projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: Mem leak in btrfs_get_acl()
[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>
4b4e25f2 32#include "compat.h"
eb60ceac
CM		 33#include "ctree.h"
34#include "disk-io.h"
e089f05c 35#include "transaction.h"
0f7d52f4 36#include "btrfs_inode.h"
0b86a832 37#include "volumes.h"
db94535d 38#include "print-tree.h"
8b712842 39#include "async-thread.h"
925baedd 40#include "locking.h"
e02119d5 41#include "tree-log.h"
fa9c0d79 42#include "free-space-cache.h"
eb60ceac 43
d1310b2e 44static struct extent_io_ops btree_extent_io_ops;
8b712842 45static void end_workqueue_fn(struct btrfs_work *work);
4df27c4d 46static void free_fs_root(struct btrfs_root *root);
ce9adaa5 47
d352ac68
CM		 48/*
49 * end_io_wq structs are used to do processing in task context when an IO is
50 * complete. This is used during reads to verify checksums, and it is used
51 * by writes to insert metadata for new file extents after IO is complete.
52 */
ce9adaa5
CM		 53struct end_io_wq {
54 struct bio *bio;
55 bio_end_io_t *end_io;
56 void *private;
57 struct btrfs_fs_info *info;
58 int error;
22c59948 59 int metadata;
ce9adaa5 60 struct list_head list;
8b712842 61 struct btrfs_work work;
ce9adaa5 62};
0da5468f 63
d352ac68
CM		 64/*
65 * async submit bios are used to offload expensive checksumming
66 * onto the worker threads. They checksum file and metadata bios
67 * just before they are sent down the IO stack.
68 */
44b8bd7e
CM		 69struct async_submit_bio {
70 struct inode *inode;
71 struct bio *bio;
72 struct list_head list;
4a69a410
CM		 73 extent_submit_bio_hook_t *submit_bio_start;
74 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM		 75 int rw;
76 int mirror_num;
c8b97818 77 unsigned long bio_flags;
eaf25d93
CM		 78 /*
79 * bio_offset is optional, can be used if the pages in the bio
80 * can't tell us where in the file the bio should go
81 */
82 u64 bio_offset;
8b712842 83 struct btrfs_work work;
44b8bd7e
CM		 84};
85
4008c04a
CM		 86/* These are used to set the lockdep class on the extent buffer locks.
87 * The class is set by the readpage_end_io_hook after the buffer has
88 * passed csum validation but before the pages are unlocked.
89 *
90 * The lockdep class is also set by btrfs_init_new_buffer on freshly
91 * allocated blocks.
92 *
93 * The class is based on the level in the tree block, which allows lockdep
94 * to know that lower nodes nest inside the locks of higher nodes.
95 *
96 * We also add a check to make sure the highest level of the tree is
97 * the same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this
98 * code needs update as well.
99 */
100#ifdef CONFIG_DEBUG_LOCK_ALLOC
101# if BTRFS_MAX_LEVEL != 8
102# error
103# endif
104static struct lock_class_key btrfs_eb_class[BTRFS_MAX_LEVEL + 1];
105static const char *btrfs_eb_name[BTRFS_MAX_LEVEL + 1] = {
106 /* leaf */
107 "btrfs-extent-00",
108 "btrfs-extent-01",
109 "btrfs-extent-02",
110 "btrfs-extent-03",
111 "btrfs-extent-04",
112 "btrfs-extent-05",
113 "btrfs-extent-06",
114 "btrfs-extent-07",
115 /* highest possible level */
116 "btrfs-extent-08",
117};
118#endif
119
d352ac68
CM		 120/*
121 * extents on the btree inode are pretty simple, there's one extent
122 * that covers the entire device
123 */
b2950863
CH		 124static struct extent_map *btree_get_extent(struct inode *inode,
125 struct page *page, size_t page_offset, u64 start, u64 len,
126 int create)
7eccb903 127{
5f39d397
CM		 128 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
129 struct extent_map *em;
130 int ret;
131
890871be 132 read_lock(&em_tree->lock);
d1310b2e 133 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 134 if (em) {
135 em->bdev =
136 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
890871be 137 read_unlock(&em_tree->lock);
5f39d397 138 goto out;
a061fc8d 139 }
890871be 140 read_unlock(&em_tree->lock);
7b13b7b1 141
5f39d397
CM		 142 em = alloc_extent_map(GFP_NOFS);
143 if (!em) {
144 em = ERR_PTR(-ENOMEM);
145 goto out;
146 }
147 em->start = 0;
0afbaf8c 148 em->len = (u64)-1;
c8b97818 149 em->block_len = (u64)-1;
5f39d397 150 em->block_start = 0;
a061fc8d 151 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e 152
890871be 153 write_lock(&em_tree->lock);
5f39d397
CM		 154 ret = add_extent_mapping(em_tree, em);
155 if (ret == -EEXIST) {
0afbaf8c
CM		 156 u64 failed_start = em->start;
157 u64 failed_len = em->len;
158
5f39d397 159 free_extent_map(em);
7b13b7b1 160 em = lookup_extent_mapping(em_tree, start, len);
0afbaf8c 161 if (em) {
7b13b7b1 162 ret = 0;
0afbaf8c
CM		 163 } else {
164 em = lookup_extent_mapping(em_tree, failed_start,
165 failed_len);
7b13b7b1 166 ret = -EIO;
0afbaf8c 167 }
5f39d397 168 } else if (ret) {
7b13b7b1
CM		 169 free_extent_map(em);
170 em = NULL;
5f39d397 171 }
890871be 172 write_unlock(&em_tree->lock);
7b13b7b1
CM		 173
174 if (ret)
175 em = ERR_PTR(ret);
5f39d397
CM		 176out:
177 return em;
7eccb903
CM		 178}
179
19c00ddc
CM		 180u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
181{
163e783e 182 return crc32c(seed, data, len);
19c00ddc
CM		 183}
184
185void btrfs_csum_final(u32 crc, char *result)
186{
187 *(__le32 *)result = ~cpu_to_le32(crc);
188}
189
d352ac68
CM		 190/*
191 * compute the csum for a btree block, and either verify it or write it
192 * into the csum field of the block.
193 */
19c00ddc
CM		 194static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
195 int verify)
196{
607d432d
JB		 197 u16 csum_size =
198 btrfs_super_csum_size(&root->fs_info->super_copy);
199 char *result = NULL;
19c00ddc
CM		 200 unsigned long len;
201 unsigned long cur_len;
202 unsigned long offset = BTRFS_CSUM_SIZE;
203 char *map_token = NULL;
204 char *kaddr;
205 unsigned long map_start;
206 unsigned long map_len;
207 int err;
208 u32 crc = ~(u32)0;
607d432d 209 unsigned long inline_result;
19c00ddc
CM		 210
211 len = buf->len - offset;
d397712b 212 while (len > 0) {
19c00ddc
CM		 213 err = map_private_extent_buffer(buf, offset, 32,
214 &map_token, &kaddr,
215 &map_start, &map_len, KM_USER0);
d397712b 216 if (err)
19c00ddc 217 return 1;
19c00ddc
CM		 218 cur_len = min(len, map_len - (offset - map_start));
219 crc = btrfs_csum_data(root, kaddr + offset - map_start,
220 crc, cur_len);
221 len -= cur_len;
222 offset += cur_len;
223 unmap_extent_buffer(buf, map_token, KM_USER0);
224 }
607d432d
JB		 225 if (csum_size > sizeof(inline_result)) {
226 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
227 if (!result)
228 return 1;
229 } else {
230 result = (char *)&inline_result;
231 }
232
19c00ddc
CM		 233 btrfs_csum_final(crc, result);
234
235 if (verify) {
607d432d 236 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 237 u32 val;
238 u32 found = 0;
607d432d 239 memcpy(&found, result, csum_size);
e4204ded 240
607d432d 241 read_extent_buffer(buf, &val, 0, csum_size);
193f284d
CM		 242 if (printk_ratelimit()) {
243 printk(KERN_INFO "btrfs: %s checksum verify "
244 "failed on %llu wanted %X found %X "
245 "level %d\n",
246 root->fs_info->sb->s_id,
247 (unsigned long long)buf->start, val, found,
248 btrfs_header_level(buf));
249 }
607d432d
JB		 250 if (result != (char *)&inline_result)
251 kfree(result);
19c00ddc
CM		 252 return 1;
253 }
254 } else {
607d432d 255 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 256 }
607d432d
JB		 257 if (result != (char *)&inline_result)
258 kfree(result);
19c00ddc
CM		 259 return 0;
260}
261
d352ac68
CM		 262/*
263 * we can't consider a given block up to date unless the transid of the
264 * block matches the transid in the parent node's pointer. This is how we
265 * detect blocks that either didn't get written at all or got written
266 * in the wrong place.
267 */
1259ab75
CM		 268static int verify_parent_transid(struct extent_io_tree *io_tree,
269 struct extent_buffer *eb, u64 parent_transid)
270{
2ac55d41 271 struct extent_state *cached_state = NULL;
1259ab75
CM		 272 int ret;
273
274 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
275 return 0;
276
2ac55d41
JB		 277 lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
278 0, &cached_state, GFP_NOFS);
279 if (extent_buffer_uptodate(io_tree, eb, cached_state) &&
1259ab75
CM		 280 btrfs_header_generation(eb) == parent_transid) {
281 ret = 0;
282 goto out;
283 }
193f284d
CM		 284 if (printk_ratelimit()) {
285 printk("parent transid verify failed on %llu wanted %llu "
286 "found %llu\n",
287 (unsigned long long)eb->start,
288 (unsigned long long)parent_transid,
289 (unsigned long long)btrfs_header_generation(eb));
290 }
1259ab75 291 ret = 1;
2ac55d41 292 clear_extent_buffer_uptodate(io_tree, eb, &cached_state);
33958dc6 293out:
2ac55d41
JB		 294 unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
295 &cached_state, GFP_NOFS);
1259ab75 296 return ret;
1259ab75
CM		 297}
298
d352ac68
CM		 299/*
300 * helper to read a given tree block, doing retries as required when
301 * the checksums don't match and we have alternate mirrors to try.
302 */
f188591e
CM		 303static int btree_read_extent_buffer_pages(struct btrfs_root *root,
304 struct extent_buffer *eb,
ca7a79ad 305 u64 start, u64 parent_transid)
f188591e
CM		 306{
307 struct extent_io_tree *io_tree;
308 int ret;
309 int num_copies = 0;
310 int mirror_num = 0;
311
312 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
313 while (1) {
314 ret = read_extent_buffer_pages(io_tree, eb, start, 1,
315 btree_get_extent, mirror_num);
1259ab75
CM		 316 if (!ret &&
317 !verify_parent_transid(io_tree, eb, parent_transid))
f188591e 318 return ret;
d397712b 319
f188591e
CM		 320 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
321 eb->start, eb->len);
4235298e 322 if (num_copies == 1)
f188591e 323 return ret;
4235298e 324
f188591e 325 mirror_num++;
4235298e 326 if (mirror_num > num_copies)
f188591e 327 return ret;
f188591e 328 }
f188591e
CM		 329 return -EIO;
330}
19c00ddc 331
d352ac68 332/*
d397712b
CM		 333 * checksum a dirty tree block before IO. This has extra checks to make sure
334 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 335 */
d397712b 336
b2950863 337static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
19c00ddc 338{
d1310b2e 339 struct extent_io_tree *tree;
35ebb934 340 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
19c00ddc 341 u64 found_start;
19c00ddc
CM		 342 unsigned long len;
343 struct extent_buffer *eb;
f188591e
CM		 344 int ret;
345
d1310b2e 346 tree = &BTRFS_I(page->mapping->host)->io_tree;
19c00ddc
CM		 347
348 if (page->private == EXTENT_PAGE_PRIVATE)
349 goto out;
350 if (!page->private)
351 goto out;
352 len = page->private >> 2;
d397712b
CM		 353 WARN_ON(len == 0);
354
19c00ddc 355 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
ca7a79ad
CM		 356 ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
357 btrfs_header_generation(eb));
f188591e 358 BUG_ON(ret);
784b4e29
CM		 359 WARN_ON(!btrfs_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN));
360
19c00ddc
CM		 361 found_start = btrfs_header_bytenr(eb);
362 if (found_start != start) {
55c69072
CM		 363 WARN_ON(1);
364 goto err;
365 }
366 if (eb->first_page != page) {
55c69072
CM		 367 WARN_ON(1);
368 goto err;
369 }
370 if (!PageUptodate(page)) {
55c69072
CM		 371 WARN_ON(1);
372 goto err;
19c00ddc 373 }
19c00ddc 374 csum_tree_block(root, eb, 0);
55c69072 375err:
19c00ddc
CM		 376 free_extent_buffer(eb);
377out:
378 return 0;
379}
380
2b82032c
YZ		 381static int check_tree_block_fsid(struct btrfs_root *root,
382 struct extent_buffer *eb)
383{
384 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
385 u8 fsid[BTRFS_UUID_SIZE];
386 int ret = 1;
387
388 read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
389 BTRFS_FSID_SIZE);
390 while (fs_devices) {
391 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
392 ret = 0;
393 break;
394 }
395 fs_devices = fs_devices->seed;
396 }
397 return ret;
398}
399
4008c04a
CM		 400#ifdef CONFIG_DEBUG_LOCK_ALLOC
401void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level)
402{
403 lockdep_set_class_and_name(&eb->lock,
404 &btrfs_eb_class[level],
405 btrfs_eb_name[level]);
406}
407#endif
408
b2950863 409static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
ce9adaa5
CM		 410 struct extent_state *state)
411{
412 struct extent_io_tree *tree;
413 u64 found_start;
414 int found_level;
415 unsigned long len;
416 struct extent_buffer *eb;
417 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 418 int ret = 0;
ce9adaa5
CM		 419
420 tree = &BTRFS_I(page->mapping->host)->io_tree;
421 if (page->private == EXTENT_PAGE_PRIVATE)
422 goto out;
423 if (!page->private)
424 goto out;
d397712b 425
ce9adaa5 426 len = page->private >> 2;
d397712b
CM		 427 WARN_ON(len == 0);
428
ce9adaa5 429 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
f188591e 430
ce9adaa5 431 found_start = btrfs_header_bytenr(eb);
23a07867 432 if (found_start != start) {
193f284d
CM		 433 if (printk_ratelimit()) {
434 printk(KERN_INFO "btrfs bad tree block start "
435 "%llu %llu\n",
436 (unsigned long long)found_start,
437 (unsigned long long)eb->start);
438 }
f188591e 439 ret = -EIO;
ce9adaa5
CM		 440 goto err;
441 }
442 if (eb->first_page != page) {
d397712b
CM		 443 printk(KERN_INFO "btrfs bad first page %lu %lu\n",
444 eb->first_page->index, page->index);
ce9adaa5 445 WARN_ON(1);
f188591e 446 ret = -EIO;
ce9adaa5
CM		 447 goto err;
448 }
2b82032c 449 if (check_tree_block_fsid(root, eb)) {
193f284d
CM		 450 if (printk_ratelimit()) {
451 printk(KERN_INFO "btrfs bad fsid on block %llu\n",
452 (unsigned long long)eb->start);
453 }
1259ab75
CM		 454 ret = -EIO;
455 goto err;
456 }
ce9adaa5
CM		 457 found_level = btrfs_header_level(eb);
458
4008c04a
CM		 459 btrfs_set_buffer_lockdep_class(eb, found_level);
460
ce9adaa5 461 ret = csum_tree_block(root, eb, 1);
f188591e
CM		 462 if (ret)
463 ret = -EIO;
ce9adaa5
CM		 464
465 end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
466 end = eb->start + end - 1;
ce9adaa5
CM		 467err:
468 free_extent_buffer(eb);
469out:
f188591e 470 return ret;
ce9adaa5
CM		 471}
472
ce9adaa5 473static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM		 474{
475 struct end_io_wq *end_io_wq = bio->bi_private;
476 struct btrfs_fs_info *fs_info;
ce9adaa5 477
ce9adaa5 478 fs_info = end_io_wq->info;
ce9adaa5 479 end_io_wq->error = err;
8b712842
CM		 480 end_io_wq->work.func = end_workqueue_fn;
481 end_io_wq->work.flags = 0;
d20f7043 482
7b6d91da 483 if (bio->bi_rw & REQ_WRITE) {
0cb59c99 484 if (end_io_wq->metadata == 1)
cad321ad
CM		 485 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
486 &end_io_wq->work);
0cb59c99
JB		 487 else if (end_io_wq->metadata == 2)
488 btrfs_queue_worker(&fs_info->endio_freespace_worker,
489 &end_io_wq->work);
cad321ad
CM		 490 else
491 btrfs_queue_worker(&fs_info->endio_write_workers,
492 &end_io_wq->work);
d20f7043
CM		 493 } else {
494 if (end_io_wq->metadata)
495 btrfs_queue_worker(&fs_info->endio_meta_workers,
496 &end_io_wq->work);
497 else
498 btrfs_queue_worker(&fs_info->endio_workers,
499 &end_io_wq->work);
500 }
ce9adaa5
CM		 501}
502
0cb59c99
JB		 503/*
504 * For the metadata arg you want
505 *
506 * 0 - if data
507 * 1 - if normal metadta
508 * 2 - if writing to the free space cache area
509 */
22c59948
CM		 510int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
511 int metadata)
0b86a832 512{
ce9adaa5 513 struct end_io_wq *end_io_wq;
ce9adaa5
CM		 514 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
515 if (!end_io_wq)
516 return -ENOMEM;
517
518 end_io_wq->private = bio->bi_private;
519 end_io_wq->end_io = bio->bi_end_io;
22c59948 520 end_io_wq->info = info;
ce9adaa5
CM		 521 end_io_wq->error = 0;
522 end_io_wq->bio = bio;
22c59948 523 end_io_wq->metadata = metadata;
ce9adaa5
CM		 524
525 bio->bi_private = end_io_wq;
526 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 527 return 0;
528}
529
b64a2851 530unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 531{
4854ddd0
CM		 532 unsigned long limit = min_t(unsigned long,
533 info->workers.max_workers,
534 info->fs_devices->open_devices);
535 return 256 * limit;
536}
0986fe9e 537
4854ddd0
CM		 538int btrfs_congested_async(struct btrfs_fs_info *info, int iodone)
539{
b64a2851
CM		 540 return atomic_read(&info->nr_async_bios) >
541 btrfs_async_submit_limit(info);
0986fe9e
CM		 542}
543
4a69a410
CM		 544static void run_one_async_start(struct btrfs_work *work)
545{
4a69a410
CM		 546 struct async_submit_bio *async;
547
548 async = container_of(work, struct async_submit_bio, work);
4a69a410 549 async->submit_bio_start(async->inode, async->rw, async->bio,
eaf25d93
CM		 550 async->mirror_num, async->bio_flags,
551 async->bio_offset);
4a69a410
CM		 552}
553
554static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 555{
556 struct btrfs_fs_info *fs_info;
557 struct async_submit_bio *async;
4854ddd0 558 int limit;
8b712842
CM		 559
560 async = container_of(work, struct async_submit_bio, work);
561 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 562
b64a2851 563 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 564 limit = limit * 2 / 3;
565
8b712842 566 atomic_dec(&fs_info->nr_async_submits);
0986fe9e 567
b64a2851
CM		 568 if (atomic_read(&fs_info->nr_async_submits) < limit &&
569 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 570 wake_up(&fs_info->async_submit_wait);
571
4a69a410 572 async->submit_bio_done(async->inode, async->rw, async->bio,
eaf25d93
CM		 573 async->mirror_num, async->bio_flags,
574 async->bio_offset);
4a69a410
CM		 575}
576
577static void run_one_async_free(struct btrfs_work *work)
578{
579 struct async_submit_bio *async;
580
581 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 582 kfree(async);
583}
584
44b8bd7e
CM		 585int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
586 int rw, struct bio *bio, int mirror_num,
c8b97818 587 unsigned long bio_flags,
eaf25d93 588 u64 bio_offset,
4a69a410
CM		 589 extent_submit_bio_hook_t *submit_bio_start,
590 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 591{
592 struct async_submit_bio *async;
593
594 async = kmalloc(sizeof(*async), GFP_NOFS);
595 if (!async)
596 return -ENOMEM;
597
598 async->inode = inode;
599 async->rw = rw;
600 async->bio = bio;
601 async->mirror_num = mirror_num;
4a69a410
CM		 602 async->submit_bio_start = submit_bio_start;
603 async->submit_bio_done = submit_bio_done;
604
605 async->work.func = run_one_async_start;
606 async->work.ordered_func = run_one_async_done;
607 async->work.ordered_free = run_one_async_free;
608
8b712842 609 async->work.flags = 0;
c8b97818 610 async->bio_flags = bio_flags;
eaf25d93 611 async->bio_offset = bio_offset;
8c8bee1d 612
cb03c743 613 atomic_inc(&fs_info->nr_async_submits);
d313d7a3 614
7b6d91da 615 if (rw & REQ_SYNC)
d313d7a3
CM		 616 btrfs_set_work_high_prio(&async->work);
617
8b712842 618 btrfs_queue_worker(&fs_info->workers, &async->work);
9473f16c 619
d397712b 620 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 621 atomic_read(&fs_info->nr_async_submits)) {
622 wait_event(fs_info->async_submit_wait,
623 (atomic_read(&fs_info->nr_async_submits) == 0));
624 }
625
44b8bd7e
CM		 626 return 0;
627}
628
ce3ed71a
CM		 629static int btree_csum_one_bio(struct bio *bio)
630{
631 struct bio_vec *bvec = bio->bi_io_vec;
632 int bio_index = 0;
633 struct btrfs_root *root;
634
635 WARN_ON(bio->bi_vcnt <= 0);
d397712b 636 while (bio_index < bio->bi_vcnt) {
ce3ed71a
CM		 637 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
638 csum_dirty_buffer(root, bvec->bv_page);
639 bio_index++;
640 bvec++;
641 }
642 return 0;
643}
644
4a69a410
CM		 645static int __btree_submit_bio_start(struct inode *inode, int rw,
646 struct bio *bio, int mirror_num,
eaf25d93
CM		 647 unsigned long bio_flags,
648 u64 bio_offset)
22c59948 649{
8b712842
CM		 650 /*
651 * when we're called for a write, we're already in the async
5443be45 652 * submission context. Just jump into btrfs_map_bio
8b712842 653 */
4a69a410
CM		 654 btree_csum_one_bio(bio);
655 return 0;
656}
22c59948 657
4a69a410 658static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 659 int mirror_num, unsigned long bio_flags,
660 u64 bio_offset)
4a69a410 661{
8b712842 662 /*
4a69a410
CM		 663 * when we're called for a write, we're already in the async
664 * submission context. Just jump into btrfs_map_bio
8b712842 665 */
8b712842 666 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
0b86a832
CM		 667}
668
44b8bd7e 669static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
eaf25d93
CM		 670 int mirror_num, unsigned long bio_flags,
671 u64 bio_offset)
44b8bd7e 672{
cad321ad
CM		 673 int ret;
674
675 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
676 bio, 1);
677 BUG_ON(ret);
678
7b6d91da 679 if (!(rw & REQ_WRITE)) {
4a69a410
CM		 680 /*
681 * called for a read, do the setup so that checksum validation
682 * can happen in the async kernel threads
683 */
4a69a410 684 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
6f3577bd 685 mirror_num, 0);
44b8bd7e 686 }
d313d7a3 687
cad321ad
CM		 688 /*
689 * kthread helpers are used to submit writes so that checksumming
690 * can happen in parallel across all CPUs
691 */
44b8bd7e 692 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
c8b97818 693 inode, rw, bio, mirror_num, 0,
eaf25d93 694 bio_offset,
4a69a410
CM		 695 __btree_submit_bio_start,
696 __btree_submit_bio_done);
44b8bd7e
CM		 697}
698
3dd1462e 699#ifdef CONFIG_MIGRATION
784b4e29
CM		 700static int btree_migratepage(struct address_space *mapping,
701 struct page *newpage, struct page *page)
702{
703 /*
704 * we can't safely write a btree page from here,
705 * we haven't done the locking hook
706 */
707 if (PageDirty(page))
708 return -EAGAIN;
709 /*
710 * Buffers may be managed in a filesystem specific way.
711 * We must have no buffers or drop them.
712 */
713 if (page_has_private(page) &&
714 !try_to_release_page(page, GFP_KERNEL))
715 return -EAGAIN;
784b4e29
CM		 716 return migrate_page(mapping, newpage, page);
717}
3dd1462e 718#endif
784b4e29 719
0da5468f
CM		 720static int btree_writepage(struct page *page, struct writeback_control *wbc)
721{
d1310b2e 722 struct extent_io_tree *tree;
b9473439
CM		 723 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
724 struct extent_buffer *eb;
725 int was_dirty;
726
d1310b2e 727 tree = &BTRFS_I(page->mapping->host)->io_tree;
b9473439
CM		 728 if (!(current->flags & PF_MEMALLOC)) {
729 return extent_write_full_page(tree, page,
730 btree_get_extent, wbc);
731 }
5443be45 732
b9473439 733 redirty_page_for_writepage(wbc, page);
784b4e29 734 eb = btrfs_find_tree_block(root, page_offset(page), PAGE_CACHE_SIZE);
b9473439
CM		 735 WARN_ON(!eb);
736
737 was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
738 if (!was_dirty) {
739 spin_lock(&root->fs_info->delalloc_lock);
740 root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE;
741 spin_unlock(&root->fs_info->delalloc_lock);
5443be45 742 }
b9473439
CM		 743 free_extent_buffer(eb);
744
745 unlock_page(page);
746 return 0;
5f39d397 747}
0da5468f
CM		 748
749static int btree_writepages(struct address_space *mapping,
750 struct writeback_control *wbc)
751{
d1310b2e
CM		 752 struct extent_io_tree *tree;
753 tree = &BTRFS_I(mapping->host)->io_tree;
d8d5f3e1 754 if (wbc->sync_mode == WB_SYNC_NONE) {
b9473439 755 struct btrfs_root *root = BTRFS_I(mapping->host)->root;
793955bc 756 u64 num_dirty;
24ab9cd8 757 unsigned long thresh = 32 * 1024 * 1024;
448d640b
CM		 758
759 if (wbc->for_kupdate)
760 return 0;
761
b9473439
CM		 762 /* this is a bit racy, but that's ok */
763 num_dirty = root->fs_info->dirty_metadata_bytes;
d397712b 764 if (num_dirty < thresh)
793955bc 765 return 0;
793955bc 766 }
0da5468f
CM		 767 return extent_writepages(tree, mapping, btree_get_extent, wbc);
768}
769
b2950863 770static int btree_readpage(struct file *file, struct page *page)
5f39d397 771{
d1310b2e
CM		 772 struct extent_io_tree *tree;
773 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 774 return extent_read_full_page(tree, page, btree_get_extent);
775}
22b0ebda 776
70dec807 777static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 778{
d1310b2e
CM		 779 struct extent_io_tree *tree;
780 struct extent_map_tree *map;
5f39d397 781 int ret;
d98237b3 782
98509cfc 783 if (PageWriteback(page) || PageDirty(page))
d397712b 784 return 0;
98509cfc 785
d1310b2e
CM		 786 tree = &BTRFS_I(page->mapping->host)->io_tree;
787 map = &BTRFS_I(page->mapping->host)->extent_tree;
6af118ce 788
7b13b7b1 789 ret = try_release_extent_state(map, tree, page, gfp_flags);
d397712b 790 if (!ret)
6af118ce 791 return 0;
6af118ce
CM		 792
793 ret = try_release_extent_buffer(tree, page);
5f39d397
CM		 794 if (ret == 1) {
795 ClearPagePrivate(page);
796 set_page_private(page, 0);
797 page_cache_release(page);
798 }
6af118ce 799
d98237b3
CM		 800 return ret;
801}
802
5f39d397 803static void btree_invalidatepage(struct page *page, unsigned long offset)
d98237b3 804{
d1310b2e
CM		 805 struct extent_io_tree *tree;
806 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 807 extent_invalidatepage(tree, page, offset);
808 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 809 if (PagePrivate(page)) {
d397712b
CM		 810 printk(KERN_WARNING "btrfs warning page private not zero "
811 "on page %llu\n", (unsigned long long)page_offset(page));
9ad6b7bc
CM		 812 ClearPagePrivate(page);
813 set_page_private(page, 0);
814 page_cache_release(page);
815 }
d98237b3
CM		 816}
817
7f09410b 818static const struct address_space_operations btree_aops = {
d98237b3
CM		 819 .readpage = btree_readpage,
820 .writepage = btree_writepage,
0da5468f 821 .writepages = btree_writepages,
5f39d397
CM		 822 .releasepage = btree_releasepage,
823 .invalidatepage = btree_invalidatepage,
d98237b3 824 .sync_page = block_sync_page,
5a92bc88 825#ifdef CONFIG_MIGRATION
784b4e29 826 .migratepage = btree_migratepage,
5a92bc88 827#endif
d98237b3
CM		 828};
829
ca7a79ad
CM		 830int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
831 u64 parent_transid)
090d1875 832{
5f39d397
CM		 833 struct extent_buffer *buf = NULL;
834 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 835 int ret = 0;
090d1875 836
db94535d 837 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 838 if (!buf)
090d1875 839 return 0;
d1310b2e 840 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
f188591e 841 buf, 0, 0, btree_get_extent, 0);
5f39d397 842 free_extent_buffer(buf);
de428b63 843 return ret;
090d1875
CM		 844}
845
0999df54
CM		 846struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
847 u64 bytenr, u32 blocksize)
848{
849 struct inode *btree_inode = root->fs_info->btree_inode;
850 struct extent_buffer *eb;
851 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
852 bytenr, blocksize, GFP_NOFS);
853 return eb;
854}
855
856struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
857 u64 bytenr, u32 blocksize)
858{
859 struct inode *btree_inode = root->fs_info->btree_inode;
860 struct extent_buffer *eb;
861
862 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
863 bytenr, blocksize, NULL, GFP_NOFS);
864 return eb;
865}
866
867
e02119d5
CM		 868int btrfs_write_tree_block(struct extent_buffer *buf)
869{
8aa38c31
CH		 870 return filemap_fdatawrite_range(buf->first_page->mapping, buf->start,
871 buf->start + buf->len - 1);
e02119d5
CM		 872}
873
874int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
875{
8aa38c31
CH		 876 return filemap_fdatawait_range(buf->first_page->mapping,
877 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 878}
879
0999df54 880struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 881 u32 blocksize, u64 parent_transid)
0999df54
CM		 882{
883 struct extent_buffer *buf = NULL;
0999df54
CM		 884 int ret;
885
0999df54
CM		 886 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
887 if (!buf)
888 return NULL;
0999df54 889
ca7a79ad 890 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
ce9adaa5 891
d397712b 892 if (ret == 0)
b4ce94de 893 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
5f39d397 894 return buf;
ce9adaa5 895
eb60ceac
CM		 896}
897
e089f05c 898int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
5f39d397 899 struct extent_buffer *buf)
ed2ff2cb 900{
5f39d397 901 struct inode *btree_inode = root->fs_info->btree_inode;
55c69072 902 if (btrfs_header_generation(buf) ==
925baedd 903 root->fs_info->running_transaction->transid) {
b9447ef8 904 btrfs_assert_tree_locked(buf);
b4ce94de 905
b9473439
CM		 906 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
907 spin_lock(&root->fs_info->delalloc_lock);
908 if (root->fs_info->dirty_metadata_bytes >= buf->len)
909 root->fs_info->dirty_metadata_bytes -= buf->len;
910 else
911 WARN_ON(1);
912 spin_unlock(&root->fs_info->delalloc_lock);
913 }
b4ce94de 914
b9473439
CM		 915 /* ugh, clear_extent_buffer_dirty needs to lock the page */
916 btrfs_set_lock_blocking(buf);
d1310b2e 917 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
55c69072 918 buf);
925baedd 919 }
5f39d397
CM		 920 return 0;
921}
922
db94535d 923static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
87ee04eb 924 u32 stripesize, struct btrfs_root *root,
9f5fae2f 925 struct btrfs_fs_info *fs_info,
e20d96d6 926 u64 objectid)
d97e63b6 927{
cfaa7295 928 root->node = NULL;
a28ec197 929 root->commit_root = NULL;
db94535d
CM		 930 root->sectorsize = sectorsize;
931 root->nodesize = nodesize;
932 root->leafsize = leafsize;
87ee04eb 933 root->stripesize = stripesize;
123abc88 934 root->ref_cows = 0;
0b86a832 935 root->track_dirty = 0;
c71bf099 936 root->in_radix = 0;
d68fc57b
YZ		 937 root->orphan_item_inserted = 0;
938 root->orphan_cleanup_state = 0;
0b86a832 939
9f5fae2f 940 root->fs_info = fs_info;
0f7d52f4
CM		 941 root->objectid = objectid;
942 root->last_trans = 0;
13a8a7c8 943 root->highest_objectid = 0;
58176a96 944 root->name = NULL;
4313b399 945 root->in_sysfs = 0;
6bef4d31 946 root->inode_tree = RB_ROOT;
f0486c68 947 root->block_rsv = NULL;
d68fc57b 948 root->orphan_block_rsv = NULL;
0b86a832
CM		 949
950 INIT_LIST_HEAD(&root->dirty_list);
7b128766 951 INIT_LIST_HEAD(&root->orphan_list);
5d4f98a2 952 INIT_LIST_HEAD(&root->root_list);
925baedd 953 spin_lock_init(&root->node_lock);
d68fc57b 954 spin_lock_init(&root->orphan_lock);
5d4f98a2 955 spin_lock_init(&root->inode_lock);
f0486c68 956 spin_lock_init(&root->accounting_lock);
a2135011 957 mutex_init(&root->objectid_mutex);
e02119d5 958 mutex_init(&root->log_mutex);
7237f183
YZ		 959 init_waitqueue_head(&root->log_writer_wait);
960 init_waitqueue_head(&root->log_commit_wait[0]);
961 init_waitqueue_head(&root->log_commit_wait[1]);
962 atomic_set(&root->log_commit[0], 0);
963 atomic_set(&root->log_commit[1], 0);
964 atomic_set(&root->log_writers, 0);
965 root->log_batch = 0;
966 root->log_transid = 0;
257c62e1 967 root->last_log_commit = 0;
d0c803c4
CM		 968 extent_io_tree_init(&root->dirty_log_pages,
969 fs_info->btree_inode->i_mapping, GFP_NOFS);
017e5369 970
3768f368
CM		 971 memset(&root->root_key, 0, sizeof(root->root_key));
972 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 973 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 974 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
3f157a2f 975 root->defrag_trans_start = fs_info->generation;
58176a96 976 init_completion(&root->kobj_unregister);
6702ed49 977 root->defrag_running = 0;
4d775673 978 root->root_key.objectid = objectid;
3394e160
CM		 979 root->anon_super.s_root = NULL;
980 root->anon_super.s_dev = 0;
981 INIT_LIST_HEAD(&root->anon_super.s_list);
982 INIT_LIST_HEAD(&root->anon_super.s_instances);
983 init_rwsem(&root->anon_super.s_umount);
984
3768f368
CM		 985 return 0;
986}
987
db94535d 988static int find_and_setup_root(struct btrfs_root *tree_root,
9f5fae2f
CM		 989 struct btrfs_fs_info *fs_info,
990 u64 objectid,
e20d96d6 991 struct btrfs_root *root)
3768f368
CM		 992{
993 int ret;
db94535d 994 u32 blocksize;
84234f3a 995 u64 generation;
3768f368 996
db94535d 997 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 998 tree_root->sectorsize, tree_root->stripesize,
999 root, fs_info, objectid);
3768f368
CM		 1000 ret = btrfs_find_last_root(tree_root, objectid,
1001 &root->root_item, &root->root_key);
4df27c4d
YZ		 1002 if (ret > 0)
1003 return -ENOENT;
3768f368
CM		 1004 BUG_ON(ret);
1005
84234f3a 1006 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1007 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1008 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1009 blocksize, generation);
68433b73
CM		 1010 if (!root->node || !btrfs_buffer_uptodate(root->node, generation)) {
1011 free_extent_buffer(root->node);
1012 return -EIO;
1013 }
4df27c4d 1014 root->commit_root = btrfs_root_node(root);
d97e63b6
CM		 1015 return 0;
1016}
1017
7237f183
YZ		 1018static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1019 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1020{
1021 struct btrfs_root *root;
1022 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1023 struct extent_buffer *leaf;
e02119d5
CM		 1024
1025 root = kzalloc(sizeof(*root), GFP_NOFS);
1026 if (!root)
7237f183 1027 return ERR_PTR(-ENOMEM);
e02119d5
CM		 1028
1029 __setup_root(tree_root->nodesize, tree_root->leafsize,
1030 tree_root->sectorsize, tree_root->stripesize,
1031 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1032
1033 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1034 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1035 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
7237f183
YZ		 1036 /*
1037 * log trees do not get reference counted because they go away
1038 * before a real commit is actually done. They do store pointers
1039 * to file data extents, and those reference counts still get
1040 * updated (along with back refs to the log tree).
1041 */
e02119d5
CM		 1042 root->ref_cows = 0;
1043
5d4f98a2
YZ		 1044 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
1045 BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0);
7237f183
YZ		 1046 if (IS_ERR(leaf)) {
1047 kfree(root);
1048 return ERR_CAST(leaf);
1049 }
e02119d5 1050
5d4f98a2
YZ		 1051 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1052 btrfs_set_header_bytenr(leaf, leaf->start);
1053 btrfs_set_header_generation(leaf, trans->transid);
1054 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1055 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
7237f183 1056 root->node = leaf;
e02119d5
CM		 1057
1058 write_extent_buffer(root->node, root->fs_info->fsid,
1059 (unsigned long)btrfs_header_fsid(root->node),
1060 BTRFS_FSID_SIZE);
1061 btrfs_mark_buffer_dirty(root->node);
1062 btrfs_tree_unlock(root->node);
7237f183
YZ		 1063 return root;
1064}
1065
1066int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1067 struct btrfs_fs_info *fs_info)
1068{
1069 struct btrfs_root *log_root;
1070
1071 log_root = alloc_log_tree(trans, fs_info);
1072 if (IS_ERR(log_root))
1073 return PTR_ERR(log_root);
1074 WARN_ON(fs_info->log_root_tree);
1075 fs_info->log_root_tree = log_root;
1076 return 0;
1077}
1078
1079int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1080 struct btrfs_root *root)
1081{
1082 struct btrfs_root *log_root;
1083 struct btrfs_inode_item *inode_item;
1084
1085 log_root = alloc_log_tree(trans, root->fs_info);
1086 if (IS_ERR(log_root))
1087 return PTR_ERR(log_root);
1088
1089 log_root->last_trans = trans->transid;
1090 log_root->root_key.offset = root->root_key.objectid;
1091
1092 inode_item = &log_root->root_item.inode;
1093 inode_item->generation = cpu_to_le64(1);
1094 inode_item->size = cpu_to_le64(3);
1095 inode_item->nlink = cpu_to_le32(1);
1096 inode_item->nbytes = cpu_to_le64(root->leafsize);
1097 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1098
5d4f98a2 1099 btrfs_set_root_node(&log_root->root_item, log_root->node);
7237f183
YZ		 1100
1101 WARN_ON(root->log_root);
1102 root->log_root = log_root;
1103 root->log_transid = 0;
257c62e1 1104 root->last_log_commit = 0;
e02119d5
CM		 1105 return 0;
1106}
1107
1108struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
1109 struct btrfs_key *location)
1110{
1111 struct btrfs_root *root;
1112 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1113 struct btrfs_path *path;
5f39d397 1114 struct extent_buffer *l;
84234f3a 1115 u64 generation;
db94535d 1116 u32 blocksize;
0f7d52f4
CM		 1117 int ret = 0;
1118
5eda7b5e 1119 root = kzalloc(sizeof(*root), GFP_NOFS);
0cf6c620 1120 if (!root)
0f7d52f4 1121 return ERR_PTR(-ENOMEM);
0f7d52f4 1122 if (location->offset == (u64)-1) {
db94535d 1123 ret = find_and_setup_root(tree_root, fs_info,
0f7d52f4
CM		 1124 location->objectid, root);
1125 if (ret) {
0f7d52f4
CM		 1126 kfree(root);
1127 return ERR_PTR(ret);
1128 }
13a8a7c8 1129 goto out;
0f7d52f4
CM		 1130 }
1131
db94535d 1132 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 1133 tree_root->sectorsize, tree_root->stripesize,
1134 root, fs_info, location->objectid);
0f7d52f4
CM		 1135
1136 path = btrfs_alloc_path();
1137 BUG_ON(!path);
1138 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
13a8a7c8
YZ		 1139 if (ret == 0) {
1140 l = path->nodes[0];
1141 read_extent_buffer(l, &root->root_item,
1142 btrfs_item_ptr_offset(l, path->slots[0]),
1143 sizeof(root->root_item));
1144 memcpy(&root->root_key, location, sizeof(*location));
0f7d52f4 1145 }
0f7d52f4
CM		 1146 btrfs_free_path(path);
1147 if (ret) {
13a8a7c8
YZ		 1148 if (ret > 0)
1149 ret = -ENOENT;
0f7d52f4
CM		 1150 return ERR_PTR(ret);
1151 }
13a8a7c8 1152
84234f3a 1153 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1154 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1155 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1156 blocksize, generation);
5d4f98a2 1157 root->commit_root = btrfs_root_node(root);
0f7d52f4 1158 BUG_ON(!root->node);
13a8a7c8
YZ		 1159out:
1160 if (location->objectid != BTRFS_TREE_LOG_OBJECTID)
e02119d5 1161 root->ref_cows = 1;
13a8a7c8 1162
5eda7b5e
CM		 1163 return root;
1164}
1165
dc17ff8f
CM		 1166struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1167 u64 root_objectid)
1168{
1169 struct btrfs_root *root;
1170
1171 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
1172 return fs_info->tree_root;
1173 if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
1174 return fs_info->extent_root;
1175
1176 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1177 (unsigned long)root_objectid);
1178 return root;
1179}
1180
edbd8d4e
CM		 1181struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
1182 struct btrfs_key *location)
5eda7b5e
CM		 1183{
1184 struct btrfs_root *root;
1185 int ret;
1186
edbd8d4e
CM		 1187 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1188 return fs_info->tree_root;
1189 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1190 return fs_info->extent_root;
8f18cf13
CM		 1191 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1192 return fs_info->chunk_root;
1193 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1194 return fs_info->dev_root;
0403e47e
YZ		 1195 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1196 return fs_info->csum_root;
4df27c4d
YZ		 1197again:
1198 spin_lock(&fs_info->fs_roots_radix_lock);
5eda7b5e
CM		 1199 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1200 (unsigned long)location->objectid);
4df27c4d 1201 spin_unlock(&fs_info->fs_roots_radix_lock);
5eda7b5e
CM		 1202 if (root)
1203 return root;
1204
e02119d5 1205 root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
5eda7b5e
CM		 1206 if (IS_ERR(root))
1207 return root;
3394e160
CM		 1208
1209 set_anon_super(&root->anon_super, NULL);
1210
d68fc57b
YZ		 1211 if (btrfs_root_refs(&root->root_item) == 0) {
1212 ret = -ENOENT;
1213 goto fail;
1214 }
1215
1216 ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
1217 if (ret < 0)
1218 goto fail;
1219 if (ret == 0)
1220 root->orphan_item_inserted = 1;
1221
4df27c4d
YZ		 1222 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1223 if (ret)
1224 goto fail;
1225
1226 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 1227 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1228 (unsigned long)root->root_key.objectid,
0f7d52f4 1229 root);
d68fc57b 1230 if (ret == 0)
4df27c4d 1231 root->in_radix = 1;
d68fc57b 1232
4df27c4d
YZ		 1233 spin_unlock(&fs_info->fs_roots_radix_lock);
1234 radix_tree_preload_end();
0f7d52f4 1235 if (ret) {
4df27c4d
YZ		 1236 if (ret == -EEXIST) {
1237 free_fs_root(root);
1238 goto again;
1239 }
1240 goto fail;
0f7d52f4 1241 }
4df27c4d
YZ		 1242
1243 ret = btrfs_find_dead_roots(fs_info->tree_root,
1244 root->root_key.objectid);
1245 WARN_ON(ret);
edbd8d4e 1246 return root;
4df27c4d
YZ		 1247fail:
1248 free_fs_root(root);
1249 return ERR_PTR(ret);
edbd8d4e
CM		 1250}
1251
1252struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
1253 struct btrfs_key *location,
1254 const char *name, int namelen)
1255{
4df27c4d
YZ		 1256 return btrfs_read_fs_root_no_name(fs_info, location);
1257#if 0
edbd8d4e
CM		 1258 struct btrfs_root *root;
1259 int ret;
1260
1261 root = btrfs_read_fs_root_no_name(fs_info, location);
1262 if (!root)
1263 return NULL;
58176a96 1264
4313b399
CM		 1265 if (root->in_sysfs)
1266 return root;
1267
58176a96
JB		 1268 ret = btrfs_set_root_name(root, name, namelen);
1269 if (ret) {
5f39d397 1270 free_extent_buffer(root->node);
58176a96
JB		 1271 kfree(root);
1272 return ERR_PTR(ret);
1273 }
4df27c4d 1274
58176a96
JB		 1275 ret = btrfs_sysfs_add_root(root);
1276 if (ret) {
5f39d397 1277 free_extent_buffer(root->node);
58176a96
JB		 1278 kfree(root->name);
1279 kfree(root);
1280 return ERR_PTR(ret);
1281 }
4313b399 1282 root->in_sysfs = 1;
0f7d52f4 1283 return root;
4df27c4d 1284#endif
0f7d52f4 1285}
04160088
CM		 1286
1287static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1288{
1289 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1290 int ret = 0;
04160088
CM		 1291 struct btrfs_device *device;
1292 struct backing_dev_info *bdi;
b7967db7 1293
c6e30871 1294 list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1295 if (!device->bdev)
1296 continue;
04160088
CM		 1297 bdi = blk_get_backing_dev_info(device->bdev);
1298 if (bdi && bdi_congested(bdi, bdi_bits)) {
1299 ret = 1;
1300 break;
1301 }
1302 }
1303 return ret;
1304}
1305
38b66988
CM		 1306/*
1307 * this unplugs every device on the box, and it is only used when page
1308 * is null
1309 */
1310static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
1311{
38b66988
CM		 1312 struct btrfs_device *device;
1313 struct btrfs_fs_info *info;
1314
1315 info = (struct btrfs_fs_info *)bdi->unplug_io_data;
c6e30871 1316 list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
d20f7043
CM		 1317 if (!device->bdev)
1318 continue;
1319
38b66988 1320 bdi = blk_get_backing_dev_info(device->bdev);
d397712b 1321 if (bdi->unplug_io_fn)
38b66988 1322 bdi->unplug_io_fn(bdi, page);
38b66988
CM		 1323 }
1324}
1325
b2950863 1326static void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
04160088 1327{
38b66988 1328 struct inode *inode;
f2d8d74d
CM		 1329 struct extent_map_tree *em_tree;
1330 struct extent_map *em;
bcbfce8a 1331 struct address_space *mapping;
38b66988
CM		 1332 u64 offset;
1333
bcbfce8a 1334 /* the generic O_DIRECT read code does this */
9f0ba5bd 1335 if (1 || !page) {
38b66988
CM		 1336 __unplug_io_fn(bdi, page);
1337 return;
1338 }
1339
bcbfce8a
CM		 1340 /*
1341 * page->mapping may change at any time. Get a consistent copy
1342 * and use that for everything below
1343 */
1344 smp_mb();
1345 mapping = page->mapping;
1346 if (!mapping)
1347 return;
1348
1349 inode = mapping->host;
240d5d48
CM		 1350
1351 /*
1352 * don't do the expensive searching for a small number of
1353 * devices
1354 */
1355 if (BTRFS_I(inode)->root->fs_info->fs_devices->open_devices <= 2) {
1356 __unplug_io_fn(bdi, page);
1357 return;
1358 }
1359
38b66988 1360 offset = page_offset(page);
04160088 1361
f2d8d74d 1362 em_tree = &BTRFS_I(inode)->extent_tree;
890871be 1363 read_lock(&em_tree->lock);
f2d8d74d 1364 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
890871be 1365 read_unlock(&em_tree->lock);
89642229
CM		 1366 if (!em) {
1367 __unplug_io_fn(bdi, page);
f2d8d74d 1368 return;
89642229 1369 }
f2d8d74d 1370
89642229
CM		 1371 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1372 free_extent_map(em);
1373 __unplug_io_fn(bdi, page);
1374 return;
1375 }
f2d8d74d
CM		 1376 offset = offset - em->start;
1377 btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree,
1378 em->block_start + offset, page);
1379 free_extent_map(em);
04160088
CM		 1380}
1381
ad081f14
JA		 1382/*
1383 * If this fails, caller must call bdi_destroy() to get rid of the
1384 * bdi again.
1385 */
04160088
CM		 1386static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1387{
ad081f14
JA		 1388 int err;
1389
1390 bdi->capabilities = BDI_CAP_MAP_COPY;
e6d086d8 1391 err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
ad081f14
JA		 1392 if (err)
1393 return err;
1394
4575c9cc 1395 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1396 bdi->unplug_io_fn = btrfs_unplug_io_fn;
1397 bdi->unplug_io_data = info;
1398 bdi->congested_fn = btrfs_congested_fn;
1399 bdi->congested_data = info;
1400 return 0;
1401}
1402
ce9adaa5
CM		 1403static int bio_ready_for_csum(struct bio *bio)
1404{
1405 u64 length = 0;
1406 u64 buf_len = 0;
1407 u64 start = 0;
1408 struct page *page;
1409 struct extent_io_tree *io_tree = NULL;
ce9adaa5
CM		 1410 struct bio_vec *bvec;
1411 int i;
1412 int ret;
1413
1414 bio_for_each_segment(bvec, bio, i) {
1415 page = bvec->bv_page;
1416 if (page->private == EXTENT_PAGE_PRIVATE) {
1417 length += bvec->bv_len;
1418 continue;
1419 }
1420 if (!page->private) {
1421 length += bvec->bv_len;
1422 continue;
1423 }
1424 length = bvec->bv_len;
1425 buf_len = page->private >> 2;
1426 start = page_offset(page) + bvec->bv_offset;
1427 io_tree = &BTRFS_I(page->mapping->host)->io_tree;
ce9adaa5
CM		 1428 }
1429 /* are we fully contained in this bio? */
1430 if (buf_len <= length)
1431 return 1;
1432
1433 ret = extent_range_uptodate(io_tree, start + length,
1434 start + buf_len - 1);
ce9adaa5
CM		 1435 return ret;
1436}
1437
8b712842
CM		 1438/*
1439 * called by the kthread helper functions to finally call the bio end_io
1440 * functions. This is where read checksum verification actually happens
1441 */
1442static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1443{
ce9adaa5 1444 struct bio *bio;
8b712842
CM		 1445 struct end_io_wq *end_io_wq;
1446 struct btrfs_fs_info *fs_info;
ce9adaa5 1447 int error;
ce9adaa5 1448
8b712842
CM		 1449 end_io_wq = container_of(work, struct end_io_wq, work);
1450 bio = end_io_wq->bio;
1451 fs_info = end_io_wq->info;
ce9adaa5 1452
cad321ad 1453 /* metadata bio reads are special because the whole tree block must
8b712842
CM		 1454 * be checksummed at once. This makes sure the entire block is in
1455 * ram and up to date before trying to verify things. For
1456 * blocksize <= pagesize, it is basically a noop
1457 */
7b6d91da 1458 if (!(bio->bi_rw & REQ_WRITE) && end_io_wq->metadata &&
cad321ad 1459 !bio_ready_for_csum(bio)) {
d20f7043 1460 btrfs_queue_worker(&fs_info->endio_meta_workers,
8b712842
CM		 1461 &end_io_wq->work);
1462 return;
1463 }
1464 error = end_io_wq->error;
1465 bio->bi_private = end_io_wq->private;
1466 bio->bi_end_io = end_io_wq->end_io;
1467 kfree(end_io_wq);
8b712842 1468 bio_endio(bio, error);
44b8bd7e
CM		 1469}
1470
a74a4b97
CM		 1471static int cleaner_kthread(void *arg)
1472{
1473 struct btrfs_root *root = arg;
1474
1475 do {
a74a4b97 1476 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
76dda93c
YZ		 1477
1478 if (!(root->fs_info->sb->s_flags & MS_RDONLY) &&
1479 mutex_trylock(&root->fs_info->cleaner_mutex)) {
24bbcf04 1480 btrfs_run_delayed_iputs(root);
76dda93c
YZ		 1481 btrfs_clean_old_snapshots(root);
1482 mutex_unlock(&root->fs_info->cleaner_mutex);
1483 }
a74a4b97
CM		 1484
1485 if (freezing(current)) {
1486 refrigerator();
1487 } else {
a74a4b97 1488 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1489 if (!kthread_should_stop())
1490 schedule();
a74a4b97
CM		 1491 __set_current_state(TASK_RUNNING);
1492 }
1493 } while (!kthread_should_stop());
1494 return 0;
1495}
1496
1497static int transaction_kthread(void *arg)
1498{
1499 struct btrfs_root *root = arg;
1500 struct btrfs_trans_handle *trans;
1501 struct btrfs_transaction *cur;
8929ecfa 1502 u64 transid;
a74a4b97
CM		 1503 unsigned long now;
1504 unsigned long delay;
1505 int ret;
1506
1507 do {
a74a4b97
CM		 1508 delay = HZ * 30;
1509 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1510 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1511
8929ecfa 1512 spin_lock(&root->fs_info->new_trans_lock);
a74a4b97
CM		 1513 cur = root->fs_info->running_transaction;
1514 if (!cur) {
8929ecfa 1515 spin_unlock(&root->fs_info->new_trans_lock);
a74a4b97
CM		 1516 goto sleep;
1517 }
31153d81 1518
a74a4b97 1519 now = get_seconds();
8929ecfa
YZ		 1520 if (!cur->blocked &&
1521 (now < cur->start_time || now - cur->start_time < 30)) {
1522 spin_unlock(&root->fs_info->new_trans_lock);
a74a4b97
CM		 1523 delay = HZ * 5;
1524 goto sleep;
1525 }
8929ecfa
YZ		 1526 transid = cur->transid;
1527 spin_unlock(&root->fs_info->new_trans_lock);
56bec294 1528
8929ecfa
YZ		 1529 trans = btrfs_join_transaction(root, 1);
1530 if (transid == trans->transid) {
1531 ret = btrfs_commit_transaction(trans, root);
1532 BUG_ON(ret);
1533 } else {
1534 btrfs_end_transaction(trans, root);
1535 }
a74a4b97
CM		 1536sleep:
1537 wake_up_process(root->fs_info->cleaner_kthread);
1538 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1539
1540 if (freezing(current)) {
1541 refrigerator();
1542 } else {
a74a4b97 1543 set_current_state(TASK_INTERRUPTIBLE);
8929ecfa
YZ		 1544 if (!kthread_should_stop() &&
1545 !btrfs_transaction_blocked(root->fs_info))
1546 schedule_timeout(delay);
a74a4b97
CM		 1547 __set_current_state(TASK_RUNNING);
1548 }
1549 } while (!kthread_should_stop());
1550 return 0;
1551}
1552
8a4b83cc 1553struct btrfs_root *open_ctree(struct super_block *sb,
dfe25020
CM		 1554 struct btrfs_fs_devices *fs_devices,
1555 char *options)
2e635a27 1556{
db94535d
CM		 1557 u32 sectorsize;
1558 u32 nodesize;
1559 u32 leafsize;
1560 u32 blocksize;
87ee04eb 1561 u32 stripesize;
84234f3a 1562 u64 generation;
f2b636e8 1563 u64 features;
3de4586c 1564 struct btrfs_key location;
a061fc8d 1565 struct buffer_head *bh;
e02119d5 1566 struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
e20d96d6 1567 GFP_NOFS);
d20f7043
CM		 1568 struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root),
1569 GFP_NOFS);
450ba0ea
JB		 1570 struct btrfs_root *tree_root = btrfs_sb(sb);
1571 struct btrfs_fs_info *fs_info = tree_root->fs_info;
e02119d5 1572 struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
0b86a832 1573 GFP_NOFS);
e02119d5 1574 struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
0b86a832 1575 GFP_NOFS);
e02119d5
CM		 1576 struct btrfs_root *log_tree_root;
1577
eb60ceac 1578 int ret;
e58ca020 1579 int err = -EINVAL;
4543df7e 1580
2c90e5d6 1581 struct btrfs_super_block *disk_super;
8790d502 1582
0463bb4e 1583 if (!extent_root || !tree_root || !fs_info ||
d20f7043 1584 !chunk_root || !dev_root || !csum_root) {
39279cc3
CM		 1585 err = -ENOMEM;
1586 goto fail;
1587 }
76dda93c
YZ		 1588
1589 ret = init_srcu_struct(&fs_info->subvol_srcu);
1590 if (ret) {
1591 err = ret;
1592 goto fail;
1593 }
1594
1595 ret = setup_bdi(fs_info, &fs_info->bdi);
1596 if (ret) {
1597 err = ret;
1598 goto fail_srcu;
1599 }
1600
1601 fs_info->btree_inode = new_inode(sb);
1602 if (!fs_info->btree_inode) {
1603 err = -ENOMEM;
1604 goto fail_bdi;
1605 }
1606
1607 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
8fd17795 1608 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 1609 INIT_LIST_HEAD(&fs_info->dead_roots);
24bbcf04 1610 INIT_LIST_HEAD(&fs_info->delayed_iputs);
19c00ddc 1611 INIT_LIST_HEAD(&fs_info->hashers);
ea8c2819 1612 INIT_LIST_HEAD(&fs_info->delalloc_inodes);
5a3f23d5 1613 INIT_LIST_HEAD(&fs_info->ordered_operations);
11833d66 1614 INIT_LIST_HEAD(&fs_info->caching_block_groups);
1832a6d5 1615 spin_lock_init(&fs_info->delalloc_lock);
cee36a03 1616 spin_lock_init(&fs_info->new_trans_lock);
31153d81 1617 spin_lock_init(&fs_info->ref_cache_lock);
76dda93c 1618 spin_lock_init(&fs_info->fs_roots_radix_lock);
24bbcf04 1619 spin_lock_init(&fs_info->delayed_iput_lock);
19c00ddc 1620
58176a96 1621 init_completion(&fs_info->kobj_unregister);
9f5fae2f
CM		 1622 fs_info->tree_root = tree_root;
1623 fs_info->extent_root = extent_root;
d20f7043 1624 fs_info->csum_root = csum_root;
0b86a832
CM		 1625 fs_info->chunk_root = chunk_root;
1626 fs_info->dev_root = dev_root;
8a4b83cc 1627 fs_info->fs_devices = fs_devices;
0b86a832 1628 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 1629 INIT_LIST_HEAD(&fs_info->space_info);
0b86a832 1630 btrfs_mapping_init(&fs_info->mapping_tree);
f0486c68
YZ		 1631 btrfs_init_block_rsv(&fs_info->global_block_rsv);
1632 btrfs_init_block_rsv(&fs_info->delalloc_block_rsv);
1633 btrfs_init_block_rsv(&fs_info->trans_block_rsv);
1634 btrfs_init_block_rsv(&fs_info->chunk_block_rsv);
1635 btrfs_init_block_rsv(&fs_info->empty_block_rsv);
1636 INIT_LIST_HEAD(&fs_info->durable_block_rsv_list);
1637 mutex_init(&fs_info->durable_block_rsv_mutex);
cb03c743 1638 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 1639 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 1640 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 1641 atomic_set(&fs_info->nr_async_bios, 0);
e20d96d6 1642 fs_info->sb = sb;
6f568d35 1643 fs_info->max_inline = 8192 * 1024;
9ed74f2d 1644 fs_info->metadata_ratio = 0;
c8b97818 1645
b34b086c
CM		 1646 fs_info->thread_pool_size = min_t(unsigned long,
1647 num_online_cpus() + 2, 8);
0afbaf8c 1648
3eaa2885
CM		 1649 INIT_LIST_HEAD(&fs_info->ordered_extents);
1650 spin_lock_init(&fs_info->ordered_extent_lock);
1651
a061fc8d
CM		 1652 sb->s_blocksize = 4096;
1653 sb->s_blocksize_bits = blksize_bits(4096);
32a88aa1 1654 sb->s_bdi = &fs_info->bdi;
a061fc8d 1655
76dda93c
YZ		 1656 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
1657 fs_info->btree_inode->i_nlink = 1;
0afbaf8c
CM		 1658 /*
1659 * we set the i_size on the btree inode to the max possible int.
1660 * the real end of the address space is determined by all of
1661 * the devices in the system
1662 */
1663 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 1664 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 1665 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
1666
5d4f98a2 1667 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
d1310b2e 1668 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
5f39d397
CM		 1669 fs_info->btree_inode->i_mapping,
1670 GFP_NOFS);
d1310b2e
CM		 1671 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
1672 GFP_NOFS);
1673
1674 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 1675
76dda93c
YZ		 1676 BTRFS_I(fs_info->btree_inode)->root = tree_root;
1677 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
1678 sizeof(struct btrfs_key));
1679 BTRFS_I(fs_info->btree_inode)->dummy_inode = 1;
c65ddb52 1680 insert_inode_hash(fs_info->btree_inode);
76dda93c 1681
0f9dd46c 1682 spin_lock_init(&fs_info->block_group_cache_lock);
6bef4d31 1683 fs_info->block_group_cache_tree = RB_ROOT;
0f9dd46c 1684
11833d66 1685 extent_io_tree_init(&fs_info->freed_extents[0],
1a5bc167 1686 fs_info->btree_inode->i_mapping, GFP_NOFS);
11833d66
YZ		 1687 extent_io_tree_init(&fs_info->freed_extents[1],
1688 fs_info->btree_inode->i_mapping, GFP_NOFS);
1689 fs_info->pinned_extents = &fs_info->freed_extents[0];
e66f709b 1690 fs_info->do_barriers = 1;
e18e4809 1691
39279cc3 1692
79154b1b 1693 mutex_init(&fs_info->trans_mutex);
5a3f23d5 1694 mutex_init(&fs_info->ordered_operations_mutex);
e02119d5 1695 mutex_init(&fs_info->tree_log_mutex);
925baedd 1696 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 1697 mutex_init(&fs_info->transaction_kthread_mutex);
1698 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 1699 mutex_init(&fs_info->volume_mutex);
276e680d 1700 init_rwsem(&fs_info->extent_commit_sem);
c71bf099 1701 init_rwsem(&fs_info->cleanup_work_sem);
76dda93c 1702 init_rwsem(&fs_info->subvol_sem);
fa9c0d79
CM		 1703
1704 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
1705 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
1706
e6dcd2dc 1707 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 1708 init_waitqueue_head(&fs_info->transaction_wait);
bb9c12c9 1709 init_waitqueue_head(&fs_info->transaction_blocked_wait);
4854ddd0 1710 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 1711
0b86a832 1712 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 1713 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 1714
a512bbf8 1715 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
a061fc8d 1716 if (!bh)
39279cc3 1717 goto fail_iput;
39279cc3 1718
a061fc8d 1719 memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy));
2d69a0f8
YZ		 1720 memcpy(&fs_info->super_for_commit, &fs_info->super_copy,
1721 sizeof(fs_info->super_for_commit));
a061fc8d 1722 brelse(bh);
5f39d397 1723
a061fc8d 1724 memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE);
0b86a832 1725
5f39d397 1726 disk_super = &fs_info->super_copy;
0f7d52f4 1727 if (!btrfs_super_root(disk_super))
c6e2bac1 1728 goto fail_iput;
0f7d52f4 1729
2b82032c
YZ		 1730 ret = btrfs_parse_options(tree_root, options);
1731 if (ret) {
1732 err = ret;
c6e2bac1 1733 goto fail_iput;
2b82032c 1734 }
dfe25020 1735
f2b636e8
JB		 1736 features = btrfs_super_incompat_flags(disk_super) &
1737 ~BTRFS_FEATURE_INCOMPAT_SUPP;
1738 if (features) {
1739 printk(KERN_ERR "BTRFS: couldn't mount because of "
1740 "unsupported optional features (%Lx).\n",
21380931 1741 (unsigned long long)features);
f2b636e8 1742 err = -EINVAL;
c6e2bac1 1743 goto fail_iput;
f2b636e8
JB		 1744 }
1745
5d4f98a2 1746 features = btrfs_super_incompat_flags(disk_super);
a6fa6fae
LZ		 1747 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
1748 if (tree_root->fs_info->compress_type & BTRFS_COMPRESS_LZO)
1749 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1750 btrfs_set_super_incompat_flags(disk_super, features);
5d4f98a2 1751
f2b636e8
JB		 1752 features = btrfs_super_compat_ro_flags(disk_super) &
1753 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
1754 if (!(sb->s_flags & MS_RDONLY) && features) {
1755 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
1756 "unsupported option features (%Lx).\n",
21380931 1757 (unsigned long long)features);
f2b636e8 1758 err = -EINVAL;
c6e2bac1 1759 goto fail_iput;
f2b636e8 1760 }
61d92c32
CM		 1761
1762 btrfs_init_workers(&fs_info->generic_worker,
1763 "genwork", 1, NULL);
1764
5443be45 1765 btrfs_init_workers(&fs_info->workers, "worker",
61d92c32
CM		 1766 fs_info->thread_pool_size,
1767 &fs_info->generic_worker);
c8b97818 1768
771ed689 1769 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
61d92c32
CM		 1770 fs_info->thread_pool_size,
1771 &fs_info->generic_worker);
771ed689 1772
5443be45 1773 btrfs_init_workers(&fs_info->submit_workers, "submit",
b720d209 1774 min_t(u64, fs_devices->num_devices,
61d92c32
CM		 1775 fs_info->thread_pool_size),
1776 &fs_info->generic_worker);
61b49440
CM		 1777
1778 /* a higher idle thresh on the submit workers makes it much more
1779 * likely that bios will be send down in a sane order to the
1780 * devices
1781 */
1782 fs_info->submit_workers.idle_thresh = 64;
53863232 1783
771ed689 1784 fs_info->workers.idle_thresh = 16;
4a69a410 1785 fs_info->workers.ordered = 1;
61b49440 1786
771ed689
CM		 1787 fs_info->delalloc_workers.idle_thresh = 2;
1788 fs_info->delalloc_workers.ordered = 1;
1789
61d92c32
CM		 1790 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
1791 &fs_info->generic_worker);
5443be45 1792 btrfs_init_workers(&fs_info->endio_workers, "endio",
61d92c32
CM		 1793 fs_info->thread_pool_size,
1794 &fs_info->generic_worker);
d20f7043 1795 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
61d92c32
CM		 1796 fs_info->thread_pool_size,
1797 &fs_info->generic_worker);
cad321ad 1798 btrfs_init_workers(&fs_info->endio_meta_write_workers,
61d92c32
CM		 1799 "endio-meta-write", fs_info->thread_pool_size,
1800 &fs_info->generic_worker);
5443be45 1801 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
61d92c32
CM		 1802 fs_info->thread_pool_size,
1803 &fs_info->generic_worker);
0cb59c99
JB		 1804 btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
1805 1, &fs_info->generic_worker);
61b49440
CM		 1806
1807 /*
1808 * endios are largely parallel and should have a very
1809 * low idle thresh
1810 */
1811 fs_info->endio_workers.idle_thresh = 4;
b51912c9
CM		 1812 fs_info->endio_meta_workers.idle_thresh = 4;
1813
9042846b
CM		 1814 fs_info->endio_write_workers.idle_thresh = 2;
1815 fs_info->endio_meta_write_workers.idle_thresh = 2;
1816
4543df7e 1817 btrfs_start_workers(&fs_info->workers, 1);
61d92c32 1818 btrfs_start_workers(&fs_info->generic_worker, 1);
1cc127b5 1819 btrfs_start_workers(&fs_info->submit_workers, 1);
771ed689 1820 btrfs_start_workers(&fs_info->delalloc_workers, 1);
247e743c 1821 btrfs_start_workers(&fs_info->fixup_workers, 1);
9042846b
CM		 1822 btrfs_start_workers(&fs_info->endio_workers, 1);
1823 btrfs_start_workers(&fs_info->endio_meta_workers, 1);
1824 btrfs_start_workers(&fs_info->endio_meta_write_workers, 1);
1825 btrfs_start_workers(&fs_info->endio_write_workers, 1);
0cb59c99 1826 btrfs_start_workers(&fs_info->endio_freespace_worker, 1);
4543df7e 1827
4575c9cc 1828 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 1829 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
1830 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 1831
db94535d
CM		 1832 nodesize = btrfs_super_nodesize(disk_super);
1833 leafsize = btrfs_super_leafsize(disk_super);
1834 sectorsize = btrfs_super_sectorsize(disk_super);
87ee04eb 1835 stripesize = btrfs_super_stripesize(disk_super);
db94535d
CM		 1836 tree_root->nodesize = nodesize;
1837 tree_root->leafsize = leafsize;
1838 tree_root->sectorsize = sectorsize;
87ee04eb 1839 tree_root->stripesize = stripesize;
a061fc8d
CM		 1840
1841 sb->s_blocksize = sectorsize;
1842 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 1843
39279cc3
CM		 1844 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
1845 sizeof(disk_super->magic))) {
d397712b 1846 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 1847 goto fail_sb_buffer;
1848 }
19c00ddc 1849
925baedd 1850 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 1851 ret = btrfs_read_sys_array(tree_root);
925baedd 1852 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 1853 if (ret) {
d397712b
CM		 1854 printk(KERN_WARNING "btrfs: failed to read the system "
1855 "array on %s\n", sb->s_id);
5d4f98a2 1856 goto fail_sb_buffer;
84eed90f 1857 }
0b86a832
CM		 1858
1859 blocksize = btrfs_level_size(tree_root,
1860 btrfs_super_chunk_root_level(disk_super));
84234f3a 1861 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 1862
1863 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1864 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1865
1866 chunk_root->node = read_tree_block(chunk_root,
1867 btrfs_super_chunk_root(disk_super),
84234f3a 1868 blocksize, generation);
0b86a832 1869 BUG_ON(!chunk_root->node);
83121942
DW		 1870 if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
1871 printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
1872 sb->s_id);
1873 goto fail_chunk_root;
1874 }
5d4f98a2
YZ		 1875 btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
1876 chunk_root->commit_root = btrfs_root_node(chunk_root);
0b86a832 1877
e17cade2 1878 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
d397712b
CM		 1879 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
1880 BTRFS_UUID_SIZE);
e17cade2 1881
925baedd 1882 mutex_lock(&fs_info->chunk_mutex);
0b86a832 1883 ret = btrfs_read_chunk_tree(chunk_root);
925baedd 1884 mutex_unlock(&fs_info->chunk_mutex);
2b82032c 1885 if (ret) {
d397712b
CM		 1886 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
1887 sb->s_id);
2b82032c
YZ		 1888 goto fail_chunk_root;
1889 }
0b86a832 1890
dfe25020
CM		 1891 btrfs_close_extra_devices(fs_devices);
1892
db94535d
CM		 1893 blocksize = btrfs_level_size(tree_root,
1894 btrfs_super_root_level(disk_super));
84234f3a 1895 generation = btrfs_super_generation(disk_super);
0b86a832 1896
e20d96d6 1897 tree_root->node = read_tree_block(tree_root,
db94535d 1898 btrfs_super_root(disk_super),
84234f3a 1899 blocksize, generation);
39279cc3 1900 if (!tree_root->node)
2b82032c 1901 goto fail_chunk_root;
83121942
DW		 1902 if (!test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
1903 printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
1904 sb->s_id);
1905 goto fail_tree_root;
1906 }
5d4f98a2
YZ		 1907 btrfs_set_root_node(&tree_root->root_item, tree_root->node);
1908 tree_root->commit_root = btrfs_root_node(tree_root);
db94535d
CM		 1909
1910 ret = find_and_setup_root(tree_root, fs_info,
e20d96d6 1911 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
0b86a832 1912 if (ret)
39279cc3 1913 goto fail_tree_root;
0b86a832
CM		 1914 extent_root->track_dirty = 1;
1915
1916 ret = find_and_setup_root(tree_root, fs_info,
1917 BTRFS_DEV_TREE_OBJECTID, dev_root);
0b86a832
CM		 1918 if (ret)
1919 goto fail_extent_root;
5d4f98a2 1920 dev_root->track_dirty = 1;
3768f368 1921
d20f7043
CM		 1922 ret = find_and_setup_root(tree_root, fs_info,
1923 BTRFS_CSUM_TREE_OBJECTID, csum_root);
1924 if (ret)
5d4f98a2 1925 goto fail_dev_root;
d20f7043
CM		 1926
1927 csum_root->track_dirty = 1;
1928
8929ecfa
YZ		 1929 fs_info->generation = generation;
1930 fs_info->last_trans_committed = generation;
1931 fs_info->data_alloc_profile = (u64)-1;
1932 fs_info->metadata_alloc_profile = (u64)-1;
1933 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
1934
1b1d1f66
JB		 1935 ret = btrfs_read_block_groups(extent_root);
1936 if (ret) {
1937 printk(KERN_ERR "Failed to read block groups: %d\n", ret);
1938 goto fail_block_groups;
1939 }
9078a3e1 1940
a74a4b97
CM		 1941 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
1942 "btrfs-cleaner");
57506d50 1943 if (IS_ERR(fs_info->cleaner_kthread))
1b1d1f66 1944 goto fail_block_groups;
a74a4b97
CM		 1945
1946 fs_info->transaction_kthread = kthread_run(transaction_kthread,
1947 tree_root,
1948 "btrfs-transaction");
57506d50 1949 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 1950 goto fail_cleaner;
a74a4b97 1951
c289811c
CM		 1952 if (!btrfs_test_opt(tree_root, SSD) &&
1953 !btrfs_test_opt(tree_root, NOSSD) &&
1954 !fs_info->fs_devices->rotating) {
1955 printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
1956 "mode\n");
1957 btrfs_set_opt(fs_info->mount_opt, SSD);
1958 }
1959
e02119d5 1960 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM		 1961 u64 bytenr = btrfs_super_log_root(disk_super);
1962
7c2ca468 1963 if (fs_devices->rw_devices == 0) {
d397712b
CM		 1964 printk(KERN_WARNING "Btrfs log replay required "
1965 "on RO media\n");
7c2ca468
CM		 1966 err = -EIO;
1967 goto fail_trans_kthread;
1968 }
e02119d5
CM		 1969 blocksize =
1970 btrfs_level_size(tree_root,
1971 btrfs_super_log_root_level(disk_super));
d18a2c44 1972
676e4c86
DC		 1973 log_tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
1974 if (!log_tree_root) {
1975 err = -ENOMEM;
1976 goto fail_trans_kthread;
1977 }
e02119d5
CM		 1978
1979 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1980 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1981
1982 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 1983 blocksize,
1984 generation + 1);
e02119d5
CM		 1985 ret = btrfs_recover_log_trees(log_tree_root);
1986 BUG_ON(ret);
e556ce2c
YZ		 1987
1988 if (sb->s_flags & MS_RDONLY) {
1989 ret = btrfs_commit_super(tree_root);
1990 BUG_ON(ret);
1991 }
e02119d5 1992 }
1a40e23b 1993
76dda93c
YZ		 1994 ret = btrfs_find_orphan_roots(tree_root);
1995 BUG_ON(ret);
1996
7c2ca468 1997 if (!(sb->s_flags & MS_RDONLY)) {
d68fc57b
YZ		 1998 ret = btrfs_cleanup_fs_roots(fs_info);
1999 BUG_ON(ret);
2000
5d4f98a2 2001 ret = btrfs_recover_relocation(tree_root);
d7ce5843
MX		 2002 if (ret < 0) {
2003 printk(KERN_WARNING
2004 "btrfs: failed to recover relocation\n");
2005 err = -EINVAL;
2006 goto fail_trans_kthread;
2007 }
7c2ca468 2008 }
1a40e23b 2009
3de4586c
CM		 2010 location.objectid = BTRFS_FS_TREE_OBJECTID;
2011 location.type = BTRFS_ROOT_ITEM_KEY;
2012 location.offset = (u64)-1;
2013
3de4586c
CM		 2014 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
2015 if (!fs_info->fs_root)
7c2ca468 2016 goto fail_trans_kthread;
3140c9a3
DC		 2017 if (IS_ERR(fs_info->fs_root)) {
2018 err = PTR_ERR(fs_info->fs_root);
2019 goto fail_trans_kthread;
2020 }
c289811c 2021
e3acc2a6
JB		 2022 if (!(sb->s_flags & MS_RDONLY)) {
2023 down_read(&fs_info->cleanup_work_sem);
2024 btrfs_orphan_cleanup(fs_info->fs_root);
0af3d00b 2025 btrfs_orphan_cleanup(fs_info->tree_root);
e3acc2a6
JB		 2026 up_read(&fs_info->cleanup_work_sem);
2027 }
2028
0f7d52f4 2029 return tree_root;
39279cc3 2030
7c2ca468
CM		 2031fail_trans_kthread:
2032 kthread_stop(fs_info->transaction_kthread);
3f157a2f 2033fail_cleaner:
a74a4b97 2034 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 2035
2036 /*
2037 * make sure we're done with the btree inode before we stop our
2038 * kthreads
2039 */
2040 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
2041 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
2042
1b1d1f66
JB		 2043fail_block_groups:
2044 btrfs_free_block_groups(fs_info);
d20f7043 2045 free_extent_buffer(csum_root->node);
5d4f98a2
YZ		 2046 free_extent_buffer(csum_root->commit_root);
2047fail_dev_root:
2048 free_extent_buffer(dev_root->node);
2049 free_extent_buffer(dev_root->commit_root);
0b86a832
CM		 2050fail_extent_root:
2051 free_extent_buffer(extent_root->node);
5d4f98a2 2052 free_extent_buffer(extent_root->commit_root);
39279cc3 2053fail_tree_root:
5f39d397 2054 free_extent_buffer(tree_root->node);
5d4f98a2 2055 free_extent_buffer(tree_root->commit_root);
2b82032c
YZ		 2056fail_chunk_root:
2057 free_extent_buffer(chunk_root->node);
5d4f98a2 2058 free_extent_buffer(chunk_root->commit_root);
39279cc3 2059fail_sb_buffer:
61d92c32 2060 btrfs_stop_workers(&fs_info->generic_worker);
247e743c 2061 btrfs_stop_workers(&fs_info->fixup_workers);
771ed689 2062 btrfs_stop_workers(&fs_info->delalloc_workers);
8b712842
CM		 2063 btrfs_stop_workers(&fs_info->workers);
2064 btrfs_stop_workers(&fs_info->endio_workers);
d20f7043 2065 btrfs_stop_workers(&fs_info->endio_meta_workers);
cad321ad 2066 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
e6dcd2dc 2067 btrfs_stop_workers(&fs_info->endio_write_workers);
0cb59c99 2068 btrfs_stop_workers(&fs_info->endio_freespace_worker);
1cc127b5 2069 btrfs_stop_workers(&fs_info->submit_workers);
4543df7e 2070fail_iput:
7c2ca468 2071 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
4543df7e 2072 iput(fs_info->btree_inode);
7e662854 2073
dfe25020 2074 btrfs_close_devices(fs_info->fs_devices);
84eed90f 2075 btrfs_mapping_tree_free(&fs_info->mapping_tree);
ad081f14 2076fail_bdi:
7e662854 2077 bdi_destroy(&fs_info->bdi);
76dda93c
YZ		 2078fail_srcu:
2079 cleanup_srcu_struct(&fs_info->subvol_srcu);
7e662854 2080fail:
39279cc3
CM		 2081 kfree(extent_root);
2082 kfree(tree_root);
2083 kfree(fs_info);
83afeac4
JM		 2084 kfree(chunk_root);
2085 kfree(dev_root);
d20f7043 2086 kfree(csum_root);
39279cc3 2087 return ERR_PTR(err);
eb60ceac
CM		 2088}
2089
f2984462
CM		 2090static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
2091{
2092 char b[BDEVNAME_SIZE];
2093
2094 if (uptodate) {
2095 set_buffer_uptodate(bh);
2096 } else {
2097 if (!buffer_eopnotsupp(bh) && printk_ratelimit()) {
2098 printk(KERN_WARNING "lost page write due to "
2099 "I/O error on %s\n",
2100 bdevname(bh->b_bdev, b));
2101 }
1259ab75
CM		 2102 /* note, we dont' set_buffer_write_io_error because we have
2103 * our own ways of dealing with the IO errors
2104 */
f2984462
CM		 2105 clear_buffer_uptodate(bh);
2106 }
2107 unlock_buffer(bh);
2108 put_bh(bh);
2109}
2110
a512bbf8
YZ		 2111struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
2112{
2113 struct buffer_head *bh;
2114 struct buffer_head *latest = NULL;
2115 struct btrfs_super_block *super;
2116 int i;
2117 u64 transid = 0;
2118 u64 bytenr;
2119
2120 /* we would like to check all the supers, but that would make
2121 * a btrfs mount succeed after a mkfs from a different FS.
2122 * So, we need to add a special mount option to scan for
2123 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2124 */
2125 for (i = 0; i < 1; i++) {
2126 bytenr = btrfs_sb_offset(i);
2127 if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
2128 break;
2129 bh = __bread(bdev, bytenr / 4096, 4096);
2130 if (!bh)
2131 continue;
2132
2133 super = (struct btrfs_super_block *)bh->b_data;
2134 if (btrfs_super_bytenr(super) != bytenr ||
2135 strncmp((char *)(&super->magic), BTRFS_MAGIC,
2136 sizeof(super->magic))) {
2137 brelse(bh);
2138 continue;
2139 }
2140
2141 if (!latest || btrfs_super_generation(super) > transid) {
2142 brelse(latest);
2143 latest = bh;
2144 transid = btrfs_super_generation(super);
2145 } else {
2146 brelse(bh);
2147 }
2148 }
2149 return latest;
2150}
2151
4eedeb75
HH		 2152/*
2153 * this should be called twice, once with wait == 0 and
2154 * once with wait == 1. When wait == 0 is done, all the buffer heads
2155 * we write are pinned.
2156 *
2157 * They are released when wait == 1 is done.
2158 * max_mirrors must be the same for both runs, and it indicates how
2159 * many supers on this one device should be written.
2160 *
2161 * max_mirrors == 0 means to write them all.
2162 */
a512bbf8
YZ		 2163static int write_dev_supers(struct btrfs_device *device,
2164 struct btrfs_super_block *sb,
2165 int do_barriers, int wait, int max_mirrors)
2166{
2167 struct buffer_head *bh;
2168 int i;
2169 int ret;
2170 int errors = 0;
2171 u32 crc;
2172 u64 bytenr;
2173 int last_barrier = 0;
2174
2175 if (max_mirrors == 0)
2176 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
2177
2178 /* make sure only the last submit_bh does a barrier */
2179 if (do_barriers) {
2180 for (i = 0; i < max_mirrors; i++) {
2181 bytenr = btrfs_sb_offset(i);
2182 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
2183 device->total_bytes)
2184 break;
2185 last_barrier = i;
2186 }
2187 }
2188
2189 for (i = 0; i < max_mirrors; i++) {
2190 bytenr = btrfs_sb_offset(i);
2191 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
2192 break;
2193
2194 if (wait) {
2195 bh = __find_get_block(device->bdev, bytenr / 4096,
2196 BTRFS_SUPER_INFO_SIZE);
2197 BUG_ON(!bh);
a512bbf8 2198 wait_on_buffer(bh);
4eedeb75
HH		 2199 if (!buffer_uptodate(bh))
2200 errors++;
2201
2202 /* drop our reference */
2203 brelse(bh);
2204
2205 /* drop the reference from the wait == 0 run */
2206 brelse(bh);
2207 continue;
a512bbf8
YZ		 2208 } else {
2209 btrfs_set_super_bytenr(sb, bytenr);
2210
2211 crc = ~(u32)0;
2212 crc = btrfs_csum_data(NULL, (char *)sb +
2213 BTRFS_CSUM_SIZE, crc,
2214 BTRFS_SUPER_INFO_SIZE -
2215 BTRFS_CSUM_SIZE);
2216 btrfs_csum_final(crc, sb->csum);
2217
4eedeb75
HH		 2218 /*
2219 * one reference for us, and we leave it for the
2220 * caller
2221 */
a512bbf8
YZ		 2222 bh = __getblk(device->bdev, bytenr / 4096,
2223 BTRFS_SUPER_INFO_SIZE);
2224 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
2225
4eedeb75 2226 /* one reference for submit_bh */
a512bbf8 2227 get_bh(bh);
4eedeb75
HH		 2228
2229 set_buffer_uptodate(bh);
a512bbf8
YZ		 2230 lock_buffer(bh);
2231 bh->b_end_io = btrfs_end_buffer_write_sync;
2232 }
2233
2234 if (i == last_barrier && do_barriers && device->barriers) {
2235 ret = submit_bh(WRITE_BARRIER, bh);
2236 if (ret == -EOPNOTSUPP) {
2237 printk("btrfs: disabling barriers on dev %s\n",
2238 device->name);
2239 set_buffer_uptodate(bh);
2240 device->barriers = 0;
4eedeb75 2241 /* one reference for submit_bh */
a512bbf8
YZ		 2242 get_bh(bh);
2243 lock_buffer(bh);
ffbd517d 2244 ret = submit_bh(WRITE_SYNC, bh);
a512bbf8
YZ		 2245 }
2246 } else {
ffbd517d 2247 ret = submit_bh(WRITE_SYNC, bh);
a512bbf8
YZ		 2248 }
2249
4eedeb75 2250 if (ret)
a512bbf8 2251 errors++;
a512bbf8
YZ		 2252 }
2253 return errors < i ? 0 : -1;
2254}
2255
2256int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 2257{
e5e9a520 2258 struct list_head *head;
f2984462 2259 struct btrfs_device *dev;
a061fc8d 2260 struct btrfs_super_block *sb;
f2984462 2261 struct btrfs_dev_item *dev_item;
f2984462
CM		 2262 int ret;
2263 int do_barriers;
a236aed1
CM		 2264 int max_errors;
2265 int total_errors = 0;
a061fc8d 2266 u64 flags;
f2984462 2267
a236aed1 2268 max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
f2984462
CM		 2269 do_barriers = !btrfs_test_opt(root, NOBARRIER);
2270
a061fc8d
CM		 2271 sb = &root->fs_info->super_for_commit;
2272 dev_item = &sb->dev_item;
e5e9a520
CM		 2273
2274 mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
2275 head = &root->fs_info->fs_devices->devices;
c6e30871 2276 list_for_each_entry(dev, head, dev_list) {
dfe25020
CM		 2277 if (!dev->bdev) {
2278 total_errors++;
2279 continue;
2280 }
2b82032c 2281 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 2282 continue;
2283
2b82032c 2284 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 2285 btrfs_set_stack_device_type(dev_item, dev->type);
2286 btrfs_set_stack_device_id(dev_item, dev->devid);
2287 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
2288 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
2289 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
2290 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
2291 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
2292 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 2293 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 2294
a061fc8d
CM		 2295 flags = btrfs_super_flags(sb);
2296 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
2297
a512bbf8 2298 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 2299 if (ret)
2300 total_errors++;
f2984462 2301 }
a236aed1 2302 if (total_errors > max_errors) {
d397712b
CM		 2303 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2304 total_errors);
a236aed1
CM		 2305 BUG();
2306 }
f2984462 2307
a512bbf8 2308 total_errors = 0;
c6e30871 2309 list_for_each_entry(dev, head, dev_list) {
dfe25020
CM		 2310 if (!dev->bdev)
2311 continue;
2b82032c 2312 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 2313 continue;
2314
a512bbf8
YZ		 2315 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
2316 if (ret)
2317 total_errors++;
f2984462 2318 }
e5e9a520 2319 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
a236aed1 2320 if (total_errors > max_errors) {
d397712b
CM		 2321 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2322 total_errors);
a236aed1
CM		 2323 BUG();
2324 }
f2984462
CM		 2325 return 0;
2326}
2327
a512bbf8
YZ		 2328int write_ctree_super(struct btrfs_trans_handle *trans,
2329 struct btrfs_root *root, int max_mirrors)
eb60ceac 2330{
e66f709b 2331 int ret;
5f39d397 2332
a512bbf8 2333 ret = write_all_supers(root, max_mirrors);
5f39d397 2334 return ret;
cfaa7295
CM		 2335}
2336
5eda7b5e 2337int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
2619ba1f 2338{
4df27c4d 2339 spin_lock(&fs_info->fs_roots_radix_lock);
2619ba1f
CM		 2340 radix_tree_delete(&fs_info->fs_roots_radix,
2341 (unsigned long)root->root_key.objectid);
4df27c4d 2342 spin_unlock(&fs_info->fs_roots_radix_lock);
76dda93c
YZ		 2343
2344 if (btrfs_root_refs(&root->root_item) == 0)
2345 synchronize_srcu(&fs_info->subvol_srcu);
2346
4df27c4d
YZ		 2347 free_fs_root(root);
2348 return 0;
2349}
2350
2351static void free_fs_root(struct btrfs_root *root)
2352{
2353 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
3394e160
CM		 2354 if (root->anon_super.s_dev) {
2355 down_write(&root->anon_super.s_umount);
2356 kill_anon_super(&root->anon_super);
2357 }
4df27c4d
YZ		 2358 free_extent_buffer(root->node);
2359 free_extent_buffer(root->commit_root);
d397712b 2360 kfree(root->name);
2619ba1f 2361 kfree(root);
2619ba1f
CM		 2362}
2363
35b7e476 2364static int del_fs_roots(struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 2365{
2366 int ret;
2367 struct btrfs_root *gang[8];
2368 int i;
2369
76dda93c
YZ		 2370 while (!list_empty(&fs_info->dead_roots)) {
2371 gang[0] = list_entry(fs_info->dead_roots.next,
2372 struct btrfs_root, root_list);
2373 list_del(&gang[0]->root_list);
2374
2375 if (gang[0]->in_radix) {
2376 btrfs_free_fs_root(fs_info, gang[0]);
2377 } else {
2378 free_extent_buffer(gang[0]->node);
2379 free_extent_buffer(gang[0]->commit_root);
2380 kfree(gang[0]);
2381 }
2382 }
2383
d397712b 2384 while (1) {
0f7d52f4
CM		 2385 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2386 (void **)gang, 0,
2387 ARRAY_SIZE(gang));
2388 if (!ret)
2389 break;
2619ba1f 2390 for (i = 0; i < ret; i++)
5eda7b5e 2391 btrfs_free_fs_root(fs_info, gang[i]);
0f7d52f4
CM		 2392 }
2393 return 0;
2394}
b4100d64 2395
c146afad 2396int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 2397{
c146afad
YZ		 2398 u64 root_objectid = 0;
2399 struct btrfs_root *gang[8];
2400 int i;
3768f368 2401 int ret;
e089f05c 2402
c146afad
YZ		 2403 while (1) {
2404 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2405 (void **)gang, root_objectid,
2406 ARRAY_SIZE(gang));
2407 if (!ret)
2408 break;
5d4f98a2
YZ		 2409
2410 root_objectid = gang[ret - 1]->root_key.objectid + 1;
c146afad
YZ		 2411 for (i = 0; i < ret; i++) {
2412 root_objectid = gang[i]->root_key.objectid;
c146afad
YZ		 2413 btrfs_orphan_cleanup(gang[i]);
2414 }
2415 root_objectid++;
2416 }
2417 return 0;
2418}
a2135011 2419
c146afad
YZ		 2420int btrfs_commit_super(struct btrfs_root *root)
2421{
2422 struct btrfs_trans_handle *trans;
2423 int ret;
a74a4b97 2424
c146afad 2425 mutex_lock(&root->fs_info->cleaner_mutex);
24bbcf04 2426 btrfs_run_delayed_iputs(root);
a74a4b97 2427 btrfs_clean_old_snapshots(root);
c146afad 2428 mutex_unlock(&root->fs_info->cleaner_mutex);
c71bf099
YZ		 2429
2430 /* wait until ongoing cleanup work done */
2431 down_write(&root->fs_info->cleanup_work_sem);
2432 up_write(&root->fs_info->cleanup_work_sem);
2433
a22285a6 2434 trans = btrfs_join_transaction(root, 1);
54aa1f4d 2435 ret = btrfs_commit_transaction(trans, root);
c146afad
YZ		 2436 BUG_ON(ret);
2437 /* run commit again to drop the original snapshot */
a22285a6 2438 trans = btrfs_join_transaction(root, 1);
79154b1b
CM		 2439 btrfs_commit_transaction(trans, root);
2440 ret = btrfs_write_and_wait_transaction(NULL, root);
3768f368 2441 BUG_ON(ret);
d6bfde87 2442
a512bbf8 2443 ret = write_ctree_super(NULL, root, 0);
c146afad
YZ		 2444 return ret;
2445}
2446
2447int close_ctree(struct btrfs_root *root)
2448{
2449 struct btrfs_fs_info *fs_info = root->fs_info;
2450 int ret;
2451
2452 fs_info->closing = 1;
2453 smp_mb();
2454
0af3d00b 2455 btrfs_put_block_group_cache(fs_info);
c146afad
YZ		 2456 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
2457 ret = btrfs_commit_super(root);
d397712b
CM		 2458 if (ret)
2459 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
c146afad 2460 }
0f7d52f4 2461
8929ecfa
YZ		 2462 kthread_stop(root->fs_info->transaction_kthread);
2463 kthread_stop(root->fs_info->cleaner_kthread);
2464
f25784b3
YZ		 2465 fs_info->closing = 2;
2466 smp_mb();
2467
b0c68f8b 2468 if (fs_info->delalloc_bytes) {
d397712b 2469 printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
21380931 2470 (unsigned long long)fs_info->delalloc_bytes);
b0c68f8b 2471 }
31153d81 2472 if (fs_info->total_ref_cache_size) {
d397712b
CM		 2473 printk(KERN_INFO "btrfs: at umount reference cache size %llu\n",
2474 (unsigned long long)fs_info->total_ref_cache_size);
31153d81 2475 }
bcc63abb 2476
5d4f98a2
YZ		 2477 free_extent_buffer(fs_info->extent_root->node);
2478 free_extent_buffer(fs_info->extent_root->commit_root);
2479 free_extent_buffer(fs_info->tree_root->node);
2480 free_extent_buffer(fs_info->tree_root->commit_root);
2481 free_extent_buffer(root->fs_info->chunk_root->node);
2482 free_extent_buffer(root->fs_info->chunk_root->commit_root);
2483 free_extent_buffer(root->fs_info->dev_root->node);
2484 free_extent_buffer(root->fs_info->dev_root->commit_root);
2485 free_extent_buffer(root->fs_info->csum_root->node);
2486 free_extent_buffer(root->fs_info->csum_root->commit_root);
d20f7043 2487
9078a3e1 2488 btrfs_free_block_groups(root->fs_info);
d10c5f31 2489
c146afad 2490 del_fs_roots(fs_info);
d10c5f31 2491
c146afad 2492 iput(fs_info->btree_inode);
9ad6b7bc 2493
61d92c32 2494 btrfs_stop_workers(&fs_info->generic_worker);
247e743c 2495 btrfs_stop_workers(&fs_info->fixup_workers);
771ed689 2496 btrfs_stop_workers(&fs_info->delalloc_workers);
8b712842
CM		 2497 btrfs_stop_workers(&fs_info->workers);
2498 btrfs_stop_workers(&fs_info->endio_workers);
d20f7043 2499 btrfs_stop_workers(&fs_info->endio_meta_workers);
cad321ad 2500 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
e6dcd2dc 2501 btrfs_stop_workers(&fs_info->endio_write_workers);
0cb59c99 2502 btrfs_stop_workers(&fs_info->endio_freespace_worker);
1cc127b5 2503 btrfs_stop_workers(&fs_info->submit_workers);
d6bfde87 2504
dfe25020 2505 btrfs_close_devices(fs_info->fs_devices);
0b86a832 2506 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 2507
04160088 2508 bdi_destroy(&fs_info->bdi);
76dda93c 2509 cleanup_srcu_struct(&fs_info->subvol_srcu);
0b86a832 2510
0f7d52f4 2511 kfree(fs_info->extent_root);
0f7d52f4 2512 kfree(fs_info->tree_root);
0b86a832
CM		 2513 kfree(fs_info->chunk_root);
2514 kfree(fs_info->dev_root);
d20f7043 2515 kfree(fs_info->csum_root);
eb60ceac
CM		 2516 return 0;
2517}
2518
1259ab75 2519int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
5f39d397 2520{
1259ab75 2521 int ret;
810191ff 2522 struct inode *btree_inode = buf->first_page->mapping->host;
1259ab75 2523
2ac55d41
JB		 2524 ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf,
2525 NULL);
1259ab75
CM		 2526 if (!ret)
2527 return ret;
2528
2529 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
2530 parent_transid);
2531 return !ret;
5f39d397
CM		 2532}
2533
2534int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 2535{
810191ff 2536 struct inode *btree_inode = buf->first_page->mapping->host;
d1310b2e 2537 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
5f39d397
CM		 2538 buf);
2539}
6702ed49 2540
5f39d397
CM		 2541void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
2542{
810191ff 2543 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
5f39d397
CM		 2544 u64 transid = btrfs_header_generation(buf);
2545 struct inode *btree_inode = root->fs_info->btree_inode;
b9473439 2546 int was_dirty;
b4ce94de 2547
b9447ef8 2548 btrfs_assert_tree_locked(buf);
ccd467d6 2549 if (transid != root->fs_info->generation) {
d397712b
CM		 2550 printk(KERN_CRIT "btrfs transid mismatch buffer %llu, "
2551 "found %llu running %llu\n",
db94535d 2552 (unsigned long long)buf->start,
d397712b
CM		 2553 (unsigned long long)transid,
2554 (unsigned long long)root->fs_info->generation);
ccd467d6
CM		 2555 WARN_ON(1);
2556 }
b9473439
CM		 2557 was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
2558 buf);
2559 if (!was_dirty) {
2560 spin_lock(&root->fs_info->delalloc_lock);
2561 root->fs_info->dirty_metadata_bytes += buf->len;
2562 spin_unlock(&root->fs_info->delalloc_lock);
2563 }
eb60ceac
CM		 2564}
2565
d3c2fdcf 2566void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
35b7e476 2567{
188de649
CM		 2568 /*
2569 * looks as though older kernels can get into trouble with
2570 * this code, they end up stuck in balance_dirty_pages forever
2571 */
d6bfde87 2572 u64 num_dirty;
771ed689 2573 unsigned long thresh = 32 * 1024 * 1024;
d6bfde87 2574
6933c02e 2575 if (current->flags & PF_MEMALLOC)
d6bfde87
CM		 2576 return;
2577
585ad2c3
CM		 2578 num_dirty = root->fs_info->dirty_metadata_bytes;
2579
d6bfde87
CM		 2580 if (num_dirty > thresh) {
2581 balance_dirty_pages_ratelimited_nr(
d7fc640e 2582 root->fs_info->btree_inode->i_mapping, 1);
d6bfde87 2583 }
188de649 2584 return;
35b7e476 2585}
6b80053d 2586
ca7a79ad 2587int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 2588{
810191ff 2589 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
ce9adaa5 2590 int ret;
ca7a79ad 2591 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
d397712b 2592 if (ret == 0)
b4ce94de 2593 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
ce9adaa5 2594 return ret;
6b80053d 2595}
0da5468f 2596
4bef0848
CM		 2597int btree_lock_page_hook(struct page *page)
2598{
2599 struct inode *inode = page->mapping->host;
b9473439 2600 struct btrfs_root *root = BTRFS_I(inode)->root;
4bef0848
CM		 2601 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2602 struct extent_buffer *eb;
2603 unsigned long len;
2604 u64 bytenr = page_offset(page);
2605
2606 if (page->private == EXTENT_PAGE_PRIVATE)
2607 goto out;
2608
2609 len = page->private >> 2;
2610 eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS);
2611 if (!eb)
2612 goto out;
2613
2614 btrfs_tree_lock(eb);
4bef0848 2615 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
b9473439
CM		 2616
2617 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
2618 spin_lock(&root->fs_info->delalloc_lock);
2619 if (root->fs_info->dirty_metadata_bytes >= eb->len)
2620 root->fs_info->dirty_metadata_bytes -= eb->len;
2621 else
2622 WARN_ON(1);
2623 spin_unlock(&root->fs_info->delalloc_lock);
2624 }
2625
4bef0848
CM		 2626 btrfs_tree_unlock(eb);
2627 free_extent_buffer(eb);
2628out:
2629 lock_page(page);
2630 return 0;
2631}
2632
d1310b2e 2633static struct extent_io_ops btree_extent_io_ops = {
4bef0848 2634 .write_cache_pages_lock_hook = btree_lock_page_hook,
ce9adaa5 2635 .readpage_end_io_hook = btree_readpage_end_io_hook,
0b86a832 2636 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 2637 /* note we're sharing with inode.c for the merge bio hook */
2638 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 2639};
Linux 6.x block layer and io_uring tree(s)