disk-io.c 54.7 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
#include <linux/kthread.h>
# include <linux/freezer.h>
30
#include "crc32c.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"
C
Chris Mason 已提交
39
#include "ref-cache.h"
40
#include "tree-log.h"
41

42 43
#if 0
static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
44
{
45 46 47 48
	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
49
		return 1;
50
	}
51
	return 0;
52
}
53
#endif
54

55
static struct extent_io_ops btree_extent_io_ops;
56
static void end_workqueue_fn(struct btrfs_work *work);
57

C
Chris Mason 已提交
58 59 60 61 62
/*
 * 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.
 */
63 64 65 66 67 68
struct end_io_wq {
	struct bio *bio;
	bio_end_io_t *end_io;
	void *private;
	struct btrfs_fs_info *info;
	int error;
69
	int metadata;
70
	struct list_head list;
71
	struct btrfs_work work;
72
};
73

C
Chris Mason 已提交
74 75 76 77 78
/*
 * 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.
 */
79 80 81 82 83 84 85
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;
86
	struct btrfs_work work;
87 88
};

C
Chris Mason 已提交
89 90 91 92
/*
 * extents on the btree inode are pretty simple, there's one extent
 * that covers the entire device
 */
93
struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
94
				    size_t page_offset, u64 start, u64 len,
95
				    int create)
96
{
97 98 99 100
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct extent_map *em;
	int ret;

101 102
	spin_lock(&em_tree->lock);
	em = lookup_extent_mapping(em_tree, start, len);
103 104 105 106
	if (em) {
		em->bdev =
			BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
		spin_unlock(&em_tree->lock);
107
		goto out;
108 109
	}
	spin_unlock(&em_tree->lock);
110

111 112 113 114 115 116
	em = alloc_extent_map(GFP_NOFS);
	if (!em) {
		em = ERR_PTR(-ENOMEM);
		goto out;
	}
	em->start = 0;
117
	em->len = (u64)-1;
118
	em->block_start = 0;
119
	em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
120 121

	spin_lock(&em_tree->lock);
122 123
	ret = add_extent_mapping(em_tree, em);
	if (ret == -EEXIST) {
124 125 126 127 128
		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);
129
		free_extent_map(em);
130
		em = lookup_extent_mapping(em_tree, start, len);
131 132 133
		if (em) {
			printk("after failing, found %Lu %Lu %Lu\n",
			       em->start, em->len, em->block_start);
134
			ret = 0;
135 136 137 138 139 140 141 142 143
		} 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);
			}
144
			ret = -EIO;
145
		}
146
	} else if (ret) {
147 148
		free_extent_map(em);
		em = NULL;
149
	}
150 151 152 153
	spin_unlock(&em_tree->lock);

	if (ret)
		em = ERR_PTR(ret);
154 155
out:
	return em;
156 157
}

158 159
u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
{
160
	return btrfs_crc32c(seed, data, len);
161 162 163 164 165 166 167
}

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

C
Chris Mason 已提交
168 169 170 171
/*
 * compute the csum for a btree block, and either verify it or write it
 * into the csum field of the block.
 */
172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
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) {
206
		/* FIXME, this is not good */
207
		if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) {
208 209 210 211 212 213
			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
214
			       "wanted %X found %X level %d\n",
215
			       root->fs_info->sb->s_id,
C
Chris Mason 已提交
216
			       buf->start, val, found, btrfs_header_level(buf));
217 218 219 220 221 222 223 224
			return 1;
		}
	} else {
		write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE);
	}
	return 0;
}

C
Chris Mason 已提交
225 226 227 228 229 230
/*
 * 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.
 */
231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250
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
251
out:
252 253 254 255 256
	unlock_extent(io_tree, eb->start, eb->start + eb->len - 1,
		      GFP_NOFS);
	return ret;
}

C
Chris Mason 已提交
257 258 259 260
/*
 * helper to read a given tree block, doing retries as required when
 * the checksums don't match and we have alternate mirrors to try.
 */
261 262
static int btree_read_extent_buffer_pages(struct btrfs_root *root,
					  struct extent_buffer *eb,
263
					  u64 start, u64 parent_transid)
264 265 266 267 268 269 270 271 272 273
{
	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);
274 275
		if (!ret &&
		    !verify_parent_transid(io_tree, eb, parent_transid))
276
			return ret;
277
printk("read extent buffer pages failed with ret %d mirror no %d\n", ret, mirror_num);
278 279
		num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
					      eb->start, eb->len);
C
Chris Mason 已提交
280
		if (num_copies == 1)
281
			return ret;
C
Chris Mason 已提交
282

283
		mirror_num++;
C
Chris Mason 已提交
284
		if (mirror_num > num_copies)
285 286 287 288
			return ret;
	}
	return -EIO;
}
289

C
Chris Mason 已提交
290 291 292 293
/*
 * 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
 */
294 295
int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
{
296
	struct extent_io_tree *tree;
297
	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
298 299 300 301
	u64 found_start;
	int found_level;
	unsigned long len;
	struct extent_buffer *eb;
302 303
	int ret;

304
	tree = &BTRFS_I(page->mapping->host)->io_tree;
305 306 307 308 309 310 311 312 313 314

	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);
315 316
	ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
					     btrfs_header_generation(eb));
317
	BUG_ON(ret);
318 319 320 321
	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);
322 323 324 325 326 327 328 329 330 331 332 333 334
		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;
335 336
	}
	found_level = btrfs_header_level(eb);
C
Chris Mason 已提交
337

338
	csum_tree_block(root, eb, 0);
339
err:
340 341 342 343 344
	free_extent_buffer(eb);
out:
	return 0;
}

345 346 347 348 349 350 351 352 353
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;
354
	int ret = 0;
355 356 357 358 359 360 361 362 363 364 365

	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);
366

367
	found_start = btrfs_header_bytenr(eb);
368
	if (found_start != start) {
369 370 371
		printk("bad tree block start %llu %llu\n",
		       (unsigned long long)found_start,
		       (unsigned long long)eb->start);
372
		ret = -EIO;
373 374 375 376 377 378
		goto err;
	}
	if (eb->first_page != page) {
		printk("bad first page %lu %lu\n", eb->first_page->index,
		       page->index);
		WARN_ON(1);
379
		ret = -EIO;
380 381
		goto err;
	}
382 383 384 385 386 387 388
	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;
	}
389 390 391
	found_level = btrfs_header_level(eb);

	ret = csum_tree_block(root, eb, 1);
392 393
	if (ret)
		ret = -EIO;
394 395 396 397 398 399

	end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
	end = eb->start + end - 1;
err:
	free_extent_buffer(eb);
out:
400
	return ret;
401 402 403 404 405 406 407 408 409
}

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;
410 411
	end_io_wq->work.func = end_workqueue_fn;
	end_io_wq->work.flags = 0;
412 413 414 415 416
	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);
417 418
}

419 420
int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
			int metadata)
421
{
422 423 424 425 426 427 428
	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;
429
	end_io_wq->info = info;
430 431
	end_io_wq->error = 0;
	end_io_wq->bio = bio;
432
	end_io_wq->metadata = metadata;
433 434 435

	bio->bi_private = end_io_wq;
	bio->bi_end_io = end_workqueue_bio;
436 437 438
	return 0;
}

439
unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
440
{
441 442 443 444 445
	unsigned long limit = min_t(unsigned long,
				    info->workers.max_workers,
				    info->fs_devices->open_devices);
	return 256 * limit;
}
446

447 448
int btrfs_congested_async(struct btrfs_fs_info *info, int iodone)
{
449 450
	return atomic_read(&info->nr_async_bios) >
		btrfs_async_submit_limit(info);
451 452
}

453 454 455 456
static void run_one_async_submit(struct btrfs_work *work)
{
	struct btrfs_fs_info *fs_info;
	struct async_submit_bio *async;
457
	int limit;
458 459 460

	async = container_of(work, struct  async_submit_bio, work);
	fs_info = BTRFS_I(async->inode)->root->fs_info;
461

462
	limit = btrfs_async_submit_limit(fs_info);
463 464
	limit = limit * 2 / 3;

465
	atomic_dec(&fs_info->nr_async_submits);
466

467 468
	if (atomic_read(&fs_info->nr_async_submits) < limit &&
	    waitqueue_active(&fs_info->async_submit_wait))
469 470
		wake_up(&fs_info->async_submit_wait);

471 472 473 474 475
	async->submit_bio_hook(async->inode, async->rw, async->bio,
			       async->mirror_num);
	kfree(async);
}

476 477 478 479 480
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;
481
	int limit = btrfs_async_submit_limit(fs_info);
482 483 484 485 486 487 488 489 490 491

	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;
492 493
	async->work.func = run_one_async_submit;
	async->work.flags = 0;
494 495 496 497 498 499 500

	while(atomic_read(&fs_info->async_submit_draining) &&
	      atomic_read(&fs_info->nr_async_submits)) {
		wait_event(fs_info->async_submit_wait,
			   (atomic_read(&fs_info->nr_async_submits) == 0));
	}

501
	atomic_inc(&fs_info->nr_async_submits);
502
	btrfs_queue_worker(&fs_info->workers, &async->work);
503

504 505
	if (atomic_read(&fs_info->nr_async_submits) > limit) {
		wait_event_timeout(fs_info->async_submit_wait,
506 507
			   (atomic_read(&fs_info->nr_async_submits) < limit),
			   HZ/10);
508 509 510 511 512

		wait_event_timeout(fs_info->async_submit_wait,
			   (atomic_read(&fs_info->nr_async_bios) < limit),
			   HZ/10);
	}
513 514 515
	return 0;
}

516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531
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);
	while(bio_index < bio->bi_vcnt) {
		root = BTRFS_I(bvec->bv_page->mapping->host)->root;
		csum_dirty_buffer(root, bvec->bv_page);
		bio_index++;
		bvec++;
	}
	return 0;
}

532
static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
533
				 int mirror_num)
534 535 536 537
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret;

538 539
	/*
	 * when we're called for a write, we're already in the async
540
	 * submission context.  Just jump into btrfs_map_bio
541
	 */
542
	if (rw & (1 << BIO_RW)) {
543
		btree_csum_one_bio(bio);
544
		return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
545
				     mirror_num, 1);
546 547
	}

548 549 550 551
	/*
	 * called for a read, do the setup so that checksum validation
	 * can happen in the async kernel threads
	 */
552 553
	ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1);
	BUG_ON(ret);
554

555
	return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
556 557
}

558 559 560
static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
				 int mirror_num)
{
561 562 563 564
	/*
	 * kthread helpers are used to submit writes so that checksumming
	 * can happen in parallel across all CPUs
	 */
565 566 567 568 569 570 571 572
	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);
}

573 574
static int btree_writepage(struct page *page, struct writeback_control *wbc)
{
575 576
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
577 578 579 580 581 582

	if (current->flags & PF_MEMALLOC) {
		redirty_page_for_writepage(wbc, page);
		unlock_page(page);
		return 0;
	}
583 584
	return extent_write_full_page(tree, page, btree_get_extent, wbc);
}
585 586 587 588

static int btree_writepages(struct address_space *mapping,
			    struct writeback_control *wbc)
{
589 590
	struct extent_io_tree *tree;
	tree = &BTRFS_I(mapping->host)->io_tree;
591
	if (wbc->sync_mode == WB_SYNC_NONE) {
592 593
		u64 num_dirty;
		u64 start = 0;
594
		unsigned long thresh = 32 * 1024 * 1024;
595 596 597 598

		if (wbc->for_kupdate)
			return 0;

599 600
		num_dirty = count_range_bits(tree, &start, (u64)-1,
					     thresh, EXTENT_DIRTY);
601 602 603 604
		if (num_dirty < thresh) {
			return 0;
		}
	}
605 606 607
	return extent_writepages(tree, mapping, btree_get_extent, wbc);
}

608 609
int btree_readpage(struct file *file, struct page *page)
{
610 611
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
612 613
	return extent_read_full_page(tree, page, btree_get_extent);
}
C
Chris Mason 已提交
614

615
static int btree_releasepage(struct page *page, gfp_t gfp_flags)
616
{
617 618
	struct extent_io_tree *tree;
	struct extent_map_tree *map;
619
	int ret;
620

621 622 623
	if (PageWriteback(page) || PageDirty(page))
	    return 0;

624 625
	tree = &BTRFS_I(page->mapping->host)->io_tree;
	map = &BTRFS_I(page->mapping->host)->extent_tree;
626

627
	ret = try_release_extent_state(map, tree, page, gfp_flags);
628 629 630 631 632
	if (!ret) {
		return 0;
	}

	ret = try_release_extent_buffer(tree, page);
633 634 635 636 637
	if (ret == 1) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
	}
638

639 640 641
	return ret;
}

642
static void btree_invalidatepage(struct page *page, unsigned long offset)
643
{
644 645
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
646 647
	extent_invalidatepage(tree, page, offset);
	btree_releasepage(page, GFP_NOFS);
648
	if (PagePrivate(page)) {
649 650
		printk("warning page private not zero on page %Lu\n",
		       page_offset(page));
651 652 653 654
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
	}
655 656
}

657
#if 0
658
static int btree_writepage(struct page *page, struct writeback_control *wbc)
659
{
C
Chris Mason 已提交
660
	struct buffer_head *bh;
661
	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
C
Chris Mason 已提交
662 663 664 665 666 667 668 669 670 671 672 673
	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);
674
	return block_write_full_page(page, btree_get_block, wbc);
675
}
676
#endif
677

678 679 680
static struct address_space_operations btree_aops = {
	.readpage	= btree_readpage,
	.writepage	= btree_writepage,
681
	.writepages	= btree_writepages,
682 683
	.releasepage	= btree_releasepage,
	.invalidatepage = btree_invalidatepage,
684 685 686
	.sync_page	= block_sync_page,
};

687 688
int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
			 u64 parent_transid)
C
Chris Mason 已提交
689
{
690 691
	struct extent_buffer *buf = NULL;
	struct inode *btree_inode = root->fs_info->btree_inode;
692
	int ret = 0;
C
Chris Mason 已提交
693

694
	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
695
	if (!buf)
C
Chris Mason 已提交
696
		return 0;
697
	read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
698
				 buf, 0, 0, btree_get_extent, 0);
699
	free_extent_buffer(buf);
700
	return ret;
C
Chris Mason 已提交
701 702
}

703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724
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;
}


725 726 727
int btrfs_write_tree_block(struct extent_buffer *buf)
{
	return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start,
728
				      buf->start + buf->len - 1, WB_SYNC_ALL);
729 730 731 732 733 734 735 736
}

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);
}

737
struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
738
				      u32 blocksize, u64 parent_transid)
739 740 741 742 743 744 745 746 747 748 749 750
{
	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;

751
	ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
752 753 754

	if (ret == 0) {
		buf->flags |= EXTENT_UPTODATE;
755 756
	} else {
		WARN_ON(1);
757
	}
758
	return buf;
759

760 761
}

762
int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
763
		     struct extent_buffer *buf)
764
{
765
	struct inode *btree_inode = root->fs_info->btree_inode;
766
	if (btrfs_header_generation(buf) ==
767 768
	    root->fs_info->running_transaction->transid) {
		WARN_ON(!btrfs_tree_locked(buf));
769
		clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
770
					  buf);
771
	}
772 773 774
	return 0;
}

775
static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
776
			u32 stripesize, struct btrfs_root *root,
777
			struct btrfs_fs_info *fs_info,
C
Chris Mason 已提交
778
			u64 objectid)
779
{
C
Chris Mason 已提交
780
	root->node = NULL;
781
	root->inode = NULL;
782
	root->commit_root = NULL;
Y
Yan Zheng 已提交
783
	root->ref_tree = NULL;
784 785 786
	root->sectorsize = sectorsize;
	root->nodesize = nodesize;
	root->leafsize = leafsize;
787
	root->stripesize = stripesize;
C
Chris Mason 已提交
788
	root->ref_cows = 0;
789 790
	root->track_dirty = 0;

791
	root->fs_info = fs_info;
792 793
	root->objectid = objectid;
	root->last_trans = 0;
C
Chris Mason 已提交
794 795
	root->highest_inode = 0;
	root->last_inode_alloc = 0;
796
	root->name = NULL;
797
	root->in_sysfs = 0;
798 799

	INIT_LIST_HEAD(&root->dirty_list);
800
	INIT_LIST_HEAD(&root->orphan_list);
801
	INIT_LIST_HEAD(&root->dead_list);
802
	spin_lock_init(&root->node_lock);
803
	spin_lock_init(&root->list_lock);
804
	mutex_init(&root->objectid_mutex);
805
	mutex_init(&root->log_mutex);
806 807
	extent_io_tree_init(&root->dirty_log_pages,
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
C
Chris Mason 已提交
808 809 810 811

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

812 813
	memset(&root->root_key, 0, sizeof(root->root_key));
	memset(&root->root_item, 0, sizeof(root->root_item));
814
	memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
815
	memset(&root->root_kobj, 0, sizeof(root->root_kobj));
816
	root->defrag_trans_start = fs_info->generation;
817
	init_completion(&root->kobj_unregister);
818 819
	root->defrag_running = 0;
	root->defrag_level = 0;
820
	root->root_key.objectid = objectid;
821 822 823
	return 0;
}

824
static int find_and_setup_root(struct btrfs_root *tree_root,
825 826
			       struct btrfs_fs_info *fs_info,
			       u64 objectid,
C
Chris Mason 已提交
827
			       struct btrfs_root *root)
828 829
{
	int ret;
830
	u32 blocksize;
831

832
	__setup_root(tree_root->nodesize, tree_root->leafsize,
833 834
		     tree_root->sectorsize, tree_root->stripesize,
		     root, fs_info, objectid);
835 836 837 838
	ret = btrfs_find_last_root(tree_root, objectid,
				   &root->root_item, &root->root_key);
	BUG_ON(ret);

839 840
	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
841
				     blocksize, 0);
842
	BUG_ON(!root->node);
843 844 845
	return 0;
}

846 847 848 849
int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
			     struct btrfs_fs_info *fs_info)
{
	struct extent_buffer *eb;
850 851 852
	struct btrfs_root *log_root_tree = fs_info->log_root_tree;
	u64 start = 0;
	u64 end = 0;
853 854
	int ret;

855
	if (!log_root_tree)
856 857
		return 0;

858 859 860 861 862 863 864 865 866
	while(1) {
		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);
	}
867 868 869 870 871
	eb = fs_info->log_root_tree->node;

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

872 873
	ret = btrfs_free_reserved_extent(fs_info->tree_root,
				eb->start, eb->len);
874 875 876 877 878 879 880 881 882 883
	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)
884 885 886
{
	struct btrfs_root *root;
	struct btrfs_root *tree_root = fs_info->tree_root;
887 888 889 890 891 892 893 894 895 896 897 898 899 900 901

	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,
Z
Zheng Yan 已提交
902 903
					    0, BTRFS_TREE_LOG_OBJECTID,
					    trans->transid, 0, 0, 0);
904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924

	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;
925
	struct btrfs_path *path;
926
	struct extent_buffer *l;
C
Chris Mason 已提交
927
	u64 highest_inode;
928
	u32 blocksize;
929 930
	int ret = 0;

931
	root = kzalloc(sizeof(*root), GFP_NOFS);
C
Chris Mason 已提交
932
	if (!root)
933 934
		return ERR_PTR(-ENOMEM);
	if (location->offset == (u64)-1) {
935
		ret = find_and_setup_root(tree_root, fs_info,
936 937 938 939 940 941 942 943
					  location->objectid, root);
		if (ret) {
			kfree(root);
			return ERR_PTR(ret);
		}
		goto insert;
	}

944
	__setup_root(tree_root->nodesize, tree_root->leafsize,
945 946
		     tree_root->sectorsize, tree_root->stripesize,
		     root, fs_info, location->objectid);
947 948 949 950 951 952 953 954 955

	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;
	}
956 957 958
	l = path->nodes[0];
	read_extent_buffer(l, &root->root_item,
	       btrfs_item_ptr_offset(l, path->slots[0]),
959
	       sizeof(root->root_item));
960
	memcpy(&root->root_key, location, sizeof(*location));
961 962 963 964 965 966 967 968
	ret = 0;
out:
	btrfs_release_path(root, path);
	btrfs_free_path(path);
	if (ret) {
		kfree(root);
		return ERR_PTR(ret);
	}
969 970
	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
971
				     blocksize, 0);
972 973
	BUG_ON(!root->node);
insert:
974 975 976 977 978 979 980
	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;
		}
981 982 983 984
	}
	return root;
}

C
Chris Mason 已提交
985 986 987 988 989 990 991 992 993 994 995 996 997 998 999
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;
}

1000 1001
struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
					      struct btrfs_key *location)
1002 1003 1004 1005
{
	struct btrfs_root *root;
	int ret;

1006 1007 1008 1009
	if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
		return fs_info->tree_root;
	if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
		return fs_info->extent_root;
1010 1011 1012 1013
	if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
		return fs_info->chunk_root;
	if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
		return fs_info->dev_root;
1014

1015 1016 1017 1018 1019
	root = radix_tree_lookup(&fs_info->fs_roots_radix,
				 (unsigned long)location->objectid);
	if (root)
		return root;

1020
	root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
1021 1022
	if (IS_ERR(root))
		return root;
C
Chris Mason 已提交
1023 1024
	ret = radix_tree_insert(&fs_info->fs_roots_radix,
				(unsigned long)root->root_key.objectid,
1025 1026
				root);
	if (ret) {
1027
		free_extent_buffer(root->node);
1028 1029 1030
		kfree(root);
		return ERR_PTR(ret);
	}
1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047
	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;
1048

1049 1050 1051
	if (root->in_sysfs)
		return root;

1052 1053
	ret = btrfs_set_root_name(root, name, namelen);
	if (ret) {
1054
		free_extent_buffer(root->node);
1055 1056 1057 1058 1059 1060
		kfree(root);
		return ERR_PTR(ret);
	}

	ret = btrfs_sysfs_add_root(root);
	if (ret) {
1061
		free_extent_buffer(root->node);
1062 1063 1064 1065
		kfree(root->name);
		kfree(root);
		return ERR_PTR(ret);
	}
1066
	root->in_sysfs = 1;
1067 1068
	return root;
}
1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086
#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
1087 1088 1089 1090 1091 1092 1093 1094 1095

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;

1096
	if ((bdi_bits & (1 << BDI_write_congested)) &&
1097
	    btrfs_congested_async(info, 0))
1098 1099
		return 1;

C
Chris Mason 已提交
1100 1101
	list_for_each(cur, &info->fs_devices->devices) {
		device = list_entry(cur, struct btrfs_device, dev_list);
1102 1103
		if (!device->bdev)
			continue;
C
Chris Mason 已提交
1104 1105 1106 1107 1108 1109 1110 1111 1112
		bdi = blk_get_backing_dev_info(device->bdev);
		if (bdi && bdi_congested(bdi, bdi_bits)) {
			ret = 1;
			break;
		}
	}
	return ret;
}

1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132
/*
 * 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
1133 1134
void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
{
1135
	struct inode *inode;
1136 1137
	struct extent_map_tree *em_tree;
	struct extent_map *em;
1138
	struct address_space *mapping;
1139 1140
	u64 offset;

1141
	/* the generic O_DIRECT read code does this */
1142 1143 1144 1145 1146
	if (!page) {
		__unplug_io_fn(bdi, page);
		return;
	}

1147 1148 1149 1150 1151 1152 1153 1154 1155 1156
	/*
	 * 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;
1157
	offset = page_offset(page);
C
Chris Mason 已提交
1158

1159 1160 1161 1162
	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);
1163 1164
	if (!em) {
		__unplug_io_fn(bdi, page);
1165
		return;
1166
	}
1167

1168 1169 1170 1171 1172
	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
		free_extent_map(em);
		__unplug_io_fn(bdi, page);
		return;
	}
1173 1174 1175 1176
	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
1177 1178 1179 1180 1181
}

static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
{
	bdi_init(bdi);
1182
	bdi->ra_pages	= default_backing_dev_info.ra_pages;
C
Chris Mason 已提交
1183 1184 1185 1186 1187 1188 1189 1190 1191
	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;
}

1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230
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;
}

1231 1232 1233 1234 1235
/*
 * 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)
1236 1237
{
	struct bio *bio;
1238 1239
	struct end_io_wq *end_io_wq;
	struct btrfs_fs_info *fs_info;
1240 1241
	int error;

1242 1243 1244
	end_io_wq = container_of(work, struct end_io_wq, work);
	bio = end_io_wq->bio;
	fs_info = end_io_wq->info;
1245

1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260
	/* 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);
	bio_endio(bio, error);
1261 1262
}

1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308
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);

1309 1310 1311 1312
		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
1313

1314 1315 1316 1317 1318 1319
		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
1320

1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346
		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;
}

1347
struct btrfs_root *open_ctree(struct super_block *sb,
1348 1349
			      struct btrfs_fs_devices *fs_devices,
			      char *options)
1350
{
1351 1352 1353 1354
	u32 sectorsize;
	u32 nodesize;
	u32 leafsize;
	u32 blocksize;
1355
	u32 stripesize;
1356
	struct buffer_head *bh;
1357
	struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
C
Chris Mason 已提交
1358
						 GFP_NOFS);
1359
	struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root),
C
Chris Mason 已提交
1360
					       GFP_NOFS);
1361
	struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
C
Chris Mason 已提交
1362
						GFP_NOFS);
1363
	struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
1364
						GFP_NOFS);
1365
	struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
1366
					      GFP_NOFS);
1367 1368
	struct btrfs_root *log_tree_root;

1369
	int ret;
1370
	int err = -EINVAL;
1371

C
Chris Mason 已提交
1372
	struct btrfs_super_block *disk_super;
1373

C
Chris Mason 已提交
1374 1375 1376 1377
	if (!extent_root || !tree_root || !fs_info) {
		err = -ENOMEM;
		goto fail;
	}
1378
	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
C
Chris Mason 已提交
1379
	INIT_LIST_HEAD(&fs_info->trans_list);
1380
	INIT_LIST_HEAD(&fs_info->dead_roots);
1381
	INIT_LIST_HEAD(&fs_info->hashers);
1382
	INIT_LIST_HEAD(&fs_info->delalloc_inodes);
1383
	spin_lock_init(&fs_info->hash_lock);
1384
	spin_lock_init(&fs_info->delalloc_lock);
1385
	spin_lock_init(&fs_info->new_trans_lock);
Y
Yan Zheng 已提交
1386
	spin_lock_init(&fs_info->ref_cache_lock);
1387

1388
	init_completion(&fs_info->kobj_unregister);
1389 1390
	fs_info->tree_root = tree_root;
	fs_info->extent_root = extent_root;
1391 1392
	fs_info->chunk_root = chunk_root;
	fs_info->dev_root = dev_root;
1393
	fs_info->fs_devices = fs_devices;
1394
	INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
1395
	INIT_LIST_HEAD(&fs_info->space_info);
1396
	btrfs_mapping_init(&fs_info->mapping_tree);
1397
	atomic_set(&fs_info->nr_async_submits, 0);
1398
	atomic_set(&fs_info->async_submit_draining, 0);
1399
	atomic_set(&fs_info->nr_async_bios, 0);
1400
	atomic_set(&fs_info->throttles, 0);
1401
	atomic_set(&fs_info->throttle_gen, 0);
C
Chris Mason 已提交
1402
	fs_info->sb = sb;
1403
	fs_info->max_extent = (u64)-1;
1404
	fs_info->max_inline = 8192 * 1024;
C
Chris Mason 已提交
1405
	setup_bdi(fs_info, &fs_info->bdi);
1406 1407
	fs_info->btree_inode = new_inode(sb);
	fs_info->btree_inode->i_ino = 1;
C
Chris Mason 已提交
1408
	fs_info->btree_inode->i_nlink = 1;
1409
	fs_info->thread_pool_size = min(num_online_cpus() + 2, 8);
1410

1411 1412 1413
	INIT_LIST_HEAD(&fs_info->ordered_extents);
	spin_lock_init(&fs_info->ordered_extent_lock);

1414 1415 1416
	sb->s_blocksize = 4096;
	sb->s_blocksize_bits = blksize_bits(4096);

1417 1418 1419 1420 1421 1422
	/*
	 * 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;
1423
	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
C
Chris Mason 已提交
1424 1425
	fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;

1426
	extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
1427 1428
			     fs_info->btree_inode->i_mapping,
			     GFP_NOFS);
1429 1430 1431 1432
	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;
1433

J
Josef Bacik 已提交
1434 1435 1436
	spin_lock_init(&fs_info->block_group_cache_lock);
	fs_info->block_group_cache_tree.rb_node = NULL;

1437
	extent_io_tree_init(&fs_info->pinned_extents,
1438
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
1439
	extent_io_tree_init(&fs_info->pending_del,
1440
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
1441
	extent_io_tree_init(&fs_info->extent_ins,
1442
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
1443
	fs_info->do_barriers = 1;
1444

Z
Zheng Yan 已提交
1445 1446 1447 1448
	extent_io_tree_init(&fs_info->reloc_mapping_tree,
			    fs_info->btree_inode->i_mapping, GFP_NOFS);
	INIT_LIST_HEAD(&fs_info->dead_reloc_roots);
	btrfs_leaf_ref_tree_init(&fs_info->reloc_ref_tree);
Z
Zheng Yan 已提交
1449 1450
	btrfs_leaf_ref_tree_init(&fs_info->shared_ref_tree);

1451 1452 1453
	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
1454
	insert_inode_hash(fs_info->btree_inode);
C
Chris Mason 已提交
1455

C
Chris Mason 已提交
1456
	mutex_init(&fs_info->trans_mutex);
1457
	mutex_init(&fs_info->tree_log_mutex);
1458
	mutex_init(&fs_info->drop_mutex);
1459 1460
	mutex_init(&fs_info->alloc_mutex);
	mutex_init(&fs_info->chunk_mutex);
1461 1462
	mutex_init(&fs_info->transaction_kthread_mutex);
	mutex_init(&fs_info->cleaner_mutex);
1463
	mutex_init(&fs_info->volume_mutex);
Z
Zheng Yan 已提交
1464
	mutex_init(&fs_info->tree_reloc_mutex);
1465
	init_waitqueue_head(&fs_info->transaction_throttle);
1466
	init_waitqueue_head(&fs_info->transaction_wait);
1467
	init_waitqueue_head(&fs_info->async_submit_wait);
1468 1469 1470 1471
	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;
1472

1473 1474 1475 1476 1477 1478 1479 1480
#if 0
	ret = add_hasher(fs_info, "crc32c");
	if (ret) {
		printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
		err = -ENOMEM;
		goto fail_iput;
	}
#endif
1481
	__setup_root(4096, 4096, 4096, 4096, tree_root,
C
Chris Mason 已提交
1482
		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
1483

1484

1485 1486 1487
	bh = __bread(fs_devices->latest_bdev,
		     BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
	if (!bh)
C
Chris Mason 已提交
1488 1489
		goto fail_iput;

1490 1491
	memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy));
	brelse(bh);
1492

1493
	memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE);
1494

1495
	disk_super = &fs_info->super_copy;
1496
	if (!btrfs_super_root(disk_super))
C
Chris Mason 已提交
1497
		goto fail_sb_buffer;
1498

1499 1500 1501
	err = btrfs_parse_options(tree_root, options);
	if (err)
		goto fail_sb_buffer;
1502

1503 1504 1505 1506 1507
	/*
	 * 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.
	 */
1508 1509 1510
	btrfs_init_workers(&fs_info->workers, "worker",
			   fs_info->thread_pool_size);
	btrfs_init_workers(&fs_info->submit_workers, "submit",
1511 1512
			   min_t(u64, fs_devices->num_devices,
			   fs_info->thread_pool_size));
1513 1514 1515 1516 1517 1518

	/* 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;
1519 1520 1521 1522 1523 1524 1525 1526

	/* 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;
1527

1528 1529 1530 1531
	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",
1532
			   fs_info->thread_pool_size);
1533 1534 1535 1536 1537 1538

	/*
	 * endios are largely parallel and should have a very
	 * low idle thresh
	 */
	fs_info->endio_workers.idle_thresh = 4;
1539
	fs_info->endio_write_workers.idle_thresh = 64;
1540

1541
	btrfs_start_workers(&fs_info->workers, 1);
1542
	btrfs_start_workers(&fs_info->submit_workers, 1);
1543
	btrfs_start_workers(&fs_info->fixup_workers, 1);
1544
	btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size);
1545 1546
	btrfs_start_workers(&fs_info->endio_write_workers,
			    fs_info->thread_pool_size);
1547

1548
	err = -EINVAL;
1549
	if (btrfs_super_num_devices(disk_super) > fs_devices->open_devices) {
1550 1551
		printk("Btrfs: wanted %llu devices, but found %llu\n",
		       (unsigned long long)btrfs_super_num_devices(disk_super),
1552
		       (unsigned long long)fs_devices->open_devices);
1553 1554 1555 1556 1557
		if (btrfs_test_opt(tree_root, DEGRADED))
			printk("continuing in degraded mode\n");
		else {
			goto fail_sb_buffer;
		}
1558
	}
1559

1560 1561
	fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);

1562 1563 1564
	nodesize = btrfs_super_nodesize(disk_super);
	leafsize = btrfs_super_leafsize(disk_super);
	sectorsize = btrfs_super_sectorsize(disk_super);
1565
	stripesize = btrfs_super_stripesize(disk_super);
1566 1567 1568
	tree_root->nodesize = nodesize;
	tree_root->leafsize = leafsize;
	tree_root->sectorsize = sectorsize;
1569
	tree_root->stripesize = stripesize;
1570 1571 1572

	sb->s_blocksize = sectorsize;
	sb->s_blocksize_bits = blksize_bits(sectorsize);
1573

C
Chris Mason 已提交
1574 1575 1576 1577 1578
	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;
	}
1579

1580
	mutex_lock(&fs_info->chunk_mutex);
1581
	ret = btrfs_read_sys_array(tree_root);
1582
	mutex_unlock(&fs_info->chunk_mutex);
1583 1584 1585 1586 1587
	if (ret) {
		printk("btrfs: failed to read the system array on %s\n",
		       sb->s_id);
		goto fail_sys_array;
	}
1588 1589 1590 1591 1592 1593 1594 1595 1596

	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),
1597
					   blocksize, 0);
1598 1599
	BUG_ON(!chunk_root->node);

1600 1601 1602 1603
	read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
	         (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
		 BTRFS_UUID_SIZE);

1604
	mutex_lock(&fs_info->chunk_mutex);
1605
	ret = btrfs_read_chunk_tree(chunk_root);
1606
	mutex_unlock(&fs_info->chunk_mutex);
1607 1608
	BUG_ON(ret);

1609 1610
	btrfs_close_extra_devices(fs_devices);

1611 1612
	blocksize = btrfs_level_size(tree_root,
				     btrfs_super_root_level(disk_super));
1613

1614

C
Chris Mason 已提交
1615
	tree_root->node = read_tree_block(tree_root,
1616
					  btrfs_super_root(disk_super),
1617
					  blocksize, 0);
C
Chris Mason 已提交
1618 1619
	if (!tree_root->node)
		goto fail_sb_buffer;
1620

1621 1622

	ret = find_and_setup_root(tree_root, fs_info,
C
Chris Mason 已提交
1623
				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
1624
	if (ret)
C
Chris Mason 已提交
1625
		goto fail_tree_root;
1626 1627 1628 1629 1630 1631 1632 1633
	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;
1634

C
Chris Mason 已提交
1635 1636
	btrfs_read_block_groups(extent_root);

1637
	fs_info->generation = btrfs_super_generation(disk_super) + 1;
C
Chris Mason 已提交
1638 1639 1640
	fs_info->data_alloc_profile = (u64)-1;
	fs_info->metadata_alloc_profile = (u64)-1;
	fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
1641 1642 1643 1644 1645 1646 1647 1648 1649
	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)
1650
		goto fail_cleaner;
1651

1652 1653 1654 1655 1656 1657 1658
	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
1659

1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670
		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);
	}
Z
Zheng Yan 已提交
1671 1672 1673 1674

	ret = btrfs_cleanup_reloc_trees(tree_root);
	BUG_ON(ret);

1675
	fs_info->last_trans_committed = btrfs_super_generation(disk_super);
1676
	return tree_root;
C
Chris Mason 已提交
1677

1678
fail_cleaner:
1679
	kthread_stop(fs_info->cleaner_kthread);
1680 1681
fail_extent_root:
	free_extent_buffer(extent_root->node);
C
Chris Mason 已提交
1682
fail_tree_root:
1683
	free_extent_buffer(tree_root->node);
1684
fail_sys_array:
C
Chris Mason 已提交
1685
fail_sb_buffer:
1686
	btrfs_stop_workers(&fs_info->fixup_workers);
1687 1688
	btrfs_stop_workers(&fs_info->workers);
	btrfs_stop_workers(&fs_info->endio_workers);
1689
	btrfs_stop_workers(&fs_info->endio_write_workers);
1690
	btrfs_stop_workers(&fs_info->submit_workers);
1691 1692
fail_iput:
	iput(fs_info->btree_inode);
C
Chris Mason 已提交
1693
fail:
1694
	btrfs_close_devices(fs_info->fs_devices);
1695 1696
	btrfs_mapping_tree_free(&fs_info->mapping_tree);

C
Chris Mason 已提交
1697 1698
	kfree(extent_root);
	kfree(tree_root);
1699
	bdi_destroy(&fs_info->bdi);
C
Chris Mason 已提交
1700 1701
	kfree(fs_info);
	return ERR_PTR(err);
1702 1703
}

1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715
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));
		}
1716 1717 1718
		/* note, we dont' set_buffer_write_io_error because we have
		 * our own ways of dealing with the IO errors
		 */
1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729
		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;
1730
	struct btrfs_super_block *sb;
1731 1732 1733 1734
	struct btrfs_dev_item *dev_item;
	struct buffer_head *bh;
	int ret;
	int do_barriers;
1735 1736
	int max_errors;
	int total_errors = 0;
1737 1738
	u32 crc;
	u64 flags;
1739

1740
	max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
1741 1742
	do_barriers = !btrfs_test_opt(root, NOBARRIER);

1743 1744
	sb = &root->fs_info->super_for_commit;
	dev_item = &sb->dev_item;
1745 1746
	list_for_each(cur, head) {
		dev = list_entry(cur, struct btrfs_device, dev_list);
1747 1748 1749 1750 1751 1752 1753
		if (!dev->bdev) {
			total_errors++;
			continue;
		}
		if (!dev->in_fs_metadata)
			continue;

1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771
		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,
1772 1773
			      BTRFS_SUPER_INFO_SIZE);

1774
		memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795
		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);
		}
1796 1797
		if (ret)
			total_errors++;
1798
	}
1799 1800 1801 1802 1803
	if (total_errors > max_errors) {
		printk("btrfs: %d errors while writing supers\n", total_errors);
		BUG();
	}
	total_errors = 0;
1804 1805 1806

	list_for_each(cur, head) {
		dev = list_entry(cur, struct btrfs_device, dev_list);
1807 1808 1809 1810 1811
		if (!dev->bdev)
			continue;
		if (!dev->in_fs_metadata)
			continue;

1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825
		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);
1826 1827
				if (!buffer_uptodate(bh))
					total_errors++;
1828
			} else {
1829
				total_errors++;
1830 1831 1832 1833 1834 1835
			}

		}
		dev->pending_io = NULL;
		brelse(bh);
	}
1836 1837 1838 1839
	if (total_errors > max_errors) {
		printk("btrfs: %d errors while writing supers\n", total_errors);
		BUG();
	}
1840 1841 1842
	return 0;
}

1843
int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
C
Chris Mason 已提交
1844
		      *root)
1845
{
1846
	int ret;
1847

1848
	ret = write_all_supers(root);
1849
	return ret;
C
Chris Mason 已提交
1850 1851
}

1852
int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
C
Chris Mason 已提交
1853 1854 1855
{
	radix_tree_delete(&fs_info->fs_roots_radix,
			  (unsigned long)root->root_key.objectid);
1856 1857
	if (root->in_sysfs)
		btrfs_sysfs_del_root(root);
C
Chris Mason 已提交
1858 1859 1860
	if (root->inode)
		iput(root->inode);
	if (root->node)
1861
		free_extent_buffer(root->node);
C
Chris Mason 已提交
1862
	if (root->commit_root)
1863
		free_extent_buffer(root->commit_root);
1864 1865
	if (root->name)
		kfree(root->name);
C
Chris Mason 已提交
1866 1867 1868 1869
	kfree(root);
	return 0;
}

C
Chris Mason 已提交
1870
static int del_fs_roots(struct btrfs_fs_info *fs_info)
1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881
{
	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
1882
		for (i = 0; i < ret; i++)
1883
			btrfs_free_fs_root(fs_info, gang[i]);
1884 1885 1886
	}
	return 0;
}
1887

C
Chris Mason 已提交
1888
int close_ctree(struct btrfs_root *root)
C
Chris Mason 已提交
1889
{
1890
	int ret;
1891
	struct btrfs_trans_handle *trans;
1892
	struct btrfs_fs_info *fs_info = root->fs_info;
1893

1894
	fs_info->closing = 1;
1895 1896
	smp_mb();

1897 1898 1899 1900
	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
1901
	trans = btrfs_start_transaction(root, 1);
1902
	ret = btrfs_commit_transaction(trans, root);
C
Chris Mason 已提交
1903 1904 1905 1906
	/* 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);
1907
	BUG_ON(ret);
1908

C
Chris Mason 已提交
1909
	write_ctree_super(NULL, root);
1910

C
Chris Mason 已提交
1911 1912 1913 1914
	if (fs_info->delalloc_bytes) {
		printk("btrfs: at unmount delalloc count %Lu\n",
		       fs_info->delalloc_bytes);
	}
Y
Yan Zheng 已提交
1915 1916 1917 1918
	if (fs_info->total_ref_cache_size) {
		printk("btrfs: at umount reference cache size %Lu\n",
			fs_info->total_ref_cache_size);
	}
1919

1920
	if (fs_info->extent_root->node)
1921
		free_extent_buffer(fs_info->extent_root->node);
1922

1923
	if (fs_info->tree_root->node)
1924
		free_extent_buffer(fs_info->tree_root->node);
1925

1926 1927 1928 1929 1930 1931
	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
1932
	btrfs_free_block_groups(root->fs_info);
1933
	fs_info->closing = 2;
1934
	del_fs_roots(fs_info);
1935 1936 1937

	filemap_write_and_wait(fs_info->btree_inode->i_mapping);

1938
	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
1939

1940
	btrfs_stop_workers(&fs_info->fixup_workers);
1941 1942
	btrfs_stop_workers(&fs_info->workers);
	btrfs_stop_workers(&fs_info->endio_workers);
1943
	btrfs_stop_workers(&fs_info->endio_write_workers);
1944
	btrfs_stop_workers(&fs_info->submit_workers);
1945

1946
	iput(fs_info->btree_inode);
1947 1948 1949 1950 1951 1952 1953 1954 1955 1956
#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
1957
	btrfs_close_devices(fs_info->fs_devices);
1958
	btrfs_mapping_tree_free(&fs_info->mapping_tree);
1959

C
Chris Mason 已提交
1960
	bdi_destroy(&fs_info->bdi);
1961

1962 1963
	kfree(fs_info->extent_root);
	kfree(fs_info->tree_root);
1964 1965
	kfree(fs_info->chunk_root);
	kfree(fs_info->dev_root);
1966 1967 1968
	return 0;
}

1969
int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
1970
{
1971
	int ret;
1972
	struct inode *btree_inode = buf->first_page->mapping->host;
1973 1974 1975 1976 1977 1978 1979 1980

	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;
1981 1982 1983
}

int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
C
Chris Mason 已提交
1984
{
1985
	struct inode *btree_inode = buf->first_page->mapping->host;
1986
	return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
1987 1988
					  buf);
}
1989

1990 1991
void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
{
1992
	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
1993 1994
	u64 transid = btrfs_header_generation(buf);
	struct inode *btree_inode = root->fs_info->btree_inode;
1995

1996
	WARN_ON(!btrfs_tree_locked(buf));
C
Chris Mason 已提交
1997 1998
	if (transid != root->fs_info->generation) {
		printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
1999
			(unsigned long long)buf->start,
C
Chris Mason 已提交
2000 2001 2002
			transid, root->fs_info->generation);
		WARN_ON(1);
	}
2003
	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf);
2004 2005
}

2006
void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
C
Chris Mason 已提交
2007
{
2008 2009 2010 2011
	/*
	 * looks as though older kernels can get into trouble with
	 * this code, they end up stuck in balance_dirty_pages forever
	 */
2012 2013 2014
	struct extent_io_tree *tree;
	u64 num_dirty;
	u64 start = 0;
2015
	unsigned long thresh = 96 * 1024 * 1024;
2016 2017
	tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;

2018
	if (current_is_pdflush() || current->flags & PF_MEMALLOC)
2019 2020 2021 2022 2023 2024
		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
2025
				   root->fs_info->btree_inode->i_mapping, 1);
2026
	}
2027
	return;
C
Chris Mason 已提交
2028
}
2029

2030
int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
2031
{
2032
	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
2033
	int ret;
2034
	ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
2035 2036 2037 2038
	if (ret == 0) {
		buf->flags |= EXTENT_UPTODATE;
	}
	return ret;
2039
}
2040

C
Chris Mason 已提交
2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068
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;
}

2069
static struct extent_io_ops btree_extent_io_ops = {
C
Chris Mason 已提交
2070
	.write_cache_pages_lock_hook = btree_lock_page_hook,
2071
	.readpage_end_io_hook = btree_readpage_end_io_hook,
2072
	.submit_bio_hook = btree_submit_bio_hook,
2073 2074
	/* note we're sharing with inode.c for the merge bio hook */
	.merge_bio_hook = btrfs_merge_bio_hook,
2075
};