projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: Fix compile warning around num_online_cpus() in a min statement
[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
a74a4b97 19#include <linux/version.h>
e20d96d6 20#include <linux/fs.h>
d98237b3 21#include <linux/blkdev.h>
87cbda5c 22#include <linux/scatterlist.h>
22b0ebda 23#include <linux/swap.h>
0f7d52f4 24#include <linux/radix-tree.h>
35b7e476 25#include <linux/writeback.h>
5f39d397 26#include <linux/buffer_head.h> // for block_sync_page
ce9adaa5 27#include <linux/workqueue.h>
a74a4b97 28#include <linux/kthread.h>
4b4e25f2
CM		 29#include <linux/freezer.h>
30#include "compat.h"
a5eb62e3 31#include "crc32c.h"
eb60ceac
CM		 32#include "ctree.h"
33#include "disk-io.h"
e089f05c 34#include "transaction.h"
0f7d52f4 35#include "btrfs_inode.h"
0b86a832 36#include "volumes.h"
db94535d 37#include "print-tree.h"
8b712842 38#include "async-thread.h"
925baedd 39#include "locking.h"
017e5369 40#include "ref-cache.h"
e02119d5 41#include "tree-log.h"
eb60ceac 42
5f39d397
CM		 43#if 0
44static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
7eccb903 45{
5f39d397
CM		 46 if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) {
47 printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n",
48 (unsigned long long)extent_buffer_blocknr(buf),
49 (unsigned long long)btrfs_header_blocknr(buf));
39279cc3 50 return 1;
d98237b3 51 }
9a8dd150 52 return 0;
eb60ceac 53}
5f39d397 54#endif
eb60ceac 55
d1310b2e 56static struct extent_io_ops btree_extent_io_ops;
8b712842 57static void end_workqueue_fn(struct btrfs_work *work);
ce9adaa5 58
d352ac68
CM		 59/*
60 * end_io_wq structs are used to do processing in task context when an IO is
61 * complete. This is used during reads to verify checksums, and it is used
62 * by writes to insert metadata for new file extents after IO is complete.
63 */
ce9adaa5
CM		 64struct end_io_wq {
65 struct bio *bio;
66 bio_end_io_t *end_io;
67 void *private;
68 struct btrfs_fs_info *info;
69 int error;
22c59948 70 int metadata;
ce9adaa5 71 struct list_head list;
8b712842 72 struct btrfs_work work;
ce9adaa5 73};
0da5468f 74
d352ac68
CM		 75/*
76 * async submit bios are used to offload expensive checksumming
77 * onto the worker threads. They checksum file and metadata bios
78 * just before they are sent down the IO stack.
79 */
44b8bd7e
CM		 80struct async_submit_bio {
81 struct inode *inode;
82 struct bio *bio;
83 struct list_head list;
4a69a410
CM		 84 extent_submit_bio_hook_t *submit_bio_start;
85 extent_submit_bio_hook_t *submit_bio_done;
44b8bd7e
CM		 86 int rw;
87 int mirror_num;
c8b97818 88 unsigned long bio_flags;
8b712842 89 struct btrfs_work work;
44b8bd7e
CM		 90};
91
d352ac68
CM		 92/*
93 * extents on the btree inode are pretty simple, there's one extent
94 * that covers the entire device
95 */
b2950863
CH		 96static struct extent_map *btree_get_extent(struct inode *inode,
97 struct page *page, size_t page_offset, u64 start, u64 len,
98 int create)
7eccb903 99{
5f39d397
CM		 100 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
101 struct extent_map *em;
102 int ret;
103
d1310b2e
CM		 104 spin_lock(&em_tree->lock);
105 em = lookup_extent_mapping(em_tree, start, len);
a061fc8d
CM		 106 if (em) {
107 em->bdev =
108 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
109 spin_unlock(&em_tree->lock);
5f39d397 110 goto out;
a061fc8d
CM		 111 }
112 spin_unlock(&em_tree->lock);
7b13b7b1 113
5f39d397
CM		 114 em = alloc_extent_map(GFP_NOFS);
115 if (!em) {
116 em = ERR_PTR(-ENOMEM);
117 goto out;
118 }
119 em->start = 0;
0afbaf8c 120 em->len = (u64)-1;
c8b97818 121 em->block_len = (u64)-1;
5f39d397 122 em->block_start = 0;
a061fc8d 123 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
d1310b2e
CM		 124
125 spin_lock(&em_tree->lock);
5f39d397
CM		 126 ret = add_extent_mapping(em_tree, em);
127 if (ret == -EEXIST) {
0afbaf8c
CM		 128 u64 failed_start = em->start;
129 u64 failed_len = em->len;
130
131 printk("failed to insert %Lu %Lu -> %Lu into tree\n",
132 em->start, em->len, em->block_start);
5f39d397 133 free_extent_map(em);
7b13b7b1 134 em = lookup_extent_mapping(em_tree, start, len);
0afbaf8c
CM		 135 if (em) {
136 printk("after failing, found %Lu %Lu %Lu\n",
137 em->start, em->len, em->block_start);
7b13b7b1 138 ret = 0;
0afbaf8c
CM		 139 } else {
140 em = lookup_extent_mapping(em_tree, failed_start,
141 failed_len);
142 if (em) {
143 printk("double failure lookup gives us "
144 "%Lu %Lu -> %Lu\n", em->start,
145 em->len, em->block_start);
146 free_extent_map(em);
147 }
7b13b7b1 148 ret = -EIO;
0afbaf8c 149 }
5f39d397 150 } else if (ret) {
7b13b7b1
CM		 151 free_extent_map(em);
152 em = NULL;
5f39d397 153 }
7b13b7b1
CM		 154 spin_unlock(&em_tree->lock);
155
156 if (ret)
157 em = ERR_PTR(ret);
5f39d397
CM		 158out:
159 return em;
7eccb903
CM		 160}
161
19c00ddc
CM		 162u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
163{
a5eb62e3 164 return btrfs_crc32c(seed, data, len);
19c00ddc
CM		 165}
166
167void btrfs_csum_final(u32 crc, char *result)
168{
169 *(__le32 *)result = ~cpu_to_le32(crc);
170}
171
d352ac68
CM		 172/*
173 * compute the csum for a btree block, and either verify it or write it
174 * into the csum field of the block.
175 */
19c00ddc
CM		 176static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
177 int verify)
178{
607d432d
JB		 179 u16 csum_size =
180 btrfs_super_csum_size(&root->fs_info->super_copy);
181 char *result = NULL;
19c00ddc
CM		 182 unsigned long len;
183 unsigned long cur_len;
184 unsigned long offset = BTRFS_CSUM_SIZE;
185 char *map_token = NULL;
186 char *kaddr;
187 unsigned long map_start;
188 unsigned long map_len;
189 int err;
190 u32 crc = ~(u32)0;
607d432d 191 unsigned long inline_result;
19c00ddc
CM		 192
193 len = buf->len - offset;
194 while(len > 0) {
195 err = map_private_extent_buffer(buf, offset, 32,
196 &map_token, &kaddr,
197 &map_start, &map_len, KM_USER0);
198 if (err) {
199 printk("failed to map extent buffer! %lu\n",
200 offset);
201 return 1;
202 }
203 cur_len = min(len, map_len - (offset - map_start));
204 crc = btrfs_csum_data(root, kaddr + offset - map_start,
205 crc, cur_len);
206 len -= cur_len;
207 offset += cur_len;
208 unmap_extent_buffer(buf, map_token, KM_USER0);
209 }
607d432d
JB		 210 if (csum_size > sizeof(inline_result)) {
211 result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
212 if (!result)
213 return 1;
214 } else {
215 result = (char *)&inline_result;
216 }
217
19c00ddc
CM		 218 btrfs_csum_final(crc, result);
219
220 if (verify) {
e4204ded 221 /* FIXME, this is not good */
607d432d 222 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
e4204ded
CM		 223 u32 val;
224 u32 found = 0;
607d432d 225 memcpy(&found, result, csum_size);
e4204ded 226
607d432d 227 read_extent_buffer(buf, &val, 0, csum_size);
e4204ded 228 printk("btrfs: %s checksum verify failed on %llu "
2dd3e67b 229 "wanted %X found %X level %d\n",
19c00ddc 230 root->fs_info->sb->s_id,
2dd3e67b 231 buf->start, val, found, btrfs_header_level(buf));
607d432d
JB		 232 if (result != (char *)&inline_result)
233 kfree(result);
19c00ddc
CM		 234 return 1;
235 }
236 } else {
607d432d 237 write_extent_buffer(buf, result, 0, csum_size);
19c00ddc 238 }
607d432d
JB		 239 if (result != (char *)&inline_result)
240 kfree(result);
19c00ddc
CM		 241 return 0;
242}
243
d352ac68
CM		 244/*
245 * we can't consider a given block up to date unless the transid of the
246 * block matches the transid in the parent node's pointer. This is how we
247 * detect blocks that either didn't get written at all or got written
248 * in the wrong place.
249 */
1259ab75
CM		 250static int verify_parent_transid(struct extent_io_tree *io_tree,
251 struct extent_buffer *eb, u64 parent_transid)
252{
253 int ret;
254
255 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
256 return 0;
257
258 lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS);
259 if (extent_buffer_uptodate(io_tree, eb) &&
260 btrfs_header_generation(eb) == parent_transid) {
261 ret = 0;
262 goto out;
263 }
264 printk("parent transid verify failed on %llu wanted %llu found %llu\n",
265 (unsigned long long)eb->start,
266 (unsigned long long)parent_transid,
267 (unsigned long long)btrfs_header_generation(eb));
268 ret = 1;
1259ab75 269 clear_extent_buffer_uptodate(io_tree, eb);
33958dc6 270out:
1259ab75
CM		 271 unlock_extent(io_tree, eb->start, eb->start + eb->len - 1,
272 GFP_NOFS);
273 return ret;
1259ab75
CM		 274}
275
d352ac68
CM		 276/*
277 * helper to read a given tree block, doing retries as required when
278 * the checksums don't match and we have alternate mirrors to try.
279 */
f188591e
CM		 280static int btree_read_extent_buffer_pages(struct btrfs_root *root,
281 struct extent_buffer *eb,
ca7a79ad 282 u64 start, u64 parent_transid)
f188591e
CM		 283{
284 struct extent_io_tree *io_tree;
285 int ret;
286 int num_copies = 0;
287 int mirror_num = 0;
288
289 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
290 while (1) {
291 ret = read_extent_buffer_pages(io_tree, eb, start, 1,
292 btree_get_extent, mirror_num);
1259ab75
CM		 293 if (!ret &&
294 !verify_parent_transid(io_tree, eb, parent_transid))
f188591e 295 return ret;
a1b32a59 296printk("read extent buffer pages failed with ret %d mirror no %d\n", ret, mirror_num);
f188591e
CM		 297 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
298 eb->start, eb->len);
4235298e 299 if (num_copies == 1)
f188591e 300 return ret;
4235298e 301
f188591e 302 mirror_num++;
4235298e 303 if (mirror_num > num_copies)
f188591e 304 return ret;
f188591e 305 }
f188591e
CM		 306 return -EIO;
307}
19c00ddc 308
d352ac68
CM		 309/*
310 * checksum a dirty tree block before IO. This has extra checks to make
311 * sure we only fill in the checksum field in the first page of a multi-page block
312 */
b2950863 313static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
19c00ddc 314{
d1310b2e 315 struct extent_io_tree *tree;
35ebb934 316 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
19c00ddc
CM		 317 u64 found_start;
318 int found_level;
319 unsigned long len;
320 struct extent_buffer *eb;
f188591e
CM		 321 int ret;
322
d1310b2e 323 tree = &BTRFS_I(page->mapping->host)->io_tree;
19c00ddc
CM		 324
325 if (page->private == EXTENT_PAGE_PRIVATE)
326 goto out;
327 if (!page->private)
328 goto out;
329 len = page->private >> 2;
330 if (len == 0) {
331 WARN_ON(1);
332 }
333 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
ca7a79ad
CM		 334 ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
335 btrfs_header_generation(eb));
f188591e 336 BUG_ON(ret);
19c00ddc
CM		 337 found_start = btrfs_header_bytenr(eb);
338 if (found_start != start) {
339 printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
340 start, found_start, len);
55c69072
CM		 341 WARN_ON(1);
342 goto err;
343 }
344 if (eb->first_page != page) {
345 printk("bad first page %lu %lu\n", eb->first_page->index,
346 page->index);
347 WARN_ON(1);
348 goto err;
349 }
350 if (!PageUptodate(page)) {
351 printk("csum not up to date page %lu\n", page->index);
352 WARN_ON(1);
353 goto err;
19c00ddc
CM		 354 }
355 found_level = btrfs_header_level(eb);
4bef0848 356
19c00ddc 357 csum_tree_block(root, eb, 0);
55c69072 358err:
19c00ddc
CM		 359 free_extent_buffer(eb);
360out:
361 return 0;
362}
363
2b82032c
YZ		 364static int check_tree_block_fsid(struct btrfs_root *root,
365 struct extent_buffer *eb)
366{
367 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
368 u8 fsid[BTRFS_UUID_SIZE];
369 int ret = 1;
370
371 read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
372 BTRFS_FSID_SIZE);
373 while (fs_devices) {
374 if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
375 ret = 0;
376 break;
377 }
378 fs_devices = fs_devices->seed;
379 }
380 return ret;
381}
382
b2950863 383static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
ce9adaa5
CM		 384 struct extent_state *state)
385{
386 struct extent_io_tree *tree;
387 u64 found_start;
388 int found_level;
389 unsigned long len;
390 struct extent_buffer *eb;
391 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
f188591e 392 int ret = 0;
ce9adaa5
CM		 393
394 tree = &BTRFS_I(page->mapping->host)->io_tree;
395 if (page->private == EXTENT_PAGE_PRIVATE)
396 goto out;
397 if (!page->private)
398 goto out;
399 len = page->private >> 2;
400 if (len == 0) {
401 WARN_ON(1);
402 }
403 eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
f188591e 404
ce9adaa5 405 found_start = btrfs_header_bytenr(eb);
23a07867 406 if (found_start != start) {
a1b32a59
CM		 407 printk("bad tree block start %llu %llu\n",
408 (unsigned long long)found_start,
409 (unsigned long long)eb->start);
f188591e 410 ret = -EIO;
ce9adaa5
CM		 411 goto err;
412 }
413 if (eb->first_page != page) {
414 printk("bad first page %lu %lu\n", eb->first_page->index,
415 page->index);
416 WARN_ON(1);
f188591e 417 ret = -EIO;
ce9adaa5
CM		 418 goto err;
419 }
2b82032c 420 if (check_tree_block_fsid(root, eb)) {
1259ab75
CM		 421 printk("bad fsid on block %Lu\n", eb->start);
422 ret = -EIO;
423 goto err;
424 }
ce9adaa5
CM		 425 found_level = btrfs_header_level(eb);
426
427 ret = csum_tree_block(root, eb, 1);
f188591e
CM		 428 if (ret)
429 ret = -EIO;
ce9adaa5
CM		 430
431 end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
432 end = eb->start + end - 1;
ce9adaa5
CM		 433err:
434 free_extent_buffer(eb);
435out:
f188591e 436 return ret;
ce9adaa5
CM		 437}
438
ce9adaa5 439static void end_workqueue_bio(struct bio *bio, int err)
ce9adaa5
CM		 440{
441 struct end_io_wq *end_io_wq = bio->bi_private;
442 struct btrfs_fs_info *fs_info;
ce9adaa5 443
ce9adaa5 444 fs_info = end_io_wq->info;
ce9adaa5 445 end_io_wq->error = err;
8b712842
CM		 446 end_io_wq->work.func = end_workqueue_fn;
447 end_io_wq->work.flags = 0;
d20f7043
CM		 448
449 if (bio->bi_rw & (1 << BIO_RW)) {
cad321ad
CM		 450 if (end_io_wq->metadata)
451 btrfs_queue_worker(&fs_info->endio_meta_write_workers,
452 &end_io_wq->work);
453 else
454 btrfs_queue_worker(&fs_info->endio_write_workers,
455 &end_io_wq->work);
d20f7043
CM		 456 } else {
457 if (end_io_wq->metadata)
458 btrfs_queue_worker(&fs_info->endio_meta_workers,
459 &end_io_wq->work);
460 else
461 btrfs_queue_worker(&fs_info->endio_workers,
462 &end_io_wq->work);
463 }
ce9adaa5
CM		 464}
465
22c59948
CM		 466int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
467 int metadata)
0b86a832 468{
ce9adaa5 469 struct end_io_wq *end_io_wq;
ce9adaa5
CM		 470 end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
471 if (!end_io_wq)
472 return -ENOMEM;
473
474 end_io_wq->private = bio->bi_private;
475 end_io_wq->end_io = bio->bi_end_io;
22c59948 476 end_io_wq->info = info;
ce9adaa5
CM		 477 end_io_wq->error = 0;
478 end_io_wq->bio = bio;
22c59948 479 end_io_wq->metadata = metadata;
ce9adaa5
CM		 480
481 bio->bi_private = end_io_wq;
482 bio->bi_end_io = end_workqueue_bio;
22c59948
CM		 483 return 0;
484}
485
b64a2851 486unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
0986fe9e 487{
4854ddd0
CM		 488 unsigned long limit = min_t(unsigned long,
489 info->workers.max_workers,
490 info->fs_devices->open_devices);
491 return 256 * limit;
492}
0986fe9e 493
4854ddd0
CM		 494int btrfs_congested_async(struct btrfs_fs_info *info, int iodone)
495{
b64a2851
CM		 496 return atomic_read(&info->nr_async_bios) >
497 btrfs_async_submit_limit(info);
0986fe9e
CM		 498}
499
4a69a410
CM		 500static void run_one_async_start(struct btrfs_work *work)
501{
502 struct btrfs_fs_info *fs_info;
503 struct async_submit_bio *async;
504
505 async = container_of(work, struct async_submit_bio, work);
506 fs_info = BTRFS_I(async->inode)->root->fs_info;
507 async->submit_bio_start(async->inode, async->rw, async->bio,
508 async->mirror_num, async->bio_flags);
509}
510
511static void run_one_async_done(struct btrfs_work *work)
8b712842
CM		 512{
513 struct btrfs_fs_info *fs_info;
514 struct async_submit_bio *async;
4854ddd0 515 int limit;
8b712842
CM		 516
517 async = container_of(work, struct async_submit_bio, work);
518 fs_info = BTRFS_I(async->inode)->root->fs_info;
4854ddd0 519
b64a2851 520 limit = btrfs_async_submit_limit(fs_info);
4854ddd0
CM		 521 limit = limit * 2 / 3;
522
8b712842 523 atomic_dec(&fs_info->nr_async_submits);
0986fe9e 524
b64a2851
CM		 525 if (atomic_read(&fs_info->nr_async_submits) < limit &&
526 waitqueue_active(&fs_info->async_submit_wait))
4854ddd0
CM		 527 wake_up(&fs_info->async_submit_wait);
528
4a69a410 529 async->submit_bio_done(async->inode, async->rw, async->bio,
c8b97818 530 async->mirror_num, async->bio_flags);
4a69a410
CM		 531}
532
533static void run_one_async_free(struct btrfs_work *work)
534{
535 struct async_submit_bio *async;
536
537 async = container_of(work, struct async_submit_bio, work);
8b712842
CM		 538 kfree(async);
539}
540
44b8bd7e
CM		 541int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
542 int rw, struct bio *bio, int mirror_num,
c8b97818 543 unsigned long bio_flags,
4a69a410
CM		 544 extent_submit_bio_hook_t *submit_bio_start,
545 extent_submit_bio_hook_t *submit_bio_done)
44b8bd7e
CM		 546{
547 struct async_submit_bio *async;
548
549 async = kmalloc(sizeof(*async), GFP_NOFS);
550 if (!async)
551 return -ENOMEM;
552
553 async->inode = inode;
554 async->rw = rw;
555 async->bio = bio;
556 async->mirror_num = mirror_num;
4a69a410
CM		 557 async->submit_bio_start = submit_bio_start;
558 async->submit_bio_done = submit_bio_done;
559
560 async->work.func = run_one_async_start;
561 async->work.ordered_func = run_one_async_done;
562 async->work.ordered_free = run_one_async_free;
563
8b712842 564 async->work.flags = 0;
c8b97818 565 async->bio_flags = bio_flags;
8c8bee1d 566
cb03c743 567 atomic_inc(&fs_info->nr_async_submits);
8b712842 568 btrfs_queue_worker(&fs_info->workers, &async->work);
d2c3f4f6 569#if 0
7c2ca468 570 int limit = btrfs_async_submit_limit(fs_info);
9473f16c
CM		 571 if (atomic_read(&fs_info->nr_async_submits) > limit) {
572 wait_event_timeout(fs_info->async_submit_wait,
4854ddd0
CM		 573 (atomic_read(&fs_info->nr_async_submits) < limit),
574 HZ/10);
9473f16c
CM		 575
576 wait_event_timeout(fs_info->async_submit_wait,
577 (atomic_read(&fs_info->nr_async_bios) < limit),
578 HZ/10);
579 }
d2c3f4f6 580#endif
771ed689
CM		 581 while(atomic_read(&fs_info->async_submit_draining) &&
582 atomic_read(&fs_info->nr_async_submits)) {
583 wait_event(fs_info->async_submit_wait,
584 (atomic_read(&fs_info->nr_async_submits) == 0));
585 }
586
44b8bd7e
CM		 587 return 0;
588}
589
ce3ed71a
CM		 590static int btree_csum_one_bio(struct bio *bio)
591{
592 struct bio_vec *bvec = bio->bi_io_vec;
593 int bio_index = 0;
594 struct btrfs_root *root;
595
596 WARN_ON(bio->bi_vcnt <= 0);
597 while(bio_index < bio->bi_vcnt) {
598 root = BTRFS_I(bvec->bv_page->mapping->host)->root;
599 csum_dirty_buffer(root, bvec->bv_page);
600 bio_index++;
601 bvec++;
602 }
603 return 0;
604}
605
4a69a410
CM		 606static int __btree_submit_bio_start(struct inode *inode, int rw,
607 struct bio *bio, int mirror_num,
608 unsigned long bio_flags)
22c59948 609{
8b712842
CM		 610 /*
611 * when we're called for a write, we're already in the async
5443be45 612 * submission context. Just jump into btrfs_map_bio
8b712842 613 */
4a69a410
CM		 614 btree_csum_one_bio(bio);
615 return 0;
616}
22c59948 617
4a69a410
CM		 618static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
619 int mirror_num, unsigned long bio_flags)
620{
8b712842 621 /*
4a69a410
CM		 622 * when we're called for a write, we're already in the async
623 * submission context. Just jump into btrfs_map_bio
8b712842 624 */
8b712842 625 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
0b86a832
CM		 626}
627
44b8bd7e 628static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
c8b97818 629 int mirror_num, unsigned long bio_flags)
44b8bd7e 630{
cad321ad
CM		 631 int ret;
632
633 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
634 bio, 1);
635 BUG_ON(ret);
636
44b8bd7e 637 if (!(rw & (1 << BIO_RW))) {
4a69a410
CM		 638 /*
639 * called for a read, do the setup so that checksum validation
640 * can happen in the async kernel threads
641 */
4a69a410 642 return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
6f3577bd 643 mirror_num, 0);
44b8bd7e 644 }
cad321ad
CM		 645 /*
646 * kthread helpers are used to submit writes so that checksumming
647 * can happen in parallel across all CPUs
648 */
44b8bd7e 649 return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
c8b97818 650 inode, rw, bio, mirror_num, 0,
4a69a410
CM		 651 __btree_submit_bio_start,
652 __btree_submit_bio_done);
44b8bd7e
CM		 653}
654
0da5468f
CM		 655static int btree_writepage(struct page *page, struct writeback_control *wbc)
656{
d1310b2e
CM		 657 struct extent_io_tree *tree;
658 tree = &BTRFS_I(page->mapping->host)->io_tree;
5443be45
CM		 659
660 if (current->flags & PF_MEMALLOC) {
661 redirty_page_for_writepage(wbc, page);
662 unlock_page(page);
663 return 0;
664 }
5f39d397
CM		 665 return extent_write_full_page(tree, page, btree_get_extent, wbc);
666}
0da5468f
CM		 667
668static int btree_writepages(struct address_space *mapping,
669 struct writeback_control *wbc)
670{
d1310b2e
CM		 671 struct extent_io_tree *tree;
672 tree = &BTRFS_I(mapping->host)->io_tree;
d8d5f3e1 673 if (wbc->sync_mode == WB_SYNC_NONE) {
793955bc
CM		 674 u64 num_dirty;
675 u64 start = 0;
24ab9cd8 676 unsigned long thresh = 32 * 1024 * 1024;
448d640b
CM		 677
678 if (wbc->for_kupdate)
679 return 0;
680
1832a6d5
CM		 681 num_dirty = count_range_bits(tree, &start, (u64)-1,
682 thresh, EXTENT_DIRTY);
793955bc
CM		 683 if (num_dirty < thresh) {
684 return 0;
685 }
686 }
0da5468f
CM		 687 return extent_writepages(tree, mapping, btree_get_extent, wbc);
688}
689
b2950863 690static int btree_readpage(struct file *file, struct page *page)
5f39d397 691{
d1310b2e
CM		 692 struct extent_io_tree *tree;
693 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 694 return extent_read_full_page(tree, page, btree_get_extent);
695}
22b0ebda 696
70dec807 697static int btree_releasepage(struct page *page, gfp_t gfp_flags)
5f39d397 698{
d1310b2e
CM		 699 struct extent_io_tree *tree;
700 struct extent_map_tree *map;
5f39d397 701 int ret;
d98237b3 702
98509cfc
CM		 703 if (PageWriteback(page) || PageDirty(page))
704 return 0;
705
d1310b2e
CM		 706 tree = &BTRFS_I(page->mapping->host)->io_tree;
707 map = &BTRFS_I(page->mapping->host)->extent_tree;
6af118ce 708
7b13b7b1 709 ret = try_release_extent_state(map, tree, page, gfp_flags);
6af118ce
CM		 710 if (!ret) {
711 return 0;
712 }
713
714 ret = try_release_extent_buffer(tree, page);
5f39d397
CM		 715 if (ret == 1) {
716 ClearPagePrivate(page);
717 set_page_private(page, 0);
718 page_cache_release(page);
719 }
6af118ce 720
d98237b3
CM		 721 return ret;
722}
723
5f39d397 724static void btree_invalidatepage(struct page *page, unsigned long offset)
d98237b3 725{
d1310b2e
CM		 726 struct extent_io_tree *tree;
727 tree = &BTRFS_I(page->mapping->host)->io_tree;
5f39d397
CM		 728 extent_invalidatepage(tree, page, offset);
729 btree_releasepage(page, GFP_NOFS);
9ad6b7bc 730 if (PagePrivate(page)) {
6af118ce
CM		 731 printk("warning page private not zero on page %Lu\n",
732 page_offset(page));
9ad6b7bc
CM		 733 ClearPagePrivate(page);
734 set_page_private(page, 0);
735 page_cache_release(page);
736 }
d98237b3
CM		 737}
738
5f39d397 739#if 0
d98237b3 740static int btree_writepage(struct page *page, struct writeback_control *wbc)
ed2ff2cb 741{
87cbda5c 742 struct buffer_head *bh;
0f7d52f4 743 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
87cbda5c 744 struct buffer_head *head;
87cbda5c
CM		 745 if (!page_has_buffers(page)) {
746 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
747 (1 << BH_Dirty)|(1 << BH_Uptodate));
748 }
749 head = page_buffers(page);
750 bh = head;
751 do {
752 if (buffer_dirty(bh))
753 csum_tree_block(root, bh, 0);
754 bh = bh->b_this_page;
755 } while (bh != head);
d98237b3 756 return block_write_full_page(page, btree_get_block, wbc);
ed2ff2cb 757}
5f39d397 758#endif
eb60ceac 759
d98237b3
CM		 760static struct address_space_operations btree_aops = {
761 .readpage = btree_readpage,
762 .writepage = btree_writepage,
0da5468f 763 .writepages = btree_writepages,
5f39d397
CM		 764 .releasepage = btree_releasepage,
765 .invalidatepage = btree_invalidatepage,
d98237b3
CM		 766 .sync_page = block_sync_page,
767};
768
ca7a79ad
CM		 769int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
770 u64 parent_transid)
090d1875 771{
5f39d397
CM		 772 struct extent_buffer *buf = NULL;
773 struct inode *btree_inode = root->fs_info->btree_inode;
de428b63 774 int ret = 0;
090d1875 775
db94535d 776 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
5f39d397 777 if (!buf)
090d1875 778 return 0;
d1310b2e 779 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
f188591e 780 buf, 0, 0, btree_get_extent, 0);
5f39d397 781 free_extent_buffer(buf);
de428b63 782 return ret;
090d1875
CM		 783}
784
0999df54
CM		 785struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
786 u64 bytenr, u32 blocksize)
787{
788 struct inode *btree_inode = root->fs_info->btree_inode;
789 struct extent_buffer *eb;
790 eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
791 bytenr, blocksize, GFP_NOFS);
792 return eb;
793}
794
795struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
796 u64 bytenr, u32 blocksize)
797{
798 struct inode *btree_inode = root->fs_info->btree_inode;
799 struct extent_buffer *eb;
800
801 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
802 bytenr, blocksize, NULL, GFP_NOFS);
803 return eb;
804}
805
806
e02119d5
CM		 807int btrfs_write_tree_block(struct extent_buffer *buf)
808{
809 return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start,
24ab9cd8 810 buf->start + buf->len - 1, WB_SYNC_ALL);
e02119d5
CM		 811}
812
813int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
814{
815 return btrfs_wait_on_page_writeback_range(buf->first_page->mapping,
816 buf->start, buf->start + buf->len -1);
817}
818
0999df54 819struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
ca7a79ad 820 u32 blocksize, u64 parent_transid)
0999df54
CM		 821{
822 struct extent_buffer *buf = NULL;
823 struct inode *btree_inode = root->fs_info->btree_inode;
824 struct extent_io_tree *io_tree;
825 int ret;
826
827 io_tree = &BTRFS_I(btree_inode)->io_tree;
828
829 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
830 if (!buf)
831 return NULL;
0999df54 832
ca7a79ad 833 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
ce9adaa5
CM		 834
835 if (ret == 0) {
836 buf->flags |= EXTENT_UPTODATE;
a1b32a59
CM		 837 } else {
838 WARN_ON(1);
ce9adaa5 839 }
5f39d397 840 return buf;
ce9adaa5 841
eb60ceac
CM		 842}
843
e089f05c 844int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
5f39d397 845 struct extent_buffer *buf)
ed2ff2cb 846{
5f39d397 847 struct inode *btree_inode = root->fs_info->btree_inode;
55c69072 848 if (btrfs_header_generation(buf) ==
925baedd
CM		 849 root->fs_info->running_transaction->transid) {
850 WARN_ON(!btrfs_tree_locked(buf));
d1310b2e 851 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
55c69072 852 buf);
925baedd 853 }
5f39d397
CM		 854 return 0;
855}
856
db94535d 857static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
87ee04eb 858 u32 stripesize, struct btrfs_root *root,
9f5fae2f 859 struct btrfs_fs_info *fs_info,
e20d96d6 860 u64 objectid)
d97e63b6 861{
cfaa7295 862 root->node = NULL;
a28ec197 863 root->commit_root = NULL;
31153d81 864 root->ref_tree = NULL;
db94535d
CM		 865 root->sectorsize = sectorsize;
866 root->nodesize = nodesize;
867 root->leafsize = leafsize;
87ee04eb 868 root->stripesize = stripesize;
123abc88 869 root->ref_cows = 0;
0b86a832
CM		 870 root->track_dirty = 0;
871
9f5fae2f 872 root->fs_info = fs_info;
0f7d52f4
CM		 873 root->objectid = objectid;
874 root->last_trans = 0;
1b05da2e
CM		 875 root->highest_inode = 0;
876 root->last_inode_alloc = 0;
58176a96 877 root->name = NULL;
4313b399 878 root->in_sysfs = 0;
0b86a832
CM		 879
880 INIT_LIST_HEAD(&root->dirty_list);
7b128766 881 INIT_LIST_HEAD(&root->orphan_list);
bcc63abb 882 INIT_LIST_HEAD(&root->dead_list);
925baedd 883 spin_lock_init(&root->node_lock);
bcc63abb 884 spin_lock_init(&root->list_lock);
a2135011 885 mutex_init(&root->objectid_mutex);
e02119d5 886 mutex_init(&root->log_mutex);
d0c803c4
CM		 887 extent_io_tree_init(&root->dirty_log_pages,
888 fs_info->btree_inode->i_mapping, GFP_NOFS);
017e5369
CM		 889
890 btrfs_leaf_ref_tree_init(&root->ref_tree_struct);
891 root->ref_tree = &root->ref_tree_struct;
892
3768f368
CM		 893 memset(&root->root_key, 0, sizeof(root->root_key));
894 memset(&root->root_item, 0, sizeof(root->root_item));
6702ed49 895 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
58176a96 896 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
3f157a2f 897 root->defrag_trans_start = fs_info->generation;
58176a96 898 init_completion(&root->kobj_unregister);
6702ed49
CM		 899 root->defrag_running = 0;
900 root->defrag_level = 0;
4d775673 901 root->root_key.objectid = objectid;
3394e160
CM		 902 root->anon_super.s_root = NULL;
903 root->anon_super.s_dev = 0;
904 INIT_LIST_HEAD(&root->anon_super.s_list);
905 INIT_LIST_HEAD(&root->anon_super.s_instances);
906 init_rwsem(&root->anon_super.s_umount);
907
3768f368
CM		 908 return 0;
909}
910
db94535d 911static int find_and_setup_root(struct btrfs_root *tree_root,
9f5fae2f
CM		 912 struct btrfs_fs_info *fs_info,
913 u64 objectid,
e20d96d6 914 struct btrfs_root *root)
3768f368
CM		 915{
916 int ret;
db94535d 917 u32 blocksize;
84234f3a 918 u64 generation;
3768f368 919
db94535d 920 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 921 tree_root->sectorsize, tree_root->stripesize,
922 root, fs_info, objectid);
3768f368
CM		 923 ret = btrfs_find_last_root(tree_root, objectid,
924 &root->root_item, &root->root_key);
925 BUG_ON(ret);
926
84234f3a 927 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 928 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
929 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 930 blocksize, generation);
3768f368 931 BUG_ON(!root->node);
d97e63b6
CM		 932 return 0;
933}
934
e02119d5
CM		 935int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
936 struct btrfs_fs_info *fs_info)
937{
938 struct extent_buffer *eb;
d0c803c4
CM		 939 struct btrfs_root *log_root_tree = fs_info->log_root_tree;
940 u64 start = 0;
941 u64 end = 0;
e02119d5
CM		 942 int ret;
943
d0c803c4 944 if (!log_root_tree)
e02119d5
CM		 945 return 0;
946
d0c803c4
CM		 947 while(1) {
948 ret = find_first_extent_bit(&log_root_tree->dirty_log_pages,
949 0, &start, &end, EXTENT_DIRTY);
950 if (ret)
951 break;
952
953 clear_extent_dirty(&log_root_tree->dirty_log_pages,
954 start, end, GFP_NOFS);
955 }
e02119d5
CM		 956 eb = fs_info->log_root_tree->node;
957
958 WARN_ON(btrfs_header_level(eb) != 0);
959 WARN_ON(btrfs_header_nritems(eb) != 0);
960
d00aff00
CM		 961 ret = btrfs_free_reserved_extent(fs_info->tree_root,
962 eb->start, eb->len);
e02119d5
CM		 963 BUG_ON(ret);
964
965 free_extent_buffer(eb);
966 kfree(fs_info->log_root_tree);
967 fs_info->log_root_tree = NULL;
968 return 0;
969}
970
971int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
972 struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 973{
974 struct btrfs_root *root;
975 struct btrfs_root *tree_root = fs_info->tree_root;
e02119d5
CM		 976
977 root = kzalloc(sizeof(*root), GFP_NOFS);
978 if (!root)
979 return -ENOMEM;
980
981 __setup_root(tree_root->nodesize, tree_root->leafsize,
982 tree_root->sectorsize, tree_root->stripesize,
983 root, fs_info, BTRFS_TREE_LOG_OBJECTID);
984
985 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
986 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
987 root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
988 root->ref_cows = 0;
989
990 root->node = btrfs_alloc_free_block(trans, root, root->leafsize,
31840ae1
ZY		 991 0, BTRFS_TREE_LOG_OBJECTID,
992 trans->transid, 0, 0, 0);
e02119d5
CM		 993
994 btrfs_set_header_nritems(root->node, 0);
995 btrfs_set_header_level(root->node, 0);
996 btrfs_set_header_bytenr(root->node, root->node->start);
997 btrfs_set_header_generation(root->node, trans->transid);
998 btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID);
999
1000 write_extent_buffer(root->node, root->fs_info->fsid,
1001 (unsigned long)btrfs_header_fsid(root->node),
1002 BTRFS_FSID_SIZE);
1003 btrfs_mark_buffer_dirty(root->node);
1004 btrfs_tree_unlock(root->node);
1005 fs_info->log_root_tree = root;
1006 return 0;
1007}
1008
1009struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
1010 struct btrfs_key *location)
1011{
1012 struct btrfs_root *root;
1013 struct btrfs_fs_info *fs_info = tree_root->fs_info;
0f7d52f4 1014 struct btrfs_path *path;
5f39d397 1015 struct extent_buffer *l;
1b05da2e 1016 u64 highest_inode;
84234f3a 1017 u64 generation;
db94535d 1018 u32 blocksize;
0f7d52f4
CM		 1019 int ret = 0;
1020
5eda7b5e 1021 root = kzalloc(sizeof(*root), GFP_NOFS);
0cf6c620 1022 if (!root)
0f7d52f4 1023 return ERR_PTR(-ENOMEM);
0f7d52f4 1024 if (location->offset == (u64)-1) {
db94535d 1025 ret = find_and_setup_root(tree_root, fs_info,
0f7d52f4
CM		 1026 location->objectid, root);
1027 if (ret) {
0f7d52f4
CM		 1028 kfree(root);
1029 return ERR_PTR(ret);
1030 }
1031 goto insert;
1032 }
1033
db94535d 1034 __setup_root(tree_root->nodesize, tree_root->leafsize,
87ee04eb
CM		 1035 tree_root->sectorsize, tree_root->stripesize,
1036 root, fs_info, location->objectid);
0f7d52f4
CM		 1037
1038 path = btrfs_alloc_path();
1039 BUG_ON(!path);
1040 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
1041 if (ret != 0) {
0f7d52f4
CM		 1042 if (ret > 0)
1043 ret = -ENOENT;
1044 goto out;
1045 }
5f39d397
CM		 1046 l = path->nodes[0];
1047 read_extent_buffer(l, &root->root_item,
1048 btrfs_item_ptr_offset(l, path->slots[0]),
0f7d52f4 1049 sizeof(root->root_item));
44b36eb2 1050 memcpy(&root->root_key, location, sizeof(*location));
0f7d52f4
CM		 1051 ret = 0;
1052out:
1053 btrfs_release_path(root, path);
1054 btrfs_free_path(path);
1055 if (ret) {
1056 kfree(root);
1057 return ERR_PTR(ret);
1058 }
84234f3a 1059 generation = btrfs_root_generation(&root->root_item);
db94535d
CM		 1060 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1061 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
84234f3a 1062 blocksize, generation);
0f7d52f4
CM		 1063 BUG_ON(!root->node);
1064insert:
e02119d5
CM		 1065 if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
1066 root->ref_cows = 1;
1067 ret = btrfs_find_highest_inode(root, &highest_inode);
1068 if (ret == 0) {
1069 root->highest_inode = highest_inode;
1070 root->last_inode_alloc = highest_inode;
1071 }
5eda7b5e
CM		 1072 }
1073 return root;
1074}
1075
dc17ff8f
CM		 1076struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1077 u64 root_objectid)
1078{
1079 struct btrfs_root *root;
1080
1081 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
1082 return fs_info->tree_root;
1083 if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
1084 return fs_info->extent_root;
1085
1086 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1087 (unsigned long)root_objectid);
1088 return root;
1089}
1090
edbd8d4e
CM		 1091struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
1092 struct btrfs_key *location)
5eda7b5e
CM		 1093{
1094 struct btrfs_root *root;
1095 int ret;
1096
edbd8d4e
CM		 1097 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1098 return fs_info->tree_root;
1099 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1100 return fs_info->extent_root;
8f18cf13
CM		 1101 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1102 return fs_info->chunk_root;
1103 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1104 return fs_info->dev_root;
0403e47e
YZ		 1105 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1106 return fs_info->csum_root;
edbd8d4e 1107
5eda7b5e
CM		 1108 root = radix_tree_lookup(&fs_info->fs_roots_radix,
1109 (unsigned long)location->objectid);
1110 if (root)
1111 return root;
1112
e02119d5 1113 root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
5eda7b5e
CM		 1114 if (IS_ERR(root))
1115 return root;
3394e160
CM		 1116
1117 set_anon_super(&root->anon_super, NULL);
1118
2619ba1f
CM		 1119 ret = radix_tree_insert(&fs_info->fs_roots_radix,
1120 (unsigned long)root->root_key.objectid,
0f7d52f4
CM		 1121 root);
1122 if (ret) {
5f39d397 1123 free_extent_buffer(root->node);
0f7d52f4
CM		 1124 kfree(root);
1125 return ERR_PTR(ret);
1126 }
c146afad
YZ		 1127 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
1128 ret = btrfs_find_dead_roots(fs_info->tree_root,
1129 root->root_key.objectid, root);
1130 BUG_ON(ret);
1131 btrfs_orphan_cleanup(root);
1132 }
edbd8d4e
CM		 1133 return root;
1134}
1135
1136struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
1137 struct btrfs_key *location,
1138 const char *name, int namelen)
1139{
1140 struct btrfs_root *root;
1141 int ret;
1142
1143 root = btrfs_read_fs_root_no_name(fs_info, location);
1144 if (!root)
1145 return NULL;
58176a96 1146
4313b399
CM		 1147 if (root->in_sysfs)
1148 return root;
1149
58176a96
JB		 1150 ret = btrfs_set_root_name(root, name, namelen);
1151 if (ret) {
5f39d397 1152 free_extent_buffer(root->node);
58176a96
JB		 1153 kfree(root);
1154 return ERR_PTR(ret);
1155 }
ea9e8b11 1156#if 0
58176a96
JB		 1157 ret = btrfs_sysfs_add_root(root);
1158 if (ret) {
5f39d397 1159 free_extent_buffer(root->node);
58176a96
JB		 1160 kfree(root->name);
1161 kfree(root);
1162 return ERR_PTR(ret);
1163 }
ea9e8b11 1164#endif
4313b399 1165 root->in_sysfs = 1;
0f7d52f4
CM		 1166 return root;
1167}
19c00ddc
CM		 1168#if 0
1169static int add_hasher(struct btrfs_fs_info *info, char *type) {
1170 struct btrfs_hasher *hasher;
1171
1172 hasher = kmalloc(sizeof(*hasher), GFP_NOFS);
1173 if (!hasher)
1174 return -ENOMEM;
1175 hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC);
1176 if (!hasher->hash_tfm) {
1177 kfree(hasher);
1178 return -EINVAL;
1179 }
1180 spin_lock(&info->hash_lock);
1181 list_add(&hasher->list, &info->hashers);
1182 spin_unlock(&info->hash_lock);
1183 return 0;
1184}
1185#endif
04160088
CM		 1186
1187static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1188{
1189 struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1190 int ret = 0;
1191 struct list_head *cur;
1192 struct btrfs_device *device;
1193 struct backing_dev_info *bdi;
6f3577bd 1194#if 0
cb03c743 1195 if ((bdi_bits & (1 << BDI_write_congested)) &&
777e6bd7 1196 btrfs_congested_async(info, 0))
cb03c743 1197 return 1;
6f3577bd 1198#endif
04160088
CM		 1199 list_for_each(cur, &info->fs_devices->devices) {
1200 device = list_entry(cur, struct btrfs_device, dev_list);
dfe25020
CM		 1201 if (!device->bdev)
1202 continue;
04160088
CM		 1203 bdi = blk_get_backing_dev_info(device->bdev);
1204 if (bdi && bdi_congested(bdi, bdi_bits)) {
1205 ret = 1;
1206 break;
1207 }
1208 }
1209 return ret;
1210}
1211
38b66988
CM		 1212/*
1213 * this unplugs every device on the box, and it is only used when page
1214 * is null
1215 */
1216static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
1217{
1218 struct list_head *cur;
1219 struct btrfs_device *device;
1220 struct btrfs_fs_info *info;
1221
1222 info = (struct btrfs_fs_info *)bdi->unplug_io_data;
1223 list_for_each(cur, &info->fs_devices->devices) {
1224 device = list_entry(cur, struct btrfs_device, dev_list);
d20f7043
CM		 1225 if (!device->bdev)
1226 continue;
1227
38b66988
CM		 1228 bdi = blk_get_backing_dev_info(device->bdev);
1229 if (bdi->unplug_io_fn) {
1230 bdi->unplug_io_fn(bdi, page);
1231 }
1232 }
1233}
1234
b2950863 1235static void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
04160088 1236{
38b66988 1237 struct inode *inode;
f2d8d74d
CM		 1238 struct extent_map_tree *em_tree;
1239 struct extent_map *em;
bcbfce8a 1240 struct address_space *mapping;
38b66988
CM		 1241 u64 offset;
1242
bcbfce8a 1243 /* the generic O_DIRECT read code does this */
9f0ba5bd 1244 if (1 || !page) {
38b66988
CM		 1245 __unplug_io_fn(bdi, page);
1246 return;
1247 }
1248
bcbfce8a
CM		 1249 /*
1250 * page->mapping may change at any time. Get a consistent copy
1251 * and use that for everything below
1252 */
1253 smp_mb();
1254 mapping = page->mapping;
1255 if (!mapping)
1256 return;
1257
1258 inode = mapping->host;
240d5d48
CM		 1259
1260 /*
1261 * don't do the expensive searching for a small number of
1262 * devices
1263 */
1264 if (BTRFS_I(inode)->root->fs_info->fs_devices->open_devices <= 2) {
1265 __unplug_io_fn(bdi, page);
1266 return;
1267 }
1268
38b66988 1269 offset = page_offset(page);
04160088 1270
f2d8d74d
CM		 1271 em_tree = &BTRFS_I(inode)->extent_tree;
1272 spin_lock(&em_tree->lock);
1273 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
1274 spin_unlock(&em_tree->lock);
89642229
CM		 1275 if (!em) {
1276 __unplug_io_fn(bdi, page);
f2d8d74d 1277 return;
89642229 1278 }
f2d8d74d 1279
89642229
CM		 1280 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1281 free_extent_map(em);
1282 __unplug_io_fn(bdi, page);
1283 return;
1284 }
f2d8d74d
CM		 1285 offset = offset - em->start;
1286 btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree,
1287 em->block_start + offset, page);
1288 free_extent_map(em);
04160088
CM		 1289}
1290
1291static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1292{
1293 bdi_init(bdi);
4575c9cc 1294 bdi->ra_pages = default_backing_dev_info.ra_pages;
04160088
CM		 1295 bdi->state = 0;
1296 bdi->capabilities = default_backing_dev_info.capabilities;
1297 bdi->unplug_io_fn = btrfs_unplug_io_fn;
1298 bdi->unplug_io_data = info;
1299 bdi->congested_fn = btrfs_congested_fn;
1300 bdi->congested_data = info;
1301 return 0;
1302}
1303
ce9adaa5
CM		 1304static int bio_ready_for_csum(struct bio *bio)
1305{
1306 u64 length = 0;
1307 u64 buf_len = 0;
1308 u64 start = 0;
1309 struct page *page;
1310 struct extent_io_tree *io_tree = NULL;
1311 struct btrfs_fs_info *info = NULL;
1312 struct bio_vec *bvec;
1313 int i;
1314 int ret;
1315
1316 bio_for_each_segment(bvec, bio, i) {
1317 page = bvec->bv_page;
1318 if (page->private == EXTENT_PAGE_PRIVATE) {
1319 length += bvec->bv_len;
1320 continue;
1321 }
1322 if (!page->private) {
1323 length += bvec->bv_len;
1324 continue;
1325 }
1326 length = bvec->bv_len;
1327 buf_len = page->private >> 2;
1328 start = page_offset(page) + bvec->bv_offset;
1329 io_tree = &BTRFS_I(page->mapping->host)->io_tree;
1330 info = BTRFS_I(page->mapping->host)->root->fs_info;
1331 }
1332 /* are we fully contained in this bio? */
1333 if (buf_len <= length)
1334 return 1;
1335
1336 ret = extent_range_uptodate(io_tree, start + length,
1337 start + buf_len - 1);
1338 if (ret == 1)
1339 return ret;
1340 return ret;
1341}
1342
8b712842
CM		 1343/*
1344 * called by the kthread helper functions to finally call the bio end_io
1345 * functions. This is where read checksum verification actually happens
1346 */
1347static void end_workqueue_fn(struct btrfs_work *work)
ce9adaa5 1348{
ce9adaa5 1349 struct bio *bio;
8b712842
CM		 1350 struct end_io_wq *end_io_wq;
1351 struct btrfs_fs_info *fs_info;
ce9adaa5 1352 int error;
ce9adaa5 1353
8b712842
CM		 1354 end_io_wq = container_of(work, struct end_io_wq, work);
1355 bio = end_io_wq->bio;
1356 fs_info = end_io_wq->info;
ce9adaa5 1357
cad321ad 1358 /* metadata bio reads are special because the whole tree block must
8b712842
CM		 1359 * be checksummed at once. This makes sure the entire block is in
1360 * ram and up to date before trying to verify things. For
1361 * blocksize <= pagesize, it is basically a noop
1362 */
cad321ad
CM		 1363 if (!(bio->bi_rw & (1 << BIO_RW)) && end_io_wq->metadata &&
1364 !bio_ready_for_csum(bio)) {
d20f7043 1365 btrfs_queue_worker(&fs_info->endio_meta_workers,
8b712842
CM		 1366 &end_io_wq->work);
1367 return;
1368 }
1369 error = end_io_wq->error;
1370 bio->bi_private = end_io_wq->private;
1371 bio->bi_end_io = end_io_wq->end_io;
1372 kfree(end_io_wq);
8b712842 1373 bio_endio(bio, error);
44b8bd7e
CM		 1374}
1375
a74a4b97
CM		 1376static int cleaner_kthread(void *arg)
1377{
1378 struct btrfs_root *root = arg;
1379
1380 do {
1381 smp_mb();
1382 if (root->fs_info->closing)
1383 break;
1384
1385 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1386 mutex_lock(&root->fs_info->cleaner_mutex);
a74a4b97 1387 btrfs_clean_old_snapshots(root);
a74a4b97
CM		 1388 mutex_unlock(&root->fs_info->cleaner_mutex);
1389
1390 if (freezing(current)) {
1391 refrigerator();
1392 } else {
1393 smp_mb();
1394 if (root->fs_info->closing)
1395 break;
1396 set_current_state(TASK_INTERRUPTIBLE);
1397 schedule();
1398 __set_current_state(TASK_RUNNING);
1399 }
1400 } while (!kthread_should_stop());
1401 return 0;
1402}
1403
1404static int transaction_kthread(void *arg)
1405{
1406 struct btrfs_root *root = arg;
1407 struct btrfs_trans_handle *trans;
1408 struct btrfs_transaction *cur;
1409 unsigned long now;
1410 unsigned long delay;
1411 int ret;
1412
1413 do {
1414 smp_mb();
1415 if (root->fs_info->closing)
1416 break;
1417
1418 delay = HZ * 30;
1419 vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
1420 mutex_lock(&root->fs_info->transaction_kthread_mutex);
1421
ab78c84d
CM		 1422 if (root->fs_info->total_ref_cache_size > 20 * 1024 * 1024) {
1423 printk("btrfs: total reference cache size %Lu\n",
1424 root->fs_info->total_ref_cache_size);
1425 }
31153d81 1426
a74a4b97
CM		 1427 mutex_lock(&root->fs_info->trans_mutex);
1428 cur = root->fs_info->running_transaction;
1429 if (!cur) {
1430 mutex_unlock(&root->fs_info->trans_mutex);
1431 goto sleep;
1432 }
31153d81 1433
a74a4b97
CM		 1434 now = get_seconds();
1435 if (now < cur->start_time || now - cur->start_time < 30) {
1436 mutex_unlock(&root->fs_info->trans_mutex);
1437 delay = HZ * 5;
1438 goto sleep;
1439 }
1440 mutex_unlock(&root->fs_info->trans_mutex);
a74a4b97
CM		 1441 trans = btrfs_start_transaction(root, 1);
1442 ret = btrfs_commit_transaction(trans, root);
1443sleep:
1444 wake_up_process(root->fs_info->cleaner_kthread);
1445 mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1446
1447 if (freezing(current)) {
1448 refrigerator();
1449 } else {
1450 if (root->fs_info->closing)
1451 break;
1452 set_current_state(TASK_INTERRUPTIBLE);
1453 schedule_timeout(delay);
1454 __set_current_state(TASK_RUNNING);
1455 }
1456 } while (!kthread_should_stop());
1457 return 0;
1458}
1459
8a4b83cc 1460struct btrfs_root *open_ctree(struct super_block *sb,
dfe25020
CM		 1461 struct btrfs_fs_devices *fs_devices,
1462 char *options)
2e635a27 1463{
db94535d
CM		 1464 u32 sectorsize;
1465 u32 nodesize;
1466 u32 leafsize;
1467 u32 blocksize;
87ee04eb 1468 u32 stripesize;
84234f3a 1469 u64 generation;
f2b636e8 1470 u64 features;
3de4586c 1471 struct btrfs_key location;
a061fc8d 1472 struct buffer_head *bh;
e02119d5 1473 struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
e20d96d6 1474 GFP_NOFS);
d20f7043
CM		 1475 struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root),
1476 GFP_NOFS);
e02119d5 1477 struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root),
e20d96d6 1478 GFP_NOFS);
8790d502 1479 struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
e20d96d6 1480 GFP_NOFS);
e02119d5 1481 struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
0b86a832 1482 GFP_NOFS);
e02119d5 1483 struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
0b86a832 1484 GFP_NOFS);
e02119d5
CM		 1485 struct btrfs_root *log_tree_root;
1486
eb60ceac 1487 int ret;
e58ca020 1488 int err = -EINVAL;
4543df7e 1489
2c90e5d6 1490 struct btrfs_super_block *disk_super;
8790d502 1491
0463bb4e 1492 if (!extent_root || !tree_root || !fs_info ||
d20f7043 1493 !chunk_root || !dev_root || !csum_root) {
39279cc3
CM		 1494 err = -ENOMEM;
1495 goto fail;
1496 }
0f7d52f4 1497 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
8fd17795 1498 INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7 1499 INIT_LIST_HEAD(&fs_info->dead_roots);
19c00ddc 1500 INIT_LIST_HEAD(&fs_info->hashers);
ea8c2819 1501 INIT_LIST_HEAD(&fs_info->delalloc_inodes);
19c00ddc 1502 spin_lock_init(&fs_info->hash_lock);
1832a6d5 1503 spin_lock_init(&fs_info->delalloc_lock);
cee36a03 1504 spin_lock_init(&fs_info->new_trans_lock);
31153d81 1505 spin_lock_init(&fs_info->ref_cache_lock);
19c00ddc 1506
58176a96 1507 init_completion(&fs_info->kobj_unregister);
9f5fae2f
CM		 1508 fs_info->tree_root = tree_root;
1509 fs_info->extent_root = extent_root;
d20f7043 1510 fs_info->csum_root = csum_root;
0b86a832
CM		 1511 fs_info->chunk_root = chunk_root;
1512 fs_info->dev_root = dev_root;
8a4b83cc 1513 fs_info->fs_devices = fs_devices;
0b86a832 1514 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
6324fbf3 1515 INIT_LIST_HEAD(&fs_info->space_info);
0b86a832 1516 btrfs_mapping_init(&fs_info->mapping_tree);
cb03c743 1517 atomic_set(&fs_info->nr_async_submits, 0);
771ed689 1518 atomic_set(&fs_info->async_delalloc_pages, 0);
8c8bee1d 1519 atomic_set(&fs_info->async_submit_draining, 0);
0986fe9e 1520 atomic_set(&fs_info->nr_async_bios, 0);
a2135011 1521 atomic_set(&fs_info->throttles, 0);
ab78c84d 1522 atomic_set(&fs_info->throttle_gen, 0);
e20d96d6 1523 fs_info->sb = sb;
c59f8951 1524 fs_info->max_extent = (u64)-1;
6f568d35 1525 fs_info->max_inline = 8192 * 1024;
04160088 1526 setup_bdi(fs_info, &fs_info->bdi);
d98237b3
CM		 1527 fs_info->btree_inode = new_inode(sb);
1528 fs_info->btree_inode->i_ino = 1;
2c90e5d6 1529 fs_info->btree_inode->i_nlink = 1;
c8b97818 1530
b34b086c
CM		 1531 fs_info->thread_pool_size = min_t(unsigned long,
1532 num_online_cpus() + 2, 8);
0afbaf8c 1533
3eaa2885
CM		 1534 INIT_LIST_HEAD(&fs_info->ordered_extents);
1535 spin_lock_init(&fs_info->ordered_extent_lock);
1536
a061fc8d
CM		 1537 sb->s_blocksize = 4096;
1538 sb->s_blocksize_bits = blksize_bits(4096);
1539
0afbaf8c
CM		 1540 /*
1541 * we set the i_size on the btree inode to the max possible int.
1542 * the real end of the address space is determined by all of
1543 * the devices in the system
1544 */
1545 fs_info->btree_inode->i_size = OFFSET_MAX;
d98237b3 1546 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
04160088
CM		 1547 fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
1548
d1310b2e 1549 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
5f39d397
CM		 1550 fs_info->btree_inode->i_mapping,
1551 GFP_NOFS);
d1310b2e
CM		 1552 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
1553 GFP_NOFS);
1554
1555 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
0da5468f 1556
0f9dd46c
JB		 1557 spin_lock_init(&fs_info->block_group_cache_lock);
1558 fs_info->block_group_cache_tree.rb_node = NULL;
1559
d1310b2e 1560 extent_io_tree_init(&fs_info->pinned_extents,
1a5bc167 1561 fs_info->btree_inode->i_mapping, GFP_NOFS);
d1310b2e 1562 extent_io_tree_init(&fs_info->pending_del,
1a5bc167 1563 fs_info->btree_inode->i_mapping, GFP_NOFS);
d1310b2e 1564 extent_io_tree_init(&fs_info->extent_ins,
1a5bc167 1565 fs_info->btree_inode->i_mapping, GFP_NOFS);
e66f709b 1566 fs_info->do_barriers = 1;
e18e4809 1567
1a40e23b
ZY		 1568 INIT_LIST_HEAD(&fs_info->dead_reloc_roots);
1569 btrfs_leaf_ref_tree_init(&fs_info->reloc_ref_tree);
e4657689
ZY		 1570 btrfs_leaf_ref_tree_init(&fs_info->shared_ref_tree);
1571
0f7d52f4
CM		 1572 BTRFS_I(fs_info->btree_inode)->root = tree_root;
1573 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
1574 sizeof(struct btrfs_key));
22b0ebda 1575 insert_inode_hash(fs_info->btree_inode);
39279cc3 1576
79154b1b 1577 mutex_init(&fs_info->trans_mutex);
e02119d5 1578 mutex_init(&fs_info->tree_log_mutex);
a2135011 1579 mutex_init(&fs_info->drop_mutex);
25179201
JB		 1580 mutex_init(&fs_info->extent_ins_mutex);
1581 mutex_init(&fs_info->pinned_mutex);
925baedd 1582 mutex_init(&fs_info->chunk_mutex);
a74a4b97
CM		 1583 mutex_init(&fs_info->transaction_kthread_mutex);
1584 mutex_init(&fs_info->cleaner_mutex);
7d9eb12c 1585 mutex_init(&fs_info->volume_mutex);
1a40e23b 1586 mutex_init(&fs_info->tree_reloc_mutex);
e6dcd2dc 1587 init_waitqueue_head(&fs_info->transaction_throttle);
f9295749 1588 init_waitqueue_head(&fs_info->transaction_wait);
4854ddd0 1589 init_waitqueue_head(&fs_info->async_submit_wait);
e02119d5
CM		 1590 init_waitqueue_head(&fs_info->tree_log_wait);
1591 atomic_set(&fs_info->tree_log_commit, 0);
1592 atomic_set(&fs_info->tree_log_writers, 0);
1593 fs_info->tree_log_transid = 0;
3768f368 1594
19c00ddc
CM		 1595#if 0
1596 ret = add_hasher(fs_info, "crc32c");
1597 if (ret) {
1598 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
1599 err = -ENOMEM;
1600 goto fail_iput;
1601 }
1602#endif
0b86a832 1603 __setup_root(4096, 4096, 4096, 4096, tree_root,
2c90e5d6 1604 fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903 1605
d98237b3 1606
a512bbf8 1607 bh = btrfs_read_dev_super(fs_devices->latest_bdev);
a061fc8d 1608 if (!bh)
39279cc3 1609 goto fail_iput;
39279cc3 1610
a061fc8d
CM		 1611 memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy));
1612 brelse(bh);
5f39d397 1613
a061fc8d 1614 memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE);
0b86a832 1615
5f39d397 1616 disk_super = &fs_info->super_copy;
0f7d52f4 1617 if (!btrfs_super_root(disk_super))
c6e2bac1 1618 goto fail_iput;
0f7d52f4 1619
2b82032c
YZ		 1620 ret = btrfs_parse_options(tree_root, options);
1621 if (ret) {
1622 err = ret;
c6e2bac1 1623 goto fail_iput;
2b82032c 1624 }
dfe25020 1625
f2b636e8
JB		 1626 features = btrfs_super_incompat_flags(disk_super) &
1627 ~BTRFS_FEATURE_INCOMPAT_SUPP;
1628 if (features) {
1629 printk(KERN_ERR "BTRFS: couldn't mount because of "
1630 "unsupported optional features (%Lx).\n",
1631 features);
1632 err = -EINVAL;
c6e2bac1 1633 goto fail_iput;
f2b636e8
JB		 1634 }
1635
1636 features = btrfs_super_compat_ro_flags(disk_super) &
1637 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
1638 if (!(sb->s_flags & MS_RDONLY) && features) {
1639 printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
1640 "unsupported option features (%Lx).\n",
1641 features);
1642 err = -EINVAL;
c6e2bac1 1643 goto fail_iput;
f2b636e8
JB		 1644 }
1645
4543df7e
CM		 1646 /*
1647 * we need to start all the end_io workers up front because the
1648 * queue work function gets called at interrupt time, and so it
1649 * cannot dynamically grow.
1650 */
5443be45
CM		 1651 btrfs_init_workers(&fs_info->workers, "worker",
1652 fs_info->thread_pool_size);
c8b97818 1653
771ed689
CM		 1654 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
1655 fs_info->thread_pool_size);
1656
5443be45 1657 btrfs_init_workers(&fs_info->submit_workers, "submit",
b720d209
CM		 1658 min_t(u64, fs_devices->num_devices,
1659 fs_info->thread_pool_size));
61b49440
CM		 1660
1661 /* a higher idle thresh on the submit workers makes it much more
1662 * likely that bios will be send down in a sane order to the
1663 * devices
1664 */
1665 fs_info->submit_workers.idle_thresh = 64;
53863232 1666
771ed689 1667 fs_info->workers.idle_thresh = 16;
4a69a410 1668 fs_info->workers.ordered = 1;
61b49440 1669
771ed689
CM		 1670 fs_info->delalloc_workers.idle_thresh = 2;
1671 fs_info->delalloc_workers.ordered = 1;
1672
5443be45
CM		 1673 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1);
1674 btrfs_init_workers(&fs_info->endio_workers, "endio",
1675 fs_info->thread_pool_size);
d20f7043
CM		 1676 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
1677 fs_info->thread_pool_size);
cad321ad
CM		 1678 btrfs_init_workers(&fs_info->endio_meta_write_workers,
1679 "endio-meta-write", fs_info->thread_pool_size);
5443be45 1680 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
e6dcd2dc 1681 fs_info->thread_pool_size);
61b49440
CM		 1682
1683 /*
1684 * endios are largely parallel and should have a very
1685 * low idle thresh
1686 */
1687 fs_info->endio_workers.idle_thresh = 4;
3a5f1d45 1688 fs_info->endio_write_workers.idle_thresh = 64;
cad321ad 1689 fs_info->endio_meta_write_workers.idle_thresh = 64;
61b49440 1690
4543df7e 1691 btrfs_start_workers(&fs_info->workers, 1);
1cc127b5 1692 btrfs_start_workers(&fs_info->submit_workers, 1);
771ed689 1693 btrfs_start_workers(&fs_info->delalloc_workers, 1);
247e743c 1694 btrfs_start_workers(&fs_info->fixup_workers, 1);
4543df7e 1695 btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size);
d20f7043
CM		 1696 btrfs_start_workers(&fs_info->endio_meta_workers,
1697 fs_info->thread_pool_size);
cad321ad
CM		 1698 btrfs_start_workers(&fs_info->endio_meta_write_workers,
1699 fs_info->thread_pool_size);
e6dcd2dc
CM		 1700 btrfs_start_workers(&fs_info->endio_write_workers,
1701 fs_info->thread_pool_size);
4543df7e 1702
4575c9cc 1703 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
c8b97818
CM		 1704 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
1705 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
4575c9cc 1706
db94535d
CM		 1707 nodesize = btrfs_super_nodesize(disk_super);
1708 leafsize = btrfs_super_leafsize(disk_super);
1709 sectorsize = btrfs_super_sectorsize(disk_super);
87ee04eb 1710 stripesize = btrfs_super_stripesize(disk_super);
db94535d
CM		 1711 tree_root->nodesize = nodesize;
1712 tree_root->leafsize = leafsize;
1713 tree_root->sectorsize = sectorsize;
87ee04eb 1714 tree_root->stripesize = stripesize;
a061fc8d
CM		 1715
1716 sb->s_blocksize = sectorsize;
1717 sb->s_blocksize_bits = blksize_bits(sectorsize);
db94535d 1718
39279cc3
CM		 1719 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
1720 sizeof(disk_super->magic))) {
1721 printk("btrfs: valid FS not found on %s\n", sb->s_id);
1722 goto fail_sb_buffer;
1723 }
19c00ddc 1724
925baedd 1725 mutex_lock(&fs_info->chunk_mutex);
e4404d6e 1726 ret = btrfs_read_sys_array(tree_root);
925baedd 1727 mutex_unlock(&fs_info->chunk_mutex);
84eed90f
CM		 1728 if (ret) {
1729 printk("btrfs: failed to read the system array on %s\n",
1730 sb->s_id);
1731 goto fail_sys_array;
1732 }
0b86a832
CM		 1733
1734 blocksize = btrfs_level_size(tree_root,
1735 btrfs_super_chunk_root_level(disk_super));
84234f3a 1736 generation = btrfs_super_chunk_root_generation(disk_super);
0b86a832
CM		 1737
1738 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1739 chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1740
1741 chunk_root->node = read_tree_block(chunk_root,
1742 btrfs_super_chunk_root(disk_super),
84234f3a 1743 blocksize, generation);
0b86a832
CM		 1744 BUG_ON(!chunk_root->node);
1745
e17cade2
CM		 1746 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
1747 (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
1748 BTRFS_UUID_SIZE);
1749
925baedd 1750 mutex_lock(&fs_info->chunk_mutex);
0b86a832 1751 ret = btrfs_read_chunk_tree(chunk_root);
925baedd 1752 mutex_unlock(&fs_info->chunk_mutex);
2b82032c
YZ		 1753 if (ret) {
1754 printk("btrfs: failed to read chunk tree on %s\n", sb->s_id);
1755 goto fail_chunk_root;
1756 }
0b86a832 1757
dfe25020
CM		 1758 btrfs_close_extra_devices(fs_devices);
1759
db94535d
CM		 1760 blocksize = btrfs_level_size(tree_root,
1761 btrfs_super_root_level(disk_super));
84234f3a 1762 generation = btrfs_super_generation(disk_super);
0b86a832 1763
e20d96d6 1764 tree_root->node = read_tree_block(tree_root,
db94535d 1765 btrfs_super_root(disk_super),
84234f3a 1766 blocksize, generation);
39279cc3 1767 if (!tree_root->node)
2b82032c 1768 goto fail_chunk_root;
3768f368 1769
db94535d
CM		 1770
1771 ret = find_and_setup_root(tree_root, fs_info,
e20d96d6 1772 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
0b86a832 1773 if (ret)
39279cc3 1774 goto fail_tree_root;
0b86a832
CM		 1775 extent_root->track_dirty = 1;
1776
1777 ret = find_and_setup_root(tree_root, fs_info,
1778 BTRFS_DEV_TREE_OBJECTID, dev_root);
1779 dev_root->track_dirty = 1;
1780
1781 if (ret)
1782 goto fail_extent_root;
3768f368 1783
d20f7043
CM		 1784 ret = find_and_setup_root(tree_root, fs_info,
1785 BTRFS_CSUM_TREE_OBJECTID, csum_root);
1786 if (ret)
1787 goto fail_extent_root;
1788
1789 csum_root->track_dirty = 1;
1790
9078a3e1
CM		 1791 btrfs_read_block_groups(extent_root);
1792
c146afad
YZ		 1793 fs_info->generation = generation + 1;
1794 fs_info->last_trans_committed = generation;
d18a2c44
CM		 1795 fs_info->data_alloc_profile = (u64)-1;
1796 fs_info->metadata_alloc_profile = (u64)-1;
1797 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
a74a4b97
CM		 1798 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
1799 "btrfs-cleaner");
1800 if (!fs_info->cleaner_kthread)
d20f7043 1801 goto fail_csum_root;
a74a4b97
CM		 1802
1803 fs_info->transaction_kthread = kthread_run(transaction_kthread,
1804 tree_root,
1805 "btrfs-transaction");
1806 if (!fs_info->transaction_kthread)
3f157a2f 1807 goto fail_cleaner;
a74a4b97 1808
e02119d5 1809 if (btrfs_super_log_root(disk_super) != 0) {
e02119d5
CM		 1810 u64 bytenr = btrfs_super_log_root(disk_super);
1811
7c2ca468
CM		 1812 if (fs_devices->rw_devices == 0) {
1813 printk("Btrfs log replay required on RO media\n");
1814 err = -EIO;
1815 goto fail_trans_kthread;
1816 }
e02119d5
CM		 1817 blocksize =
1818 btrfs_level_size(tree_root,
1819 btrfs_super_log_root_level(disk_super));
d18a2c44 1820
e02119d5
CM		 1821 log_tree_root = kzalloc(sizeof(struct btrfs_root),
1822 GFP_NOFS);
1823
1824 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1825 log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1826
1827 log_tree_root->node = read_tree_block(tree_root, bytenr,
84234f3a
YZ		 1828 blocksize,
1829 generation + 1);
e02119d5
CM		 1830 ret = btrfs_recover_log_trees(log_tree_root);
1831 BUG_ON(ret);
e556ce2c
YZ		 1832
1833 if (sb->s_flags & MS_RDONLY) {
1834 ret = btrfs_commit_super(tree_root);
1835 BUG_ON(ret);
1836 }
e02119d5 1837 }
1a40e23b 1838
7c2ca468
CM		 1839 if (!(sb->s_flags & MS_RDONLY)) {
1840 ret = btrfs_cleanup_reloc_trees(tree_root);
1841 BUG_ON(ret);
1842 }
1a40e23b 1843
3de4586c
CM		 1844 location.objectid = BTRFS_FS_TREE_OBJECTID;
1845 location.type = BTRFS_ROOT_ITEM_KEY;
1846 location.offset = (u64)-1;
1847
3de4586c
CM		 1848 fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
1849 if (!fs_info->fs_root)
7c2ca468 1850 goto fail_trans_kthread;
0f7d52f4 1851 return tree_root;
39279cc3 1852
7c2ca468
CM		 1853fail_trans_kthread:
1854 kthread_stop(fs_info->transaction_kthread);
3f157a2f 1855fail_cleaner:
a74a4b97 1856 kthread_stop(fs_info->cleaner_kthread);
7c2ca468
CM		 1857
1858 /*
1859 * make sure we're done with the btree inode before we stop our
1860 * kthreads
1861 */
1862 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
1863 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
1864
d20f7043
CM		 1865fail_csum_root:
1866 free_extent_buffer(csum_root->node);
0b86a832
CM		 1867fail_extent_root:
1868 free_extent_buffer(extent_root->node);
39279cc3 1869fail_tree_root:
5f39d397 1870 free_extent_buffer(tree_root->node);
2b82032c
YZ		 1871fail_chunk_root:
1872 free_extent_buffer(chunk_root->node);
84eed90f 1873fail_sys_array:
7c2ca468 1874 free_extent_buffer(dev_root->node);
39279cc3 1875fail_sb_buffer:
247e743c 1876 btrfs_stop_workers(&fs_info->fixup_workers);
771ed689 1877 btrfs_stop_workers(&fs_info->delalloc_workers);
8b712842
CM		 1878 btrfs_stop_workers(&fs_info->workers);
1879 btrfs_stop_workers(&fs_info->endio_workers);
d20f7043 1880 btrfs_stop_workers(&fs_info->endio_meta_workers);
cad321ad 1881 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
e6dcd2dc 1882 btrfs_stop_workers(&fs_info->endio_write_workers);
1cc127b5 1883 btrfs_stop_workers(&fs_info->submit_workers);
4543df7e 1884fail_iput:
7c2ca468 1885 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
4543df7e 1886 iput(fs_info->btree_inode);
39279cc3 1887fail:
dfe25020 1888 btrfs_close_devices(fs_info->fs_devices);
84eed90f
CM		 1889 btrfs_mapping_tree_free(&fs_info->mapping_tree);
1890
39279cc3
CM		 1891 kfree(extent_root);
1892 kfree(tree_root);
2d2ae547 1893 bdi_destroy(&fs_info->bdi);
39279cc3 1894 kfree(fs_info);
83afeac4
JM		 1895 kfree(chunk_root);
1896 kfree(dev_root);
d20f7043 1897 kfree(csum_root);
39279cc3 1898 return ERR_PTR(err);
eb60ceac
CM		 1899}
1900
f2984462
CM		 1901static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
1902{
1903 char b[BDEVNAME_SIZE];
1904
1905 if (uptodate) {
1906 set_buffer_uptodate(bh);
1907 } else {
1908 if (!buffer_eopnotsupp(bh) && printk_ratelimit()) {
1909 printk(KERN_WARNING "lost page write due to "
1910 "I/O error on %s\n",
1911 bdevname(bh->b_bdev, b));
1912 }
1259ab75
CM		 1913 /* note, we dont' set_buffer_write_io_error because we have
1914 * our own ways of dealing with the IO errors
1915 */
f2984462
CM		 1916 clear_buffer_uptodate(bh);
1917 }
1918 unlock_buffer(bh);
1919 put_bh(bh);
1920}
1921
a512bbf8
YZ		 1922struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
1923{
1924 struct buffer_head *bh;
1925 struct buffer_head *latest = NULL;
1926 struct btrfs_super_block *super;
1927 int i;
1928 u64 transid = 0;
1929 u64 bytenr;
1930
1931 /* we would like to check all the supers, but that would make
1932 * a btrfs mount succeed after a mkfs from a different FS.
1933 * So, we need to add a special mount option to scan for
1934 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
1935 */
1936 for (i = 0; i < 1; i++) {
1937 bytenr = btrfs_sb_offset(i);
1938 if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
1939 break;
1940 bh = __bread(bdev, bytenr / 4096, 4096);
1941 if (!bh)
1942 continue;
1943
1944 super = (struct btrfs_super_block *)bh->b_data;
1945 if (btrfs_super_bytenr(super) != bytenr ||
1946 strncmp((char *)(&super->magic), BTRFS_MAGIC,
1947 sizeof(super->magic))) {
1948 brelse(bh);
1949 continue;
1950 }
1951
1952 if (!latest || btrfs_super_generation(super) > transid) {
1953 brelse(latest);
1954 latest = bh;
1955 transid = btrfs_super_generation(super);
1956 } else {
1957 brelse(bh);
1958 }
1959 }
1960 return latest;
1961}
1962
1963static int write_dev_supers(struct btrfs_device *device,
1964 struct btrfs_super_block *sb,
1965 int do_barriers, int wait, int max_mirrors)
1966{
1967 struct buffer_head *bh;
1968 int i;
1969 int ret;
1970 int errors = 0;
1971 u32 crc;
1972 u64 bytenr;
1973 int last_barrier = 0;
1974
1975 if (max_mirrors == 0)
1976 max_mirrors = BTRFS_SUPER_MIRROR_MAX;
1977
1978 /* make sure only the last submit_bh does a barrier */
1979 if (do_barriers) {
1980 for (i = 0; i < max_mirrors; i++) {
1981 bytenr = btrfs_sb_offset(i);
1982 if (bytenr + BTRFS_SUPER_INFO_SIZE >=
1983 device->total_bytes)
1984 break;
1985 last_barrier = i;
1986 }
1987 }
1988
1989 for (i = 0; i < max_mirrors; i++) {
1990 bytenr = btrfs_sb_offset(i);
1991 if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
1992 break;
1993
1994 if (wait) {
1995 bh = __find_get_block(device->bdev, bytenr / 4096,
1996 BTRFS_SUPER_INFO_SIZE);
1997 BUG_ON(!bh);
1998 brelse(bh);
1999 wait_on_buffer(bh);
2000 if (buffer_uptodate(bh)) {
2001 brelse(bh);
2002 continue;
2003 }
2004 } else {
2005 btrfs_set_super_bytenr(sb, bytenr);
2006
2007 crc = ~(u32)0;
2008 crc = btrfs_csum_data(NULL, (char *)sb +
2009 BTRFS_CSUM_SIZE, crc,
2010 BTRFS_SUPER_INFO_SIZE -
2011 BTRFS_CSUM_SIZE);
2012 btrfs_csum_final(crc, sb->csum);
2013
2014 bh = __getblk(device->bdev, bytenr / 4096,
2015 BTRFS_SUPER_INFO_SIZE);
2016 memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
2017
2018 set_buffer_uptodate(bh);
2019 get_bh(bh);
2020 lock_buffer(bh);
2021 bh->b_end_io = btrfs_end_buffer_write_sync;
2022 }
2023
2024 if (i == last_barrier && do_barriers && device->barriers) {
2025 ret = submit_bh(WRITE_BARRIER, bh);
2026 if (ret == -EOPNOTSUPP) {
2027 printk("btrfs: disabling barriers on dev %s\n",
2028 device->name);
2029 set_buffer_uptodate(bh);
2030 device->barriers = 0;
2031 get_bh(bh);
2032 lock_buffer(bh);
2033 ret = submit_bh(WRITE, bh);
2034 }
2035 } else {
2036 ret = submit_bh(WRITE, bh);
2037 }
2038
2039 if (!ret && wait) {
2040 wait_on_buffer(bh);
2041 if (!buffer_uptodate(bh))
2042 errors++;
2043 } else if (ret) {
2044 errors++;
2045 }
2046 if (wait)
2047 brelse(bh);
2048 }
2049 return errors < i ? 0 : -1;
2050}
2051
2052int write_all_supers(struct btrfs_root *root, int max_mirrors)
f2984462
CM		 2053{
2054 struct list_head *cur;
2055 struct list_head *head = &root->fs_info->fs_devices->devices;
2056 struct btrfs_device *dev;
a061fc8d 2057 struct btrfs_super_block *sb;
f2984462 2058 struct btrfs_dev_item *dev_item;
f2984462
CM		 2059 int ret;
2060 int do_barriers;
a236aed1
CM		 2061 int max_errors;
2062 int total_errors = 0;
a061fc8d 2063 u64 flags;
f2984462 2064
a236aed1 2065 max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
f2984462
CM		 2066 do_barriers = !btrfs_test_opt(root, NOBARRIER);
2067
a061fc8d
CM		 2068 sb = &root->fs_info->super_for_commit;
2069 dev_item = &sb->dev_item;
f2984462
CM		 2070 list_for_each(cur, head) {
2071 dev = list_entry(cur, struct btrfs_device, dev_list);
dfe25020
CM		 2072 if (!dev->bdev) {
2073 total_errors++;
2074 continue;
2075 }
2b82032c 2076 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 2077 continue;
2078
2b82032c 2079 btrfs_set_stack_device_generation(dev_item, 0);
a061fc8d
CM		 2080 btrfs_set_stack_device_type(dev_item, dev->type);
2081 btrfs_set_stack_device_id(dev_item, dev->devid);
2082 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
2083 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
2084 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
2085 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
2086 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
2087 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
2b82032c 2088 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
a512bbf8 2089
a061fc8d
CM		 2090 flags = btrfs_super_flags(sb);
2091 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
2092
a512bbf8 2093 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
a236aed1
CM		 2094 if (ret)
2095 total_errors++;
f2984462 2096 }
a236aed1
CM		 2097 if (total_errors > max_errors) {
2098 printk("btrfs: %d errors while writing supers\n", total_errors);
2099 BUG();
2100 }
f2984462 2101
a512bbf8 2102 total_errors = 0;
f2984462
CM		 2103 list_for_each(cur, head) {
2104 dev = list_entry(cur, struct btrfs_device, dev_list);
dfe25020
CM		 2105 if (!dev->bdev)
2106 continue;
2b82032c 2107 if (!dev->in_fs_metadata || !dev->writeable)
dfe25020
CM		 2108 continue;
2109
a512bbf8
YZ		 2110 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
2111 if (ret)
2112 total_errors++;
f2984462 2113 }
a236aed1
CM		 2114 if (total_errors > max_errors) {
2115 printk("btrfs: %d errors while writing supers\n", total_errors);
2116 BUG();
2117 }
f2984462
CM		 2118 return 0;
2119}
2120
a512bbf8
YZ		 2121int write_ctree_super(struct btrfs_trans_handle *trans,
2122 struct btrfs_root *root, int max_mirrors)
eb60ceac 2123{
e66f709b 2124 int ret;
5f39d397 2125
a512bbf8 2126 ret = write_all_supers(root, max_mirrors);
5f39d397 2127 return ret;
cfaa7295
CM		 2128}
2129
5eda7b5e 2130int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
2619ba1f
CM		 2131{
2132 radix_tree_delete(&fs_info->fs_roots_radix,
2133 (unsigned long)root->root_key.objectid);
3394e160
CM		 2134 if (root->anon_super.s_dev) {
2135 down_write(&root->anon_super.s_umount);
2136 kill_anon_super(&root->anon_super);
2137 }
ea9e8b11 2138#if 0
b99aa6cb
CM		 2139 if (root->in_sysfs)
2140 btrfs_sysfs_del_root(root);
ea9e8b11 2141#endif
2619ba1f 2142 if (root->node)
5f39d397 2143 free_extent_buffer(root->node);
2619ba1f 2144 if (root->commit_root)
5f39d397 2145 free_extent_buffer(root->commit_root);
58176a96
JB		 2146 if (root->name)
2147 kfree(root->name);
2619ba1f
CM		 2148 kfree(root);
2149 return 0;
2150}
2151
35b7e476 2152static int del_fs_roots(struct btrfs_fs_info *fs_info)
0f7d52f4
CM		 2153{
2154 int ret;
2155 struct btrfs_root *gang[8];
2156 int i;
2157
2158 while(1) {
2159 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2160 (void **)gang, 0,
2161 ARRAY_SIZE(gang));
2162 if (!ret)
2163 break;
2619ba1f 2164 for (i = 0; i < ret; i++)
5eda7b5e 2165 btrfs_free_fs_root(fs_info, gang[i]);
0f7d52f4
CM		 2166 }
2167 return 0;
2168}
b4100d64 2169
c146afad 2170int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
cfaa7295 2171{
c146afad
YZ		 2172 u64 root_objectid = 0;
2173 struct btrfs_root *gang[8];
2174 int i;
3768f368 2175 int ret;
e089f05c 2176
c146afad
YZ		 2177 while (1) {
2178 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2179 (void **)gang, root_objectid,
2180 ARRAY_SIZE(gang));
2181 if (!ret)
2182 break;
2183 for (i = 0; i < ret; i++) {
2184 root_objectid = gang[i]->root_key.objectid;
2185 ret = btrfs_find_dead_roots(fs_info->tree_root,
2186 root_objectid, gang[i]);
2187 BUG_ON(ret);
2188 btrfs_orphan_cleanup(gang[i]);
2189 }
2190 root_objectid++;
2191 }
2192 return 0;
2193}
a2135011 2194
c146afad
YZ		 2195int btrfs_commit_super(struct btrfs_root *root)
2196{
2197 struct btrfs_trans_handle *trans;
2198 int ret;
a74a4b97 2199
c146afad 2200 mutex_lock(&root->fs_info->cleaner_mutex);
a74a4b97 2201 btrfs_clean_old_snapshots(root);
c146afad 2202 mutex_unlock(&root->fs_info->cleaner_mutex);
79154b1b 2203 trans = btrfs_start_transaction(root, 1);
54aa1f4d 2204 ret = btrfs_commit_transaction(trans, root);
c146afad
YZ		 2205 BUG_ON(ret);
2206 /* run commit again to drop the original snapshot */
79154b1b
CM		 2207 trans = btrfs_start_transaction(root, 1);
2208 btrfs_commit_transaction(trans, root);
2209 ret = btrfs_write_and_wait_transaction(NULL, root);
3768f368 2210 BUG_ON(ret);
d6bfde87 2211
a512bbf8 2212 ret = write_ctree_super(NULL, root, 0);
c146afad
YZ		 2213 return ret;
2214}
2215
2216int close_ctree(struct btrfs_root *root)
2217{
2218 struct btrfs_fs_info *fs_info = root->fs_info;
2219 int ret;
2220
2221 fs_info->closing = 1;
2222 smp_mb();
2223
2224 kthread_stop(root->fs_info->transaction_kthread);
2225 kthread_stop(root->fs_info->cleaner_kthread);
2226
2227 if (!(fs_info->sb->s_flags & MS_RDONLY)) {
2228 ret = btrfs_commit_super(root);
2229 if (ret) {
2230 printk("btrfs: commit super returns %d\n", ret);
2231 }
2232 }
0f7d52f4 2233
b0c68f8b
CM		 2234 if (fs_info->delalloc_bytes) {
2235 printk("btrfs: at unmount delalloc count %Lu\n",
2236 fs_info->delalloc_bytes);
2237 }
31153d81
YZ		 2238 if (fs_info->total_ref_cache_size) {
2239 printk("btrfs: at umount reference cache size %Lu\n",
2240 fs_info->total_ref_cache_size);
2241 }
bcc63abb 2242
0f7d52f4 2243 if (fs_info->extent_root->node)
5f39d397 2244 free_extent_buffer(fs_info->extent_root->node);
f510cfec 2245
0f7d52f4 2246 if (fs_info->tree_root->node)
5f39d397 2247 free_extent_buffer(fs_info->tree_root->node);
f510cfec 2248
0b86a832
CM		 2249 if (root->fs_info->chunk_root->node);
2250 free_extent_buffer(root->fs_info->chunk_root->node);
2251
2252 if (root->fs_info->dev_root->node);
2253 free_extent_buffer(root->fs_info->dev_root->node);
2254
d20f7043
CM		 2255 if (root->fs_info->csum_root->node);
2256 free_extent_buffer(root->fs_info->csum_root->node);
2257
9078a3e1 2258 btrfs_free_block_groups(root->fs_info);
d10c5f31 2259
c146afad 2260 del_fs_roots(fs_info);
d10c5f31 2261
c146afad 2262 iput(fs_info->btree_inode);
9ad6b7bc 2263
247e743c 2264 btrfs_stop_workers(&fs_info->fixup_workers);
771ed689 2265 btrfs_stop_workers(&fs_info->delalloc_workers);
8b712842
CM		 2266 btrfs_stop_workers(&fs_info->workers);
2267 btrfs_stop_workers(&fs_info->endio_workers);
d20f7043 2268 btrfs_stop_workers(&fs_info->endio_meta_workers);
cad321ad 2269 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
e6dcd2dc 2270 btrfs_stop_workers(&fs_info->endio_write_workers);
1cc127b5 2271 btrfs_stop_workers(&fs_info->submit_workers);
d6bfde87 2272
19c00ddc
CM		 2273#if 0
2274 while(!list_empty(&fs_info->hashers)) {
2275 struct btrfs_hasher *hasher;
2276 hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher,
2277 hashers);
2278 list_del(&hasher->hashers);
2279 crypto_free_hash(&fs_info->hash_tfm);
2280 kfree(hasher);
2281 }
2282#endif
dfe25020 2283 btrfs_close_devices(fs_info->fs_devices);
0b86a832 2284 btrfs_mapping_tree_free(&fs_info->mapping_tree);
b248a415 2285
04160088 2286 bdi_destroy(&fs_info->bdi);
0b86a832 2287
0f7d52f4 2288 kfree(fs_info->extent_root);
0f7d52f4 2289 kfree(fs_info->tree_root);
0b86a832
CM		 2290 kfree(fs_info->chunk_root);
2291 kfree(fs_info->dev_root);
d20f7043 2292 kfree(fs_info->csum_root);
eb60ceac
CM		 2293 return 0;
2294}
2295
1259ab75 2296int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
5f39d397 2297{
1259ab75 2298 int ret;
810191ff 2299 struct inode *btree_inode = buf->first_page->mapping->host;
1259ab75
CM		 2300
2301 ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf);
2302 if (!ret)
2303 return ret;
2304
2305 ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
2306 parent_transid);
2307 return !ret;
5f39d397
CM		 2308}
2309
2310int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
ccd467d6 2311{
810191ff 2312 struct inode *btree_inode = buf->first_page->mapping->host;
d1310b2e 2313 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
5f39d397
CM		 2314 buf);
2315}
6702ed49 2316
5f39d397
CM		 2317void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
2318{
810191ff 2319 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
5f39d397
CM		 2320 u64 transid = btrfs_header_generation(buf);
2321 struct inode *btree_inode = root->fs_info->btree_inode;
6702ed49 2322
925baedd 2323 WARN_ON(!btrfs_tree_locked(buf));
ccd467d6
CM		 2324 if (transid != root->fs_info->generation) {
2325 printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
db94535d 2326 (unsigned long long)buf->start,
ccd467d6
CM		 2327 transid, root->fs_info->generation);
2328 WARN_ON(1);
2329 }
d1310b2e 2330 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf);
eb60ceac
CM		 2331}
2332
d3c2fdcf 2333void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
35b7e476 2334{
188de649
CM		 2335 /*
2336 * looks as though older kernels can get into trouble with
2337 * this code, they end up stuck in balance_dirty_pages forever
2338 */
d6bfde87
CM		 2339 struct extent_io_tree *tree;
2340 u64 num_dirty;
2341 u64 start = 0;
771ed689 2342 unsigned long thresh = 32 * 1024 * 1024;
d6bfde87
CM		 2343 tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
2344
b64a2851 2345 if (current_is_pdflush() || current->flags & PF_MEMALLOC)
d6bfde87
CM		 2346 return;
2347
2348 num_dirty = count_range_bits(tree, &start, (u64)-1,
2349 thresh, EXTENT_DIRTY);
2350 if (num_dirty > thresh) {
2351 balance_dirty_pages_ratelimited_nr(
d7fc640e 2352 root->fs_info->btree_inode->i_mapping, 1);
d6bfde87 2353 }
188de649 2354 return;
35b7e476 2355}
6b80053d 2356
ca7a79ad 2357int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
6b80053d 2358{
810191ff 2359 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
ce9adaa5 2360 int ret;
ca7a79ad 2361 ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
ce9adaa5
CM		 2362 if (ret == 0) {
2363 buf->flags |= EXTENT_UPTODATE;
2364 }
2365 return ret;
6b80053d 2366}
0da5468f 2367
4bef0848
CM		 2368int btree_lock_page_hook(struct page *page)
2369{
2370 struct inode *inode = page->mapping->host;
2371 struct btrfs_root *root = BTRFS_I(inode)->root;
2372 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2373 struct extent_buffer *eb;
2374 unsigned long len;
2375 u64 bytenr = page_offset(page);
2376
2377 if (page->private == EXTENT_PAGE_PRIVATE)
2378 goto out;
2379
2380 len = page->private >> 2;
2381 eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS);
2382 if (!eb)
2383 goto out;
2384
2385 btrfs_tree_lock(eb);
2386 spin_lock(&root->fs_info->hash_lock);
2387 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
2388 spin_unlock(&root->fs_info->hash_lock);
2389 btrfs_tree_unlock(eb);
2390 free_extent_buffer(eb);
2391out:
2392 lock_page(page);
2393 return 0;
2394}
2395
d1310b2e 2396static struct extent_io_ops btree_extent_io_ops = {
4bef0848 2397 .write_cache_pages_lock_hook = btree_lock_page_hook,
ce9adaa5 2398 .readpage_end_io_hook = btree_readpage_end_io_hook,
0b86a832 2399 .submit_bio_hook = btree_submit_bio_hook,
239b14b3
CM		 2400 /* note we're sharing with inode.c for the merge bio hook */
2401 .merge_bio_hook = btrfs_merge_bio_hook,
0da5468f 2402};
Linux 6.x block layer and io_uring tree(s)