projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: handle errors returned from get_tree_block_key
[GitHub/mt8127/android_kernel_alcatel_ttab.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>
7e75bf3f 34#include <asm/unaligned.h>
4b4e25f2 35#include "compat.h"
eb60ceac
CM		 36#include "ctree.h"
37#include "disk-io.h"
e089f05c 38#include "transaction.h"
0f7d52f4 39#include "btrfs_inode.h"
0b86a832 40#include "volumes.h"
db94535d 41#include "print-tree.h"
8b712842 42#include "async-thread.h"
925baedd 43#include "locking.h"
e02119d5 44#include "tree-log.h"
fa9c0d79 45#include "free-space-cache.h"
581bb050 46#include "inode-map.h"
21adbd5c 47#include "check-integrity.h"
606686ee 48#include "rcu-string.h"
8dabb742 49#include "dev-replace.h"
53b381b3 50#include "raid56.h"
eb60ceac 51
de0022b9
JB		 52#ifdef CONFIG_X86
53#include <asm/cpufeature.h>
54#endif
55
d1310b2e 56static struct extent_io_ops btree_extent_io_ops;
8b712842 57static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 58static void free_fs_root(struct btrfs_root *root);
fcd1f065 59static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 60 int read_only);
569e0f35
JB		 61static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
62 struct btrfs_root *root);
143bede5 63static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
acce952b 64static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
65 struct btrfs_root *root);
aec8030a 66static void btrfs_evict_pending_snapshots(struct btrfs_transaction *t);
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" },
155 { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" },
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" },
159 { .id = 0, .name_stem = "tree" },
4008c04a 160};
85d4e461
CM		 161
162void __init btrfs_init_lockdep(void)
163{
164 int i, j;
165
166 /* initialize lockdep class names */
167 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
168 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
169
170 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
171 snprintf(ks->names[j], sizeof(ks->names[j]),
172 "btrfs-%s-%02d", ks->name_stem, j);
173 }
174}
175
176void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
177 int level)
178{
179 struct btrfs_lockdep_keyset *ks;
180
181 BUG_ON(level >= ARRAY_SIZE(ks->keys));
182
183 /* find the matching keyset, id 0 is the default entry */
184 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
185 if (ks->id == objectid)
186 break;
187
188 lockdep_set_class_and_name(&eb->lock,
189 &ks->keys[level], ks->names[level]);
190}
191
4008c04a
CM		 192#endif
193
d352ac68
CM		 194/*
195 * extents on the btree inode are pretty simple, there's one extent
196 * that covers the entire device
197 */
b2950863 198static struct extent_map *btree_get_extent(struct inode *inode,
306e16ce 199 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 200 int create)
7eccb903 201{
5f39d397
CM		 202 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
203 struct extent_map *em;
204 int ret;
205
890871be 206 read_lock(&em_tree->lock);
d1310b2e 207 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 208 if (em) {
209 em->bdev =
210 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 211 read_unlock(&em_tree->lock);
5f39d397 212 goto out;
a061fc8d 213 }
890871be 214 read_unlock(&em_tree->lock);
7b13b7b1 215
172ddd60 216 em = alloc_extent_map();
5f39d397
CM		 217 if (!em) {
218 em = ERR_PTR(-ENOMEM);
219 goto out;
220 }
221 em->start = 0;
0afbaf8c 222 em->len = (u64)-1;
c8b97818 223 em->block_len = (u64)-1;
5f39d397 224 em->block_start = 0;
a061fc8d 225 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 226
890871be 227 write_lock(&em_tree->lock);
09a2a8f9 228 ret = add_extent_mapping(em_tree, em, 0);
5f39d397
CM		 229 if (ret == -EEXIST) {
230 free_extent_map(em);
7b13b7b1 231 em = lookup_extent_mapping(em_tree, start, len);
b4f359ab 232 if (!em)
0433f20d 233 em = ERR_PTR(-EIO);
5f39d397 234 } else if (ret) {
7b13b7b1 235 free_extent_map(em);
0433f20d 236 em = ERR_PTR(ret);
5f39d397 237 }
890871be 238 write_unlock(&em_tree->lock);
7b13b7b1 239
5f39d397
CM		 240out:
241 return em;
7eccb903
CM		 242}
243
b0496686 244u32 btrfs_csum_data(char *data, u32 seed, size_t len)
19c00ddc 245{
163e783e 246 return crc32c(seed, data, len);
19c00ddc
CM		 247}
248
249void btrfs_csum_final(u32 crc, char *result)
250{
7e75bf3f 251 put_unaligned_le32(~crc, result);
19c00ddc
CM		 252}
253
d352ac68
CM		 254/*
255 * compute the csum for a btree block, and either verify it or write it
256 * into the csum field of the block.
257 */
19c00ddc
CM		 258static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
259 int verify)
260{
6c41761f 261 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
607d432d 262 char *result = NULL;
19c00ddc
CM		 263 unsigned long len;
264 unsigned long cur_len;
265 unsigned long offset = BTRFS_CSUM_SIZE;
19c00ddc
CM		 266 char *kaddr;
267 unsigned long map_start;
268 unsigned long map_len;
269 int err;
270 u32 crc = ~(u32)0;
607d432d 271 unsigned long inline_result;
19c00ddc
CM		 272
273 len = buf->len - offset;
d397712b 274 while (len > 0) {
19c00ddc 275 err = map_private_extent_buffer(buf, offset, 32,
a6591715 276 &kaddr, &map_start, &map_len);
d397712b 277 if (err)
19c00ddc 278 return 1;
19c00ddc 279 cur_len = min(len, map_len - (offset - map_start));
b0496686 280 crc = btrfs_csum_data(kaddr + offset - map_start,
19c00ddc
CM		 281 crc, cur_len);
282 len -= cur_len;
283 offset += cur_len;
19c00ddc 284 }
607d432d
JB		 285 if (csum_size > sizeof(inline_result)) {
286 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
287 if (!result)
288 return 1;
289 } else {
290 result = (char *)&inline_result;
291 }
292
19c00ddc
CM		 293 btrfs_csum_final(crc, result);
294
295 if (verify) {
607d432d 296 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 297 u32 val;
298 u32 found = 0;
607d432d 299 memcpy(&found, result, csum_size);
e4204ded 300
607d432d 301 read_extent_buffer(buf, &val, 0, csum_size);
7a36ddec 302 printk_ratelimited(KERN_INFO "btrfs: %s checksum verify "
193f284d
CM		 303 "failed on %llu wanted %X found %X "
304 "level %d\n",
305 root->fs_info->sb->s_id,
306 (unsigned long long)buf->start, val, found,
307 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
CM		 347 "found %llu\n",
348 (unsigned long long)eb->start,
349 (unsigned long long)parent_transid,
350 (unsigned long long)btrfs_header_generation(eb));
1259ab75 351 ret = 1;
0b32f4bb 352 clear_extent_buffer_uptodate(eb);
33958dc6 353out:
2ac55d41
JB		 354 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
355 &cached_state, GFP_NOFS);
1259ab75 356 return ret;
1259ab75
CM		 357}
358
d352ac68
CM		 359/*
360 * helper to read a given tree block, doing retries as required when
361 * the checksums don't match and we have alternate mirrors to try.
362 */
f188591e
CM		 363static int btree_read_extent_buffer_pages(struct btrfs_root *root,
364 struct extent_buffer *eb,
ca7a79ad 365 u64 start, u64 parent_transid)
f188591e
CM		 366{
367 struct extent_io_tree *io_tree;
ea466794 368 int failed = 0;
f188591e
CM		 369 int ret;
370 int num_copies = 0;
371 int mirror_num = 0;
ea466794 372 int failed_mirror = 0;
f188591e 373
a826d6dc 374 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
f188591e
CM		 375 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
376 while (1) {
bb82ab88
AJ		 377 ret = read_extent_buffer_pages(io_tree, eb, start,
378 WAIT_COMPLETE,
f188591e 379 btree_get_extent, mirror_num);
256dd1bb
SB		 380 if (!ret) {
381 if (!verify_parent_transid(io_tree, eb,
b9fab919 382 parent_transid, 0))
256dd1bb
SB		 383 break;
384 else
385 ret = -EIO;
386 }
d397712b 387
a826d6dc
JB		 388 /*
389 * This buffer's crc is fine, but its contents are corrupted, so
390 * there is no reason to read the other copies, they won't be
391 * any less wrong.
392 */
393 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
ea466794
JB		 394 break;
395
5d964051 396 num_copies = btrfs_num_copies(root->fs_info,
f188591e 397 eb->start, eb->len);
4235298e 398 if (num_copies == 1)
ea466794 399 break;
4235298e 400
5cf1ab56
JB		 401 if (!failed_mirror) {
402 failed = 1;
403 failed_mirror = eb->read_mirror;
404 }
405
f188591e 406 mirror_num++;
ea466794
JB		 407 if (mirror_num == failed_mirror)
408 mirror_num++;
409
4235298e 410 if (mirror_num > num_copies)
ea466794 411 break;
f188591e 412 }
ea466794 413
c0901581 414 if (failed && !ret && failed_mirror)
ea466794
JB		 415 repair_eb_io_failure(root, eb, failed_mirror);
416
417 return ret;
f188591e 418}
19c00ddc 419
d352ac68 420/*
d397712b
CM		 421 * checksum a dirty tree block before IO. This has extra checks to make sure
422 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 423 */
d397712b 424
b2950863 425static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
19c00ddc 426{
d1310b2e 427 struct extent_io_tree *tree;
4eee4fa4 428 u64 start = page_offset(page);
19c00ddc 429 u64 found_start;
19c00ddc 430 struct extent_buffer *eb;
f188591e 431
d1310b2e 432 tree = &BTRFS_I(page->mapping->host)->io_tree;
19c00ddc 433
4f2de97a
JB		 434 eb = (struct extent_buffer *)page->private;
435 if (page != eb->pages[0])
436 return 0;
19c00ddc
CM		 437 found_start = btrfs_header_bytenr(eb);
438 if (found_start != start) {
55c69072 439 WARN_ON(1);
4f2de97a 440 return 0;
55c69072 441 }
55c69072 442 if (!PageUptodate(page)) {
55c69072 443 WARN_ON(1);
4f2de97a 444 return 0;
19c00ddc 445 }
19c00ddc 446 csum_tree_block(root, eb, 0);
19c00ddc
CM		 447 return 0;
448}
449
2b82032c
YZ		 450static int check_tree_block_fsid(struct btrfs_root *root,
451 struct extent_buffer *eb)
452{
453 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
454 u8 fsid[BTRFS_UUID_SIZE];
455 int ret = 1;
456
457 read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
458 BTRFS_FSID_SIZE);
459 while (fs_devices) {
460 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
461 ret = 0;
462 break;
463 }
464 fs_devices = fs_devices->seed;
465 }
466 return ret;
467}
468
a826d6dc
JB		 469#define CORRUPT(reason, eb, root, slot) \
470 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
471 "root=%llu, slot=%d\n", reason, \
472 (unsigned long long)btrfs_header_bytenr(eb), \
473 (unsigned long long)root->objectid, slot)
474
475static noinline int check_leaf(struct btrfs_root *root,
476 struct extent_buffer *leaf)
477{
478 struct btrfs_key key;
479 struct btrfs_key leaf_key;
480 u32 nritems = btrfs_header_nritems(leaf);
481 int slot;
482
483 if (nritems == 0)
484 return 0;
485
486 /* Check the 0 item */
487 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
488 BTRFS_LEAF_DATA_SIZE(root)) {
489 CORRUPT("invalid item offset size pair", leaf, root, 0);
490 return -EIO;
491 }
492
493 /*
494 * Check to make sure each items keys are in the correct order and their
495 * offsets make sense. We only have to loop through nritems-1 because
496 * we check the current slot against the next slot, which verifies the
497 * next slot's offset+size makes sense and that the current's slot
498 * offset is correct.
499 */
500 for (slot = 0; slot < nritems - 1; slot++) {
501 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
502 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
503
504 /* Make sure the keys are in the right order */
505 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
506 CORRUPT("bad key order", leaf, root, slot);
507 return -EIO;
508 }
509
510 /*
511 * Make sure the offset and ends are right, remember that the
512 * item data starts at the end of the leaf and grows towards the
513 * front.
514 */
515 if (btrfs_item_offset_nr(leaf, slot) !=
516 btrfs_item_end_nr(leaf, slot + 1)) {
517 CORRUPT("slot offset bad", leaf, root, slot);
518 return -EIO;
519 }
520
521 /*
522 * Check to make sure that we don't point outside of the leaf,
523 * just incase all the items are consistent to eachother, but
524 * all point outside of the leaf.
525 */
526 if (btrfs_item_end_nr(leaf, slot) >
527 BTRFS_LEAF_DATA_SIZE(root)) {
528 CORRUPT("slot end outside of leaf", leaf, root, slot);
529 return -EIO;
530 }
531 }
532
533 return 0;
534}
535
b2950863 536static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
5cf1ab56 537 struct extent_state *state, int mirror)
ce9adaa5
CM		 538{
539 struct extent_io_tree *tree;
540 u64 found_start;
541 int found_level;
ce9adaa5
CM		 542 struct extent_buffer *eb;
543 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 544 int ret = 0;
727011e0 545 int reads_done;
ce9adaa5 546
ce9adaa5
CM		 547 if (!page->private)
548 goto out;
d397712b 549
727011e0 550 tree = &BTRFS_I(page->mapping->host)->io_tree;
4f2de97a 551 eb = (struct extent_buffer *)page->private;
d397712b 552
0b32f4bb
JB		 553 /* the pending IO might have been the only thing that kept this buffer
554 * in memory. Make sure we have a ref for all this other checks
555 */
556 extent_buffer_get(eb);
557
558 reads_done = atomic_dec_and_test(&eb->io_pages);
727011e0
CM		 559 if (!reads_done)
560 goto err;
f188591e 561
5cf1ab56 562 eb->read_mirror = mirror;
ea466794
JB		 563 if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
564 ret = -EIO;
565 goto err;
566 }
567
ce9adaa5 568 found_start = btrfs_header_bytenr(eb);
727011e0 569 if (found_start != eb->start) {
7a36ddec 570 printk_ratelimited(KERN_INFO "btrfs bad tree block start "
193f284d
CM		 571 "%llu %llu\n",
572 (unsigned long long)found_start,
573 (unsigned long long)eb->start);
f188591e 574 ret = -EIO;
ce9adaa5
CM		 575 goto err;
576 }
2b82032c 577 if (check_tree_block_fsid(root, eb)) {
7a36ddec 578 printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n",
193f284d 579 (unsigned long long)eb->start);
1259ab75
CM		 580 ret = -EIO;
581 goto err;
582 }
ce9adaa5 583 found_level = btrfs_header_level(eb);
1c24c3ce
JB		 584 if (found_level >= BTRFS_MAX_LEVEL) {
585 btrfs_info(root->fs_info, "bad tree block level %d\n",
586 (int)btrfs_header_level(eb));
587 ret = -EIO;
588 goto err;
589 }
ce9adaa5 590
85d4e461
CM		 591 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
592 eb, found_level);
4008c04a 593
ce9adaa5 594 ret = csum_tree_block(root, eb, 1);
a826d6dc 595 if (ret) {
f188591e 596 ret = -EIO;
a826d6dc
JB		 597 goto err;
598 }
599
600 /*
601 * If this is a leaf block and it is corrupt, set the corrupt bit so
602 * that we don't try and read the other copies of this block, just
603 * return -EIO.
604 */
605 if (found_level == 0 && check_leaf(root, eb)) {
606 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
607 ret = -EIO;
608 }
ce9adaa5 609
0b32f4bb
JB		 610 if (!ret)
611 set_extent_buffer_uptodate(eb);
ce9adaa5 612err:
79fb65a1
JB		 613 if (reads_done &&
614 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 615 btree_readahead_hook(root, eb, eb->start, ret);
4bb31e92 616
53b381b3
DW		 617 if (ret) {
618 /*
619 * our io error hook is going to dec the io pages
620 * again, we have to make sure it has something
621 * to decrement
622 */
623 atomic_inc(&eb->io_pages);
0b32f4bb 624 clear_extent_buffer_uptodate(eb);
53b381b3 625 }
0b32f4bb 626 free_extent_buffer(eb);
ce9adaa5 627out:
f188591e 628 return ret;
ce9adaa5
CM		 629}
630
ea466794 631static int btree_io_failed_hook(struct page *page, int failed_mirror)
4bb31e92 632{
4bb31e92
AJ		 633 struct extent_buffer *eb;
634 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
635
4f2de97a 636 eb = (struct extent_buffer *)page->private;
ea466794 637 set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
5cf1ab56 638 eb->read_mirror = failed_mirror;
53b381b3 639 atomic_dec(&eb->io_pages);
ea466794 640 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
4bb31e92 641 btree_readahead_hook(root, eb, eb->start, -EIO);
4bb31e92
AJ		 642 return -EIO; /* we fixed nothing */
643}
644
ce9adaa5 645static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM		 646{
647 struct end_io_wq *end_io_wq = bio->bi_private;
648 struct btrfs_fs_info *fs_info;
ce9adaa5 649
ce9adaa5 650 fs_info = end_io_wq->info;
ce9adaa5 651 end_io_wq->error = err;
8b712842
CM		 652 end_io_wq->work.func = end_workqueue_fn;
653 end_io_wq->work.flags = 0;
d20f7043 654
7b6d91da 655 if (bio->bi_rw & REQ_WRITE) {
53b381b3 656 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA)
cad321ad
CM		 657 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
658 &end_io_wq->work);
53b381b3 659 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE)
0cb59c99
JB		 660 btrfs_queue_worker(&fs_info->endio_freespace_worker,
661 &end_io_wq->work);
53b381b3
DW		 662 else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
663 btrfs_queue_worker(&fs_info->endio_raid56_workers,
664 &end_io_wq->work);
cad321ad
CM		 665 else
666 btrfs_queue_worker(&fs_info->endio_write_workers,
667 &end_io_wq->work);
d20f7043 668 } else {
53b381b3
DW		 669 if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
670 btrfs_queue_worker(&fs_info->endio_raid56_workers,
671 &end_io_wq->work);
672 else if (end_io_wq->metadata)
d20f7043
CM		 673 btrfs_queue_worker(&fs_info->endio_meta_workers,
674 &end_io_wq->work);
675 else
676 btrfs_queue_worker(&fs_info->endio_workers,
677 &end_io_wq->work);
678 }
ce9adaa5
CM		 679}
680
0cb59c99
JB		 681/*
682 * For the metadata arg you want
683 *
684 * 0 - if data
685 * 1 - if normal metadta
686 * 2 - if writing to the free space cache area
53b381b3 687 * 3 - raid parity work
0cb59c99 688 */
22c59948
CM		 689int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
690 int metadata)
0b86a832 691{
ce9adaa5 692 struct end_io_wq *end_io_wq;
ce9adaa5
CM		 693 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
694 if (!end_io_wq)
695 return -ENOMEM;
696
697 end_io_wq->private = bio->bi_private;
698 end_io_wq->end_io = bio->bi_end_io;
22c59948 699 end_io_wq->info = info;
ce9adaa5
CM		 700 end_io_wq->error = 0;
701 end_io_wq->bio = bio;
22c59948 702 end_io_wq->metadata = metadata;
ce9adaa5
CM		 703
704 bio->bi_private = end_io_wq;
705 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 706 return 0;
707}
708
b64a2851 709unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 710{
4854ddd0
CM		 711 unsigned long limit = min_t(unsigned long,
712 info->workers.max_workers,
713 info->fs_devices->open_devices);
714 return 256 * limit;
715}
0986fe9e 716
4a69a410
CM		 717static void run_one_async_start(struct btrfs_work *work)
718{
4a69a410 719 struct async_submit_bio *async;
79787eaa 720 int ret;
4a69a410
CM		 721
722 async = container_of(work, struct async_submit_bio, work);
79787eaa
JM		 723 ret = async->submit_bio_start(async->inode, async->rw, async->bio,
724 async->mirror_num, async->bio_flags,
725 async->bio_offset);
726 if (ret)
727 async->error = ret;
4a69a410
CM		 728}
729
730static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 731{
732 struct btrfs_fs_info *fs_info;
733 struct async_submit_bio *async;
4854ddd0 734 int limit;
8b712842
CM		 735
736 async = container_of(work, struct async_submit_bio, work);
737 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 738
b64a2851 739 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 740 limit = limit * 2 / 3;
741
66657b31 742 if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
b64a2851 743 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 744 wake_up(&fs_info->async_submit_wait);
745
79787eaa
JM		 746 /* If an error occured we just want to clean up the bio and move on */
747 if (async->error) {
748 bio_endio(async->bio, async->error);
749 return;
750 }
751
4a69a410 752 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 753 async->mirror_num, async->bio_flags,
754 async->bio_offset);
4a69a410
CM		 755}
756
757static void run_one_async_free(struct btrfs_work *work)
758{
759 struct async_submit_bio *async;
760
761 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 762 kfree(async);
763}
764
44b8bd7e
CM		 765int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
766 int rw, struct bio *bio, int mirror_num,
c8b97818 767 unsigned long bio_flags,
eaf25d93 768 u64 bio_offset,
4a69a410
CM		 769 extent_submit_bio_hook_t *submit_bio_start,
770 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 771{
772 struct async_submit_bio *async;
773
774 async = kmalloc(sizeof(*async), GFP_NOFS);
775 if (!async)
776 return -ENOMEM;
777
778 async->inode = inode;
779 async->rw = rw;
780 async->bio = bio;
781 async->mirror_num = mirror_num;
4a69a410
CM		 782 async->submit_bio_start = submit_bio_start;
783 async->submit_bio_done = submit_bio_done;
784
785 async->work.func = run_one_async_start;
786 async->work.ordered_func = run_one_async_done;
787 async->work.ordered_free = run_one_async_free;
788
8b712842 789 async->work.flags = 0;
c8b97818 790 async->bio_flags = bio_flags;
eaf25d93 791 async->bio_offset = bio_offset;
8c8bee1d 792
79787eaa
JM		 793 async->error = 0;
794
cb03c743 795 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 796
7b6d91da 797 if (rw & REQ_SYNC)
d313d7a3
CM		 798 btrfs_set_work_high_prio(&async->work);
799
8b712842 800 btrfs_queue_worker(&fs_info->workers, &async->work);
9473f16c 801
d397712b 802 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 803 atomic_read(&fs_info->nr_async_submits)) {
804 wait_event(fs_info->async_submit_wait,
805 (atomic_read(&fs_info->nr_async_submits) == 0));
806 }
807
44b8bd7e
CM		 808 return 0;
809}
810
ce3ed71a
CM		 811static int btree_csum_one_bio(struct bio *bio)
812{
813 struct bio_vec *bvec = bio->bi_io_vec;
814 int bio_index = 0;
815 struct btrfs_root *root;
79787eaa 816 int ret = 0;
ce3ed71a
CM		 817
818 WARN_ON(bio->bi_vcnt <= 0);
d397712b 819 while (bio_index < bio->bi_vcnt) {
ce3ed71a 820 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
79787eaa
JM		 821 ret = csum_dirty_buffer(root, bvec->bv_page);
822 if (ret)
823 break;
ce3ed71a
CM		 824 bio_index++;
825 bvec++;
826 }
79787eaa 827 return ret;
ce3ed71a
CM		 828}
829
4a69a410
CM		 830static int __btree_submit_bio_start(struct inode *inode, int rw,
831 struct bio *bio, int mirror_num,
eaf25d93
CM		 832 unsigned long bio_flags,
833 u64 bio_offset)
22c59948 834{
8b712842
CM		 835 /*
836 * when we're called for a write, we're already in the async
5443be45 837 * submission context. Just jump into btrfs_map_bio
8b712842 838 */
79787eaa 839 return btree_csum_one_bio(bio);
4a69a410 840}
22c59948 841
4a69a410 842static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 843 int mirror_num, unsigned long bio_flags,
844 u64 bio_offset)
4a69a410 845{
61891923
SB		 846 int ret;
847
8b712842 848 /*
4a69a410
CM		 849 * when we're called for a write, we're already in the async
850 * submission context. Just jump into btrfs_map_bio
8b712842 851 */
61891923
SB		 852 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
853 if (ret)
854 bio_endio(bio, ret);
855 return ret;
0b86a832
CM		 856}
857
de0022b9
JB		 858static int check_async_write(struct inode *inode, unsigned long bio_flags)
859{
860 if (bio_flags & EXTENT_BIO_TREE_LOG)
861 return 0;
862#ifdef CONFIG_X86
863 if (cpu_has_xmm4_2)
864 return 0;
865#endif
866 return 1;
867}
868
44b8bd7e 869static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 870 int mirror_num, unsigned long bio_flags,
871 u64 bio_offset)
44b8bd7e 872{
de0022b9 873 int async = check_async_write(inode, bio_flags);
cad321ad
CM		 874 int ret;
875
7b6d91da 876 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 877 /*
878 * called for a read, do the setup so that checksum validation
879 * can happen in the async kernel threads
880 */
f3f266ab
CM		 881 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
882 bio, 1);
1d4284bd 883 if (ret)
61891923
SB		 884 goto out_w_error;
885 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
886 mirror_num, 0);
de0022b9
JB		 887 } else if (!async) {
888 ret = btree_csum_one_bio(bio);
889 if (ret)
61891923
SB		 890 goto out_w_error;
891 ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
892 mirror_num, 0);
893 } else {
894 /*
895 * kthread helpers are used to submit writes so that
896 * checksumming can happen in parallel across all CPUs
897 */
898 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
899 inode, rw, bio, mirror_num, 0,
900 bio_offset,
901 __btree_submit_bio_start,
902 __btree_submit_bio_done);
44b8bd7e 903 }
d313d7a3 904
61891923
SB		 905 if (ret) {
906out_w_error:
907 bio_endio(bio, ret);
908 }
909 return ret;
44b8bd7e
CM		 910}
911
3dd1462e 912#ifdef CONFIG_MIGRATION
784b4e29 913static int btree_migratepage(struct address_space *mapping,
a6bc32b8
MG		 914 struct page *newpage, struct page *page,
915 enum migrate_mode mode)
784b4e29
CM		 916{
917 /*
918 * we can't safely write a btree page from here,
919 * we haven't done the locking hook
920 */
921 if (PageDirty(page))
922 return -EAGAIN;
923 /*
924 * Buffers may be managed in a filesystem specific way.
925 * We must have no buffers or drop them.
926 */
927 if (page_has_private(page) &&
928 !try_to_release_page(page, GFP_KERNEL))
929 return -EAGAIN;
a6bc32b8 930 return migrate_page(mapping, newpage, page, mode);
784b4e29 931}
3dd1462e 932#endif
784b4e29 933
0da5468f
CM		 934
935static int btree_writepages(struct address_space *mapping,
936 struct writeback_control *wbc)
937{
d1310b2e 938 struct extent_io_tree *tree;
e2d84521
MX		 939 struct btrfs_fs_info *fs_info;
940 int ret;
941
d1310b2e 942 tree = &BTRFS_I(mapping->host)->io_tree;
d8d5f3e1 943 if (wbc->sync_mode == WB_SYNC_NONE) {
448d640b
CM		 944
945 if (wbc->for_kupdate)
946 return 0;
947
e2d84521 948 fs_info = BTRFS_I(mapping->host)->root->fs_info;
b9473439 949 /* this is a bit racy, but that's ok */
e2d84521
MX		 950 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
951 BTRFS_DIRTY_METADATA_THRESH);
952 if (ret < 0)
793955bc 953 return 0;
793955bc 954 }
0b32f4bb 955 return btree_write_cache_pages(mapping, wbc);
0da5468f
CM		 956}
957
b2950863 958static int btree_readpage(struct file *file, struct page *page)
5f39d397 959{
d1310b2e
CM		 960 struct extent_io_tree *tree;
961 tree = &BTRFS_I(page->mapping->host)->io_tree;
8ddc7d9c 962 return extent_read_full_page(tree, page, btree_get_extent, 0);
5f39d397 963}
22b0ebda 964
70dec807 965static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 966{
98509cfc 967 if (PageWriteback(page) || PageDirty(page))
d397712b 968 return 0;
0c4e538b
DS		 969 /*
970 * We need to mask out eg. __GFP_HIGHMEM and __GFP_DMA32 as we're doing
971 * slab allocation from alloc_extent_state down the callchain where
972 * it'd hit a BUG_ON as those flags are not allowed.
973 */
974 gfp_flags &= ~GFP_SLAB_BUG_MASK;
975
3083ee2e 976 return try_release_extent_buffer(page, gfp_flags);
d98237b3
CM		 977}
978
5f39d397 979static void btree_invalidatepage(struct page *page, unsigned long offset)
d98237b3 980{
d1310b2e
CM		 981 struct extent_io_tree *tree;
982 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 983 extent_invalidatepage(tree, page, offset);
984 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 985 if (PagePrivate(page)) {
d397712b
CM		 986 printk(KERN_WARNING "btrfs warning page private not zero "
987 "on page %llu\n", (unsigned long long)page_offset(page));
9ad6b7bc
CM		 988 ClearPagePrivate(page);
989 set_page_private(page, 0);
990 page_cache_release(page);
991 }
d98237b3
CM		 992}
993
0b32f4bb
JB		 994static int btree_set_page_dirty(struct page *page)
995{
bb146eb2 996#ifdef DEBUG
0b32f4bb
JB		 997 struct extent_buffer *eb;
998
999 BUG_ON(!PagePrivate(page));
1000 eb = (struct extent_buffer *)page->private;
1001 BUG_ON(!eb);
1002 BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1003 BUG_ON(!atomic_read(&eb->refs));
1004 btrfs_assert_tree_locked(eb);
bb146eb2 1005#endif
0b32f4bb
JB		 1006 return __set_page_dirty_nobuffers(page);
1007}
1008
7f09410b 1009static const struct address_space_operations btree_aops = {
d98237b3 1010 .readpage = btree_readpage,
0da5468f 1011 .writepages = btree_writepages,
5f39d397
CM		 1012 .releasepage = btree_releasepage,
1013 .invalidatepage = btree_invalidatepage,
5a92bc88 1014#ifdef CONFIG_MIGRATION
784b4e29 1015 .migratepage = btree_migratepage,
5a92bc88 1016#endif
0b32f4bb 1017 .set_page_dirty = btree_set_page_dirty,
d98237b3
CM		 1018};
1019
ca7a79ad
CM		 1020int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1021 u64 parent_transid)
090d1875 1022{
5f39d397
CM		 1023 struct extent_buffer *buf = NULL;
1024 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 1025 int ret = 0;
090d1875 1026
db94535d 1027 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 1028 if (!buf)
090d1875 1029 return 0;
d1310b2e 1030 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
bb82ab88 1031 buf, 0, WAIT_NONE, btree_get_extent, 0);
5f39d397 1032 free_extent_buffer(buf);
de428b63 1033 return ret;
090d1875
CM		 1034}
1035
ab0fff03
AJ		 1036int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1037 int mirror_num, struct extent_buffer **eb)
1038{
1039 struct extent_buffer *buf = NULL;
1040 struct inode *btree_inode = root->fs_info->btree_inode;
1041 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1042 int ret;
1043
1044 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1045 if (!buf)
1046 return 0;
1047
1048 set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1049
1050 ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1051 btree_get_extent, mirror_num);
1052 if (ret) {
1053 free_extent_buffer(buf);
1054 return ret;
1055 }
1056
1057 if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1058 free_extent_buffer(buf);
1059 return -EIO;
0b32f4bb 1060 } else if (extent_buffer_uptodate(buf)) {
ab0fff03
AJ		 1061 *eb = buf;
1062 } else {
1063 free_extent_buffer(buf);
1064 }
1065 return 0;
1066}
1067
0999df54
CM		 1068struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1069 u64 bytenr, u32 blocksize)
1070{
1071 struct inode *btree_inode = root->fs_info->btree_inode;
1072 struct extent_buffer *eb;
1073 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
f09d1f60 1074 bytenr, blocksize);
0999df54
CM		 1075 return eb;
1076}
1077
1078struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1079 u64 bytenr, u32 blocksize)
1080{
1081 struct inode *btree_inode = root->fs_info->btree_inode;
1082 struct extent_buffer *eb;
1083
1084 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
727011e0 1085 bytenr, blocksize);
0999df54
CM		 1086 return eb;
1087}
1088
1089
e02119d5
CM		 1090int btrfs_write_tree_block(struct extent_buffer *buf)
1091{
727011e0 1092 return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
8aa38c31 1093 buf->start + buf->len - 1);
e02119d5
CM		 1094}
1095
1096int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1097{
727011e0 1098 return filemap_fdatawait_range(buf->pages[0]->mapping,
8aa38c31 1099 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1100}
1101
0999df54 1102struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 1103 u32 blocksize, u64 parent_transid)
0999df54
CM		 1104{
1105 struct extent_buffer *buf = NULL;
0999df54
CM		 1106 int ret;
1107
0999df54
CM		 1108 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1109 if (!buf)
1110 return NULL;
0999df54 1111
ca7a79ad 1112 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
5f39d397 1113 return buf;
ce9adaa5 1114
eb60ceac
CM		 1115}
1116
d5c13f92
JM		 1117void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1118 struct extent_buffer *buf)
ed2ff2cb 1119{
e2d84521
MX		 1120 struct btrfs_fs_info *fs_info = root->fs_info;
1121
55c69072 1122 if (btrfs_header_generation(buf) ==
e2d84521 1123 fs_info->running_transaction->transid) {
b9447ef8 1124 btrfs_assert_tree_locked(buf);
b4ce94de 1125
b9473439 1126 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
e2d84521
MX		 1127 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1128 -buf->len,
1129 fs_info->dirty_metadata_batch);
ed7b63eb
JB		 1130 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1131 btrfs_set_lock_blocking(buf);
1132 clear_extent_buffer_dirty(buf);
1133 }
925baedd 1134 }
5f39d397
CM		 1135}
1136
143bede5
JM		 1137static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1138 u32 stripesize, struct btrfs_root *root,
1139 struct btrfs_fs_info *fs_info,
1140 u64 objectid)
d97e63b6 1141{
cfaa7295 1142 root->node = NULL;
a28ec197 1143 root->commit_root = NULL;
db94535d
CM		 1144 root->sectorsize = sectorsize;
1145 root->nodesize = nodesize;
1146 root->leafsize = leafsize;
87ee04eb 1147 root->stripesize = stripesize;
123abc88 1148 root->ref_cows = 0;
0b86a832 1149 root->track_dirty = 0;
c71bf099 1150 root->in_radix = 0;
d68fc57b
YZ		 1151 root->orphan_item_inserted = 0;
1152 root->orphan_cleanup_state = 0;
0b86a832 1153
0f7d52f4
CM		 1154 root->objectid = objectid;
1155 root->last_trans = 0;
13a8a7c8 1156 root->highest_objectid = 0;
58176a96 1157 root->name = NULL;
6bef4d31 1158 root->inode_tree = RB_ROOT;
16cdcec7 1159 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1160 root->block_rsv = NULL;
d68fc57b 1161 root->orphan_block_rsv = NULL;
0b86a832
CM		 1162
1163 INIT_LIST_HEAD(&root->dirty_list);
5d4f98a2 1164 INIT_LIST_HEAD(&root->root_list);
2ab28f32
JB		 1165 INIT_LIST_HEAD(&root->logged_list[0]);
1166 INIT_LIST_HEAD(&root->logged_list[1]);
d68fc57b 1167 spin_lock_init(&root->orphan_lock);
5d4f98a2 1168 spin_lock_init(&root->inode_lock);
f0486c68 1169 spin_lock_init(&root->accounting_lock);
2ab28f32
JB		 1170 spin_lock_init(&root->log_extents_lock[0]);
1171 spin_lock_init(&root->log_extents_lock[1]);
a2135011 1172 mutex_init(&root->objectid_mutex);
e02119d5 1173 mutex_init(&root->log_mutex);
7237f183
YZ		 1174 init_waitqueue_head(&root->log_writer_wait);
1175 init_waitqueue_head(&root->log_commit_wait[0]);
1176 init_waitqueue_head(&root->log_commit_wait[1]);
1177 atomic_set(&root->log_commit[0], 0);
1178 atomic_set(&root->log_commit[1], 0);
1179 atomic_set(&root->log_writers, 0);
2ecb7923 1180 atomic_set(&root->log_batch, 0);
8a35d95f 1181 atomic_set(&root->orphan_inodes, 0);
7237f183 1182 root->log_transid = 0;
257c62e1 1183 root->last_log_commit = 0;
d0c803c4 1184 extent_io_tree_init(&root->dirty_log_pages,
f993c883 1185 fs_info->btree_inode->i_mapping);
017e5369 1186
3768f368
CM		 1187 memset(&root->root_key, 0, sizeof(root->root_key));
1188 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1189 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1190 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
3f157a2f 1191 root->defrag_trans_start = fs_info->generation;
58176a96 1192 init_completion(&root->kobj_unregister);
6702ed49 1193 root->defrag_running = 0;
4d775673 1194 root->root_key.objectid = objectid;
0ee5dc67 1195 root->anon_dev = 0;
8ea05e3a 1196
5f3ab90a 1197 spin_lock_init(&root->root_item_lock);
3768f368
CM		 1198}
1199
200a5c17
JM		 1200static int __must_check find_and_setup_root(struct btrfs_root *tree_root,
1201 struct btrfs_fs_info *fs_info,
1202 u64 objectid,
1203 struct btrfs_root *root)
3768f368
CM		 1204{
1205 int ret;
db94535d 1206 u32 blocksize;
84234f3a 1207 u64 generation;
3768f368 1208
db94535d 1209 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 1210 tree_root->sectorsize, tree_root->stripesize,
1211 root, fs_info, objectid);
3768f368
CM		 1212 ret = btrfs_find_last_root(tree_root, objectid,
1213 &root->root_item, &root->root_key);
4df27c4d
YZ		 1214 if (ret > 0)
1215 return -ENOENT;
200a5c17
JM		 1216 else if (ret < 0)
1217 return ret;
3768f368 1218
84234f3a 1219 generation = btrfs_root_generation(&root->root_item);
db94535d 1220 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
af31f5e5 1221 root->commit_root = NULL;
db94535d 1222 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1223 blocksize, generation);
b9fab919 1224 if (!root->node || !btrfs_buffer_uptodate(root->node, generation, 0)) {
68433b73 1225 free_extent_buffer(root->node);
af31f5e5 1226 root->node = NULL;
68433b73
CM		 1227 return -EIO;
1228 }
4df27c4d 1229 root->commit_root = btrfs_root_node(root);
d97e63b6
CM		 1230 return 0;
1231}
1232
f84a8bd6 1233static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
6f07e42e
AV		 1234{
1235 struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1236 if (root)
1237 root->fs_info = fs_info;
1238 return root;
1239}
1240
20897f5c
AJ		 1241struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1242 struct btrfs_fs_info *fs_info,
1243 u64 objectid)
1244{
1245 struct extent_buffer *leaf;
1246 struct btrfs_root *tree_root = fs_info->tree_root;
1247 struct btrfs_root *root;
1248 struct btrfs_key key;
1249 int ret = 0;
1250 u64 bytenr;
6463fe58 1251 uuid_le uuid;
20897f5c
AJ		 1252
1253 root = btrfs_alloc_root(fs_info);
1254 if (!root)
1255 return ERR_PTR(-ENOMEM);
1256
1257 __setup_root(tree_root->nodesize, tree_root->leafsize,
1258 tree_root->sectorsize, tree_root->stripesize,
1259 root, fs_info, objectid);
1260 root->root_key.objectid = objectid;
1261 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1262 root->root_key.offset = 0;
1263
1264 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1265 0, objectid, NULL, 0, 0, 0);
1266 if (IS_ERR(leaf)) {
1267 ret = PTR_ERR(leaf);
1dd05682 1268 leaf = NULL;
20897f5c
AJ		 1269 goto fail;
1270 }
1271
1272 bytenr = leaf->start;
1273 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1274 btrfs_set_header_bytenr(leaf, leaf->start);
1275 btrfs_set_header_generation(leaf, trans->transid);
1276 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1277 btrfs_set_header_owner(leaf, objectid);
1278 root->node = leaf;
1279
1280 write_extent_buffer(leaf, fs_info->fsid,
1281 (unsigned long)btrfs_header_fsid(leaf),
1282 BTRFS_FSID_SIZE);
1283 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
1284 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
1285 BTRFS_UUID_SIZE);
1286 btrfs_mark_buffer_dirty(leaf);
1287
1288 root->commit_root = btrfs_root_node(root);
1289 root->track_dirty = 1;
1290
1291
1292 root->root_item.flags = 0;
1293 root->root_item.byte_limit = 0;
1294 btrfs_set_root_bytenr(&root->root_item, leaf->start);
1295 btrfs_set_root_generation(&root->root_item, trans->transid);
1296 btrfs_set_root_level(&root->root_item, 0);
1297 btrfs_set_root_refs(&root->root_item, 1);
1298 btrfs_set_root_used(&root->root_item, leaf->len);
1299 btrfs_set_root_last_snapshot(&root->root_item, 0);
1300 btrfs_set_root_dirid(&root->root_item, 0);
6463fe58
SB		 1301 uuid_le_gen(&uuid);
1302 memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
20897f5c
AJ		 1303 root->root_item.drop_level = 0;
1304
1305 key.objectid = objectid;
1306 key.type = BTRFS_ROOT_ITEM_KEY;
1307 key.offset = 0;
1308 ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1309 if (ret)
1310 goto fail;
1311
1312 btrfs_tree_unlock(leaf);
1313
1dd05682
TI		 1314 return root;
1315
20897f5c 1316fail:
1dd05682
TI		 1317 if (leaf) {
1318 btrfs_tree_unlock(leaf);
1319 free_extent_buffer(leaf);
1320 }
1321 kfree(root);
20897f5c 1322
1dd05682 1323 return ERR_PTR(ret);
20897f5c
AJ		 1324}
1325
7237f183
YZ		 1326static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1327 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1328{
1329 struct btrfs_root *root;
1330 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1331 struct extent_buffer *leaf;
e02119d5 1332
6f07e42e 1333 root = btrfs_alloc_root(fs_info);
e02119d5 1334 if (!root)
7237f183 1335 return ERR_PTR(-ENOMEM);
e02119d5
CM		 1336
1337 __setup_root(tree_root->nodesize, tree_root->leafsize,
1338 tree_root->sectorsize, tree_root->stripesize,
1339 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1340
1341 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1342 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1343 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
7237f183
YZ		 1344 /*
1345 * log trees do not get reference counted because they go away
1346 * before a real commit is actually done. They do store pointers
1347 * to file data extents, and those reference counts still get
1348 * updated (along with back refs to the log tree).
1349 */
e02119d5
CM		 1350 root->ref_cows = 0;
1351
5d4f98a2 1352 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
66d7e7f0 1353 BTRFS_TREE_LOG_OBJECTID, NULL,
5581a51a 1354 0, 0, 0);
7237f183
YZ		 1355 if (IS_ERR(leaf)) {
1356 kfree(root);
1357 return ERR_CAST(leaf);
1358 }
e02119d5 1359
5d4f98a2
YZ		 1360 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1361 btrfs_set_header_bytenr(leaf, leaf->start);
1362 btrfs_set_header_generation(leaf, trans->transid);
1363 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1364 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1365 root->node = leaf;
e02119d5
CM		 1366
1367 write_extent_buffer(root->node, root->fs_info->fsid,
1368 (unsigned long)btrfs_header_fsid(root->node),
1369 BTRFS_FSID_SIZE);
1370 btrfs_mark_buffer_dirty(root->node);
1371 btrfs_tree_unlock(root->node);
7237f183
YZ		 1372 return root;
1373}
1374
1375int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1376 struct btrfs_fs_info *fs_info)
1377{
1378 struct btrfs_root *log_root;
1379
1380 log_root = alloc_log_tree(trans, fs_info);
1381 if (IS_ERR(log_root))
1382 return PTR_ERR(log_root);
1383 WARN_ON(fs_info->log_root_tree);
1384 fs_info->log_root_tree = log_root;
1385 return 0;
1386}
1387
1388int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1389 struct btrfs_root *root)
1390{
1391 struct btrfs_root *log_root;
1392 struct btrfs_inode_item *inode_item;
1393
1394 log_root = alloc_log_tree(trans, root->fs_info);
1395 if (IS_ERR(log_root))
1396 return PTR_ERR(log_root);
1397
1398 log_root->last_trans = trans->transid;
1399 log_root->root_key.offset = root->root_key.objectid;
1400
1401 inode_item = &log_root->root_item.inode;
1402 inode_item->generation = cpu_to_le64(1);
1403 inode_item->size = cpu_to_le64(3);
1404 inode_item->nlink = cpu_to_le32(1);
1405 inode_item->nbytes = cpu_to_le64(root->leafsize);
1406 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1407
5d4f98a2 1408 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1409
1410 WARN_ON(root->log_root);
1411 root->log_root = log_root;
1412 root->log_transid = 0;
257c62e1 1413 root->last_log_commit = 0;
e02119d5
CM		 1414 return 0;
1415}
1416
1417struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
1418 struct btrfs_key *location)
1419{
1420 struct btrfs_root *root;
1421 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1422 struct btrfs_path *path;
5f39d397 1423 struct extent_buffer *l;
84234f3a 1424 u64 generation;
db94535d 1425 u32 blocksize;
0f7d52f4 1426 int ret = 0;
8ea05e3a 1427 int slot;
0f7d52f4 1428
6f07e42e 1429 root = btrfs_alloc_root(fs_info);
0cf6c620 1430 if (!root)
0f7d52f4 1431 return ERR_PTR(-ENOMEM);
0f7d52f4 1432 if (location->offset == (u64)-1) {
db94535d 1433 ret = find_and_setup_root(tree_root, fs_info,
0f7d52f4
CM		 1434 location->objectid, root);
1435 if (ret) {
0f7d52f4
CM		 1436 kfree(root);
1437 return ERR_PTR(ret);
1438 }
13a8a7c8 1439 goto out;
0f7d52f4
CM		 1440 }
1441
db94535d 1442 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 1443 tree_root->sectorsize, tree_root->stripesize,
1444 root, fs_info, location->objectid);
0f7d52f4
CM		 1445
1446 path = btrfs_alloc_path();
db5b493a
TI		 1447 if (!path) {
1448 kfree(root);
1449 return ERR_PTR(-ENOMEM);
1450 }
0f7d52f4 1451 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
13a8a7c8
YZ		 1452 if (ret == 0) {
1453 l = path->nodes[0];
8ea05e3a 1454 slot = path->slots[0];
5fbf83c1 1455 btrfs_read_root_item(l, slot, &root->root_item);
13a8a7c8 1456 memcpy(&root->root_key, location, sizeof(*location));
0f7d52f4 1457 }
0f7d52f4
CM		 1458 btrfs_free_path(path);
1459 if (ret) {
5e540f77 1460 kfree(root);
13a8a7c8
YZ		 1461 if (ret > 0)
1462 ret = -ENOENT;
0f7d52f4
CM		 1463 return ERR_PTR(ret);
1464 }
13a8a7c8 1465
84234f3a 1466 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1467 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1468 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1469 blocksize, generation);
416bc658
JB		 1470 if (!root->node || !extent_buffer_uptodate(root->node)) {
1471 ret = (!root->node) ? -ENOMEM : -EIO;
1472
1473 free_extent_buffer(root->node);
1474 kfree(root);
1475 return ERR_PTR(ret);
1476 }
1477
5d4f98a2 1478 root->commit_root = btrfs_root_node(root);
79787eaa 1479 BUG_ON(!root->node); /* -ENOMEM */
13a8a7c8 1480out:
08fe4db1 1481 if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
e02119d5 1482 root->ref_cows = 1;
08fe4db1
LZ		 1483 btrfs_check_and_init_root_item(&root->root_item);
1484 }
13a8a7c8 1485
5eda7b5e
CM		 1486 return root;
1487}
1488
edbd8d4e
CM		 1489struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
1490 struct btrfs_key *location)
5eda7b5e
CM		 1491{
1492 struct btrfs_root *root;
1493 int ret;
1494
edbd8d4e
CM		 1495 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1496 return fs_info->tree_root;
1497 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1498 return fs_info->extent_root;
8f18cf13
CM		 1499 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1500 return fs_info->chunk_root;
1501 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1502 return fs_info->dev_root;
0403e47e
YZ		 1503 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1504 return fs_info->csum_root;
bcef60f2
AJ		 1505 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1506 return fs_info->quota_root ? fs_info->quota_root :
1507 ERR_PTR(-ENOENT);
4df27c4d
YZ		 1508again:
1509 spin_lock(&fs_info->fs_roots_radix_lock);
5eda7b5e
CM		 1510 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1511 (unsigned long)location->objectid);
4df27c4d 1512 spin_unlock(&fs_info->fs_roots_radix_lock);
5eda7b5e
CM		 1513 if (root)
1514 return root;
1515
e02119d5 1516 root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
5eda7b5e
CM		 1517 if (IS_ERR(root))
1518 return root;
3394e160 1519
581bb050 1520 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
581bb050
LZ		 1521 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1522 GFP_NOFS);
35a30d7c
DS		 1523 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1524 ret = -ENOMEM;
581bb050 1525 goto fail;
35a30d7c 1526 }
581bb050
LZ		 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
0ee5dc67 1533 ret = get_anon_bdev(&root->anon_dev);
ac08aedf
CM		 1534 if (ret)
1535 goto fail;
3394e160 1536
d68fc57b
YZ		 1537 if (btrfs_root_refs(&root->root_item) == 0) {
1538 ret = -ENOENT;
1539 goto fail;
1540 }
1541
1542 ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
1543 if (ret < 0)
1544 goto fail;
1545 if (ret == 0)
1546 root->orphan_item_inserted = 1;
1547
4df27c4d
YZ		 1548 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1549 if (ret)
1550 goto fail;
1551
1552 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 1553 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1554 (unsigned long)root->root_key.objectid,
0f7d52f4 1555 root);
d68fc57b 1556 if (ret == 0)
4df27c4d 1557 root->in_radix = 1;
d68fc57b 1558
4df27c4d
YZ		 1559 spin_unlock(&fs_info->fs_roots_radix_lock);
1560 radix_tree_preload_end();
0f7d52f4 1561 if (ret) {
4df27c4d
YZ		 1562 if (ret == -EEXIST) {
1563 free_fs_root(root);
1564 goto again;
1565 }
1566 goto fail;
0f7d52f4 1567 }
4df27c4d
YZ		 1568
1569 ret = btrfs_find_dead_roots(fs_info->tree_root,
1570 root->root_key.objectid);
1571 WARN_ON(ret);
edbd8d4e 1572 return root;
4df27c4d
YZ		 1573fail:
1574 free_fs_root(root);
1575 return ERR_PTR(ret);
edbd8d4e
CM		 1576}
1577
04160088
CM		 1578static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1579{
1580 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1581 int ret = 0;
04160088
CM		 1582 struct btrfs_device *device;
1583 struct backing_dev_info *bdi;
b7967db7 1584
1f78160c
XG		 1585 rcu_read_lock();
1586 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1587 if (!device->bdev)
1588 continue;
04160088
CM		 1589 bdi = blk_get_backing_dev_info(device->bdev);
1590 if (bdi && bdi_congested(bdi, bdi_bits)) {
1591 ret = 1;
1592 break;
1593 }
1594 }
1f78160c 1595 rcu_read_unlock();
04160088
CM		 1596 return ret;
1597}
1598
ad081f14
JA		 1599/*
1600 * If this fails, caller must call bdi_destroy() to get rid of the
1601 * bdi again.
1602 */
04160088
CM		 1603static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1604{
ad081f14
JA		 1605 int err;
1606
1607 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1608 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1609 if (err)
1610 return err;
1611
4575c9cc 1612 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1613 bdi->congested_fn = btrfs_congested_fn;
1614 bdi->congested_data = info;
1615 return 0;
1616}
1617
8b712842
CM		 1618/*
1619 * called by the kthread helper functions to finally call the bio end_io
1620 * functions. This is where read checksum verification actually happens
1621 */
1622static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1623{
ce9adaa5 1624 struct bio *bio;
8b712842
CM		 1625 struct end_io_wq *end_io_wq;
1626 struct btrfs_fs_info *fs_info;
ce9adaa5 1627 int error;
ce9adaa5 1628
8b712842
CM		 1629 end_io_wq = container_of(work, struct end_io_wq, work);
1630 bio = end_io_wq->bio;
1631 fs_info = end_io_wq->info;
ce9adaa5 1632
8b712842
CM		 1633 error = end_io_wq->error;
1634 bio->bi_private = end_io_wq->private;
1635 bio->bi_end_io = end_io_wq->end_io;
1636 kfree(end_io_wq);
8b712842 1637 bio_endio(bio, error);
44b8bd7e
CM		 1638}
1639
a74a4b97
CM		 1640static int cleaner_kthread(void *arg)
1641{
1642 struct btrfs_root *root = arg;
1643
1644 do {
9d1a2a3a
DS		 1645 int again = 0;
1646
76dda93c 1647 if (!(root->fs_info->sb->s_flags & MS_RDONLY) &&
9d1a2a3a
DS		 1648 down_read_trylock(&root->fs_info->sb->s_umount)) {
1649 if (mutex_trylock(&root->fs_info->cleaner_mutex)) {
1650 btrfs_run_delayed_iputs(root);
1651 again = btrfs_clean_one_deleted_snapshot(root);
1652 mutex_unlock(&root->fs_info->cleaner_mutex);
1653 }
4cb5300b 1654 btrfs_run_defrag_inodes(root->fs_info);
9d1a2a3a 1655 up_read(&root->fs_info->sb->s_umount);
76dda93c 1656 }
a74a4b97 1657
9d1a2a3a 1658 if (!try_to_freeze() && !again) {
a74a4b97 1659 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1660 if (!kthread_should_stop())
1661 schedule();
a74a4b97
CM		 1662 __set_current_state(TASK_RUNNING);
1663 }
1664 } while (!kthread_should_stop());
1665 return 0;
1666}
1667
1668static int transaction_kthread(void *arg)
1669{
1670 struct btrfs_root *root = arg;
1671 struct btrfs_trans_handle *trans;
1672 struct btrfs_transaction *cur;
8929ecfa 1673 u64 transid;
a74a4b97
CM		 1674 unsigned long now;
1675 unsigned long delay;
914b2007 1676 bool cannot_commit;
a74a4b97
CM		 1677
1678 do {
914b2007 1679 cannot_commit = false;
a74a4b97 1680 delay = HZ * 30;
a74a4b97
CM		 1681 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1682
a4abeea4 1683 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1684 cur = root->fs_info->running_transaction;
1685 if (!cur) {
a4abeea4 1686 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1687 goto sleep;
1688 }
31153d81 1689
a74a4b97 1690 now = get_seconds();
8929ecfa
YZ		 1691 if (!cur->blocked &&
1692 (now < cur->start_time || now - cur->start_time < 30)) {
a4abeea4 1693 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1694 delay = HZ * 5;
1695 goto sleep;
1696 }
8929ecfa 1697 transid = cur->transid;
a4abeea4 1698 spin_unlock(&root->fs_info->trans_lock);
56bec294 1699
79787eaa 1700 /* If the file system is aborted, this will always fail. */
354aa0fb 1701 trans = btrfs_attach_transaction(root);
914b2007 1702 if (IS_ERR(trans)) {
354aa0fb
MX		 1703 if (PTR_ERR(trans) != -ENOENT)
1704 cannot_commit = true;
79787eaa 1705 goto sleep;
914b2007 1706 }
8929ecfa 1707 if (transid == trans->transid) {
79787eaa 1708 btrfs_commit_transaction(trans, root);
8929ecfa
YZ		 1709 } else {
1710 btrfs_end_transaction(trans, root);
1711 }
a74a4b97
CM		 1712sleep:
1713 wake_up_process(root->fs_info->cleaner_kthread);
1714 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1715
a0acae0e 1716 if (!try_to_freeze()) {
a74a4b97 1717 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa 1718 if (!kthread_should_stop() &&
914b2007
JK		 1719 (!btrfs_transaction_blocked(root->fs_info) ||
1720 cannot_commit))
8929ecfa 1721 schedule_timeout(delay);
a74a4b97
CM		 1722 __set_current_state(TASK_RUNNING);
1723 }
1724 } while (!kthread_should_stop());
1725 return 0;
1726}
1727
af31f5e5
CM		 1728/*
1729 * this will find the highest generation in the array of
1730 * root backups. The index of the highest array is returned,
1731 * or -1 if we can't find anything.
1732 *
1733 * We check to make sure the array is valid by comparing the
1734 * generation of the latest root in the array with the generation
1735 * in the super block. If they don't match we pitch it.
1736 */
1737static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1738{
1739 u64 cur;
1740 int newest_index = -1;
1741 struct btrfs_root_backup *root_backup;
1742 int i;
1743
1744 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1745 root_backup = info->super_copy->super_roots + i;
1746 cur = btrfs_backup_tree_root_gen(root_backup);
1747 if (cur == newest_gen)
1748 newest_index = i;
1749 }
1750
1751 /* check to see if we actually wrapped around */
1752 if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1753 root_backup = info->super_copy->super_roots;
1754 cur = btrfs_backup_tree_root_gen(root_backup);
1755 if (cur == newest_gen)
1756 newest_index = 0;
1757 }
1758 return newest_index;
1759}
1760
1761
1762/*
1763 * find the oldest backup so we know where to store new entries
1764 * in the backup array. This will set the backup_root_index
1765 * field in the fs_info struct
1766 */
1767static void find_oldest_super_backup(struct btrfs_fs_info *info,
1768 u64 newest_gen)
1769{
1770 int newest_index = -1;
1771
1772 newest_index = find_newest_super_backup(info, newest_gen);
1773 /* if there was garbage in there, just move along */
1774 if (newest_index == -1) {
1775 info->backup_root_index = 0;
1776 } else {
1777 info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1778 }
1779}
1780
1781/*
1782 * copy all the root pointers into the super backup array.
1783 * this will bump the backup pointer by one when it is
1784 * done
1785 */
1786static void backup_super_roots(struct btrfs_fs_info *info)
1787{
1788 int next_backup;
1789 struct btrfs_root_backup *root_backup;
1790 int last_backup;
1791
1792 next_backup = info->backup_root_index;
1793 last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1794 BTRFS_NUM_BACKUP_ROOTS;
1795
1796 /*
1797 * just overwrite the last backup if we're at the same generation
1798 * this happens only at umount
1799 */
1800 root_backup = info->super_for_commit->super_roots + last_backup;
1801 if (btrfs_backup_tree_root_gen(root_backup) ==
1802 btrfs_header_generation(info->tree_root->node))
1803 next_backup = last_backup;
1804
1805 root_backup = info->super_for_commit->super_roots + next_backup;
1806
1807 /*
1808 * make sure all of our padding and empty slots get zero filled
1809 * regardless of which ones we use today
1810 */
1811 memset(root_backup, 0, sizeof(*root_backup));
1812
1813 info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1814
1815 btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1816 btrfs_set_backup_tree_root_gen(root_backup,
1817 btrfs_header_generation(info->tree_root->node));
1818
1819 btrfs_set_backup_tree_root_level(root_backup,
1820 btrfs_header_level(info->tree_root->node));
1821
1822 btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1823 btrfs_set_backup_chunk_root_gen(root_backup,
1824 btrfs_header_generation(info->chunk_root->node));
1825 btrfs_set_backup_chunk_root_level(root_backup,
1826 btrfs_header_level(info->chunk_root->node));
1827
1828 btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1829 btrfs_set_backup_extent_root_gen(root_backup,
1830 btrfs_header_generation(info->extent_root->node));
1831 btrfs_set_backup_extent_root_level(root_backup,
1832 btrfs_header_level(info->extent_root->node));
1833
7c7e82a7
CM		 1834 /*
1835 * we might commit during log recovery, which happens before we set
1836 * the fs_root. Make sure it is valid before we fill it in.
1837 */
1838 if (info->fs_root && info->fs_root->node) {
1839 btrfs_set_backup_fs_root(root_backup,
1840 info->fs_root->node->start);
1841 btrfs_set_backup_fs_root_gen(root_backup,
af31f5e5 1842 btrfs_header_generation(info->fs_root->node));
7c7e82a7 1843 btrfs_set_backup_fs_root_level(root_backup,
af31f5e5 1844 btrfs_header_level(info->fs_root->node));
7c7e82a7 1845 }
af31f5e5
CM		 1846
1847 btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1848 btrfs_set_backup_dev_root_gen(root_backup,
1849 btrfs_header_generation(info->dev_root->node));
1850 btrfs_set_backup_dev_root_level(root_backup,
1851 btrfs_header_level(info->dev_root->node));
1852
1853 btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1854 btrfs_set_backup_csum_root_gen(root_backup,
1855 btrfs_header_generation(info->csum_root->node));
1856 btrfs_set_backup_csum_root_level(root_backup,
1857 btrfs_header_level(info->csum_root->node));
1858
1859 btrfs_set_backup_total_bytes(root_backup,
1860 btrfs_super_total_bytes(info->super_copy));
1861 btrfs_set_backup_bytes_used(root_backup,
1862 btrfs_super_bytes_used(info->super_copy));
1863 btrfs_set_backup_num_devices(root_backup,
1864 btrfs_super_num_devices(info->super_copy));
1865
1866 /*
1867 * if we don't copy this out to the super_copy, it won't get remembered
1868 * for the next commit
1869 */
1870 memcpy(&info->super_copy->super_roots,
1871 &info->super_for_commit->super_roots,
1872 sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
1873}
1874
1875/*
1876 * this copies info out of the root backup array and back into
1877 * the in-memory super block. It is meant to help iterate through
1878 * the array, so you send it the number of backups you've already
1879 * tried and the last backup index you used.
1880 *
1881 * this returns -1 when it has tried all the backups
1882 */
1883static noinline int next_root_backup(struct btrfs_fs_info *info,
1884 struct btrfs_super_block *super,
1885 int *num_backups_tried, int *backup_index)
1886{
1887 struct btrfs_root_backup *root_backup;
1888 int newest = *backup_index;
1889
1890 if (*num_backups_tried == 0) {
1891 u64 gen = btrfs_super_generation(super);
1892
1893 newest = find_newest_super_backup(info, gen);
1894 if (newest == -1)
1895 return -1;
1896
1897 *backup_index = newest;
1898 *num_backups_tried = 1;
1899 } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
1900 /* we've tried all the backups, all done */
1901 return -1;
1902 } else {
1903 /* jump to the next oldest backup */
1904 newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
1905 BTRFS_NUM_BACKUP_ROOTS;
1906 *backup_index = newest;
1907 *num_backups_tried += 1;
1908 }
1909 root_backup = super->super_roots + newest;
1910
1911 btrfs_set_super_generation(super,
1912 btrfs_backup_tree_root_gen(root_backup));
1913 btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
1914 btrfs_set_super_root_level(super,
1915 btrfs_backup_tree_root_level(root_backup));
1916 btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
1917
1918 /*
1919 * fixme: the total bytes and num_devices need to match or we should
1920 * need a fsck
1921 */
1922 btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
1923 btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
1924 return 0;
1925}
1926
7abadb64
LB		 1927/* helper to cleanup workers */
1928static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
1929{
1930 btrfs_stop_workers(&fs_info->generic_worker);
1931 btrfs_stop_workers(&fs_info->fixup_workers);
1932 btrfs_stop_workers(&fs_info->delalloc_workers);
1933 btrfs_stop_workers(&fs_info->workers);
1934 btrfs_stop_workers(&fs_info->endio_workers);
1935 btrfs_stop_workers(&fs_info->endio_meta_workers);
1936 btrfs_stop_workers(&fs_info->endio_raid56_workers);
1937 btrfs_stop_workers(&fs_info->rmw_workers);
1938 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
1939 btrfs_stop_workers(&fs_info->endio_write_workers);
1940 btrfs_stop_workers(&fs_info->endio_freespace_worker);
1941 btrfs_stop_workers(&fs_info->submit_workers);
1942 btrfs_stop_workers(&fs_info->delayed_workers);
1943 btrfs_stop_workers(&fs_info->caching_workers);
1944 btrfs_stop_workers(&fs_info->readahead_workers);
1945 btrfs_stop_workers(&fs_info->flush_workers);
2f232036 1946 btrfs_stop_workers(&fs_info->qgroup_rescan_workers);
7abadb64
LB		 1947}
1948
af31f5e5
CM		 1949/* helper to cleanup tree roots */
1950static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
1951{
1952 free_extent_buffer(info->tree_root->node);
1953 free_extent_buffer(info->tree_root->commit_root);
1954 free_extent_buffer(info->dev_root->node);
1955 free_extent_buffer(info->dev_root->commit_root);
1956 free_extent_buffer(info->extent_root->node);
1957 free_extent_buffer(info->extent_root->commit_root);
1958 free_extent_buffer(info->csum_root->node);
1959 free_extent_buffer(info->csum_root->commit_root);
bcef60f2
AJ		 1960 if (info->quota_root) {
1961 free_extent_buffer(info->quota_root->node);
1962 free_extent_buffer(info->quota_root->commit_root);
1963 }
af31f5e5
CM		 1964
1965 info->tree_root->node = NULL;
1966 info->tree_root->commit_root = NULL;
1967 info->dev_root->node = NULL;
1968 info->dev_root->commit_root = NULL;
1969 info->extent_root->node = NULL;
1970 info->extent_root->commit_root = NULL;
1971 info->csum_root->node = NULL;
1972 info->csum_root->commit_root = NULL;
bcef60f2
AJ		 1973 if (info->quota_root) {
1974 info->quota_root->node = NULL;
1975 info->quota_root->commit_root = NULL;
1976 }
af31f5e5
CM		 1977
1978 if (chunk_root) {
1979 free_extent_buffer(info->chunk_root->node);
1980 free_extent_buffer(info->chunk_root->commit_root);
1981 info->chunk_root->node = NULL;
1982 info->chunk_root->commit_root = NULL;
1983 }
1984}
1985
171f6537
JB		 1986static void del_fs_roots(struct btrfs_fs_info *fs_info)
1987{
1988 int ret;
1989 struct btrfs_root *gang[8];
1990 int i;
1991
1992 while (!list_empty(&fs_info->dead_roots)) {
1993 gang[0] = list_entry(fs_info->dead_roots.next,
1994 struct btrfs_root, root_list);
1995 list_del(&gang[0]->root_list);
1996
1997 if (gang[0]->in_radix) {
1998 btrfs_free_fs_root(fs_info, gang[0]);
1999 } else {
2000 free_extent_buffer(gang[0]->node);
2001 free_extent_buffer(gang[0]->commit_root);
2002 kfree(gang[0]);
2003 }
2004 }
2005
2006 while (1) {
2007 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2008 (void **)gang, 0,
2009 ARRAY_SIZE(gang));
2010 if (!ret)
2011 break;
2012 for (i = 0; i < ret; i++)
2013 btrfs_free_fs_root(fs_info, gang[i]);
2014 }
2015}
af31f5e5 2016
ad2b2c80
AV		 2017int open_ctree(struct super_block *sb,
2018 struct btrfs_fs_devices *fs_devices,
2019 char *options)
2e635a27 2020{
db94535d
CM		 2021 u32 sectorsize;
2022 u32 nodesize;
2023 u32 leafsize;
2024 u32 blocksize;
87ee04eb 2025 u32 stripesize;
84234f3a 2026 u64 generation;
f2b636e8 2027 u64 features;
3de4586c 2028 struct btrfs_key location;
a061fc8d 2029 struct buffer_head *bh;
4d34b278 2030 struct btrfs_super_block *disk_super;
815745cf 2031 struct btrfs_fs_info *fs_info = btrfs_sb(sb);
f84a8bd6 2032 struct btrfs_root *tree_root;
4d34b278
ID		 2033 struct btrfs_root *extent_root;
2034 struct btrfs_root *csum_root;
2035 struct btrfs_root *chunk_root;
2036 struct btrfs_root *dev_root;
bcef60f2 2037 struct btrfs_root *quota_root;
e02119d5 2038 struct btrfs_root *log_tree_root;
eb60ceac 2039 int ret;
e58ca020 2040 int err = -EINVAL;
af31f5e5
CM		 2041 int num_backups_tried = 0;
2042 int backup_index = 0;
4543df7e 2043
f84a8bd6 2044 tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
6f07e42e
AV		 2045 extent_root = fs_info->extent_root = btrfs_alloc_root(fs_info);
2046 csum_root = fs_info->csum_root = btrfs_alloc_root(fs_info);
2047 chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
2048 dev_root = fs_info->dev_root = btrfs_alloc_root(fs_info);
bcef60f2 2049 quota_root = fs_info->quota_root = btrfs_alloc_root(fs_info);
8790d502 2050
f84a8bd6 2051 if (!tree_root || !extent_root || !csum_root ||
bcef60f2 2052 !chunk_root || !dev_root || !quota_root) {
39279cc3
CM		 2053 err = -ENOMEM;
2054 goto fail;
2055 }
76dda93c
YZ		 2056
2057 ret = init_srcu_struct(&fs_info->subvol_srcu);
2058 if (ret) {
2059 err = ret;
2060 goto fail;
2061 }
2062
2063 ret = setup_bdi(fs_info, &fs_info->bdi);
2064 if (ret) {
2065 err = ret;
2066 goto fail_srcu;
2067 }
2068
e2d84521
MX		 2069 ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2070 if (ret) {
2071 err = ret;
2072 goto fail_bdi;
2073 }
2074 fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2075 (1 + ilog2(nr_cpu_ids));
2076
963d678b
MX		 2077 ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2078 if (ret) {
2079 err = ret;
2080 goto fail_dirty_metadata_bytes;
2081 }
2082
76dda93c
YZ		 2083 fs_info->btree_inode = new_inode(sb);
2084 if (!fs_info->btree_inode) {
2085 err = -ENOMEM;
963d678b 2086 goto fail_delalloc_bytes;
76dda93c
YZ		 2087 }
2088
a6591715 2089 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 2090
76dda93c 2091 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
8fd17795 2092 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 2093 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 2094 INIT_LIST_HEAD(&fs_info->delayed_iputs);
ea8c2819 2095 INIT_LIST_HEAD(&fs_info->delalloc_inodes);
11833d66 2096 INIT_LIST_HEAD(&fs_info->caching_block_groups);
1832a6d5 2097 spin_lock_init(&fs_info->delalloc_lock);
a4abeea4 2098 spin_lock_init(&fs_info->trans_lock);
76dda93c 2099 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 2100 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 2101 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 2102 spin_lock_init(&fs_info->free_chunk_lock);
f29021b2 2103 spin_lock_init(&fs_info->tree_mod_seq_lock);
ceda0864 2104 spin_lock_init(&fs_info->super_lock);
f29021b2 2105 rwlock_init(&fs_info->tree_mod_log_lock);
7585717f 2106 mutex_init(&fs_info->reloc_mutex);
de98ced9 2107 seqlock_init(&fs_info->profiles_lock);
19c00ddc 2108
58176a96 2109 init_completion(&fs_info->kobj_unregister);
0b86a832 2110 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 2111 INIT_LIST_HEAD(&fs_info->space_info);
f29021b2 2112 INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
0b86a832 2113 btrfs_mapping_init(&fs_info->mapping_tree);
66d8f3dd
MX		 2114 btrfs_init_block_rsv(&fs_info->global_block_rsv,
2115 BTRFS_BLOCK_RSV_GLOBAL);
2116 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2117 BTRFS_BLOCK_RSV_DELALLOC);
2118 btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2119 btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2120 btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2121 btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2122 BTRFS_BLOCK_RSV_DELOPS);
cb03c743 2123 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 2124 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 2125 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 2126 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 2127 atomic_set(&fs_info->defrag_running, 0);
fc36ed7e 2128 atomic64_set(&fs_info->tree_mod_seq, 0);
e20d96d6 2129 fs_info->sb = sb;
6f568d35 2130 fs_info->max_inline = 8192 * 1024;
9ed74f2d 2131 fs_info->metadata_ratio = 0;
4cb5300b 2132 fs_info->defrag_inodes = RB_ROOT;
a4abeea4 2133 fs_info->trans_no_join = 0;
2bf64758 2134 fs_info->free_chunk_space = 0;
f29021b2 2135 fs_info->tree_mod_log = RB_ROOT;
c8b97818 2136
90519d66
AJ		 2137 /* readahead state */
2138 INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2139 spin_lock_init(&fs_info->reada_lock);
c8b97818 2140
b34b086c
CM		 2141 fs_info->thread_pool_size = min_t(unsigned long,
2142 num_online_cpus() + 2, 8);
0afbaf8c 2143
3eaa2885
CM		 2144 INIT_LIST_HEAD(&fs_info->ordered_extents);
2145 spin_lock_init(&fs_info->ordered_extent_lock);
16cdcec7
MX		 2146 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2147 GFP_NOFS);
2148 if (!fs_info->delayed_root) {
2149 err = -ENOMEM;
2150 goto fail_iput;
2151 }
2152 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 2153
a2de733c
AJ		 2154 mutex_init(&fs_info->scrub_lock);
2155 atomic_set(&fs_info->scrubs_running, 0);
2156 atomic_set(&fs_info->scrub_pause_req, 0);
2157 atomic_set(&fs_info->scrubs_paused, 0);
2158 atomic_set(&fs_info->scrub_cancel_req, 0);
2159 init_waitqueue_head(&fs_info->scrub_pause_wait);
2160 init_rwsem(&fs_info->scrub_super_lock);
2161 fs_info->scrub_workers_refcnt = 0;
21adbd5c
SB		 2162#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2163 fs_info->check_integrity_print_mask = 0;
2164#endif
a2de733c 2165
c9e9f97b
ID		 2166 spin_lock_init(&fs_info->balance_lock);
2167 mutex_init(&fs_info->balance_mutex);
837d5b6e
ID		 2168 atomic_set(&fs_info->balance_running, 0);
2169 atomic_set(&fs_info->balance_pause_req, 0);
a7e99c69 2170 atomic_set(&fs_info->balance_cancel_req, 0);
c9e9f97b 2171 fs_info->balance_ctl = NULL;
837d5b6e 2172 init_waitqueue_head(&fs_info->balance_wait_q);
a2de733c 2173
a061fc8d
CM		 2174 sb->s_blocksize = 4096;
2175 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 2176 sb->s_bdi = &fs_info->bdi;
a061fc8d 2177
76dda93c 2178 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
bfe86848 2179 set_nlink(fs_info->btree_inode, 1);
0afbaf8c
CM		 2180 /*
2181 * we set the i_size on the btree inode to the max possible int.
2182 * the real end of the address space is determined by all of
2183 * the devices in the system
2184 */
2185 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 2186 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 2187 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2188
5d4f98a2 2189 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 2190 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 2191 fs_info->btree_inode->i_mapping);
0b32f4bb 2192 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
a8067e02 2193 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM		 2194
2195 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 2196
76dda93c
YZ		 2197 BTRFS_I(fs_info->btree_inode)->root = tree_root;
2198 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2199 sizeof(struct btrfs_key));
72ac3c0d
JB		 2200 set_bit(BTRFS_INODE_DUMMY,
2201 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
c65ddb52 2202 insert_inode_hash(fs_info->btree_inode);
76dda93c 2203
0f9dd46c 2204 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 2205 fs_info->block_group_cache_tree = RB_ROOT;
a1897fdd 2206 fs_info->first_logical_byte = (u64)-1;
0f9dd46c 2207
11833d66 2208 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 2209 fs_info->btree_inode->i_mapping);
11833d66 2210 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 2211 fs_info->btree_inode->i_mapping);
11833d66 2212 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 2213 fs_info->do_barriers = 1;
e18e4809 2214
39279cc3 2215
5a3f23d5 2216 mutex_init(&fs_info->ordered_operations_mutex);
e02119d5 2217 mutex_init(&fs_info->tree_log_mutex);
925baedd 2218 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 2219 mutex_init(&fs_info->transaction_kthread_mutex);
2220 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 2221 mutex_init(&fs_info->volume_mutex);
276e680d 2222 init_rwsem(&fs_info->extent_commit_sem);
c71bf099 2223 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 2224 init_rwsem(&fs_info->subvol_sem);
e922e087
SB		 2225 fs_info->dev_replace.lock_owner = 0;
2226 atomic_set(&fs_info->dev_replace.nesting_level, 0);
2227 mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2228 mutex_init(&fs_info->dev_replace.lock_management_lock);
2229 mutex_init(&fs_info->dev_replace.lock);
fa9c0d79 2230
416ac51d 2231 spin_lock_init(&fs_info->qgroup_lock);
f2f6ed3d 2232 mutex_init(&fs_info->qgroup_ioctl_lock);
416ac51d
AJ		 2233 fs_info->qgroup_tree = RB_ROOT;
2234 INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2235 fs_info->qgroup_seq = 1;
2236 fs_info->quota_enabled = 0;
2237 fs_info->pending_quota_state = 0;
2f232036 2238 mutex_init(&fs_info->qgroup_rescan_lock);
416ac51d 2239
fa9c0d79
CM		 2240 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2241 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2242
e6dcd2dc 2243 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 2244 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 2245 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 2246 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 2247
53b381b3
DW		 2248 ret = btrfs_alloc_stripe_hash_table(fs_info);
2249 if (ret) {
83c8266a 2250 err = ret;
53b381b3
DW		 2251 goto fail_alloc;
2252 }
2253
0b86a832 2254 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 2255 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 2256
3c4bb26b 2257 invalidate_bdev(fs_devices->latest_bdev);
a512bbf8 2258 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY		 2259 if (!bh) {
2260 err = -EINVAL;
16cdcec7 2261 goto fail_alloc;
20b45077 2262 }
39279cc3 2263
6c41761f
DS		 2264 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2265 memcpy(fs_info->super_for_commit, fs_info->super_copy,
2266 sizeof(*fs_info->super_for_commit));
a061fc8d 2267 brelse(bh);
5f39d397 2268
6c41761f 2269 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 2270
6c41761f 2271 disk_super = fs_info->super_copy;
0f7d52f4 2272 if (!btrfs_super_root(disk_super))
16cdcec7 2273 goto fail_alloc;
0f7d52f4 2274
acce952b 2275 /* check FS state, whether FS is broken. */
87533c47
MX		 2276 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2277 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
acce952b 2278
fcd1f065
DS		 2279 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2280 if (ret) {
2281 printk(KERN_ERR "btrfs: superblock contains fatal errors\n");
2282 err = ret;
2283 goto fail_alloc;
2284 }
acce952b 2285
af31f5e5
CM		 2286 /*
2287 * run through our array of backup supers and setup
2288 * our ring pointer to the oldest one
2289 */
2290 generation = btrfs_super_generation(disk_super);
2291 find_oldest_super_backup(fs_info, generation);
2292
75e7cb7f
LB		 2293 /*
2294 * In the long term, we'll store the compression type in the super
2295 * block, and it'll be used for per file compression control.
2296 */
2297 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2298
2b82032c
YZ		 2299 ret = btrfs_parse_options(tree_root, options);
2300 if (ret) {
2301 err = ret;
16cdcec7 2302 goto fail_alloc;
2b82032c 2303 }
dfe25020 2304
f2b636e8
JB		 2305 features = btrfs_super_incompat_flags(disk_super) &
2306 ~BTRFS_FEATURE_INCOMPAT_SUPP;
2307 if (features) {
2308 printk(KERN_ERR "BTRFS: couldn't mount because of "
2309 "unsupported optional features (%Lx).\n",
21380931 2310 (unsigned long long)features);
f2b636e8 2311 err = -EINVAL;
16cdcec7 2312 goto fail_alloc;
f2b636e8
JB		 2313 }
2314
727011e0
CM		 2315 if (btrfs_super_leafsize(disk_super) !=
2316 btrfs_super_nodesize(disk_super)) {
2317 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2318 "blocksizes don't match. node %d leaf %d\n",
2319 btrfs_super_nodesize(disk_super),
2320 btrfs_super_leafsize(disk_super));
2321 err = -EINVAL;
2322 goto fail_alloc;
2323 }
2324 if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2325 printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2326 "blocksize (%d) was too large\n",
2327 btrfs_super_leafsize(disk_super));
2328 err = -EINVAL;
2329 goto fail_alloc;
2330 }
2331
5d4f98a2 2332 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae 2333 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
69e380d1 2334 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
a6fa6fae 2335 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
727011e0 2336
3173a18f
JB		 2337 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
2338 printk(KERN_ERR "btrfs: has skinny extents\n");
2339
727011e0
CM		 2340 /*
2341 * flag our filesystem as having big metadata blocks if
2342 * they are bigger than the page size
2343 */
2344 if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2345 if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
2346 printk(KERN_INFO "btrfs flagging fs with big metadata feature\n");
2347 features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2348 }
2349
bc3f116f
CM		 2350 nodesize = btrfs_super_nodesize(disk_super);
2351 leafsize = btrfs_super_leafsize(disk_super);
2352 sectorsize = btrfs_super_sectorsize(disk_super);
2353 stripesize = btrfs_super_stripesize(disk_super);
e2d84521 2354 fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids));
963d678b 2355 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
bc3f116f
CM		 2356
2357 /*
2358 * mixed block groups end up with duplicate but slightly offset
2359 * extent buffers for the same range. It leads to corruptions
2360 */
2361 if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2362 (sectorsize != leafsize)) {
2363 printk(KERN_WARNING "btrfs: unequal leaf/node/sector sizes "
2364 "are not allowed for mixed block groups on %s\n",
2365 sb->s_id);
2366 goto fail_alloc;
2367 }
2368
ceda0864
MX		 2369 /*
2370 * Needn't use the lock because there is no other task which will
2371 * update the flag.
2372 */
a6fa6fae 2373 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 2374
f2b636e8
JB		 2375 features = btrfs_super_compat_ro_flags(disk_super) &
2376 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2377 if (!(sb->s_flags & MS_RDONLY) && features) {
2378 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2379 "unsupported option features (%Lx).\n",
21380931 2380 (unsigned long long)features);
f2b636e8 2381 err = -EINVAL;
16cdcec7 2382 goto fail_alloc;
f2b636e8 2383 }
61d92c32
CM		 2384
2385 btrfs_init_workers(&fs_info->generic_worker,
2386 "genwork", 1, NULL);
2387
5443be45 2388 btrfs_init_workers(&fs_info->workers, "worker",
61d92c32
CM		 2389 fs_info->thread_pool_size,
2390 &fs_info->generic_worker);
c8b97818 2391
771ed689 2392 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
61d92c32
CM		 2393 fs_info->thread_pool_size,
2394 &fs_info->generic_worker);
771ed689 2395
8ccf6f19
MX		 2396 btrfs_init_workers(&fs_info->flush_workers, "flush_delalloc",
2397 fs_info->thread_pool_size,
2398 &fs_info->generic_worker);
2399
5443be45 2400 btrfs_init_workers(&fs_info->submit_workers, "submit",
b720d209 2401 min_t(u64, fs_devices->num_devices,
61d92c32
CM		 2402 fs_info->thread_pool_size),
2403 &fs_info->generic_worker);
61b49440 2404
bab39bf9
JB		 2405 btrfs_init_workers(&fs_info->caching_workers, "cache",
2406 2, &fs_info->generic_worker);
2407
61b49440
CM		 2408 /* a higher idle thresh on the submit workers makes it much more
2409 * likely that bios will be send down in a sane order to the
2410 * devices
2411 */
2412 fs_info->submit_workers.idle_thresh = 64;
53863232 2413
771ed689 2414 fs_info->workers.idle_thresh = 16;
4a69a410 2415 fs_info->workers.ordered = 1;
61b49440 2416
771ed689
CM		 2417 fs_info->delalloc_workers.idle_thresh = 2;
2418 fs_info->delalloc_workers.ordered = 1;
2419
61d92c32
CM		 2420 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
2421 &fs_info->generic_worker);
5443be45 2422 btrfs_init_workers(&fs_info->endio_workers, "endio",
61d92c32
CM		 2423 fs_info->thread_pool_size,
2424 &fs_info->generic_worker);
d20f7043 2425 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
61d92c32
CM		 2426 fs_info->thread_pool_size,
2427 &fs_info->generic_worker);
cad321ad 2428 btrfs_init_workers(&fs_info->endio_meta_write_workers,
61d92c32
CM		 2429 "endio-meta-write", fs_info->thread_pool_size,
2430 &fs_info->generic_worker);
53b381b3
DW		 2431 btrfs_init_workers(&fs_info->endio_raid56_workers,
2432 "endio-raid56", fs_info->thread_pool_size,
2433 &fs_info->generic_worker);
2434 btrfs_init_workers(&fs_info->rmw_workers,
2435 "rmw", fs_info->thread_pool_size,
2436 &fs_info->generic_worker);
5443be45 2437 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
61d92c32
CM		 2438 fs_info->thread_pool_size,
2439 &fs_info->generic_worker);
0cb59c99
JB		 2440 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
2441 1, &fs_info->generic_worker);
16cdcec7
MX		 2442 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
2443 fs_info->thread_pool_size,
2444 &fs_info->generic_worker);
90519d66
AJ		 2445 btrfs_init_workers(&fs_info->readahead_workers, "readahead",
2446 fs_info->thread_pool_size,
2447 &fs_info->generic_worker);
2f232036
JS		 2448 btrfs_init_workers(&fs_info->qgroup_rescan_workers, "qgroup-rescan", 1,
2449 &fs_info->generic_worker);
61b49440
CM		 2450
2451 /*
2452 * endios are largely parallel and should have a very
2453 * low idle thresh
2454 */
2455 fs_info->endio_workers.idle_thresh = 4;
b51912c9 2456 fs_info->endio_meta_workers.idle_thresh = 4;
53b381b3
DW		 2457 fs_info->endio_raid56_workers.idle_thresh = 4;
2458 fs_info->rmw_workers.idle_thresh = 2;
b51912c9 2459
9042846b
CM		 2460 fs_info->endio_write_workers.idle_thresh = 2;
2461 fs_info->endio_meta_write_workers.idle_thresh = 2;
90519d66 2462 fs_info->readahead_workers.idle_thresh = 2;
9042846b 2463
0dc3b84a
JB		 2464 /*
2465 * btrfs_start_workers can really only fail because of ENOMEM so just
2466 * return -ENOMEM if any of these fail.
2467 */
2468 ret = btrfs_start_workers(&fs_info->workers);
2469 ret |= btrfs_start_workers(&fs_info->generic_worker);
2470 ret |= btrfs_start_workers(&fs_info->submit_workers);
2471 ret |= btrfs_start_workers(&fs_info->delalloc_workers);
2472 ret |= btrfs_start_workers(&fs_info->fixup_workers);
2473 ret |= btrfs_start_workers(&fs_info->endio_workers);
2474 ret |= btrfs_start_workers(&fs_info->endio_meta_workers);
53b381b3
DW		 2475 ret |= btrfs_start_workers(&fs_info->rmw_workers);
2476 ret |= btrfs_start_workers(&fs_info->endio_raid56_workers);
0dc3b84a
JB		 2477 ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers);
2478 ret |= btrfs_start_workers(&fs_info->endio_write_workers);
2479 ret |= btrfs_start_workers(&fs_info->endio_freespace_worker);
2480 ret |= btrfs_start_workers(&fs_info->delayed_workers);
2481 ret |= btrfs_start_workers(&fs_info->caching_workers);
2482 ret |= btrfs_start_workers(&fs_info->readahead_workers);
8ccf6f19 2483 ret |= btrfs_start_workers(&fs_info->flush_workers);
2f232036 2484 ret |= btrfs_start_workers(&fs_info->qgroup_rescan_workers);
0dc3b84a 2485 if (ret) {
fed425c7 2486 err = -ENOMEM;
0dc3b84a
JB		 2487 goto fail_sb_buffer;
2488 }
4543df7e 2489
4575c9cc 2490 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 2491 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2492 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 2493
db94535d
CM		 2494 tree_root->nodesize = nodesize;
2495 tree_root->leafsize = leafsize;
2496 tree_root->sectorsize = sectorsize;
87ee04eb 2497 tree_root->stripesize = stripesize;
a061fc8d
CM		 2498
2499 sb->s_blocksize = sectorsize;
2500 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 2501
cdb4c574 2502 if (disk_super->magic != cpu_to_le64(BTRFS_MAGIC)) {
d397712b 2503 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 2504 goto fail_sb_buffer;
2505 }
19c00ddc 2506
8d082fb7
LB		 2507 if (sectorsize != PAGE_SIZE) {
2508 printk(KERN_WARNING "btrfs: Incompatible sector size(%lu) "
2509 "found on %s\n", (unsigned long)sectorsize, sb->s_id);
941b2ddf
KM		 2510 goto fail_sb_buffer;
2511 }
2512
925baedd 2513 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 2514 ret = btrfs_read_sys_array(tree_root);
925baedd 2515 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 2516 if (ret) {
d397712b
CM		 2517 printk(KERN_WARNING "btrfs: failed to read the system "
2518 "array on %s\n", sb->s_id);
5d4f98a2 2519 goto fail_sb_buffer;
84eed90f 2520 }
0b86a832
CM		 2521
2522 blocksize = btrfs_level_size(tree_root,
2523 btrfs_super_chunk_root_level(disk_super));
84234f3a 2524 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 2525
2526 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2527 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2528
2529 chunk_root->node = read_tree_block(chunk_root,
2530 btrfs_super_chunk_root(disk_super),
84234f3a 2531 blocksize, generation);
416bc658
JB		 2532 if (!chunk_root->node ||
2533 !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
83121942
DW		 2534 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
2535 sb->s_id);
af31f5e5 2536 goto fail_tree_roots;
83121942 2537 }
5d4f98a2
YZ		 2538 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2539 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 2540
e17cade2 2541 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
d397712b
CM		 2542 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
2543 BTRFS_UUID_SIZE);
e17cade2 2544
0b86a832 2545 ret = btrfs_read_chunk_tree(chunk_root);
2b82032c 2546 if (ret) {
d397712b
CM		 2547 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
2548 sb->s_id);
af31f5e5 2549 goto fail_tree_roots;
2b82032c 2550 }
0b86a832 2551
8dabb742
SB		 2552 /*
2553 * keep the device that is marked to be the target device for the
2554 * dev_replace procedure
2555 */
2556 btrfs_close_extra_devices(fs_info, fs_devices, 0);
dfe25020 2557
a6b0d5c8
CM		 2558 if (!fs_devices->latest_bdev) {
2559 printk(KERN_CRIT "btrfs: failed to read devices on %s\n",
2560 sb->s_id);
2561 goto fail_tree_roots;
2562 }
2563
af31f5e5 2564retry_root_backup:
db94535d
CM		 2565 blocksize = btrfs_level_size(tree_root,
2566 btrfs_super_root_level(disk_super));
84234f3a 2567 generation = btrfs_super_generation(disk_super);
0b86a832 2568
e20d96d6 2569 tree_root->node = read_tree_block(tree_root,
db94535d 2570 btrfs_super_root(disk_super),
84234f3a 2571 blocksize, generation);
af31f5e5
CM		 2572 if (!tree_root->node ||
2573 !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
83121942
DW		 2574 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
2575 sb->s_id);
af31f5e5
CM		 2576
2577 goto recovery_tree_root;
83121942 2578 }
af31f5e5 2579
5d4f98a2
YZ		 2580 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2581 tree_root->commit_root = btrfs_root_node(tree_root);
db94535d
CM		 2582
2583 ret = find_and_setup_root(tree_root, fs_info,
e20d96d6 2584 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
0b86a832 2585 if (ret)
af31f5e5 2586 goto recovery_tree_root;
0b86a832
CM		 2587 extent_root->track_dirty = 1;
2588
2589 ret = find_and_setup_root(tree_root, fs_info,
2590 BTRFS_DEV_TREE_OBJECTID, dev_root);
0b86a832 2591 if (ret)
af31f5e5 2592 goto recovery_tree_root;
5d4f98a2 2593 dev_root->track_dirty = 1;
3768f368 2594
d20f7043
CM		 2595 ret = find_and_setup_root(tree_root, fs_info,
2596 BTRFS_CSUM_TREE_OBJECTID, csum_root);
2597 if (ret)
af31f5e5 2598 goto recovery_tree_root;
d20f7043
CM		 2599 csum_root->track_dirty = 1;
2600
bcef60f2
AJ		 2601 ret = find_and_setup_root(tree_root, fs_info,
2602 BTRFS_QUOTA_TREE_OBJECTID, quota_root);
2603 if (ret) {
2604 kfree(quota_root);
2605 quota_root = fs_info->quota_root = NULL;
2606 } else {
2607 quota_root->track_dirty = 1;
2608 fs_info->quota_enabled = 1;
2609 fs_info->pending_quota_state = 1;
2610 }
2611
8929ecfa
YZ		 2612 fs_info->generation = generation;
2613 fs_info->last_trans_committed = generation;
8929ecfa 2614
68310a5e
ID		 2615 ret = btrfs_recover_balance(fs_info);
2616 if (ret) {
2617 printk(KERN_WARNING "btrfs: failed to recover balance\n");
2618 goto fail_block_groups;
2619 }
2620
733f4fbb
SB		 2621 ret = btrfs_init_dev_stats(fs_info);
2622 if (ret) {
2623 printk(KERN_ERR "btrfs: failed to init dev_stats: %d\n",
2624 ret);
2625 goto fail_block_groups;
2626 }
2627
8dabb742
SB		 2628 ret = btrfs_init_dev_replace(fs_info);
2629 if (ret) {
2630 pr_err("btrfs: failed to init dev_replace: %d\n", ret);
2631 goto fail_block_groups;
2632 }
2633
2634 btrfs_close_extra_devices(fs_info, fs_devices, 1);
2635
c59021f8 2636 ret = btrfs_init_space_info(fs_info);
2637 if (ret) {
2638 printk(KERN_ERR "Failed to initial space info: %d\n", ret);
2639 goto fail_block_groups;
2640 }
2641
1b1d1f66
JB		 2642 ret = btrfs_read_block_groups(extent_root);
2643 if (ret) {
2644 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
2645 goto fail_block_groups;
2646 }
5af3e8cc
SB		 2647 fs_info->num_tolerated_disk_barrier_failures =
2648 btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
292fd7fc
SB		 2649 if (fs_info->fs_devices->missing_devices >
2650 fs_info->num_tolerated_disk_barrier_failures &&
2651 !(sb->s_flags & MS_RDONLY)) {
2652 printk(KERN_WARNING
2653 "Btrfs: too many missing devices, writeable mount is not allowed\n");
2654 goto fail_block_groups;
2655 }
9078a3e1 2656
a74a4b97
CM		 2657 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2658 "btrfs-cleaner");
57506d50 2659 if (IS_ERR(fs_info->cleaner_kthread))
1b1d1f66 2660 goto fail_block_groups;
a74a4b97
CM		 2661
2662 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2663 tree_root,
2664 "btrfs-transaction");
57506d50 2665 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2666 goto fail_cleaner;
a74a4b97 2667
c289811c
CM		 2668 if (!btrfs_test_opt(tree_root, SSD) &&
2669 !btrfs_test_opt(tree_root, NOSSD) &&
2670 !fs_info->fs_devices->rotating) {
2671 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
2672 "mode\n");
2673 btrfs_set_opt(fs_info->mount_opt, SSD);
2674 }
2675
21adbd5c
SB		 2676#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2677 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2678 ret = btrfsic_mount(tree_root, fs_devices,
2679 btrfs_test_opt(tree_root,
2680 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2681 1 : 0,
2682 fs_info->check_integrity_print_mask);
2683 if (ret)
2684 printk(KERN_WARNING "btrfs: failed to initialize"
2685 " integrity check module %s\n", sb->s_id);
2686 }
2687#endif
bcef60f2
AJ		 2688 ret = btrfs_read_qgroup_config(fs_info);
2689 if (ret)
2690 goto fail_trans_kthread;
21adbd5c 2691
acce952b 2692 /* do not make disk changes in broken FS */
68ce9682 2693 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM		 2694 u64 bytenr = btrfs_super_log_root(disk_super);
2695
7c2ca468 2696 if (fs_devices->rw_devices == 0) {
d397712b
CM		 2697 printk(KERN_WARNING "Btrfs log replay required "
2698 "on RO media\n");
7c2ca468 2699 err = -EIO;
bcef60f2 2700 goto fail_qgroup;
7c2ca468 2701 }
e02119d5
CM		 2702 blocksize =
2703 btrfs_level_size(tree_root,
2704 btrfs_super_log_root_level(disk_super));
d18a2c44 2705
6f07e42e 2706 log_tree_root = btrfs_alloc_root(fs_info);
676e4c86
DC		 2707 if (!log_tree_root) {
2708 err = -ENOMEM;
bcef60f2 2709 goto fail_qgroup;
676e4c86 2710 }
e02119d5
CM		 2711
2712 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2713 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2714
2715 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 2716 blocksize,
2717 generation + 1);
416bc658
JB		 2718 if (!log_tree_root->node ||
2719 !extent_buffer_uptodate(log_tree_root->node)) {
2720 printk(KERN_ERR "btrfs: failed to read log tree\n");
2721 free_extent_buffer(log_tree_root->node);
2722 kfree(log_tree_root);
2723 goto fail_trans_kthread;
2724 }
79787eaa 2725 /* returns with log_tree_root freed on success */
e02119d5 2726 ret = btrfs_recover_log_trees(log_tree_root);
79787eaa
JM		 2727 if (ret) {
2728 btrfs_error(tree_root->fs_info, ret,
2729 "Failed to recover log tree");
2730 free_extent_buffer(log_tree_root->node);
2731 kfree(log_tree_root);
2732 goto fail_trans_kthread;
2733 }
e556ce2c
YZ		 2734
2735 if (sb->s_flags & MS_RDONLY) {
79787eaa
JM		 2736 ret = btrfs_commit_super(tree_root);
2737 if (ret)
2738 goto fail_trans_kthread;
e556ce2c 2739 }
e02119d5 2740 }
1a40e23b 2741
76dda93c 2742 ret = btrfs_find_orphan_roots(tree_root);
79787eaa
JM		 2743 if (ret)
2744 goto fail_trans_kthread;
76dda93c 2745
7c2ca468 2746 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b 2747 ret = btrfs_cleanup_fs_roots(fs_info);
44c44af2
ID		 2748 if (ret)
2749 goto fail_trans_kthread;
d68fc57b 2750
5d4f98a2 2751 ret = btrfs_recover_relocation(tree_root);
d7ce5843
MX		 2752 if (ret < 0) {
2753 printk(KERN_WARNING
2754 "btrfs: failed to recover relocation\n");
2755 err = -EINVAL;
bcef60f2 2756 goto fail_qgroup;
d7ce5843 2757 }
7c2ca468 2758 }
1a40e23b 2759
3de4586c
CM		 2760 location.objectid = BTRFS_FS_TREE_OBJECTID;
2761 location.type = BTRFS_ROOT_ITEM_KEY;
2762 location.offset = (u64)-1;
2763
3de4586c
CM		 2764 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
2765 if (!fs_info->fs_root)
bcef60f2 2766 goto fail_qgroup;
3140c9a3
DC		 2767 if (IS_ERR(fs_info->fs_root)) {
2768 err = PTR_ERR(fs_info->fs_root);
bcef60f2 2769 goto fail_qgroup;
3140c9a3 2770 }
c289811c 2771
2b6ba629
ID		 2772 if (sb->s_flags & MS_RDONLY)
2773 return 0;
59641015 2774
2b6ba629
ID		 2775 down_read(&fs_info->cleanup_work_sem);
2776 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2777 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
e3acc2a6 2778 up_read(&fs_info->cleanup_work_sem);
2b6ba629
ID		 2779 close_ctree(tree_root);
2780 return ret;
2781 }
2782 up_read(&fs_info->cleanup_work_sem);
59641015 2783
2b6ba629
ID		 2784 ret = btrfs_resume_balance_async(fs_info);
2785 if (ret) {
2786 printk(KERN_WARNING "btrfs: failed to resume balance\n");
2787 close_ctree(tree_root);
2788 return ret;
e3acc2a6
JB		 2789 }
2790
8dabb742
SB		 2791 ret = btrfs_resume_dev_replace_async(fs_info);
2792 if (ret) {
2793 pr_warn("btrfs: failed to resume dev_replace\n");
2794 close_ctree(tree_root);
2795 return ret;
2796 }
2797
ad2b2c80 2798 return 0;
39279cc3 2799
bcef60f2
AJ		 2800fail_qgroup:
2801 btrfs_free_qgroup_config(fs_info);
7c2ca468
CM		 2802fail_trans_kthread:
2803 kthread_stop(fs_info->transaction_kthread);
171f6537 2804 del_fs_roots(fs_info);
54067ae9 2805 btrfs_cleanup_transaction(fs_info->tree_root);
3f157a2f 2806fail_cleaner:
a74a4b97 2807 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2808
2809 /*
2810 * make sure we're done with the btree inode before we stop our
2811 * kthreads
2812 */
2813 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
7c2ca468 2814
1b1d1f66 2815fail_block_groups:
54067ae9 2816 btrfs_put_block_group_cache(fs_info);
1b1d1f66 2817 btrfs_free_block_groups(fs_info);
af31f5e5
CM		 2818
2819fail_tree_roots:
2820 free_root_pointers(fs_info, 1);
2b8195bb 2821 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
af31f5e5 2822
39279cc3 2823fail_sb_buffer:
7abadb64 2824 btrfs_stop_all_workers(fs_info);
16cdcec7 2825fail_alloc:
4543df7e 2826fail_iput:
586e46e2
ID		 2827 btrfs_mapping_tree_free(&fs_info->mapping_tree);
2828
4543df7e 2829 iput(fs_info->btree_inode);
963d678b
MX		 2830fail_delalloc_bytes:
2831 percpu_counter_destroy(&fs_info->delalloc_bytes);
e2d84521
MX		 2832fail_dirty_metadata_bytes:
2833 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
ad081f14 2834fail_bdi:
7e662854 2835 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 2836fail_srcu:
2837 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 2838fail:
53b381b3 2839 btrfs_free_stripe_hash_table(fs_info);
586e46e2 2840 btrfs_close_devices(fs_info->fs_devices);
ad2b2c80 2841 return err;
af31f5e5
CM		 2842
2843recovery_tree_root:
af31f5e5
CM		 2844 if (!btrfs_test_opt(tree_root, RECOVERY))
2845 goto fail_tree_roots;
2846
2847 free_root_pointers(fs_info, 0);
2848
2849 /* don't use the log in recovery mode, it won't be valid */
2850 btrfs_set_super_log_root(disk_super, 0);
2851
2852 /* we can't trust the free space cache either */
2853 btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
2854
2855 ret = next_root_backup(fs_info, fs_info->super_copy,
2856 &num_backups_tried, &backup_index);
2857 if (ret == -1)
2858 goto fail_block_groups;
2859 goto retry_root_backup;
eb60ceac
CM		 2860}
2861
f2984462
CM		 2862static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2863{
f2984462
CM		 2864 if (uptodate) {
2865 set_buffer_uptodate(bh);
2866 } else {
442a4f63
SB		 2867 struct btrfs_device *device = (struct btrfs_device *)
2868 bh->b_private;
2869
606686ee
JB		 2870 printk_ratelimited_in_rcu(KERN_WARNING "lost page write due to "
2871 "I/O error on %s\n",
2872 rcu_str_deref(device->name));
1259ab75
CM		 2873 /* note, we dont' set_buffer_write_io_error because we have
2874 * our own ways of dealing with the IO errors
2875 */
f2984462 2876 clear_buffer_uptodate(bh);
442a4f63 2877 btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
f2984462
CM		 2878 }
2879 unlock_buffer(bh);
2880 put_bh(bh);
2881}
2882
a512bbf8
YZ		 2883struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
2884{
2885 struct buffer_head *bh;
2886 struct buffer_head *latest = NULL;
2887 struct btrfs_super_block *super;
2888 int i;
2889 u64 transid = 0;
2890 u64 bytenr;
2891
2892 /* we would like to check all the supers, but that would make
2893 * a btrfs mount succeed after a mkfs from a different FS.
2894 * So, we need to add a special mount option to scan for
2895 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2896 */
2897 for (i = 0; i < 1; i++) {
2898 bytenr = btrfs_sb_offset(i);
2899 if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
2900 break;
2901 bh = __bread(bdev, bytenr / 4096, 4096);
2902 if (!bh)
2903 continue;
2904
2905 super = (struct btrfs_super_block *)bh->b_data;
2906 if (btrfs_super_bytenr(super) != bytenr ||
cdb4c574 2907 super->magic != cpu_to_le64(BTRFS_MAGIC)) {
a512bbf8
YZ		 2908 brelse(bh);
2909 continue;
2910 }
2911
2912 if (!latest || btrfs_super_generation(super) > transid) {
2913 brelse(latest);
2914 latest = bh;
2915 transid = btrfs_super_generation(super);
2916 } else {
2917 brelse(bh);
2918 }
2919 }
2920 return latest;
2921}
2922
4eedeb75
HH		 2923/*
2924 * this should be called twice, once with wait == 0 and
2925 * once with wait == 1. When wait == 0 is done, all the buffer heads
2926 * we write are pinned.
2927 *
2928 * They are released when wait == 1 is done.
2929 * max_mirrors must be the same for both runs, and it indicates how
2930 * many supers on this one device should be written.
2931 *
2932 * max_mirrors == 0 means to write them all.
2933 */
a512bbf8
YZ		 2934static int write_dev_supers(struct btrfs_device *device,
2935 struct btrfs_super_block *sb,
2936 int do_barriers, int wait, int max_mirrors)
2937{
2938 struct buffer_head *bh;
2939 int i;
2940 int ret;
2941 int errors = 0;
2942 u32 crc;
2943 u64 bytenr;
a512bbf8
YZ		 2944
2945 if (max_mirrors == 0)
2946 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
2947
a512bbf8
YZ		 2948 for (i = 0; i < max_mirrors; i++) {
2949 bytenr = btrfs_sb_offset(i);
2950 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
2951 break;
2952
2953 if (wait) {
2954 bh = __find_get_block(device->bdev, bytenr / 4096,
2955 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 2956 if (!bh) {
2957 errors++;
2958 continue;
2959 }
a512bbf8 2960 wait_on_buffer(bh);
4eedeb75
HH		 2961 if (!buffer_uptodate(bh))
2962 errors++;
2963
2964 /* drop our reference */
2965 brelse(bh);
2966
2967 /* drop the reference from the wait == 0 run */
2968 brelse(bh);
2969 continue;
a512bbf8
YZ		 2970 } else {
2971 btrfs_set_super_bytenr(sb, bytenr);
2972
2973 crc = ~(u32)0;
b0496686 2974 crc = btrfs_csum_data((char *)sb +
a512bbf8
YZ		 2975 BTRFS_CSUM_SIZE, crc,
2976 BTRFS_SUPER_INFO_SIZE -
2977 BTRFS_CSUM_SIZE);
2978 btrfs_csum_final(crc, sb->csum);
2979
4eedeb75
HH		 2980 /*
2981 * one reference for us, and we leave it for the
2982 * caller
2983 */
a512bbf8
YZ		 2984 bh = __getblk(device->bdev, bytenr / 4096,
2985 BTRFS_SUPER_INFO_SIZE);
634554dc
JB		 2986 if (!bh) {
2987 printk(KERN_ERR "btrfs: couldn't get super "
2988 "buffer head for bytenr %Lu\n", bytenr);
2989 errors++;
2990 continue;
2991 }
2992
a512bbf8
YZ		 2993 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
2994
4eedeb75 2995 /* one reference for submit_bh */
a512bbf8 2996 get_bh(bh);
4eedeb75
HH		 2997
2998 set_buffer_uptodate(bh);
a512bbf8
YZ		 2999 lock_buffer(bh);
3000 bh->b_end_io = btrfs_end_buffer_write_sync;
442a4f63 3001 bh->b_private = device;
a512bbf8
YZ		 3002 }
3003
387125fc
CM		 3004 /*
3005 * we fua the first super. The others we allow
3006 * to go down lazy.
3007 */
21adbd5c 3008 ret = btrfsic_submit_bh(WRITE_FUA, bh);
4eedeb75 3009 if (ret)
a512bbf8 3010 errors++;
a512bbf8
YZ		 3011 }
3012 return errors < i ? 0 : -1;
3013}
3014
387125fc
CM		 3015/*
3016 * endio for the write_dev_flush, this will wake anyone waiting
3017 * for the barrier when it is done
3018 */
3019static void btrfs_end_empty_barrier(struct bio *bio, int err)
3020{
3021 if (err) {
3022 if (err == -EOPNOTSUPP)
3023 set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3024 clear_bit(BIO_UPTODATE, &bio->bi_flags);
3025 }
3026 if (bio->bi_private)
3027 complete(bio->bi_private);
3028 bio_put(bio);
3029}
3030
3031/*
3032 * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3033 * sent down. With wait == 1, it waits for the previous flush.
3034 *
3035 * any device where the flush fails with eopnotsupp are flagged as not-barrier
3036 * capable
3037 */
3038static int write_dev_flush(struct btrfs_device *device, int wait)
3039{
3040 struct bio *bio;
3041 int ret = 0;
3042
3043 if (device->nobarriers)
3044 return 0;
3045
3046 if (wait) {
3047 bio = device->flush_bio;
3048 if (!bio)
3049 return 0;
3050
3051 wait_for_completion(&device->flush_wait);
3052
3053 if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
606686ee
JB		 3054 printk_in_rcu("btrfs: disabling barriers on dev %s\n",
3055 rcu_str_deref(device->name));
387125fc 3056 device->nobarriers = 1;
5af3e8cc 3057 } else if (!bio_flagged(bio, BIO_UPTODATE)) {
387125fc 3058 ret = -EIO;
5af3e8cc
SB		 3059 btrfs_dev_stat_inc_and_print(device,
3060 BTRFS_DEV_STAT_FLUSH_ERRS);
387125fc
CM		 3061 }
3062
3063 /* drop the reference from the wait == 0 run */
3064 bio_put(bio);
3065 device->flush_bio = NULL;
3066
3067 return ret;
3068 }
3069
3070 /*
3071 * one reference for us, and we leave it for the
3072 * caller
3073 */
9c017abc 3074 device->flush_bio = NULL;
387125fc
CM		 3075 bio = bio_alloc(GFP_NOFS, 0);
3076 if (!bio)
3077 return -ENOMEM;
3078
3079 bio->bi_end_io = btrfs_end_empty_barrier;
3080 bio->bi_bdev = device->bdev;
3081 init_completion(&device->flush_wait);
3082 bio->bi_private = &device->flush_wait;
3083 device->flush_bio = bio;
3084
3085 bio_get(bio);
21adbd5c 3086 btrfsic_submit_bio(WRITE_FLUSH, bio);
387125fc
CM		 3087
3088 return 0;
3089}
3090
3091/*
3092 * send an empty flush down to each device in parallel,
3093 * then wait for them
3094 */
3095static int barrier_all_devices(struct btrfs_fs_info *info)
3096{
3097 struct list_head *head;
3098 struct btrfs_device *dev;
5af3e8cc
SB		 3099 int errors_send = 0;
3100 int errors_wait = 0;
387125fc
CM		 3101 int ret;
3102
3103 /* send down all the barriers */
3104 head = &info->fs_devices->devices;
3105 list_for_each_entry_rcu(dev, head, dev_list) {
3106 if (!dev->bdev) {
5af3e8cc 3107 errors_send++;
387125fc
CM		 3108 continue;
3109 }
3110 if (!dev->in_fs_metadata || !dev->writeable)
3111 continue;
3112
3113 ret = write_dev_flush(dev, 0);
3114 if (ret)
5af3e8cc 3115 errors_send++;
387125fc
CM		 3116 }
3117
3118 /* wait for all the barriers */
3119 list_for_each_entry_rcu(dev, head, dev_list) {
3120 if (!dev->bdev) {
5af3e8cc 3121 errors_wait++;
387125fc
CM		 3122 continue;
3123 }
3124 if (!dev->in_fs_metadata || !dev->writeable)
3125 continue;
3126
3127 ret = write_dev_flush(dev, 1);
3128 if (ret)
5af3e8cc 3129 errors_wait++;
387125fc 3130 }
5af3e8cc
SB		 3131 if (errors_send > info->num_tolerated_disk_barrier_failures ||
3132 errors_wait > info->num_tolerated_disk_barrier_failures)
387125fc
CM		 3133 return -EIO;
3134 return 0;
3135}
3136
5af3e8cc
SB		 3137int btrfs_calc_num_tolerated_disk_barrier_failures(
3138 struct btrfs_fs_info *fs_info)
3139{
3140 struct btrfs_ioctl_space_info space;
3141 struct btrfs_space_info *sinfo;
3142 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3143 BTRFS_BLOCK_GROUP_SYSTEM,
3144 BTRFS_BLOCK_GROUP_METADATA,
3145 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3146 int num_types = 4;
3147 int i;
3148 int c;
3149 int num_tolerated_disk_barrier_failures =
3150 (int)fs_info->fs_devices->num_devices;
3151
3152 for (i = 0; i < num_types; i++) {
3153 struct btrfs_space_info *tmp;
3154
3155 sinfo = NULL;
3156 rcu_read_lock();
3157 list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3158 if (tmp->flags == types[i]) {
3159 sinfo = tmp;
3160 break;
3161 }
3162 }
3163 rcu_read_unlock();
3164
3165 if (!sinfo)
3166 continue;
3167
3168 down_read(&sinfo->groups_sem);
3169 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3170 if (!list_empty(&sinfo->block_groups[c])) {
3171 u64 flags;
3172
3173 btrfs_get_block_group_info(
3174 &sinfo->block_groups[c], &space);
3175 if (space.total_bytes == 0 ||
3176 space.used_bytes == 0)
3177 continue;
3178 flags = space.flags;
3179 /*
3180 * return
3181 * 0: if dup, single or RAID0 is configured for
3182 * any of metadata, system or data, else
3183 * 1: if RAID5 is configured, or if RAID1 or
3184 * RAID10 is configured and only two mirrors
3185 * are used, else
3186 * 2: if RAID6 is configured, else
3187 * num_mirrors - 1: if RAID1 or RAID10 is
3188 * configured and more than
3189 * 2 mirrors are used.
3190 */
3191 if (num_tolerated_disk_barrier_failures > 0 &&
3192 ((flags & (BTRFS_BLOCK_GROUP_DUP |
3193 BTRFS_BLOCK_GROUP_RAID0)) ||
3194 ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3195 == 0)))
3196 num_tolerated_disk_barrier_failures = 0;
53b381b3
DW		 3197 else if (num_tolerated_disk_barrier_failures > 1) {
3198 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3199 BTRFS_BLOCK_GROUP_RAID5 |
3200 BTRFS_BLOCK_GROUP_RAID10)) {
3201 num_tolerated_disk_barrier_failures = 1;
3202 } else if (flags &
3203 BTRFS_BLOCK_GROUP_RAID5) {
3204 num_tolerated_disk_barrier_failures = 2;
3205 }
3206 }
5af3e8cc
SB		 3207 }
3208 }
3209 up_read(&sinfo->groups_sem);
3210 }
3211
3212 return num_tolerated_disk_barrier_failures;
3213}
3214
48a3b636 3215static int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 3216{
e5e9a520 3217 struct list_head *head;
f2984462 3218 struct btrfs_device *dev;
a061fc8d 3219 struct btrfs_super_block *sb;
f2984462 3220 struct btrfs_dev_item *dev_item;
f2984462
CM		 3221 int ret;
3222 int do_barriers;
a236aed1
CM		 3223 int max_errors;
3224 int total_errors = 0;
a061fc8d 3225 u64 flags;
f2984462 3226
6c41761f 3227 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
f2984462 3228 do_barriers = !btrfs_test_opt(root, NOBARRIER);
af31f5e5 3229 backup_super_roots(root->fs_info);
f2984462 3230
6c41761f 3231 sb = root->fs_info->super_for_commit;
a061fc8d 3232 dev_item = &sb->dev_item;
e5e9a520 3233
174ba509 3234 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 3235 head = &root->fs_info->fs_devices->devices;
387125fc 3236
5af3e8cc
SB		 3237 if (do_barriers) {
3238 ret = barrier_all_devices(root->fs_info);
3239 if (ret) {
3240 mutex_unlock(
3241 &root->fs_info->fs_devices->device_list_mutex);
3242 btrfs_error(root->fs_info, ret,
3243 "errors while submitting device barriers.");
3244 return ret;
3245 }
3246 }
387125fc 3247
1f78160c 3248 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3249 if (!dev->bdev) {
3250 total_errors++;
3251 continue;
3252 }
2b82032c 3253 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3254 continue;
3255
2b82032c 3256 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 3257 btrfs_set_stack_device_type(dev_item, dev->type);
3258 btrfs_set_stack_device_id(dev_item, dev->devid);
3259 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
3260 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
3261 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3262 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3263 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3264 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 3265 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 3266
a061fc8d
CM		 3267 flags = btrfs_super_flags(sb);
3268 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3269
a512bbf8 3270 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 3271 if (ret)
3272 total_errors++;
f2984462 3273 }
a236aed1 3274 if (total_errors > max_errors) {
d397712b
CM		 3275 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
3276 total_errors);
79787eaa
JM		 3277
3278 /* This shouldn't happen. FUA is masked off if unsupported */
a236aed1
CM		 3279 BUG();
3280 }
f2984462 3281
a512bbf8 3282 total_errors = 0;
1f78160c 3283 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 3284 if (!dev->bdev)
3285 continue;
2b82032c 3286 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 3287 continue;
3288
a512bbf8
YZ		 3289 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3290 if (ret)
3291 total_errors++;
f2984462 3292 }
174ba509 3293 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 3294 if (total_errors > max_errors) {
79787eaa
JM		 3295 btrfs_error(root->fs_info, -EIO,
3296 "%d errors while writing supers", total_errors);
3297 return -EIO;
a236aed1 3298 }
f2984462
CM		 3299 return 0;
3300}
3301
a512bbf8
YZ		 3302int write_ctree_super(struct btrfs_trans_handle *trans,
3303 struct btrfs_root *root, int max_mirrors)
eb60ceac 3304{
e66f709b 3305 int ret;
5f39d397 3306
a512bbf8 3307 ret = write_all_supers(root, max_mirrors);
5f39d397 3308 return ret;
cfaa7295
CM		 3309}
3310
143bede5 3311void btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
2619ba1f 3312{
4df27c4d 3313 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 3314 radix_tree_delete(&fs_info->fs_roots_radix,
3315 (unsigned long)root->root_key.objectid);
4df27c4d 3316 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 3317
3318 if (btrfs_root_refs(&root->root_item) == 0)
3319 synchronize_srcu(&fs_info->subvol_srcu);
3320
d7634482 3321 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
3321719e
LB		 3322 btrfs_free_log(NULL, root);
3323 btrfs_free_log_root_tree(NULL, fs_info);
3324 }
3325
581bb050
LZ		 3326 __btrfs_remove_free_space_cache(root->free_ino_pinned);
3327 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d 3328 free_fs_root(root);
4df27c4d
YZ		 3329}
3330
3331static void free_fs_root(struct btrfs_root *root)
3332{
82d5902d 3333 iput(root->cache_inode);
4df27c4d 3334 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
0ee5dc67
AV		 3335 if (root->anon_dev)
3336 free_anon_bdev(root->anon_dev);
4df27c4d
YZ		 3337 free_extent_buffer(root->node);
3338 free_extent_buffer(root->commit_root);
581bb050
LZ		 3339 kfree(root->free_ino_ctl);
3340 kfree(root->free_ino_pinned);
d397712b 3341 kfree(root->name);
2619ba1f 3342 kfree(root);
2619ba1f
CM		 3343}
3344
c146afad 3345int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 3346{
c146afad
YZ		 3347 u64 root_objectid = 0;
3348 struct btrfs_root *gang[8];
3349 int i;
3768f368 3350 int ret;
e089f05c 3351
c146afad
YZ		 3352 while (1) {
3353 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3354 (void **)gang, root_objectid,
3355 ARRAY_SIZE(gang));
3356 if (!ret)
3357 break;
5d4f98a2
YZ		 3358
3359 root_objectid = gang[ret - 1]->root_key.objectid + 1;
c146afad 3360 for (i = 0; i < ret; i++) {
66b4ffd1
JB		 3361 int err;
3362
c146afad 3363 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 3364 err = btrfs_orphan_cleanup(gang[i]);
3365 if (err)
3366 return err;
c146afad
YZ		 3367 }
3368 root_objectid++;
3369 }
3370 return 0;
3371}
a2135011 3372
c146afad
YZ		 3373int btrfs_commit_super(struct btrfs_root *root)
3374{
3375 struct btrfs_trans_handle *trans;
3376 int ret;
a74a4b97 3377
c146afad 3378 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 3379 btrfs_run_delayed_iputs(root);
c146afad 3380 mutex_unlock(&root->fs_info->cleaner_mutex);
9d1a2a3a 3381 wake_up_process(root->fs_info->cleaner_kthread);
c71bf099
YZ		 3382
3383 /* wait until ongoing cleanup work done */
3384 down_write(&root->fs_info->cleanup_work_sem);
3385 up_write(&root->fs_info->cleanup_work_sem);
3386
7a7eaa40 3387 trans = btrfs_join_transaction(root);
3612b495
TI		 3388 if (IS_ERR(trans))
3389 return PTR_ERR(trans);
54aa1f4d 3390 ret = btrfs_commit_transaction(trans, root);
79787eaa
JM		 3391 if (ret)
3392 return ret;
c146afad 3393 /* run commit again to drop the original snapshot */
7a7eaa40 3394 trans = btrfs_join_transaction(root);
3612b495
TI		 3395 if (IS_ERR(trans))
3396 return PTR_ERR(trans);
79787eaa
JM		 3397 ret = btrfs_commit_transaction(trans, root);
3398 if (ret)
3399 return ret;
79154b1b 3400 ret = btrfs_write_and_wait_transaction(NULL, root);
79787eaa
JM		 3401 if (ret) {
3402 btrfs_error(root->fs_info, ret,
3403 "Failed to sync btree inode to disk.");
3404 return ret;
3405 }
d6bfde87 3406
a512bbf8 3407 ret = write_ctree_super(NULL, root, 0);
c146afad
YZ		 3408 return ret;
3409}
3410
3411int close_ctree(struct btrfs_root *root)
3412{
3413 struct btrfs_fs_info *fs_info = root->fs_info;
3414 int ret;
3415
3416 fs_info->closing = 1;
3417 smp_mb();
3418
837d5b6e 3419 /* pause restriper - we want to resume on mount */
aa1b8cd4 3420 btrfs_pause_balance(fs_info);
837d5b6e 3421
8dabb742
SB		 3422 btrfs_dev_replace_suspend_for_unmount(fs_info);
3423
aa1b8cd4 3424 btrfs_scrub_cancel(fs_info);
4cb5300b
CM		 3425
3426 /* wait for any defraggers to finish */
3427 wait_event(fs_info->transaction_wait,
3428 (atomic_read(&fs_info->defrag_running) == 0));
3429
3430 /* clear out the rbtree of defraggable inodes */
26176e7c 3431 btrfs_cleanup_defrag_inodes(fs_info);
4cb5300b 3432
c146afad 3433 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 3434 ret = btrfs_commit_super(root);
3435 if (ret)
3436 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
3437 }
3438
87533c47 3439 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
68ce9682 3440 btrfs_error_commit_super(root);
0f7d52f4 3441
300e4f8a
JB		 3442 btrfs_put_block_group_cache(fs_info);
3443
e3029d9f
AV		 3444 kthread_stop(fs_info->transaction_kthread);
3445 kthread_stop(fs_info->cleaner_kthread);
8929ecfa 3446
f25784b3
YZ		 3447 fs_info->closing = 2;
3448 smp_mb();
3449
bcef60f2
AJ		 3450 btrfs_free_qgroup_config(root->fs_info);
3451
963d678b
MX		 3452 if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
3453 printk(KERN_INFO "btrfs: at unmount delalloc count %lld\n",
3454 percpu_counter_sum(&fs_info->delalloc_bytes));
b0c68f8b 3455 }
bcc63abb 3456
be283b2e 3457 free_root_pointers(fs_info, 1);
d20f7043 3458
e3029d9f 3459 btrfs_free_block_groups(fs_info);
d10c5f31 3460
c146afad 3461 del_fs_roots(fs_info);
d10c5f31 3462
c146afad 3463 iput(fs_info->btree_inode);
9ad6b7bc 3464
7abadb64 3465 btrfs_stop_all_workers(fs_info);
d6bfde87 3466
21adbd5c
SB		 3467#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3468 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3469 btrfsic_unmount(root, fs_info->fs_devices);
3470#endif
3471
dfe25020 3472 btrfs_close_devices(fs_info->fs_devices);
0b86a832 3473 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 3474
e2d84521 3475 percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
963d678b 3476 percpu_counter_destroy(&fs_info->delalloc_bytes);
04160088 3477 bdi_destroy(&fs_info->bdi);
76dda93c 3478 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 3479
53b381b3
DW		 3480 btrfs_free_stripe_hash_table(fs_info);
3481
eb60ceac
CM		 3482 return 0;
3483}
3484
b9fab919
CM		 3485int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3486 int atomic)
5f39d397 3487{
1259ab75 3488 int ret;
727011e0 3489 struct inode *btree_inode = buf->pages[0]->mapping->host;
1259ab75 3490
0b32f4bb 3491 ret = extent_buffer_uptodate(buf);
1259ab75
CM		 3492 if (!ret)
3493 return ret;
3494
3495 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
b9fab919
CM		 3496 parent_transid, atomic);
3497 if (ret == -EAGAIN)
3498 return ret;
1259ab75 3499 return !ret;
5f39d397
CM		 3500}
3501
3502int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 3503{
0b32f4bb 3504 return set_extent_buffer_uptodate(buf);
5f39d397 3505}
6702ed49 3506
5f39d397
CM		 3507void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3508{
727011e0 3509 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
5f39d397 3510 u64 transid = btrfs_header_generation(buf);
b9473439 3511 int was_dirty;
b4ce94de 3512
b9447ef8 3513 btrfs_assert_tree_locked(buf);
31b1a2bd
JL		 3514 if (transid != root->fs_info->generation)
3515 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
d397712b 3516 "found %llu running %llu\n",
db94535d 3517 (unsigned long long)buf->start,
d397712b
CM		 3518 (unsigned long long)transid,
3519 (unsigned long long)root->fs_info->generation);
0b32f4bb 3520 was_dirty = set_extent_buffer_dirty(buf);
e2d84521
MX		 3521 if (!was_dirty)
3522 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3523 buf->len,
3524 root->fs_info->dirty_metadata_batch);
eb60ceac
CM		 3525}
3526
b53d3f5d
LB		 3527static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3528 int flush_delayed)
16cdcec7
MX		 3529{
3530 /*
3531 * looks as though older kernels can get into trouble with
3532 * this code, they end up stuck in balance_dirty_pages forever
3533 */
e2d84521 3534 int ret;
16cdcec7
MX		 3535
3536 if (current->flags & PF_MEMALLOC)
3537 return;
3538
b53d3f5d
LB		 3539 if (flush_delayed)
3540 btrfs_balance_delayed_items(root);
16cdcec7 3541
e2d84521
MX		 3542 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3543 BTRFS_DIRTY_METADATA_THRESH);
3544 if (ret > 0) {
d0e1d66b
NJ		 3545 balance_dirty_pages_ratelimited(
3546 root->fs_info->btree_inode->i_mapping);
16cdcec7
MX		 3547 }
3548 return;
3549}
3550
b53d3f5d 3551void btrfs_btree_balance_dirty(struct btrfs_root *root)
35b7e476 3552{
b53d3f5d
LB		 3553 __btrfs_btree_balance_dirty(root, 1);
3554}
585ad2c3 3555
b53d3f5d
LB		 3556void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3557{
3558 __btrfs_btree_balance_dirty(root, 0);
35b7e476 3559}
6b80053d 3560
ca7a79ad 3561int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 3562{
727011e0 3563 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
0b32f4bb 3564 return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
6b80053d 3565}
0da5468f 3566
fcd1f065 3567static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
acce952b 3568 int read_only)
3569{
fcd1f065
DS		 3570 if (btrfs_super_csum_type(fs_info->super_copy) >= ARRAY_SIZE(btrfs_csum_sizes)) {
3571 printk(KERN_ERR "btrfs: unsupported checksum algorithm\n");
3572 return -EINVAL;
3573 }
3574
acce952b 3575 if (read_only)
fcd1f065 3576 return 0;
acce952b 3577
fcd1f065 3578 return 0;
acce952b 3579}
3580
48a3b636 3581static void btrfs_error_commit_super(struct btrfs_root *root)
acce952b 3582{
acce952b 3583 mutex_lock(&root->fs_info->cleaner_mutex);
3584 btrfs_run_delayed_iputs(root);
3585 mutex_unlock(&root->fs_info->cleaner_mutex);
3586
3587 down_write(&root->fs_info->cleanup_work_sem);
3588 up_write(&root->fs_info->cleanup_work_sem);
3589
3590 /* cleanup FS via transaction */
3591 btrfs_cleanup_transaction(root);
acce952b 3592}
3593
569e0f35
JB		 3594static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
3595 struct btrfs_root *root)
acce952b 3596{
3597 struct btrfs_inode *btrfs_inode;
3598 struct list_head splice;
3599
3600 INIT_LIST_HEAD(&splice);
3601
3602 mutex_lock(&root->fs_info->ordered_operations_mutex);
3603 spin_lock(&root->fs_info->ordered_extent_lock);
3604
569e0f35 3605 list_splice_init(&t->ordered_operations, &splice);
acce952b 3606 while (!list_empty(&splice)) {
3607 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3608 ordered_operations);
3609
3610 list_del_init(&btrfs_inode->ordered_operations);
3611
3612 btrfs_invalidate_inodes(btrfs_inode->root);
3613 }
3614
3615 spin_unlock(&root->fs_info->ordered_extent_lock);
3616 mutex_unlock(&root->fs_info->ordered_operations_mutex);
acce952b 3617}
3618
143bede5 3619static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
acce952b 3620{
acce952b 3621 struct btrfs_ordered_extent *ordered;
acce952b 3622
3623 spin_lock(&root->fs_info->ordered_extent_lock);
779880ef
JB		 3624 /*
3625 * This will just short circuit the ordered completion stuff which will
3626 * make sure the ordered extent gets properly cleaned up.
3627 */
3628 list_for_each_entry(ordered, &root->fs_info->ordered_extents,
3629 root_extent_list)
3630 set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
acce952b 3631 spin_unlock(&root->fs_info->ordered_extent_lock);
acce952b 3632}
3633
79787eaa
JM		 3634int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3635 struct btrfs_root *root)
acce952b 3636{
3637 struct rb_node *node;
3638 struct btrfs_delayed_ref_root *delayed_refs;
3639 struct btrfs_delayed_ref_node *ref;
3640 int ret = 0;
3641
3642 delayed_refs = &trans->delayed_refs;
3643
3644 spin_lock(&delayed_refs->lock);
3645 if (delayed_refs->num_entries == 0) {
cfece4db 3646 spin_unlock(&delayed_refs->lock);
acce952b 3647 printk(KERN_INFO "delayed_refs has NO entry\n");
3648 return ret;
3649 }
3650
b939d1ab 3651 while ((node = rb_first(&delayed_refs->root)) != NULL) {
eb12db69 3652 struct btrfs_delayed_ref_head *head = NULL;
acce952b 3653
eb12db69 3654 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
acce952b 3655 atomic_set(&ref->refs, 1);
3656 if (btrfs_delayed_ref_is_head(ref)) {
acce952b 3657
3658 head = btrfs_delayed_node_to_head(ref);
b939d1ab
JB		 3659 if (!mutex_trylock(&head->mutex)) {
3660 atomic_inc(&ref->refs);
3661 spin_unlock(&delayed_refs->lock);
3662
3663 /* Need to wait for the delayed ref to run */
3664 mutex_lock(&head->mutex);
3665 mutex_unlock(&head->mutex);
3666 btrfs_put_delayed_ref(ref);
3667
e18fca73 3668 spin_lock(&delayed_refs->lock);
b939d1ab
JB		 3669 continue;
3670 }
3671
54067ae9
JB		 3672 if (head->must_insert_reserved)
3673 btrfs_pin_extent(root, ref->bytenr,
3674 ref->num_bytes, 1);
78a6184a 3675 btrfs_free_delayed_extent_op(head->extent_op);
acce952b 3676 delayed_refs->num_heads--;
3677 if (list_empty(&head->cluster))
3678 delayed_refs->num_heads_ready--;
3679 list_del_init(&head->cluster);
acce952b 3680 }
eb12db69 3681
b939d1ab
JB		 3682 ref->in_tree = 0;
3683 rb_erase(&ref->rb_node, &delayed_refs->root);
3684 delayed_refs->num_entries--;
eb12db69
JB		 3685 if (head)
3686 mutex_unlock(&head->mutex);
acce952b 3687 spin_unlock(&delayed_refs->lock);
3688 btrfs_put_delayed_ref(ref);
3689
3690 cond_resched();
3691 spin_lock(&delayed_refs->lock);
3692 }
3693
3694 spin_unlock(&delayed_refs->lock);
3695
3696 return ret;
3697}
3698
aec8030a 3699static void btrfs_evict_pending_snapshots(struct btrfs_transaction *t)
acce952b 3700{
3701 struct btrfs_pending_snapshot *snapshot;
3702 struct list_head splice;
3703
3704 INIT_LIST_HEAD(&splice);
3705
3706 list_splice_init(&t->pending_snapshots, &splice);
3707
3708 while (!list_empty(&splice)) {
3709 snapshot = list_entry(splice.next,
3710 struct btrfs_pending_snapshot,
3711 list);
aec8030a 3712 snapshot->error = -ECANCELED;
acce952b 3713 list_del_init(&snapshot->list);
acce952b 3714 }
acce952b 3715}
3716
143bede5 3717static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
acce952b 3718{
3719 struct btrfs_inode *btrfs_inode;
3720 struct list_head splice;
3721
3722 INIT_LIST_HEAD(&splice);
3723
acce952b 3724 spin_lock(&root->fs_info->delalloc_lock);
5be76758 3725 list_splice_init(&root->fs_info->delalloc_inodes, &splice);
acce952b 3726
3727 while (!list_empty(&splice)) {
3728 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
3729 delalloc_inodes);
3730
3731 list_del_init(&btrfs_inode->delalloc_inodes);
df0af1a5
MX		 3732 clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3733 &btrfs_inode->runtime_flags);
acce952b 3734
3735 btrfs_invalidate_inodes(btrfs_inode->root);
3736 }
3737
3738 spin_unlock(&root->fs_info->delalloc_lock);
acce952b 3739}
3740
3741static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3742 struct extent_io_tree *dirty_pages,
3743 int mark)
3744{
3745 int ret;
acce952b 3746 struct extent_buffer *eb;
3747 u64 start = 0;
3748 u64 end;
acce952b 3749
3750 while (1) {
3751 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
e6138876 3752 mark, NULL);
acce952b 3753 if (ret)
3754 break;
3755
3756 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
3757 while (start <= end) {
fd8b2b61
JB		 3758 eb = btrfs_find_tree_block(root, start,
3759 root->leafsize);
3760 start += eb->len;
3761 if (!eb)
acce952b 3762 continue;
fd8b2b61 3763 wait_on_extent_buffer_writeback(eb);
acce952b 3764
fd8b2b61
JB		 3765 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
3766 &eb->bflags))
3767 clear_extent_buffer_dirty(eb);
3768 free_extent_buffer_stale(eb);
acce952b 3769 }
3770 }
3771
3772 return ret;
3773}
3774
3775static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3776 struct extent_io_tree *pinned_extents)
3777{
3778 struct extent_io_tree *unpin;
3779 u64 start;
3780 u64 end;
3781 int ret;
ed0eaa14 3782 bool loop = true;
acce952b 3783
3784 unpin = pinned_extents;
ed0eaa14 3785again:
acce952b 3786 while (1) {
3787 ret = find_first_extent_bit(unpin, 0, &start, &end,
e6138876 3788 EXTENT_DIRTY, NULL);
acce952b 3789 if (ret)
3790 break;
3791
3792 /* opt_discard */
5378e607
LD		 3793 if (btrfs_test_opt(root, DISCARD))
3794 ret = btrfs_error_discard_extent(root, start,
3795 end + 1 - start,
3796 NULL);
acce952b 3797
3798 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3799 btrfs_error_unpin_extent_range(root, start, end);
3800 cond_resched();
3801 }
3802
ed0eaa14
LB		 3803 if (loop) {
3804 if (unpin == &root->fs_info->freed_extents[0])
3805 unpin = &root->fs_info->freed_extents[1];
3806 else
3807 unpin = &root->fs_info->freed_extents[0];
3808 loop = false;
3809 goto again;
3810 }
3811
acce952b 3812 return 0;
3813}
3814
49b25e05
JM		 3815void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
3816 struct btrfs_root *root)
3817{
3818 btrfs_destroy_delayed_refs(cur_trans, root);
3819 btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv,
3820 cur_trans->dirty_pages.dirty_bytes);
3821
3822 /* FIXME: cleanup wait for commit */
3823 cur_trans->in_commit = 1;
3824 cur_trans->blocked = 1;
d7096fc3 3825 wake_up(&root->fs_info->transaction_blocked_wait);
49b25e05 3826
aec8030a
MX		 3827 btrfs_evict_pending_snapshots(cur_trans);
3828
49b25e05 3829 cur_trans->blocked = 0;
d7096fc3 3830 wake_up(&root->fs_info->transaction_wait);
49b25e05
JM		 3831
3832 cur_trans->commit_done = 1;
d7096fc3 3833 wake_up(&cur_trans->commit_wait);
49b25e05 3834
67cde344
MX		 3835 btrfs_destroy_delayed_inodes(root);
3836 btrfs_assert_delayed_root_empty(root);
49b25e05 3837
49b25e05
JM		 3838 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
3839 EXTENT_DIRTY);
6e841e32
LB		 3840 btrfs_destroy_pinned_extent(root,
3841 root->fs_info->pinned_extents);
49b25e05
JM		 3842
3843 /*
3844 memset(cur_trans, 0, sizeof(*cur_trans));
3845 kmem_cache_free(btrfs_transaction_cachep, cur_trans);
3846 */
3847}
3848
48a3b636 3849static int btrfs_cleanup_transaction(struct btrfs_root *root)
acce952b 3850{
3851 struct btrfs_transaction *t;
3852 LIST_HEAD(list);
3853
acce952b 3854 mutex_lock(&root->fs_info->transaction_kthread_mutex);
3855
a4abeea4 3856 spin_lock(&root->fs_info->trans_lock);
acce952b 3857 list_splice_init(&root->fs_info->trans_list, &list);
a4abeea4
JB		 3858 root->fs_info->trans_no_join = 1;
3859 spin_unlock(&root->fs_info->trans_lock);
3860
acce952b 3861 while (!list_empty(&list)) {
3862 t = list_entry(list.next, struct btrfs_transaction, list);
acce952b 3863
569e0f35 3864 btrfs_destroy_ordered_operations(t, root);
acce952b 3865
3866 btrfs_destroy_ordered_extents(root);
3867
3868 btrfs_destroy_delayed_refs(t, root);
3869
acce952b 3870 /* FIXME: cleanup wait for commit */
3871 t->in_commit = 1;
3872 t->blocked = 1;
66657b31 3873 smp_mb();
acce952b 3874 if (waitqueue_active(&root->fs_info->transaction_blocked_wait))
3875 wake_up(&root->fs_info->transaction_blocked_wait);
3876
aec8030a
MX		 3877 btrfs_evict_pending_snapshots(t);
3878
acce952b 3879 t->blocked = 0;
66657b31 3880 smp_mb();
acce952b 3881 if (waitqueue_active(&root->fs_info->transaction_wait))
3882 wake_up(&root->fs_info->transaction_wait);
acce952b 3883
acce952b 3884 t->commit_done = 1;
66657b31 3885 smp_mb();
acce952b 3886 if (waitqueue_active(&t->commit_wait))
3887 wake_up(&t->commit_wait);
acce952b 3888
67cde344
MX		 3889 btrfs_destroy_delayed_inodes(root);
3890 btrfs_assert_delayed_root_empty(root);
3891
acce952b 3892 btrfs_destroy_delalloc_inodes(root);
3893
a4abeea4 3894 spin_lock(&root->fs_info->trans_lock);
acce952b 3895 root->fs_info->running_transaction = NULL;
a4abeea4 3896 spin_unlock(&root->fs_info->trans_lock);
acce952b 3897
3898 btrfs_destroy_marked_extents(root, &t->dirty_pages,
3899 EXTENT_DIRTY);
3900
3901 btrfs_destroy_pinned_extent(root,
3902 root->fs_info->pinned_extents);
3903
13c5a93e 3904 atomic_set(&t->use_count, 0);
acce952b 3905 list_del_init(&t->list);
3906 memset(t, 0, sizeof(*t));
3907 kmem_cache_free(btrfs_transaction_cachep, t);
3908 }
3909
a4abeea4
JB		 3910 spin_lock(&root->fs_info->trans_lock);
3911 root->fs_info->trans_no_join = 0;
3912 spin_unlock(&root->fs_info->trans_lock);
acce952b 3913 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 3914
3915 return 0;
3916}
3917
d1310b2e 3918static struct extent_io_ops btree_extent_io_ops = {
ce9adaa5 3919 .readpage_end_io_hook = btree_readpage_end_io_hook,
4bb31e92 3920 .readpage_io_failed_hook = btree_io_failed_hook,
0b86a832 3921 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 3922 /* note we're sharing with inode.c for the merge bio hook */
3923 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 3924};
kernel source for telekom puls (alcatel/mediatek)