projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Linux 2.6.30-rc2
[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
CM		 28#include <linux/freezer.h>
29#include "compat.h"
a5eb62e3 30#include "crc32c.h"
eb60ceac
CM		 31#include "ctree.h"
32#include "disk-io.h"
e089f05c 33#include "transaction.h"
0f7d52f4 34#include "btrfs_inode.h"
0b86a832 35#include "volumes.h"
db94535d 36#include "print-tree.h"
8b712842 37#include "async-thread.h"
925baedd 38#include "locking.h"
017e5369 39#include "ref-cache.h"
e02119d5 40#include "tree-log.h"
fa9c0d79 41#include "free-space-cache.h"
eb60ceac 42
d1310b2e 43static struct extent_io_ops btree_extent_io_ops;
8b712842 44static void end_workqueue_fn(struct btrfs_work *work);
ce9adaa5 45
d352ac68
CM		 46/*
47 * end_io_wq structs are used to do processing in task context when an IO is
48 * complete. This is used during reads to verify checksums, and it is used
49 * by writes to insert metadata for new file extents after IO is complete.
50 */
ce9adaa5
CM		 51struct end_io_wq {
52 struct bio *bio;
53 bio_end_io_t *end_io;
54 void *private;
55 struct btrfs_fs_info *info;
56 int error;
22c59948 57 int metadata;
ce9adaa5 58 struct list_head list;
8b712842 59 struct btrfs_work work;
ce9adaa5 60};
0da5468f 61
d352ac68
CM		 62/*
63 * async submit bios are used to offload expensive checksumming
64 * onto the worker threads. They checksum file and metadata bios
65 * just before they are sent down the IO stack.
66 */
44b8bd7e
CM		 67struct async_submit_bio {
68 struct inode *inode;
69 struct bio *bio;
70 struct list_head list;
4a69a410
CM		 71 extent_submit_bio_hook_t *submit_bio_start;
72 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM		 73 int rw;
74 int mirror_num;
c8b97818 75 unsigned long bio_flags;
8b712842 76 struct btrfs_work work;
44b8bd7e
CM		 77};
78
4008c04a
CM		 79/* These are used to set the lockdep class on the extent buffer locks.
80 * The class is set by the readpage_end_io_hook after the buffer has
81 * passed csum validation but before the pages are unlocked.
82 *
83 * The lockdep class is also set by btrfs_init_new_buffer on freshly
84 * allocated blocks.
85 *
86 * The class is based on the level in the tree block, which allows lockdep
87 * to know that lower nodes nest inside the locks of higher nodes.
88 *
89 * We also add a check to make sure the highest level of the tree is
90 * the same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this
91 * code needs update as well.
92 */
93#ifdef CONFIG_DEBUG_LOCK_ALLOC
94# if BTRFS_MAX_LEVEL != 8
95# error
96# endif
97static struct lock_class_key btrfs_eb_class[BTRFS_MAX_LEVEL + 1];
98static const char *btrfs_eb_name[BTRFS_MAX_LEVEL + 1] = {
99 /* leaf */
100 "btrfs-extent-00",
101 "btrfs-extent-01",
102 "btrfs-extent-02",
103 "btrfs-extent-03",
104 "btrfs-extent-04",
105 "btrfs-extent-05",
106 "btrfs-extent-06",
107 "btrfs-extent-07",
108 /* highest possible level */
109 "btrfs-extent-08",
110};
111#endif
112
d352ac68
CM		 113/*
114 * extents on the btree inode are pretty simple, there's one extent
115 * that covers the entire device
116 */
b2950863
CH		 117static struct extent_map *btree_get_extent(struct inode *inode,
118 struct page *page, size_t page_offset, u64 start, u64 len,
119 int create)
7eccb903 120{
5f39d397
CM		 121 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
122 struct extent_map *em;
123 int ret;
124
d1310b2e
CM		 125 spin_lock(&em_tree->lock);
126 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 127 if (em) {
128 em->bdev =
129 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
130 spin_unlock(&em_tree->lock);
5f39d397 131 goto out;
a061fc8d
CM		 132 }
133 spin_unlock(&em_tree->lock);
7b13b7b1 134
5f39d397
CM		 135 em = alloc_extent_map(GFP_NOFS);
136 if (!em) {
137 em = ERR_PTR(-ENOMEM);
138 goto out;
139 }
140 em->start = 0;
0afbaf8c 141 em->len = (u64)-1;
c8b97818 142 em->block_len = (u64)-1;
5f39d397 143 em->block_start = 0;
a061fc8d 144 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e
CM		 145
146 spin_lock(&em_tree->lock);
5f39d397
CM		 147 ret = add_extent_mapping(em_tree, em);
148 if (ret == -EEXIST) {
0afbaf8c
CM		 149 u64 failed_start = em->start;
150 u64 failed_len = em->len;
151
5f39d397 152 free_extent_map(em);
7b13b7b1 153 em = lookup_extent_mapping(em_tree, start, len);
0afbaf8c 154 if (em) {
7b13b7b1 155 ret = 0;
0afbaf8c
CM		 156 } else {
157 em = lookup_extent_mapping(em_tree, failed_start,
158 failed_len);
7b13b7b1 159 ret = -EIO;
0afbaf8c 160 }
5f39d397 161 } else if (ret) {
7b13b7b1
CM		 162 free_extent_map(em);
163 em = NULL;
5f39d397 164 }
7b13b7b1
CM		 165 spin_unlock(&em_tree->lock);
166
167 if (ret)
168 em = ERR_PTR(ret);
5f39d397
CM		 169out:
170 return em;
7eccb903
CM		 171}
172
19c00ddc
CM		 173u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
174{
a5eb62e3 175 return btrfs_crc32c(seed, data, len);
19c00ddc
CM		 176}
177
178void btrfs_csum_final(u32 crc, char *result)
179{
180 *(__le32 *)result = ~cpu_to_le32(crc);
181}
182
d352ac68
CM		 183/*
184 * compute the csum for a btree block, and either verify it or write it
185 * into the csum field of the block.
186 */
19c00ddc
CM		 187static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
188 int verify)
189{
607d432d
JB		 190 u16 csum_size =
191 btrfs_super_csum_size(&root->fs_info->super_copy);
192 char *result = NULL;
19c00ddc
CM		 193 unsigned long len;
194 unsigned long cur_len;
195 unsigned long offset = BTRFS_CSUM_SIZE;
196 char *map_token = NULL;
197 char *kaddr;
198 unsigned long map_start;
199 unsigned long map_len;
200 int err;
201 u32 crc = ~(u32)0;
607d432d 202 unsigned long inline_result;
19c00ddc
CM		 203
204 len = buf->len - offset;
d397712b 205 while (len > 0) {
19c00ddc
CM		 206 err = map_private_extent_buffer(buf, offset, 32,
207 &map_token, &kaddr,
208 &map_start, &map_len, KM_USER0);
d397712b 209 if (err)
19c00ddc 210 return 1;
19c00ddc
CM		 211 cur_len = min(len, map_len - (offset - map_start));
212 crc = btrfs_csum_data(root, kaddr + offset - map_start,
213 crc, cur_len);
214 len -= cur_len;
215 offset += cur_len;
216 unmap_extent_buffer(buf, map_token, KM_USER0);
217 }
607d432d
JB		 218 if (csum_size > sizeof(inline_result)) {
219 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
220 if (!result)
221 return 1;
222 } else {
223 result = (char *)&inline_result;
224 }
225
19c00ddc
CM		 226 btrfs_csum_final(crc, result);
227
228 if (verify) {
607d432d 229 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 230 u32 val;
231 u32 found = 0;
607d432d 232 memcpy(&found, result, csum_size);
e4204ded 233
607d432d 234 read_extent_buffer(buf, &val, 0, csum_size);
d397712b
CM		 235 printk(KERN_INFO "btrfs: %s checksum verify failed "
236 "on %llu wanted %X found %X level %d\n",
19c00ddc 237 root->fs_info->sb->s_id,
2dd3e67b 238 buf->start, val, found, btrfs_header_level(buf));
607d432d
JB		 239 if (result != (char *)&inline_result)
240 kfree(result);
19c00ddc
CM		 241 return 1;
242 }
243 } else {
607d432d 244 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 245 }
607d432d
JB		 246 if (result != (char *)&inline_result)
247 kfree(result);
19c00ddc
CM		 248 return 0;
249}
250
d352ac68
CM		 251/*
252 * we can't consider a given block up to date unless the transid of the
253 * block matches the transid in the parent node's pointer. This is how we
254 * detect blocks that either didn't get written at all or got written
255 * in the wrong place.
256 */
1259ab75
CM		 257static int verify_parent_transid(struct extent_io_tree *io_tree,
258 struct extent_buffer *eb, u64 parent_transid)
259{
260 int ret;
261
262 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
263 return 0;
264
265 lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS);
266 if (extent_buffer_uptodate(io_tree, eb) &&
267 btrfs_header_generation(eb) == parent_transid) {
268 ret = 0;
269 goto out;
270 }
271 printk("parent transid verify failed on %llu wanted %llu found %llu\n",
272 (unsigned long long)eb->start,
273 (unsigned long long)parent_transid,
274 (unsigned long long)btrfs_header_generation(eb));
275 ret = 1;
1259ab75 276 clear_extent_buffer_uptodate(io_tree, eb);
33958dc6 277out:
1259ab75
CM		 278 unlock_extent(io_tree, eb->start, eb->start + eb->len - 1,
279 GFP_NOFS);
280 return ret;
1259ab75
CM		 281}
282
d352ac68
CM		 283/*
284 * helper to read a given tree block, doing retries as required when
285 * the checksums don't match and we have alternate mirrors to try.
286 */
f188591e
CM		 287static int btree_read_extent_buffer_pages(struct btrfs_root *root,
288 struct extent_buffer *eb,
ca7a79ad 289 u64 start, u64 parent_transid)
f188591e
CM		 290{
291 struct extent_io_tree *io_tree;
292 int ret;
293 int num_copies = 0;
294 int mirror_num = 0;
295
296 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
297 while (1) {
298 ret = read_extent_buffer_pages(io_tree, eb, start, 1,
299 btree_get_extent, mirror_num);
1259ab75
CM		 300 if (!ret &&
301 !verify_parent_transid(io_tree, eb, parent_transid))
f188591e 302 return ret;
d397712b 303
f188591e
CM		 304 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
305 eb->start, eb->len);
4235298e 306 if (num_copies == 1)
f188591e 307 return ret;
4235298e 308
f188591e 309 mirror_num++;
4235298e 310 if (mirror_num > num_copies)
f188591e 311 return ret;
f188591e 312 }
f188591e
CM		 313 return -EIO;
314}
19c00ddc 315
d352ac68 316/*
d397712b
CM		 317 * checksum a dirty tree block before IO. This has extra checks to make sure
318 * we only fill in the checksum field in the first page of a multi-page block
d352ac68 319 */
d397712b 320
b2950863 321static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
19c00ddc 322{
d1310b2e 323 struct extent_io_tree *tree;
35ebb934 324 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
19c00ddc
CM		 325 u64 found_start;
326 int found_level;
327 unsigned long len;
328 struct extent_buffer *eb;
f188591e
CM		 329 int ret;
330
d1310b2e 331 tree = &BTRFS_I(page->mapping->host)->io_tree;
19c00ddc
CM		 332
333 if (page->private == EXTENT_PAGE_PRIVATE)
334 goto out;
335 if (!page->private)
336 goto out;
337 len = page->private >> 2;
d397712b
CM		 338 WARN_ON(len == 0);
339
19c00ddc 340 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
ca7a79ad
CM		 341 ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
342 btrfs_header_generation(eb));
f188591e 343 BUG_ON(ret);
19c00ddc
CM		 344 found_start = btrfs_header_bytenr(eb);
345 if (found_start != start) {
55c69072
CM		 346 WARN_ON(1);
347 goto err;
348 }
349 if (eb->first_page != page) {
55c69072
CM		 350 WARN_ON(1);
351 goto err;
352 }
353 if (!PageUptodate(page)) {
55c69072
CM		 354 WARN_ON(1);
355 goto err;
19c00ddc
CM		 356 }
357 found_level = btrfs_header_level(eb);
4bef0848 358
19c00ddc 359 csum_tree_block(root, eb, 0);
55c69072 360err:
19c00ddc
CM		 361 free_extent_buffer(eb);
362out:
363 return 0;
364}
365
2b82032c
YZ		 366static int check_tree_block_fsid(struct btrfs_root *root,
367 struct extent_buffer *eb)
368{
369 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
370 u8 fsid[BTRFS_UUID_SIZE];
371 int ret = 1;
372
373 read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
374 BTRFS_FSID_SIZE);
375 while (fs_devices) {
376 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
377 ret = 0;
378 break;
379 }
380 fs_devices = fs_devices->seed;
381 }
382 return ret;
383}
384
4008c04a
CM		 385#ifdef CONFIG_DEBUG_LOCK_ALLOC
386void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level)
387{
388 lockdep_set_class_and_name(&eb->lock,
389 &btrfs_eb_class[level],
390 btrfs_eb_name[level]);
391}
392#endif
393
b2950863 394static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
ce9adaa5
CM		 395 struct extent_state *state)
396{
397 struct extent_io_tree *tree;
398 u64 found_start;
399 int found_level;
400 unsigned long len;
401 struct extent_buffer *eb;
402 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 403 int ret = 0;
ce9adaa5
CM		 404
405 tree = &BTRFS_I(page->mapping->host)->io_tree;
406 if (page->private == EXTENT_PAGE_PRIVATE)
407 goto out;
408 if (!page->private)
409 goto out;
d397712b 410
ce9adaa5 411 len = page->private >> 2;
d397712b
CM		 412 WARN_ON(len == 0);
413
ce9adaa5 414 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
f188591e 415
ce9adaa5 416 found_start = btrfs_header_bytenr(eb);
23a07867 417 if (found_start != start) {
d397712b 418 printk(KERN_INFO "btrfs bad tree block start %llu %llu\n",
a1b32a59
CM		 419 (unsigned long long)found_start,
420 (unsigned long long)eb->start);
f188591e 421 ret = -EIO;
ce9adaa5
CM		 422 goto err;
423 }
424 if (eb->first_page != page) {
d397712b
CM		 425 printk(KERN_INFO "btrfs bad first page %lu %lu\n",
426 eb->first_page->index, page->index);
ce9adaa5 427 WARN_ON(1);
f188591e 428 ret = -EIO;
ce9adaa5
CM		 429 goto err;
430 }
2b82032c 431 if (check_tree_block_fsid(root, eb)) {
d397712b
CM		 432 printk(KERN_INFO "btrfs bad fsid on block %llu\n",
433 (unsigned long long)eb->start);
1259ab75
CM		 434 ret = -EIO;
435 goto err;
436 }
ce9adaa5
CM		 437 found_level = btrfs_header_level(eb);
438
4008c04a
CM		 439 btrfs_set_buffer_lockdep_class(eb, found_level);
440
ce9adaa5 441 ret = csum_tree_block(root, eb, 1);
f188591e
CM		 442 if (ret)
443 ret = -EIO;
ce9adaa5
CM		 444
445 end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
446 end = eb->start + end - 1;
ce9adaa5
CM		 447err:
448 free_extent_buffer(eb);
449out:
f188591e 450 return ret;
ce9adaa5
CM		 451}
452
ce9adaa5 453static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM		 454{
455 struct end_io_wq *end_io_wq = bio->bi_private;
456 struct btrfs_fs_info *fs_info;
ce9adaa5 457
ce9adaa5 458 fs_info = end_io_wq->info;
ce9adaa5 459 end_io_wq->error = err;
8b712842
CM		 460 end_io_wq->work.func = end_workqueue_fn;
461 end_io_wq->work.flags = 0;
d20f7043
CM		 462
463 if (bio->bi_rw & (1 << BIO_RW)) {
cad321ad
CM		 464 if (end_io_wq->metadata)
465 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
466 &end_io_wq->work);
467 else
468 btrfs_queue_worker(&fs_info->endio_write_workers,
469 &end_io_wq->work);
d20f7043
CM		 470 } else {
471 if (end_io_wq->metadata)
472 btrfs_queue_worker(&fs_info->endio_meta_workers,
473 &end_io_wq->work);
474 else
475 btrfs_queue_worker(&fs_info->endio_workers,
476 &end_io_wq->work);
477 }
ce9adaa5
CM		 478}
479
22c59948
CM		 480int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
481 int metadata)
0b86a832 482{
ce9adaa5 483 struct end_io_wq *end_io_wq;
ce9adaa5
CM		 484 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
485 if (!end_io_wq)
486 return -ENOMEM;
487
488 end_io_wq->private = bio->bi_private;
489 end_io_wq->end_io = bio->bi_end_io;
22c59948 490 end_io_wq->info = info;
ce9adaa5
CM		 491 end_io_wq->error = 0;
492 end_io_wq->bio = bio;
22c59948 493 end_io_wq->metadata = metadata;
ce9adaa5
CM		 494
495 bio->bi_private = end_io_wq;
496 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 497 return 0;
498}
499
b64a2851 500unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 501{
4854ddd0
CM		 502 unsigned long limit = min_t(unsigned long,
503 info->workers.max_workers,
504 info->fs_devices->open_devices);
505 return 256 * limit;
506}
0986fe9e 507
4854ddd0
CM		 508int btrfs_congested_async(struct btrfs_fs_info *info, int iodone)
509{
b64a2851
CM		 510 return atomic_read(&info->nr_async_bios) >
511 btrfs_async_submit_limit(info);
0986fe9e
CM		 512}
513
4a69a410
CM		 514static void run_one_async_start(struct btrfs_work *work)
515{
516 struct btrfs_fs_info *fs_info;
517 struct async_submit_bio *async;
518
519 async = container_of(work, struct async_submit_bio, work);
520 fs_info = BTRFS_I(async->inode)->root->fs_info;
521 async->submit_bio_start(async->inode, async->rw, async->bio,
522 async->mirror_num, async->bio_flags);
523}
524
525static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 526{
527 struct btrfs_fs_info *fs_info;
528 struct async_submit_bio *async;
4854ddd0 529 int limit;
8b712842
CM		 530
531 async = container_of(work, struct async_submit_bio, work);
532 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 533
b64a2851 534 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 535 limit = limit * 2 / 3;
536
8b712842 537 atomic_dec(&fs_info->nr_async_submits);
0986fe9e 538
b64a2851
CM		 539 if (atomic_read(&fs_info->nr_async_submits) < limit &&
540 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 541 wake_up(&fs_info->async_submit_wait);
542
4a69a410 543 async->submit_bio_done(async->inode, async->rw, async->bio,
c8b97818 544 async->mirror_num, async->bio_flags);
4a69a410
CM		 545}
546
547static void run_one_async_free(struct btrfs_work *work)
548{
549 struct async_submit_bio *async;
550
551 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 552 kfree(async);
553}
554
44b8bd7e
CM		 555int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
556 int rw, struct bio *bio, int mirror_num,
c8b97818 557 unsigned long bio_flags,
4a69a410
CM		 558 extent_submit_bio_hook_t *submit_bio_start,
559 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 560{
561 struct async_submit_bio *async;
562
563 async = kmalloc(sizeof(*async), GFP_NOFS);
564 if (!async)
565 return -ENOMEM;
566
567 async->inode = inode;
568 async->rw = rw;
569 async->bio = bio;
570 async->mirror_num = mirror_num;
4a69a410
CM		 571 async->submit_bio_start = submit_bio_start;
572 async->submit_bio_done = submit_bio_done;
573
574 async->work.func = run_one_async_start;
575 async->work.ordered_func = run_one_async_done;
576 async->work.ordered_free = run_one_async_free;
577
8b712842 578 async->work.flags = 0;
c8b97818 579 async->bio_flags = bio_flags;
8c8bee1d 580
cb03c743 581 atomic_inc(&fs_info->nr_async_submits);
8b712842 582 btrfs_queue_worker(&fs_info->workers, &async->work);
d2c3f4f6 583#if 0
7c2ca468 584 int limit = btrfs_async_submit_limit(fs_info);
9473f16c
CM		 585 if (atomic_read(&fs_info->nr_async_submits) > limit) {
586 wait_event_timeout(fs_info->async_submit_wait,
4854ddd0
CM		 587 (atomic_read(&fs_info->nr_async_submits) < limit),
588 HZ/10);
9473f16c
CM		 589
590 wait_event_timeout(fs_info->async_submit_wait,
591 (atomic_read(&fs_info->nr_async_bios) < limit),
592 HZ/10);
593 }
d2c3f4f6 594#endif
d397712b 595 while (atomic_read(&fs_info->async_submit_draining) &&
771ed689
CM		 596 atomic_read(&fs_info->nr_async_submits)) {
597 wait_event(fs_info->async_submit_wait,
598 (atomic_read(&fs_info->nr_async_submits) == 0));
599 }
600
44b8bd7e
CM		 601 return 0;
602}
603
ce3ed71a
CM		 604static int btree_csum_one_bio(struct bio *bio)
605{
606 struct bio_vec *bvec = bio->bi_io_vec;
607 int bio_index = 0;
608 struct btrfs_root *root;
609
610 WARN_ON(bio->bi_vcnt <= 0);
d397712b 611 while (bio_index < bio->bi_vcnt) {
ce3ed71a
CM		 612 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
613 csum_dirty_buffer(root, bvec->bv_page);
614 bio_index++;
615 bvec++;
616 }
617 return 0;
618}
619
4a69a410
CM		 620static int __btree_submit_bio_start(struct inode *inode, int rw,
621 struct bio *bio, int mirror_num,
622 unsigned long bio_flags)
22c59948 623{
8b712842
CM		 624 /*
625 * when we're called for a write, we're already in the async
5443be45 626 * submission context. Just jump into btrfs_map_bio
8b712842 627 */
4a69a410
CM		 628 btree_csum_one_bio(bio);
629 return 0;
630}
22c59948 631
4a69a410
CM		 632static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
633 int mirror_num, unsigned long bio_flags)
634{
8b712842 635 /*
4a69a410
CM		 636 * when we're called for a write, we're already in the async
637 * submission context. Just jump into btrfs_map_bio
8b712842 638 */
8b712842 639 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
0b86a832
CM		 640}
641
44b8bd7e 642static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
c8b97818 643 int mirror_num, unsigned long bio_flags)
44b8bd7e 644{
cad321ad
CM		 645 int ret;
646
647 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
648 bio, 1);
649 BUG_ON(ret);
650
44b8bd7e 651 if (!(rw & (1 << BIO_RW))) {
4a69a410
CM		 652 /*
653 * called for a read, do the setup so that checksum validation
654 * can happen in the async kernel threads
655 */
4a69a410 656 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
6f3577bd 657 mirror_num, 0);
44b8bd7e 658 }
cad321ad
CM		 659 /*
660 * kthread helpers are used to submit writes so that checksumming
661 * can happen in parallel across all CPUs
662 */
44b8bd7e 663 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
c8b97818 664 inode, rw, bio, mirror_num, 0,
4a69a410
CM		 665 __btree_submit_bio_start,
666 __btree_submit_bio_done);
44b8bd7e
CM		 667}
668
0da5468f
CM		 669static int btree_writepage(struct page *page, struct writeback_control *wbc)
670{
d1310b2e 671 struct extent_io_tree *tree;
b9473439
CM		 672 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
673 struct extent_buffer *eb;
674 int was_dirty;
675
d1310b2e 676 tree = &BTRFS_I(page->mapping->host)->io_tree;
b9473439
CM		 677 if (!(current->flags & PF_MEMALLOC)) {
678 return extent_write_full_page(tree, page,
679 btree_get_extent, wbc);
680 }
5443be45 681
b9473439
CM		 682 redirty_page_for_writepage(wbc, page);
683 eb = btrfs_find_tree_block(root, page_offset(page),
684 PAGE_CACHE_SIZE);
685 WARN_ON(!eb);
686
687 was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
688 if (!was_dirty) {
689 spin_lock(&root->fs_info->delalloc_lock);
690 root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE;
691 spin_unlock(&root->fs_info->delalloc_lock);
5443be45 692 }
b9473439
CM		 693 free_extent_buffer(eb);
694
695 unlock_page(page);
696 return 0;
5f39d397 697}
0da5468f
CM		 698
699static int btree_writepages(struct address_space *mapping,
700 struct writeback_control *wbc)
701{
d1310b2e
CM		 702 struct extent_io_tree *tree;
703 tree = &BTRFS_I(mapping->host)->io_tree;
d8d5f3e1 704 if (wbc->sync_mode == WB_SYNC_NONE) {
b9473439 705 struct btrfs_root *root = BTRFS_I(mapping->host)->root;
793955bc 706 u64 num_dirty;
24ab9cd8 707 unsigned long thresh = 32 * 1024 * 1024;
448d640b
CM		 708
709 if (wbc->for_kupdate)
710 return 0;
711
b9473439
CM		 712 /* this is a bit racy, but that's ok */
713 num_dirty = root->fs_info->dirty_metadata_bytes;
d397712b 714 if (num_dirty < thresh)
793955bc 715 return 0;
793955bc 716 }
0da5468f
CM		 717 return extent_writepages(tree, mapping, btree_get_extent, wbc);
718}
719
b2950863 720static int btree_readpage(struct file *file, struct page *page)
5f39d397 721{
d1310b2e
CM		 722 struct extent_io_tree *tree;
723 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 724 return extent_read_full_page(tree, page, btree_get_extent);
725}
22b0ebda 726
70dec807 727static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 728{
d1310b2e
CM		 729 struct extent_io_tree *tree;
730 struct extent_map_tree *map;
5f39d397 731 int ret;
d98237b3 732
98509cfc 733 if (PageWriteback(page) || PageDirty(page))
d397712b 734 return 0;
98509cfc 735
d1310b2e
CM		 736 tree = &BTRFS_I(page->mapping->host)->io_tree;
737 map = &BTRFS_I(page->mapping->host)->extent_tree;
6af118ce 738
7b13b7b1 739 ret = try_release_extent_state(map, tree, page, gfp_flags);
d397712b 740 if (!ret)
6af118ce 741 return 0;
6af118ce
CM		 742
743 ret = try_release_extent_buffer(tree, page);
5f39d397
CM		 744 if (ret == 1) {
745 ClearPagePrivate(page);
746 set_page_private(page, 0);
747 page_cache_release(page);
748 }
6af118ce 749
d98237b3
CM		 750 return ret;
751}
752
5f39d397 753static void btree_invalidatepage(struct page *page, unsigned long offset)
d98237b3 754{
d1310b2e
CM		 755 struct extent_io_tree *tree;
756 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 757 extent_invalidatepage(tree, page, offset);
758 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 759 if (PagePrivate(page)) {
d397712b
CM		 760 printk(KERN_WARNING "btrfs warning page private not zero "
761 "on page %llu\n", (unsigned long long)page_offset(page));
9ad6b7bc
CM		 762 ClearPagePrivate(page);
763 set_page_private(page, 0);
764 page_cache_release(page);
765 }
d98237b3
CM		 766}
767
5f39d397 768#if 0
d98237b3 769static int btree_writepage(struct page *page, struct writeback_control *wbc)
ed2ff2cb 770{
87cbda5c 771 struct buffer_head *bh;
0f7d52f4 772 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
87cbda5c 773 struct buffer_head *head;
87cbda5c
CM		 774 if (!page_has_buffers(page)) {
775 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
776 (1 << BH_Dirty)|(1 << BH_Uptodate));
777 }
778 head = page_buffers(page);
779 bh = head;
780 do {
781 if (buffer_dirty(bh))
782 csum_tree_block(root, bh, 0);
783 bh = bh->b_this_page;
784 } while (bh != head);
d98237b3 785 return block_write_full_page(page, btree_get_block, wbc);
ed2ff2cb 786}
5f39d397 787#endif
eb60ceac 788
d98237b3
CM		 789static struct address_space_operations btree_aops = {
790 .readpage = btree_readpage,
791 .writepage = btree_writepage,
0da5468f 792 .writepages = btree_writepages,
5f39d397
CM		 793 .releasepage = btree_releasepage,
794 .invalidatepage = btree_invalidatepage,
d98237b3
CM		 795 .sync_page = block_sync_page,
796};
797
ca7a79ad
CM		 798int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
799 u64 parent_transid)
090d1875 800{
5f39d397
CM		 801 struct extent_buffer *buf = NULL;
802 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 803 int ret = 0;
090d1875 804
db94535d 805 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 806 if (!buf)
090d1875 807 return 0;
d1310b2e 808 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
f188591e 809 buf, 0, 0, btree_get_extent, 0);
5f39d397 810 free_extent_buffer(buf);
de428b63 811 return ret;
090d1875
CM		 812}
813
0999df54
CM		 814struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
815 u64 bytenr, u32 blocksize)
816{
817 struct inode *btree_inode = root->fs_info->btree_inode;
818 struct extent_buffer *eb;
819 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
820 bytenr, blocksize, GFP_NOFS);
821 return eb;
822}
823
824struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
825 u64 bytenr, u32 blocksize)
826{
827 struct inode *btree_inode = root->fs_info->btree_inode;
828 struct extent_buffer *eb;
829
830 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
831 bytenr, blocksize, NULL, GFP_NOFS);
832 return eb;
833}
834
835
e02119d5
CM		 836int btrfs_write_tree_block(struct extent_buffer *buf)
837{
838 return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start,
24ab9cd8 839 buf->start + buf->len - 1, WB_SYNC_ALL);
e02119d5
CM		 840}
841
842int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
843{
844 return btrfs_wait_on_page_writeback_range(buf->first_page->mapping,
d397712b 845 buf->start, buf->start + buf->len - 1);
e02119d5
CM		 846}
847
0999df54 848struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 849 u32 blocksize, u64 parent_transid)
0999df54
CM		 850{
851 struct extent_buffer *buf = NULL;
852 struct inode *btree_inode = root->fs_info->btree_inode;
853 struct extent_io_tree *io_tree;
854 int ret;
855
856 io_tree = &BTRFS_I(btree_inode)->io_tree;
857
858 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
859 if (!buf)
860 return NULL;
0999df54 861
ca7a79ad 862 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
ce9adaa5 863
d397712b 864 if (ret == 0)
b4ce94de 865 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
d397712b 866 else
a1b32a59 867 WARN_ON(1);
5f39d397 868 return buf;
ce9adaa5 869
eb60ceac
CM		 870}
871
e089f05c 872int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
5f39d397 873 struct extent_buffer *buf)
ed2ff2cb 874{
5f39d397 875 struct inode *btree_inode = root->fs_info->btree_inode;
55c69072 876 if (btrfs_header_generation(buf) ==
925baedd 877 root->fs_info->running_transaction->transid) {
b9447ef8 878 btrfs_assert_tree_locked(buf);
b4ce94de 879
b9473439
CM		 880 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
881 spin_lock(&root->fs_info->delalloc_lock);
882 if (root->fs_info->dirty_metadata_bytes >= buf->len)
883 root->fs_info->dirty_metadata_bytes -= buf->len;
884 else
885 WARN_ON(1);
886 spin_unlock(&root->fs_info->delalloc_lock);
887 }
b4ce94de 888
b9473439
CM		 889 /* ugh, clear_extent_buffer_dirty needs to lock the page */
890 btrfs_set_lock_blocking(buf);
d1310b2e 891 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
55c69072 892 buf);
925baedd 893 }
5f39d397
CM		 894 return 0;
895}
896
db94535d 897static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
87ee04eb 898 u32 stripesize, struct btrfs_root *root,
9f5fae2f 899 struct btrfs_fs_info *fs_info,
e20d96d6 900 u64 objectid)
d97e63b6 901{
cfaa7295 902 root->node = NULL;
a28ec197 903 root->commit_root = NULL;
31153d81 904 root->ref_tree = NULL;
db94535d
CM		 905 root->sectorsize = sectorsize;
906 root->nodesize = nodesize;
907 root->leafsize = leafsize;
87ee04eb 908 root->stripesize = stripesize;
123abc88 909 root->ref_cows = 0;
0b86a832
CM		 910 root->track_dirty = 0;
911
9f5fae2f 912 root->fs_info = fs_info;
0f7d52f4
CM		 913 root->objectid = objectid;
914 root->last_trans = 0;
1b05da2e
CM		 915 root->highest_inode = 0;
916 root->last_inode_alloc = 0;
58176a96 917 root->name = NULL;
4313b399 918 root->in_sysfs = 0;
0b86a832
CM		 919
920 INIT_LIST_HEAD(&root->dirty_list);
7b128766 921 INIT_LIST_HEAD(&root->orphan_list);
bcc63abb 922 INIT_LIST_HEAD(&root->dead_list);
925baedd 923 spin_lock_init(&root->node_lock);
bcc63abb 924 spin_lock_init(&root->list_lock);
a2135011 925 mutex_init(&root->objectid_mutex);
e02119d5 926 mutex_init(&root->log_mutex);
7237f183
YZ		 927 init_waitqueue_head(&root->log_writer_wait);
928 init_waitqueue_head(&root->log_commit_wait[0]);
929 init_waitqueue_head(&root->log_commit_wait[1]);
930 atomic_set(&root->log_commit[0], 0);
931 atomic_set(&root->log_commit[1], 0);
932 atomic_set(&root->log_writers, 0);
933 root->log_batch = 0;
934 root->log_transid = 0;
d0c803c4
CM		 935 extent_io_tree_init(&root->dirty_log_pages,
936 fs_info->btree_inode->i_mapping, GFP_NOFS);
017e5369
CM		 937
938 btrfs_leaf_ref_tree_init(&root->ref_tree_struct);
939 root->ref_tree = &root->ref_tree_struct;
940
3768f368
CM		 941 memset(&root->root_key, 0, sizeof(root->root_key));
942 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 943 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 944 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
3f157a2f 945 root->defrag_trans_start = fs_info->generation;
58176a96 946 init_completion(&root->kobj_unregister);
6702ed49
CM		 947 root->defrag_running = 0;
948 root->defrag_level = 0;
4d775673 949 root->root_key.objectid = objectid;
3394e160
CM		 950 root->anon_super.s_root = NULL;
951 root->anon_super.s_dev = 0;
952 INIT_LIST_HEAD(&root->anon_super.s_list);
953 INIT_LIST_HEAD(&root->anon_super.s_instances);
954 init_rwsem(&root->anon_super.s_umount);
955
3768f368
CM		 956 return 0;
957}
958
db94535d 959static int find_and_setup_root(struct btrfs_root *tree_root,
9f5fae2f
CM		 960 struct btrfs_fs_info *fs_info,
961 u64 objectid,
e20d96d6 962 struct btrfs_root *root)
3768f368
CM		 963{
964 int ret;
db94535d 965 u32 blocksize;
84234f3a 966 u64 generation;
3768f368 967
db94535d 968 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 969 tree_root->sectorsize, tree_root->stripesize,
970 root, fs_info, objectid);
3768f368
CM		 971 ret = btrfs_find_last_root(tree_root, objectid,
972 &root->root_item, &root->root_key);
973 BUG_ON(ret);
974
84234f3a 975 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 976 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
977 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 978 blocksize, generation);
3768f368 979 BUG_ON(!root->node);
d97e63b6
CM		 980 return 0;
981}
982
e02119d5
CM		 983int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
984 struct btrfs_fs_info *fs_info)
985{
986 struct extent_buffer *eb;
d0c803c4
CM		 987 struct btrfs_root *log_root_tree = fs_info->log_root_tree;
988 u64 start = 0;
989 u64 end = 0;
e02119d5
CM		 990 int ret;
991
d0c803c4 992 if (!log_root_tree)
e02119d5
CM		 993 return 0;
994
d397712b 995 while (1) {
d0c803c4
CM		 996 ret = find_first_extent_bit(&log_root_tree->dirty_log_pages,
997 0, &start, &end, EXTENT_DIRTY);
998 if (ret)
999 break;
1000
1001 clear_extent_dirty(&log_root_tree->dirty_log_pages,
1002 start, end, GFP_NOFS);
1003 }
e02119d5
CM		 1004 eb = fs_info->log_root_tree->node;
1005
1006 WARN_ON(btrfs_header_level(eb) != 0);
1007 WARN_ON(btrfs_header_nritems(eb) != 0);
1008
d00aff00
CM		 1009 ret = btrfs_free_reserved_extent(fs_info->tree_root,
1010 eb->start, eb->len);
e02119d5
CM		 1011 BUG_ON(ret);
1012
1013 free_extent_buffer(eb);
1014 kfree(fs_info->log_root_tree);
1015 fs_info->log_root_tree = NULL;
1016 return 0;
1017}
1018
7237f183
YZ		 1019static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1020 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 1021{
1022 struct btrfs_root *root;
1023 struct btrfs_root *tree_root = fs_info->tree_root;
7237f183 1024 struct extent_buffer *leaf;
e02119d5
CM		 1025
1026 root = kzalloc(sizeof(*root), GFP_NOFS);
1027 if (!root)
7237f183 1028 return ERR_PTR(-ENOMEM);
e02119d5
CM		 1029
1030 __setup_root(tree_root->nodesize, tree_root->leafsize,
1031 tree_root->sectorsize, tree_root->stripesize,
1032 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1033
1034 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1035 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1036 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
7237f183
YZ		 1037 /*
1038 * log trees do not get reference counted because they go away
1039 * before a real commit is actually done. They do store pointers
1040 * to file data extents, and those reference counts still get
1041 * updated (along with back refs to the log tree).
1042 */
e02119d5
CM		 1043 root->ref_cows = 0;
1044
7237f183
YZ		 1045 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1046 0, BTRFS_TREE_LOG_OBJECTID,
1047 trans->transid, 0, 0, 0);
1048 if (IS_ERR(leaf)) {
1049 kfree(root);
1050 return ERR_CAST(leaf);
1051 }
e02119d5 1052
7237f183 1053 root->node = leaf;
e02119d5
CM		 1054 btrfs_set_header_nritems(root->node, 0);
1055 btrfs_set_header_level(root->node, 0);
1056 btrfs_set_header_bytenr(root->node, root->node->start);
1057 btrfs_set_header_generation(root->node, trans->transid);
1058 btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID);
1059
1060 write_extent_buffer(root->node, root->fs_info->fsid,
1061 (unsigned long)btrfs_header_fsid(root->node),
1062 BTRFS_FSID_SIZE);
1063 btrfs_mark_buffer_dirty(root->node);
1064 btrfs_tree_unlock(root->node);
7237f183
YZ		 1065 return root;
1066}
1067
1068int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1069 struct btrfs_fs_info *fs_info)
1070{
1071 struct btrfs_root *log_root;
1072
1073 log_root = alloc_log_tree(trans, fs_info);
1074 if (IS_ERR(log_root))
1075 return PTR_ERR(log_root);
1076 WARN_ON(fs_info->log_root_tree);
1077 fs_info->log_root_tree = log_root;
1078 return 0;
1079}
1080
1081int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1082 struct btrfs_root *root)
1083{
1084 struct btrfs_root *log_root;
1085 struct btrfs_inode_item *inode_item;
1086
1087 log_root = alloc_log_tree(trans, root->fs_info);
1088 if (IS_ERR(log_root))
1089 return PTR_ERR(log_root);
1090
1091 log_root->last_trans = trans->transid;
1092 log_root->root_key.offset = root->root_key.objectid;
1093
1094 inode_item = &log_root->root_item.inode;
1095 inode_item->generation = cpu_to_le64(1);
1096 inode_item->size = cpu_to_le64(3);
1097 inode_item->nlink = cpu_to_le32(1);
1098 inode_item->nbytes = cpu_to_le64(root->leafsize);
1099 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1100
1101 btrfs_set_root_bytenr(&log_root->root_item, log_root->node->start);
1102 btrfs_set_root_generation(&log_root->root_item, trans->transid);
1103
1104 WARN_ON(root->log_root);
1105 root->log_root = log_root;
1106 root->log_transid = 0;
e02119d5
CM		 1107 return 0;
1108}
1109
1110struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
1111 struct btrfs_key *location)
1112{
1113 struct btrfs_root *root;
1114 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1115 struct btrfs_path *path;
5f39d397 1116 struct extent_buffer *l;
1b05da2e 1117 u64 highest_inode;
84234f3a 1118 u64 generation;
db94535d 1119 u32 blocksize;
0f7d52f4
CM		 1120 int ret = 0;
1121
5eda7b5e 1122 root = kzalloc(sizeof(*root), GFP_NOFS);
0cf6c620 1123 if (!root)
0f7d52f4 1124 return ERR_PTR(-ENOMEM);
0f7d52f4 1125 if (location->offset == (u64)-1) {
db94535d 1126 ret = find_and_setup_root(tree_root, fs_info,
0f7d52f4
CM		 1127 location->objectid, root);
1128 if (ret) {
0f7d52f4
CM		 1129 kfree(root);
1130 return ERR_PTR(ret);
1131 }
1132 goto insert;
1133 }
1134
db94535d 1135 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 1136 tree_root->sectorsize, tree_root->stripesize,
1137 root, fs_info, location->objectid);
0f7d52f4
CM		 1138
1139 path = btrfs_alloc_path();
1140 BUG_ON(!path);
1141 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
1142 if (ret != 0) {
0f7d52f4
CM		 1143 if (ret > 0)
1144 ret = -ENOENT;
1145 goto out;
1146 }
5f39d397
CM		 1147 l = path->nodes[0];
1148 read_extent_buffer(l, &root->root_item,
1149 btrfs_item_ptr_offset(l, path->slots[0]),
0f7d52f4 1150 sizeof(root->root_item));
44b36eb2 1151 memcpy(&root->root_key, location, sizeof(*location));
0f7d52f4
CM		 1152 ret = 0;
1153out:
1154 btrfs_release_path(root, path);
1155 btrfs_free_path(path);
1156 if (ret) {
1157 kfree(root);
1158 return ERR_PTR(ret);
1159 }
84234f3a 1160 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1161 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1162 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1163 blocksize, generation);
0f7d52f4
CM		 1164 BUG_ON(!root->node);
1165insert:
e02119d5
CM		 1166 if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
1167 root->ref_cows = 1;
1168 ret = btrfs_find_highest_inode(root, &highest_inode);
1169 if (ret == 0) {
1170 root->highest_inode = highest_inode;
1171 root->last_inode_alloc = highest_inode;
1172 }
5eda7b5e
CM		 1173 }
1174 return root;
1175}
1176
dc17ff8f
CM		 1177struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1178 u64 root_objectid)
1179{
1180 struct btrfs_root *root;
1181
1182 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
1183 return fs_info->tree_root;
1184 if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
1185 return fs_info->extent_root;
1186
1187 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1188 (unsigned long)root_objectid);
1189 return root;
1190}
1191
edbd8d4e
CM		 1192struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
1193 struct btrfs_key *location)
5eda7b5e
CM		 1194{
1195 struct btrfs_root *root;
1196 int ret;
1197
edbd8d4e
CM		 1198 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1199 return fs_info->tree_root;
1200 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1201 return fs_info->extent_root;
8f18cf13
CM		 1202 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1203 return fs_info->chunk_root;
1204 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1205 return fs_info->dev_root;
0403e47e
YZ		 1206 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1207 return fs_info->csum_root;
edbd8d4e 1208
5eda7b5e
CM		 1209 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1210 (unsigned long)location->objectid);
1211 if (root)
1212 return root;
1213
e02119d5 1214 root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
5eda7b5e
CM		 1215 if (IS_ERR(root))
1216 return root;
3394e160
CM		 1217
1218 set_anon_super(&root->anon_super, NULL);
1219
2619ba1f
CM		 1220 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1221 (unsigned long)root->root_key.objectid,
0f7d52f4
CM		 1222 root);
1223 if (ret) {
5f39d397 1224 free_extent_buffer(root->node);
0f7d52f4
CM		 1225 kfree(root);
1226 return ERR_PTR(ret);
1227 }
c146afad
YZ		 1228 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
1229 ret = btrfs_find_dead_roots(fs_info->tree_root,
1230 root->root_key.objectid, root);
1231 BUG_ON(ret);
1232 btrfs_orphan_cleanup(root);
1233 }
edbd8d4e
CM		 1234 return root;
1235}
1236
1237struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
1238 struct btrfs_key *location,
1239 const char *name, int namelen)
1240{
1241 struct btrfs_root *root;
1242 int ret;
1243
1244 root = btrfs_read_fs_root_no_name(fs_info, location);
1245 if (!root)
1246 return NULL;
58176a96 1247
4313b399
CM		 1248 if (root->in_sysfs)
1249 return root;
1250
58176a96
JB		 1251 ret = btrfs_set_root_name(root, name, namelen);
1252 if (ret) {
5f39d397 1253 free_extent_buffer(root->node);
58176a96
JB		 1254 kfree(root);
1255 return ERR_PTR(ret);
1256 }
ea9e8b11 1257#if 0
58176a96
JB		 1258 ret = btrfs_sysfs_add_root(root);
1259 if (ret) {
5f39d397 1260 free_extent_buffer(root->node);
58176a96
JB		 1261 kfree(root->name);
1262 kfree(root);
1263 return ERR_PTR(ret);
1264 }
ea9e8b11 1265#endif
4313b399 1266 root->in_sysfs = 1;
0f7d52f4
CM		 1267 return root;
1268}
04160088
CM		 1269
1270static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1271{
1272 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1273 int ret = 0;
04160088
CM		 1274 struct btrfs_device *device;
1275 struct backing_dev_info *bdi;
6f3577bd 1276#if 0
cb03c743 1277 if ((bdi_bits & (1 << BDI_write_congested)) &&
777e6bd7 1278 btrfs_congested_async(info, 0))
cb03c743 1279 return 1;
6f3577bd 1280#endif
c6e30871 1281 list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
dfe25020
CM		 1282 if (!device->bdev)
1283 continue;
04160088
CM		 1284 bdi = blk_get_backing_dev_info(device->bdev);
1285 if (bdi && bdi_congested(bdi, bdi_bits)) {
1286 ret = 1;
1287 break;
1288 }
1289 }
1290 return ret;
1291}
1292
38b66988
CM		 1293/*
1294 * this unplugs every device on the box, and it is only used when page
1295 * is null
1296 */
1297static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
1298{
38b66988
CM		 1299 struct btrfs_device *device;
1300 struct btrfs_fs_info *info;
1301
1302 info = (struct btrfs_fs_info *)bdi->unplug_io_data;
c6e30871 1303 list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
d20f7043
CM		 1304 if (!device->bdev)
1305 continue;
1306
38b66988 1307 bdi = blk_get_backing_dev_info(device->bdev);
d397712b 1308 if (bdi->unplug_io_fn)
38b66988 1309 bdi->unplug_io_fn(bdi, page);
38b66988
CM		 1310 }
1311}
1312
b2950863 1313static void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
04160088 1314{
38b66988 1315 struct inode *inode;
f2d8d74d
CM		 1316 struct extent_map_tree *em_tree;
1317 struct extent_map *em;
bcbfce8a 1318 struct address_space *mapping;
38b66988
CM		 1319 u64 offset;
1320
bcbfce8a 1321 /* the generic O_DIRECT read code does this */
9f0ba5bd 1322 if (1 || !page) {
38b66988
CM		 1323 __unplug_io_fn(bdi, page);
1324 return;
1325 }
1326
bcbfce8a
CM		 1327 /*
1328 * page->mapping may change at any time. Get a consistent copy
1329 * and use that for everything below
1330 */
1331 smp_mb();
1332 mapping = page->mapping;
1333 if (!mapping)
1334 return;
1335
1336 inode = mapping->host;
240d5d48
CM		 1337
1338 /*
1339 * don't do the expensive searching for a small number of
1340 * devices
1341 */
1342 if (BTRFS_I(inode)->root->fs_info->fs_devices->open_devices <= 2) {
1343 __unplug_io_fn(bdi, page);
1344 return;
1345 }
1346
38b66988 1347 offset = page_offset(page);
04160088 1348
f2d8d74d
CM		 1349 em_tree = &BTRFS_I(inode)->extent_tree;
1350 spin_lock(&em_tree->lock);
1351 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
1352 spin_unlock(&em_tree->lock);
89642229
CM		 1353 if (!em) {
1354 __unplug_io_fn(bdi, page);
f2d8d74d 1355 return;
89642229 1356 }
f2d8d74d 1357
89642229
CM		 1358 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1359 free_extent_map(em);
1360 __unplug_io_fn(bdi, page);
1361 return;
1362 }
f2d8d74d
CM		 1363 offset = offset - em->start;
1364 btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree,
1365 em->block_start + offset, page);
1366 free_extent_map(em);
04160088
CM		 1367}
1368
1369static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1370{
1371 bdi_init(bdi);
4575c9cc 1372 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1373 bdi->state = 0;
1374 bdi->capabilities = default_backing_dev_info.capabilities;
1375 bdi->unplug_io_fn = btrfs_unplug_io_fn;
1376 bdi->unplug_io_data = info;
1377 bdi->congested_fn = btrfs_congested_fn;
1378 bdi->congested_data = info;
1379 return 0;
1380}
1381
ce9adaa5
CM		 1382static int bio_ready_for_csum(struct bio *bio)
1383{
1384 u64 length = 0;
1385 u64 buf_len = 0;
1386 u64 start = 0;
1387 struct page *page;
1388 struct extent_io_tree *io_tree = NULL;
1389 struct btrfs_fs_info *info = NULL;
1390 struct bio_vec *bvec;
1391 int i;
1392 int ret;
1393
1394 bio_for_each_segment(bvec, bio, i) {
1395 page = bvec->bv_page;
1396 if (page->private == EXTENT_PAGE_PRIVATE) {
1397 length += bvec->bv_len;
1398 continue;
1399 }
1400 if (!page->private) {
1401 length += bvec->bv_len;
1402 continue;
1403 }
1404 length = bvec->bv_len;
1405 buf_len = page->private >> 2;
1406 start = page_offset(page) + bvec->bv_offset;
1407 io_tree = &BTRFS_I(page->mapping->host)->io_tree;
1408 info = BTRFS_I(page->mapping->host)->root->fs_info;
1409 }
1410 /* are we fully contained in this bio? */
1411 if (buf_len <= length)
1412 return 1;
1413
1414 ret = extent_range_uptodate(io_tree, start + length,
1415 start + buf_len - 1);
ce9adaa5
CM		 1416 return ret;
1417}
1418
8b712842
CM		 1419/*
1420 * called by the kthread helper functions to finally call the bio end_io
1421 * functions. This is where read checksum verification actually happens
1422 */
1423static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1424{
ce9adaa5 1425 struct bio *bio;
8b712842
CM		 1426 struct end_io_wq *end_io_wq;
1427 struct btrfs_fs_info *fs_info;
ce9adaa5 1428 int error;
ce9adaa5 1429
8b712842
CM		 1430 end_io_wq = container_of(work, struct end_io_wq, work);
1431 bio = end_io_wq->bio;
1432 fs_info = end_io_wq->info;
ce9adaa5 1433
cad321ad 1434 /* metadata bio reads are special because the whole tree block must
8b712842
CM		 1435 * be checksummed at once. This makes sure the entire block is in
1436 * ram and up to date before trying to verify things. For
1437 * blocksize <= pagesize, it is basically a noop
1438 */
cad321ad
CM		 1439 if (!(bio->bi_rw & (1 << BIO_RW)) && end_io_wq->metadata &&
1440 !bio_ready_for_csum(bio)) {
d20f7043 1441 btrfs_queue_worker(&fs_info->endio_meta_workers,
8b712842
CM		 1442 &end_io_wq->work);
1443 return;
1444 }
1445 error = end_io_wq->error;
1446 bio->bi_private = end_io_wq->private;
1447 bio->bi_end_io = end_io_wq->end_io;
1448 kfree(end_io_wq);
8b712842 1449 bio_endio(bio, error);
44b8bd7e
CM		 1450}
1451
a74a4b97
CM		 1452static int cleaner_kthread(void *arg)
1453{
1454 struct btrfs_root *root = arg;
1455
1456 do {
1457 smp_mb();
1458 if (root->fs_info->closing)
1459 break;
1460
1461 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1462 mutex_lock(&root->fs_info->cleaner_mutex);
a74a4b97 1463 btrfs_clean_old_snapshots(root);
a74a4b97
CM		 1464 mutex_unlock(&root->fs_info->cleaner_mutex);
1465
1466 if (freezing(current)) {
1467 refrigerator();
1468 } else {
1469 smp_mb();
1470 if (root->fs_info->closing)
1471 break;
1472 set_current_state(TASK_INTERRUPTIBLE);
1473 schedule();
1474 __set_current_state(TASK_RUNNING);
1475 }
1476 } while (!kthread_should_stop());
1477 return 0;
1478}
1479
1480static int transaction_kthread(void *arg)
1481{
1482 struct btrfs_root *root = arg;
1483 struct btrfs_trans_handle *trans;
1484 struct btrfs_transaction *cur;
1485 unsigned long now;
1486 unsigned long delay;
1487 int ret;
1488
1489 do {
1490 smp_mb();
1491 if (root->fs_info->closing)
1492 break;
1493
1494 delay = HZ * 30;
1495 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1496 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1497
1498 mutex_lock(&root->fs_info->trans_mutex);
1499 cur = root->fs_info->running_transaction;
1500 if (!cur) {
1501 mutex_unlock(&root->fs_info->trans_mutex);
1502 goto sleep;
1503 }
31153d81 1504
a74a4b97
CM		 1505 now = get_seconds();
1506 if (now < cur->start_time || now - cur->start_time < 30) {
1507 mutex_unlock(&root->fs_info->trans_mutex);
1508 delay = HZ * 5;
1509 goto sleep;
1510 }
1511 mutex_unlock(&root->fs_info->trans_mutex);
a74a4b97
CM		 1512 trans = btrfs_start_transaction(root, 1);
1513 ret = btrfs_commit_transaction(trans, root);
56bec294 1514
a74a4b97
CM		 1515sleep:
1516 wake_up_process(root->fs_info->cleaner_kthread);
1517 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1518
1519 if (freezing(current)) {
1520 refrigerator();
1521 } else {
1522 if (root->fs_info->closing)
1523 break;
1524 set_current_state(TASK_INTERRUPTIBLE);
1525 schedule_timeout(delay);
1526 __set_current_state(TASK_RUNNING);
1527 }
1528 } while (!kthread_should_stop());
1529 return 0;
1530}
1531
8a4b83cc 1532struct btrfs_root *open_ctree(struct super_block *sb,
dfe25020
CM		 1533 struct btrfs_fs_devices *fs_devices,
1534 char *options)
2e635a27 1535{
db94535d
CM		 1536 u32 sectorsize;
1537 u32 nodesize;
1538 u32 leafsize;
1539 u32 blocksize;
87ee04eb 1540 u32 stripesize;
84234f3a 1541 u64 generation;
f2b636e8 1542 u64 features;
3de4586c 1543 struct btrfs_key location;
a061fc8d 1544 struct buffer_head *bh;
e02119d5 1545 struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
e20d96d6 1546 GFP_NOFS);
d20f7043
CM		 1547 struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root),
1548 GFP_NOFS);
e02119d5 1549 struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root),
e20d96d6 1550 GFP_NOFS);
8790d502 1551 struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
e20d96d6 1552 GFP_NOFS);
e02119d5 1553 struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
0b86a832 1554 GFP_NOFS);
e02119d5 1555 struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
0b86a832 1556 GFP_NOFS);
e02119d5
CM		 1557 struct btrfs_root *log_tree_root;
1558
eb60ceac 1559 int ret;
e58ca020 1560 int err = -EINVAL;
4543df7e 1561
2c90e5d6 1562 struct btrfs_super_block *disk_super;
8790d502 1563
0463bb4e 1564 if (!extent_root || !tree_root || !fs_info ||
d20f7043 1565 !chunk_root || !dev_root || !csum_root) {
39279cc3
CM		 1566 err = -ENOMEM;
1567 goto fail;
1568 }
0f7d52f4 1569 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
8fd17795 1570 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 1571 INIT_LIST_HEAD(&fs_info->dead_roots);
19c00ddc 1572 INIT_LIST_HEAD(&fs_info->hashers);
ea8c2819 1573 INIT_LIST_HEAD(&fs_info->delalloc_inodes);
5a3f23d5 1574 INIT_LIST_HEAD(&fs_info->ordered_operations);
1832a6d5 1575 spin_lock_init(&fs_info->delalloc_lock);
cee36a03 1576 spin_lock_init(&fs_info->new_trans_lock);
31153d81 1577 spin_lock_init(&fs_info->ref_cache_lock);
19c00ddc 1578
58176a96 1579 init_completion(&fs_info->kobj_unregister);
9f5fae2f
CM		 1580 fs_info->tree_root = tree_root;
1581 fs_info->extent_root = extent_root;
d20f7043 1582 fs_info->csum_root = csum_root;
0b86a832
CM		 1583 fs_info->chunk_root = chunk_root;
1584 fs_info->dev_root = dev_root;
8a4b83cc 1585 fs_info->fs_devices = fs_devices;
0b86a832 1586 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 1587 INIT_LIST_HEAD(&fs_info->space_info);
0b86a832 1588 btrfs_mapping_init(&fs_info->mapping_tree);
cb03c743 1589 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 1590 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 1591 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 1592 atomic_set(&fs_info->nr_async_bios, 0);
a2135011 1593 atomic_set(&fs_info->throttles, 0);
ab78c84d 1594 atomic_set(&fs_info->throttle_gen, 0);
e20d96d6 1595 fs_info->sb = sb;
c59f8951 1596 fs_info->max_extent = (u64)-1;
6f568d35 1597 fs_info->max_inline = 8192 * 1024;
04160088 1598 setup_bdi(fs_info, &fs_info->bdi);
d98237b3
CM		 1599 fs_info->btree_inode = new_inode(sb);
1600 fs_info->btree_inode->i_ino = 1;
2c90e5d6 1601 fs_info->btree_inode->i_nlink = 1;
c8b97818 1602
b34b086c
CM		 1603 fs_info->thread_pool_size = min_t(unsigned long,
1604 num_online_cpus() + 2, 8);
0afbaf8c 1605
3eaa2885
CM		 1606 INIT_LIST_HEAD(&fs_info->ordered_extents);
1607 spin_lock_init(&fs_info->ordered_extent_lock);
1608
a061fc8d
CM		 1609 sb->s_blocksize = 4096;
1610 sb->s_blocksize_bits = blksize_bits(4096);
1611
0afbaf8c
CM		 1612 /*
1613 * we set the i_size on the btree inode to the max possible int.
1614 * the real end of the address space is determined by all of
1615 * the devices in the system
1616 */
1617 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 1618 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 1619 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
1620
d1310b2e 1621 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
5f39d397
CM		 1622 fs_info->btree_inode->i_mapping,
1623 GFP_NOFS);
d1310b2e
CM		 1624 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
1625 GFP_NOFS);
1626
1627 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 1628
0f9dd46c
JB		 1629 spin_lock_init(&fs_info->block_group_cache_lock);
1630 fs_info->block_group_cache_tree.rb_node = NULL;
1631
d1310b2e 1632 extent_io_tree_init(&fs_info->pinned_extents,
1a5bc167 1633 fs_info->btree_inode->i_mapping, GFP_NOFS);
e66f709b 1634 fs_info->do_barriers = 1;
e18e4809 1635
1a40e23b
ZY		 1636 INIT_LIST_HEAD(&fs_info->dead_reloc_roots);
1637 btrfs_leaf_ref_tree_init(&fs_info->reloc_ref_tree);
e4657689
ZY		 1638 btrfs_leaf_ref_tree_init(&fs_info->shared_ref_tree);
1639
0f7d52f4
CM		 1640 BTRFS_I(fs_info->btree_inode)->root = tree_root;
1641 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
1642 sizeof(struct btrfs_key));
22b0ebda 1643 insert_inode_hash(fs_info->btree_inode);
39279cc3 1644
79154b1b 1645 mutex_init(&fs_info->trans_mutex);
5a3f23d5 1646 mutex_init(&fs_info->ordered_operations_mutex);
e02119d5 1647 mutex_init(&fs_info->tree_log_mutex);
a2135011 1648 mutex_init(&fs_info->drop_mutex);
925baedd 1649 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 1650 mutex_init(&fs_info->transaction_kthread_mutex);
1651 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 1652 mutex_init(&fs_info->volume_mutex);
1a40e23b 1653 mutex_init(&fs_info->tree_reloc_mutex);
fa9c0d79
CM		 1654
1655 btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
1656 btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
1657
e6dcd2dc 1658 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 1659 init_waitqueue_head(&fs_info->transaction_wait);
4854ddd0 1660 init_waitqueue_head(&fs_info->async_submit_wait);
3768f368 1661
0b86a832 1662 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 1663 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 1664
d98237b3 1665
a512bbf8 1666 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
a061fc8d 1667 if (!bh)
39279cc3 1668 goto fail_iput;
39279cc3 1669
a061fc8d 1670 memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy));
2d69a0f8
YZ		 1671 memcpy(&fs_info->super_for_commit, &fs_info->super_copy,
1672 sizeof(fs_info->super_for_commit));
a061fc8d 1673 brelse(bh);
5f39d397 1674
a061fc8d 1675 memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE);
0b86a832 1676
5f39d397 1677 disk_super = &fs_info->super_copy;
0f7d52f4 1678 if (!btrfs_super_root(disk_super))
c6e2bac1 1679 goto fail_iput;
0f7d52f4 1680
2b82032c
YZ		 1681 ret = btrfs_parse_options(tree_root, options);
1682 if (ret) {
1683 err = ret;
c6e2bac1 1684 goto fail_iput;
2b82032c 1685 }
dfe25020 1686
f2b636e8
JB		 1687 features = btrfs_super_incompat_flags(disk_super) &
1688 ~BTRFS_FEATURE_INCOMPAT_SUPP;
1689 if (features) {
1690 printk(KERN_ERR "BTRFS: couldn't mount because of "
1691 "unsupported optional features (%Lx).\n",
1692 features);
1693 err = -EINVAL;
c6e2bac1 1694 goto fail_iput;
f2b636e8
JB		 1695 }
1696
1697 features = btrfs_super_compat_ro_flags(disk_super) &
1698 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
1699 if (!(sb->s_flags & MS_RDONLY) && features) {
1700 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
1701 "unsupported option features (%Lx).\n",
1702 features);
1703 err = -EINVAL;
c6e2bac1 1704 goto fail_iput;
f2b636e8
JB		 1705 }
1706
4543df7e
CM		 1707 /*
1708 * we need to start all the end_io workers up front because the
1709 * queue work function gets called at interrupt time, and so it
1710 * cannot dynamically grow.
1711 */
5443be45
CM		 1712 btrfs_init_workers(&fs_info->workers, "worker",
1713 fs_info->thread_pool_size);
c8b97818 1714
771ed689
CM		 1715 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
1716 fs_info->thread_pool_size);
1717
5443be45 1718 btrfs_init_workers(&fs_info->submit_workers, "submit",
b720d209
CM		 1719 min_t(u64, fs_devices->num_devices,
1720 fs_info->thread_pool_size));
61b49440
CM		 1721
1722 /* a higher idle thresh on the submit workers makes it much more
1723 * likely that bios will be send down in a sane order to the
1724 * devices
1725 */
1726 fs_info->submit_workers.idle_thresh = 64;
53863232 1727
771ed689 1728 fs_info->workers.idle_thresh = 16;
4a69a410 1729 fs_info->workers.ordered = 1;
61b49440 1730
771ed689
CM		 1731 fs_info->delalloc_workers.idle_thresh = 2;
1732 fs_info->delalloc_workers.ordered = 1;
1733
5443be45
CM		 1734 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1);
1735 btrfs_init_workers(&fs_info->endio_workers, "endio",
1736 fs_info->thread_pool_size);
d20f7043
CM		 1737 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
1738 fs_info->thread_pool_size);
cad321ad
CM		 1739 btrfs_init_workers(&fs_info->endio_meta_write_workers,
1740 "endio-meta-write", fs_info->thread_pool_size);
5443be45 1741 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
e6dcd2dc 1742 fs_info->thread_pool_size);
61b49440
CM		 1743
1744 /*
1745 * endios are largely parallel and should have a very
1746 * low idle thresh
1747 */
1748 fs_info->endio_workers.idle_thresh = 4;
b51912c9
CM		 1749 fs_info->endio_meta_workers.idle_thresh = 4;
1750
3a5f1d45 1751 fs_info->endio_write_workers.idle_thresh = 64;
cad321ad 1752 fs_info->endio_meta_write_workers.idle_thresh = 64;
61b49440 1753
4543df7e 1754 btrfs_start_workers(&fs_info->workers, 1);
1cc127b5 1755 btrfs_start_workers(&fs_info->submit_workers, 1);
771ed689 1756 btrfs_start_workers(&fs_info->delalloc_workers, 1);
247e743c 1757 btrfs_start_workers(&fs_info->fixup_workers, 1);
4543df7e 1758 btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size);
d20f7043
CM		 1759 btrfs_start_workers(&fs_info->endio_meta_workers,
1760 fs_info->thread_pool_size);
cad321ad
CM		 1761 btrfs_start_workers(&fs_info->endio_meta_write_workers,
1762 fs_info->thread_pool_size);
e6dcd2dc
CM		 1763 btrfs_start_workers(&fs_info->endio_write_workers,
1764 fs_info->thread_pool_size);
4543df7e 1765
4575c9cc 1766 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 1767 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
1768 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 1769
db94535d
CM		 1770 nodesize = btrfs_super_nodesize(disk_super);
1771 leafsize = btrfs_super_leafsize(disk_super);
1772 sectorsize = btrfs_super_sectorsize(disk_super);
87ee04eb 1773 stripesize = btrfs_super_stripesize(disk_super);
db94535d
CM		 1774 tree_root->nodesize = nodesize;
1775 tree_root->leafsize = leafsize;
1776 tree_root->sectorsize = sectorsize;
87ee04eb 1777 tree_root->stripesize = stripesize;
a061fc8d
CM		 1778
1779 sb->s_blocksize = sectorsize;
1780 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 1781
39279cc3
CM		 1782 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
1783 sizeof(disk_super->magic))) {
d397712b 1784 printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
39279cc3
CM		 1785 goto fail_sb_buffer;
1786 }
19c00ddc 1787
925baedd 1788 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 1789 ret = btrfs_read_sys_array(tree_root);
925baedd 1790 mutex_unlock(&fs_info->chunk_mutex);
84eed90f 1791 if (ret) {
d397712b
CM		 1792 printk(KERN_WARNING "btrfs: failed to read the system "
1793 "array on %s\n", sb->s_id);
84eed90f
CM		 1794 goto fail_sys_array;
1795 }
0b86a832
CM		 1796
1797 blocksize = btrfs_level_size(tree_root,
1798 btrfs_super_chunk_root_level(disk_super));
84234f3a 1799 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 1800
1801 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1802 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1803
1804 chunk_root->node = read_tree_block(chunk_root,
1805 btrfs_super_chunk_root(disk_super),
84234f3a 1806 blocksize, generation);
0b86a832
CM		 1807 BUG_ON(!chunk_root->node);
1808
e17cade2 1809 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
d397712b
CM		 1810 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
1811 BTRFS_UUID_SIZE);
e17cade2 1812
925baedd 1813 mutex_lock(&fs_info->chunk_mutex);
0b86a832 1814 ret = btrfs_read_chunk_tree(chunk_root);
925baedd 1815 mutex_unlock(&fs_info->chunk_mutex);
2b82032c 1816 if (ret) {
d397712b
CM		 1817 printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
1818 sb->s_id);
2b82032c
YZ		 1819 goto fail_chunk_root;
1820 }
0b86a832 1821
dfe25020
CM		 1822 btrfs_close_extra_devices(fs_devices);
1823
db94535d
CM		 1824 blocksize = btrfs_level_size(tree_root,
1825 btrfs_super_root_level(disk_super));
84234f3a 1826 generation = btrfs_super_generation(disk_super);
0b86a832 1827
e20d96d6 1828 tree_root->node = read_tree_block(tree_root,
db94535d 1829 btrfs_super_root(disk_super),
84234f3a 1830 blocksize, generation);
39279cc3 1831 if (!tree_root->node)
2b82032c 1832 goto fail_chunk_root;
3768f368 1833
db94535d
CM		 1834
1835 ret = find_and_setup_root(tree_root, fs_info,
e20d96d6 1836 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
0b86a832 1837 if (ret)
39279cc3 1838 goto fail_tree_root;
0b86a832
CM		 1839 extent_root->track_dirty = 1;
1840
1841 ret = find_and_setup_root(tree_root, fs_info,
1842 BTRFS_DEV_TREE_OBJECTID, dev_root);
1843 dev_root->track_dirty = 1;
0b86a832
CM		 1844 if (ret)
1845 goto fail_extent_root;
3768f368 1846
d20f7043
CM		 1847 ret = find_and_setup_root(tree_root, fs_info,
1848 BTRFS_CSUM_TREE_OBJECTID, csum_root);
1849 if (ret)
1850 goto fail_extent_root;
1851
1852 csum_root->track_dirty = 1;
1853
9078a3e1
CM		 1854 btrfs_read_block_groups(extent_root);
1855
2d69a0f8 1856 fs_info->generation = generation;
c146afad 1857 fs_info->last_trans_committed = generation;
d18a2c44
CM		 1858 fs_info->data_alloc_profile = (u64)-1;
1859 fs_info->metadata_alloc_profile = (u64)-1;
1860 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
a74a4b97
CM		 1861 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
1862 "btrfs-cleaner");
57506d50 1863 if (IS_ERR(fs_info->cleaner_kthread))
d20f7043 1864 goto fail_csum_root;
a74a4b97
CM		 1865
1866 fs_info->transaction_kthread = kthread_run(transaction_kthread,
1867 tree_root,
1868 "btrfs-transaction");
57506d50 1869 if (IS_ERR(fs_info->transaction_kthread))
3f157a2f 1870 goto fail_cleaner;
a74a4b97 1871
e02119d5 1872 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM		 1873 u64 bytenr = btrfs_super_log_root(disk_super);
1874
7c2ca468 1875 if (fs_devices->rw_devices == 0) {
d397712b
CM		 1876 printk(KERN_WARNING "Btrfs log replay required "
1877 "on RO media\n");
7c2ca468
CM		 1878 err = -EIO;
1879 goto fail_trans_kthread;
1880 }
e02119d5
CM		 1881 blocksize =
1882 btrfs_level_size(tree_root,
1883 btrfs_super_log_root_level(disk_super));
d18a2c44 1884
e02119d5
CM		 1885 log_tree_root = kzalloc(sizeof(struct btrfs_root),
1886 GFP_NOFS);
1887
1888 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1889 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1890
1891 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 1892 blocksize,
1893 generation + 1);
e02119d5
CM		 1894 ret = btrfs_recover_log_trees(log_tree_root);
1895 BUG_ON(ret);
e556ce2c
YZ		 1896
1897 if (sb->s_flags & MS_RDONLY) {
1898 ret = btrfs_commit_super(tree_root);
1899 BUG_ON(ret);
1900 }
e02119d5 1901 }
1a40e23b 1902
7c2ca468
CM		 1903 if (!(sb->s_flags & MS_RDONLY)) {
1904 ret = btrfs_cleanup_reloc_trees(tree_root);
1905 BUG_ON(ret);
1906 }
1a40e23b 1907
3de4586c
CM		 1908 location.objectid = BTRFS_FS_TREE_OBJECTID;
1909 location.type = BTRFS_ROOT_ITEM_KEY;
1910 location.offset = (u64)-1;
1911
3de4586c
CM		 1912 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
1913 if (!fs_info->fs_root)
7c2ca468 1914 goto fail_trans_kthread;
0f7d52f4 1915 return tree_root;
39279cc3 1916
7c2ca468
CM		 1917fail_trans_kthread:
1918 kthread_stop(fs_info->transaction_kthread);
3f157a2f 1919fail_cleaner:
a74a4b97 1920 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 1921
1922 /*
1923 * make sure we're done with the btree inode before we stop our
1924 * kthreads
1925 */
1926 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
1927 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
1928
d20f7043
CM		 1929fail_csum_root:
1930 free_extent_buffer(csum_root->node);
0b86a832
CM		 1931fail_extent_root:
1932 free_extent_buffer(extent_root->node);
39279cc3 1933fail_tree_root:
5f39d397 1934 free_extent_buffer(tree_root->node);
2b82032c
YZ		 1935fail_chunk_root:
1936 free_extent_buffer(chunk_root->node);
84eed90f 1937fail_sys_array:
7c2ca468 1938 free_extent_buffer(dev_root->node);
39279cc3 1939fail_sb_buffer:
247e743c 1940 btrfs_stop_workers(&fs_info->fixup_workers);
771ed689 1941 btrfs_stop_workers(&fs_info->delalloc_workers);
8b712842
CM		 1942 btrfs_stop_workers(&fs_info->workers);
1943 btrfs_stop_workers(&fs_info->endio_workers);
d20f7043 1944 btrfs_stop_workers(&fs_info->endio_meta_workers);
cad321ad 1945 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
e6dcd2dc 1946 btrfs_stop_workers(&fs_info->endio_write_workers);
1cc127b5 1947 btrfs_stop_workers(&fs_info->submit_workers);
4543df7e 1948fail_iput:
7c2ca468 1949 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
4543df7e 1950 iput(fs_info->btree_inode);
7e662854 1951
dfe25020 1952 btrfs_close_devices(fs_info->fs_devices);
84eed90f 1953 btrfs_mapping_tree_free(&fs_info->mapping_tree);
7e662854 1954 bdi_destroy(&fs_info->bdi);
84eed90f 1955
7e662854 1956fail:
39279cc3
CM		 1957 kfree(extent_root);
1958 kfree(tree_root);
1959 kfree(fs_info);
83afeac4
JM		 1960 kfree(chunk_root);
1961 kfree(dev_root);
d20f7043 1962 kfree(csum_root);
39279cc3 1963 return ERR_PTR(err);
eb60ceac
CM		 1964}
1965
f2984462
CM		 1966static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
1967{
1968 char b[BDEVNAME_SIZE];
1969
1970 if (uptodate) {
1971 set_buffer_uptodate(bh);
1972 } else {
1973 if (!buffer_eopnotsupp(bh) && printk_ratelimit()) {
1974 printk(KERN_WARNING "lost page write due to "
1975 "I/O error on %s\n",
1976 bdevname(bh->b_bdev, b));
1977 }
1259ab75
CM		 1978 /* note, we dont' set_buffer_write_io_error because we have
1979 * our own ways of dealing with the IO errors
1980 */
f2984462
CM		 1981 clear_buffer_uptodate(bh);
1982 }
1983 unlock_buffer(bh);
1984 put_bh(bh);
1985}
1986
a512bbf8
YZ		 1987struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
1988{
1989 struct buffer_head *bh;
1990 struct buffer_head *latest = NULL;
1991 struct btrfs_super_block *super;
1992 int i;
1993 u64 transid = 0;
1994 u64 bytenr;
1995
1996 /* we would like to check all the supers, but that would make
1997 * a btrfs mount succeed after a mkfs from a different FS.
1998 * So, we need to add a special mount option to scan for
1999 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
2000 */
2001 for (i = 0; i < 1; i++) {
2002 bytenr = btrfs_sb_offset(i);
2003 if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
2004 break;
2005 bh = __bread(bdev, bytenr / 4096, 4096);
2006 if (!bh)
2007 continue;
2008
2009 super = (struct btrfs_super_block *)bh->b_data;
2010 if (btrfs_super_bytenr(super) != bytenr ||
2011 strncmp((char *)(&super->magic), BTRFS_MAGIC,
2012 sizeof(super->magic))) {
2013 brelse(bh);
2014 continue;
2015 }
2016
2017 if (!latest || btrfs_super_generation(super) > transid) {
2018 brelse(latest);
2019 latest = bh;
2020 transid = btrfs_super_generation(super);
2021 } else {
2022 brelse(bh);
2023 }
2024 }
2025 return latest;
2026}
2027
2028static int write_dev_supers(struct btrfs_device *device,
2029 struct btrfs_super_block *sb,
2030 int do_barriers, int wait, int max_mirrors)
2031{
2032 struct buffer_head *bh;
2033 int i;
2034 int ret;
2035 int errors = 0;
2036 u32 crc;
2037 u64 bytenr;
2038 int last_barrier = 0;
2039
2040 if (max_mirrors == 0)
2041 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
2042
2043 /* make sure only the last submit_bh does a barrier */
2044 if (do_barriers) {
2045 for (i = 0; i < max_mirrors; i++) {
2046 bytenr = btrfs_sb_offset(i);
2047 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
2048 device->total_bytes)
2049 break;
2050 last_barrier = i;
2051 }
2052 }
2053
2054 for (i = 0; i < max_mirrors; i++) {
2055 bytenr = btrfs_sb_offset(i);
2056 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
2057 break;
2058
2059 if (wait) {
2060 bh = __find_get_block(device->bdev, bytenr / 4096,
2061 BTRFS_SUPER_INFO_SIZE);
2062 BUG_ON(!bh);
2063 brelse(bh);
2064 wait_on_buffer(bh);
2065 if (buffer_uptodate(bh)) {
2066 brelse(bh);
2067 continue;
2068 }
2069 } else {
2070 btrfs_set_super_bytenr(sb, bytenr);
2071
2072 crc = ~(u32)0;
2073 crc = btrfs_csum_data(NULL, (char *)sb +
2074 BTRFS_CSUM_SIZE, crc,
2075 BTRFS_SUPER_INFO_SIZE -
2076 BTRFS_CSUM_SIZE);
2077 btrfs_csum_final(crc, sb->csum);
2078
2079 bh = __getblk(device->bdev, bytenr / 4096,
2080 BTRFS_SUPER_INFO_SIZE);
2081 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
2082
2083 set_buffer_uptodate(bh);
2084 get_bh(bh);
2085 lock_buffer(bh);
2086 bh->b_end_io = btrfs_end_buffer_write_sync;
2087 }
2088
2089 if (i == last_barrier && do_barriers && device->barriers) {
2090 ret = submit_bh(WRITE_BARRIER, bh);
2091 if (ret == -EOPNOTSUPP) {
2092 printk("btrfs: disabling barriers on dev %s\n",
2093 device->name);
2094 set_buffer_uptodate(bh);
2095 device->barriers = 0;
2096 get_bh(bh);
2097 lock_buffer(bh);
2098 ret = submit_bh(WRITE, bh);
2099 }
2100 } else {
2101 ret = submit_bh(WRITE, bh);
2102 }
2103
2104 if (!ret && wait) {
2105 wait_on_buffer(bh);
2106 if (!buffer_uptodate(bh))
2107 errors++;
2108 } else if (ret) {
2109 errors++;
2110 }
2111 if (wait)
2112 brelse(bh);
2113 }
2114 return errors < i ? 0 : -1;
2115}
2116
2117int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462 2118{
f2984462
CM		 2119 struct list_head *head = &root->fs_info->fs_devices->devices;
2120 struct btrfs_device *dev;
a061fc8d 2121 struct btrfs_super_block *sb;
f2984462 2122 struct btrfs_dev_item *dev_item;
f2984462
CM		 2123 int ret;
2124 int do_barriers;
a236aed1
CM		 2125 int max_errors;
2126 int total_errors = 0;
a061fc8d 2127 u64 flags;
f2984462 2128
a236aed1 2129 max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
f2984462
CM		 2130 do_barriers = !btrfs_test_opt(root, NOBARRIER);
2131
a061fc8d
CM		 2132 sb = &root->fs_info->super_for_commit;
2133 dev_item = &sb->dev_item;
c6e30871 2134 list_for_each_entry(dev, head, dev_list) {
dfe25020
CM		 2135 if (!dev->bdev) {
2136 total_errors++;
2137 continue;
2138 }
2b82032c 2139 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 2140 continue;
2141
2b82032c 2142 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 2143 btrfs_set_stack_device_type(dev_item, dev->type);
2144 btrfs_set_stack_device_id(dev_item, dev->devid);
2145 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
2146 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
2147 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
2148 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
2149 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
2150 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 2151 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 2152
a061fc8d
CM		 2153 flags = btrfs_super_flags(sb);
2154 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
2155
a512bbf8 2156 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 2157 if (ret)
2158 total_errors++;
f2984462 2159 }
a236aed1 2160 if (total_errors > max_errors) {
d397712b
CM		 2161 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2162 total_errors);
a236aed1
CM		 2163 BUG();
2164 }
f2984462 2165
a512bbf8 2166 total_errors = 0;
c6e30871 2167 list_for_each_entry(dev, head, dev_list) {
dfe25020
CM		 2168 if (!dev->bdev)
2169 continue;
2b82032c 2170 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 2171 continue;
2172
a512bbf8
YZ		 2173 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
2174 if (ret)
2175 total_errors++;
f2984462 2176 }
a236aed1 2177 if (total_errors > max_errors) {
d397712b
CM		 2178 printk(KERN_ERR "btrfs: %d errors while writing supers\n",
2179 total_errors);
a236aed1
CM		 2180 BUG();
2181 }
f2984462
CM		 2182 return 0;
2183}
2184
a512bbf8
YZ		 2185int write_ctree_super(struct btrfs_trans_handle *trans,
2186 struct btrfs_root *root, int max_mirrors)
eb60ceac 2187{
e66f709b 2188 int ret;
5f39d397 2189
a512bbf8 2190 ret = write_all_supers(root, max_mirrors);
5f39d397 2191 return ret;
cfaa7295
CM		 2192}
2193
5eda7b5e 2194int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
2619ba1f
CM		 2195{
2196 radix_tree_delete(&fs_info->fs_roots_radix,
2197 (unsigned long)root->root_key.objectid);
3394e160
CM		 2198 if (root->anon_super.s_dev) {
2199 down_write(&root->anon_super.s_umount);
2200 kill_anon_super(&root->anon_super);
2201 }
2619ba1f 2202 if (root->node)
5f39d397 2203 free_extent_buffer(root->node);
2619ba1f 2204 if (root->commit_root)
5f39d397 2205 free_extent_buffer(root->commit_root);
d397712b 2206 kfree(root->name);
2619ba1f
CM		 2207 kfree(root);
2208 return 0;
2209}
2210
35b7e476 2211static int del_fs_roots(struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 2212{
2213 int ret;
2214 struct btrfs_root *gang[8];
2215 int i;
2216
d397712b 2217 while (1) {
0f7d52f4
CM		 2218 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2219 (void **)gang, 0,
2220 ARRAY_SIZE(gang));
2221 if (!ret)
2222 break;
2619ba1f 2223 for (i = 0; i < ret; i++)
5eda7b5e 2224 btrfs_free_fs_root(fs_info, gang[i]);
0f7d52f4
CM		 2225 }
2226 return 0;
2227}
b4100d64 2228
c146afad 2229int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 2230{
c146afad
YZ		 2231 u64 root_objectid = 0;
2232 struct btrfs_root *gang[8];
2233 int i;
3768f368 2234 int ret;
e089f05c 2235
c146afad
YZ		 2236 while (1) {
2237 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2238 (void **)gang, root_objectid,
2239 ARRAY_SIZE(gang));
2240 if (!ret)
2241 break;
2242 for (i = 0; i < ret; i++) {
2243 root_objectid = gang[i]->root_key.objectid;
2244 ret = btrfs_find_dead_roots(fs_info->tree_root,
2245 root_objectid, gang[i]);
2246 BUG_ON(ret);
2247 btrfs_orphan_cleanup(gang[i]);
2248 }
2249 root_objectid++;
2250 }
2251 return 0;
2252}
a2135011 2253
c146afad
YZ		 2254int btrfs_commit_super(struct btrfs_root *root)
2255{
2256 struct btrfs_trans_handle *trans;
2257 int ret;
a74a4b97 2258
c146afad 2259 mutex_lock(&root->fs_info->cleaner_mutex);
a74a4b97 2260 btrfs_clean_old_snapshots(root);
c146afad 2261 mutex_unlock(&root->fs_info->cleaner_mutex);
79154b1b 2262 trans = btrfs_start_transaction(root, 1);
54aa1f4d 2263 ret = btrfs_commit_transaction(trans, root);
c146afad
YZ		 2264 BUG_ON(ret);
2265 /* run commit again to drop the original snapshot */
79154b1b
CM		 2266 trans = btrfs_start_transaction(root, 1);
2267 btrfs_commit_transaction(trans, root);
2268 ret = btrfs_write_and_wait_transaction(NULL, root);
3768f368 2269 BUG_ON(ret);
d6bfde87 2270
a512bbf8 2271 ret = write_ctree_super(NULL, root, 0);
c146afad
YZ		 2272 return ret;
2273}
2274
2275int close_ctree(struct btrfs_root *root)
2276{
2277 struct btrfs_fs_info *fs_info = root->fs_info;
2278 int ret;
2279
2280 fs_info->closing = 1;
2281 smp_mb();
2282
2283 kthread_stop(root->fs_info->transaction_kthread);
2284 kthread_stop(root->fs_info->cleaner_kthread);
2285
2286 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
2287 ret = btrfs_commit_super(root);
d397712b
CM		 2288 if (ret)
2289 printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
c146afad 2290 }
0f7d52f4 2291
b0c68f8b 2292 if (fs_info->delalloc_bytes) {
d397712b 2293 printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
b0c68f8b
CM		 2294 fs_info->delalloc_bytes);
2295 }
31153d81 2296 if (fs_info->total_ref_cache_size) {
d397712b
CM		 2297 printk(KERN_INFO "btrfs: at umount reference cache size %llu\n",
2298 (unsigned long long)fs_info->total_ref_cache_size);
31153d81 2299 }
bcc63abb 2300
0f7d52f4 2301 if (fs_info->extent_root->node)
5f39d397 2302 free_extent_buffer(fs_info->extent_root->node);
f510cfec 2303
0f7d52f4 2304 if (fs_info->tree_root->node)
5f39d397 2305 free_extent_buffer(fs_info->tree_root->node);
f510cfec 2306
d397712b 2307 if (root->fs_info->chunk_root->node)
0b86a832
CM		 2308 free_extent_buffer(root->fs_info->chunk_root->node);
2309
d397712b 2310 if (root->fs_info->dev_root->node)
0b86a832
CM		 2311 free_extent_buffer(root->fs_info->dev_root->node);
2312
d397712b 2313 if (root->fs_info->csum_root->node)
d20f7043
CM		 2314 free_extent_buffer(root->fs_info->csum_root->node);
2315
9078a3e1 2316 btrfs_free_block_groups(root->fs_info);
d10c5f31 2317
c146afad 2318 del_fs_roots(fs_info);
d10c5f31 2319
c146afad 2320 iput(fs_info->btree_inode);
9ad6b7bc 2321
247e743c 2322 btrfs_stop_workers(&fs_info->fixup_workers);
771ed689 2323 btrfs_stop_workers(&fs_info->delalloc_workers);
8b712842
CM		 2324 btrfs_stop_workers(&fs_info->workers);
2325 btrfs_stop_workers(&fs_info->endio_workers);
d20f7043 2326 btrfs_stop_workers(&fs_info->endio_meta_workers);
cad321ad 2327 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
e6dcd2dc 2328 btrfs_stop_workers(&fs_info->endio_write_workers);
1cc127b5 2329 btrfs_stop_workers(&fs_info->submit_workers);
d6bfde87 2330
19c00ddc 2331#if 0
d397712b 2332 while (!list_empty(&fs_info->hashers)) {
19c00ddc
CM		 2333 struct btrfs_hasher *hasher;
2334 hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher,
2335 hashers);
2336 list_del(&hasher->hashers);
2337 crypto_free_hash(&fs_info->hash_tfm);
2338 kfree(hasher);
2339 }
2340#endif
dfe25020 2341 btrfs_close_devices(fs_info->fs_devices);
0b86a832 2342 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 2343
04160088 2344 bdi_destroy(&fs_info->bdi);
0b86a832 2345
0f7d52f4 2346 kfree(fs_info->extent_root);
0f7d52f4 2347 kfree(fs_info->tree_root);
0b86a832
CM		 2348 kfree(fs_info->chunk_root);
2349 kfree(fs_info->dev_root);
d20f7043 2350 kfree(fs_info->csum_root);
eb60ceac
CM		 2351 return 0;
2352}
2353
1259ab75 2354int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
5f39d397 2355{
1259ab75 2356 int ret;
810191ff 2357 struct inode *btree_inode = buf->first_page->mapping->host;
1259ab75
CM		 2358
2359 ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf);
2360 if (!ret)
2361 return ret;
2362
2363 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
2364 parent_transid);
2365 return !ret;
5f39d397
CM		 2366}
2367
2368int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 2369{
810191ff 2370 struct inode *btree_inode = buf->first_page->mapping->host;
d1310b2e 2371 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
5f39d397
CM		 2372 buf);
2373}
6702ed49 2374
5f39d397
CM		 2375void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
2376{
810191ff 2377 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
5f39d397
CM		 2378 u64 transid = btrfs_header_generation(buf);
2379 struct inode *btree_inode = root->fs_info->btree_inode;
b9473439 2380 int was_dirty;
b4ce94de 2381
b9447ef8 2382 btrfs_assert_tree_locked(buf);
ccd467d6 2383 if (transid != root->fs_info->generation) {
d397712b
CM		 2384 printk(KERN_CRIT "btrfs transid mismatch buffer %llu, "
2385 "found %llu running %llu\n",
db94535d 2386 (unsigned long long)buf->start,
d397712b
CM		 2387 (unsigned long long)transid,
2388 (unsigned long long)root->fs_info->generation);
ccd467d6
CM		 2389 WARN_ON(1);
2390 }
b9473439
CM		 2391 was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
2392 buf);
2393 if (!was_dirty) {
2394 spin_lock(&root->fs_info->delalloc_lock);
2395 root->fs_info->dirty_metadata_bytes += buf->len;
2396 spin_unlock(&root->fs_info->delalloc_lock);
2397 }
eb60ceac
CM		 2398}
2399
d3c2fdcf 2400void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
35b7e476 2401{
188de649
CM		 2402 /*
2403 * looks as though older kernels can get into trouble with
2404 * this code, they end up stuck in balance_dirty_pages forever
2405 */
d6bfde87
CM		 2406 struct extent_io_tree *tree;
2407 u64 num_dirty;
2408 u64 start = 0;
771ed689 2409 unsigned long thresh = 32 * 1024 * 1024;
d6bfde87
CM		 2410 tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
2411
6933c02e 2412 if (current->flags & PF_MEMALLOC)
d6bfde87
CM		 2413 return;
2414
2415 num_dirty = count_range_bits(tree, &start, (u64)-1,
2416 thresh, EXTENT_DIRTY);
2417 if (num_dirty > thresh) {
2418 balance_dirty_pages_ratelimited_nr(
d7fc640e 2419 root->fs_info->btree_inode->i_mapping, 1);
d6bfde87 2420 }
188de649 2421 return;
35b7e476 2422}
6b80053d 2423
ca7a79ad 2424int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 2425{
810191ff 2426 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
ce9adaa5 2427 int ret;
ca7a79ad 2428 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
d397712b 2429 if (ret == 0)
b4ce94de 2430 set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
ce9adaa5 2431 return ret;
6b80053d 2432}
0da5468f 2433
4bef0848
CM		 2434int btree_lock_page_hook(struct page *page)
2435{
2436 struct inode *inode = page->mapping->host;
b9473439 2437 struct btrfs_root *root = BTRFS_I(inode)->root;
4bef0848
CM		 2438 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2439 struct extent_buffer *eb;
2440 unsigned long len;
2441 u64 bytenr = page_offset(page);
2442
2443 if (page->private == EXTENT_PAGE_PRIVATE)
2444 goto out;
2445
2446 len = page->private >> 2;
2447 eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS);
2448 if (!eb)
2449 goto out;
2450
2451 btrfs_tree_lock(eb);
4bef0848 2452 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
b9473439
CM		 2453
2454 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
2455 spin_lock(&root->fs_info->delalloc_lock);
2456 if (root->fs_info->dirty_metadata_bytes >= eb->len)
2457 root->fs_info->dirty_metadata_bytes -= eb->len;
2458 else
2459 WARN_ON(1);
2460 spin_unlock(&root->fs_info->delalloc_lock);
2461 }
2462
4bef0848
CM		 2463 btrfs_tree_unlock(eb);
2464 free_extent_buffer(eb);
2465out:
2466 lock_page(page);
2467 return 0;
2468}
2469
d1310b2e 2470static struct extent_io_ops btree_extent_io_ops = {
4bef0848 2471 .write_cache_pages_lock_hook = btree_lock_page_hook,
ce9adaa5 2472 .readpage_end_io_hook = btree_readpage_end_io_hook,
0b86a832 2473 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 2474 /* note we're sharing with inode.c for the merge bio hook */
2475 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 2476};
Linux 6.x block layer and io_uring tree(s)