disk-io.c 53.2 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.
 */

19
#include <linux/version.h>
C
Chris Mason 已提交
20
#include <linux/fs.h>
21
#include <linux/blkdev.h>
C
Chris Mason 已提交
22
#include <linux/scatterlist.h>
C
Chris Mason 已提交
23
#include <linux/swap.h>
24
#include <linux/radix-tree.h>
C
Chris Mason 已提交
25
#include <linux/writeback.h>
26
#include <linux/buffer_head.h> // for block_sync_page
27
#include <linux/workqueue.h>
28 29 30 31 32 33
#include <linux/kthread.h>
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20)
# include <linux/freezer.h>
#else
# include <linux/sched.h>
#endif
34
#include "crc32c.h"
35 36
#include "ctree.h"
#include "disk-io.h"
37
#include "transaction.h"
38
#include "btrfs_inode.h"
39
#include "volumes.h"
40
#include "print-tree.h"
41
#include "async-thread.h"
42
#include "locking.h"
C
Chris Mason 已提交
43
#include "ref-cache.h"
44
#include "tree-log.h"
45

46 47
#if 0
static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
48
{
49 50 51 52
	if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) {
		printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n",
		       (unsigned long long)extent_buffer_blocknr(buf),
		       (unsigned long long)btrfs_header_blocknr(buf));
C
Chris Mason 已提交
53
		return 1;
54
	}
55
	return 0;
56
}
57
#endif
58

59
static struct extent_io_ops btree_extent_io_ops;
60
static void end_workqueue_fn(struct btrfs_work *work);
61 62 63 64 65 66 67

struct end_io_wq {
	struct bio *bio;
	bio_end_io_t *end_io;
	void *private;
	struct btrfs_fs_info *info;
	int error;
68
	int metadata;
69
	struct list_head list;
70
	struct btrfs_work work;
71
};
72

73 74 75 76 77 78 79
struct async_submit_bio {
	struct inode *inode;
	struct bio *bio;
	struct list_head list;
	extent_submit_bio_hook_t *submit_bio_hook;
	int rw;
	int mirror_num;
80
	struct btrfs_work work;
81 82
};

83
struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
84
				    size_t page_offset, u64 start, u64 len,
85
				    int create)
86
{
87 88 89 90
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct extent_map *em;
	int ret;

91 92
	spin_lock(&em_tree->lock);
	em = lookup_extent_mapping(em_tree, start, len);
93 94 95 96
	if (em) {
		em->bdev =
			BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
		spin_unlock(&em_tree->lock);
97
		goto out;
98 99
	}
	spin_unlock(&em_tree->lock);
100

101 102 103 104 105 106
	em = alloc_extent_map(GFP_NOFS);
	if (!em) {
		em = ERR_PTR(-ENOMEM);
		goto out;
	}
	em->start = 0;
107
	em->len = (u64)-1;
108
	em->block_start = 0;
109
	em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
110 111

	spin_lock(&em_tree->lock);
112 113
	ret = add_extent_mapping(em_tree, em);
	if (ret == -EEXIST) {
114 115 116 117 118
		u64 failed_start = em->start;
		u64 failed_len = em->len;

		printk("failed to insert %Lu %Lu -> %Lu into tree\n",
		       em->start, em->len, em->block_start);
119
		free_extent_map(em);
120
		em = lookup_extent_mapping(em_tree, start, len);
121 122 123
		if (em) {
			printk("after failing, found %Lu %Lu %Lu\n",
			       em->start, em->len, em->block_start);
124
			ret = 0;
125 126 127 128 129 130 131 132 133
		} else {
			em = lookup_extent_mapping(em_tree, failed_start,
						   failed_len);
			if (em) {
				printk("double failure lookup gives us "
				       "%Lu %Lu -> %Lu\n", em->start,
				       em->len, em->block_start);
				free_extent_map(em);
			}
134
			ret = -EIO;
135
		}
136
	} else if (ret) {
137 138
		free_extent_map(em);
		em = NULL;
139
	}
140 141 142 143
	spin_unlock(&em_tree->lock);

	if (ret)
		em = ERR_PTR(ret);
144 145
out:
	return em;
146 147
}

148 149
u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
{
150
	return btrfs_crc32c(seed, data, len);
151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
}

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

static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
			   int verify)
{
	char result[BTRFS_CRC32_SIZE];
	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;

	len = buf->len - offset;
	while(len > 0) {
		err = map_private_extent_buffer(buf, offset, 32,
					&map_token, &kaddr,
					&map_start, &map_len, KM_USER0);
		if (err) {
			printk("failed to map extent buffer! %lu\n",
			       offset);
			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);
	}
	btrfs_csum_final(crc, result);

	if (verify) {
192
		/* FIXME, this is not good */
193
		if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) {
194 195 196 197 198 199
			u32 val;
			u32 found = 0;
			memcpy(&found, result, BTRFS_CRC32_SIZE);

			read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE);
			printk("btrfs: %s checksum verify failed on %llu "
C
Chris Mason 已提交
200
			       "wanted %X found %X level %d\n",
201
			       root->fs_info->sb->s_id,
C
Chris Mason 已提交
202
			       buf->start, val, found, btrfs_header_level(buf));
203 204 205 206 207 208 209 210
			return 1;
		}
	} else {
		write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE);
	}
	return 0;
}

211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230
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;
	}
	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));
	ret = 1;
	clear_extent_buffer_uptodate(io_tree, eb);
C
Chris Mason 已提交
231
out:
232 233 234 235 236 237
	unlock_extent(io_tree, eb->start, eb->start + eb->len - 1,
		      GFP_NOFS);
	return ret;

}

238 239
static int btree_read_extent_buffer_pages(struct btrfs_root *root,
					  struct extent_buffer *eb,
240
					  u64 start, u64 parent_transid)
241 242 243 244 245 246 247 248 249 250
{
	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);
251 252
		if (!ret &&
		    !verify_parent_transid(io_tree, eb, parent_transid))
253
			return ret;
254
printk("read extent buffer pages failed with ret %d mirror no %d\n", ret, mirror_num);
255 256
		num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
					      eb->start, eb->len);
C
Chris Mason 已提交
257
		if (num_copies == 1)
258
			return ret;
C
Chris Mason 已提交
259

260
		mirror_num++;
C
Chris Mason 已提交
261
		if (mirror_num > num_copies)
262 263 264 265
			return ret;
	}
	return -EIO;
}
266 267 268

int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
{
269
	struct extent_io_tree *tree;
270
	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
271 272 273 274
	u64 found_start;
	int found_level;
	unsigned long len;
	struct extent_buffer *eb;
275 276
	int ret;

277
	tree = &BTRFS_I(page->mapping->host)->io_tree;
278 279 280 281 282 283 284 285 286 287

	if (page->private == EXTENT_PAGE_PRIVATE)
		goto out;
	if (!page->private)
		goto out;
	len = page->private >> 2;
	if (len == 0) {
		WARN_ON(1);
	}
	eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
288 289
	ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
					     btrfs_header_generation(eb));
290
	BUG_ON(ret);
291 292 293 294
	found_start = btrfs_header_bytenr(eb);
	if (found_start != start) {
		printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
		       start, found_start, len);
295 296 297 298 299 300 301 302 303 304 305 306 307
		WARN_ON(1);
		goto err;
	}
	if (eb->first_page != page) {
		printk("bad first page %lu %lu\n", eb->first_page->index,
		       page->index);
		WARN_ON(1);
		goto err;
	}
	if (!PageUptodate(page)) {
		printk("csum not up to date page %lu\n", page->index);
		WARN_ON(1);
		goto err;
308 309
	}
	found_level = btrfs_header_level(eb);
C
Chris Mason 已提交
310

311
	csum_tree_block(root, eb, 0);
312
err:
313 314 315 316 317
	free_extent_buffer(eb);
out:
	return 0;
}

318
static int btree_writepage_io_hook(struct page *page, u64 start, u64 end)
319
{
320 321 322
	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;

	csum_dirty_buffer(root, page);
323 324 325
	return 0;
}

326 327 328 329 330 331 332 333 334
int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
			       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;
335
	int ret = 0;
336 337 338 339 340 341 342 343 344 345 346

	tree = &BTRFS_I(page->mapping->host)->io_tree;
	if (page->private == EXTENT_PAGE_PRIVATE)
		goto out;
	if (!page->private)
		goto out;
	len = page->private >> 2;
	if (len == 0) {
		WARN_ON(1);
	}
	eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
347

348
	found_start = btrfs_header_bytenr(eb);
349
	if (0 && found_start != start) {
350 351 352
		printk("bad tree block start %llu %llu\n",
		       (unsigned long long)found_start,
		       (unsigned long long)eb->start);
353
		ret = -EIO;
354 355 356 357 358 359
		goto err;
	}
	if (eb->first_page != page) {
		printk("bad first page %lu %lu\n", eb->first_page->index,
		       page->index);
		WARN_ON(1);
360
		ret = -EIO;
361 362
		goto err;
	}
363 364 365 366 367 368 369
	if (memcmp_extent_buffer(eb, root->fs_info->fsid,
				 (unsigned long)btrfs_header_fsid(eb),
				 BTRFS_FSID_SIZE)) {
		printk("bad fsid on block %Lu\n", eb->start);
		ret = -EIO;
		goto err;
	}
370 371 372
	found_level = btrfs_header_level(eb);

	ret = csum_tree_block(root, eb, 1);
373 374
	if (ret)
		ret = -EIO;
375 376 377 378 379 380

	end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
	end = eb->start + end - 1;
err:
	free_extent_buffer(eb);
out:
381
	return ret;
382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400
}

#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
static void end_workqueue_bio(struct bio *bio, int err)
#else
static int end_workqueue_bio(struct bio *bio,
				   unsigned int bytes_done, int err)
#endif
{
	struct end_io_wq *end_io_wq = bio->bi_private;
	struct btrfs_fs_info *fs_info;

#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
	if (bio->bi_size)
		return 1;
#endif

	fs_info = end_io_wq->info;
	end_io_wq->error = err;
401 402
	end_io_wq->work.func = end_workqueue_fn;
	end_io_wq->work.flags = 0;
403 404 405 406 407
	if (bio->bi_rw & (1 << BIO_RW))
		btrfs_queue_worker(&fs_info->endio_write_workers,
				   &end_io_wq->work);
	else
		btrfs_queue_worker(&fs_info->endio_workers, &end_io_wq->work);
408 409 410 411 412 413

#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
	return 0;
#endif
}

414 415
int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
			int metadata)
416
{
417 418 419 420 421 422 423
	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;
424
	end_io_wq->info = info;
425 426
	end_io_wq->error = 0;
	end_io_wq->bio = bio;
427
	end_io_wq->metadata = metadata;
428 429 430

	bio->bi_private = end_io_wq;
	bio->bi_end_io = end_workqueue_bio;
431 432 433
	return 0;
}

434
unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
435
{
436 437 438 439 440
	unsigned long limit = min_t(unsigned long,
				    info->workers.max_workers,
				    info->fs_devices->open_devices);
	return 256 * limit;
}
441

442 443
int btrfs_congested_async(struct btrfs_fs_info *info, int iodone)
{
444 445
	return atomic_read(&info->nr_async_bios) >
		btrfs_async_submit_limit(info);
446 447
}

448 449 450 451
static void run_one_async_submit(struct btrfs_work *work)
{
	struct btrfs_fs_info *fs_info;
	struct async_submit_bio *async;
452
	int limit;
453 454 455

	async = container_of(work, struct  async_submit_bio, work);
	fs_info = BTRFS_I(async->inode)->root->fs_info;
456

457
	limit = btrfs_async_submit_limit(fs_info);
458 459
	limit = limit * 2 / 3;

460
	atomic_dec(&fs_info->nr_async_submits);
461

462 463
	if (atomic_read(&fs_info->nr_async_submits) < limit &&
	    waitqueue_active(&fs_info->async_submit_wait))
464 465
		wake_up(&fs_info->async_submit_wait);

466 467 468 469 470
	async->submit_bio_hook(async->inode, async->rw, async->bio,
			       async->mirror_num);
	kfree(async);
}

471 472 473 474 475
int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
			int rw, struct bio *bio, int mirror_num,
			extent_submit_bio_hook_t *submit_bio_hook)
{
	struct async_submit_bio *async;
476
	int limit = btrfs_async_submit_limit(fs_info);
477 478 479 480 481 482 483 484 485 486

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

	async->inode = inode;
	async->rw = rw;
	async->bio = bio;
	async->mirror_num = mirror_num;
	async->submit_bio_hook = submit_bio_hook;
487 488
	async->work.func = run_one_async_submit;
	async->work.flags = 0;
489
	atomic_inc(&fs_info->nr_async_submits);
490
	btrfs_queue_worker(&fs_info->workers, &async->work);
491

492 493
	if (atomic_read(&fs_info->nr_async_submits) > limit) {
		wait_event_timeout(fs_info->async_submit_wait,
494 495
			   (atomic_read(&fs_info->nr_async_submits) < limit),
			   HZ/10);
496 497 498 499 500

		wait_event_timeout(fs_info->async_submit_wait,
			   (atomic_read(&fs_info->nr_async_bios) < limit),
			   HZ/10);
	}
501 502 503 504
	return 0;
}

static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
505
				 int mirror_num)
506 507 508 509 510 511 512
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u64 offset;
	int ret;

	offset = bio->bi_sector << 9;

513 514
	/*
	 * when we're called for a write, we're already in the async
515
	 * submission context.  Just jump into btrfs_map_bio
516
	 */
517
	if (rw & (1 << BIO_RW)) {
518
		return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
519
				     mirror_num, 1);
520 521
	}

522 523 524 525
	/*
	 * called for a read, do the setup so that checksum validation
	 * can happen in the async kernel threads
	 */
526 527
	ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1);
	BUG_ON(ret);
528

529
	return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
530 531
}

532 533 534
static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
				 int mirror_num)
{
535 536 537 538
	/*
	 * kthread helpers are used to submit writes so that checksumming
	 * can happen in parallel across all CPUs
	 */
539 540 541 542 543 544 545 546
	if (!(rw & (1 << BIO_RW))) {
		return __btree_submit_bio_hook(inode, rw, bio, mirror_num);
	}
	return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
				   inode, rw, bio, mirror_num,
				   __btree_submit_bio_hook);
}

547 548
static int btree_writepage(struct page *page, struct writeback_control *wbc)
{
549 550
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
551 552 553 554 555 556

	if (current->flags & PF_MEMALLOC) {
		redirty_page_for_writepage(wbc, page);
		unlock_page(page);
		return 0;
	}
557 558
	return extent_write_full_page(tree, page, btree_get_extent, wbc);
}
559 560 561 562

static int btree_writepages(struct address_space *mapping,
			    struct writeback_control *wbc)
{
563 564
	struct extent_io_tree *tree;
	tree = &BTRFS_I(mapping->host)->io_tree;
565
	if (wbc->sync_mode == WB_SYNC_NONE) {
566 567
		u64 num_dirty;
		u64 start = 0;
568
		unsigned long thresh = 8 * 1024 * 1024;
569 570 571 572

		if (wbc->for_kupdate)
			return 0;

573 574
		num_dirty = count_range_bits(tree, &start, (u64)-1,
					     thresh, EXTENT_DIRTY);
575 576 577 578
		if (num_dirty < thresh) {
			return 0;
		}
	}
579 580 581
	return extent_writepages(tree, mapping, btree_get_extent, wbc);
}

582 583
int btree_readpage(struct file *file, struct page *page)
{
584 585
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
586 587
	return extent_read_full_page(tree, page, btree_get_extent);
}
C
Chris Mason 已提交
588

589
static int btree_releasepage(struct page *page, gfp_t gfp_flags)
590
{
591 592
	struct extent_io_tree *tree;
	struct extent_map_tree *map;
593
	int ret;
594

595 596 597
	if (PageWriteback(page) || PageDirty(page))
	    return 0;

598 599
	tree = &BTRFS_I(page->mapping->host)->io_tree;
	map = &BTRFS_I(page->mapping->host)->extent_tree;
600

601
	ret = try_release_extent_state(map, tree, page, gfp_flags);
602 603 604 605 606
	if (!ret) {
		return 0;
	}

	ret = try_release_extent_buffer(tree, page);
607 608 609 610 611
	if (ret == 1) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
	}
612

613 614 615
	return ret;
}

616
static void btree_invalidatepage(struct page *page, unsigned long offset)
617
{
618 619
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
620 621
	extent_invalidatepage(tree, page, offset);
	btree_releasepage(page, GFP_NOFS);
622
	if (PagePrivate(page)) {
623 624
		printk("warning page private not zero on page %Lu\n",
		       page_offset(page));
625 626 627 628
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
	}
629 630
}

631
#if 0
632
static int btree_writepage(struct page *page, struct writeback_control *wbc)
633
{
C
Chris Mason 已提交
634
	struct buffer_head *bh;
635
	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
C
Chris Mason 已提交
636 637 638 639 640 641 642 643 644 645 646 647
	struct buffer_head *head;
	if (!page_has_buffers(page)) {
		create_empty_buffers(page, root->fs_info->sb->s_blocksize,
					(1 << BH_Dirty)|(1 << BH_Uptodate));
	}
	head = page_buffers(page);
	bh = head;
	do {
		if (buffer_dirty(bh))
			csum_tree_block(root, bh, 0);
		bh = bh->b_this_page;
	} while (bh != head);
648
	return block_write_full_page(page, btree_get_block, wbc);
649
}
650
#endif
651

652 653 654
static struct address_space_operations btree_aops = {
	.readpage	= btree_readpage,
	.writepage	= btree_writepage,
655
	.writepages	= btree_writepages,
656 657
	.releasepage	= btree_releasepage,
	.invalidatepage = btree_invalidatepage,
658 659 660
	.sync_page	= block_sync_page,
};

661 662
int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
			 u64 parent_transid)
C
Chris Mason 已提交
663
{
664 665
	struct extent_buffer *buf = NULL;
	struct inode *btree_inode = root->fs_info->btree_inode;
666
	int ret = 0;
C
Chris Mason 已提交
667

668
	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
669
	if (!buf)
C
Chris Mason 已提交
670
		return 0;
671
	read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
672
				 buf, 0, 0, btree_get_extent, 0);
673
	free_extent_buffer(buf);
674
	return ret;
C
Chris Mason 已提交
675 676
}

677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698
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;
}


699 700 701 702 703 704 705 706 707 708 709 710
int btrfs_write_tree_block(struct extent_buffer *buf)
{
	return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start,
				      buf->start + buf->len - 1, WB_SYNC_NONE);
}

int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
{
	return btrfs_wait_on_page_writeback_range(buf->first_page->mapping,
				  buf->start, buf->start + buf->len -1);
}

711
struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
712
				      u32 blocksize, u64 parent_transid)
713 714 715 716 717 718 719 720 721 722 723 724
{
	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;

725
	ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
726 727 728

	if (ret == 0) {
		buf->flags |= EXTENT_UPTODATE;
729 730
	} else {
		WARN_ON(1);
731
	}
732
	return buf;
733

734 735
}

736
int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
737
		     struct extent_buffer *buf)
738
{
739
	struct inode *btree_inode = root->fs_info->btree_inode;
740
	if (btrfs_header_generation(buf) ==
741 742
	    root->fs_info->running_transaction->transid) {
		WARN_ON(!btrfs_tree_locked(buf));
743
		clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
744
					  buf);
745
	}
746 747 748
	return 0;
}

749
static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
750
			u32 stripesize, struct btrfs_root *root,
751
			struct btrfs_fs_info *fs_info,
C
Chris Mason 已提交
752
			u64 objectid)
753
{
C
Chris Mason 已提交
754
	root->node = NULL;
755
	root->inode = NULL;
756
	root->commit_root = NULL;
Y
Yan Zheng 已提交
757
	root->ref_tree = NULL;
758 759 760
	root->sectorsize = sectorsize;
	root->nodesize = nodesize;
	root->leafsize = leafsize;
761
	root->stripesize = stripesize;
C
Chris Mason 已提交
762
	root->ref_cows = 0;
763 764
	root->track_dirty = 0;

765
	root->fs_info = fs_info;
766 767
	root->objectid = objectid;
	root->last_trans = 0;
C
Chris Mason 已提交
768 769
	root->highest_inode = 0;
	root->last_inode_alloc = 0;
770
	root->name = NULL;
771
	root->in_sysfs = 0;
772 773

	INIT_LIST_HEAD(&root->dirty_list);
774
	INIT_LIST_HEAD(&root->orphan_list);
775
	INIT_LIST_HEAD(&root->dead_list);
776
	spin_lock_init(&root->node_lock);
777
	spin_lock_init(&root->list_lock);
778
	mutex_init(&root->objectid_mutex);
779
	mutex_init(&root->log_mutex);
C
Chris Mason 已提交
780 781 782 783

	btrfs_leaf_ref_tree_init(&root->ref_tree_struct);
	root->ref_tree = &root->ref_tree_struct;

784 785
	memset(&root->root_key, 0, sizeof(root->root_key));
	memset(&root->root_item, 0, sizeof(root->root_item));
786
	memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
787
	memset(&root->root_kobj, 0, sizeof(root->root_kobj));
788
	root->defrag_trans_start = fs_info->generation;
789
	init_completion(&root->kobj_unregister);
790 791
	root->defrag_running = 0;
	root->defrag_level = 0;
792
	root->root_key.objectid = objectid;
793 794 795
	return 0;
}

796
static int find_and_setup_root(struct btrfs_root *tree_root,
797 798
			       struct btrfs_fs_info *fs_info,
			       u64 objectid,
C
Chris Mason 已提交
799
			       struct btrfs_root *root)
800 801
{
	int ret;
802
	u32 blocksize;
803

804
	__setup_root(tree_root->nodesize, tree_root->leafsize,
805 806
		     tree_root->sectorsize, tree_root->stripesize,
		     root, fs_info, objectid);
807 808 809 810
	ret = btrfs_find_last_root(tree_root, objectid,
				   &root->root_item, &root->root_key);
	BUG_ON(ret);

811 812
	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
813
				     blocksize, 0);
814
	BUG_ON(!root->node);
815 816 817
	return 0;
}

818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844
int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
			     struct btrfs_fs_info *fs_info)
{
	struct extent_buffer *eb;
	int ret;

	if (!fs_info->log_root_tree)
		return 0;

	eb = fs_info->log_root_tree->node;

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

	ret = btrfs_free_extent(trans, fs_info->tree_root,
				eb->start, eb->len,
				BTRFS_TREE_LOG_OBJECTID, 0, 0, 0, 1);
	BUG_ON(ret);

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

int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
			     struct btrfs_fs_info *fs_info)
845 846 847
{
	struct btrfs_root *root;
	struct btrfs_root *tree_root = fs_info->tree_root;
848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885

	root = kzalloc(sizeof(*root), GFP_NOFS);
	if (!root)
		return -ENOMEM;

	__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;
	root->ref_cows = 0;

	root->node = btrfs_alloc_free_block(trans, root, root->leafsize,
					    BTRFS_TREE_LOG_OBJECTID,
					    0, 0, 0, 0, 0);

	btrfs_set_header_nritems(root->node, 0);
	btrfs_set_header_level(root->node, 0);
	btrfs_set_header_bytenr(root->node, root->node->start);
	btrfs_set_header_generation(root->node, trans->transid);
	btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID);

	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);
	fs_info->log_root_tree = root;
	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;
886
	struct btrfs_path *path;
887
	struct extent_buffer *l;
C
Chris Mason 已提交
888
	u64 highest_inode;
889
	u32 blocksize;
890 891
	int ret = 0;

892
	root = kzalloc(sizeof(*root), GFP_NOFS);
C
Chris Mason 已提交
893
	if (!root)
894 895
		return ERR_PTR(-ENOMEM);
	if (location->offset == (u64)-1) {
896
		ret = find_and_setup_root(tree_root, fs_info,
897 898 899 900 901 902 903 904
					  location->objectid, root);
		if (ret) {
			kfree(root);
			return ERR_PTR(ret);
		}
		goto insert;
	}

905
	__setup_root(tree_root->nodesize, tree_root->leafsize,
906 907
		     tree_root->sectorsize, tree_root->stripesize,
		     root, fs_info, location->objectid);
908 909 910 911 912 913 914 915 916

	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;
	}
917 918 919
	l = path->nodes[0];
	read_extent_buffer(l, &root->root_item,
	       btrfs_item_ptr_offset(l, path->slots[0]),
920
	       sizeof(root->root_item));
921
	memcpy(&root->root_key, location, sizeof(*location));
922 923 924 925 926 927 928 929
	ret = 0;
out:
	btrfs_release_path(root, path);
	btrfs_free_path(path);
	if (ret) {
		kfree(root);
		return ERR_PTR(ret);
	}
930 931
	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
932
				     blocksize, 0);
933 934
	BUG_ON(!root->node);
insert:
935 936 937 938 939 940 941
	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;
		}
942 943 944 945
	}
	return root;
}

C
Chris Mason 已提交
946 947 948 949 950 951 952 953 954 955 956 957 958 959 960
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;
}

961 962
struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
					      struct btrfs_key *location)
963 964 965 966
{
	struct btrfs_root *root;
	int ret;

967 968 969 970
	if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
		return fs_info->tree_root;
	if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
		return fs_info->extent_root;
971 972 973 974
	if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
		return fs_info->chunk_root;
	if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
		return fs_info->dev_root;
975

976 977 978 979 980
	root = radix_tree_lookup(&fs_info->fs_roots_radix,
				 (unsigned long)location->objectid);
	if (root)
		return root;

981
	root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
982 983
	if (IS_ERR(root))
		return root;
C
Chris Mason 已提交
984 985
	ret = radix_tree_insert(&fs_info->fs_roots_radix,
				(unsigned long)root->root_key.objectid,
986 987
				root);
	if (ret) {
988
		free_extent_buffer(root->node);
989 990 991
		kfree(root);
		return ERR_PTR(ret);
	}
992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008
	ret = btrfs_find_dead_roots(fs_info->tree_root,
				    root->root_key.objectid, root);
	BUG_ON(ret);

	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;
1009

1010 1011 1012
	if (root->in_sysfs)
		return root;

1013 1014
	ret = btrfs_set_root_name(root, name, namelen);
	if (ret) {
1015
		free_extent_buffer(root->node);
1016 1017 1018 1019 1020 1021
		kfree(root);
		return ERR_PTR(ret);
	}

	ret = btrfs_sysfs_add_root(root);
	if (ret) {
1022
		free_extent_buffer(root->node);
1023 1024 1025 1026
		kfree(root->name);
		kfree(root);
		return ERR_PTR(ret);
	}
1027
	root->in_sysfs = 1;
1028 1029
	return root;
}
1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047
#if 0
static int add_hasher(struct btrfs_fs_info *info, char *type) {
	struct btrfs_hasher *hasher;

	hasher = kmalloc(sizeof(*hasher), GFP_NOFS);
	if (!hasher)
		return -ENOMEM;
	hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC);
	if (!hasher->hash_tfm) {
		kfree(hasher);
		return -EINVAL;
	}
	spin_lock(&info->hash_lock);
	list_add(&hasher->list, &info->hashers);
	spin_unlock(&info->hash_lock);
	return 0;
}
#endif
C
Chris Mason 已提交
1048 1049 1050 1051 1052 1053 1054 1055 1056

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 list_head *cur;
	struct btrfs_device *device;
	struct backing_dev_info *bdi;

1057
	if ((bdi_bits & (1 << BDI_write_congested)) &&
1058
	    btrfs_congested_async(info, 0))
1059 1060
		return 1;

C
Chris Mason 已提交
1061 1062
	list_for_each(cur, &info->fs_devices->devices) {
		device = list_entry(cur, struct btrfs_device, dev_list);
1063 1064
		if (!device->bdev)
			continue;
C
Chris Mason 已提交
1065 1066 1067 1068 1069 1070 1071 1072 1073
		bdi = blk_get_backing_dev_info(device->bdev);
		if (bdi && bdi_congested(bdi, bdi_bits)) {
			ret = 1;
			break;
		}
	}
	return ret;
}

1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093
/*
 * 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 list_head *cur;
	struct btrfs_device *device;
	struct btrfs_fs_info *info;

	info = (struct btrfs_fs_info *)bdi->unplug_io_data;
	list_for_each(cur, &info->fs_devices->devices) {
		device = list_entry(cur, struct btrfs_device, dev_list);
		bdi = blk_get_backing_dev_info(device->bdev);
		if (bdi->unplug_io_fn) {
			bdi->unplug_io_fn(bdi, page);
		}
	}
}

C
Chris Mason 已提交
1094 1095
void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
{
1096
	struct inode *inode;
1097 1098
	struct extent_map_tree *em_tree;
	struct extent_map *em;
1099
	struct address_space *mapping;
1100 1101
	u64 offset;

1102
	/* the generic O_DIRECT read code does this */
1103 1104 1105 1106 1107
	if (!page) {
		__unplug_io_fn(bdi, page);
		return;
	}

1108 1109 1110 1111 1112 1113 1114 1115 1116 1117
	/*
	 * 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;
1118
	offset = page_offset(page);
C
Chris Mason 已提交
1119

1120 1121 1122 1123
	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);
1124 1125
	if (!em) {
		__unplug_io_fn(bdi, page);
1126
		return;
1127
	}
1128

1129 1130 1131 1132 1133
	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
		free_extent_map(em);
		__unplug_io_fn(bdi, page);
		return;
	}
1134 1135 1136 1137
	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
1138 1139 1140 1141
}

static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
{
1142
#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
C
Chris Mason 已提交
1143
	bdi_init(bdi);
1144
#endif
1145
	bdi->ra_pages	= default_backing_dev_info.ra_pages;
C
Chris Mason 已提交
1146 1147 1148 1149 1150 1151 1152 1153 1154
	bdi->state		= 0;
	bdi->capabilities	= default_backing_dev_info.capabilities;
	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;
}

1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193
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);
	if (ret == 1)
		return ret;
	return ret;
}

1194 1195 1196 1197 1198
/*
 * 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)
1199 1200
{
	struct bio *bio;
1201 1202
	struct end_io_wq *end_io_wq;
	struct btrfs_fs_info *fs_info;
1203 1204
	int error;

1205 1206 1207
	end_io_wq = container_of(work, struct end_io_wq, work);
	bio = end_io_wq->bio;
	fs_info = end_io_wq->info;
1208

1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224
	/* metadata bios are special because the whole tree block must
	 * 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
	 */
	if (end_io_wq->metadata && !bio_ready_for_csum(bio)) {
		btrfs_queue_worker(&fs_info->endio_workers,
				   &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);
#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
	bio_endio(bio, bio->bi_size, error);
1225
#else
1226
	bio_endio(bio, error);
1227 1228 1229
#endif
}

1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275
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);

1276 1277 1278 1279
		if (root->fs_info->total_ref_cache_size > 20 * 1024 * 1024) {
			printk("btrfs: total reference cache size %Lu\n",
				root->fs_info->total_ref_cache_size);
		}
Y
Yan Zheng 已提交
1280

1281 1282 1283 1284 1285 1286
		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
1287

1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313
		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);
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;
}

1314
struct btrfs_root *open_ctree(struct super_block *sb,
1315 1316
			      struct btrfs_fs_devices *fs_devices,
			      char *options)
1317
{
1318 1319 1320 1321
	u32 sectorsize;
	u32 nodesize;
	u32 leafsize;
	u32 blocksize;
1322
	u32 stripesize;
1323
	struct buffer_head *bh;
1324
	struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
C
Chris Mason 已提交
1325
						 GFP_NOFS);
1326
	struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root),
C
Chris Mason 已提交
1327
					       GFP_NOFS);
1328
	struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
C
Chris Mason 已提交
1329
						GFP_NOFS);
1330
	struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
1331
						GFP_NOFS);
1332
	struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
1333
					      GFP_NOFS);
1334 1335
	struct btrfs_root *log_tree_root;

1336
	int ret;
1337
	int err = -EINVAL;
1338

C
Chris Mason 已提交
1339
	struct btrfs_super_block *disk_super;
1340

C
Chris Mason 已提交
1341 1342 1343 1344
	if (!extent_root || !tree_root || !fs_info) {
		err = -ENOMEM;
		goto fail;
	}
1345
	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
C
Chris Mason 已提交
1346
	INIT_LIST_HEAD(&fs_info->trans_list);
1347
	INIT_LIST_HEAD(&fs_info->dead_roots);
1348
	INIT_LIST_HEAD(&fs_info->hashers);
1349
	INIT_LIST_HEAD(&fs_info->delalloc_inodes);
1350
	spin_lock_init(&fs_info->hash_lock);
1351
	spin_lock_init(&fs_info->delalloc_lock);
1352
	spin_lock_init(&fs_info->new_trans_lock);
Y
Yan Zheng 已提交
1353
	spin_lock_init(&fs_info->ref_cache_lock);
1354

1355
	init_completion(&fs_info->kobj_unregister);
1356 1357
	fs_info->tree_root = tree_root;
	fs_info->extent_root = extent_root;
1358 1359
	fs_info->chunk_root = chunk_root;
	fs_info->dev_root = dev_root;
1360
	fs_info->fs_devices = fs_devices;
1361
	INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
1362
	INIT_LIST_HEAD(&fs_info->space_info);
1363
	btrfs_mapping_init(&fs_info->mapping_tree);
1364
	atomic_set(&fs_info->nr_async_submits, 0);
1365
	atomic_set(&fs_info->nr_async_bios, 0);
1366
	atomic_set(&fs_info->throttles, 0);
1367
	atomic_set(&fs_info->throttle_gen, 0);
C
Chris Mason 已提交
1368
	fs_info->sb = sb;
1369
	fs_info->max_extent = (u64)-1;
1370
	fs_info->max_inline = 8192 * 1024;
C
Chris Mason 已提交
1371
	setup_bdi(fs_info, &fs_info->bdi);
1372 1373
	fs_info->btree_inode = new_inode(sb);
	fs_info->btree_inode->i_ino = 1;
C
Chris Mason 已提交
1374
	fs_info->btree_inode->i_nlink = 1;
1375
	fs_info->thread_pool_size = min(num_online_cpus() + 2, 8);
1376

1377 1378 1379
	INIT_LIST_HEAD(&fs_info->ordered_extents);
	spin_lock_init(&fs_info->ordered_extent_lock);

1380 1381 1382
	sb->s_blocksize = 4096;
	sb->s_blocksize_bits = blksize_bits(4096);

1383 1384 1385 1386 1387 1388
	/*
	 * 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;
1389
	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
C
Chris Mason 已提交
1390 1391
	fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;

1392
	extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
1393 1394
			     fs_info->btree_inode->i_mapping,
			     GFP_NOFS);
1395 1396 1397 1398
	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;
1399

1400
	extent_io_tree_init(&fs_info->free_space_cache,
1401
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
1402
	extent_io_tree_init(&fs_info->block_group_cache,
1403
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
1404
	extent_io_tree_init(&fs_info->pinned_extents,
1405
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
1406
	extent_io_tree_init(&fs_info->pending_del,
1407
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
1408
	extent_io_tree_init(&fs_info->extent_ins,
1409
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
1410
	fs_info->do_barriers = 1;
1411

1412 1413 1414
	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
1415
	insert_inode_hash(fs_info->btree_inode);
1416
	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
C
Chris Mason 已提交
1417

C
Chris Mason 已提交
1418
	mutex_init(&fs_info->trans_mutex);
1419
	mutex_init(&fs_info->tree_log_mutex);
1420
	mutex_init(&fs_info->drop_mutex);
1421 1422
	mutex_init(&fs_info->alloc_mutex);
	mutex_init(&fs_info->chunk_mutex);
1423 1424
	mutex_init(&fs_info->transaction_kthread_mutex);
	mutex_init(&fs_info->cleaner_mutex);
1425
	mutex_init(&fs_info->volume_mutex);
1426
	init_waitqueue_head(&fs_info->transaction_throttle);
1427
	init_waitqueue_head(&fs_info->transaction_wait);
1428
	init_waitqueue_head(&fs_info->async_submit_wait);
1429 1430 1431 1432
	init_waitqueue_head(&fs_info->tree_log_wait);
	atomic_set(&fs_info->tree_log_commit, 0);
	atomic_set(&fs_info->tree_log_writers, 0);
	fs_info->tree_log_transid = 0;
1433

1434 1435 1436 1437 1438 1439 1440 1441
#if 0
	ret = add_hasher(fs_info, "crc32c");
	if (ret) {
		printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
		err = -ENOMEM;
		goto fail_iput;
	}
#endif
1442
	__setup_root(4096, 4096, 4096, 4096, tree_root,
C
Chris Mason 已提交
1443
		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
1444

1445

1446 1447 1448
	bh = __bread(fs_devices->latest_bdev,
		     BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
	if (!bh)
C
Chris Mason 已提交
1449 1450
		goto fail_iput;

1451 1452
	memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy));
	brelse(bh);
1453

1454
	memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE);
1455

1456
	disk_super = &fs_info->super_copy;
1457
	if (!btrfs_super_root(disk_super))
C
Chris Mason 已提交
1458
		goto fail_sb_buffer;
1459

1460 1461 1462
	err = btrfs_parse_options(tree_root, options);
	if (err)
		goto fail_sb_buffer;
1463

1464 1465 1466 1467 1468
	/*
	 * 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.
	 */
1469 1470 1471
	btrfs_init_workers(&fs_info->workers, "worker",
			   fs_info->thread_pool_size);
	btrfs_init_workers(&fs_info->submit_workers, "submit",
1472 1473
			   min_t(u64, fs_devices->num_devices,
			   fs_info->thread_pool_size));
1474 1475 1476 1477 1478 1479

	/* 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;
1480 1481 1482 1483 1484 1485 1486 1487

	/* fs_info->workers is responsible for checksumming file data
	 * blocks and metadata.  Using a larger idle thresh allows each
	 * worker thread to operate on things in roughly the order they
	 * were sent by the writeback daemons, improving overall locality
	 * of the IO going down the pipe.
	 */
	fs_info->workers.idle_thresh = 128;
1488

1489 1490 1491 1492
	btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1);
	btrfs_init_workers(&fs_info->endio_workers, "endio",
			   fs_info->thread_pool_size);
	btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
1493
			   fs_info->thread_pool_size);
1494 1495 1496 1497 1498 1499 1500 1501

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

1502
	btrfs_start_workers(&fs_info->workers, 1);
1503
	btrfs_start_workers(&fs_info->submit_workers, 1);
1504
	btrfs_start_workers(&fs_info->fixup_workers, 1);
1505
	btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size);
1506 1507
	btrfs_start_workers(&fs_info->endio_write_workers,
			    fs_info->thread_pool_size);
1508

1509
	err = -EINVAL;
1510
	if (btrfs_super_num_devices(disk_super) > fs_devices->open_devices) {
1511 1512
		printk("Btrfs: wanted %llu devices, but found %llu\n",
		       (unsigned long long)btrfs_super_num_devices(disk_super),
1513
		       (unsigned long long)fs_devices->open_devices);
1514 1515 1516 1517 1518
		if (btrfs_test_opt(tree_root, DEGRADED))
			printk("continuing in degraded mode\n");
		else {
			goto fail_sb_buffer;
		}
1519
	}
1520

1521 1522
	fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);

1523 1524 1525
	nodesize = btrfs_super_nodesize(disk_super);
	leafsize = btrfs_super_leafsize(disk_super);
	sectorsize = btrfs_super_sectorsize(disk_super);
1526
	stripesize = btrfs_super_stripesize(disk_super);
1527 1528 1529
	tree_root->nodesize = nodesize;
	tree_root->leafsize = leafsize;
	tree_root->sectorsize = sectorsize;
1530
	tree_root->stripesize = stripesize;
1531 1532 1533

	sb->s_blocksize = sectorsize;
	sb->s_blocksize_bits = blksize_bits(sectorsize);
1534

C
Chris Mason 已提交
1535 1536 1537 1538 1539
	if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
		    sizeof(disk_super->magic))) {
		printk("btrfs: valid FS not found on %s\n", sb->s_id);
		goto fail_sb_buffer;
	}
1540

1541
	mutex_lock(&fs_info->chunk_mutex);
1542
	ret = btrfs_read_sys_array(tree_root);
1543
	mutex_unlock(&fs_info->chunk_mutex);
1544 1545 1546 1547 1548
	if (ret) {
		printk("btrfs: failed to read the system array on %s\n",
		       sb->s_id);
		goto fail_sys_array;
	}
1549 1550 1551 1552 1553 1554 1555 1556 1557

	blocksize = btrfs_level_size(tree_root,
				     btrfs_super_chunk_root_level(disk_super));

	__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),
1558
					   blocksize, 0);
1559 1560
	BUG_ON(!chunk_root->node);

1561 1562 1563 1564
	read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
	         (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
		 BTRFS_UUID_SIZE);

1565
	mutex_lock(&fs_info->chunk_mutex);
1566
	ret = btrfs_read_chunk_tree(chunk_root);
1567
	mutex_unlock(&fs_info->chunk_mutex);
1568 1569
	BUG_ON(ret);

1570 1571
	btrfs_close_extra_devices(fs_devices);

1572 1573
	blocksize = btrfs_level_size(tree_root,
				     btrfs_super_root_level(disk_super));
1574

1575

C
Chris Mason 已提交
1576
	tree_root->node = read_tree_block(tree_root,
1577
					  btrfs_super_root(disk_super),
1578
					  blocksize, 0);
C
Chris Mason 已提交
1579 1580
	if (!tree_root->node)
		goto fail_sb_buffer;
1581

1582 1583

	ret = find_and_setup_root(tree_root, fs_info,
C
Chris Mason 已提交
1584
				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
1585
	if (ret)
C
Chris Mason 已提交
1586
		goto fail_tree_root;
1587 1588 1589 1590 1591 1592 1593 1594
	extent_root->track_dirty = 1;

	ret = find_and_setup_root(tree_root, fs_info,
				  BTRFS_DEV_TREE_OBJECTID, dev_root);
	dev_root->track_dirty = 1;

	if (ret)
		goto fail_extent_root;
1595

C
Chris Mason 已提交
1596 1597
	btrfs_read_block_groups(extent_root);

1598
	fs_info->generation = btrfs_super_generation(disk_super) + 1;
C
Chris Mason 已提交
1599 1600 1601
	fs_info->data_alloc_profile = (u64)-1;
	fs_info->metadata_alloc_profile = (u64)-1;
	fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
1602 1603 1604 1605 1606 1607 1608 1609 1610
	fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
					       "btrfs-cleaner");
	if (!fs_info->cleaner_kthread)
		goto fail_extent_root;

	fs_info->transaction_kthread = kthread_run(transaction_kthread,
						   tree_root,
						   "btrfs-transaction");
	if (!fs_info->transaction_kthread)
1611
		goto fail_cleaner;
1612

1613 1614 1615 1616 1617 1618 1619
	if (btrfs_super_log_root(disk_super) != 0) {
		u32 blocksize;
		u64 bytenr = btrfs_super_log_root(disk_super);

		blocksize =
		     btrfs_level_size(tree_root,
				      btrfs_super_log_root_level(disk_super));
C
Chris Mason 已提交
1620

1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632
		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,
						      blocksize, 0);
		ret = btrfs_recover_log_trees(log_tree_root);
		BUG_ON(ret);
	}
	fs_info->last_trans_committed = btrfs_super_generation(disk_super);
1633
	return tree_root;
C
Chris Mason 已提交
1634

1635
fail_cleaner:
1636
	kthread_stop(fs_info->cleaner_kthread);
1637 1638
fail_extent_root:
	free_extent_buffer(extent_root->node);
C
Chris Mason 已提交
1639
fail_tree_root:
1640
	free_extent_buffer(tree_root->node);
1641
fail_sys_array:
C
Chris Mason 已提交
1642
fail_sb_buffer:
1643
	btrfs_stop_workers(&fs_info->fixup_workers);
1644 1645
	btrfs_stop_workers(&fs_info->workers);
	btrfs_stop_workers(&fs_info->endio_workers);
1646
	btrfs_stop_workers(&fs_info->endio_write_workers);
1647
	btrfs_stop_workers(&fs_info->submit_workers);
1648 1649
fail_iput:
	iput(fs_info->btree_inode);
C
Chris Mason 已提交
1650
fail:
1651
	btrfs_close_devices(fs_info->fs_devices);
1652 1653
	btrfs_mapping_tree_free(&fs_info->mapping_tree);

C
Chris Mason 已提交
1654 1655
	kfree(extent_root);
	kfree(tree_root);
1656
#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
1657
	bdi_destroy(&fs_info->bdi);
1658
#endif
C
Chris Mason 已提交
1659 1660
	kfree(fs_info);
	return ERR_PTR(err);
1661 1662
}

1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674
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));
		}
1675 1676 1677
		/* note, we dont' set_buffer_write_io_error because we have
		 * our own ways of dealing with the IO errors
		 */
1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688
		clear_buffer_uptodate(bh);
	}
	unlock_buffer(bh);
	put_bh(bh);
}

int write_all_supers(struct btrfs_root *root)
{
	struct list_head *cur;
	struct list_head *head = &root->fs_info->fs_devices->devices;
	struct btrfs_device *dev;
1689
	struct btrfs_super_block *sb;
1690 1691 1692 1693
	struct btrfs_dev_item *dev_item;
	struct buffer_head *bh;
	int ret;
	int do_barriers;
1694 1695
	int max_errors;
	int total_errors = 0;
1696 1697
	u32 crc;
	u64 flags;
1698

1699
	max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
1700 1701
	do_barriers = !btrfs_test_opt(root, NOBARRIER);

1702 1703
	sb = &root->fs_info->super_for_commit;
	dev_item = &sb->dev_item;
1704 1705
	list_for_each(cur, head) {
		dev = list_entry(cur, struct btrfs_device, dev_list);
1706 1707 1708 1709 1710 1711 1712
		if (!dev->bdev) {
			total_errors++;
			continue;
		}
		if (!dev->in_fs_metadata)
			continue;

1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730
		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);
		flags = btrfs_super_flags(sb);
		btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);


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

		bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / 4096,
1731 1732
			      BTRFS_SUPER_INFO_SIZE);

1733
		memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754
		dev->pending_io = bh;

		get_bh(bh);
		set_buffer_uptodate(bh);
		lock_buffer(bh);
		bh->b_end_io = btrfs_end_buffer_write_sync;

		if (do_barriers && dev->barriers) {
			ret = submit_bh(WRITE_BARRIER, bh);
			if (ret == -EOPNOTSUPP) {
				printk("btrfs: disabling barriers on dev %s\n",
				       dev->name);
				set_buffer_uptodate(bh);
				dev->barriers = 0;
				get_bh(bh);
				lock_buffer(bh);
				ret = submit_bh(WRITE, bh);
			}
		} else {
			ret = submit_bh(WRITE, bh);
		}
1755 1756
		if (ret)
			total_errors++;
1757
	}
1758 1759 1760 1761 1762
	if (total_errors > max_errors) {
		printk("btrfs: %d errors while writing supers\n", total_errors);
		BUG();
	}
	total_errors = 0;
1763 1764 1765

	list_for_each(cur, head) {
		dev = list_entry(cur, struct btrfs_device, dev_list);
1766 1767 1768 1769 1770
		if (!dev->bdev)
			continue;
		if (!dev->in_fs_metadata)
			continue;

1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784
		BUG_ON(!dev->pending_io);
		bh = dev->pending_io;
		wait_on_buffer(bh);
		if (!buffer_uptodate(dev->pending_io)) {
			if (do_barriers && dev->barriers) {
				printk("btrfs: disabling barriers on dev %s\n",
				       dev->name);
				set_buffer_uptodate(bh);
				get_bh(bh);
				lock_buffer(bh);
				dev->barriers = 0;
				ret = submit_bh(WRITE, bh);
				BUG_ON(ret);
				wait_on_buffer(bh);
1785 1786
				if (!buffer_uptodate(bh))
					total_errors++;
1787
			} else {
1788
				total_errors++;
1789 1790 1791 1792 1793 1794
			}

		}
		dev->pending_io = NULL;
		brelse(bh);
	}
1795 1796 1797 1798
	if (total_errors > max_errors) {
		printk("btrfs: %d errors while writing supers\n", total_errors);
		BUG();
	}
1799 1800 1801
	return 0;
}

1802
int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
C
Chris Mason 已提交
1803
		      *root)
1804
{
1805
	int ret;
1806

1807
	ret = write_all_supers(root);
1808
	return ret;
C
Chris Mason 已提交
1809 1810
}

1811
int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
C
Chris Mason 已提交
1812 1813 1814
{
	radix_tree_delete(&fs_info->fs_roots_radix,
			  (unsigned long)root->root_key.objectid);
1815 1816
	if (root->in_sysfs)
		btrfs_sysfs_del_root(root);
C
Chris Mason 已提交
1817 1818 1819
	if (root->inode)
		iput(root->inode);
	if (root->node)
1820
		free_extent_buffer(root->node);
C
Chris Mason 已提交
1821
	if (root->commit_root)
1822
		free_extent_buffer(root->commit_root);
1823 1824
	if (root->name)
		kfree(root->name);
C
Chris Mason 已提交
1825 1826 1827 1828
	kfree(root);
	return 0;
}

C
Chris Mason 已提交
1829
static int del_fs_roots(struct btrfs_fs_info *fs_info)
1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840
{
	int ret;
	struct btrfs_root *gang[8];
	int i;

	while(1) {
		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
1841
		for (i = 0; i < ret; i++)
1842
			btrfs_free_fs_root(fs_info, gang[i]);
1843 1844 1845
	}
	return 0;
}
1846

C
Chris Mason 已提交
1847
int close_ctree(struct btrfs_root *root)
C
Chris Mason 已提交
1848
{
1849
	int ret;
1850
	struct btrfs_trans_handle *trans;
1851
	struct btrfs_fs_info *fs_info = root->fs_info;
1852

1853
	fs_info->closing = 1;
1854 1855
	smp_mb();

1856 1857 1858 1859
	kthread_stop(root->fs_info->transaction_kthread);
	kthread_stop(root->fs_info->cleaner_kthread);

	btrfs_clean_old_snapshots(root);
C
Chris Mason 已提交
1860
	trans = btrfs_start_transaction(root, 1);
1861
	ret = btrfs_commit_transaction(trans, root);
C
Chris Mason 已提交
1862 1863 1864 1865
	/* run commit again to  drop the original snapshot */
	trans = btrfs_start_transaction(root, 1);
	btrfs_commit_transaction(trans, root);
	ret = btrfs_write_and_wait_transaction(NULL, root);
1866
	BUG_ON(ret);
1867

C
Chris Mason 已提交
1868
	write_ctree_super(NULL, root);
1869

C
Chris Mason 已提交
1870 1871 1872 1873
	if (fs_info->delalloc_bytes) {
		printk("btrfs: at unmount delalloc count %Lu\n",
		       fs_info->delalloc_bytes);
	}
Y
Yan Zheng 已提交
1874 1875 1876 1877
	if (fs_info->total_ref_cache_size) {
		printk("btrfs: at umount reference cache size %Lu\n",
			fs_info->total_ref_cache_size);
	}
1878

1879
	if (fs_info->extent_root->node)
1880
		free_extent_buffer(fs_info->extent_root->node);
1881

1882
	if (fs_info->tree_root->node)
1883
		free_extent_buffer(fs_info->tree_root->node);
1884

1885 1886 1887 1888 1889 1890
	if (root->fs_info->chunk_root->node);
		free_extent_buffer(root->fs_info->chunk_root->node);

	if (root->fs_info->dev_root->node);
		free_extent_buffer(root->fs_info->dev_root->node);

C
Chris Mason 已提交
1891
	btrfs_free_block_groups(root->fs_info);
1892
	fs_info->closing = 2;
1893
	del_fs_roots(fs_info);
1894 1895 1896

	filemap_write_and_wait(fs_info->btree_inode->i_mapping);

1897
	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
1898

1899
	btrfs_stop_workers(&fs_info->fixup_workers);
1900 1901
	btrfs_stop_workers(&fs_info->workers);
	btrfs_stop_workers(&fs_info->endio_workers);
1902
	btrfs_stop_workers(&fs_info->endio_write_workers);
1903
	btrfs_stop_workers(&fs_info->submit_workers);
1904

1905
	iput(fs_info->btree_inode);
1906 1907 1908 1909 1910 1911 1912 1913 1914 1915
#if 0
	while(!list_empty(&fs_info->hashers)) {
		struct btrfs_hasher *hasher;
		hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher,
				    hashers);
		list_del(&hasher->hashers);
		crypto_free_hash(&fs_info->hash_tfm);
		kfree(hasher);
	}
#endif
1916
	btrfs_close_devices(fs_info->fs_devices);
1917
	btrfs_mapping_tree_free(&fs_info->mapping_tree);
1918

1919
#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
C
Chris Mason 已提交
1920
	bdi_destroy(&fs_info->bdi);
1921
#endif
1922

1923 1924
	kfree(fs_info->extent_root);
	kfree(fs_info->tree_root);
1925 1926
	kfree(fs_info->chunk_root);
	kfree(fs_info->dev_root);
1927 1928 1929
	return 0;
}

1930
int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
1931
{
1932
	int ret;
1933
	struct inode *btree_inode = buf->first_page->mapping->host;
1934 1935 1936 1937 1938 1939 1940 1941

	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;
1942 1943 1944
}

int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
C
Chris Mason 已提交
1945
{
1946
	struct inode *btree_inode = buf->first_page->mapping->host;
1947
	return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
1948 1949
					  buf);
}
1950

1951 1952
void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
{
1953
	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1954 1955
	u64 transid = btrfs_header_generation(buf);
	struct inode *btree_inode = root->fs_info->btree_inode;
1956

1957
	WARN_ON(!btrfs_tree_locked(buf));
C
Chris Mason 已提交
1958 1959
	if (transid != root->fs_info->generation) {
		printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
1960
			(unsigned long long)buf->start,
C
Chris Mason 已提交
1961 1962 1963
			transid, root->fs_info->generation);
		WARN_ON(1);
	}
1964
	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf);
1965 1966
}

1967
void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
C
Chris Mason 已提交
1968
{
1969 1970 1971 1972
	/*
	 * looks as though older kernels can get into trouble with
	 * this code, they end up stuck in balance_dirty_pages forever
	 */
1973 1974 1975
	struct extent_io_tree *tree;
	u64 num_dirty;
	u64 start = 0;
1976
	unsigned long thresh = 96 * 1024 * 1024;
1977 1978
	tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;

1979
	if (current_is_pdflush() || current->flags & PF_MEMALLOC)
1980 1981 1982 1983 1984 1985
		return;

	num_dirty = count_range_bits(tree, &start, (u64)-1,
				     thresh, EXTENT_DIRTY);
	if (num_dirty > thresh) {
		balance_dirty_pages_ratelimited_nr(
C
Chris Mason 已提交
1986
				   root->fs_info->btree_inode->i_mapping, 1);
1987
	}
1988
	return;
C
Chris Mason 已提交
1989
}
1990

1991
int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
1992
{
1993
	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1994
	int ret;
1995
	ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
1996 1997 1998 1999
	if (ret == 0) {
		buf->flags |= EXTENT_UPTODATE;
	}
	return ret;
2000
}
2001

C
Chris Mason 已提交
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029
int btree_lock_page_hook(struct page *page)
{
	struct inode *inode = page->mapping->host;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	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);
	spin_lock(&root->fs_info->hash_lock);
	btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
	spin_unlock(&root->fs_info->hash_lock);
	btrfs_tree_unlock(eb);
	free_extent_buffer(eb);
out:
	lock_page(page);
	return 0;
}

2030
static struct extent_io_ops btree_extent_io_ops = {
C
Chris Mason 已提交
2031
	.write_cache_pages_lock_hook = btree_lock_page_hook,
2032
	.writepage_io_hook = btree_writepage_io_hook,
2033
	.readpage_end_io_hook = btree_readpage_end_io_hook,
2034
	.submit_bio_hook = btree_submit_bio_hook,
2035 2036
	/* note we're sharing with inode.c for the merge bio hook */
	.merge_bio_hook = btrfs_merge_bio_hook,
2037
};