disk-io.c 66.5 KB
Newer Older
C
Chris Mason 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/*
 * Copyright (C) 2007 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

C
Chris Mason 已提交
19
#include <linux/fs.h>
20
#include <linux/blkdev.h>
C
Chris Mason 已提交
21
#include <linux/scatterlist.h>
C
Chris Mason 已提交
22
#include <linux/swap.h>
23
#include <linux/radix-tree.h>
C
Chris Mason 已提交
24
#include <linux/writeback.h>
C
Chris Mason 已提交
25
#include <linux/buffer_head.h>
26
#include <linux/workqueue.h>
27
#include <linux/kthread.h>
C
Chris Mason 已提交
28
#include <linux/freezer.h>
29
#include <linux/crc32c.h>
C
Chris Mason 已提交
30
#include "compat.h"
31 32
#include "ctree.h"
#include "disk-io.h"
33
#include "transaction.h"
34
#include "btrfs_inode.h"
35
#include "volumes.h"
36
#include "print-tree.h"
37
#include "async-thread.h"
38
#include "locking.h"
39
#include "tree-log.h"
40
#include "free-space-cache.h"
41

42
static struct extent_io_ops btree_extent_io_ops;
43
static void end_workqueue_fn(struct btrfs_work *work);
44

45 46
static atomic_t btrfs_bdi_num = ATOMIC_INIT(0);

C
Chris Mason 已提交
47 48 49 50 51
/*
 * end_io_wq structs are used to do processing in task context when an IO is
 * complete.  This is used during reads to verify checksums, and it is used
 * by writes to insert metadata for new file extents after IO is complete.
 */
52 53 54 55 56 57
struct end_io_wq {
	struct bio *bio;
	bio_end_io_t *end_io;
	void *private;
	struct btrfs_fs_info *info;
	int error;
58
	int metadata;
59
	struct list_head list;
60
	struct btrfs_work work;
61
};
62

C
Chris Mason 已提交
63 64 65 66 67
/*
 * async submit bios are used to offload expensive checksumming
 * onto the worker threads.  They checksum file and metadata bios
 * just before they are sent down the IO stack.
 */
68 69 70 71
struct async_submit_bio {
	struct inode *inode;
	struct bio *bio;
	struct list_head list;
C
Chris Mason 已提交
72 73
	extent_submit_bio_hook_t *submit_bio_start;
	extent_submit_bio_hook_t *submit_bio_done;
74 75
	int rw;
	int mirror_num;
C
Chris Mason 已提交
76
	unsigned long bio_flags;
77
	struct btrfs_work work;
78 79
};

80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113
/* These are used to set the lockdep class on the extent buffer locks.
 * The class is set by the readpage_end_io_hook after the buffer has
 * passed csum validation but before the pages are unlocked.
 *
 * The lockdep class is also set by btrfs_init_new_buffer on freshly
 * allocated blocks.
 *
 * The class is based on the level in the tree block, which allows lockdep
 * to know that lower nodes nest inside the locks of higher nodes.
 *
 * We also add a check to make sure the highest level of the tree is
 * the same as our lockdep setup here.  If BTRFS_MAX_LEVEL changes, this
 * code needs update as well.
 */
#ifdef CONFIG_DEBUG_LOCK_ALLOC
# if BTRFS_MAX_LEVEL != 8
#  error
# endif
static struct lock_class_key btrfs_eb_class[BTRFS_MAX_LEVEL + 1];
static const char *btrfs_eb_name[BTRFS_MAX_LEVEL + 1] = {
	/* leaf */
	"btrfs-extent-00",
	"btrfs-extent-01",
	"btrfs-extent-02",
	"btrfs-extent-03",
	"btrfs-extent-04",
	"btrfs-extent-05",
	"btrfs-extent-06",
	"btrfs-extent-07",
	/* highest possible level */
	"btrfs-extent-08",
};
#endif

C
Chris Mason 已提交
114 115 116 117
/*
 * extents on the btree inode are pretty simple, there's one extent
 * that covers the entire device
 */
118 119 120
static struct extent_map *btree_get_extent(struct inode *inode,
		struct page *page, size_t page_offset, u64 start, u64 len,
		int create)
121
{
122 123 124 125
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct extent_map *em;
	int ret;

126 127
	spin_lock(&em_tree->lock);
	em = lookup_extent_mapping(em_tree, start, len);
128 129 130 131
	if (em) {
		em->bdev =
			BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
		spin_unlock(&em_tree->lock);
132
		goto out;
133 134
	}
	spin_unlock(&em_tree->lock);
135

136 137 138 139 140 141
	em = alloc_extent_map(GFP_NOFS);
	if (!em) {
		em = ERR_PTR(-ENOMEM);
		goto out;
	}
	em->start = 0;
142
	em->len = (u64)-1;
C
Chris Mason 已提交
143
	em->block_len = (u64)-1;
144
	em->block_start = 0;
145
	em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
146 147

	spin_lock(&em_tree->lock);
148 149
	ret = add_extent_mapping(em_tree, em);
	if (ret == -EEXIST) {
150 151 152
		u64 failed_start = em->start;
		u64 failed_len = em->len;

153
		free_extent_map(em);
154
		em = lookup_extent_mapping(em_tree, start, len);
155
		if (em) {
156
			ret = 0;
157 158 159
		} else {
			em = lookup_extent_mapping(em_tree, failed_start,
						   failed_len);
160
			ret = -EIO;
161
		}
162
	} else if (ret) {
163 164
		free_extent_map(em);
		em = NULL;
165
	}
166 167 168 169
	spin_unlock(&em_tree->lock);

	if (ret)
		em = ERR_PTR(ret);
170 171
out:
	return em;
172 173
}

174 175
u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
{
176
	return crc32c(seed, data, len);
177 178 179 180 181 182 183
}

void btrfs_csum_final(u32 crc, char *result)
{
	*(__le32 *)result = ~cpu_to_le32(crc);
}

C
Chris Mason 已提交
184 185 186 187
/*
 * compute the csum for a btree block, and either verify it or write it
 * into the csum field of the block.
 */
188 189 190
static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
			   int verify)
{
191 192 193
	u16 csum_size =
		btrfs_super_csum_size(&root->fs_info->super_copy);
	char *result = NULL;
194 195 196 197 198 199 200 201 202
	unsigned long len;
	unsigned long cur_len;
	unsigned long offset = BTRFS_CSUM_SIZE;
	char *map_token = NULL;
	char *kaddr;
	unsigned long map_start;
	unsigned long map_len;
	int err;
	u32 crc = ~(u32)0;
203
	unsigned long inline_result;
204 205

	len = buf->len - offset;
C
Chris Mason 已提交
206
	while (len > 0) {
207 208 209
		err = map_private_extent_buffer(buf, offset, 32,
					&map_token, &kaddr,
					&map_start, &map_len, KM_USER0);
C
Chris Mason 已提交
210
		if (err)
211 212 213 214 215 216 217 218
			return 1;
		cur_len = min(len, map_len - (offset - map_start));
		crc = btrfs_csum_data(root, kaddr + offset - map_start,
				      crc, cur_len);
		len -= cur_len;
		offset += cur_len;
		unmap_extent_buffer(buf, map_token, KM_USER0);
	}
219 220 221 222 223 224 225 226
	if (csum_size > sizeof(inline_result)) {
		result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
		if (!result)
			return 1;
	} else {
		result = (char *)&inline_result;
	}

227 228 229
	btrfs_csum_final(crc, result);

	if (verify) {
230
		if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
231 232
			u32 val;
			u32 found = 0;
233
			memcpy(&found, result, csum_size);
234

235
			read_extent_buffer(buf, &val, 0, csum_size);
C
Chris Mason 已提交
236 237 238 239 240 241 242 243
			if (printk_ratelimit()) {
				printk(KERN_INFO "btrfs: %s checksum verify "
				       "failed on %llu wanted %X found %X "
				       "level %d\n",
				       root->fs_info->sb->s_id,
				       (unsigned long long)buf->start, val, found,
				       btrfs_header_level(buf));
			}
244 245
			if (result != (char *)&inline_result)
				kfree(result);
246 247 248
			return 1;
		}
	} else {
249
		write_extent_buffer(buf, result, 0, csum_size);
250
	}
251 252
	if (result != (char *)&inline_result)
		kfree(result);
253 254 255
	return 0;
}

C
Chris Mason 已提交
256 257 258 259 260 261
/*
 * we can't consider a given block up to date unless the transid of the
 * block matches the transid in the parent node's pointer.  This is how we
 * detect blocks that either didn't get written at all or got written
 * in the wrong place.
 */
262 263 264 265 266 267 268 269 270 271 272 273 274 275
static int verify_parent_transid(struct extent_io_tree *io_tree,
				 struct extent_buffer *eb, u64 parent_transid)
{
	int ret;

	if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
		return 0;

	lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS);
	if (extent_buffer_uptodate(io_tree, eb) &&
	    btrfs_header_generation(eb) == parent_transid) {
		ret = 0;
		goto out;
	}
C
Chris Mason 已提交
276 277 278 279 280 281 282
	if (printk_ratelimit()) {
		printk("parent transid verify failed on %llu wanted %llu "
		       "found %llu\n",
		       (unsigned long long)eb->start,
		       (unsigned long long)parent_transid,
		       (unsigned long long)btrfs_header_generation(eb));
	}
283 284
	ret = 1;
	clear_extent_buffer_uptodate(io_tree, eb);
C
Chris Mason 已提交
285
out:
286 287 288 289 290
	unlock_extent(io_tree, eb->start, eb->start + eb->len - 1,
		      GFP_NOFS);
	return ret;
}

C
Chris Mason 已提交
291 292 293 294
/*
 * helper to read a given tree block, doing retries as required when
 * the checksums don't match and we have alternate mirrors to try.
 */
295 296
static int btree_read_extent_buffer_pages(struct btrfs_root *root,
					  struct extent_buffer *eb,
297
					  u64 start, u64 parent_transid)
298 299 300 301 302 303 304 305 306 307
{
	struct extent_io_tree *io_tree;
	int ret;
	int num_copies = 0;
	int mirror_num = 0;

	io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
	while (1) {
		ret = read_extent_buffer_pages(io_tree, eb, start, 1,
					       btree_get_extent, mirror_num);
308 309
		if (!ret &&
		    !verify_parent_transid(io_tree, eb, parent_transid))
310
			return ret;
C
Chris Mason 已提交
311

312 313
		num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
					      eb->start, eb->len);
C
Chris Mason 已提交
314
		if (num_copies == 1)
315
			return ret;
C
Chris Mason 已提交
316

317
		mirror_num++;
C
Chris Mason 已提交
318
		if (mirror_num > num_copies)
319 320 321 322
			return ret;
	}
	return -EIO;
}
323

C
Chris Mason 已提交
324
/*
C
Chris Mason 已提交
325 326
 * checksum a dirty tree block before IO.  This has extra checks to make sure
 * we only fill in the checksum field in the first page of a multi-page block
C
Chris Mason 已提交
327
 */
C
Chris Mason 已提交
328

329
static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
330
{
331
	struct extent_io_tree *tree;
332
	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
333 334 335 336
	u64 found_start;
	int found_level;
	unsigned long len;
	struct extent_buffer *eb;
337 338
	int ret;

339
	tree = &BTRFS_I(page->mapping->host)->io_tree;
340 341 342 343 344 345

	if (page->private == EXTENT_PAGE_PRIVATE)
		goto out;
	if (!page->private)
		goto out;
	len = page->private >> 2;
C
Chris Mason 已提交
346 347
	WARN_ON(len == 0);

348
	eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
349 350
	ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
					     btrfs_header_generation(eb));
351
	BUG_ON(ret);
352 353
	found_start = btrfs_header_bytenr(eb);
	if (found_start != start) {
354 355 356 357 358 359 360 361 362 363
		WARN_ON(1);
		goto err;
	}
	if (eb->first_page != page) {
		WARN_ON(1);
		goto err;
	}
	if (!PageUptodate(page)) {
		WARN_ON(1);
		goto err;
364 365
	}
	found_level = btrfs_header_level(eb);
C
Chris Mason 已提交
366

367
	csum_tree_block(root, eb, 0);
368
err:
369 370 371 372 373
	free_extent_buffer(eb);
out:
	return 0;
}

Y
Yan Zheng 已提交
374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
static int check_tree_block_fsid(struct btrfs_root *root,
				 struct extent_buffer *eb)
{
	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
	u8 fsid[BTRFS_UUID_SIZE];
	int ret = 1;

	read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
			   BTRFS_FSID_SIZE);
	while (fs_devices) {
		if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
			ret = 0;
			break;
		}
		fs_devices = fs_devices->seed;
	}
	return ret;
}

393 394 395 396 397 398 399 400 401
#ifdef CONFIG_DEBUG_LOCK_ALLOC
void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level)
{
	lockdep_set_class_and_name(&eb->lock,
			   &btrfs_eb_class[level],
			   btrfs_eb_name[level]);
}
#endif

402
static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
403 404 405 406 407 408 409 410
			       struct extent_state *state)
{
	struct extent_io_tree *tree;
	u64 found_start;
	int found_level;
	unsigned long len;
	struct extent_buffer *eb;
	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
411
	int ret = 0;
412 413 414 415 416 417

	tree = &BTRFS_I(page->mapping->host)->io_tree;
	if (page->private == EXTENT_PAGE_PRIVATE)
		goto out;
	if (!page->private)
		goto out;
C
Chris Mason 已提交
418

419
	len = page->private >> 2;
C
Chris Mason 已提交
420 421
	WARN_ON(len == 0);

422
	eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
423

424
	found_start = btrfs_header_bytenr(eb);
425
	if (found_start != start) {
C
Chris Mason 已提交
426 427 428 429 430 431
		if (printk_ratelimit()) {
			printk(KERN_INFO "btrfs bad tree block start "
			       "%llu %llu\n",
			       (unsigned long long)found_start,
			       (unsigned long long)eb->start);
		}
432
		ret = -EIO;
433 434 435
		goto err;
	}
	if (eb->first_page != page) {
C
Chris Mason 已提交
436 437
		printk(KERN_INFO "btrfs bad first page %lu %lu\n",
		       eb->first_page->index, page->index);
438
		WARN_ON(1);
439
		ret = -EIO;
440 441
		goto err;
	}
Y
Yan Zheng 已提交
442
	if (check_tree_block_fsid(root, eb)) {
C
Chris Mason 已提交
443 444 445 446
		if (printk_ratelimit()) {
			printk(KERN_INFO "btrfs bad fsid on block %llu\n",
			       (unsigned long long)eb->start);
		}
447 448 449
		ret = -EIO;
		goto err;
	}
450 451
	found_level = btrfs_header_level(eb);

452 453
	btrfs_set_buffer_lockdep_class(eb, found_level);

454
	ret = csum_tree_block(root, eb, 1);
455 456
	if (ret)
		ret = -EIO;
457 458 459 460 461 462

	end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
	end = eb->start + end - 1;
err:
	free_extent_buffer(eb);
out:
463
	return ret;
464 465 466 467 468 469 470 471 472
}

static void end_workqueue_bio(struct bio *bio, int err)
{
	struct end_io_wq *end_io_wq = bio->bi_private;
	struct btrfs_fs_info *fs_info;

	fs_info = end_io_wq->info;
	end_io_wq->error = err;
473 474
	end_io_wq->work.func = end_workqueue_fn;
	end_io_wq->work.flags = 0;
475 476

	if (bio->bi_rw & (1 << BIO_RW)) {
477 478 479 480 481 482
		if (end_io_wq->metadata)
			btrfs_queue_worker(&fs_info->endio_meta_write_workers,
					   &end_io_wq->work);
		else
			btrfs_queue_worker(&fs_info->endio_write_workers,
					   &end_io_wq->work);
483 484 485 486 487 488 489 490
	} else {
		if (end_io_wq->metadata)
			btrfs_queue_worker(&fs_info->endio_meta_workers,
					   &end_io_wq->work);
		else
			btrfs_queue_worker(&fs_info->endio_workers,
					   &end_io_wq->work);
	}
491 492
}

493 494
int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
			int metadata)
495
{
496 497 498 499 500 501 502
	struct end_io_wq *end_io_wq;
	end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
	if (!end_io_wq)
		return -ENOMEM;

	end_io_wq->private = bio->bi_private;
	end_io_wq->end_io = bio->bi_end_io;
503
	end_io_wq->info = info;
504 505
	end_io_wq->error = 0;
	end_io_wq->bio = bio;
506
	end_io_wq->metadata = metadata;
507 508 509

	bio->bi_private = end_io_wq;
	bio->bi_end_io = end_workqueue_bio;
510 511 512
	return 0;
}

513
unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
514
{
515 516 517 518 519
	unsigned long limit = min_t(unsigned long,
				    info->workers.max_workers,
				    info->fs_devices->open_devices);
	return 256 * limit;
}
520

521 522
int btrfs_congested_async(struct btrfs_fs_info *info, int iodone)
{
523 524
	return atomic_read(&info->nr_async_bios) >
		btrfs_async_submit_limit(info);
525 526
}

C
Chris Mason 已提交
527 528 529 530 531 532 533 534 535 536 537 538
static void run_one_async_start(struct btrfs_work *work)
{
	struct btrfs_fs_info *fs_info;
	struct async_submit_bio *async;

	async = container_of(work, struct  async_submit_bio, work);
	fs_info = BTRFS_I(async->inode)->root->fs_info;
	async->submit_bio_start(async->inode, async->rw, async->bio,
			       async->mirror_num, async->bio_flags);
}

static void run_one_async_done(struct btrfs_work *work)
539 540 541
{
	struct btrfs_fs_info *fs_info;
	struct async_submit_bio *async;
542
	int limit;
543 544 545

	async = container_of(work, struct  async_submit_bio, work);
	fs_info = BTRFS_I(async->inode)->root->fs_info;
546

547
	limit = btrfs_async_submit_limit(fs_info);
548 549
	limit = limit * 2 / 3;

550
	atomic_dec(&fs_info->nr_async_submits);
551

552 553
	if (atomic_read(&fs_info->nr_async_submits) < limit &&
	    waitqueue_active(&fs_info->async_submit_wait))
554 555
		wake_up(&fs_info->async_submit_wait);

C
Chris Mason 已提交
556
	async->submit_bio_done(async->inode, async->rw, async->bio,
C
Chris Mason 已提交
557
			       async->mirror_num, async->bio_flags);
C
Chris Mason 已提交
558 559 560 561 562 563 564
}

static void run_one_async_free(struct btrfs_work *work)
{
	struct async_submit_bio *async;

	async = container_of(work, struct  async_submit_bio, work);
565 566 567
	kfree(async);
}

568 569
int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
			int rw, struct bio *bio, int mirror_num,
C
Chris Mason 已提交
570
			unsigned long bio_flags,
C
Chris Mason 已提交
571 572
			extent_submit_bio_hook_t *submit_bio_start,
			extent_submit_bio_hook_t *submit_bio_done)
573 574 575 576 577 578 579 580 581 582 583
{
	struct async_submit_bio *async;

	async = kmalloc(sizeof(*async), GFP_NOFS);
	if (!async)
		return -ENOMEM;

	async->inode = inode;
	async->rw = rw;
	async->bio = bio;
	async->mirror_num = mirror_num;
C
Chris Mason 已提交
584 585 586 587 588 589 590
	async->submit_bio_start = submit_bio_start;
	async->submit_bio_done = submit_bio_done;

	async->work.func = run_one_async_start;
	async->work.ordered_func = run_one_async_done;
	async->work.ordered_free = run_one_async_free;

591
	async->work.flags = 0;
C
Chris Mason 已提交
592
	async->bio_flags = bio_flags;
593

594
	atomic_inc(&fs_info->nr_async_submits);
595 596 597 598

	if (rw & (1 << BIO_RW_SYNCIO))
		btrfs_set_work_high_prio(&async->work);

599
	btrfs_queue_worker(&fs_info->workers, &async->work);
600

C
Chris Mason 已提交
601
	while (atomic_read(&fs_info->async_submit_draining) &&
602 603 604 605 606
	      atomic_read(&fs_info->nr_async_submits)) {
		wait_event(fs_info->async_submit_wait,
			   (atomic_read(&fs_info->nr_async_submits) == 0));
	}

607 608 609
	return 0;
}

610 611 612 613 614 615 616
static int btree_csum_one_bio(struct bio *bio)
{
	struct bio_vec *bvec = bio->bi_io_vec;
	int bio_index = 0;
	struct btrfs_root *root;

	WARN_ON(bio->bi_vcnt <= 0);
C
Chris Mason 已提交
617
	while (bio_index < bio->bi_vcnt) {
618 619 620 621 622 623 624 625
		root = BTRFS_I(bvec->bv_page->mapping->host)->root;
		csum_dirty_buffer(root, bvec->bv_page);
		bio_index++;
		bvec++;
	}
	return 0;
}

C
Chris Mason 已提交
626 627 628
static int __btree_submit_bio_start(struct inode *inode, int rw,
				    struct bio *bio, int mirror_num,
				    unsigned long bio_flags)
629
{
630 631
	/*
	 * when we're called for a write, we're already in the async
632
	 * submission context.  Just jump into btrfs_map_bio
633
	 */
C
Chris Mason 已提交
634 635 636
	btree_csum_one_bio(bio);
	return 0;
}
637

C
Chris Mason 已提交
638 639 640
static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
				 int mirror_num, unsigned long bio_flags)
{
641
	/*
C
Chris Mason 已提交
642 643
	 * when we're called for a write, we're already in the async
	 * submission context.  Just jump into btrfs_map_bio
644 645
	 */
	return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
646 647
}

648
static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
C
Chris Mason 已提交
649
				 int mirror_num, unsigned long bio_flags)
650
{
651 652 653 654 655 656
	int ret;

	ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
					  bio, 1);
	BUG_ON(ret);

657
	if (!(rw & (1 << BIO_RW))) {
C
Chris Mason 已提交
658 659 660 661 662
		/*
		 * called for a read, do the setup so that checksum validation
		 * can happen in the async kernel threads
		 */
		return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
663
				     mirror_num, 0);
664
	}
665

666 667 668 669
	/*
	 * kthread helpers are used to submit writes so that checksumming
	 * can happen in parallel across all CPUs
	 */
670
	return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
C
Chris Mason 已提交
671
				   inode, rw, bio, mirror_num, 0,
C
Chris Mason 已提交
672 673
				   __btree_submit_bio_start,
				   __btree_submit_bio_done);
674 675
}

676 677
static int btree_writepage(struct page *page, struct writeback_control *wbc)
{
678
	struct extent_io_tree *tree;
679 680 681 682
	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
	struct extent_buffer *eb;
	int was_dirty;

683
	tree = &BTRFS_I(page->mapping->host)->io_tree;
684 685 686 687
	if (!(current->flags & PF_MEMALLOC)) {
		return extent_write_full_page(tree, page,
					      btree_get_extent, wbc);
	}
688

689 690 691 692 693 694 695 696 697 698
	redirty_page_for_writepage(wbc, page);
	eb = btrfs_find_tree_block(root, page_offset(page),
				      PAGE_CACHE_SIZE);
	WARN_ON(!eb);

	was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
	if (!was_dirty) {
		spin_lock(&root->fs_info->delalloc_lock);
		root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE;
		spin_unlock(&root->fs_info->delalloc_lock);
699
	}
700 701 702 703
	free_extent_buffer(eb);

	unlock_page(page);
	return 0;
704
}
705 706 707 708

static int btree_writepages(struct address_space *mapping,
			    struct writeback_control *wbc)
{
709 710
	struct extent_io_tree *tree;
	tree = &BTRFS_I(mapping->host)->io_tree;
711
	if (wbc->sync_mode == WB_SYNC_NONE) {
712
		struct btrfs_root *root = BTRFS_I(mapping->host)->root;
713
		u64 num_dirty;
714
		unsigned long thresh = 32 * 1024 * 1024;
715 716 717 718

		if (wbc->for_kupdate)
			return 0;

719 720
		/* this is a bit racy, but that's ok */
		num_dirty = root->fs_info->dirty_metadata_bytes;
C
Chris Mason 已提交
721
		if (num_dirty < thresh)
722 723
			return 0;
	}
724 725 726
	return extent_writepages(tree, mapping, btree_get_extent, wbc);
}

727
static int btree_readpage(struct file *file, struct page *page)
728
{
729 730
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
731 732
	return extent_read_full_page(tree, page, btree_get_extent);
}
C
Chris Mason 已提交
733

734
static int btree_releasepage(struct page *page, gfp_t gfp_flags)
735
{
736 737
	struct extent_io_tree *tree;
	struct extent_map_tree *map;
738
	int ret;
739

740
	if (PageWriteback(page) || PageDirty(page))
C
Chris Mason 已提交
741
		return 0;
742

743 744
	tree = &BTRFS_I(page->mapping->host)->io_tree;
	map = &BTRFS_I(page->mapping->host)->extent_tree;
745

746
	ret = try_release_extent_state(map, tree, page, gfp_flags);
C
Chris Mason 已提交
747
	if (!ret)
748 749 750
		return 0;

	ret = try_release_extent_buffer(tree, page);
751 752 753 754 755
	if (ret == 1) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
	}
756

757 758 759
	return ret;
}

760
static void btree_invalidatepage(struct page *page, unsigned long offset)
761
{
762 763
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
764 765
	extent_invalidatepage(tree, page, offset);
	btree_releasepage(page, GFP_NOFS);
766
	if (PagePrivate(page)) {
C
Chris Mason 已提交
767 768
		printk(KERN_WARNING "btrfs warning page private not zero "
		       "on page %llu\n", (unsigned long long)page_offset(page));
769 770 771 772
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
	}
773 774 775 776 777
}

static struct address_space_operations btree_aops = {
	.readpage	= btree_readpage,
	.writepage	= btree_writepage,
778
	.writepages	= btree_writepages,
779 780
	.releasepage	= btree_releasepage,
	.invalidatepage = btree_invalidatepage,
781 782 783
	.sync_page	= block_sync_page,
};

784 785
int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
			 u64 parent_transid)
C
Chris Mason 已提交
786
{
787 788
	struct extent_buffer *buf = NULL;
	struct inode *btree_inode = root->fs_info->btree_inode;
789
	int ret = 0;
C
Chris Mason 已提交
790

791
	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
792
	if (!buf)
C
Chris Mason 已提交
793
		return 0;
794
	read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
795
				 buf, 0, 0, btree_get_extent, 0);
796
	free_extent_buffer(buf);
797
	return ret;
C
Chris Mason 已提交
798 799
}

800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821
struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
					    u64 bytenr, u32 blocksize)
{
	struct inode *btree_inode = root->fs_info->btree_inode;
	struct extent_buffer *eb;
	eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
				bytenr, blocksize, GFP_NOFS);
	return eb;
}

struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
						 u64 bytenr, u32 blocksize)
{
	struct inode *btree_inode = root->fs_info->btree_inode;
	struct extent_buffer *eb;

	eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
				 bytenr, blocksize, NULL, GFP_NOFS);
	return eb;
}


822 823 824
int btrfs_write_tree_block(struct extent_buffer *buf)
{
	return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start,
825
				      buf->start + buf->len - 1, WB_SYNC_ALL);
826 827 828 829 830
}

int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
{
	return btrfs_wait_on_page_writeback_range(buf->first_page->mapping,
C
Chris Mason 已提交
831
				  buf->start, buf->start + buf->len - 1);
832 833
}

834
struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
835
				      u32 blocksize, u64 parent_transid)
836 837 838 839 840 841 842 843 844 845 846 847
{
	struct extent_buffer *buf = NULL;
	struct inode *btree_inode = root->fs_info->btree_inode;
	struct extent_io_tree *io_tree;
	int ret;

	io_tree = &BTRFS_I(btree_inode)->io_tree;

	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
	if (!buf)
		return NULL;

848
	ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
849

C
Chris Mason 已提交
850
	if (ret == 0)
851
		set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
852
	return buf;
853

854 855
}

856
int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
857
		     struct extent_buffer *buf)
858
{
859
	struct inode *btree_inode = root->fs_info->btree_inode;
860
	if (btrfs_header_generation(buf) ==
861
	    root->fs_info->running_transaction->transid) {
862
		btrfs_assert_tree_locked(buf);
863

864 865 866 867 868 869 870 871
		if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
			spin_lock(&root->fs_info->delalloc_lock);
			if (root->fs_info->dirty_metadata_bytes >= buf->len)
				root->fs_info->dirty_metadata_bytes -= buf->len;
			else
				WARN_ON(1);
			spin_unlock(&root->fs_info->delalloc_lock);
		}
872

873 874
		/* ugh, clear_extent_buffer_dirty needs to lock the page */
		btrfs_set_lock_blocking(buf);
875
		clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
876
					  buf);
877
	}
878 879 880
	return 0;
}

881
static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
882
			u32 stripesize, struct btrfs_root *root,
883
			struct btrfs_fs_info *fs_info,
C
Chris Mason 已提交
884
			u64 objectid)
885
{
C
Chris Mason 已提交
886
	root->node = NULL;
887
	root->commit_root = NULL;
888 889 890
	root->sectorsize = sectorsize;
	root->nodesize = nodesize;
	root->leafsize = leafsize;
891
	root->stripesize = stripesize;
C
Chris Mason 已提交
892
	root->ref_cows = 0;
893 894
	root->track_dirty = 0;

895
	root->fs_info = fs_info;
896 897
	root->objectid = objectid;
	root->last_trans = 0;
C
Chris Mason 已提交
898 899
	root->highest_inode = 0;
	root->last_inode_alloc = 0;
900
	root->name = NULL;
901
	root->in_sysfs = 0;
902
	root->inode_tree.rb_node = NULL;
903 904

	INIT_LIST_HEAD(&root->dirty_list);
905
	INIT_LIST_HEAD(&root->orphan_list);
906
	INIT_LIST_HEAD(&root->root_list);
907
	spin_lock_init(&root->node_lock);
908
	spin_lock_init(&root->list_lock);
909
	spin_lock_init(&root->inode_lock);
910
	mutex_init(&root->objectid_mutex);
911
	mutex_init(&root->log_mutex);
Y
Yan Zheng 已提交
912 913 914 915 916 917 918 919
	init_waitqueue_head(&root->log_writer_wait);
	init_waitqueue_head(&root->log_commit_wait[0]);
	init_waitqueue_head(&root->log_commit_wait[1]);
	atomic_set(&root->log_commit[0], 0);
	atomic_set(&root->log_commit[1], 0);
	atomic_set(&root->log_writers, 0);
	root->log_batch = 0;
	root->log_transid = 0;
920 921
	extent_io_tree_init(&root->dirty_log_pages,
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
C
Chris Mason 已提交
922

923 924
	memset(&root->root_key, 0, sizeof(root->root_key));
	memset(&root->root_item, 0, sizeof(root->root_item));
925
	memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
926
	memset(&root->root_kobj, 0, sizeof(root->root_kobj));
927
	root->defrag_trans_start = fs_info->generation;
928
	init_completion(&root->kobj_unregister);
929 930
	root->defrag_running = 0;
	root->defrag_level = 0;
931
	root->root_key.objectid = objectid;
932 933 934 935 936 937
	root->anon_super.s_root = NULL;
	root->anon_super.s_dev = 0;
	INIT_LIST_HEAD(&root->anon_super.s_list);
	INIT_LIST_HEAD(&root->anon_super.s_instances);
	init_rwsem(&root->anon_super.s_umount);

938 939 940
	return 0;
}

941
static int find_and_setup_root(struct btrfs_root *tree_root,
942 943
			       struct btrfs_fs_info *fs_info,
			       u64 objectid,
C
Chris Mason 已提交
944
			       struct btrfs_root *root)
945 946
{
	int ret;
947
	u32 blocksize;
948
	u64 generation;
949

950
	__setup_root(tree_root->nodesize, tree_root->leafsize,
951 952
		     tree_root->sectorsize, tree_root->stripesize,
		     root, fs_info, objectid);
953 954 955 956
	ret = btrfs_find_last_root(tree_root, objectid,
				   &root->root_item, &root->root_key);
	BUG_ON(ret);

957
	generation = btrfs_root_generation(&root->root_item);
958 959
	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
960
				     blocksize, generation);
961
	root->commit_root = btrfs_root_node(root);
962
	BUG_ON(!root->node);
963 964 965
	return 0;
}

966 967 968 969
int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
			     struct btrfs_fs_info *fs_info)
{
	struct extent_buffer *eb;
970 971 972
	struct btrfs_root *log_root_tree = fs_info->log_root_tree;
	u64 start = 0;
	u64 end = 0;
973 974
	int ret;

975
	if (!log_root_tree)
976 977
		return 0;

C
Chris Mason 已提交
978
	while (1) {
979 980 981 982 983 984 985 986
		ret = find_first_extent_bit(&log_root_tree->dirty_log_pages,
				    0, &start, &end, EXTENT_DIRTY);
		if (ret)
			break;

		clear_extent_dirty(&log_root_tree->dirty_log_pages,
				   start, end, GFP_NOFS);
	}
987 988 989 990 991
	eb = fs_info->log_root_tree->node;

	WARN_ON(btrfs_header_level(eb) != 0);
	WARN_ON(btrfs_header_nritems(eb) != 0);

992 993
	ret = btrfs_free_reserved_extent(fs_info->tree_root,
				eb->start, eb->len);
994 995 996 997 998 999 1000 1001
	BUG_ON(ret);

	free_extent_buffer(eb);
	kfree(fs_info->log_root_tree);
	fs_info->log_root_tree = NULL;
	return 0;
}

Y
Yan Zheng 已提交
1002 1003
static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
					 struct btrfs_fs_info *fs_info)
1004 1005 1006
{
	struct btrfs_root *root;
	struct btrfs_root *tree_root = fs_info->tree_root;
Y
Yan Zheng 已提交
1007
	struct extent_buffer *leaf;
1008 1009 1010

	root = kzalloc(sizeof(*root), GFP_NOFS);
	if (!root)
Y
Yan Zheng 已提交
1011
		return ERR_PTR(-ENOMEM);
1012 1013 1014 1015 1016 1017 1018 1019

	__setup_root(tree_root->nodesize, tree_root->leafsize,
		     tree_root->sectorsize, tree_root->stripesize,
		     root, fs_info, BTRFS_TREE_LOG_OBJECTID);

	root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
	root->root_key.type = BTRFS_ROOT_ITEM_KEY;
	root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
Y
Yan Zheng 已提交
1020 1021 1022 1023 1024 1025
	/*
	 * log trees do not get reference counted because they go away
	 * before a real commit is actually done.  They do store pointers
	 * to file data extents, and those reference counts still get
	 * updated (along with back refs to the log tree).
	 */
1026 1027
	root->ref_cows = 0;

1028 1029
	leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
				      BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0);
Y
Yan Zheng 已提交
1030 1031 1032 1033
	if (IS_ERR(leaf)) {
		kfree(root);
		return ERR_CAST(leaf);
	}
1034

1035 1036 1037 1038 1039
	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
	btrfs_set_header_bytenr(leaf, leaf->start);
	btrfs_set_header_generation(leaf, trans->transid);
	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
	btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
Y
Yan Zheng 已提交
1040
	root->node = leaf;
1041 1042 1043 1044 1045 1046

	write_extent_buffer(root->node, root->fs_info->fsid,
			    (unsigned long)btrfs_header_fsid(root->node),
			    BTRFS_FSID_SIZE);
	btrfs_mark_buffer_dirty(root->node);
	btrfs_tree_unlock(root->node);
Y
Yan Zheng 已提交
1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082
	return root;
}

int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
			     struct btrfs_fs_info *fs_info)
{
	struct btrfs_root *log_root;

	log_root = alloc_log_tree(trans, fs_info);
	if (IS_ERR(log_root))
		return PTR_ERR(log_root);
	WARN_ON(fs_info->log_root_tree);
	fs_info->log_root_tree = log_root;
	return 0;
}

int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
		       struct btrfs_root *root)
{
	struct btrfs_root *log_root;
	struct btrfs_inode_item *inode_item;

	log_root = alloc_log_tree(trans, root->fs_info);
	if (IS_ERR(log_root))
		return PTR_ERR(log_root);

	log_root->last_trans = trans->transid;
	log_root->root_key.offset = root->root_key.objectid;

	inode_item = &log_root->root_item.inode;
	inode_item->generation = cpu_to_le64(1);
	inode_item->size = cpu_to_le64(3);
	inode_item->nlink = cpu_to_le32(1);
	inode_item->nbytes = cpu_to_le64(root->leafsize);
	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);

1083
	btrfs_set_root_node(&log_root->root_item, log_root->node);
Y
Yan Zheng 已提交
1084 1085 1086 1087

	WARN_ON(root->log_root);
	root->log_root = log_root;
	root->log_transid = 0;
1088 1089 1090 1091 1092 1093 1094 1095
	return 0;
}

struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
					       struct btrfs_key *location)
{
	struct btrfs_root *root;
	struct btrfs_fs_info *fs_info = tree_root->fs_info;
1096
	struct btrfs_path *path;
1097
	struct extent_buffer *l;
C
Chris Mason 已提交
1098
	u64 highest_inode;
1099
	u64 generation;
1100
	u32 blocksize;
1101 1102
	int ret = 0;

1103
	root = kzalloc(sizeof(*root), GFP_NOFS);
C
Chris Mason 已提交
1104
	if (!root)
1105 1106
		return ERR_PTR(-ENOMEM);
	if (location->offset == (u64)-1) {
1107
		ret = find_and_setup_root(tree_root, fs_info,
1108 1109 1110 1111 1112 1113 1114 1115
					  location->objectid, root);
		if (ret) {
			kfree(root);
			return ERR_PTR(ret);
		}
		goto insert;
	}

1116
	__setup_root(tree_root->nodesize, tree_root->leafsize,
1117 1118
		     tree_root->sectorsize, tree_root->stripesize,
		     root, fs_info, location->objectid);
1119 1120 1121 1122 1123 1124 1125 1126 1127

	path = btrfs_alloc_path();
	BUG_ON(!path);
	ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
	if (ret != 0) {
		if (ret > 0)
			ret = -ENOENT;
		goto out;
	}
1128 1129 1130
	l = path->nodes[0];
	read_extent_buffer(l, &root->root_item,
	       btrfs_item_ptr_offset(l, path->slots[0]),
1131
	       sizeof(root->root_item));
1132
	memcpy(&root->root_key, location, sizeof(*location));
1133 1134 1135 1136 1137 1138 1139 1140
	ret = 0;
out:
	btrfs_release_path(root, path);
	btrfs_free_path(path);
	if (ret) {
		kfree(root);
		return ERR_PTR(ret);
	}
1141
	generation = btrfs_root_generation(&root->root_item);
1142 1143
	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
1144
				     blocksize, generation);
1145
	root->commit_root = btrfs_root_node(root);
1146 1147
	BUG_ON(!root->node);
insert:
1148 1149 1150 1151 1152 1153 1154
	if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
		root->ref_cows = 1;
		ret = btrfs_find_highest_inode(root, &highest_inode);
		if (ret == 0) {
			root->highest_inode = highest_inode;
			root->last_inode_alloc = highest_inode;
		}
1155 1156 1157 1158
	}
	return root;
}

C
Chris Mason 已提交
1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173
struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
					u64 root_objectid)
{
	struct btrfs_root *root;

	if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
		return fs_info->tree_root;
	if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
		return fs_info->extent_root;

	root = radix_tree_lookup(&fs_info->fs_roots_radix,
				 (unsigned long)root_objectid);
	return root;
}

1174 1175
struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
					      struct btrfs_key *location)
1176 1177 1178 1179
{
	struct btrfs_root *root;
	int ret;

1180 1181 1182 1183
	if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
		return fs_info->tree_root;
	if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
		return fs_info->extent_root;
1184 1185 1186 1187
	if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
		return fs_info->chunk_root;
	if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
		return fs_info->dev_root;
1188 1189
	if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
		return fs_info->csum_root;
1190

1191 1192 1193 1194 1195
	root = radix_tree_lookup(&fs_info->fs_roots_radix,
				 (unsigned long)location->objectid);
	if (root)
		return root;

1196
	root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
1197 1198
	if (IS_ERR(root))
		return root;
1199 1200 1201

	set_anon_super(&root->anon_super, NULL);

C
Chris Mason 已提交
1202 1203
	ret = radix_tree_insert(&fs_info->fs_roots_radix,
				(unsigned long)root->root_key.objectid,
1204 1205
				root);
	if (ret) {
1206
		free_extent_buffer(root->node);
1207 1208 1209
		kfree(root);
		return ERR_PTR(ret);
	}
Y
Yan Zheng 已提交
1210 1211
	if (!(fs_info->sb->s_flags & MS_RDONLY)) {
		ret = btrfs_find_dead_roots(fs_info->tree_root,
1212
					    root->root_key.objectid);
Y
Yan Zheng 已提交
1213 1214 1215
		BUG_ON(ret);
		btrfs_orphan_cleanup(root);
	}
1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228
	return root;
}

struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
				      struct btrfs_key *location,
				      const char *name, int namelen)
{
	struct btrfs_root *root;
	int ret;

	root = btrfs_read_fs_root_no_name(fs_info, location);
	if (!root)
		return NULL;
1229

1230 1231 1232
	if (root->in_sysfs)
		return root;

1233 1234
	ret = btrfs_set_root_name(root, name, namelen);
	if (ret) {
1235
		free_extent_buffer(root->node);
1236 1237 1238
		kfree(root);
		return ERR_PTR(ret);
	}
1239
#if 0
1240 1241
	ret = btrfs_sysfs_add_root(root);
	if (ret) {
1242
		free_extent_buffer(root->node);
1243 1244 1245 1246
		kfree(root->name);
		kfree(root);
		return ERR_PTR(ret);
	}
1247
#endif
1248
	root->in_sysfs = 1;
1249 1250
	return root;
}
C
Chris Mason 已提交
1251 1252 1253 1254 1255 1256 1257

static int btrfs_congested_fn(void *congested_data, int bdi_bits)
{
	struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
	int ret = 0;
	struct btrfs_device *device;
	struct backing_dev_info *bdi;
C
Chris Mason 已提交
1258

Q
Qinghuang Feng 已提交
1259
	list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
1260 1261
		if (!device->bdev)
			continue;
C
Chris Mason 已提交
1262 1263 1264 1265 1266 1267 1268 1269 1270
		bdi = blk_get_backing_dev_info(device->bdev);
		if (bdi && bdi_congested(bdi, bdi_bits)) {
			ret = 1;
			break;
		}
	}
	return ret;
}

1271 1272 1273 1274 1275 1276 1277 1278 1279 1280
/*
 * this unplugs every device on the box, and it is only used when page
 * is null
 */
static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
{
	struct btrfs_device *device;
	struct btrfs_fs_info *info;

	info = (struct btrfs_fs_info *)bdi->unplug_io_data;
Q
Qinghuang Feng 已提交
1281
	list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
1282 1283 1284
		if (!device->bdev)
			continue;

1285
		bdi = blk_get_backing_dev_info(device->bdev);
C
Chris Mason 已提交
1286
		if (bdi->unplug_io_fn)
1287 1288 1289 1290
			bdi->unplug_io_fn(bdi, page);
	}
}

1291
static void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
C
Chris Mason 已提交
1292
{
1293
	struct inode *inode;
1294 1295
	struct extent_map_tree *em_tree;
	struct extent_map *em;
1296
	struct address_space *mapping;
1297 1298
	u64 offset;

1299
	/* the generic O_DIRECT read code does this */
1300
	if (1 || !page) {
1301 1302 1303 1304
		__unplug_io_fn(bdi, page);
		return;
	}

1305 1306 1307 1308 1309 1310 1311 1312 1313 1314
	/*
	 * page->mapping may change at any time.  Get a consistent copy
	 * and use that for everything below
	 */
	smp_mb();
	mapping = page->mapping;
	if (!mapping)
		return;

	inode = mapping->host;
1315 1316 1317 1318 1319 1320 1321 1322 1323 1324

	/*
	 * don't do the expensive searching for a small number of
	 * devices
	 */
	if (BTRFS_I(inode)->root->fs_info->fs_devices->open_devices <= 2) {
		__unplug_io_fn(bdi, page);
		return;
	}

1325
	offset = page_offset(page);
C
Chris Mason 已提交
1326

1327 1328 1329 1330
	em_tree = &BTRFS_I(inode)->extent_tree;
	spin_lock(&em_tree->lock);
	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
	spin_unlock(&em_tree->lock);
1331 1332
	if (!em) {
		__unplug_io_fn(bdi, page);
1333
		return;
1334
	}
1335

1336 1337 1338 1339 1340
	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
		free_extent_map(em);
		__unplug_io_fn(bdi, page);
		return;
	}
1341 1342 1343 1344
	offset = offset - em->start;
	btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree,
			  em->block_start + offset, page);
	free_extent_map(em);
C
Chris Mason 已提交
1345 1346
}

1347 1348 1349 1350
/*
 * If this fails, caller must call bdi_destroy() to get rid of the
 * bdi again.
 */
C
Chris Mason 已提交
1351 1352
static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
{
1353 1354
	int err;

1355
	bdi->name = "btrfs";
1356 1357 1358 1359 1360 1361 1362 1363 1364 1365
	bdi->capabilities = BDI_CAP_MAP_COPY;
	err = bdi_init(bdi);
	if (err)
		return err;

	err = bdi_register(bdi, NULL, "btrfs-%d",
				atomic_inc_return(&btrfs_bdi_num));
	if (err)
		return err;

1366
	bdi->ra_pages	= default_backing_dev_info.ra_pages;
C
Chris Mason 已提交
1367 1368 1369 1370 1371 1372 1373
	bdi->unplug_io_fn	= btrfs_unplug_io_fn;
	bdi->unplug_io_data	= info;
	bdi->congested_fn	= btrfs_congested_fn;
	bdi->congested_data	= info;
	return 0;
}

1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410
static int bio_ready_for_csum(struct bio *bio)
{
	u64 length = 0;
	u64 buf_len = 0;
	u64 start = 0;
	struct page *page;
	struct extent_io_tree *io_tree = NULL;
	struct btrfs_fs_info *info = NULL;
	struct bio_vec *bvec;
	int i;
	int ret;

	bio_for_each_segment(bvec, bio, i) {
		page = bvec->bv_page;
		if (page->private == EXTENT_PAGE_PRIVATE) {
			length += bvec->bv_len;
			continue;
		}
		if (!page->private) {
			length += bvec->bv_len;
			continue;
		}
		length = bvec->bv_len;
		buf_len = page->private >> 2;
		start = page_offset(page) + bvec->bv_offset;
		io_tree = &BTRFS_I(page->mapping->host)->io_tree;
		info = BTRFS_I(page->mapping->host)->root->fs_info;
	}
	/* are we fully contained in this bio? */
	if (buf_len <= length)
		return 1;

	ret = extent_range_uptodate(io_tree, start + length,
				    start + buf_len - 1);
	return ret;
}

1411 1412 1413 1414 1415
/*
 * called by the kthread helper functions to finally call the bio end_io
 * functions.  This is where read checksum verification actually happens
 */
static void end_workqueue_fn(struct btrfs_work *work)
1416 1417
{
	struct bio *bio;
1418 1419
	struct end_io_wq *end_io_wq;
	struct btrfs_fs_info *fs_info;
1420 1421
	int error;

1422 1423 1424
	end_io_wq = container_of(work, struct end_io_wq, work);
	bio = end_io_wq->bio;
	fs_info = end_io_wq->info;
1425

1426
	/* metadata bio reads are special because the whole tree block must
1427 1428 1429 1430
	 * be checksummed at once.  This makes sure the entire block is in
	 * ram and up to date before trying to verify things.  For
	 * blocksize <= pagesize, it is basically a noop
	 */
1431 1432
	if (!(bio->bi_rw & (1 << BIO_RW)) && end_io_wq->metadata &&
	    !bio_ready_for_csum(bio)) {
1433
		btrfs_queue_worker(&fs_info->endio_meta_workers,
1434 1435 1436 1437 1438 1439 1440 1441
				   &end_io_wq->work);
		return;
	}
	error = end_io_wq->error;
	bio->bi_private = end_io_wq->private;
	bio->bi_end_io = end_io_wq->end_io;
	kfree(end_io_wq);
	bio_endio(bio, error);
1442 1443
}

1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495
static int cleaner_kthread(void *arg)
{
	struct btrfs_root *root = arg;

	do {
		smp_mb();
		if (root->fs_info->closing)
			break;

		vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
		mutex_lock(&root->fs_info->cleaner_mutex);
		btrfs_clean_old_snapshots(root);
		mutex_unlock(&root->fs_info->cleaner_mutex);

		if (freezing(current)) {
			refrigerator();
		} else {
			smp_mb();
			if (root->fs_info->closing)
				break;
			set_current_state(TASK_INTERRUPTIBLE);
			schedule();
			__set_current_state(TASK_RUNNING);
		}
	} while (!kthread_should_stop());
	return 0;
}

static int transaction_kthread(void *arg)
{
	struct btrfs_root *root = arg;
	struct btrfs_trans_handle *trans;
	struct btrfs_transaction *cur;
	unsigned long now;
	unsigned long delay;
	int ret;

	do {
		smp_mb();
		if (root->fs_info->closing)
			break;

		delay = HZ * 30;
		vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
		mutex_lock(&root->fs_info->transaction_kthread_mutex);

		mutex_lock(&root->fs_info->trans_mutex);
		cur = root->fs_info->running_transaction;
		if (!cur) {
			mutex_unlock(&root->fs_info->trans_mutex);
			goto sleep;
		}
Y
Yan Zheng 已提交
1496

1497 1498 1499 1500 1501 1502 1503 1504 1505
		now = get_seconds();
		if (now < cur->start_time || now - cur->start_time < 30) {
			mutex_unlock(&root->fs_info->trans_mutex);
			delay = HZ * 5;
			goto sleep;
		}
		mutex_unlock(&root->fs_info->trans_mutex);
		trans = btrfs_start_transaction(root, 1);
		ret = btrfs_commit_transaction(trans, root);
1506

1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523
sleep:
		wake_up_process(root->fs_info->cleaner_kthread);
		mutex_unlock(&root->fs_info->transaction_kthread_mutex);

		if (freezing(current)) {
			refrigerator();
		} else {
			if (root->fs_info->closing)
				break;
			set_current_state(TASK_INTERRUPTIBLE);
			schedule_timeout(delay);
			__set_current_state(TASK_RUNNING);
		}
	} while (!kthread_should_stop());
	return 0;
}

1524
struct btrfs_root *open_ctree(struct super_block *sb,
1525 1526
			      struct btrfs_fs_devices *fs_devices,
			      char *options)
1527
{
1528 1529 1530 1531
	u32 sectorsize;
	u32 nodesize;
	u32 leafsize;
	u32 blocksize;
1532
	u32 stripesize;
1533
	u64 generation;
1534
	u64 features;
1535
	struct btrfs_key location;
1536
	struct buffer_head *bh;
1537
	struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
C
Chris Mason 已提交
1538
						 GFP_NOFS);
1539 1540
	struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root),
						 GFP_NOFS);
1541
	struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root),
C
Chris Mason 已提交
1542
					       GFP_NOFS);
1543
	struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
C
Chris Mason 已提交
1544
						GFP_NOFS);
1545
	struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
1546
						GFP_NOFS);
1547
	struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
1548
					      GFP_NOFS);
1549 1550
	struct btrfs_root *log_tree_root;

1551
	int ret;
1552
	int err = -EINVAL;
1553

C
Chris Mason 已提交
1554
	struct btrfs_super_block *disk_super;
1555

1556
	if (!extent_root || !tree_root || !fs_info ||
1557
	    !chunk_root || !dev_root || !csum_root) {
C
Chris Mason 已提交
1558 1559 1560
		err = -ENOMEM;
		goto fail;
	}
1561
	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
C
Chris Mason 已提交
1562
	INIT_LIST_HEAD(&fs_info->trans_list);
1563
	INIT_LIST_HEAD(&fs_info->dead_roots);
1564
	INIT_LIST_HEAD(&fs_info->hashers);
1565
	INIT_LIST_HEAD(&fs_info->delalloc_inodes);
1566
	INIT_LIST_HEAD(&fs_info->ordered_operations);
1567
	spin_lock_init(&fs_info->delalloc_lock);
1568
	spin_lock_init(&fs_info->new_trans_lock);
Y
Yan Zheng 已提交
1569
	spin_lock_init(&fs_info->ref_cache_lock);
1570

1571
	init_completion(&fs_info->kobj_unregister);
1572 1573
	fs_info->tree_root = tree_root;
	fs_info->extent_root = extent_root;
1574
	fs_info->csum_root = csum_root;
1575 1576
	fs_info->chunk_root = chunk_root;
	fs_info->dev_root = dev_root;
1577
	fs_info->fs_devices = fs_devices;
1578
	INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
1579
	INIT_LIST_HEAD(&fs_info->space_info);
1580
	btrfs_mapping_init(&fs_info->mapping_tree);
1581
	atomic_set(&fs_info->nr_async_submits, 0);
1582
	atomic_set(&fs_info->async_delalloc_pages, 0);
1583
	atomic_set(&fs_info->async_submit_draining, 0);
1584
	atomic_set(&fs_info->nr_async_bios, 0);
C
Chris Mason 已提交
1585
	fs_info->sb = sb;
1586
	fs_info->max_extent = (u64)-1;
1587
	fs_info->max_inline = 8192 * 1024;
1588 1589
	if (setup_bdi(fs_info, &fs_info->bdi))
		goto fail_bdi;
1590 1591
	fs_info->btree_inode = new_inode(sb);
	fs_info->btree_inode->i_ino = 1;
C
Chris Mason 已提交
1592
	fs_info->btree_inode->i_nlink = 1;
1593
	fs_info->metadata_ratio = 8;
C
Chris Mason 已提交
1594

1595 1596
	fs_info->thread_pool_size = min_t(unsigned long,
					  num_online_cpus() + 2, 8);
1597

1598 1599 1600
	INIT_LIST_HEAD(&fs_info->ordered_extents);
	spin_lock_init(&fs_info->ordered_extent_lock);

1601 1602 1603
	sb->s_blocksize = 4096;
	sb->s_blocksize_bits = blksize_bits(4096);

1604 1605 1606 1607 1608 1609
	/*
	 * we set the i_size on the btree inode to the max possible int.
	 * the real end of the address space is determined by all of
	 * the devices in the system
	 */
	fs_info->btree_inode->i_size = OFFSET_MAX;
1610
	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
C
Chris Mason 已提交
1611 1612
	fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;

1613
	RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
1614
	extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
1615 1616
			     fs_info->btree_inode->i_mapping,
			     GFP_NOFS);
1617 1618 1619 1620
	extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
			     GFP_NOFS);

	BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
1621

J
Josef Bacik 已提交
1622 1623 1624
	spin_lock_init(&fs_info->block_group_cache_lock);
	fs_info->block_group_cache_tree.rb_node = NULL;

1625
	extent_io_tree_init(&fs_info->pinned_extents,
1626
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
1627
	fs_info->do_barriers = 1;
1628

1629 1630 1631
	BTRFS_I(fs_info->btree_inode)->root = tree_root;
	memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
	       sizeof(struct btrfs_key));
C
Chris Mason 已提交
1632
	insert_inode_hash(fs_info->btree_inode);
C
Chris Mason 已提交
1633

C
Chris Mason 已提交
1634
	mutex_init(&fs_info->trans_mutex);
1635
	mutex_init(&fs_info->ordered_operations_mutex);
1636
	mutex_init(&fs_info->tree_log_mutex);
1637
	mutex_init(&fs_info->drop_mutex);
1638
	mutex_init(&fs_info->chunk_mutex);
1639 1640
	mutex_init(&fs_info->transaction_kthread_mutex);
	mutex_init(&fs_info->cleaner_mutex);
1641
	mutex_init(&fs_info->volume_mutex);
Z
Zheng Yan 已提交
1642
	mutex_init(&fs_info->tree_reloc_mutex);
1643
	init_rwsem(&fs_info->extent_commit_sem);
1644 1645 1646 1647

	btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
	btrfs_init_free_cluster(&fs_info->data_alloc_cluster);

1648
	init_waitqueue_head(&fs_info->transaction_throttle);
1649
	init_waitqueue_head(&fs_info->transaction_wait);
1650
	init_waitqueue_head(&fs_info->async_submit_wait);
1651

1652
	__setup_root(4096, 4096, 4096, 4096, tree_root,
C
Chris Mason 已提交
1653
		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
1654

1655

Y
Yan Zheng 已提交
1656
	bh = btrfs_read_dev_super(fs_devices->latest_bdev);
1657
	if (!bh)
C
Chris Mason 已提交
1658 1659
		goto fail_iput;

1660
	memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy));
1661 1662
	memcpy(&fs_info->super_for_commit, &fs_info->super_copy,
	       sizeof(fs_info->super_for_commit));
1663
	brelse(bh);
1664

1665
	memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE);
1666

1667
	disk_super = &fs_info->super_copy;
1668
	if (!btrfs_super_root(disk_super))
1669
		goto fail_iput;
1670

Y
Yan Zheng 已提交
1671 1672 1673
	ret = btrfs_parse_options(tree_root, options);
	if (ret) {
		err = ret;
1674
		goto fail_iput;
Y
Yan Zheng 已提交
1675
	}
1676

1677 1678 1679 1680 1681
	features = btrfs_super_incompat_flags(disk_super) &
		~BTRFS_FEATURE_INCOMPAT_SUPP;
	if (features) {
		printk(KERN_ERR "BTRFS: couldn't mount because of "
		       "unsupported optional features (%Lx).\n",
1682
		       (unsigned long long)features);
1683
		err = -EINVAL;
1684
		goto fail_iput;
1685 1686
	}

1687 1688 1689 1690 1691 1692
	features = btrfs_super_incompat_flags(disk_super);
	if (!(features & BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF)) {
		features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
		btrfs_set_super_incompat_flags(disk_super, features);
	}

1693 1694 1695 1696 1697
	features = btrfs_super_compat_ro_flags(disk_super) &
		~BTRFS_FEATURE_COMPAT_RO_SUPP;
	if (!(sb->s_flags & MS_RDONLY) && features) {
		printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
		       "unsupported option features (%Lx).\n",
1698
		       (unsigned long long)features);
1699
		err = -EINVAL;
1700
		goto fail_iput;
1701 1702
	}

1703 1704 1705 1706 1707
	/*
	 * we need to start all the end_io workers up front because the
	 * queue work function gets called at interrupt time, and so it
	 * cannot dynamically grow.
	 */
1708 1709
	btrfs_init_workers(&fs_info->workers, "worker",
			   fs_info->thread_pool_size);
C
Chris Mason 已提交
1710

1711 1712 1713
	btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
			   fs_info->thread_pool_size);

1714
	btrfs_init_workers(&fs_info->submit_workers, "submit",
1715 1716
			   min_t(u64, fs_devices->num_devices,
			   fs_info->thread_pool_size));
1717 1718 1719 1720 1721 1722

	/* a higher idle thresh on the submit workers makes it much more
	 * likely that bios will be send down in a sane order to the
	 * devices
	 */
	fs_info->submit_workers.idle_thresh = 64;
1723

1724
	fs_info->workers.idle_thresh = 16;
C
Chris Mason 已提交
1725
	fs_info->workers.ordered = 1;
1726

1727 1728 1729
	fs_info->delalloc_workers.idle_thresh = 2;
	fs_info->delalloc_workers.ordered = 1;

1730 1731 1732
	btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1);
	btrfs_init_workers(&fs_info->endio_workers, "endio",
			   fs_info->thread_pool_size);
1733 1734
	btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
			   fs_info->thread_pool_size);
1735 1736
	btrfs_init_workers(&fs_info->endio_meta_write_workers,
			   "endio-meta-write", fs_info->thread_pool_size);
1737
	btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
1738
			   fs_info->thread_pool_size);
1739 1740 1741 1742 1743 1744

	/*
	 * endios are largely parallel and should have a very
	 * low idle thresh
	 */
	fs_info->endio_workers.idle_thresh = 4;
1745 1746
	fs_info->endio_meta_workers.idle_thresh = 4;

1747
	fs_info->endio_write_workers.idle_thresh = 64;
1748
	fs_info->endio_meta_write_workers.idle_thresh = 64;
1749

1750
	btrfs_start_workers(&fs_info->workers, 1);
1751
	btrfs_start_workers(&fs_info->submit_workers, 1);
1752
	btrfs_start_workers(&fs_info->delalloc_workers, 1);
1753
	btrfs_start_workers(&fs_info->fixup_workers, 1);
1754
	btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size);
1755 1756
	btrfs_start_workers(&fs_info->endio_meta_workers,
			    fs_info->thread_pool_size);
1757 1758
	btrfs_start_workers(&fs_info->endio_meta_write_workers,
			    fs_info->thread_pool_size);
1759 1760
	btrfs_start_workers(&fs_info->endio_write_workers,
			    fs_info->thread_pool_size);
1761

1762
	fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
C
Chris Mason 已提交
1763 1764
	fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
				    4 * 1024 * 1024 / PAGE_CACHE_SIZE);
1765

1766 1767 1768
	nodesize = btrfs_super_nodesize(disk_super);
	leafsize = btrfs_super_leafsize(disk_super);
	sectorsize = btrfs_super_sectorsize(disk_super);
1769
	stripesize = btrfs_super_stripesize(disk_super);
1770 1771 1772
	tree_root->nodesize = nodesize;
	tree_root->leafsize = leafsize;
	tree_root->sectorsize = sectorsize;
1773
	tree_root->stripesize = stripesize;
1774 1775 1776

	sb->s_blocksize = sectorsize;
	sb->s_blocksize_bits = blksize_bits(sectorsize);
1777

C
Chris Mason 已提交
1778 1779
	if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
		    sizeof(disk_super->magic))) {
C
Chris Mason 已提交
1780
		printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
C
Chris Mason 已提交
1781 1782
		goto fail_sb_buffer;
	}
1783

1784
	mutex_lock(&fs_info->chunk_mutex);
Y
Yan Zheng 已提交
1785
	ret = btrfs_read_sys_array(tree_root);
1786
	mutex_unlock(&fs_info->chunk_mutex);
1787
	if (ret) {
C
Chris Mason 已提交
1788 1789
		printk(KERN_WARNING "btrfs: failed to read the system "
		       "array on %s\n", sb->s_id);
1790
		goto fail_sb_buffer;
1791
	}
1792 1793 1794

	blocksize = btrfs_level_size(tree_root,
				     btrfs_super_chunk_root_level(disk_super));
1795
	generation = btrfs_super_chunk_root_generation(disk_super);
1796 1797 1798 1799 1800 1801

	__setup_root(nodesize, leafsize, sectorsize, stripesize,
		     chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);

	chunk_root->node = read_tree_block(chunk_root,
					   btrfs_super_chunk_root(disk_super),
1802
					   blocksize, generation);
1803
	BUG_ON(!chunk_root->node);
1804 1805 1806 1807 1808
	if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
		printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
		       sb->s_id);
		goto fail_chunk_root;
	}
1809 1810
	btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
	chunk_root->commit_root = btrfs_root_node(chunk_root);
1811

1812
	read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
C
Chris Mason 已提交
1813 1814
	   (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
	   BTRFS_UUID_SIZE);
1815

1816
	mutex_lock(&fs_info->chunk_mutex);
1817
	ret = btrfs_read_chunk_tree(chunk_root);
1818
	mutex_unlock(&fs_info->chunk_mutex);
Y
Yan Zheng 已提交
1819
	if (ret) {
C
Chris Mason 已提交
1820 1821
		printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
		       sb->s_id);
Y
Yan Zheng 已提交
1822 1823
		goto fail_chunk_root;
	}
1824

1825 1826
	btrfs_close_extra_devices(fs_devices);

1827 1828
	blocksize = btrfs_level_size(tree_root,
				     btrfs_super_root_level(disk_super));
1829
	generation = btrfs_super_generation(disk_super);
1830

C
Chris Mason 已提交
1831
	tree_root->node = read_tree_block(tree_root,
1832
					  btrfs_super_root(disk_super),
1833
					  blocksize, generation);
C
Chris Mason 已提交
1834
	if (!tree_root->node)
Y
Yan Zheng 已提交
1835
		goto fail_chunk_root;
1836 1837 1838 1839 1840
	if (!test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
		printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
		       sb->s_id);
		goto fail_tree_root;
	}
1841 1842
	btrfs_set_root_node(&tree_root->root_item, tree_root->node);
	tree_root->commit_root = btrfs_root_node(tree_root);
1843 1844

	ret = find_and_setup_root(tree_root, fs_info,
C
Chris Mason 已提交
1845
				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
1846
	if (ret)
C
Chris Mason 已提交
1847
		goto fail_tree_root;
1848 1849 1850 1851 1852 1853
	extent_root->track_dirty = 1;

	ret = find_and_setup_root(tree_root, fs_info,
				  BTRFS_DEV_TREE_OBJECTID, dev_root);
	if (ret)
		goto fail_extent_root;
1854
	dev_root->track_dirty = 1;
1855

1856 1857 1858
	ret = find_and_setup_root(tree_root, fs_info,
				  BTRFS_CSUM_TREE_OBJECTID, csum_root);
	if (ret)
1859
		goto fail_dev_root;
1860 1861 1862

	csum_root->track_dirty = 1;

C
Chris Mason 已提交
1863 1864
	btrfs_read_block_groups(extent_root);

1865
	fs_info->generation = generation;
Y
Yan Zheng 已提交
1866
	fs_info->last_trans_committed = generation;
C
Chris Mason 已提交
1867 1868 1869
	fs_info->data_alloc_profile = (u64)-1;
	fs_info->metadata_alloc_profile = (u64)-1;
	fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
1870 1871
	fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
					       "btrfs-cleaner");
1872
	if (IS_ERR(fs_info->cleaner_kthread))
1873
		goto fail_csum_root;
1874 1875 1876 1877

	fs_info->transaction_kthread = kthread_run(transaction_kthread,
						   tree_root,
						   "btrfs-transaction");
1878
	if (IS_ERR(fs_info->transaction_kthread))
1879
		goto fail_cleaner;
1880

C
Chris Mason 已提交
1881 1882 1883 1884 1885 1886 1887 1888
	if (!btrfs_test_opt(tree_root, SSD) &&
	    !btrfs_test_opt(tree_root, NOSSD) &&
	    !fs_info->fs_devices->rotating) {
		printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
		       "mode\n");
		btrfs_set_opt(fs_info->mount_opt, SSD);
	}

1889 1890 1891
	if (btrfs_super_log_root(disk_super) != 0) {
		u64 bytenr = btrfs_super_log_root(disk_super);

1892
		if (fs_devices->rw_devices == 0) {
C
Chris Mason 已提交
1893 1894
			printk(KERN_WARNING "Btrfs log replay required "
			       "on RO media\n");
1895 1896 1897
			err = -EIO;
			goto fail_trans_kthread;
		}
1898 1899 1900
		blocksize =
		     btrfs_level_size(tree_root,
				      btrfs_super_log_root_level(disk_super));
C
Chris Mason 已提交
1901

1902 1903 1904 1905 1906 1907 1908
		log_tree_root = kzalloc(sizeof(struct btrfs_root),
						      GFP_NOFS);

		__setup_root(nodesize, leafsize, sectorsize, stripesize,
			     log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);

		log_tree_root->node = read_tree_block(tree_root, bytenr,
1909 1910
						      blocksize,
						      generation + 1);
1911 1912
		ret = btrfs_recover_log_trees(log_tree_root);
		BUG_ON(ret);
1913 1914 1915 1916 1917

		if (sb->s_flags & MS_RDONLY) {
			ret =  btrfs_commit_super(tree_root);
			BUG_ON(ret);
		}
1918
	}
Z
Zheng Yan 已提交
1919

1920
	if (!(sb->s_flags & MS_RDONLY)) {
1921
		ret = btrfs_recover_relocation(tree_root);
1922 1923
		BUG_ON(ret);
	}
Z
Zheng Yan 已提交
1924

1925 1926 1927 1928 1929 1930
	location.objectid = BTRFS_FS_TREE_OBJECTID;
	location.type = BTRFS_ROOT_ITEM_KEY;
	location.offset = (u64)-1;

	fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
	if (!fs_info->fs_root)
1931
		goto fail_trans_kthread;
C
Chris Mason 已提交
1932

1933
	return tree_root;
C
Chris Mason 已提交
1934

1935 1936
fail_trans_kthread:
	kthread_stop(fs_info->transaction_kthread);
1937
fail_cleaner:
1938
	kthread_stop(fs_info->cleaner_kthread);
1939 1940 1941 1942 1943 1944 1945 1946

	/*
	 * make sure we're done with the btree inode before we stop our
	 * kthreads
	 */
	filemap_write_and_wait(fs_info->btree_inode->i_mapping);
	invalidate_inode_pages2(fs_info->btree_inode->i_mapping);

1947 1948
fail_csum_root:
	free_extent_buffer(csum_root->node);
1949 1950 1951 1952
	free_extent_buffer(csum_root->commit_root);
fail_dev_root:
	free_extent_buffer(dev_root->node);
	free_extent_buffer(dev_root->commit_root);
1953 1954
fail_extent_root:
	free_extent_buffer(extent_root->node);
1955
	free_extent_buffer(extent_root->commit_root);
C
Chris Mason 已提交
1956
fail_tree_root:
1957
	free_extent_buffer(tree_root->node);
1958
	free_extent_buffer(tree_root->commit_root);
Y
Yan Zheng 已提交
1959 1960
fail_chunk_root:
	free_extent_buffer(chunk_root->node);
1961
	free_extent_buffer(chunk_root->commit_root);
C
Chris Mason 已提交
1962
fail_sb_buffer:
1963
	btrfs_stop_workers(&fs_info->fixup_workers);
1964
	btrfs_stop_workers(&fs_info->delalloc_workers);
1965 1966
	btrfs_stop_workers(&fs_info->workers);
	btrfs_stop_workers(&fs_info->endio_workers);
1967
	btrfs_stop_workers(&fs_info->endio_meta_workers);
1968
	btrfs_stop_workers(&fs_info->endio_meta_write_workers);
1969
	btrfs_stop_workers(&fs_info->endio_write_workers);
1970
	btrfs_stop_workers(&fs_info->submit_workers);
1971
fail_iput:
1972
	invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
1973
	iput(fs_info->btree_inode);
1974

1975
	btrfs_close_devices(fs_info->fs_devices);
1976
	btrfs_mapping_tree_free(&fs_info->mapping_tree);
1977
fail_bdi:
1978 1979
	bdi_destroy(&fs_info->bdi);
fail:
C
Chris Mason 已提交
1980 1981 1982
	kfree(extent_root);
	kfree(tree_root);
	kfree(fs_info);
1983 1984
	kfree(chunk_root);
	kfree(dev_root);
1985
	kfree(csum_root);
C
Chris Mason 已提交
1986
	return ERR_PTR(err);
1987 1988
}

1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
{
	char b[BDEVNAME_SIZE];

	if (uptodate) {
		set_buffer_uptodate(bh);
	} else {
		if (!buffer_eopnotsupp(bh) && printk_ratelimit()) {
			printk(KERN_WARNING "lost page write due to "
					"I/O error on %s\n",
				       bdevname(bh->b_bdev, b));
		}
2001 2002 2003
		/* note, we dont' set_buffer_write_io_error because we have
		 * our own ways of dealing with the IO errors
		 */
2004 2005 2006 2007 2008 2009
		clear_buffer_uptodate(bh);
	}
	unlock_buffer(bh);
	put_bh(bh);
}

Y
Yan Zheng 已提交
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050
struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
{
	struct buffer_head *bh;
	struct buffer_head *latest = NULL;
	struct btrfs_super_block *super;
	int i;
	u64 transid = 0;
	u64 bytenr;

	/* we would like to check all the supers, but that would make
	 * a btrfs mount succeed after a mkfs from a different FS.
	 * So, we need to add a special mount option to scan for
	 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
	 */
	for (i = 0; i < 1; i++) {
		bytenr = btrfs_sb_offset(i);
		if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
			break;
		bh = __bread(bdev, bytenr / 4096, 4096);
		if (!bh)
			continue;

		super = (struct btrfs_super_block *)bh->b_data;
		if (btrfs_super_bytenr(super) != bytenr ||
		    strncmp((char *)(&super->magic), BTRFS_MAGIC,
			    sizeof(super->magic))) {
			brelse(bh);
			continue;
		}

		if (!latest || btrfs_super_generation(super) > transid) {
			brelse(latest);
			latest = bh;
			transid = btrfs_super_generation(super);
		} else {
			brelse(bh);
		}
	}
	return latest;
}

2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061
/*
 * this should be called twice, once with wait == 0 and
 * once with wait == 1.  When wait == 0 is done, all the buffer heads
 * we write are pinned.
 *
 * They are released when wait == 1 is done.
 * max_mirrors must be the same for both runs, and it indicates how
 * many supers on this one device should be written.
 *
 * max_mirrors == 0 means to write them all.
 */
Y
Yan Zheng 已提交
2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097
static int write_dev_supers(struct btrfs_device *device,
			    struct btrfs_super_block *sb,
			    int do_barriers, int wait, int max_mirrors)
{
	struct buffer_head *bh;
	int i;
	int ret;
	int errors = 0;
	u32 crc;
	u64 bytenr;
	int last_barrier = 0;

	if (max_mirrors == 0)
		max_mirrors = BTRFS_SUPER_MIRROR_MAX;

	/* make sure only the last submit_bh does a barrier */
	if (do_barriers) {
		for (i = 0; i < max_mirrors; i++) {
			bytenr = btrfs_sb_offset(i);
			if (bytenr + BTRFS_SUPER_INFO_SIZE >=
			    device->total_bytes)
				break;
			last_barrier = i;
		}
	}

	for (i = 0; i < max_mirrors; i++) {
		bytenr = btrfs_sb_offset(i);
		if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
			break;

		if (wait) {
			bh = __find_get_block(device->bdev, bytenr / 4096,
					      BTRFS_SUPER_INFO_SIZE);
			BUG_ON(!bh);
			wait_on_buffer(bh);
2098 2099 2100 2101 2102 2103 2104 2105 2106
			if (!buffer_uptodate(bh))
				errors++;

			/* drop our reference */
			brelse(bh);

			/* drop the reference from the wait == 0 run */
			brelse(bh);
			continue;
Y
Yan Zheng 已提交
2107 2108 2109 2110 2111 2112 2113 2114 2115 2116
		} else {
			btrfs_set_super_bytenr(sb, bytenr);

			crc = ~(u32)0;
			crc = btrfs_csum_data(NULL, (char *)sb +
					      BTRFS_CSUM_SIZE, crc,
					      BTRFS_SUPER_INFO_SIZE -
					      BTRFS_CSUM_SIZE);
			btrfs_csum_final(crc, sb->csum);

2117 2118 2119 2120
			/*
			 * one reference for us, and we leave it for the
			 * caller
			 */
Y
Yan Zheng 已提交
2121 2122 2123 2124
			bh = __getblk(device->bdev, bytenr / 4096,
				      BTRFS_SUPER_INFO_SIZE);
			memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);

2125
			/* one reference for submit_bh */
Y
Yan Zheng 已提交
2126
			get_bh(bh);
2127 2128

			set_buffer_uptodate(bh);
Y
Yan Zheng 已提交
2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139
			lock_buffer(bh);
			bh->b_end_io = btrfs_end_buffer_write_sync;
		}

		if (i == last_barrier && do_barriers && device->barriers) {
			ret = submit_bh(WRITE_BARRIER, bh);
			if (ret == -EOPNOTSUPP) {
				printk("btrfs: disabling barriers on dev %s\n",
				       device->name);
				set_buffer_uptodate(bh);
				device->barriers = 0;
2140
				/* one reference for submit_bh */
Y
Yan Zheng 已提交
2141 2142
				get_bh(bh);
				lock_buffer(bh);
2143
				ret = submit_bh(WRITE_SYNC, bh);
Y
Yan Zheng 已提交
2144 2145
			}
		} else {
2146
			ret = submit_bh(WRITE_SYNC, bh);
Y
Yan Zheng 已提交
2147 2148
		}

2149
		if (ret)
Y
Yan Zheng 已提交
2150 2151 2152 2153 2154 2155
			errors++;
	}
	return errors < i ? 0 : -1;
}

int write_all_supers(struct btrfs_root *root, int max_mirrors)
2156
{
2157
	struct list_head *head;
2158
	struct btrfs_device *dev;
2159
	struct btrfs_super_block *sb;
2160 2161 2162
	struct btrfs_dev_item *dev_item;
	int ret;
	int do_barriers;
2163 2164
	int max_errors;
	int total_errors = 0;
2165
	u64 flags;
2166

2167
	max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
2168 2169
	do_barriers = !btrfs_test_opt(root, NOBARRIER);

2170 2171
	sb = &root->fs_info->super_for_commit;
	dev_item = &sb->dev_item;
2172 2173 2174

	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
	head = &root->fs_info->fs_devices->devices;
Q
Qinghuang Feng 已提交
2175
	list_for_each_entry(dev, head, dev_list) {
2176 2177 2178 2179
		if (!dev->bdev) {
			total_errors++;
			continue;
		}
Y
Yan Zheng 已提交
2180
		if (!dev->in_fs_metadata || !dev->writeable)
2181 2182
			continue;

Y
Yan Zheng 已提交
2183
		btrfs_set_stack_device_generation(dev_item, 0);
2184 2185 2186 2187 2188 2189 2190 2191
		btrfs_set_stack_device_type(dev_item, dev->type);
		btrfs_set_stack_device_id(dev_item, dev->devid);
		btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
		btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
		btrfs_set_stack_device_io_align(dev_item, dev->io_align);
		btrfs_set_stack_device_io_width(dev_item, dev->io_width);
		btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
		memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Y
Yan Zheng 已提交
2192
		memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
Y
Yan Zheng 已提交
2193

2194 2195 2196
		flags = btrfs_super_flags(sb);
		btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);

Y
Yan Zheng 已提交
2197
		ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
2198 2199
		if (ret)
			total_errors++;
2200
	}
2201
	if (total_errors > max_errors) {
C
Chris Mason 已提交
2202 2203
		printk(KERN_ERR "btrfs: %d errors while writing supers\n",
		       total_errors);
2204 2205
		BUG();
	}
2206

Y
Yan Zheng 已提交
2207
	total_errors = 0;
Q
Qinghuang Feng 已提交
2208
	list_for_each_entry(dev, head, dev_list) {
2209 2210
		if (!dev->bdev)
			continue;
Y
Yan Zheng 已提交
2211
		if (!dev->in_fs_metadata || !dev->writeable)
2212 2213
			continue;

Y
Yan Zheng 已提交
2214 2215 2216
		ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
		if (ret)
			total_errors++;
2217
	}
2218
	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
2219
	if (total_errors > max_errors) {
C
Chris Mason 已提交
2220 2221
		printk(KERN_ERR "btrfs: %d errors while writing supers\n",
		       total_errors);
2222 2223
		BUG();
	}
2224 2225 2226
	return 0;
}

Y
Yan Zheng 已提交
2227 2228
int write_ctree_super(struct btrfs_trans_handle *trans,
		      struct btrfs_root *root, int max_mirrors)
2229
{
2230
	int ret;
2231

Y
Yan Zheng 已提交
2232
	ret = write_all_supers(root, max_mirrors);
2233
	return ret;
C
Chris Mason 已提交
2234 2235
}

2236
int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
C
Chris Mason 已提交
2237
{
2238
	WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
C
Chris Mason 已提交
2239 2240
	radix_tree_delete(&fs_info->fs_roots_radix,
			  (unsigned long)root->root_key.objectid);
2241 2242 2243 2244
	if (root->anon_super.s_dev) {
		down_write(&root->anon_super.s_umount);
		kill_anon_super(&root->anon_super);
	}
C
Chris Mason 已提交
2245
	if (root->node)
2246
		free_extent_buffer(root->node);
C
Chris Mason 已提交
2247
	if (root->commit_root)
2248
		free_extent_buffer(root->commit_root);
C
Chris Mason 已提交
2249
	kfree(root->name);
C
Chris Mason 已提交
2250 2251 2252 2253
	kfree(root);
	return 0;
}

C
Chris Mason 已提交
2254
static int del_fs_roots(struct btrfs_fs_info *fs_info)
2255 2256 2257 2258 2259
{
	int ret;
	struct btrfs_root *gang[8];
	int i;

C
Chris Mason 已提交
2260
	while (1) {
2261 2262 2263 2264 2265
		ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
					     (void **)gang, 0,
					     ARRAY_SIZE(gang));
		if (!ret)
			break;
C
Chris Mason 已提交
2266
		for (i = 0; i < ret; i++)
2267
			btrfs_free_fs_root(fs_info, gang[i]);
2268 2269 2270
	}
	return 0;
}
2271

Y
Yan Zheng 已提交
2272
int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
C
Chris Mason 已提交
2273
{
Y
Yan Zheng 已提交
2274 2275 2276
	u64 root_objectid = 0;
	struct btrfs_root *gang[8];
	int i;
2277
	int ret;
2278

Y
Yan Zheng 已提交
2279 2280 2281 2282 2283 2284
	while (1) {
		ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
					     (void **)gang, root_objectid,
					     ARRAY_SIZE(gang));
		if (!ret)
			break;
2285 2286

		root_objectid = gang[ret - 1]->root_key.objectid + 1;
Y
Yan Zheng 已提交
2287 2288 2289
		for (i = 0; i < ret; i++) {
			root_objectid = gang[i]->root_key.objectid;
			ret = btrfs_find_dead_roots(fs_info->tree_root,
2290
						    root_objectid);
Y
Yan Zheng 已提交
2291 2292 2293 2294 2295 2296 2297
			BUG_ON(ret);
			btrfs_orphan_cleanup(gang[i]);
		}
		root_objectid++;
	}
	return 0;
}
2298

Y
Yan Zheng 已提交
2299 2300 2301 2302
int btrfs_commit_super(struct btrfs_root *root)
{
	struct btrfs_trans_handle *trans;
	int ret;
2303

Y
Yan Zheng 已提交
2304
	mutex_lock(&root->fs_info->cleaner_mutex);
2305
	btrfs_clean_old_snapshots(root);
Y
Yan Zheng 已提交
2306
	mutex_unlock(&root->fs_info->cleaner_mutex);
C
Chris Mason 已提交
2307
	trans = btrfs_start_transaction(root, 1);
2308
	ret = btrfs_commit_transaction(trans, root);
Y
Yan Zheng 已提交
2309 2310
	BUG_ON(ret);
	/* run commit again to drop the original snapshot */
C
Chris Mason 已提交
2311 2312 2313
	trans = btrfs_start_transaction(root, 1);
	btrfs_commit_transaction(trans, root);
	ret = btrfs_write_and_wait_transaction(NULL, root);
2314
	BUG_ON(ret);
2315

Y
Yan Zheng 已提交
2316
	ret = write_ctree_super(NULL, root, 0);
Y
Yan Zheng 已提交
2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332
	return ret;
}

int close_ctree(struct btrfs_root *root)
{
	struct btrfs_fs_info *fs_info = root->fs_info;
	int ret;

	fs_info->closing = 1;
	smp_mb();

	kthread_stop(root->fs_info->transaction_kthread);
	kthread_stop(root->fs_info->cleaner_kthread);

	if (!(fs_info->sb->s_flags & MS_RDONLY)) {
		ret =  btrfs_commit_super(root);
C
Chris Mason 已提交
2333 2334
		if (ret)
			printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
Y
Yan Zheng 已提交
2335
	}
2336

2337 2338 2339
	fs_info->closing = 2;
	smp_mb();

C
Chris Mason 已提交
2340
	if (fs_info->delalloc_bytes) {
C
Chris Mason 已提交
2341
		printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
2342
		       (unsigned long long)fs_info->delalloc_bytes);
C
Chris Mason 已提交
2343
	}
Y
Yan Zheng 已提交
2344
	if (fs_info->total_ref_cache_size) {
C
Chris Mason 已提交
2345 2346
		printk(KERN_INFO "btrfs: at umount reference cache size %llu\n",
		       (unsigned long long)fs_info->total_ref_cache_size);
Y
Yan Zheng 已提交
2347
	}
2348

2349 2350 2351 2352 2353 2354 2355 2356 2357 2358
	free_extent_buffer(fs_info->extent_root->node);
	free_extent_buffer(fs_info->extent_root->commit_root);
	free_extent_buffer(fs_info->tree_root->node);
	free_extent_buffer(fs_info->tree_root->commit_root);
	free_extent_buffer(root->fs_info->chunk_root->node);
	free_extent_buffer(root->fs_info->chunk_root->commit_root);
	free_extent_buffer(root->fs_info->dev_root->node);
	free_extent_buffer(root->fs_info->dev_root->commit_root);
	free_extent_buffer(root->fs_info->csum_root->node);
	free_extent_buffer(root->fs_info->csum_root->commit_root);
2359

C
Chris Mason 已提交
2360
	btrfs_free_block_groups(root->fs_info);
J
Josef Bacik 已提交
2361
	btrfs_free_pinned_extents(root->fs_info);
2362

Y
Yan Zheng 已提交
2363
	del_fs_roots(fs_info);
2364

Y
Yan Zheng 已提交
2365
	iput(fs_info->btree_inode);
2366

2367
	btrfs_stop_workers(&fs_info->fixup_workers);
2368
	btrfs_stop_workers(&fs_info->delalloc_workers);
2369 2370
	btrfs_stop_workers(&fs_info->workers);
	btrfs_stop_workers(&fs_info->endio_workers);
2371
	btrfs_stop_workers(&fs_info->endio_meta_workers);
2372
	btrfs_stop_workers(&fs_info->endio_meta_write_workers);
2373
	btrfs_stop_workers(&fs_info->endio_write_workers);
2374
	btrfs_stop_workers(&fs_info->submit_workers);
2375

2376
	btrfs_close_devices(fs_info->fs_devices);
2377
	btrfs_mapping_tree_free(&fs_info->mapping_tree);
2378

C
Chris Mason 已提交
2379
	bdi_destroy(&fs_info->bdi);
2380

2381 2382
	kfree(fs_info->extent_root);
	kfree(fs_info->tree_root);
2383 2384
	kfree(fs_info->chunk_root);
	kfree(fs_info->dev_root);
2385
	kfree(fs_info->csum_root);
2386 2387 2388
	return 0;
}

2389
int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
2390
{
2391
	int ret;
2392
	struct inode *btree_inode = buf->first_page->mapping->host;
2393 2394 2395 2396 2397 2398 2399 2400

	ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf);
	if (!ret)
		return ret;

	ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
				    parent_transid);
	return !ret;
2401 2402 2403
}

int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
C
Chris Mason 已提交
2404
{
2405
	struct inode *btree_inode = buf->first_page->mapping->host;
2406
	return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
2407 2408
					  buf);
}
2409

2410 2411
void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
{
2412
	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
2413 2414
	u64 transid = btrfs_header_generation(buf);
	struct inode *btree_inode = root->fs_info->btree_inode;
2415
	int was_dirty;
2416

2417
	btrfs_assert_tree_locked(buf);
C
Chris Mason 已提交
2418
	if (transid != root->fs_info->generation) {
C
Chris Mason 已提交
2419 2420
		printk(KERN_CRIT "btrfs transid mismatch buffer %llu, "
		       "found %llu running %llu\n",
2421
			(unsigned long long)buf->start,
C
Chris Mason 已提交
2422 2423
			(unsigned long long)transid,
			(unsigned long long)root->fs_info->generation);
C
Chris Mason 已提交
2424 2425
		WARN_ON(1);
	}
2426 2427 2428 2429 2430 2431 2432
	was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
					    buf);
	if (!was_dirty) {
		spin_lock(&root->fs_info->delalloc_lock);
		root->fs_info->dirty_metadata_bytes += buf->len;
		spin_unlock(&root->fs_info->delalloc_lock);
	}
2433 2434
}

2435
void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
C
Chris Mason 已提交
2436
{
2437 2438 2439 2440
	/*
	 * looks as though older kernels can get into trouble with
	 * this code, they end up stuck in balance_dirty_pages forever
	 */
2441
	u64 num_dirty;
2442
	unsigned long thresh = 32 * 1024 * 1024;
2443

2444
	if (current->flags & PF_MEMALLOC)
2445 2446
		return;

2447 2448
	num_dirty = root->fs_info->dirty_metadata_bytes;

2449 2450
	if (num_dirty > thresh) {
		balance_dirty_pages_ratelimited_nr(
C
Chris Mason 已提交
2451
				   root->fs_info->btree_inode->i_mapping, 1);
2452
	}
2453
	return;
C
Chris Mason 已提交
2454
}
2455

2456
int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
2457
{
2458
	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
2459
	int ret;
2460
	ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
C
Chris Mason 已提交
2461
	if (ret == 0)
2462
		set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
2463
	return ret;
2464
}
2465

C
Chris Mason 已提交
2466 2467 2468
int btree_lock_page_hook(struct page *page)
{
	struct inode *inode = page->mapping->host;
2469
	struct btrfs_root *root = BTRFS_I(inode)->root;
C
Chris Mason 已提交
2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct extent_buffer *eb;
	unsigned long len;
	u64 bytenr = page_offset(page);

	if (page->private == EXTENT_PAGE_PRIVATE)
		goto out;

	len = page->private >> 2;
	eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS);
	if (!eb)
		goto out;

	btrfs_tree_lock(eb);
	btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
2485 2486 2487 2488 2489 2490 2491 2492 2493 2494

	if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
		spin_lock(&root->fs_info->delalloc_lock);
		if (root->fs_info->dirty_metadata_bytes >= eb->len)
			root->fs_info->dirty_metadata_bytes -= eb->len;
		else
			WARN_ON(1);
		spin_unlock(&root->fs_info->delalloc_lock);
	}

C
Chris Mason 已提交
2495 2496 2497 2498 2499 2500 2501
	btrfs_tree_unlock(eb);
	free_extent_buffer(eb);
out:
	lock_page(page);
	return 0;
}

2502
static struct extent_io_ops btree_extent_io_ops = {
C
Chris Mason 已提交
2503
	.write_cache_pages_lock_hook = btree_lock_page_hook,
2504
	.readpage_end_io_hook = btree_readpage_end_io_hook,
2505
	.submit_bio_hook = btree_submit_bio_hook,
2506 2507
	/* note we're sharing with inode.c for the merge bio hook */
	.merge_bio_hook = btrfs_merge_bio_hook,
2508
};