs_csum_final(crc, sb->csum);
3121
4eedeb75
HH		 3122 /*
3123 * one reference for us, and we leave it for the
3124 * caller
3125 */
a512bbf8
YZ		 3126 bh = __getblk(device->bdev, bytenr / 4096,
3127 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 3128 if (!bh) {
3129 printk(KERN_ERR "btrfs: couldn't get super "
3130 "buffer head for bytenr %Lu\n", bytenr);
3131 errors++;
3132 continue;
3133 }
3134
a512bbf8
YZ		 3135 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3136
4eedeb75 3137 /* one reference for submit_bh */
a512bbf8 3138 get_bh(bh);
4eedeb75
HH		 3139
3140 set_buffer_uptodate(bh);
a512bbf8
YZ		 3141 lock_buffer(bh);
3142 bh->b_end_io = btrfs_end_buffer_write_sync;
442a4f63 3143 bh->b_private = device;
a512bbf8
YZ		 3144 }
3145
387125fc
CM		 3146 /*
3147 * we fua the first super. The others we allow
3148 * to go down lazy.
3149 */
21adbd5c 3150 ret = btrfsic_submit_bh(WRITE_FUA, bh);
4eedeb75 3151 if (ret)
a512bbf8 3152 errors++;
a512bbf8
YZ		 3153 }
3154 return errors < i ? 0 : -1;
3155}
3156
387125fc
CM		 3157/*
3158 * endio for the write_dev_flush, this will wake anyone waiting
3159 * for the barrier when it is done
3160 */
3161static void btrfs_end_empty_barrier(struct bio *bio, int err)
3162{
3163 if (err) {
3164 if (err == -EOPNOTSUPP)
3165 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3166 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3167 }
3168 if (bio->bi_private)
3169 complete(bio->bi_private);
3170 bio_put(bio);
3171}
3172
3173/*
3174 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3175 * sent down. With wait == 1, it waits for the previous flush.
3176 *
3177 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3178 * capable
3179 */
3180static int write_dev_flush(struct btrfs_device *device, int wait)
3181{
3182 struct bio *bio;
3183 int ret = 0;
3184
3185 if (device->nobarriers)
3186 return 0;
3187
3188 if (wait) {
3189 bio = device->flush_bio;
3190 if (!bio)
3191 return 0;
3192
3193 wait_for_completion(&device->flush_wait);
3194
3195 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
606686ee
JB		 3196 printk_in_rcu("btrfs: disabling barriers on dev %s\n",
3197 rcu_str_deref(device->name));
387125fc 3198 device->nobarriers = 1;
5af3e8cc 3199 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
387125fc 3200 ret = -EIO;
5af3e8cc
SB		 3201 btrfs_dev_stat_inc_and_print(device,
3202 BTRFS_DEV_STAT_FLUSH_ERRS);
387125fc
CM		 3203 }
3204
3205 /* drop the reference from the wait == 0 run */
3206 bio_put(bio);
3207 device->flush_bio = NULL;
3208
3209 return ret;
3210 }
3211
3212 /*
3213 * one reference for us, and we leave it for the
3214 * caller
3215 */
9c017abc 3216 device->flush_bio = NULL;
9be3395b 3217 bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
387125fc
CM		 3218 if (!bio)
3219 return -ENOMEM;
3220
3221 bio->bi_end_io = btrfs_end_empty_barrier;
3222 bio->bi_bdev = device->bdev;
3223 init_completion(&device->flush_wait);
3224 bio->bi_private = &device->flush_wait;
3225 device->flush_bio = bio;
3226
3227 bio_get(bio);
21adbd5c 3228 btrfsic_submit_bio(WRITE_FLUSH, bio);
387125fc
CM		 3229
3230 return 0;
3231}
3232
3233/*
3234 * send an empty flush down to each device in parallel,
3235 * then wait for them
3236 */
3237static int barrier_all_devices(struct btrfs_fs_info *info)
3238{
3239 struct list_head *head;
3240 struct btrfs_device *dev;
5af3e8cc
SB		 3241 int errors_send = 0;
3242 int errors_wait = 0;
387125fc
CM		 3243 int ret;
3244
3245 /* send down all the barriers */
3246 head = &info->fs_devices->devices;
3247 list_for_each_entry_rcu(dev, head, dev_list) {
3248 if (!dev->bdev) {
5af3e8cc 3249 errors_send++;
387125fc
CM		 3250 continue;
3251 }
3252 if (!dev->in_fs_metadata || !dev->writeable)
3253 continue;
3254
3255 ret = write_dev_flush(dev, 0);
3256 if (ret)
5af3e8cc 3257 errors_send++;
387125fc
CM		 3258 }
3259
3260 /* wait for all the barriers */
3261 list_for_each_entry_rcu(dev, head, dev_list) {
3262 if (!dev->bdev) {
5af3e8cc 3263 errors_wait++;
387125fc
CM		 3264 continue;
3265 }
3266 if (!dev->in_fs_metadata || !dev->writeable)
3267 continue;
3268
3269 ret = write_dev_flush(dev, 1);
3270 if (ret)
5af3e8cc 3271 errors_wait++;
387125fc 3272 }
5af3e8cc
SB		 3273 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3274 errors_wait > info->num_tolerated_disk_barrier_failures)
387125fc
CM		 3275 return -EIO;
3276 return 0;
3277}
3278
5af3e8cc
SB		 3279int btrfs_calc_num_tolerated_disk_barrier_failures(
3280 struct btrfs_fs_info *fs_info)
3281{
3282 struct btrfs_ioctl_space_info space;
3283 struct btrfs_space_info *sinfo;
3284 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3285 BTRFS_BLOCK_GROUP_SYSTEM,
3286 BTRFS_BLOCK_GROUP_METADATA,
3287 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3288 int num_types = 4;
3289 int i;
3290 int c;
3291 int num_tolerated_disk_barrier_failures =
3292 (int)fs_info->fs_devices->num_devices;
3293
3294 for (i = 0; i < num_types; i++) {
3295 struct btrfs_space_info *tmp;
3296
3297 sinfo = NULL;
3298 rcu_read_lock();
3299 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3300 if (tmp->flags == types[i]) {
3301 sinfo = tmp;
3302 break;
3303 }
3304 }
3305 rcu_read_unlock();
3306
3307 if (!sinfo)
3308 continue;
3309
3310 down_read(&sinfo->groups_sem);
3311 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3312 if (!list_empty(&sinfo->block_groups[c])) {
3313 u64 flags;
3314
3315 btrfs_get_block_group_info(
3316 &sinfo->block_groups[c], &space);
3317 if (space.total_bytes == 0 ||
3318 space.used_bytes == 0)
3319 continue;
3320 flags = space.flags;
3321 /*
3322 * return
3323 * 0: if dup, single or RAID0 is configured for
3324 * any of metadata, system or data, else
3325 * 1: if RAID5 is configured, or if RAID1 or
3326 * RAID10 is configured and only two mirrors
3327 * are used, else
3328 * 2: if RAID6 is configured, else
3329 * num_mirrors - 1: if RAID1 or RAID10 is
3330 * configured and more than
3331 * 2 mirrors are used.
3332 */
3333 if (num_tolerated_disk_barrier_failures > 0 &&
3334 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3335 BTRFS_BLOCK_GROUP_RAID0)) ||
3336 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3337 == 0)))
3338 num_tolerated_disk_barrier_failures = 0;
53b381b3
DW		 3339 else if (num_tolerated_disk_barrier_failures > 1) {
3340 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3341 BTRFS_BLOCK_GROUP_RAID5 |
3342 BTRFS_BLOCK_GROUP_RAID10)) {
3343 num_tolerated_disk_barrier_failures = 1;
3344 } else if (flags &
15b0a89d 3345 BTRFS_BLOCK_GROUP_RAID6) {
53b381b3
DW		 3346 num_tolerated_disk_barrier_failures = 2;
3347 }
3348 }
5af3e8cc
SB		 3349 }
3350 }
3351 up_read(&sinfo->groups_sem);
3352 }
3353
3354 return num_tolerated_disk_barrier_failures;
3355}
3356
48a3b636 3357static int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 3358{
e5e9a520 3359 struct list_head *head;
f2984462 3360 struct btrfs_device *dev;
a061fc8d 3361 struct btrfs_super_block *sb;
f2984462 3362 struct btrfs_dev_item *dev_item;
f2984462
CM		 3363 int ret;
3364 int do_barriers;
a236aed1
CM		 3365 int max_errors;
3366 int total_errors = 0;
a061fc8d 3367 u64 flags;
f2984462
CM		 3368
3369 do_barriers = !btrfs_test_opt(root, NOBARRIER);
af31f5e5 3370 backup_super_roots(root->fs_info);
f2984462 3371
6c41761f 3372 sb = root->fs_info->super_for_commit;
a061fc8d 3373 dev_item = &sb->dev_item;
e5e9a520 3374
174ba509 3375 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 3376 head = &root->fs_info->fs_devices->devices;
d7306801 3377 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
387125fc 3378
5af3e8cc
SB		 3379 if (do_barriers) {
3380 ret = barrier_all_devices(root->fs_info);
3381 if (ret) {
3382 mutex_unlock(
3383 &root->fs_info->fs_devices->device_list_mutex);
3384 btrfs_error(root->fs_info, ret,
3385 "errors while submitting device barriers.");
3386 return ret;
3387 }
3388 }
387125fc 3389
1f78160c 3390 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3391 if (!dev->bdev) {
3392 total_errors++;
3393 continue;
3394 }
2b82032c 3395 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3396 continue;
3397
2b82032c 3398 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 3399 btrfs_set_stack_device_type(dev_item, dev->type);
3400 btrfs_set_stack_device_id(dev_item, dev->devid);
3401 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
3402 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
3403 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3404 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3405 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3406 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 3407 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 3408
a061fc8d
CM		 3409 flags = btrfs_super_flags(sb);
3410 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3411
a512bbf8 3412 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 3413 if (ret)
3414 total_errors++;
f2984462 3415 }
a236aed1 3416 if (total_errors > max_errors) {
d397712b
CM		 3417 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
3418 total_errors);
a724b436 3419 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
79787eaa 3420
9d565ba4
SB		 3421 /* FUA is masked off if unsupported and can't be the reason */
3422 btrfs_error(root->fs_info, -EIO,
3423 "%d errors while writing supers", total_errors);
3424 return -EIO;
a236aed1 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{
f570e757 3450 return write_all_supers(root, max_mirrors);
cfaa7295
CM		 3451}
3452
cb517eab
MX		 3453/* Drop a fs root from the radix tree and free it. */
3454void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3455 struct btrfs_root *root)
2619ba1f 3456{
4df27c4d 3457 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 3458 radix_tree_delete(&fs_info->fs_roots_radix,
3459 (unsigned long)root->root_key.objectid);
4df27c4d 3460 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 3461
3462 if (btrfs_root_refs(&root->root_item) == 0)
3463 synchronize_srcu(&fs_info->subvol_srcu);
3464
d7634482 3465 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
3321719e
LB		 3466 btrfs_free_log(NULL, root);
3467 btrfs_free_log_root_tree(NULL, fs_info);
3468 }
3469
581bb050
LZ		 3470 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3471 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3472 free_fs_root(root);
4df27c4d
YZ		 3473}
3474
3475static void free_fs_root(struct btrfs_root *root)
3476{
82d5902d 3477 iput(root->cache_inode);
4df27c4d 3478 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
1cb048f5
FDBM		 3479 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3480 root->orphan_block_rsv = NULL;
0ee5dc67
AV		 3481 if (root->anon_dev)
3482 free_anon_bdev(root->anon_dev);
4df27c4d
YZ		 3483 free_extent_buffer(root->node);
3484 free_extent_buffer(root->commit_root);
581bb050
LZ		 3485 kfree(root->free_ino_ctl);
3486 kfree(root->free_ino_pinned);
d397712b 3487 kfree(root->name);
b0feb9d9 3488 btrfs_put_fs_root(root);
2619ba1f
CM		 3489}
3490
cb517eab
MX		 3491void btrfs_free_fs_root(struct btrfs_root *root)
3492{
3493 free_fs_root(root);
2619ba1f
CM		 3494}
3495
c146afad 3496int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3497{
c146afad
YZ		 3498 u64 root_objectid = 0;
3499 struct btrfs_root *gang[8];
3500 int i;
3768f368 3501 int ret;
e089f05c 3502
c146afad
YZ		 3503 while (1) {
3504 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3505 (void **)gang, root_objectid,
3506 ARRAY_SIZE(gang));
3507 if (!ret)
3508 break;
5d4f98a2
YZ		 3509
3510 root_objectid = gang[ret - 1]->root_key.objectid + 1;
c146afad 3511 for (i = 0; i < ret; i++) {
66b4ffd1
JB		 3512 int err;
3513
c146afad 3514 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 3515 err = btrfs_orphan_cleanup(gang[i]);
3516 if (err)
3517 return err;
c146afad
YZ		 3518 }
3519 root_objectid++;
3520 }
3521 return 0;
3522}
a2135011 3523
c146afad
YZ		 3524int btrfs_commit_super(struct btrfs_root *root)
3525{
3526 struct btrfs_trans_handle *trans;
a74a4b97 3527
c146afad 3528 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 3529 btrfs_run_delayed_iputs(root);
c146afad 3530 mutex_unlock(&root->fs_info->cleaner_mutex);
9d1a2a3a 3531 wake_up_process(root->fs_info->cleaner_kthread);
c71bf099
YZ		 3532
3533 /* wait until ongoing cleanup work done */
3534 down_write(&root->fs_info->cleanup_work_sem);
3535 up_write(&root->fs_info->cleanup_work_sem);
3536
7a7eaa40 3537 trans = btrfs_join_transaction(root);
3612b495
TI		 3538 if (IS_ERR(trans))
3539 return PTR_ERR(trans);
d52c1bcc 3540 return btrfs_commit_transaction(trans, root);
c146afad
YZ		 3541}
3542
3543int close_ctree(struct btrfs_root *root)
3544{
3545 struct btrfs_fs_info *fs_info = root->fs_info;
3546 int ret;
3547
3548 fs_info->closing = 1;
3549 smp_mb();
3550
803b2f54
SB		 3551 /* wait for the uuid_scan task to finish */
3552 down(&fs_info->uuid_tree_rescan_sem);
3553 /* avoid complains from lockdep et al., set sem back to initial state */
3554 up(&fs_info->uuid_tree_rescan_sem);
3555
837d5b6e 3556 /* pause restriper - we want to resume on mount */
aa1b8cd4 3557 btrfs_pause_balance(fs_info);
837d5b6e 3558
8dabb742
SB		 3559 btrfs_dev_replace_suspend_for_unmount(fs_info);
3560
aa1b8cd4 3561 btrfs_scrub_cancel(fs_info);
4cb5300b
CM		 3562
3563 /* wait for any defraggers to finish */
3564 wait_event(fs_info->transaction_wait,
3565 (atomic_read(&fs_info->defrag_running) == 0));
3566
3567 /* clear out the rbtree of defraggable inodes */
26176e7c 3568 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3569
c146afad 3570 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 3571 ret = btrfs_commit_super(root);
3572 if (ret)
3573 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
3574 }
3575
87533c47 3576 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
68ce9682 3577 btrfs_error_commit_super(root);
0f7d52f4 3578
300e4f8a
JB		 3579 btrfs_put_block_group_cache(fs_info);
3580
e3029d9f
AV		 3581 kthread_stop(fs_info->transaction_kthread);
3582 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3583
f25784b3
YZ		 3584 fs_info->closing = 2;
3585 smp_mb();
3586
bcef60f2
AJ		 3587 btrfs_free_qgroup_config(root->fs_info);
3588
963d678b
MX		 3589 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
3590 printk(KERN_INFO "btrfs: at unmount delalloc count %lld\n",
3591 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3592 }
bcc63abb 3593
5ac1d209
JM		 3594 btrfs_sysfs_remove_one(fs_info);
3595
c146afad 3596 del_fs_roots(fs_info);
d10c5f31 3597
2b1360da
JB		 3598 btrfs_free_block_groups(fs_info);
3599
96192499
JB		 3600 btrfs_stop_all_workers(fs_info);
3601
13e6c37b 3602 free_root_pointers(fs_info, 1);
9ad6b7bc 3603
13e6c37b 3604 iput(fs_info->btree_inode);
d6bfde87 3605
21adbd5c
SB		 3606#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3607 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3608 btrfsic_unmount(root, fs_info->fs_devices);
3609#endif
3610
dfe25020 3611 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3612 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3613
e2d84521 3614 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 3615 percpu_counter_destroy(&fs_info->delalloc_bytes);
04160088 3616 bdi_destroy(&fs_info->bdi);
76dda93c 3617 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 3618
53b381b3
DW		 3619 btrfs_free_stripe_hash_table(fs_info);
3620
1cb048f5
FDBM		 3621 btrfs_free_block_rsv(root, root->orphan_block_rsv);
3622 root->orphan_block_rsv = NULL;
3623
eb60ceac
CM		 3624 return 0;
3625}
3626
b9fab919
CM		 3627int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3628 int atomic)
5f39d397 3629{
1259ab75 3630 int ret;
727011e0 3631 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 3632
0b32f4bb 3633 ret = extent_buffer_uptodate(buf);
1259ab75
CM		 3634 if (!ret)
3635 return ret;
3636
3637 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM		 3638 parent_transid, atomic);
3639 if (ret == -EAGAIN)
3640 return ret;
1259ab75 3641 return !ret;
5f39d397
CM		 3642}
3643
3644int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 3645{
0b32f4bb 3646 return set_extent_buffer_uptodate(buf);
5f39d397 3647}
6702ed49 3648
5f39d397
CM		 3649void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3650{
06ea65a3 3651 struct btrfs_root *root;
5f39d397 3652 u64 transid = btrfs_header_generation(buf);
b9473439 3653 int was_dirty;
b4ce94de 3654
06ea65a3
JB		 3655#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3656 /*
3657 * This is a fast path so only do this check if we have sanity tests
3658 * enabled. Normal people shouldn't be marking dummy buffers as dirty
3659 * outside of the sanity tests.
3660 */
3661 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3662 return;
3663#endif
3664 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
b9447ef8 3665 btrfs_assert_tree_locked(buf);
31b1a2bd
JL		 3666 if (transid != root->fs_info->generation)
3667 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
d397712b 3668 "found %llu running %llu\n",
c1c9ff7c 3669 buf->start, transid, root->fs_info->generation);
0b32f4bb 3670 was_dirty = set_extent_buffer_dirty(buf);
e2d84521
MX		 3671 if (!was_dirty)
3672 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3673 buf->len,
3674 root->fs_info->dirty_metadata_batch);
eb60ceac
CM		 3675}
3676
b53d3f5d
LB		 3677static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3678 int flush_delayed)
16cdcec7
MX		 3679{
3680 /*
3681 * looks as though older kernels can get into trouble with
3682 * this code, they end up stuck in balance_dirty_pages forever
3683 */
e2d84521 3684 int ret;
16cdcec7
MX		 3685
3686 if (current->flags & PF_MEMALLOC)
3687 return;
3688
b53d3f5d
LB		 3689 if (flush_delayed)
3690 btrfs_balance_delayed_items(root);
16cdcec7 3691
e2d84521
MX		 3692 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3693 BTRFS_DIRTY_METADATA_THRESH);
3694 if (ret > 0) {
d0e1d66b
NJ		 3695 balance_dirty_pages_ratelimited(
3696 root->fs_info->btree_inode->i_mapping);
16cdcec7
MX		 3697 }
3698 return;
3699}
3700
b53d3f5d 3701void btrfs_btree_balance_dirty(struct btrfs_root *root)
35b7e476 3702{
b53d3f5d
LB		 3703 __btrfs_btree_balance_dirty(root, 1);
3704}
585ad2c3 3705
b53d3f5d
LB		 3706void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3707{
3708 __btrfs_btree_balance_dirty(root, 0);
35b7e476 3709}
6b80053d 3710
ca7a79ad 3711int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 3712{
727011e0 3713 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
0b32f4bb 3714 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
6b80053d 3715}
0da5468f 3716
fcd1f065 3717static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 3718 int read_only)
3719{
1104a885
DS		 3720 /*
3721 * Placeholder for checks
3722 */
fcd1f065 3723 return 0;
acce952b 3724}
3725
48a3b636 3726static void btrfs_error_commit_super(struct btrfs_root *root)
acce952b 3727{
acce952b 3728 mutex_lock(&root->fs_info->cleaner_mutex);
3729 btrfs_run_delayed_iputs(root);
3730 mutex_unlock(&root->fs_info->cleaner_mutex);
3731
3732 down_write(&root->fs_info->cleanup_work_sem);
3733 up_write(&root->fs_info->cleanup_work_sem);
3734
3735 /* cleanup FS via transaction */
3736 btrfs_cleanup_transaction(root);
acce952b 3737}
3738
569e0f35
JB		 3739static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
3740 struct btrfs_root *root)
acce952b 3741{
3742 struct btrfs_inode *btrfs_inode;
3743 struct list_head splice;
3744
3745 INIT_LIST_HEAD(&splice);
3746
3747 mutex_lock(&root->fs_info->ordered_operations_mutex);
199c2a9c 3748 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3749
569e0f35 3750 list_splice_init(&t->ordered_operations, &splice);
acce952b 3751 while (!list_empty(&splice)) {
3752 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3753 ordered_operations);
3754
3755 list_del_init(&btrfs_inode->ordered_operations);
199c2a9c 3756 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3757
3758 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3759
199c2a9c 3760 spin_lock(&root->fs_info->ordered_root_lock);
acce952b 3761 }
3762
199c2a9c 3763 spin_unlock(&root->fs_info->ordered_root_lock);
acce952b 3764 mutex_unlock(&root->fs_info->ordered_operations_mutex);
acce952b 3765}
3766
143bede5 3767static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 3768{
acce952b 3769 struct btrfs_ordered_extent *ordered;
acce952b 3770
199c2a9c 3771 spin_lock(&root->ordered_extent_lock);
779880ef
JB		 3772 /*
3773 * This will just short circuit the ordered completion stuff which will
3774 * make sure the ordered extent gets properly cleaned up.
3775 */
199c2a9c 3776 list_for_each_entry(ordered, &root->ordered_extents,
779880ef
JB		 3777 root_extent_list)
3778 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
199c2a9c
MX		 3779 spin_unlock(&root->ordered_extent_lock);
3780}
3781
3782static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3783{
3784 struct btrfs_root *root;
3785 struct list_head splice;
3786
3787 INIT_LIST_HEAD(&splice);
3788
3789 spin_lock(&fs_info->ordered_root_lock);
3790 list_splice_init(&fs_info->ordered_roots, &splice);
3791 while (!list_empty(&splice)) {
3792 root = list_first_entry(&splice, struct btrfs_root,
3793 ordered_root);
1de2cfde
JB		 3794 list_move_tail(&root->ordered_root,
3795 &fs_info->ordered_roots);
199c2a9c
MX		 3796
3797 btrfs_destroy_ordered_extents(root);
3798
3799 cond_resched_lock(&fs_info->ordered_root_lock);
3800 }
3801 spin_unlock(&fs_info->ordered_root_lock);
acce952b 3802}
3803
35a3621b
SB		 3804static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3805 struct btrfs_root *root)
acce952b 3806{
3807 struct rb_node *node;
3808 struct btrfs_delayed_ref_root *delayed_refs;
3809 struct btrfs_delayed_ref_node *ref;
3810 int ret = 0;
3811
3812 delayed_refs = &trans->delayed_refs;
3813
3814 spin_lock(&delayed_refs->lock);
3815 if (delayed_refs->num_entries == 0) {
cfece4db 3816 spin_unlock(&delayed_refs->lock);
acce952b 3817 printk(KERN_INFO "delayed_refs has NO entry\n");
3818 return ret;
3819 }
3820
b939d1ab 3821 while ((node = rb_first(&delayed_refs->root)) != NULL) {
eb12db69 3822 struct btrfs_delayed_ref_head *head = NULL;
e78417d1 3823 bool pin_bytes = false;
acce952b 3824
eb12db69 3825 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
acce952b 3826 atomic_set(&ref->refs, 1);
3827 if (btrfs_delayed_ref_is_head(ref)) {
acce952b 3828
3829 head = btrfs_delayed_node_to_head(ref);
b939d1ab
JB		 3830 if (!mutex_trylock(&head->mutex)) {
3831 atomic_inc(&ref->refs);
3832 spin_unlock(&delayed_refs->lock);
3833
3834 /* Need to wait for the delayed ref to run */
3835 mutex_lock(&head->mutex);
3836 mutex_unlock(&head->mutex);
3837 btrfs_put_delayed_ref(ref);
3838
e18fca73 3839 spin_lock(&delayed_refs->lock);
b939d1ab
JB		 3840 continue;
3841 }
3842
54067ae9 3843 if (head->must_insert_reserved)
e78417d1 3844 pin_bytes = true;
78a6184a 3845 btrfs_free_delayed_extent_op(head->extent_op);
acce952b 3846 delayed_refs->num_heads--;
3847 if (list_empty(&head->cluster))
3848 delayed_refs->num_heads_ready--;
3849 list_del_init(&head->cluster);
acce952b 3850 }
eb12db69 3851
b939d1ab
JB		 3852 ref->in_tree = 0;
3853 rb_erase(&ref->rb_node, &delayed_refs->root);
c46effa6
LB		 3854 if (head)
3855 rb_erase(&head->href_node, &delayed_refs->href_root);
3856
b939d1ab 3857 delayed_refs->num_entries--;
acce952b 3858 spin_unlock(&delayed_refs->lock);
e78417d1
JB		 3859 if (head) {
3860 if (pin_bytes)
3861 btrfs_pin_extent(root, ref->bytenr,
3862 ref->num_bytes, 1);
3863 mutex_unlock(&head->mutex);
3864 }
acce952b 3865 btrfs_put_delayed_ref(ref);
3866
3867 cond_resched();
3868 spin_lock(&delayed_refs->lock);
3869 }
3870
3871 spin_unlock(&delayed_refs->lock);
3872
3873 return ret;
3874}
3875
143bede5 3876static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 3877{
3878 struct btrfs_inode *btrfs_inode;
3879 struct list_head splice;
3880
3881 INIT_LIST_HEAD(&splice);
3882
eb73c1b7
MX		 3883 spin_lock(&root->delalloc_lock);
3884 list_splice_init(&root->delalloc_inodes, &splice);
acce952b 3885
3886 while (!list_empty(&splice)) {
eb73c1b7
MX		 3887 btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3888 delalloc_inodes);
acce952b 3889
3890 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX		 3891 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3892 &btrfs_inode->runtime_flags);
eb73c1b7 3893 spin_unlock(&root->delalloc_lock);
acce952b 3894
3895 btrfs_invalidate_inodes(btrfs_inode->root);
b216cbfb 3896
eb73c1b7 3897 spin_lock(&root->delalloc_lock);
acce952b 3898 }
3899
eb73c1b7
MX		 3900 spin_unlock(&root->delalloc_lock);
3901}
3902
3903static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3904{
3905 struct btrfs_root *root;
3906 struct list_head splice;
3907
3908 INIT_LIST_HEAD(&splice);
3909
3910 spin_lock(&fs_info->delalloc_root_lock);
3911 list_splice_init(&fs_info->delalloc_roots, &splice);
3912 while (!list_empty(&splice)) {
3913 root = list_first_entry(&splice, struct btrfs_root,
3914 delalloc_root);
3915 list_del_init(&root->delalloc_root);
3916 root = btrfs_grab_fs_root(root);
3917 BUG_ON(!root);
3918 spin_unlock(&fs_info->delalloc_root_lock);
3919
3920 btrfs_destroy_delalloc_inodes(root);
3921 btrfs_put_fs_root(root);
3922
3923 spin_lock(&fs_info->delalloc_root_lock);
3924 }
3925 spin_unlock(&fs_info->delalloc_root_lock);
acce952b 3926}
3927
3928static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3929 struct extent_io_tree *dirty_pages,
3930 int mark)
3931{
3932 int ret;
acce952b 3933 struct extent_buffer *eb;
3934 u64 start = 0;
3935 u64 end;
acce952b 3936
3937 while (1) {
3938 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 3939 mark, NULL);
acce952b 3940 if (ret)
3941 break;
3942
3943 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
3944 while (start <= end) {
fd8b2b61
JB		 3945 eb = btrfs_find_tree_block(root, start,
3946 root->leafsize);
69a85bd8 3947 start += root->leafsize;
fd8b2b61 3948 if (!eb)
acce952b 3949 continue;
fd8b2b61 3950 wait_on_extent_buffer_writeback(eb);
acce952b 3951
fd8b2b61
JB		 3952 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
3953 &eb->bflags))
3954 clear_extent_buffer_dirty(eb);
3955 free_extent_buffer_stale(eb);
acce952b 3956 }
3957 }
3958
3959 return ret;
3960}
3961
3962static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3963 struct extent_io_tree *pinned_extents)
3964{
3965 struct extent_io_tree *unpin;
3966 u64 start;
3967 u64 end;
3968 int ret;
ed0eaa14 3969 bool loop = true;
acce952b 3970
3971 unpin = pinned_extents;
ed0eaa14 3972again:
acce952b 3973 while (1) {
3974 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 3975 EXTENT_DIRTY, NULL);
acce952b 3976 if (ret)
3977 break;
3978
3979 /* opt_discard */
5378e607
LD		 3980 if (btrfs_test_opt(root, DISCARD))
3981 ret = btrfs_error_discard_extent(root, start,
3982 end + 1 - start,
3983 NULL);
acce952b 3984
3985 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3986 btrfs_error_unpin_extent_range(root, start, end);
3987 cond_resched();
3988 }
3989
ed0eaa14
LB		 3990 if (loop) {
3991 if (unpin == &root->fs_info->freed_extents[0])
3992 unpin = &root->fs_info->freed_extents[1];
3993 else
3994 unpin = &root->fs_info->freed_extents[0];
3995 loop = false;
3996 goto again;
3997 }
3998
acce952b 3999 return 0;
4000}
4001
49b25e05
JM		 4002void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4003 struct btrfs_root *root)
4004{
724e2315
JB		 4005 btrfs_destroy_ordered_operations(cur_trans, root);
4006
49b25e05 4007 btrfs_destroy_delayed_refs(cur_trans, root);
49b25e05 4008
4a9d8bde 4009 cur_trans->state = TRANS_STATE_COMMIT_START;
d7096fc3 4010 wake_up(&root->fs_info->transaction_blocked_wait);
49b25e05 4011
4a9d8bde 4012 cur_trans->state = TRANS_STATE_UNBLOCKED;
d7096fc3 4013 wake_up(&root->fs_info->transaction_wait);
49b25e05 4014
67cde344
MX		 4015 btrfs_destroy_delayed_inodes(root);
4016 btrfs_assert_delayed_root_empty(root);
49b25e05 4017
49b25e05
JM		 4018 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4019 EXTENT_DIRTY);
6e841e32
LB		 4020 btrfs_destroy_pinned_extent(root,
4021 root->fs_info->pinned_extents);
49b25e05 4022
4a9d8bde
MX		 4023 cur_trans->state =TRANS_STATE_COMPLETED;
4024 wake_up(&cur_trans->commit_wait);
4025
49b25e05
JM		 4026 /*
4027 memset(cur_trans, 0, sizeof(*cur_trans));
4028 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4029 */
4030}
4031
48a3b636 4032static int btrfs_cleanup_transaction(struct btrfs_root *root)
acce952b 4033{
4034 struct btrfs_transaction *t;
acce952b 4035
acce952b 4036 mutex_lock(&root->fs_info->transaction_kthread_mutex);
4037
a4abeea4 4038 spin_lock(&root->fs_info->trans_lock);
724e2315
JB		 4039 while (!list_empty(&root->fs_info->trans_list)) {
4040 t = list_first_entry(&root->fs_info->trans_list,
4041 struct btrfs_transaction, list);
4042 if (t->state >= TRANS_STATE_COMMIT_START) {
4043 atomic_inc(&t->use_count);
4044 spin_unlock(&root->fs_info->trans_lock);
4045 btrfs_wait_for_commit(root, t->transid);
4046 btrfs_put_transaction(t);
4047 spin_lock(&root->fs_info->trans_lock);
4048 continue;
4049 }
4050 if (t == root->fs_info->running_transaction) {
4051 t->state = TRANS_STATE_COMMIT_DOING;
4052 spin_unlock(&root->fs_info->trans_lock);
4053 /*
4054 * We wait for 0 num_writers since we don't hold a trans
4055 * handle open currently for this transaction.
4056 */
4057 wait_event(t->writer_wait,
4058 atomic_read(&t->num_writers) == 0);
4059 } else {
4060 spin_unlock(&root->fs_info->trans_lock);
4061 }
4062 btrfs_cleanup_one_transaction(t, root);
4a9d8bde 4063
724e2315
JB		 4064 spin_lock(&root->fs_info->trans_lock);
4065 if (t == root->fs_info->running_transaction)
4066 root->fs_info->running_transaction = NULL;
acce952b 4067 list_del_init(&t->list);
724e2315 4068 spin_unlock(&root->fs_info->trans_lock);
acce952b 4069
724e2315
JB		 4070 btrfs_put_transaction(t);
4071 trace_btrfs_transaction_commit(root);
4072 spin_lock(&root->fs_info->trans_lock);
4073 }
4074 spin_unlock(&root->fs_info->trans_lock);
4075 btrfs_destroy_all_ordered_extents(root->fs_info);
4076 btrfs_destroy_delayed_inodes(root);
4077 btrfs_assert_delayed_root_empty(root);
4078 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4079 btrfs_destroy_all_delalloc_inodes(root->fs_info);
acce952b 4080 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 4081
4082 return 0;
4083}
4084
d1310b2e 4085static struct extent_io_ops btree_extent_io_ops = {
ce9adaa5 4086 .readpage_end_io_hook = btree_readpage_end_io_hook,
4bb31e92 4087 .readpage_io_failed_hook = btree_io_failed_hook,
0b86a832 4088 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 4089 /* note we're sharing with inode.c for the merge bio hook */
4090 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 4091};
Linux kernel - Deliverables
This page took 0.637299 seconds and 5 git commands to generate.