projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: separate superblock items out of fs_info
[linux-2.6-block.git] / fs / btrfs / disk-io.c
CommitLineData
6cbd5570
CM		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
e20d96d6 19#include <linux/fs.h>
d98237b3 20#include <linux/blkdev.h>
87cbda5c 21#include <linux/scatterlist.h>
22b0ebda 22#include <linux/swap.h>
0f7d52f4 23#include <linux/radix-tree.h>
35b7e476 24#include <linux/writeback.h>
d397712b 25#include <linux/buffer_head.h>
ce9adaa5 26#include <linux/workqueue.h>
a74a4b97 27#include <linux/kthread.h>
4b4e25f2 28#include <linux/freezer.h>
163e783e 29#include <linux/crc32c.h>
5a0e3ad6 30#include <linux/slab.h>
784b4e29 31#include <linux/migrate.h>
7a36ddec 32#include <linux/ratelimit.h>
7e75bf3f 33#include <asm/unaligned.h>
4b4e25f2 34#include "compat.h"
eb60ceac
CM		 35#include "ctree.h"
36#include "disk-io.h"
e089f05c 37#include "transaction.h"
0f7d52f4 38#include "btrfs_inode.h"
0b86a832 39#include "volumes.h"
db94535d 40#include "print-tree.h"
8b712842 41#include "async-thread.h"
925baedd 42#include "locking.h"
e02119d5 43#include "tree-log.h"
fa9c0d79 44#include "free-space-cache.h"
581bb050 45#include "inode-map.h"
eb60ceac 46
d1310b2e 47static struct extent_io_ops btree_extent_io_ops;
8b712842 48static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 49static void free_fs_root(struct btrfs_root *root);
acce952b 50static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
51 int read_only);
52static int btrfs_destroy_ordered_operations(struct btrfs_root *root);
53static int btrfs_destroy_ordered_extents(struct btrfs_root *root);
54static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
55 struct btrfs_root *root);
56static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t);
57static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
58static int btrfs_destroy_marked_extents(struct btrfs_root *root,
59 struct extent_io_tree *dirty_pages,
60 int mark);
61static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
62 struct extent_io_tree *pinned_extents);
63static int btrfs_cleanup_transaction(struct btrfs_root *root);
ce9adaa5 64
d352ac68
CM		 65/*
66 * end_io_wq structs are used to do processing in task context when an IO is
67 * complete. This is used during reads to verify checksums, and it is used
68 * by writes to insert metadata for new file extents after IO is complete.
69 */
ce9adaa5
CM		 70struct end_io_wq {
71 struct bio *bio;
72 bio_end_io_t *end_io;
73 void *private;
74 struct btrfs_fs_info *info;
75 int error;
22c59948 76 int metadata;
ce9adaa5 77 struct list_head list;
8b712842 78 struct btrfs_work work;
ce9adaa5 79};
0da5468f 80
d352ac68
CM		 81/*
82 * async submit bios are used to offload expensive checksumming
83 * onto the worker threads. They checksum file and metadata bios
84 * just before they are sent down the IO stack.
85 */
44b8bd7e
CM		 86struct async_submit_bio {
87 struct inode *inode;
88 struct bio *bio;
89 struct list_head list;
4a69a410
CM		 90 extent_submit_bio_hook_t *submit_bio_start;
91 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM		 92 int rw;
93 int mirror_num;
c8b97818 94 unsigned long bio_flags;
eaf25d93
CM		 95 /*
96 * bio_offset is optional, can be used if the pages in the bio
97 * can't tell us where in the file the bio should go
98 */
99 u64 bio_offset;
8b712842 100 struct btrfs_work work;
44b8bd7e
CM		 101};
102
85d4e461
CM		 103/*
104 * Lockdep class keys for extent_buffer->lock's in this root. For a given
105 * eb, the lockdep key is determined by the btrfs_root it belongs to and
106 * the level the eb occupies in the tree.
107 *
108 * Different roots are used for different purposes and may nest inside each
109 * other and they require separate keysets. As lockdep keys should be
110 * static, assign keysets according to the purpose of the root as indicated
111 * by btrfs_root->objectid. This ensures that all special purpose roots
112 * have separate keysets.
4008c04a 113 *
85d4e461
CM		 114 * Lock-nesting across peer nodes is always done with the immediate parent
115 * node locked thus preventing deadlock. As lockdep doesn't know this, use
116 * subclass to avoid triggering lockdep warning in such cases.
4008c04a 117 *
85d4e461
CM		 118 * The key is set by the readpage_end_io_hook after the buffer has passed
119 * csum validation but before the pages are unlocked. It is also set by
120 * btrfs_init_new_buffer on freshly allocated blocks.
4008c04a 121 *
85d4e461
CM		 122 * We also add a check to make sure the highest level of the tree is the
123 * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
124 * needs update as well.
4008c04a
CM		 125 */
126#ifdef CONFIG_DEBUG_LOCK_ALLOC
127# if BTRFS_MAX_LEVEL != 8
128# error
129# endif
85d4e461
CM		 130
131static struct btrfs_lockdep_keyset {
132 u64 id; /* root objectid */
133 const char *name_stem; /* lock name stem */
134 char names[BTRFS_MAX_LEVEL + 1][20];
135 struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
136} btrfs_lockdep_keysets[] = {
137 { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
138 { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
139 { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
140 { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
141 { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
142 { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
143 { .id = BTRFS_ORPHAN_OBJECTID, .name_stem = "orphan" },
144 { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
145 { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
146 { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
147 { .id = 0, .name_stem = "tree" },
4008c04a 148};
85d4e461
CM		 149
150void __init btrfs_init_lockdep(void)
151{
152 int i, j;
153
154 /* initialize lockdep class names */
155 for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
156 struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
157
158 for (j = 0; j < ARRAY_SIZE(ks->names); j++)
159 snprintf(ks->names[j], sizeof(ks->names[j]),
160 "btrfs-%s-%02d", ks->name_stem, j);
161 }
162}
163
164void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
165 int level)
166{
167 struct btrfs_lockdep_keyset *ks;
168
169 BUG_ON(level >= ARRAY_SIZE(ks->keys));
170
171 /* find the matching keyset, id 0 is the default entry */
172 for (ks = btrfs_lockdep_keysets; ks->id; ks++)
173 if (ks->id == objectid)
174 break;
175
176 lockdep_set_class_and_name(&eb->lock,
177 &ks->keys[level], ks->names[level]);
178}
179
4008c04a
CM		 180#endif
181
d352ac68
CM		 182/*
183 * extents on the btree inode are pretty simple, there's one extent
184 * that covers the entire device
185 */
b2950863 186static struct extent_map *btree_get_extent(struct inode *inode,
306e16ce 187 struct page *page, size_t pg_offset, u64 start, u64 len,
b2950863 188 int create)
7eccb903 189{
5f39d397
CM		 190 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
191 struct extent_map *em;
192 int ret;
193
890871be 194 read_lock(&em_tree->lock);
d1310b2e 195 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 196 if (em) {
197 em->bdev =
198 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 199 read_unlock(&em_tree->lock);
5f39d397 200 goto out;
a061fc8d 201 }
890871be 202 read_unlock(&em_tree->lock);
7b13b7b1 203
172ddd60 204 em = alloc_extent_map();
5f39d397
CM		 205 if (!em) {
206 em = ERR_PTR(-ENOMEM);
207 goto out;
208 }
209 em->start = 0;
0afbaf8c 210 em->len = (u64)-1;
c8b97818 211 em->block_len = (u64)-1;
5f39d397 212 em->block_start = 0;
a061fc8d 213 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 214
890871be 215 write_lock(&em_tree->lock);
5f39d397
CM		 216 ret = add_extent_mapping(em_tree, em);
217 if (ret == -EEXIST) {
0afbaf8c
CM		 218 u64 failed_start = em->start;
219 u64 failed_len = em->len;
220
5f39d397 221 free_extent_map(em);
7b13b7b1 222 em = lookup_extent_mapping(em_tree, start, len);
0afbaf8c 223 if (em) {
7b13b7b1 224 ret = 0;
0afbaf8c
CM		 225 } else {
226 em = lookup_extent_mapping(em_tree, failed_start,
227 failed_len);
7b13b7b1 228 ret = -EIO;
0afbaf8c 229 }
5f39d397 230 } else if (ret) {
7b13b7b1
CM		 231 free_extent_map(em);
232 em = NULL;
5f39d397 233 }
890871be 234 write_unlock(&em_tree->lock);
7b13b7b1
CM		 235
236 if (ret)
237 em = ERR_PTR(ret);
5f39d397
CM		 238out:
239 return em;
7eccb903
CM		 240}
241
19c00ddc
CM		 242u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
243{
163e783e 244 return crc32c(seed, data, len);
19c00ddc
CM		 245}
246
247void btrfs_csum_final(u32 crc, char *result)
248{
7e75bf3f 249 put_unaligned_le32(~crc, result);
19c00ddc
CM		 250}
251
d352ac68
CM		 252/*
253 * compute the csum for a btree block, and either verify it or write it
254 * into the csum field of the block.
255 */
19c00ddc
CM		 256static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
257 int verify)
258{
6c41761f 259 u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
607d432d 260 char *result = NULL;
19c00ddc
CM		 261 unsigned long len;
262 unsigned long cur_len;
263 unsigned long offset = BTRFS_CSUM_SIZE;
19c00ddc
CM		 264 char *kaddr;
265 unsigned long map_start;
266 unsigned long map_len;
267 int err;
268 u32 crc = ~(u32)0;
607d432d 269 unsigned long inline_result;
19c00ddc
CM		 270
271 len = buf->len - offset;
d397712b 272 while (len > 0) {
19c00ddc 273 err = map_private_extent_buffer(buf, offset, 32,
a6591715 274 &kaddr, &map_start, &map_len);
d397712b 275 if (err)
19c00ddc 276 return 1;
19c00ddc
CM		 277 cur_len = min(len, map_len - (offset - map_start));
278 crc = btrfs_csum_data(root, kaddr + offset - map_start,
279 crc, cur_len);
280 len -= cur_len;
281 offset += cur_len;
19c00ddc 282 }
607d432d
JB		 283 if (csum_size > sizeof(inline_result)) {
284 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
285 if (!result)
286 return 1;
287 } else {
288 result = (char *)&inline_result;
289 }
290
19c00ddc
CM		 291 btrfs_csum_final(crc, result);
292
293 if (verify) {
607d432d 294 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 295 u32 val;
296 u32 found = 0;
607d432d 297 memcpy(&found, result, csum_size);
e4204ded 298
607d432d 299 read_extent_buffer(buf, &val, 0, csum_size);
7a36ddec 300 printk_ratelimited(KERN_INFO "btrfs: %s checksum verify "
193f284d
CM		 301 "failed on %llu wanted %X found %X "
302 "level %d\n",
303 root->fs_info->sb->s_id,
304 (unsigned long long)buf->start, val, found,
305 btrfs_header_level(buf));
607d432d
JB		 306 if (result != (char *)&inline_result)
307 kfree(result);
19c00ddc
CM		 308 return 1;
309 }
310 } else {
607d432d 311 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 312 }
607d432d
JB		 313 if (result != (char *)&inline_result)
314 kfree(result);
19c00ddc
CM		 315 return 0;
316}
317
d352ac68
CM		 318/*
319 * we can't consider a given block up to date unless the transid of the
320 * block matches the transid in the parent node's pointer. This is how we
321 * detect blocks that either didn't get written at all or got written
322 * in the wrong place.
323 */
1259ab75
CM		 324static int verify_parent_transid(struct extent_io_tree *io_tree,
325 struct extent_buffer *eb, u64 parent_transid)
326{
2ac55d41 327 struct extent_state *cached_state = NULL;
1259ab75
CM		 328 int ret;
329
330 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
331 return 0;
332
2ac55d41
JB		 333 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
334 0, &cached_state, GFP_NOFS);
335 if (extent_buffer_uptodate(io_tree, eb, cached_state) &&
1259ab75
CM		 336 btrfs_header_generation(eb) == parent_transid) {
337 ret = 0;
338 goto out;
339 }
7a36ddec 340 printk_ratelimited("parent transid verify failed on %llu wanted %llu "
193f284d
CM		 341 "found %llu\n",
342 (unsigned long long)eb->start,
343 (unsigned long long)parent_transid,
344 (unsigned long long)btrfs_header_generation(eb));
1259ab75 345 ret = 1;
2ac55d41 346 clear_extent_buffer_uptodate(io_tree, eb, &cached_state);
33958dc6 347out:
2ac55d41
JB		 348 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
349 &cached_state, GFP_NOFS);
1259ab75 350 return ret;
1259ab75
CM		 351}
352
d352ac68
CM		 353/*
354 * helper to read a given tree block, doing retries as required when
355 * the checksums don't match and we have alternate mirrors to try.
356 */
f188591e
CM		 357static int btree_read_extent_buffer_pages(struct btrfs_root *root,
358 struct extent_buffer *eb,
ca7a79ad 359 u64 start, u64 parent_transid)
f188591e
CM		 360{
361 struct extent_io_tree *io_tree;
362 int ret;
363 int num_copies = 0;
364 int mirror_num = 0;
365
a826d6dc 366 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
f188591e
CM		 367 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
368 while (1) {
369 ret = read_extent_buffer_pages(io_tree, eb, start, 1,
370 btree_get_extent, mirror_num);
1259ab75
CM		 371 if (!ret &&
372 !verify_parent_transid(io_tree, eb, parent_transid))
f188591e 373 return ret;
d397712b 374
a826d6dc
JB		 375 /*
376 * This buffer's crc is fine, but its contents are corrupted, so
377 * there is no reason to read the other copies, they won't be
378 * any less wrong.
379 */
380 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
381 return ret;
382
f188591e
CM		 383 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
384 eb->start, eb->len);
4235298e 385 if (num_copies == 1)
f188591e 386 return ret;
4235298e 387
f188591e 388 mirror_num++;
4235298e 389 if (mirror_num > num_copies)
f188591e 390 return ret;
f188591e 391 }
f188591e
CM		 392 return -EIO;
393}
19c00ddc 394
d352ac68 395/*
d397712b
CM		 396 * checksum a dirty tree block before IO. This has extra checks to make sure
397 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 398 */
d397712b 399
b2950863 400static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
19c00ddc 401{
d1310b2e 402 struct extent_io_tree *tree;
35ebb934 403 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
19c00ddc 404 u64 found_start;
19c00ddc
CM		 405 unsigned long len;
406 struct extent_buffer *eb;
f188591e
CM		 407 int ret;
408
d1310b2e 409 tree = &BTRFS_I(page->mapping->host)->io_tree;
19c00ddc 410
eb14ab8e
CM		 411 if (page->private == EXTENT_PAGE_PRIVATE) {
412 WARN_ON(1);
19c00ddc 413 goto out;
eb14ab8e
CM		 414 }
415 if (!page->private) {
416 WARN_ON(1);
19c00ddc 417 goto out;
eb14ab8e 418 }
19c00ddc 419 len = page->private >> 2;
d397712b
CM		 420 WARN_ON(len == 0);
421
ba144192 422 eb = alloc_extent_buffer(tree, start, len, page);
91ca338d
TI		 423 if (eb == NULL) {
424 WARN_ON(1);
425 goto out;
426 }
ca7a79ad
CM		 427 ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
428 btrfs_header_generation(eb));
f188591e 429 BUG_ON(ret);
784b4e29
CM		 430 WARN_ON(!btrfs_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN));
431
19c00ddc
CM		 432 found_start = btrfs_header_bytenr(eb);
433 if (found_start != start) {
55c69072
CM		 434 WARN_ON(1);
435 goto err;
436 }
437 if (eb->first_page != page) {
55c69072
CM		 438 WARN_ON(1);
439 goto err;
440 }
441 if (!PageUptodate(page)) {
55c69072
CM		 442 WARN_ON(1);
443 goto err;
19c00ddc 444 }
19c00ddc 445 csum_tree_block(root, eb, 0);
55c69072 446err:
19c00ddc
CM		 447 free_extent_buffer(eb);
448out:
449 return 0;
450}
451
2b82032c
YZ		 452static int check_tree_block_fsid(struct btrfs_root *root,
453 struct extent_buffer *eb)
454{
455 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
456 u8 fsid[BTRFS_UUID_SIZE];
457 int ret = 1;
458
459 read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
460 BTRFS_FSID_SIZE);
461 while (fs_devices) {
462 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
463 ret = 0;
464 break;
465 }
466 fs_devices = fs_devices->seed;
467 }
468 return ret;
469}
470
a826d6dc
JB		 471#define CORRUPT(reason, eb, root, slot) \
472 printk(KERN_CRIT "btrfs: corrupt leaf, %s: block=%llu," \
473 "root=%llu, slot=%d\n", reason, \
474 (unsigned long long)btrfs_header_bytenr(eb), \
475 (unsigned long long)root->objectid, slot)
476
477static noinline int check_leaf(struct btrfs_root *root,
478 struct extent_buffer *leaf)
479{
480 struct btrfs_key key;
481 struct btrfs_key leaf_key;
482 u32 nritems = btrfs_header_nritems(leaf);
483 int slot;
484
485 if (nritems == 0)
486 return 0;
487
488 /* Check the 0 item */
489 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
490 BTRFS_LEAF_DATA_SIZE(root)) {
491 CORRUPT("invalid item offset size pair", leaf, root, 0);
492 return -EIO;
493 }
494
495 /*
496 * Check to make sure each items keys are in the correct order and their
497 * offsets make sense. We only have to loop through nritems-1 because
498 * we check the current slot against the next slot, which verifies the
499 * next slot's offset+size makes sense and that the current's slot
500 * offset is correct.
501 */
502 for (slot = 0; slot < nritems - 1; slot++) {
503 btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
504 btrfs_item_key_to_cpu(leaf, &key, slot + 1);
505
506 /* Make sure the keys are in the right order */
507 if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
508 CORRUPT("bad key order", leaf, root, slot);
509 return -EIO;
510 }
511
512 /*
513 * Make sure the offset and ends are right, remember that the
514 * item data starts at the end of the leaf and grows towards the
515 * front.
516 */
517 if (btrfs_item_offset_nr(leaf, slot) !=
518 btrfs_item_end_nr(leaf, slot + 1)) {
519 CORRUPT("slot offset bad", leaf, root, slot);
520 return -EIO;
521 }
522
523 /*
524 * Check to make sure that we don't point outside of the leaf,
525 * just incase all the items are consistent to eachother, but
526 * all point outside of the leaf.
527 */
528 if (btrfs_item_end_nr(leaf, slot) >
529 BTRFS_LEAF_DATA_SIZE(root)) {
530 CORRUPT("slot end outside of leaf", leaf, root, slot);
531 return -EIO;
532 }
533 }
534
535 return 0;
536}
537
b2950863 538static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
ce9adaa5
CM		 539 struct extent_state *state)
540{
541 struct extent_io_tree *tree;
542 u64 found_start;
543 int found_level;
544 unsigned long len;
545 struct extent_buffer *eb;
546 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 547 int ret = 0;
ce9adaa5
CM		 548
549 tree = &BTRFS_I(page->mapping->host)->io_tree;
550 if (page->private == EXTENT_PAGE_PRIVATE)
551 goto out;
552 if (!page->private)
553 goto out;
d397712b 554
ce9adaa5 555 len = page->private >> 2;
d397712b
CM		 556 WARN_ON(len == 0);
557
ba144192 558 eb = alloc_extent_buffer(tree, start, len, page);
91ca338d
TI		 559 if (eb == NULL) {
560 ret = -EIO;
561 goto out;
562 }
f188591e 563
ce9adaa5 564 found_start = btrfs_header_bytenr(eb);
23a07867 565 if (found_start != start) {
7a36ddec 566 printk_ratelimited(KERN_INFO "btrfs bad tree block start "
193f284d
CM		 567 "%llu %llu\n",
568 (unsigned long long)found_start,
569 (unsigned long long)eb->start);
f188591e 570 ret = -EIO;
ce9adaa5
CM		 571 goto err;
572 }
573 if (eb->first_page != page) {
d397712b
CM		 574 printk(KERN_INFO "btrfs bad first page %lu %lu\n",
575 eb->first_page->index, page->index);
ce9adaa5 576 WARN_ON(1);
f188591e 577 ret = -EIO;
ce9adaa5
CM		 578 goto err;
579 }
2b82032c 580 if (check_tree_block_fsid(root, eb)) {
7a36ddec 581 printk_ratelimited(KERN_INFO "btrfs bad fsid on block %llu\n",
193f284d 582 (unsigned long long)eb->start);
1259ab75
CM		 583 ret = -EIO;
584 goto err;
585 }
ce9adaa5
CM		 586 found_level = btrfs_header_level(eb);
587
85d4e461
CM		 588 btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
589 eb, found_level);
4008c04a 590
ce9adaa5 591 ret = csum_tree_block(root, eb, 1);
a826d6dc 592 if (ret) {
f188591e 593 ret = -EIO;
a826d6dc
JB		 594 goto err;
595 }
596
597 /*
598 * If this is a leaf block and it is corrupt, set the corrupt bit so
599 * that we don't try and read the other copies of this block, just
600 * return -EIO.
601 */
602 if (found_level == 0 && check_leaf(root, eb)) {
603 set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
604 ret = -EIO;
605 }
ce9adaa5
CM		 606
607 end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
608 end = eb->start + end - 1;
ce9adaa5
CM		 609err:
610 free_extent_buffer(eb);
611out:
f188591e 612 return ret;
ce9adaa5
CM		 613}
614
ce9adaa5 615static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM		 616{
617 struct end_io_wq *end_io_wq = bio->bi_private;
618 struct btrfs_fs_info *fs_info;
ce9adaa5 619
ce9adaa5 620 fs_info = end_io_wq->info;
ce9adaa5 621 end_io_wq->error = err;
8b712842
CM		 622 end_io_wq->work.func = end_workqueue_fn;
623 end_io_wq->work.flags = 0;
d20f7043 624
7b6d91da 625 if (bio->bi_rw & REQ_WRITE) {
0cb59c99 626 if (end_io_wq->metadata == 1)
cad321ad
CM		 627 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
628 &end_io_wq->work);
0cb59c99
JB		 629 else if (end_io_wq->metadata == 2)
630 btrfs_queue_worker(&fs_info->endio_freespace_worker,
631 &end_io_wq->work);
cad321ad
CM		 632 else
633 btrfs_queue_worker(&fs_info->endio_write_workers,
634 &end_io_wq->work);
d20f7043
CM		 635 } else {
636 if (end_io_wq->metadata)
637 btrfs_queue_worker(&fs_info->endio_meta_workers,
638 &end_io_wq->work);
639 else
640 btrfs_queue_worker(&fs_info->endio_workers,
641 &end_io_wq->work);
642 }
ce9adaa5
CM		 643}
644
0cb59c99
JB		 645/*
646 * For the metadata arg you want
647 *
648 * 0 - if data
649 * 1 - if normal metadta
650 * 2 - if writing to the free space cache area
651 */
22c59948
CM		 652int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
653 int metadata)
0b86a832 654{
ce9adaa5 655 struct end_io_wq *end_io_wq;
ce9adaa5
CM		 656 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
657 if (!end_io_wq)
658 return -ENOMEM;
659
660 end_io_wq->private = bio->bi_private;
661 end_io_wq->end_io = bio->bi_end_io;
22c59948 662 end_io_wq->info = info;
ce9adaa5
CM		 663 end_io_wq->error = 0;
664 end_io_wq->bio = bio;
22c59948 665 end_io_wq->metadata = metadata;
ce9adaa5
CM		 666
667 bio->bi_private = end_io_wq;
668 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 669 return 0;
670}
671
b64a2851 672unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 673{
4854ddd0
CM		 674 unsigned long limit = min_t(unsigned long,
675 info->workers.max_workers,
676 info->fs_devices->open_devices);
677 return 256 * limit;
678}
0986fe9e 679
4a69a410
CM		 680static void run_one_async_start(struct btrfs_work *work)
681{
4a69a410
CM		 682 struct async_submit_bio *async;
683
684 async = container_of(work, struct async_submit_bio, work);
4a69a410 685 async->submit_bio_start(async->inode, async->rw, async->bio,
eaf25d93
CM		 686 async->mirror_num, async->bio_flags,
687 async->bio_offset);
4a69a410
CM		 688}
689
690static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 691{
692 struct btrfs_fs_info *fs_info;
693 struct async_submit_bio *async;
4854ddd0 694 int limit;
8b712842
CM		 695
696 async = container_of(work, struct async_submit_bio, work);
697 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 698
b64a2851 699 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 700 limit = limit * 2 / 3;
701
8b712842 702 atomic_dec(&fs_info->nr_async_submits);
0986fe9e 703
b64a2851
CM		 704 if (atomic_read(&fs_info->nr_async_submits) < limit &&
705 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 706 wake_up(&fs_info->async_submit_wait);
707
4a69a410 708 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 709 async->mirror_num, async->bio_flags,
710 async->bio_offset);
4a69a410
CM		 711}
712
713static void run_one_async_free(struct btrfs_work *work)
714{
715 struct async_submit_bio *async;
716
717 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 718 kfree(async);
719}
720
44b8bd7e
CM		 721int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
722 int rw, struct bio *bio, int mirror_num,
c8b97818 723 unsigned long bio_flags,
eaf25d93 724 u64 bio_offset,
4a69a410
CM		 725 extent_submit_bio_hook_t *submit_bio_start,
726 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 727{
728 struct async_submit_bio *async;
729
730 async = kmalloc(sizeof(*async), GFP_NOFS);
731 if (!async)
732 return -ENOMEM;
733
734 async->inode = inode;
735 async->rw = rw;
736 async->bio = bio;
737 async->mirror_num = mirror_num;
4a69a410
CM		 738 async->submit_bio_start = submit_bio_start;
739 async->submit_bio_done = submit_bio_done;
740
741 async->work.func = run_one_async_start;
742 async->work.ordered_func = run_one_async_done;
743 async->work.ordered_free = run_one_async_free;
744
8b712842 745 async->work.flags = 0;
c8b97818 746 async->bio_flags = bio_flags;
eaf25d93 747 async->bio_offset = bio_offset;
8c8bee1d 748
cb03c743 749 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 750
7b6d91da 751 if (rw & REQ_SYNC)
d313d7a3
CM		 752 btrfs_set_work_high_prio(&async->work);
753
8b712842 754 btrfs_queue_worker(&fs_info->workers, &async->work);
9473f16c 755
d397712b 756 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 757 atomic_read(&fs_info->nr_async_submits)) {
758 wait_event(fs_info->async_submit_wait,
759 (atomic_read(&fs_info->nr_async_submits) == 0));
760 }
761
44b8bd7e
CM		 762 return 0;
763}
764
ce3ed71a
CM		 765static int btree_csum_one_bio(struct bio *bio)
766{
767 struct bio_vec *bvec = bio->bi_io_vec;
768 int bio_index = 0;
769 struct btrfs_root *root;
770
771 WARN_ON(bio->bi_vcnt <= 0);
d397712b 772 while (bio_index < bio->bi_vcnt) {
ce3ed71a
CM		 773 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
774 csum_dirty_buffer(root, bvec->bv_page);
775 bio_index++;
776 bvec++;
777 }
778 return 0;
779}
780
4a69a410
CM		 781static int __btree_submit_bio_start(struct inode *inode, int rw,
782 struct bio *bio, int mirror_num,
eaf25d93
CM		 783 unsigned long bio_flags,
784 u64 bio_offset)
22c59948 785{
8b712842
CM		 786 /*
787 * when we're called for a write, we're already in the async
5443be45 788 * submission context. Just jump into btrfs_map_bio
8b712842 789 */
4a69a410
CM		 790 btree_csum_one_bio(bio);
791 return 0;
792}
22c59948 793
4a69a410 794static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 795 int mirror_num, unsigned long bio_flags,
796 u64 bio_offset)
4a69a410 797{
8b712842 798 /*
4a69a410
CM		 799 * when we're called for a write, we're already in the async
800 * submission context. Just jump into btrfs_map_bio
8b712842 801 */
8b712842 802 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
0b86a832
CM		 803}
804
44b8bd7e 805static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 806 int mirror_num, unsigned long bio_flags,
807 u64 bio_offset)
44b8bd7e 808{
cad321ad
CM		 809 int ret;
810
811 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
812 bio, 1);
813 BUG_ON(ret);
814
7b6d91da 815 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 816 /*
817 * called for a read, do the setup so that checksum validation
818 * can happen in the async kernel threads
819 */
4a69a410 820 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
6f3577bd 821 mirror_num, 0);
44b8bd7e 822 }
d313d7a3 823
cad321ad
CM		 824 /*
825 * kthread helpers are used to submit writes so that checksumming
826 * can happen in parallel across all CPUs
827 */
44b8bd7e 828 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
c8b97818 829 inode, rw, bio, mirror_num, 0,
eaf25d93 830 bio_offset,
4a69a410
CM		 831 __btree_submit_bio_start,
832 __btree_submit_bio_done);
44b8bd7e
CM		 833}
834
3dd1462e 835#ifdef CONFIG_MIGRATION
784b4e29
CM		 836static int btree_migratepage(struct address_space *mapping,
837 struct page *newpage, struct page *page)
838{
839 /*
840 * we can't safely write a btree page from here,
841 * we haven't done the locking hook
842 */
843 if (PageDirty(page))
844 return -EAGAIN;
845 /*
846 * Buffers may be managed in a filesystem specific way.
847 * We must have no buffers or drop them.
848 */
849 if (page_has_private(page) &&
850 !try_to_release_page(page, GFP_KERNEL))
851 return -EAGAIN;
784b4e29
CM		 852 return migrate_page(mapping, newpage, page);
853}
3dd1462e 854#endif
784b4e29 855
0da5468f
CM		 856static int btree_writepage(struct page *page, struct writeback_control *wbc)
857{
d1310b2e 858 struct extent_io_tree *tree;
b9473439
CM		 859 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
860 struct extent_buffer *eb;
861 int was_dirty;
862
d1310b2e 863 tree = &BTRFS_I(page->mapping->host)->io_tree;
b9473439
CM		 864 if (!(current->flags & PF_MEMALLOC)) {
865 return extent_write_full_page(tree, page,
866 btree_get_extent, wbc);
867 }
5443be45 868
b9473439 869 redirty_page_for_writepage(wbc, page);
784b4e29 870 eb = btrfs_find_tree_block(root, page_offset(page), PAGE_CACHE_SIZE);
b9473439
CM		 871 WARN_ON(!eb);
872
873 was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
874 if (!was_dirty) {
875 spin_lock(&root->fs_info->delalloc_lock);
876 root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE;
877 spin_unlock(&root->fs_info->delalloc_lock);
5443be45 878 }
b9473439
CM		 879 free_extent_buffer(eb);
880
881 unlock_page(page);
882 return 0;
5f39d397 883}
0da5468f
CM		 884
885static int btree_writepages(struct address_space *mapping,
886 struct writeback_control *wbc)
887{
d1310b2e
CM		 888 struct extent_io_tree *tree;
889 tree = &BTRFS_I(mapping->host)->io_tree;
d8d5f3e1 890 if (wbc->sync_mode == WB_SYNC_NONE) {
b9473439 891 struct btrfs_root *root = BTRFS_I(mapping->host)->root;
793955bc 892 u64 num_dirty;
24ab9cd8 893 unsigned long thresh = 32 * 1024 * 1024;
448d640b
CM		 894
895 if (wbc->for_kupdate)
896 return 0;
897
b9473439
CM		 898 /* this is a bit racy, but that's ok */
899 num_dirty = root->fs_info->dirty_metadata_bytes;
d397712b 900 if (num_dirty < thresh)
793955bc 901 return 0;
793955bc 902 }
0da5468f
CM		 903 return extent_writepages(tree, mapping, btree_get_extent, wbc);
904}
905
b2950863 906static int btree_readpage(struct file *file, struct page *page)
5f39d397 907{
d1310b2e
CM		 908 struct extent_io_tree *tree;
909 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 910 return extent_read_full_page(tree, page, btree_get_extent);
911}
22b0ebda 912
70dec807 913static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 914{
d1310b2e
CM		 915 struct extent_io_tree *tree;
916 struct extent_map_tree *map;
5f39d397 917 int ret;
d98237b3 918
98509cfc 919 if (PageWriteback(page) || PageDirty(page))
d397712b 920 return 0;
98509cfc 921
d1310b2e
CM		 922 tree = &BTRFS_I(page->mapping->host)->io_tree;
923 map = &BTRFS_I(page->mapping->host)->extent_tree;
6af118ce 924
7b13b7b1 925 ret = try_release_extent_state(map, tree, page, gfp_flags);
d397712b 926 if (!ret)
6af118ce 927 return 0;
6af118ce
CM		 928
929 ret = try_release_extent_buffer(tree, page);
5f39d397
CM		 930 if (ret == 1) {
931 ClearPagePrivate(page);
932 set_page_private(page, 0);
933 page_cache_release(page);
934 }
6af118ce 935
d98237b3
CM		 936 return ret;
937}
938
5f39d397 939static void btree_invalidatepage(struct page *page, unsigned long offset)
d98237b3 940{
d1310b2e
CM		 941 struct extent_io_tree *tree;
942 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 943 extent_invalidatepage(tree, page, offset);
944 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 945 if (PagePrivate(page)) {
d397712b
CM		 946 printk(KERN_WARNING "btrfs warning page private not zero "
947 "on page %llu\n", (unsigned long long)page_offset(page));
9ad6b7bc
CM		 948 ClearPagePrivate(page);
949 set_page_private(page, 0);
950 page_cache_release(page);
951 }
d98237b3
CM		 952}
953
7f09410b 954static const struct address_space_operations btree_aops = {
d98237b3
CM		 955 .readpage = btree_readpage,
956 .writepage = btree_writepage,
0da5468f 957 .writepages = btree_writepages,
5f39d397
CM		 958 .releasepage = btree_releasepage,
959 .invalidatepage = btree_invalidatepage,
5a92bc88 960#ifdef CONFIG_MIGRATION
784b4e29 961 .migratepage = btree_migratepage,
5a92bc88 962#endif
d98237b3
CM		 963};
964
ca7a79ad
CM		 965int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
966 u64 parent_transid)
090d1875 967{
5f39d397
CM		 968 struct extent_buffer *buf = NULL;
969 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 970 int ret = 0;
090d1875 971
db94535d 972 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 973 if (!buf)
090d1875 974 return 0;
d1310b2e 975 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
f188591e 976 buf, 0, 0, btree_get_extent, 0);
5f39d397 977 free_extent_buffer(buf);
de428b63 978 return ret;
090d1875
CM		 979}
980
0999df54
CM		 981struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
982 u64 bytenr, u32 blocksize)
983{
984 struct inode *btree_inode = root->fs_info->btree_inode;
985 struct extent_buffer *eb;
986 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
f09d1f60 987 bytenr, blocksize);
0999df54
CM		 988 return eb;
989}
990
991struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
992 u64 bytenr, u32 blocksize)
993{
994 struct inode *btree_inode = root->fs_info->btree_inode;
995 struct extent_buffer *eb;
996
997 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
ba144192 998 bytenr, blocksize, NULL);
0999df54
CM		 999 return eb;
1000}
1001
1002
e02119d5
CM		 1003int btrfs_write_tree_block(struct extent_buffer *buf)
1004{
8aa38c31
CH		 1005 return filemap_fdatawrite_range(buf->first_page->mapping, buf->start,
1006 buf->start + buf->len - 1);
e02119d5
CM		 1007}
1008
1009int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1010{
8aa38c31
CH		 1011 return filemap_fdatawait_range(buf->first_page->mapping,
1012 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 1013}
1014
0999df54 1015struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 1016 u32 blocksize, u64 parent_transid)
0999df54
CM		 1017{
1018 struct extent_buffer *buf = NULL;
0999df54
CM		 1019 int ret;
1020
0999df54
CM		 1021 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1022 if (!buf)
1023 return NULL;
0999df54 1024
ca7a79ad 1025 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
ce9adaa5 1026
d397712b 1027 if (ret == 0)
b4ce94de 1028 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
5f39d397 1029 return buf;
ce9adaa5 1030
eb60ceac
CM		 1031}
1032
e089f05c 1033int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
5f39d397 1034 struct extent_buffer *buf)
ed2ff2cb 1035{
5f39d397 1036 struct inode *btree_inode = root->fs_info->btree_inode;
55c69072 1037 if (btrfs_header_generation(buf) ==
925baedd 1038 root->fs_info->running_transaction->transid) {
b9447ef8 1039 btrfs_assert_tree_locked(buf);
b4ce94de 1040
b9473439
CM		 1041 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
1042 spin_lock(&root->fs_info->delalloc_lock);
1043 if (root->fs_info->dirty_metadata_bytes >= buf->len)
1044 root->fs_info->dirty_metadata_bytes -= buf->len;
1045 else
1046 WARN_ON(1);
1047 spin_unlock(&root->fs_info->delalloc_lock);
1048 }
b4ce94de 1049
b9473439
CM		 1050 /* ugh, clear_extent_buffer_dirty needs to lock the page */
1051 btrfs_set_lock_blocking(buf);
d1310b2e 1052 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
55c69072 1053 buf);
925baedd 1054 }
5f39d397
CM		 1055 return 0;
1056}
1057
db94535d 1058static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
87ee04eb 1059 u32 stripesize, struct btrfs_root *root,
9f5fae2f 1060 struct btrfs_fs_info *fs_info,
e20d96d6 1061 u64 objectid)
d97e63b6 1062{
cfaa7295 1063 root->node = NULL;
a28ec197 1064 root->commit_root = NULL;
db94535d
CM		 1065 root->sectorsize = sectorsize;
1066 root->nodesize = nodesize;
1067 root->leafsize = leafsize;
87ee04eb 1068 root->stripesize = stripesize;
123abc88 1069 root->ref_cows = 0;
0b86a832 1070 root->track_dirty = 0;
c71bf099 1071 root->in_radix = 0;
d68fc57b
YZ		 1072 root->orphan_item_inserted = 0;
1073 root->orphan_cleanup_state = 0;
0b86a832 1074
9f5fae2f 1075 root->fs_info = fs_info;
0f7d52f4
CM		 1076 root->objectid = objectid;
1077 root->last_trans = 0;
13a8a7c8 1078 root->highest_objectid = 0;
58176a96 1079 root->name = NULL;
6bef4d31 1080 root->inode_tree = RB_ROOT;
16cdcec7 1081 INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
f0486c68 1082 root->block_rsv = NULL;
d68fc57b 1083 root->orphan_block_rsv = NULL;
0b86a832
CM		 1084
1085 INIT_LIST_HEAD(&root->dirty_list);
7b128766 1086 INIT_LIST_HEAD(&root->orphan_list);
5d4f98a2 1087 INIT_LIST_HEAD(&root->root_list);
d68fc57b 1088 spin_lock_init(&root->orphan_lock);
5d4f98a2 1089 spin_lock_init(&root->inode_lock);
f0486c68 1090 spin_lock_init(&root->accounting_lock);
a2135011 1091 mutex_init(&root->objectid_mutex);
e02119d5 1092 mutex_init(&root->log_mutex);
7237f183
YZ		 1093 init_waitqueue_head(&root->log_writer_wait);
1094 init_waitqueue_head(&root->log_commit_wait[0]);
1095 init_waitqueue_head(&root->log_commit_wait[1]);
1096 atomic_set(&root->log_commit[0], 0);
1097 atomic_set(&root->log_commit[1], 0);
1098 atomic_set(&root->log_writers, 0);
1099 root->log_batch = 0;
1100 root->log_transid = 0;
257c62e1 1101 root->last_log_commit = 0;
d0c803c4 1102 extent_io_tree_init(&root->dirty_log_pages,
f993c883 1103 fs_info->btree_inode->i_mapping);
017e5369 1104
3768f368
CM		 1105 memset(&root->root_key, 0, sizeof(root->root_key));
1106 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 1107 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 1108 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
3f157a2f 1109 root->defrag_trans_start = fs_info->generation;
58176a96 1110 init_completion(&root->kobj_unregister);
6702ed49 1111 root->defrag_running = 0;
4d775673 1112 root->root_key.objectid = objectid;
0ee5dc67 1113 root->anon_dev = 0;
3768f368
CM		 1114 return 0;
1115}
1116
db94535d 1117static int find_and_setup_root(struct btrfs_root *tree_root,
9f5fae2f
CM		 1118 struct btrfs_fs_info *fs_info,
1119 u64 objectid,
e20d96d6 1120 struct btrfs_root *root)
3768f368
CM		 1121{
1122 int ret;
db94535d 1123 u32 blocksize;
84234f3a 1124 u64 generation;
3768f368 1125
db94535d 1126 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 1127 tree_root->sectorsize, tree_root->stripesize,
1128 root, fs_info, objectid);
3768f368
CM		 1129 ret = btrfs_find_last_root(tree_root, objectid,
1130 &root->root_item, &root->root_key);
4df27c4d
YZ		 1131 if (ret > 0)
1132 return -ENOENT;
3768f368
CM		 1133 BUG_ON(ret);
1134
84234f3a 1135 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1136 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1137 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1138 blocksize, generation);
68433b73
CM		 1139 if (!root->node || !btrfs_buffer_uptodate(root->node, generation)) {
1140 free_extent_buffer(root->node);
1141 return -EIO;
1142 }
4df27c4d 1143 root->commit_root = btrfs_root_node(root);
d97e63b6
CM		 1144 return 0;
1145}
1146
7237f183
YZ		 1147static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1148 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1149{
1150 struct btrfs_root *root;
1151 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1152 struct extent_buffer *leaf;
e02119d5
CM		 1153
1154 root = kzalloc(sizeof(*root), GFP_NOFS);
1155 if (!root)
7237f183 1156 return ERR_PTR(-ENOMEM);
e02119d5
CM		 1157
1158 __setup_root(tree_root->nodesize, tree_root->leafsize,
1159 tree_root->sectorsize, tree_root->stripesize,
1160 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1161
1162 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1163 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1164 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
7237f183
YZ		 1165 /*
1166 * log trees do not get reference counted because they go away
1167 * before a real commit is actually done. They do store pointers
1168 * to file data extents, and those reference counts still get
1169 * updated (along with back refs to the log tree).
1170 */
e02119d5
CM		 1171 root->ref_cows = 0;
1172
5d4f98a2
YZ		 1173 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
1174 BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0);
7237f183
YZ		 1175 if (IS_ERR(leaf)) {
1176 kfree(root);
1177 return ERR_CAST(leaf);
1178 }
e02119d5 1179
5d4f98a2
YZ		 1180 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1181 btrfs_set_header_bytenr(leaf, leaf->start);
1182 btrfs_set_header_generation(leaf, trans->transid);
1183 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1184 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1185 root->node = leaf;
e02119d5
CM		 1186
1187 write_extent_buffer(root->node, root->fs_info->fsid,
1188 (unsigned long)btrfs_header_fsid(root->node),
1189 BTRFS_FSID_SIZE);
1190 btrfs_mark_buffer_dirty(root->node);
1191 btrfs_tree_unlock(root->node);
7237f183
YZ		 1192 return root;
1193}
1194
1195int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1196 struct btrfs_fs_info *fs_info)
1197{
1198 struct btrfs_root *log_root;
1199
1200 log_root = alloc_log_tree(trans, fs_info);
1201 if (IS_ERR(log_root))
1202 return PTR_ERR(log_root);
1203 WARN_ON(fs_info->log_root_tree);
1204 fs_info->log_root_tree = log_root;
1205 return 0;
1206}
1207
1208int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1209 struct btrfs_root *root)
1210{
1211 struct btrfs_root *log_root;
1212 struct btrfs_inode_item *inode_item;
1213
1214 log_root = alloc_log_tree(trans, root->fs_info);
1215 if (IS_ERR(log_root))
1216 return PTR_ERR(log_root);
1217
1218 log_root->last_trans = trans->transid;
1219 log_root->root_key.offset = root->root_key.objectid;
1220
1221 inode_item = &log_root->root_item.inode;
1222 inode_item->generation = cpu_to_le64(1);
1223 inode_item->size = cpu_to_le64(3);
1224 inode_item->nlink = cpu_to_le32(1);
1225 inode_item->nbytes = cpu_to_le64(root->leafsize);
1226 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1227
5d4f98a2 1228 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1229
1230 WARN_ON(root->log_root);
1231 root->log_root = log_root;
1232 root->log_transid = 0;
257c62e1 1233 root->last_log_commit = 0;
e02119d5
CM		 1234 return 0;
1235}
1236
1237struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
1238 struct btrfs_key *location)
1239{
1240 struct btrfs_root *root;
1241 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1242 struct btrfs_path *path;
5f39d397 1243 struct extent_buffer *l;
84234f3a 1244 u64 generation;
db94535d 1245 u32 blocksize;
0f7d52f4
CM		 1246 int ret = 0;
1247
5eda7b5e 1248 root = kzalloc(sizeof(*root), GFP_NOFS);
0cf6c620 1249 if (!root)
0f7d52f4 1250 return ERR_PTR(-ENOMEM);
0f7d52f4 1251 if (location->offset == (u64)-1) {
db94535d 1252 ret = find_and_setup_root(tree_root, fs_info,
0f7d52f4
CM		 1253 location->objectid, root);
1254 if (ret) {
0f7d52f4
CM		 1255 kfree(root);
1256 return ERR_PTR(ret);
1257 }
13a8a7c8 1258 goto out;
0f7d52f4
CM		 1259 }
1260
db94535d 1261 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 1262 tree_root->sectorsize, tree_root->stripesize,
1263 root, fs_info, location->objectid);
0f7d52f4
CM		 1264
1265 path = btrfs_alloc_path();
db5b493a
TI		 1266 if (!path) {
1267 kfree(root);
1268 return ERR_PTR(-ENOMEM);
1269 }
0f7d52f4 1270 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
13a8a7c8
YZ		 1271 if (ret == 0) {
1272 l = path->nodes[0];
1273 read_extent_buffer(l, &root->root_item,
1274 btrfs_item_ptr_offset(l, path->slots[0]),
1275 sizeof(root->root_item));
1276 memcpy(&root->root_key, location, sizeof(*location));
0f7d52f4 1277 }
0f7d52f4
CM		 1278 btrfs_free_path(path);
1279 if (ret) {
5e540f77 1280 kfree(root);
13a8a7c8
YZ		 1281 if (ret > 0)
1282 ret = -ENOENT;
0f7d52f4
CM		 1283 return ERR_PTR(ret);
1284 }
13a8a7c8 1285
84234f3a 1286 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1287 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1288 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1289 blocksize, generation);
5d4f98a2 1290 root->commit_root = btrfs_root_node(root);
0f7d52f4 1291 BUG_ON(!root->node);
13a8a7c8 1292out:
08fe4db1 1293 if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
e02119d5 1294 root->ref_cows = 1;
08fe4db1
LZ		 1295 btrfs_check_and_init_root_item(&root->root_item);
1296 }
13a8a7c8 1297
5eda7b5e
CM		 1298 return root;
1299}
1300
edbd8d4e
CM		 1301struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
1302 struct btrfs_key *location)
5eda7b5e
CM		 1303{
1304 struct btrfs_root *root;
1305 int ret;
1306
edbd8d4e
CM		 1307 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1308 return fs_info->tree_root;
1309 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1310 return fs_info->extent_root;
8f18cf13
CM		 1311 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1312 return fs_info->chunk_root;
1313 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1314 return fs_info->dev_root;
0403e47e
YZ		 1315 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1316 return fs_info->csum_root;
4df27c4d
YZ		 1317again:
1318 spin_lock(&fs_info->fs_roots_radix_lock);
5eda7b5e
CM		 1319 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1320 (unsigned long)location->objectid);
4df27c4d 1321 spin_unlock(&fs_info->fs_roots_radix_lock);
5eda7b5e
CM		 1322 if (root)
1323 return root;
1324
e02119d5 1325 root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
5eda7b5e
CM		 1326 if (IS_ERR(root))
1327 return root;
3394e160 1328
581bb050 1329 root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
581bb050
LZ		 1330 root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1331 GFP_NOFS);
35a30d7c
DS		 1332 if (!root->free_ino_pinned || !root->free_ino_ctl) {
1333 ret = -ENOMEM;
581bb050 1334 goto fail;
35a30d7c 1335 }
581bb050
LZ		 1336
1337 btrfs_init_free_ino_ctl(root);
1338 mutex_init(&root->fs_commit_mutex);
1339 spin_lock_init(&root->cache_lock);
1340 init_waitqueue_head(&root->cache_wait);
1341
0ee5dc67 1342 ret = get_anon_bdev(&root->anon_dev);
ac08aedf
CM		 1343 if (ret)
1344 goto fail;
3394e160 1345
d68fc57b
YZ		 1346 if (btrfs_root_refs(&root->root_item) == 0) {
1347 ret = -ENOENT;
1348 goto fail;
1349 }
1350
1351 ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
1352 if (ret < 0)
1353 goto fail;
1354 if (ret == 0)
1355 root->orphan_item_inserted = 1;
1356
4df27c4d
YZ		 1357 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1358 if (ret)
1359 goto fail;
1360
1361 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 1362 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1363 (unsigned long)root->root_key.objectid,
0f7d52f4 1364 root);
d68fc57b 1365 if (ret == 0)
4df27c4d 1366 root->in_radix = 1;
d68fc57b 1367
4df27c4d
YZ		 1368 spin_unlock(&fs_info->fs_roots_radix_lock);
1369 radix_tree_preload_end();
0f7d52f4 1370 if (ret) {
4df27c4d
YZ		 1371 if (ret == -EEXIST) {
1372 free_fs_root(root);
1373 goto again;
1374 }
1375 goto fail;
0f7d52f4 1376 }
4df27c4d
YZ		 1377
1378 ret = btrfs_find_dead_roots(fs_info->tree_root,
1379 root->root_key.objectid);
1380 WARN_ON(ret);
edbd8d4e 1381 return root;
4df27c4d
YZ		 1382fail:
1383 free_fs_root(root);
1384 return ERR_PTR(ret);
edbd8d4e
CM		 1385}
1386
04160088
CM		 1387static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1388{
1389 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1390 int ret = 0;
04160088
CM		 1391 struct btrfs_device *device;
1392 struct backing_dev_info *bdi;
b7967db7 1393
1f78160c
XG		 1394 rcu_read_lock();
1395 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1396 if (!device->bdev)
1397 continue;
04160088
CM		 1398 bdi = blk_get_backing_dev_info(device->bdev);
1399 if (bdi && bdi_congested(bdi, bdi_bits)) {
1400 ret = 1;
1401 break;
1402 }
1403 }
1f78160c 1404 rcu_read_unlock();
04160088
CM		 1405 return ret;
1406}
1407
ad081f14
JA		 1408/*
1409 * If this fails, caller must call bdi_destroy() to get rid of the
1410 * bdi again.
1411 */
04160088
CM		 1412static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1413{
ad081f14
JA		 1414 int err;
1415
1416 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1417 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1418 if (err)
1419 return err;
1420
4575c9cc 1421 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1422 bdi->congested_fn = btrfs_congested_fn;
1423 bdi->congested_data = info;
1424 return 0;
1425}
1426
ce9adaa5
CM		 1427static int bio_ready_for_csum(struct bio *bio)
1428{
1429 u64 length = 0;
1430 u64 buf_len = 0;
1431 u64 start = 0;
1432 struct page *page;
1433 struct extent_io_tree *io_tree = NULL;
ce9adaa5
CM		 1434 struct bio_vec *bvec;
1435 int i;
1436 int ret;
1437
1438 bio_for_each_segment(bvec, bio, i) {
1439 page = bvec->bv_page;
1440 if (page->private == EXTENT_PAGE_PRIVATE) {
1441 length += bvec->bv_len;
1442 continue;
1443 }
1444 if (!page->private) {
1445 length += bvec->bv_len;
1446 continue;
1447 }
1448 length = bvec->bv_len;
1449 buf_len = page->private >> 2;
1450 start = page_offset(page) + bvec->bv_offset;
1451 io_tree = &BTRFS_I(page->mapping->host)->io_tree;
ce9adaa5
CM		 1452 }
1453 /* are we fully contained in this bio? */
1454 if (buf_len <= length)
1455 return 1;
1456
1457 ret = extent_range_uptodate(io_tree, start + length,
1458 start + buf_len - 1);
ce9adaa5
CM		 1459 return ret;
1460}
1461
8b712842
CM		 1462/*
1463 * called by the kthread helper functions to finally call the bio end_io
1464 * functions. This is where read checksum verification actually happens
1465 */
1466static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1467{
ce9adaa5 1468 struct bio *bio;
8b712842
CM		 1469 struct end_io_wq *end_io_wq;
1470 struct btrfs_fs_info *fs_info;
ce9adaa5 1471 int error;
ce9adaa5 1472
8b712842
CM		 1473 end_io_wq = container_of(work, struct end_io_wq, work);
1474 bio = end_io_wq->bio;
1475 fs_info = end_io_wq->info;
ce9adaa5 1476
cad321ad 1477 /* metadata bio reads are special because the whole tree block must
8b712842
CM		 1478 * be checksummed at once. This makes sure the entire block is in
1479 * ram and up to date before trying to verify things. For
1480 * blocksize <= pagesize, it is basically a noop
1481 */
7b6d91da 1482 if (!(bio->bi_rw & REQ_WRITE) && end_io_wq->metadata &&
cad321ad 1483 !bio_ready_for_csum(bio)) {
d20f7043 1484 btrfs_queue_worker(&fs_info->endio_meta_workers,
8b712842
CM		 1485 &end_io_wq->work);
1486 return;
1487 }
1488 error = end_io_wq->error;
1489 bio->bi_private = end_io_wq->private;
1490 bio->bi_end_io = end_io_wq->end_io;
1491 kfree(end_io_wq);
8b712842 1492 bio_endio(bio, error);
44b8bd7e
CM		 1493}
1494
a74a4b97
CM		 1495static int cleaner_kthread(void *arg)
1496{
1497 struct btrfs_root *root = arg;
1498
1499 do {
a74a4b97 1500 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
76dda93c
YZ		 1501
1502 if (!(root->fs_info->sb->s_flags & MS_RDONLY) &&
1503 mutex_trylock(&root->fs_info->cleaner_mutex)) {
24bbcf04 1504 btrfs_run_delayed_iputs(root);
76dda93c
YZ		 1505 btrfs_clean_old_snapshots(root);
1506 mutex_unlock(&root->fs_info->cleaner_mutex);
4cb5300b 1507 btrfs_run_defrag_inodes(root->fs_info);
76dda93c 1508 }
a74a4b97
CM		 1509
1510 if (freezing(current)) {
1511 refrigerator();
1512 } else {
a74a4b97 1513 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1514 if (!kthread_should_stop())
1515 schedule();
a74a4b97
CM		 1516 __set_current_state(TASK_RUNNING);
1517 }
1518 } while (!kthread_should_stop());
1519 return 0;
1520}
1521
1522static int transaction_kthread(void *arg)
1523{
1524 struct btrfs_root *root = arg;
1525 struct btrfs_trans_handle *trans;
1526 struct btrfs_transaction *cur;
8929ecfa 1527 u64 transid;
a74a4b97
CM		 1528 unsigned long now;
1529 unsigned long delay;
1530 int ret;
1531
1532 do {
a74a4b97
CM		 1533 delay = HZ * 30;
1534 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1535 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1536
a4abeea4 1537 spin_lock(&root->fs_info->trans_lock);
a74a4b97
CM		 1538 cur = root->fs_info->running_transaction;
1539 if (!cur) {
a4abeea4 1540 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1541 goto sleep;
1542 }
31153d81 1543
a74a4b97 1544 now = get_seconds();
8929ecfa
YZ		 1545 if (!cur->blocked &&
1546 (now < cur->start_time || now - cur->start_time < 30)) {
a4abeea4 1547 spin_unlock(&root->fs_info->trans_lock);
a74a4b97
CM		 1548 delay = HZ * 5;
1549 goto sleep;
1550 }
8929ecfa 1551 transid = cur->transid;
a4abeea4 1552 spin_unlock(&root->fs_info->trans_lock);
56bec294 1553
7a7eaa40 1554 trans = btrfs_join_transaction(root);
3612b495 1555 BUG_ON(IS_ERR(trans));
8929ecfa
YZ		 1556 if (transid == trans->transid) {
1557 ret = btrfs_commit_transaction(trans, root);
1558 BUG_ON(ret);
1559 } else {
1560 btrfs_end_transaction(trans, root);
1561 }
a74a4b97
CM		 1562sleep:
1563 wake_up_process(root->fs_info->cleaner_kthread);
1564 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1565
1566 if (freezing(current)) {
1567 refrigerator();
1568 } else {
a74a4b97 1569 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1570 if (!kthread_should_stop() &&
1571 !btrfs_transaction_blocked(root->fs_info))
1572 schedule_timeout(delay);
a74a4b97
CM		 1573 __set_current_state(TASK_RUNNING);
1574 }
1575 } while (!kthread_should_stop());
1576 return 0;
1577}
1578
8a4b83cc 1579struct btrfs_root *open_ctree(struct super_block *sb,
dfe25020
CM		 1580 struct btrfs_fs_devices *fs_devices,
1581 char *options)
2e635a27 1582{
db94535d
CM		 1583 u32 sectorsize;
1584 u32 nodesize;
1585 u32 leafsize;
1586 u32 blocksize;
87ee04eb 1587 u32 stripesize;
84234f3a 1588 u64 generation;
f2b636e8 1589 u64 features;
3de4586c 1590 struct btrfs_key location;
a061fc8d 1591 struct buffer_head *bh;
e02119d5 1592 struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
e20d96d6 1593 GFP_NOFS);
d20f7043
CM		 1594 struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root),
1595 GFP_NOFS);
450ba0ea 1596 struct btrfs_root *tree_root = btrfs_sb(sb);
4891aca2 1597 struct btrfs_fs_info *fs_info = NULL;
e02119d5 1598 struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
0b86a832 1599 GFP_NOFS);
e02119d5 1600 struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
0b86a832 1601 GFP_NOFS);
e02119d5
CM		 1602 struct btrfs_root *log_tree_root;
1603
eb60ceac 1604 int ret;
e58ca020 1605 int err = -EINVAL;
4543df7e 1606
2c90e5d6 1607 struct btrfs_super_block *disk_super;
8790d502 1608
4891aca2 1609 if (!extent_root || !tree_root || !tree_root->fs_info ||
d20f7043 1610 !chunk_root || !dev_root || !csum_root) {
39279cc3
CM		 1611 err = -ENOMEM;
1612 goto fail;
1613 }
4891aca2 1614 fs_info = tree_root->fs_info;
76dda93c
YZ		 1615
1616 ret = init_srcu_struct(&fs_info->subvol_srcu);
1617 if (ret) {
1618 err = ret;
1619 goto fail;
1620 }
1621
1622 ret = setup_bdi(fs_info, &fs_info->bdi);
1623 if (ret) {
1624 err = ret;
1625 goto fail_srcu;
1626 }
1627
1628 fs_info->btree_inode = new_inode(sb);
1629 if (!fs_info->btree_inode) {
1630 err = -ENOMEM;
1631 goto fail_bdi;
1632 }
1633
a6591715 1634 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
1561deda 1635
76dda93c 1636 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
8fd17795 1637 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 1638 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 1639 INIT_LIST_HEAD(&fs_info->delayed_iputs);
19c00ddc 1640 INIT_LIST_HEAD(&fs_info->hashers);
ea8c2819 1641 INIT_LIST_HEAD(&fs_info->delalloc_inodes);
5a3f23d5 1642 INIT_LIST_HEAD(&fs_info->ordered_operations);
11833d66 1643 INIT_LIST_HEAD(&fs_info->caching_block_groups);
1832a6d5 1644 spin_lock_init(&fs_info->delalloc_lock);
a4abeea4 1645 spin_lock_init(&fs_info->trans_lock);
31153d81 1646 spin_lock_init(&fs_info->ref_cache_lock);
76dda93c 1647 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 1648 spin_lock_init(&fs_info->delayed_iput_lock);
4cb5300b 1649 spin_lock_init(&fs_info->defrag_inodes_lock);
2bf64758 1650 spin_lock_init(&fs_info->free_chunk_lock);
7585717f 1651 mutex_init(&fs_info->reloc_mutex);
19c00ddc 1652
58176a96 1653 init_completion(&fs_info->kobj_unregister);
9f5fae2f
CM		 1654 fs_info->tree_root = tree_root;
1655 fs_info->extent_root = extent_root;
d20f7043 1656 fs_info->csum_root = csum_root;
0b86a832
CM		 1657 fs_info->chunk_root = chunk_root;
1658 fs_info->dev_root = dev_root;
8a4b83cc 1659 fs_info->fs_devices = fs_devices;
0b86a832 1660 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 1661 INIT_LIST_HEAD(&fs_info->space_info);
0b86a832 1662 btrfs_mapping_init(&fs_info->mapping_tree);
f0486c68
YZ		 1663 btrfs_init_block_rsv(&fs_info->global_block_rsv);
1664 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv);
1665 btrfs_init_block_rsv(&fs_info->trans_block_rsv);
1666 btrfs_init_block_rsv(&fs_info->chunk_block_rsv);
1667 btrfs_init_block_rsv(&fs_info->empty_block_rsv);
cb03c743 1668 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 1669 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 1670 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 1671 atomic_set(&fs_info->nr_async_bios, 0);
4cb5300b 1672 atomic_set(&fs_info->defrag_running, 0);
e20d96d6 1673 fs_info->sb = sb;
6f568d35 1674 fs_info->max_inline = 8192 * 1024;
9ed74f2d 1675 fs_info->metadata_ratio = 0;
4cb5300b 1676 fs_info->defrag_inodes = RB_ROOT;
a4abeea4 1677 fs_info->trans_no_join = 0;
2bf64758 1678 fs_info->free_chunk_space = 0;
c8b97818 1679
b34b086c
CM		 1680 fs_info->thread_pool_size = min_t(unsigned long,
1681 num_online_cpus() + 2, 8);
0afbaf8c 1682
3eaa2885
CM		 1683 INIT_LIST_HEAD(&fs_info->ordered_extents);
1684 spin_lock_init(&fs_info->ordered_extent_lock);
16cdcec7
MX		 1685 fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
1686 GFP_NOFS);
1687 if (!fs_info->delayed_root) {
1688 err = -ENOMEM;
1689 goto fail_iput;
1690 }
1691 btrfs_init_delayed_root(fs_info->delayed_root);
3eaa2885 1692
a2de733c
AJ		 1693 mutex_init(&fs_info->scrub_lock);
1694 atomic_set(&fs_info->scrubs_running, 0);
1695 atomic_set(&fs_info->scrub_pause_req, 0);
1696 atomic_set(&fs_info->scrubs_paused, 0);
1697 atomic_set(&fs_info->scrub_cancel_req, 0);
1698 init_waitqueue_head(&fs_info->scrub_pause_wait);
1699 init_rwsem(&fs_info->scrub_super_lock);
1700 fs_info->scrub_workers_refcnt = 0;
a2de733c 1701
a061fc8d
CM		 1702 sb->s_blocksize = 4096;
1703 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 1704 sb->s_bdi = &fs_info->bdi;
a061fc8d 1705
76dda93c
YZ		 1706 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
1707 fs_info->btree_inode->i_nlink = 1;
0afbaf8c
CM		 1708 /*
1709 * we set the i_size on the btree inode to the max possible int.
1710 * the real end of the address space is determined by all of
1711 * the devices in the system
1712 */
1713 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 1714 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 1715 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
1716
5d4f98a2 1717 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 1718 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
f993c883 1719 fs_info->btree_inode->i_mapping);
a8067e02 1720 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
d1310b2e
CM		 1721
1722 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 1723
76dda93c
YZ		 1724 BTRFS_I(fs_info->btree_inode)->root = tree_root;
1725 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
1726 sizeof(struct btrfs_key));
1727 BTRFS_I(fs_info->btree_inode)->dummy_inode = 1;
c65ddb52 1728 insert_inode_hash(fs_info->btree_inode);
76dda93c 1729
0f9dd46c 1730 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 1731 fs_info->block_group_cache_tree = RB_ROOT;
0f9dd46c 1732
11833d66 1733 extent_io_tree_init(&fs_info->freed_extents[0],
f993c883 1734 fs_info->btree_inode->i_mapping);
11833d66 1735 extent_io_tree_init(&fs_info->freed_extents[1],
f993c883 1736 fs_info->btree_inode->i_mapping);
11833d66 1737 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 1738 fs_info->do_barriers = 1;
e18e4809 1739
39279cc3 1740
5a3f23d5 1741 mutex_init(&fs_info->ordered_operations_mutex);
e02119d5 1742 mutex_init(&fs_info->tree_log_mutex);
925baedd 1743 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 1744 mutex_init(&fs_info->transaction_kthread_mutex);
1745 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 1746 mutex_init(&fs_info->volume_mutex);
276e680d 1747 init_rwsem(&fs_info->extent_commit_sem);
c71bf099 1748 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 1749 init_rwsem(&fs_info->subvol_sem);
fa9c0d79
CM		 1750
1751 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
1752 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
1753
e6dcd2dc 1754 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 1755 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 1756 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 1757 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 1758
0b86a832 1759 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 1760 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 1761
a512bbf8 1762 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
20b45077
DY		 1763 if (!bh) {
1764 err = -EINVAL;
16cdcec7 1765 goto fail_alloc;
20b45077 1766 }
39279cc3 1767
6c41761f
DS		 1768 memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
1769 memcpy(fs_info->super_for_commit, fs_info->super_copy,
1770 sizeof(*fs_info->super_for_commit));
a061fc8d 1771 brelse(bh);
5f39d397 1772
6c41761f 1773 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
0b86a832 1774
6c41761f 1775 disk_super = fs_info->super_copy;
0f7d52f4 1776 if (!btrfs_super_root(disk_super))
16cdcec7 1777 goto fail_alloc;
0f7d52f4 1778
acce952b 1779 /* check FS state, whether FS is broken. */
1780 fs_info->fs_state |= btrfs_super_flags(disk_super);
1781
1782 btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
1783
75e7cb7f
LB		 1784 /*
1785 * In the long term, we'll store the compression type in the super
1786 * block, and it'll be used for per file compression control.
1787 */
1788 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
1789
2b82032c
YZ		 1790 ret = btrfs_parse_options(tree_root, options);
1791 if (ret) {
1792 err = ret;
16cdcec7 1793 goto fail_alloc;
2b82032c 1794 }
dfe25020 1795
f2b636e8
JB		 1796 features = btrfs_super_incompat_flags(disk_super) &
1797 ~BTRFS_FEATURE_INCOMPAT_SUPP;
1798 if (features) {
1799 printk(KERN_ERR "BTRFS: couldn't mount because of "
1800 "unsupported optional features (%Lx).\n",
21380931 1801 (unsigned long long)features);
f2b636e8 1802 err = -EINVAL;
16cdcec7 1803 goto fail_alloc;
f2b636e8
JB		 1804 }
1805
5d4f98a2 1806 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae
LZ		 1807 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
1808 if (tree_root->fs_info->compress_type & BTRFS_COMPRESS_LZO)
1809 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1810 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 1811
f2b636e8
JB		 1812 features = btrfs_super_compat_ro_flags(disk_super) &
1813 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
1814 if (!(sb->s_flags & MS_RDONLY) && features) {
1815 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
1816 "unsupported option features (%Lx).\n",
21380931 1817 (unsigned long long)features);
f2b636e8 1818 err = -EINVAL;
16cdcec7 1819 goto fail_alloc;
f2b636e8 1820 }
61d92c32
CM		 1821
1822 btrfs_init_workers(&fs_info->generic_worker,
1823 "genwork", 1, NULL);
1824
5443be45 1825 btrfs_init_workers(&fs_info->workers, "worker",
61d92c32
CM		 1826 fs_info->thread_pool_size,
1827 &fs_info->generic_worker);
c8b97818 1828
771ed689 1829 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
61d92c32
CM		 1830 fs_info->thread_pool_size,
1831 &fs_info->generic_worker);
771ed689 1832
5443be45 1833 btrfs_init_workers(&fs_info->submit_workers, "submit",
b720d209 1834 min_t(u64, fs_devices->num_devices,
61d92c32
CM		 1835 fs_info->thread_pool_size),
1836 &fs_info->generic_worker);
61b49440 1837
bab39bf9
JB		 1838 btrfs_init_workers(&fs_info->caching_workers, "cache",
1839 2, &fs_info->generic_worker);
1840
61b49440
CM		 1841 /* a higher idle thresh on the submit workers makes it much more
1842 * likely that bios will be send down in a sane order to the
1843 * devices
1844 */
1845 fs_info->submit_workers.idle_thresh = 64;
53863232 1846
771ed689 1847 fs_info->workers.idle_thresh = 16;
4a69a410 1848 fs_info->workers.ordered = 1;
61b49440 1849
771ed689
CM		 1850 fs_info->delalloc_workers.idle_thresh = 2;
1851 fs_info->delalloc_workers.ordered = 1;
1852
61d92c32
CM		 1853 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
1854 &fs_info->generic_worker);
5443be45 1855 btrfs_init_workers(&fs_info->endio_workers, "endio",
61d92c32
CM		 1856 fs_info->thread_pool_size,
1857 &fs_info->generic_worker);
d20f7043 1858 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
61d92c32
CM		 1859 fs_info->thread_pool_size,
1860 &fs_info->generic_worker);
cad321ad 1861 btrfs_init_workers(&fs_info->endio_meta_write_workers,
61d92c32
CM		 1862 "endio-meta-write", fs_info->thread_pool_size,
1863 &fs_info->generic_worker);
5443be45 1864 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
61d92c32
CM		 1865 fs_info->thread_pool_size,
1866 &fs_info->generic_worker);
0cb59c99
JB		 1867 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
1868 1, &fs_info->generic_worker);
16cdcec7
MX		 1869 btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
1870 fs_info->thread_pool_size,
1871 &fs_info->generic_worker);
61b49440
CM		 1872
1873 /*
1874 * endios are largely parallel and should have a very
1875 * low idle thresh
1876 */
1877 fs_info->endio_workers.idle_thresh = 4;
b51912c9
CM		 1878 fs_info->endio_meta_workers.idle_thresh = 4;
1879
9042846b
CM		 1880 fs_info->endio_write_workers.idle_thresh = 2;
1881 fs_info->endio_meta_write_workers.idle_thresh = 2;
1882
4543df7e 1883 btrfs_start_workers(&fs_info->workers, 1);
61d92c32 1884 btrfs_start_workers(&fs_info->generic_worker, 1);
1cc127b5 1885 btrfs_start_workers(&fs_info->submit_workers, 1);
771ed689 1886 btrfs_start_workers(&fs_info->delalloc_workers, 1);
247e743c 1887 btrfs_start_workers(&fs_info->fixup_workers, 1);
9042846b
CM		 1888 btrfs_start_workers(&fs_info->endio_workers, 1);
1889 btrfs_start_workers(&fs_info->endio_meta_workers, 1);
1890 btrfs_start_workers(&fs_info->endio_meta_write_workers, 1);
1891 btrfs_start_workers(&fs_info->endio_write_workers, 1);
0cb59c99 1892 btrfs_start_workers(&fs_info->endio_freespace_worker, 1);
16cdcec7 1893 btrfs_start_workers(&fs_info->delayed_workers, 1);
bab39bf9 1894 btrfs_start_workers(&fs_info->caching_workers, 1);
4543df7e 1895
4575c9cc 1896 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 1897 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
1898 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 1899
db94535d
CM		 1900 nodesize = btrfs_super_nodesize(disk_super);
1901 leafsize = btrfs_super_leafsize(disk_super);
1902 sectorsize = btrfs_super_sectorsize(disk_super);
87ee04eb 1903 stripesize = btrfs_super_stripesize(disk_super);
db94535d
CM		 1904 tree_root->nodesize = nodesize;
1905 tree_root->leafsize = leafsize;
1906 tree_root->sectorsize = sectorsize;
87ee04eb 1907 tree_root->stripesize = stripesize;
a061fc8d
CM		 1908
1909 sb->s_blocksize = sectorsize;
1910 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 1911
39279cc3
CM		 1912 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
1913 sizeof(disk_super->magic))) {
d397712b 1914 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 1915 goto fail_sb_buffer;
1916 }
19c00ddc 1917
925baedd 1918 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 1919 ret = btrfs_read_sys_array(tree_root);
925baedd 1920 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 1921 if (ret) {
d397712b
CM		 1922 printk(KERN_WARNING "btrfs: failed to read the system "
1923 "array on %s\n", sb->s_id);
5d4f98a2 1924 goto fail_sb_buffer;
84eed90f 1925 }
0b86a832
CM		 1926
1927 blocksize = btrfs_level_size(tree_root,
1928 btrfs_super_chunk_root_level(disk_super));
84234f3a 1929 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 1930
1931 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1932 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1933
1934 chunk_root->node = read_tree_block(chunk_root,
1935 btrfs_super_chunk_root(disk_super),
84234f3a 1936 blocksize, generation);
0b86a832 1937 BUG_ON(!chunk_root->node);
83121942
DW		 1938 if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
1939 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
1940 sb->s_id);
1941 goto fail_chunk_root;
1942 }
5d4f98a2
YZ		 1943 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
1944 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 1945
e17cade2 1946 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
d397712b
CM		 1947 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
1948 BTRFS_UUID_SIZE);
e17cade2 1949
925baedd 1950 mutex_lock(&fs_info->chunk_mutex);
0b86a832 1951 ret = btrfs_read_chunk_tree(chunk_root);
925baedd 1952 mutex_unlock(&fs_info->chunk_mutex);
2b82032c 1953 if (ret) {
d397712b
CM		 1954 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
1955 sb->s_id);
2b82032c
YZ		 1956 goto fail_chunk_root;
1957 }
0b86a832 1958
dfe25020
CM		 1959 btrfs_close_extra_devices(fs_devices);
1960
db94535d
CM		 1961 blocksize = btrfs_level_size(tree_root,
1962 btrfs_super_root_level(disk_super));
84234f3a 1963 generation = btrfs_super_generation(disk_super);
0b86a832 1964
e20d96d6 1965 tree_root->node = read_tree_block(tree_root,
db94535d 1966 btrfs_super_root(disk_super),
84234f3a 1967 blocksize, generation);
39279cc3 1968 if (!tree_root->node)
2b82032c 1969 goto fail_chunk_root;
83121942
DW		 1970 if (!test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
1971 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
1972 sb->s_id);
1973 goto fail_tree_root;
1974 }
5d4f98a2
YZ		 1975 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
1976 tree_root->commit_root = btrfs_root_node(tree_root);
db94535d
CM		 1977
1978 ret = find_and_setup_root(tree_root, fs_info,
e20d96d6 1979 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
0b86a832 1980 if (ret)
39279cc3 1981 goto fail_tree_root;
0b86a832
CM		 1982 extent_root->track_dirty = 1;
1983
1984 ret = find_and_setup_root(tree_root, fs_info,
1985 BTRFS_DEV_TREE_OBJECTID, dev_root);
0b86a832
CM		 1986 if (ret)
1987 goto fail_extent_root;
5d4f98a2 1988 dev_root->track_dirty = 1;
3768f368 1989
d20f7043
CM		 1990 ret = find_and_setup_root(tree_root, fs_info,
1991 BTRFS_CSUM_TREE_OBJECTID, csum_root);
1992 if (ret)
5d4f98a2 1993 goto fail_dev_root;
d20f7043
CM		 1994
1995 csum_root->track_dirty = 1;
1996
8929ecfa
YZ		 1997 fs_info->generation = generation;
1998 fs_info->last_trans_committed = generation;
1999 fs_info->data_alloc_profile = (u64)-1;
2000 fs_info->metadata_alloc_profile = (u64)-1;
2001 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
2002
c59021f8 2003 ret = btrfs_init_space_info(fs_info);
2004 if (ret) {
2005 printk(KERN_ERR "Failed to initial space info: %d\n", ret);
2006 goto fail_block_groups;
2007 }
2008
1b1d1f66
JB		 2009 ret = btrfs_read_block_groups(extent_root);
2010 if (ret) {
2011 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
2012 goto fail_block_groups;
2013 }
9078a3e1 2014
a74a4b97
CM		 2015 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2016 "btrfs-cleaner");
57506d50 2017 if (IS_ERR(fs_info->cleaner_kthread))
1b1d1f66 2018 goto fail_block_groups;
a74a4b97
CM		 2019
2020 fs_info->transaction_kthread = kthread_run(transaction_kthread,
2021 tree_root,
2022 "btrfs-transaction");
57506d50 2023 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 2024 goto fail_cleaner;
a74a4b97 2025
c289811c
CM		 2026 if (!btrfs_test_opt(tree_root, SSD) &&
2027 !btrfs_test_opt(tree_root, NOSSD) &&
2028 !fs_info->fs_devices->rotating) {
2029 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
2030 "mode\n");
2031 btrfs_set_opt(fs_info->mount_opt, SSD);
2032 }
2033
acce952b 2034 /* do not make disk changes in broken FS */
2035 if (btrfs_super_log_root(disk_super) != 0 &&
2036 !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) {
e02119d5
CM		 2037 u64 bytenr = btrfs_super_log_root(disk_super);
2038
7c2ca468 2039 if (fs_devices->rw_devices == 0) {
d397712b
CM		 2040 printk(KERN_WARNING "Btrfs log replay required "
2041 "on RO media\n");
7c2ca468
CM		 2042 err = -EIO;
2043 goto fail_trans_kthread;
2044 }
e02119d5
CM		 2045 blocksize =
2046 btrfs_level_size(tree_root,
2047 btrfs_super_log_root_level(disk_super));
d18a2c44 2048
676e4c86
DC		 2049 log_tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
2050 if (!log_tree_root) {
2051 err = -ENOMEM;
2052 goto fail_trans_kthread;
2053 }
e02119d5
CM		 2054
2055 __setup_root(nodesize, leafsize, sectorsize, stripesize,
2056 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2057
2058 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 2059 blocksize,
2060 generation + 1);
e02119d5
CM		 2061 ret = btrfs_recover_log_trees(log_tree_root);
2062 BUG_ON(ret);
e556ce2c
YZ		 2063
2064 if (sb->s_flags & MS_RDONLY) {
2065 ret = btrfs_commit_super(tree_root);
2066 BUG_ON(ret);
2067 }
e02119d5 2068 }
1a40e23b 2069
76dda93c
YZ		 2070 ret = btrfs_find_orphan_roots(tree_root);
2071 BUG_ON(ret);
2072
7c2ca468 2073 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b
YZ		 2074 ret = btrfs_cleanup_fs_roots(fs_info);
2075 BUG_ON(ret);
2076
5d4f98a2 2077 ret = btrfs_recover_relocation(tree_root);
d7ce5843
MX		 2078 if (ret < 0) {
2079 printk(KERN_WARNING
2080 "btrfs: failed to recover relocation\n");
2081 err = -EINVAL;
2082 goto fail_trans_kthread;
2083 }
7c2ca468 2084 }
1a40e23b 2085
3de4586c
CM		 2086 location.objectid = BTRFS_FS_TREE_OBJECTID;
2087 location.type = BTRFS_ROOT_ITEM_KEY;
2088 location.offset = (u64)-1;
2089
3de4586c
CM		 2090 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
2091 if (!fs_info->fs_root)
7c2ca468 2092 goto fail_trans_kthread;
3140c9a3
DC		 2093 if (IS_ERR(fs_info->fs_root)) {
2094 err = PTR_ERR(fs_info->fs_root);
2095 goto fail_trans_kthread;
2096 }
c289811c 2097
e3acc2a6
JB		 2098 if (!(sb->s_flags & MS_RDONLY)) {
2099 down_read(&fs_info->cleanup_work_sem);
66b4ffd1
JB		 2100 err = btrfs_orphan_cleanup(fs_info->fs_root);
2101 if (!err)
2102 err = btrfs_orphan_cleanup(fs_info->tree_root);
e3acc2a6 2103 up_read(&fs_info->cleanup_work_sem);
66b4ffd1
JB		 2104 if (err) {
2105 close_ctree(tree_root);
2106 return ERR_PTR(err);
2107 }
e3acc2a6
JB		 2108 }
2109
0f7d52f4 2110 return tree_root;
39279cc3 2111
7c2ca468
CM		 2112fail_trans_kthread:
2113 kthread_stop(fs_info->transaction_kthread);
3f157a2f 2114fail_cleaner:
a74a4b97 2115 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2116
2117 /*
2118 * make sure we're done with the btree inode before we stop our
2119 * kthreads
2120 */
2121 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
2122 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
2123
1b1d1f66
JB		 2124fail_block_groups:
2125 btrfs_free_block_groups(fs_info);
d20f7043 2126 free_extent_buffer(csum_root->node);
5d4f98a2
YZ		 2127 free_extent_buffer(csum_root->commit_root);
2128fail_dev_root:
2129 free_extent_buffer(dev_root->node);
2130 free_extent_buffer(dev_root->commit_root);
0b86a832
CM		 2131fail_extent_root:
2132 free_extent_buffer(extent_root->node);
5d4f98a2 2133 free_extent_buffer(extent_root->commit_root);
39279cc3 2134fail_tree_root:
5f39d397 2135 free_extent_buffer(tree_root->node);
5d4f98a2 2136 free_extent_buffer(tree_root->commit_root);
2b82032c
YZ		 2137fail_chunk_root:
2138 free_extent_buffer(chunk_root->node);
5d4f98a2 2139 free_extent_buffer(chunk_root->commit_root);
39279cc3 2140fail_sb_buffer:
61d92c32 2141 btrfs_stop_workers(&fs_info->generic_worker);
247e743c 2142 btrfs_stop_workers(&fs_info->fixup_workers);
771ed689 2143 btrfs_stop_workers(&fs_info->delalloc_workers);
8b712842
CM		 2144 btrfs_stop_workers(&fs_info->workers);
2145 btrfs_stop_workers(&fs_info->endio_workers);
d20f7043 2146 btrfs_stop_workers(&fs_info->endio_meta_workers);
cad321ad 2147 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
e6dcd2dc 2148 btrfs_stop_workers(&fs_info->endio_write_workers);
0cb59c99 2149 btrfs_stop_workers(&fs_info->endio_freespace_worker);
1cc127b5 2150 btrfs_stop_workers(&fs_info->submit_workers);
16cdcec7 2151 btrfs_stop_workers(&fs_info->delayed_workers);
bab39bf9 2152 btrfs_stop_workers(&fs_info->caching_workers);
16cdcec7 2153fail_alloc:
4543df7e 2154fail_iput:
7c2ca468 2155 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
4543df7e 2156 iput(fs_info->btree_inode);
7e662854 2157
dfe25020 2158 btrfs_close_devices(fs_info->fs_devices);
84eed90f 2159 btrfs_mapping_tree_free(&fs_info->mapping_tree);
ad081f14 2160fail_bdi:
7e662854 2161 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 2162fail_srcu:
2163 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 2164fail:
6c41761f 2165 free_fs_info(fs_info);
39279cc3 2166 return ERR_PTR(err);
eb60ceac
CM		 2167}
2168
f2984462
CM		 2169static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2170{
2171 char b[BDEVNAME_SIZE];
2172
2173 if (uptodate) {
2174 set_buffer_uptodate(bh);
2175 } else {
7a36ddec 2176 printk_ratelimited(KERN_WARNING "lost page write due to "
f2984462
CM		 2177 "I/O error on %s\n",
2178 bdevname(bh->b_bdev, b));
1259ab75
CM		 2179 /* note, we dont' set_buffer_write_io_error because we have
2180 * our own ways of dealing with the IO errors
2181 */
f2984462
CM		 2182 clear_buffer_uptodate(bh);
2183 }
2184 unlock_buffer(bh);
2185 put_bh(bh);
2186}
2187
a512bbf8
YZ		 2188struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
2189{
2190 struct buffer_head *bh;
2191 struct buffer_head *latest = NULL;
2192 struct btrfs_super_block *super;
2193 int i;
2194 u64 transid = 0;
2195 u64 bytenr;
2196
2197 /* we would like to check all the supers, but that would make
2198 * a btrfs mount succeed after a mkfs from a different FS.
2199 * So, we need to add a special mount option to scan for
2200 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2201 */
2202 for (i = 0; i < 1; i++) {
2203 bytenr = btrfs_sb_offset(i);
2204 if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
2205 break;
2206 bh = __bread(bdev, bytenr / 4096, 4096);
2207 if (!bh)
2208 continue;
2209
2210 super = (struct btrfs_super_block *)bh->b_data;
2211 if (btrfs_super_bytenr(super) != bytenr ||
2212 strncmp((char *)(&super->magic), BTRFS_MAGIC,
2213 sizeof(super->magic))) {
2214 brelse(bh);
2215 continue;
2216 }
2217
2218 if (!latest || btrfs_super_generation(super) > transid) {
2219 brelse(latest);
2220 latest = bh;
2221 transid = btrfs_super_generation(super);
2222 } else {
2223 brelse(bh);
2224 }
2225 }
2226 return latest;
2227}
2228
4eedeb75
HH		 2229/*
2230 * this should be called twice, once with wait == 0 and
2231 * once with wait == 1. When wait == 0 is done, all the buffer heads
2232 * we write are pinned.
2233 *
2234 * They are released when wait == 1 is done.
2235 * max_mirrors must be the same for both runs, and it indicates how
2236 * many supers on this one device should be written.
2237 *
2238 * max_mirrors == 0 means to write them all.
2239 */
a512bbf8
YZ		 2240static int write_dev_supers(struct btrfs_device *device,
2241 struct btrfs_super_block *sb,
2242 int do_barriers, int wait, int max_mirrors)
2243{
2244 struct buffer_head *bh;
2245 int i;
2246 int ret;
2247 int errors = 0;
2248 u32 crc;
2249 u64 bytenr;
2250 int last_barrier = 0;
2251
2252 if (max_mirrors == 0)
2253 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
2254
2255 /* make sure only the last submit_bh does a barrier */
2256 if (do_barriers) {
2257 for (i = 0; i < max_mirrors; i++) {
2258 bytenr = btrfs_sb_offset(i);
2259 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
2260 device->total_bytes)
2261 break;
2262 last_barrier = i;
2263 }
2264 }
2265
2266 for (i = 0; i < max_mirrors; i++) {
2267 bytenr = btrfs_sb_offset(i);
2268 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
2269 break;
2270
2271 if (wait) {
2272 bh = __find_get_block(device->bdev, bytenr / 4096,
2273 BTRFS_SUPER_INFO_SIZE);
2274 BUG_ON(!bh);
a512bbf8 2275 wait_on_buffer(bh);
4eedeb75
HH		 2276 if (!buffer_uptodate(bh))
2277 errors++;
2278
2279 /* drop our reference */
2280 brelse(bh);
2281
2282 /* drop the reference from the wait == 0 run */
2283 brelse(bh);
2284 continue;
a512bbf8
YZ		 2285 } else {
2286 btrfs_set_super_bytenr(sb, bytenr);
2287
2288 crc = ~(u32)0;
2289 crc = btrfs_csum_data(NULL, (char *)sb +
2290 BTRFS_CSUM_SIZE, crc,
2291 BTRFS_SUPER_INFO_SIZE -
2292 BTRFS_CSUM_SIZE);
2293 btrfs_csum_final(crc, sb->csum);
2294
4eedeb75
HH		 2295 /*
2296 * one reference for us, and we leave it for the
2297 * caller
2298 */
a512bbf8
YZ		 2299 bh = __getblk(device->bdev, bytenr / 4096,
2300 BTRFS_SUPER_INFO_SIZE);
2301 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
2302
4eedeb75 2303 /* one reference for submit_bh */
a512bbf8 2304 get_bh(bh);
4eedeb75
HH		 2305
2306 set_buffer_uptodate(bh);
a512bbf8
YZ		 2307 lock_buffer(bh);
2308 bh->b_end_io = btrfs_end_buffer_write_sync;
2309 }
2310
c3b9a62c
CH		 2311 if (i == last_barrier && do_barriers)
2312 ret = submit_bh(WRITE_FLUSH_FUA, bh);
2313 else
ffbd517d 2314 ret = submit_bh(WRITE_SYNC, bh);
a512bbf8 2315
4eedeb75 2316 if (ret)
a512bbf8 2317 errors++;
a512bbf8
YZ		 2318 }
2319 return errors < i ? 0 : -1;
2320}
2321
2322int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 2323{
e5e9a520 2324 struct list_head *head;
f2984462 2325 struct btrfs_device *dev;
a061fc8d 2326 struct btrfs_super_block *sb;
f2984462 2327 struct btrfs_dev_item *dev_item;
f2984462
CM		 2328 int ret;
2329 int do_barriers;
a236aed1
CM		 2330 int max_errors;
2331 int total_errors = 0;
a061fc8d 2332 u64 flags;
f2984462 2333
6c41761f 2334 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
f2984462
CM		 2335 do_barriers = !btrfs_test_opt(root, NOBARRIER);
2336
6c41761f 2337 sb = root->fs_info->super_for_commit;
a061fc8d 2338 dev_item = &sb->dev_item;
e5e9a520 2339
174ba509 2340 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
e5e9a520 2341 head = &root->fs_info->fs_devices->devices;
1f78160c 2342 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 2343 if (!dev->bdev) {
2344 total_errors++;
2345 continue;
2346 }
2b82032c 2347 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 2348 continue;
2349
2b82032c 2350 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 2351 btrfs_set_stack_device_type(dev_item, dev->type);
2352 btrfs_set_stack_device_id(dev_item, dev->devid);
2353 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
2354 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
2355 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
2356 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
2357 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
2358 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 2359 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 2360
a061fc8d
CM		 2361 flags = btrfs_super_flags(sb);
2362 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
2363
a512bbf8 2364 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 2365 if (ret)
2366 total_errors++;
f2984462 2367 }
a236aed1 2368 if (total_errors > max_errors) {
d397712b
CM		 2369 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2370 total_errors);
a236aed1
CM		 2371 BUG();
2372 }
f2984462 2373
a512bbf8 2374 total_errors = 0;
1f78160c 2375 list_for_each_entry_rcu(dev, head, dev_list) {
dfe25020
CM		 2376 if (!dev->bdev)
2377 continue;
2b82032c 2378 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 2379 continue;
2380
a512bbf8
YZ		 2381 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
2382 if (ret)
2383 total_errors++;
f2984462 2384 }
174ba509 2385 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 2386 if (total_errors > max_errors) {
d397712b
CM		 2387 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2388 total_errors);
a236aed1
CM		 2389 BUG();
2390 }
f2984462
CM		 2391 return 0;
2392}
2393
a512bbf8
YZ		 2394int write_ctree_super(struct btrfs_trans_handle *trans,
2395 struct btrfs_root *root, int max_mirrors)
eb60ceac 2396{
e66f709b 2397 int ret;
5f39d397 2398
a512bbf8 2399 ret = write_all_supers(root, max_mirrors);
5f39d397 2400 return ret;
cfaa7295
CM		 2401}
2402
5eda7b5e 2403int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
2619ba1f 2404{
4df27c4d 2405 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 2406 radix_tree_delete(&fs_info->fs_roots_radix,
2407 (unsigned long)root->root_key.objectid);
4df27c4d 2408 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 2409
2410 if (btrfs_root_refs(&root->root_item) == 0)
2411 synchronize_srcu(&fs_info->subvol_srcu);
2412
581bb050
LZ		 2413 __btrfs_remove_free_space_cache(root->free_ino_pinned);
2414 __btrfs_remove_free_space_cache(root->free_ino_ctl);
4df27c4d
YZ		 2415 free_fs_root(root);
2416 return 0;
2417}
2418
2419static void free_fs_root(struct btrfs_root *root)
2420{
82d5902d 2421 iput(root->cache_inode);
4df27c4d 2422 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
0ee5dc67
AV		 2423 if (root->anon_dev)
2424 free_anon_bdev(root->anon_dev);
4df27c4d
YZ		 2425 free_extent_buffer(root->node);
2426 free_extent_buffer(root->commit_root);
581bb050
LZ		 2427 kfree(root->free_ino_ctl);
2428 kfree(root->free_ino_pinned);
d397712b 2429 kfree(root->name);
2619ba1f 2430 kfree(root);
2619ba1f
CM		 2431}
2432
35b7e476 2433static int del_fs_roots(struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 2434{
2435 int ret;
2436 struct btrfs_root *gang[8];
2437 int i;
2438
76dda93c
YZ		 2439 while (!list_empty(&fs_info->dead_roots)) {
2440 gang[0] = list_entry(fs_info->dead_roots.next,
2441 struct btrfs_root, root_list);
2442 list_del(&gang[0]->root_list);
2443
2444 if (gang[0]->in_radix) {
2445 btrfs_free_fs_root(fs_info, gang[0]);
2446 } else {
2447 free_extent_buffer(gang[0]->node);
2448 free_extent_buffer(gang[0]->commit_root);
2449 kfree(gang[0]);
2450 }
2451 }
2452
d397712b 2453 while (1) {
0f7d52f4
CM		 2454 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2455 (void **)gang, 0,
2456 ARRAY_SIZE(gang));
2457 if (!ret)
2458 break;
2619ba1f 2459 for (i = 0; i < ret; i++)
5eda7b5e 2460 btrfs_free_fs_root(fs_info, gang[i]);
0f7d52f4
CM		 2461 }
2462 return 0;
2463}
b4100d64 2464
c146afad 2465int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 2466{
c146afad
YZ		 2467 u64 root_objectid = 0;
2468 struct btrfs_root *gang[8];
2469 int i;
3768f368 2470 int ret;
e089f05c 2471
c146afad
YZ		 2472 while (1) {
2473 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2474 (void **)gang, root_objectid,
2475 ARRAY_SIZE(gang));
2476 if (!ret)
2477 break;
5d4f98a2
YZ		 2478
2479 root_objectid = gang[ret - 1]->root_key.objectid + 1;
c146afad 2480 for (i = 0; i < ret; i++) {
66b4ffd1
JB		 2481 int err;
2482
c146afad 2483 root_objectid = gang[i]->root_key.objectid;
66b4ffd1
JB		 2484 err = btrfs_orphan_cleanup(gang[i]);
2485 if (err)
2486 return err;
c146afad
YZ		 2487 }
2488 root_objectid++;
2489 }
2490 return 0;
2491}
a2135011 2492
c146afad
YZ		 2493int btrfs_commit_super(struct btrfs_root *root)
2494{
2495 struct btrfs_trans_handle *trans;
2496 int ret;
a74a4b97 2497
c146afad 2498 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 2499 btrfs_run_delayed_iputs(root);
a74a4b97 2500 btrfs_clean_old_snapshots(root);
c146afad 2501 mutex_unlock(&root->fs_info->cleaner_mutex);
c71bf099
YZ		 2502
2503 /* wait until ongoing cleanup work done */
2504 down_write(&root->fs_info->cleanup_work_sem);
2505 up_write(&root->fs_info->cleanup_work_sem);
2506
7a7eaa40 2507 trans = btrfs_join_transaction(root);
3612b495
TI		 2508 if (IS_ERR(trans))
2509 return PTR_ERR(trans);
54aa1f4d 2510 ret = btrfs_commit_transaction(trans, root);
c146afad
YZ		 2511 BUG_ON(ret);
2512 /* run commit again to drop the original snapshot */
7a7eaa40 2513 trans = btrfs_join_transaction(root);
3612b495
TI		 2514 if (IS_ERR(trans))
2515 return PTR_ERR(trans);
79154b1b
CM		 2516 btrfs_commit_transaction(trans, root);
2517 ret = btrfs_write_and_wait_transaction(NULL, root);
3768f368 2518 BUG_ON(ret);
d6bfde87 2519
a512bbf8 2520 ret = write_ctree_super(NULL, root, 0);
c146afad
YZ		 2521 return ret;
2522}
2523
2524int close_ctree(struct btrfs_root *root)
2525{
2526 struct btrfs_fs_info *fs_info = root->fs_info;
2527 int ret;
2528
2529 fs_info->closing = 1;
2530 smp_mb();
2531
a2de733c 2532 btrfs_scrub_cancel(root);
4cb5300b
CM		 2533
2534 /* wait for any defraggers to finish */
2535 wait_event(fs_info->transaction_wait,
2536 (atomic_read(&fs_info->defrag_running) == 0));
2537
2538 /* clear out the rbtree of defraggable inodes */
2539 btrfs_run_defrag_inodes(root->fs_info);
2540
acce952b 2541 /*
2542 * Here come 2 situations when btrfs is broken to flip readonly:
2543 *
2544 * 1. when btrfs flips readonly somewhere else before
2545 * btrfs_commit_super, sb->s_flags has MS_RDONLY flag,
2546 * and btrfs will skip to write sb directly to keep
2547 * ERROR state on disk.
2548 *
2549 * 2. when btrfs flips readonly just in btrfs_commit_super,
ae0e47f0 2550 * and in such case, btrfs cannot write sb via btrfs_commit_super,
acce952b 2551 * and since fs_state has been set BTRFS_SUPER_FLAG_ERROR flag,
2552 * btrfs will cleanup all FS resources first and write sb then.
2553 */
c146afad 2554 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
acce952b 2555 ret = btrfs_commit_super(root);
2556 if (ret)
2557 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
2558 }
2559
2560 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
2561 ret = btrfs_error_commit_super(root);
d397712b
CM		 2562 if (ret)
2563 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
c146afad 2564 }
0f7d52f4 2565
300e4f8a
JB		 2566 btrfs_put_block_group_cache(fs_info);
2567
8929ecfa
YZ		 2568 kthread_stop(root->fs_info->transaction_kthread);
2569 kthread_stop(root->fs_info->cleaner_kthread);
2570
f25784b3
YZ		 2571 fs_info->closing = 2;
2572 smp_mb();
2573
b0c68f8b 2574 if (fs_info->delalloc_bytes) {
d397712b 2575 printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
21380931 2576 (unsigned long long)fs_info->delalloc_bytes);
b0c68f8b 2577 }
31153d81 2578 if (fs_info->total_ref_cache_size) {
d397712b
CM		 2579 printk(KERN_INFO "btrfs: at umount reference cache size %llu\n",
2580 (unsigned long long)fs_info->total_ref_cache_size);
31153d81 2581 }
bcc63abb 2582
5d4f98a2
YZ		 2583 free_extent_buffer(fs_info->extent_root->node);
2584 free_extent_buffer(fs_info->extent_root->commit_root);
2585 free_extent_buffer(fs_info->tree_root->node);
2586 free_extent_buffer(fs_info->tree_root->commit_root);
2587 free_extent_buffer(root->fs_info->chunk_root->node);
2588 free_extent_buffer(root->fs_info->chunk_root->commit_root);
2589 free_extent_buffer(root->fs_info->dev_root->node);
2590 free_extent_buffer(root->fs_info->dev_root->commit_root);
2591 free_extent_buffer(root->fs_info->csum_root->node);
2592 free_extent_buffer(root->fs_info->csum_root->commit_root);
d20f7043 2593
9078a3e1 2594 btrfs_free_block_groups(root->fs_info);
d10c5f31 2595
c146afad 2596 del_fs_roots(fs_info);
d10c5f31 2597
c146afad 2598 iput(fs_info->btree_inode);
9ad6b7bc 2599
61d92c32 2600 btrfs_stop_workers(&fs_info->generic_worker);
247e743c 2601 btrfs_stop_workers(&fs_info->fixup_workers);
771ed689 2602 btrfs_stop_workers(&fs_info->delalloc_workers);
8b712842
CM		 2603 btrfs_stop_workers(&fs_info->workers);
2604 btrfs_stop_workers(&fs_info->endio_workers);
d20f7043 2605 btrfs_stop_workers(&fs_info->endio_meta_workers);
cad321ad 2606 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
e6dcd2dc 2607 btrfs_stop_workers(&fs_info->endio_write_workers);
0cb59c99 2608 btrfs_stop_workers(&fs_info->endio_freespace_worker);
1cc127b5 2609 btrfs_stop_workers(&fs_info->submit_workers);
16cdcec7 2610 btrfs_stop_workers(&fs_info->delayed_workers);
bab39bf9 2611 btrfs_stop_workers(&fs_info->caching_workers);
d6bfde87 2612
dfe25020 2613 btrfs_close_devices(fs_info->fs_devices);
0b86a832 2614 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 2615
04160088 2616 bdi_destroy(&fs_info->bdi);
76dda93c 2617 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 2618
6c41761f 2619 free_fs_info(fs_info);
83a4d548 2620
eb60ceac
CM		 2621 return 0;
2622}
2623
1259ab75 2624int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
5f39d397 2625{
1259ab75 2626 int ret;
810191ff 2627 struct inode *btree_inode = buf->first_page->mapping->host;
1259ab75 2628
2ac55d41
JB		 2629 ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf,
2630 NULL);
1259ab75
CM		 2631 if (!ret)
2632 return ret;
2633
2634 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
2635 parent_transid);
2636 return !ret;
5f39d397
CM		 2637}
2638
2639int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 2640{
810191ff 2641 struct inode *btree_inode = buf->first_page->mapping->host;
d1310b2e 2642 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
5f39d397
CM		 2643 buf);
2644}
6702ed49 2645
5f39d397
CM		 2646void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
2647{
810191ff 2648 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
5f39d397
CM		 2649 u64 transid = btrfs_header_generation(buf);
2650 struct inode *btree_inode = root->fs_info->btree_inode;
b9473439 2651 int was_dirty;
b4ce94de 2652
b9447ef8 2653 btrfs_assert_tree_locked(buf);
ccd467d6 2654 if (transid != root->fs_info->generation) {
d397712b
CM		 2655 printk(KERN_CRIT "btrfs transid mismatch buffer %llu, "
2656 "found %llu running %llu\n",
db94535d 2657 (unsigned long long)buf->start,
d397712b
CM		 2658 (unsigned long long)transid,
2659 (unsigned long long)root->fs_info->generation);
ccd467d6
CM		 2660 WARN_ON(1);
2661 }
b9473439
CM		 2662 was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
2663 buf);
2664 if (!was_dirty) {
2665 spin_lock(&root->fs_info->delalloc_lock);
2666 root->fs_info->dirty_metadata_bytes += buf->len;
2667 spin_unlock(&root->fs_info->delalloc_lock);
2668 }
eb60ceac
CM		 2669}
2670
d3c2fdcf 2671void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
16cdcec7
MX		 2672{
2673 /*
2674 * looks as though older kernels can get into trouble with
2675 * this code, they end up stuck in balance_dirty_pages forever
2676 */
2677 u64 num_dirty;
2678 unsigned long thresh = 32 * 1024 * 1024;
2679
2680 if (current->flags & PF_MEMALLOC)
2681 return;
2682
2683 btrfs_balance_delayed_items(root);
2684
2685 num_dirty = root->fs_info->dirty_metadata_bytes;
2686
2687 if (num_dirty > thresh) {
2688 balance_dirty_pages_ratelimited_nr(
2689 root->fs_info->btree_inode->i_mapping, 1);
2690 }
2691 return;
2692}
2693
2694void __btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
35b7e476 2695{
188de649
CM		 2696 /*
2697 * looks as though older kernels can get into trouble with
2698 * this code, they end up stuck in balance_dirty_pages forever
2699 */
d6bfde87 2700 u64 num_dirty;
771ed689 2701 unsigned long thresh = 32 * 1024 * 1024;
d6bfde87 2702
6933c02e 2703 if (current->flags & PF_MEMALLOC)
d6bfde87
CM		 2704 return;
2705
585ad2c3
CM		 2706 num_dirty = root->fs_info->dirty_metadata_bytes;
2707
d6bfde87
CM		 2708 if (num_dirty > thresh) {
2709 balance_dirty_pages_ratelimited_nr(
d7fc640e 2710 root->fs_info->btree_inode->i_mapping, 1);
d6bfde87 2711 }
188de649 2712 return;
35b7e476 2713}
6b80053d 2714
ca7a79ad 2715int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 2716{
810191ff 2717 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
ce9adaa5 2718 int ret;
ca7a79ad 2719 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
d397712b 2720 if (ret == 0)
b4ce94de 2721 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
ce9adaa5 2722 return ret;
6b80053d 2723}
0da5468f 2724
01d658f2
CM		 2725static int btree_lock_page_hook(struct page *page, void *data,
2726 void (*flush_fn)(void *))
4bef0848
CM		 2727{
2728 struct inode *inode = page->mapping->host;
b9473439 2729 struct btrfs_root *root = BTRFS_I(inode)->root;
4bef0848
CM		 2730 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2731 struct extent_buffer *eb;
2732 unsigned long len;
2733 u64 bytenr = page_offset(page);
2734
2735 if (page->private == EXTENT_PAGE_PRIVATE)
2736 goto out;
2737
2738 len = page->private >> 2;
f09d1f60 2739 eb = find_extent_buffer(io_tree, bytenr, len);
4bef0848
CM		 2740 if (!eb)
2741 goto out;
2742
01d658f2
CM		 2743 if (!btrfs_try_tree_write_lock(eb)) {
2744 flush_fn(data);
2745 btrfs_tree_lock(eb);
2746 }
4bef0848 2747 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
b9473439
CM		 2748
2749 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
2750 spin_lock(&root->fs_info->delalloc_lock);
2751 if (root->fs_info->dirty_metadata_bytes >= eb->len)
2752 root->fs_info->dirty_metadata_bytes -= eb->len;
2753 else
2754 WARN_ON(1);
2755 spin_unlock(&root->fs_info->delalloc_lock);
2756 }
2757
4bef0848
CM		 2758 btrfs_tree_unlock(eb);
2759 free_extent_buffer(eb);
2760out:
01d658f2
CM		 2761 if (!trylock_page(page)) {
2762 flush_fn(data);
2763 lock_page(page);
2764 }
4bef0848
CM		 2765 return 0;
2766}
2767
acce952b 2768static void btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
2769 int read_only)
2770{
2771 if (read_only)
2772 return;
2773
2774 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)
2775 printk(KERN_WARNING "warning: mount fs with errors, "
2776 "running btrfsck is recommended\n");
2777}
2778
2779int btrfs_error_commit_super(struct btrfs_root *root)
2780{
2781 int ret;
2782
2783 mutex_lock(&root->fs_info->cleaner_mutex);
2784 btrfs_run_delayed_iputs(root);
2785 mutex_unlock(&root->fs_info->cleaner_mutex);
2786
2787 down_write(&root->fs_info->cleanup_work_sem);
2788 up_write(&root->fs_info->cleanup_work_sem);
2789
2790 /* cleanup FS via transaction */
2791 btrfs_cleanup_transaction(root);
2792
2793 ret = write_ctree_super(NULL, root, 0);
2794
2795 return ret;
2796}
2797
2798static int btrfs_destroy_ordered_operations(struct btrfs_root *root)
2799{
2800 struct btrfs_inode *btrfs_inode;
2801 struct list_head splice;
2802
2803 INIT_LIST_HEAD(&splice);
2804
2805 mutex_lock(&root->fs_info->ordered_operations_mutex);
2806 spin_lock(&root->fs_info->ordered_extent_lock);
2807
2808 list_splice_init(&root->fs_info->ordered_operations, &splice);
2809 while (!list_empty(&splice)) {
2810 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
2811 ordered_operations);
2812
2813 list_del_init(&btrfs_inode->ordered_operations);
2814
2815 btrfs_invalidate_inodes(btrfs_inode->root);
2816 }
2817
2818 spin_unlock(&root->fs_info->ordered_extent_lock);
2819 mutex_unlock(&root->fs_info->ordered_operations_mutex);
2820
2821 return 0;
2822}
2823
2824static int btrfs_destroy_ordered_extents(struct btrfs_root *root)
2825{
2826 struct list_head splice;
2827 struct btrfs_ordered_extent *ordered;
2828 struct inode *inode;
2829
2830 INIT_LIST_HEAD(&splice);
2831
2832 spin_lock(&root->fs_info->ordered_extent_lock);
2833
2834 list_splice_init(&root->fs_info->ordered_extents, &splice);
2835 while (!list_empty(&splice)) {
2836 ordered = list_entry(splice.next, struct btrfs_ordered_extent,
2837 root_extent_list);
2838
2839 list_del_init(&ordered->root_extent_list);
2840 atomic_inc(&ordered->refs);
2841
2842 /* the inode may be getting freed (in sys_unlink path). */
2843 inode = igrab(ordered->inode);
2844
2845 spin_unlock(&root->fs_info->ordered_extent_lock);
2846 if (inode)
2847 iput(inode);
2848
2849 atomic_set(&ordered->refs, 1);
2850 btrfs_put_ordered_extent(ordered);
2851
2852 spin_lock(&root->fs_info->ordered_extent_lock);
2853 }
2854
2855 spin_unlock(&root->fs_info->ordered_extent_lock);
2856
2857 return 0;
2858}
2859
2860static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
2861 struct btrfs_root *root)
2862{
2863 struct rb_node *node;
2864 struct btrfs_delayed_ref_root *delayed_refs;
2865 struct btrfs_delayed_ref_node *ref;
2866 int ret = 0;
2867
2868 delayed_refs = &trans->delayed_refs;
2869
2870 spin_lock(&delayed_refs->lock);
2871 if (delayed_refs->num_entries == 0) {
cfece4db 2872 spin_unlock(&delayed_refs->lock);
acce952b 2873 printk(KERN_INFO "delayed_refs has NO entry\n");
2874 return ret;
2875 }
2876
2877 node = rb_first(&delayed_refs->root);
2878 while (node) {
2879 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2880 node = rb_next(node);
2881
2882 ref->in_tree = 0;
2883 rb_erase(&ref->rb_node, &delayed_refs->root);
2884 delayed_refs->num_entries--;
2885
2886 atomic_set(&ref->refs, 1);
2887 if (btrfs_delayed_ref_is_head(ref)) {
2888 struct btrfs_delayed_ref_head *head;
2889
2890 head = btrfs_delayed_node_to_head(ref);
2891 mutex_lock(&head->mutex);
2892 kfree(head->extent_op);
2893 delayed_refs->num_heads--;
2894 if (list_empty(&head->cluster))
2895 delayed_refs->num_heads_ready--;
2896 list_del_init(&head->cluster);
2897 mutex_unlock(&head->mutex);
2898 }
2899
2900 spin_unlock(&delayed_refs->lock);
2901 btrfs_put_delayed_ref(ref);
2902
2903 cond_resched();
2904 spin_lock(&delayed_refs->lock);
2905 }
2906
2907 spin_unlock(&delayed_refs->lock);
2908
2909 return ret;
2910}
2911
2912static int btrfs_destroy_pending_snapshots(struct btrfs_transaction *t)
2913{
2914 struct btrfs_pending_snapshot *snapshot;
2915 struct list_head splice;
2916
2917 INIT_LIST_HEAD(&splice);
2918
2919 list_splice_init(&t->pending_snapshots, &splice);
2920
2921 while (!list_empty(&splice)) {
2922 snapshot = list_entry(splice.next,
2923 struct btrfs_pending_snapshot,
2924 list);
2925
2926 list_del_init(&snapshot->list);
2927
2928 kfree(snapshot);
2929 }
2930
2931 return 0;
2932}
2933
2934static int btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
2935{
2936 struct btrfs_inode *btrfs_inode;
2937 struct list_head splice;
2938
2939 INIT_LIST_HEAD(&splice);
2940
acce952b 2941 spin_lock(&root->fs_info->delalloc_lock);
5be76758 2942 list_splice_init(&root->fs_info->delalloc_inodes, &splice);
acce952b 2943
2944 while (!list_empty(&splice)) {
2945 btrfs_inode = list_entry(splice.next, struct btrfs_inode,
2946 delalloc_inodes);
2947
2948 list_del_init(&btrfs_inode->delalloc_inodes);
2949
2950 btrfs_invalidate_inodes(btrfs_inode->root);
2951 }
2952
2953 spin_unlock(&root->fs_info->delalloc_lock);
2954
2955 return 0;
2956}
2957
2958static int btrfs_destroy_marked_extents(struct btrfs_root *root,
2959 struct extent_io_tree *dirty_pages,
2960 int mark)
2961{
2962 int ret;
2963 struct page *page;
2964 struct inode *btree_inode = root->fs_info->btree_inode;
2965 struct extent_buffer *eb;
2966 u64 start = 0;
2967 u64 end;
2968 u64 offset;
2969 unsigned long index;
2970
2971 while (1) {
2972 ret = find_first_extent_bit(dirty_pages, start, &start, &end,
2973 mark);
2974 if (ret)
2975 break;
2976
2977 clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
2978 while (start <= end) {
2979 index = start >> PAGE_CACHE_SHIFT;
2980 start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
2981 page = find_get_page(btree_inode->i_mapping, index);
2982 if (!page)
2983 continue;
2984 offset = page_offset(page);
2985
2986 spin_lock(&dirty_pages->buffer_lock);
2987 eb = radix_tree_lookup(
2988 &(&BTRFS_I(page->mapping->host)->io_tree)->buffer,
2989 offset >> PAGE_CACHE_SHIFT);
2990 spin_unlock(&dirty_pages->buffer_lock);
2991 if (eb) {
2992 ret = test_and_clear_bit(EXTENT_BUFFER_DIRTY,
2993 &eb->bflags);
2994 atomic_set(&eb->refs, 1);
2995 }
2996 if (PageWriteback(page))
2997 end_page_writeback(page);
2998
2999 lock_page(page);
3000 if (PageDirty(page)) {
3001 clear_page_dirty_for_io(page);
3002 spin_lock_irq(&page->mapping->tree_lock);
3003 radix_tree_tag_clear(&page->mapping->page_tree,
3004 page_index(page),
3005 PAGECACHE_TAG_DIRTY);
3006 spin_unlock_irq(&page->mapping->tree_lock);
3007 }
3008
3009 page->mapping->a_ops->invalidatepage(page, 0);
3010 unlock_page(page);
3011 }
3012 }
3013
3014 return ret;
3015}
3016
3017static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
3018 struct extent_io_tree *pinned_extents)
3019{
3020 struct extent_io_tree *unpin;
3021 u64 start;
3022 u64 end;
3023 int ret;
3024
3025 unpin = pinned_extents;
3026 while (1) {
3027 ret = find_first_extent_bit(unpin, 0, &start, &end,
3028 EXTENT_DIRTY);
3029 if (ret)
3030 break;
3031
3032 /* opt_discard */
5378e607
LD		 3033 if (btrfs_test_opt(root, DISCARD))
3034 ret = btrfs_error_discard_extent(root, start,
3035 end + 1 - start,
3036 NULL);
acce952b 3037
3038 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3039 btrfs_error_unpin_extent_range(root, start, end);
3040 cond_resched();
3041 }
3042
3043 return 0;
3044}
3045
3046static int btrfs_cleanup_transaction(struct btrfs_root *root)
3047{
3048 struct btrfs_transaction *t;
3049 LIST_HEAD(list);
3050
3051 WARN_ON(1);
3052
acce952b 3053 mutex_lock(&root->fs_info->transaction_kthread_mutex);
3054
a4abeea4 3055 spin_lock(&root->fs_info->trans_lock);
acce952b 3056 list_splice_init(&root->fs_info->trans_list, &list);
a4abeea4
JB		 3057 root->fs_info->trans_no_join = 1;
3058 spin_unlock(&root->fs_info->trans_lock);
3059
acce952b 3060 while (!list_empty(&list)) {
3061 t = list_entry(list.next, struct btrfs_transaction, list);
3062 if (!t)
3063 break;
3064
3065 btrfs_destroy_ordered_operations(root);
3066
3067 btrfs_destroy_ordered_extents(root);
3068
3069 btrfs_destroy_delayed_refs(t, root);
3070
3071 btrfs_block_rsv_release(root,
3072 &root->fs_info->trans_block_rsv,
3073 t->dirty_pages.dirty_bytes);
3074
3075 /* FIXME: cleanup wait for commit */
3076 t->in_commit = 1;
3077 t->blocked = 1;
3078 if (waitqueue_active(&root->fs_info->transaction_blocked_wait))
3079 wake_up(&root->fs_info->transaction_blocked_wait);
3080
3081 t->blocked = 0;
3082 if (waitqueue_active(&root->fs_info->transaction_wait))
3083 wake_up(&root->fs_info->transaction_wait);
acce952b 3084
acce952b 3085 t->commit_done = 1;
3086 if (waitqueue_active(&t->commit_wait))
3087 wake_up(&t->commit_wait);
acce952b 3088
3089 btrfs_destroy_pending_snapshots(t);
3090
3091 btrfs_destroy_delalloc_inodes(root);
3092
a4abeea4 3093 spin_lock(&root->fs_info->trans_lock);
acce952b 3094 root->fs_info->running_transaction = NULL;
a4abeea4 3095 spin_unlock(&root->fs_info->trans_lock);
acce952b 3096
3097 btrfs_destroy_marked_extents(root, &t->dirty_pages,
3098 EXTENT_DIRTY);
3099
3100 btrfs_destroy_pinned_extent(root,
3101 root->fs_info->pinned_extents);
3102
13c5a93e 3103 atomic_set(&t->use_count, 0);
acce952b 3104 list_del_init(&t->list);
3105 memset(t, 0, sizeof(*t));
3106 kmem_cache_free(btrfs_transaction_cachep, t);
3107 }
3108
a4abeea4
JB		 3109 spin_lock(&root->fs_info->trans_lock);
3110 root->fs_info->trans_no_join = 0;
3111 spin_unlock(&root->fs_info->trans_lock);
acce952b 3112 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
acce952b 3113
3114 return 0;
3115}
3116
d1310b2e 3117static struct extent_io_ops btree_extent_io_ops = {
4bef0848 3118 .write_cache_pages_lock_hook = btree_lock_page_hook,
ce9adaa5 3119 .readpage_end_io_hook = btree_readpage_end_io_hook,
0b86a832 3120 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 3121 /* note we're sharing with inode.c for the merge bio hook */
3122 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 3123};
Linux 6.x block layer and io_uring tree(s)