disk-io.c 59.3 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
#include <linux/kthread.h>
C
Chris Mason 已提交
29 30
#include <linux/freezer.h>
#include "compat.h"
31
#include "crc32c.h"
32 33
#include "ctree.h"
#include "disk-io.h"
34
#include "transaction.h"
35
#include "btrfs_inode.h"
36
#include "volumes.h"
37
#include "print-tree.h"
38
#include "async-thread.h"
39
#include "locking.h"
C
Chris Mason 已提交
40
#include "ref-cache.h"
41
#include "tree-log.h"
42

43 44
#if 0
static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
45
{
46 47 48 49
	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
50
		return 1;
51
	}
52
	return 0;
53
}
54
#endif
55

56
static struct extent_io_ops btree_extent_io_ops;
57
static void end_workqueue_fn(struct btrfs_work *work);
58

C
Chris Mason 已提交
59 60 61 62 63
/*
 * 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.
 */
64 65 66 67 68 69
struct end_io_wq {
	struct bio *bio;
	bio_end_io_t *end_io;
	void *private;
	struct btrfs_fs_info *info;
	int error;
70
	int metadata;
71
	struct list_head list;
72
	struct btrfs_work work;
73
};
74

C
Chris Mason 已提交
75 76 77 78 79
/*
 * 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.
 */
80 81 82 83
struct async_submit_bio {
	struct inode *inode;
	struct bio *bio;
	struct list_head list;
C
Chris Mason 已提交
84 85
	extent_submit_bio_hook_t *submit_bio_start;
	extent_submit_bio_hook_t *submit_bio_done;
86 87
	int rw;
	int mirror_num;
C
Chris Mason 已提交
88
	unsigned long bio_flags;
89
	struct btrfs_work work;
90 91
};

C
Chris Mason 已提交
92 93 94 95
/*
 * extents on the btree inode are pretty simple, there's one extent
 * that covers the entire device
 */
96 97 98
static struct extent_map *btree_get_extent(struct inode *inode,
		struct page *page, size_t page_offset, u64 start, u64 len,
		int create)
99
{
100 101 102 103
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct extent_map *em;
	int ret;

104 105
	spin_lock(&em_tree->lock);
	em = lookup_extent_mapping(em_tree, start, len);
106 107 108 109
	if (em) {
		em->bdev =
			BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
		spin_unlock(&em_tree->lock);
110
		goto out;
111 112
	}
	spin_unlock(&em_tree->lock);
113

114 115 116 117 118 119
	em = alloc_extent_map(GFP_NOFS);
	if (!em) {
		em = ERR_PTR(-ENOMEM);
		goto out;
	}
	em->start = 0;
120
	em->len = (u64)-1;
C
Chris Mason 已提交
121
	em->block_len = (u64)-1;
122
	em->block_start = 0;
123
	em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
124 125

	spin_lock(&em_tree->lock);
126 127
	ret = add_extent_mapping(em_tree, em);
	if (ret == -EEXIST) {
128 129 130 131 132
		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);
133
		free_extent_map(em);
134
		em = lookup_extent_mapping(em_tree, start, len);
135 136 137
		if (em) {
			printk("after failing, found %Lu %Lu %Lu\n",
			       em->start, em->len, em->block_start);
138
			ret = 0;
139 140 141 142 143 144 145 146 147
		} 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);
			}
148
			ret = -EIO;
149
		}
150
	} else if (ret) {
151 152
		free_extent_map(em);
		em = NULL;
153
	}
154 155 156 157
	spin_unlock(&em_tree->lock);

	if (ret)
		em = ERR_PTR(ret);
158 159
out:
	return em;
160 161
}

162 163
u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
{
164
	return btrfs_crc32c(seed, data, len);
165 166 167 168 169 170 171
}

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

C
Chris Mason 已提交
172 173 174 175
/*
 * compute the csum for a btree block, and either verify it or write it
 * into the csum field of the block.
 */
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209
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) {
210
		/* FIXME, this is not good */
211
		if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) {
212 213 214 215 216 217
			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
218
			       "wanted %X found %X level %d\n",
219
			       root->fs_info->sb->s_id,
C
Chris Mason 已提交
220
			       buf->start, val, found, btrfs_header_level(buf));
221 222 223 224 225 226 227 228
			return 1;
		}
	} else {
		write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE);
	}
	return 0;
}

C
Chris Mason 已提交
229 230 231 232 233 234
/*
 * 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.
 */
235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254
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
255
out:
256 257 258 259 260
	unlock_extent(io_tree, eb->start, eb->start + eb->len - 1,
		      GFP_NOFS);
	return ret;
}

C
Chris Mason 已提交
261 262 263 264
/*
 * helper to read a given tree block, doing retries as required when
 * the checksums don't match and we have alternate mirrors to try.
 */
265 266
static int btree_read_extent_buffer_pages(struct btrfs_root *root,
					  struct extent_buffer *eb,
267
					  u64 start, u64 parent_transid)
268 269 270 271 272 273 274 275 276 277
{
	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);
278 279
		if (!ret &&
		    !verify_parent_transid(io_tree, eb, parent_transid))
280
			return ret;
281
printk("read extent buffer pages failed with ret %d mirror no %d\n", ret, mirror_num);
282 283
		num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
					      eb->start, eb->len);
C
Chris Mason 已提交
284
		if (num_copies == 1)
285
			return ret;
C
Chris Mason 已提交
286

287
		mirror_num++;
C
Chris Mason 已提交
288
		if (mirror_num > num_copies)
289 290 291 292
			return ret;
	}
	return -EIO;
}
293

C
Chris Mason 已提交
294 295 296 297
/*
 * 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
 */
298
static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
299
{
300
	struct extent_io_tree *tree;
301
	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
302 303 304 305
	u64 found_start;
	int found_level;
	unsigned long len;
	struct extent_buffer *eb;
306 307
	int ret;

308
	tree = &BTRFS_I(page->mapping->host)->io_tree;
309 310 311 312 313 314 315 316 317 318

	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);
319 320
	ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
					     btrfs_header_generation(eb));
321
	BUG_ON(ret);
322 323 324 325
	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);
326 327 328 329 330 331 332 333 334 335 336 337 338
		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;
339 340
	}
	found_level = btrfs_header_level(eb);
C
Chris Mason 已提交
341

342
	csum_tree_block(root, eb, 0);
343
err:
344 345 346 347 348
	free_extent_buffer(eb);
out:
	return 0;
}

Y
Yan Zheng 已提交
349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367
static int check_tree_block_fsid(struct btrfs_root *root,
				 struct extent_buffer *eb)
{
	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
	u8 fsid[BTRFS_UUID_SIZE];
	int ret = 1;

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

368
static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
369 370 371 372 373 374 375 376
			       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;
377
	int ret = 0;
378 379 380 381 382 383 384 385 386 387 388

	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);
389

390
	found_start = btrfs_header_bytenr(eb);
391
	if (found_start != start) {
392 393 394
		printk("bad tree block start %llu %llu\n",
		       (unsigned long long)found_start,
		       (unsigned long long)eb->start);
395
		ret = -EIO;
396 397 398 399 400 401
		goto err;
	}
	if (eb->first_page != page) {
		printk("bad first page %lu %lu\n", eb->first_page->index,
		       page->index);
		WARN_ON(1);
402
		ret = -EIO;
403 404
		goto err;
	}
Y
Yan Zheng 已提交
405
	if (check_tree_block_fsid(root, eb)) {
406 407 408 409
		printk("bad fsid on block %Lu\n", eb->start);
		ret = -EIO;
		goto err;
	}
410 411 412
	found_level = btrfs_header_level(eb);

	ret = csum_tree_block(root, eb, 1);
413 414
	if (ret)
		ret = -EIO;
415 416 417 418 419 420

	end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
	end = eb->start + end - 1;
err:
	free_extent_buffer(eb);
out:
421
	return ret;
422 423 424 425 426 427 428 429 430
}

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;
431 432
	end_io_wq->work.func = end_workqueue_fn;
	end_io_wq->work.flags = 0;
433 434 435 436 437
	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);
438 439
}

440 441
int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
			int metadata)
442
{
443 444 445 446 447 448 449
	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;
450
	end_io_wq->info = info;
451 452
	end_io_wq->error = 0;
	end_io_wq->bio = bio;
453
	end_io_wq->metadata = metadata;
454 455 456

	bio->bi_private = end_io_wq;
	bio->bi_end_io = end_workqueue_bio;
457 458 459
	return 0;
}

460
unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
461
{
462 463 464 465 466
	unsigned long limit = min_t(unsigned long,
				    info->workers.max_workers,
				    info->fs_devices->open_devices);
	return 256 * limit;
}
467

468 469
int btrfs_congested_async(struct btrfs_fs_info *info, int iodone)
{
470 471
	return atomic_read(&info->nr_async_bios) >
		btrfs_async_submit_limit(info);
472 473
}

C
Chris Mason 已提交
474 475 476 477 478 479 480 481 482 483 484 485
static void run_one_async_start(struct btrfs_work *work)
{
	struct btrfs_fs_info *fs_info;
	struct async_submit_bio *async;

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

static void run_one_async_done(struct btrfs_work *work)
486 487 488
{
	struct btrfs_fs_info *fs_info;
	struct async_submit_bio *async;
489
	int limit;
490 491 492

	async = container_of(work, struct  async_submit_bio, work);
	fs_info = BTRFS_I(async->inode)->root->fs_info;
493

494
	limit = btrfs_async_submit_limit(fs_info);
495 496
	limit = limit * 2 / 3;

497
	atomic_dec(&fs_info->nr_async_submits);
498

499 500
	if (atomic_read(&fs_info->nr_async_submits) < limit &&
	    waitqueue_active(&fs_info->async_submit_wait))
501 502
		wake_up(&fs_info->async_submit_wait);

C
Chris Mason 已提交
503
	async->submit_bio_done(async->inode, async->rw, async->bio,
C
Chris Mason 已提交
504
			       async->mirror_num, async->bio_flags);
C
Chris Mason 已提交
505 506 507 508 509 510 511
}

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

	async = container_of(work, struct  async_submit_bio, work);
512 513 514
	kfree(async);
}

515 516
int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
			int rw, struct bio *bio, int mirror_num,
C
Chris Mason 已提交
517
			unsigned long bio_flags,
C
Chris Mason 已提交
518 519
			extent_submit_bio_hook_t *submit_bio_start,
			extent_submit_bio_hook_t *submit_bio_done)
520 521 522 523 524 525 526 527 528 529 530
{
	struct async_submit_bio *async;

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

	async->inode = inode;
	async->rw = rw;
	async->bio = bio;
	async->mirror_num = mirror_num;
C
Chris Mason 已提交
531 532 533 534 535 536 537
	async->submit_bio_start = submit_bio_start;
	async->submit_bio_done = submit_bio_done;

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

538
	async->work.flags = 0;
C
Chris Mason 已提交
539
	async->bio_flags = bio_flags;
540

541
	atomic_inc(&fs_info->nr_async_submits);
542
	btrfs_queue_worker(&fs_info->workers, &async->work);
C
Chris Mason 已提交
543
#if 0
544
	int limit = btrfs_async_submit_limit(fs_info);
545 546
	if (atomic_read(&fs_info->nr_async_submits) > limit) {
		wait_event_timeout(fs_info->async_submit_wait,
547 548
			   (atomic_read(&fs_info->nr_async_submits) < limit),
			   HZ/10);
549 550 551 552 553

		wait_event_timeout(fs_info->async_submit_wait,
			   (atomic_read(&fs_info->nr_async_bios) < limit),
			   HZ/10);
	}
C
Chris Mason 已提交
554
#endif
555 556 557 558 559 560
	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));
	}

561 562 563
	return 0;
}

564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579
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;
}

C
Chris Mason 已提交
580 581 582
static int __btree_submit_bio_start(struct inode *inode, int rw,
				    struct bio *bio, int mirror_num,
				    unsigned long bio_flags)
583
{
584 585
	/*
	 * when we're called for a write, we're already in the async
586
	 * submission context.  Just jump into btrfs_map_bio
587
	 */
C
Chris Mason 已提交
588 589 590
	btree_csum_one_bio(bio);
	return 0;
}
591

C
Chris Mason 已提交
592 593 594
static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
				 int mirror_num, unsigned long bio_flags)
{
595
	/*
C
Chris Mason 已提交
596 597
	 * when we're called for a write, we're already in the async
	 * submission context.  Just jump into btrfs_map_bio
598 599
	 */
	return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
600 601
}

602
static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
C
Chris Mason 已提交
603
				 int mirror_num, unsigned long bio_flags)
604
{
605 606 607 608
	/*
	 * kthread helpers are used to submit writes so that checksumming
	 * can happen in parallel across all CPUs
	 */
609
	if (!(rw & (1 << BIO_RW))) {
C
Chris Mason 已提交
610 611 612 613 614 615 616 617 618 619
		int ret;
		/*
		 * called for a read, do the setup so that checksum validation
		 * can happen in the async kernel threads
		 */
		ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
					  bio, 1);
		BUG_ON(ret);

		return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
620
				     mirror_num, 0);
621 622
	}
	return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
C
Chris Mason 已提交
623
				   inode, rw, bio, mirror_num, 0,
C
Chris Mason 已提交
624 625
				   __btree_submit_bio_start,
				   __btree_submit_bio_done);
626 627
}

628 629
static int btree_writepage(struct page *page, struct writeback_control *wbc)
{
630 631
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
632 633 634 635 636 637

	if (current->flags & PF_MEMALLOC) {
		redirty_page_for_writepage(wbc, page);
		unlock_page(page);
		return 0;
	}
638 639
	return extent_write_full_page(tree, page, btree_get_extent, wbc);
}
640 641 642 643

static int btree_writepages(struct address_space *mapping,
			    struct writeback_control *wbc)
{
644 645
	struct extent_io_tree *tree;
	tree = &BTRFS_I(mapping->host)->io_tree;
646
	if (wbc->sync_mode == WB_SYNC_NONE) {
647 648
		u64 num_dirty;
		u64 start = 0;
649
		unsigned long thresh = 32 * 1024 * 1024;
650 651 652 653

		if (wbc->for_kupdate)
			return 0;

654 655
		num_dirty = count_range_bits(tree, &start, (u64)-1,
					     thresh, EXTENT_DIRTY);
656 657 658 659
		if (num_dirty < thresh) {
			return 0;
		}
	}
660 661 662
	return extent_writepages(tree, mapping, btree_get_extent, wbc);
}

663
static int btree_readpage(struct file *file, struct page *page)
664
{
665 666
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
667 668
	return extent_read_full_page(tree, page, btree_get_extent);
}
C
Chris Mason 已提交
669

670
static int btree_releasepage(struct page *page, gfp_t gfp_flags)
671
{
672 673
	struct extent_io_tree *tree;
	struct extent_map_tree *map;
674
	int ret;
675

676 677 678
	if (PageWriteback(page) || PageDirty(page))
	    return 0;

679 680
	tree = &BTRFS_I(page->mapping->host)->io_tree;
	map = &BTRFS_I(page->mapping->host)->extent_tree;
681

682
	ret = try_release_extent_state(map, tree, page, gfp_flags);
683 684 685 686 687
	if (!ret) {
		return 0;
	}

	ret = try_release_extent_buffer(tree, page);
688 689 690 691 692
	if (ret == 1) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
	}
693

694 695 696
	return ret;
}

697
static void btree_invalidatepage(struct page *page, unsigned long offset)
698
{
699 700
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
701 702
	extent_invalidatepage(tree, page, offset);
	btree_releasepage(page, GFP_NOFS);
703
	if (PagePrivate(page)) {
704 705
		printk("warning page private not zero on page %Lu\n",
		       page_offset(page));
706 707 708 709
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
	}
710 711
}

712
#if 0
713
static int btree_writepage(struct page *page, struct writeback_control *wbc)
714
{
C
Chris Mason 已提交
715
	struct buffer_head *bh;
716
	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
C
Chris Mason 已提交
717 718 719 720 721 722 723 724 725 726 727 728
	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);
729
	return block_write_full_page(page, btree_get_block, wbc);
730
}
731
#endif
732

733 734 735
static struct address_space_operations btree_aops = {
	.readpage	= btree_readpage,
	.writepage	= btree_writepage,
736
	.writepages	= btree_writepages,
737 738
	.releasepage	= btree_releasepage,
	.invalidatepage = btree_invalidatepage,
739 740 741
	.sync_page	= block_sync_page,
};

742 743
int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
			 u64 parent_transid)
C
Chris Mason 已提交
744
{
745 746
	struct extent_buffer *buf = NULL;
	struct inode *btree_inode = root->fs_info->btree_inode;
747
	int ret = 0;
C
Chris Mason 已提交
748

749
	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
750
	if (!buf)
C
Chris Mason 已提交
751
		return 0;
752
	read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
753
				 buf, 0, 0, btree_get_extent, 0);
754
	free_extent_buffer(buf);
755
	return ret;
C
Chris Mason 已提交
756 757
}

758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779
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;
}


780 781 782
int btrfs_write_tree_block(struct extent_buffer *buf)
{
	return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start,
783
				      buf->start + buf->len - 1, WB_SYNC_ALL);
784 785 786 787 788 789 790 791
}

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);
}

792
struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
793
				      u32 blocksize, u64 parent_transid)
794 795 796 797 798 799 800 801 802 803 804 805
{
	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;

806
	ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
807 808 809

	if (ret == 0) {
		buf->flags |= EXTENT_UPTODATE;
810 811
	} else {
		WARN_ON(1);
812
	}
813
	return buf;
814

815 816
}

817
int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
818
		     struct extent_buffer *buf)
819
{
820
	struct inode *btree_inode = root->fs_info->btree_inode;
821
	if (btrfs_header_generation(buf) ==
822 823
	    root->fs_info->running_transaction->transid) {
		WARN_ON(!btrfs_tree_locked(buf));
824
		clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
825
					  buf);
826
	}
827 828 829
	return 0;
}

830
static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
831
			u32 stripesize, struct btrfs_root *root,
832
			struct btrfs_fs_info *fs_info,
C
Chris Mason 已提交
833
			u64 objectid)
834
{
C
Chris Mason 已提交
835
	root->node = NULL;
836
	root->commit_root = NULL;
Y
Yan Zheng 已提交
837
	root->ref_tree = NULL;
838 839 840
	root->sectorsize = sectorsize;
	root->nodesize = nodesize;
	root->leafsize = leafsize;
841
	root->stripesize = stripesize;
C
Chris Mason 已提交
842
	root->ref_cows = 0;
843 844
	root->track_dirty = 0;

845
	root->fs_info = fs_info;
846 847
	root->objectid = objectid;
	root->last_trans = 0;
C
Chris Mason 已提交
848 849
	root->highest_inode = 0;
	root->last_inode_alloc = 0;
850
	root->name = NULL;
851
	root->in_sysfs = 0;
852 853

	INIT_LIST_HEAD(&root->dirty_list);
854
	INIT_LIST_HEAD(&root->orphan_list);
855
	INIT_LIST_HEAD(&root->dead_list);
856
	spin_lock_init(&root->node_lock);
857
	spin_lock_init(&root->list_lock);
858
	mutex_init(&root->objectid_mutex);
859
	mutex_init(&root->log_mutex);
860 861
	extent_io_tree_init(&root->dirty_log_pages,
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
C
Chris Mason 已提交
862 863 864 865

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

866 867
	memset(&root->root_key, 0, sizeof(root->root_key));
	memset(&root->root_item, 0, sizeof(root->root_item));
868
	memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
869
	memset(&root->root_kobj, 0, sizeof(root->root_kobj));
870
	root->defrag_trans_start = fs_info->generation;
871
	init_completion(&root->kobj_unregister);
872 873
	root->defrag_running = 0;
	root->defrag_level = 0;
874
	root->root_key.objectid = objectid;
875 876 877 878 879 880
	root->anon_super.s_root = NULL;
	root->anon_super.s_dev = 0;
	INIT_LIST_HEAD(&root->anon_super.s_list);
	INIT_LIST_HEAD(&root->anon_super.s_instances);
	init_rwsem(&root->anon_super.s_umount);

881 882 883
	return 0;
}

884
static int find_and_setup_root(struct btrfs_root *tree_root,
885 886
			       struct btrfs_fs_info *fs_info,
			       u64 objectid,
C
Chris Mason 已提交
887
			       struct btrfs_root *root)
888 889
{
	int ret;
890
	u32 blocksize;
891
	u64 generation;
892

893
	__setup_root(tree_root->nodesize, tree_root->leafsize,
894 895
		     tree_root->sectorsize, tree_root->stripesize,
		     root, fs_info, objectid);
896 897 898 899
	ret = btrfs_find_last_root(tree_root, objectid,
				   &root->root_item, &root->root_key);
	BUG_ON(ret);

900
	generation = btrfs_root_generation(&root->root_item);
901 902
	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
903
				     blocksize, generation);
904
	BUG_ON(!root->node);
905 906 907
	return 0;
}

908 909 910 911
int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
			     struct btrfs_fs_info *fs_info)
{
	struct extent_buffer *eb;
912 913 914
	struct btrfs_root *log_root_tree = fs_info->log_root_tree;
	u64 start = 0;
	u64 end = 0;
915 916
	int ret;

917
	if (!log_root_tree)
918 919
		return 0;

920 921 922 923 924 925 926 927 928
	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);
	}
929 930 931 932 933
	eb = fs_info->log_root_tree->node;

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

934 935
	ret = btrfs_free_reserved_extent(fs_info->tree_root,
				eb->start, eb->len);
936 937 938 939 940 941 942 943 944 945
	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)
946 947 948
{
	struct btrfs_root *root;
	struct btrfs_root *tree_root = fs_info->tree_root;
949 950 951 952 953 954 955 956 957 958 959 960 961 962 963

	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
964 965
					    0, BTRFS_TREE_LOG_OBJECTID,
					    trans->transid, 0, 0, 0);
966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986

	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;
987
	struct btrfs_path *path;
988
	struct extent_buffer *l;
C
Chris Mason 已提交
989
	u64 highest_inode;
990
	u64 generation;
991
	u32 blocksize;
992 993
	int ret = 0;

994
	root = kzalloc(sizeof(*root), GFP_NOFS);
C
Chris Mason 已提交
995
	if (!root)
996 997
		return ERR_PTR(-ENOMEM);
	if (location->offset == (u64)-1) {
998
		ret = find_and_setup_root(tree_root, fs_info,
999 1000 1001 1002 1003 1004 1005 1006
					  location->objectid, root);
		if (ret) {
			kfree(root);
			return ERR_PTR(ret);
		}
		goto insert;
	}

1007
	__setup_root(tree_root->nodesize, tree_root->leafsize,
1008 1009
		     tree_root->sectorsize, tree_root->stripesize,
		     root, fs_info, location->objectid);
1010 1011 1012 1013 1014 1015 1016 1017 1018

	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;
	}
1019 1020 1021
	l = path->nodes[0];
	read_extent_buffer(l, &root->root_item,
	       btrfs_item_ptr_offset(l, path->slots[0]),
1022
	       sizeof(root->root_item));
1023
	memcpy(&root->root_key, location, sizeof(*location));
1024 1025 1026 1027 1028 1029 1030 1031
	ret = 0;
out:
	btrfs_release_path(root, path);
	btrfs_free_path(path);
	if (ret) {
		kfree(root);
		return ERR_PTR(ret);
	}
1032
	generation = btrfs_root_generation(&root->root_item);
1033 1034
	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
1035
				     blocksize, generation);
1036 1037
	BUG_ON(!root->node);
insert:
1038 1039 1040 1041 1042 1043 1044
	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;
		}
1045 1046 1047 1048
	}
	return root;
}

C
Chris Mason 已提交
1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063
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;
}

1064 1065
struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
					      struct btrfs_key *location)
1066 1067 1068 1069
{
	struct btrfs_root *root;
	int ret;

1070 1071 1072 1073
	if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
		return fs_info->tree_root;
	if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
		return fs_info->extent_root;
1074 1075 1076 1077
	if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
		return fs_info->chunk_root;
	if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
		return fs_info->dev_root;
1078

1079 1080 1081 1082 1083
	root = radix_tree_lookup(&fs_info->fs_roots_radix,
				 (unsigned long)location->objectid);
	if (root)
		return root;

1084
	root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
1085 1086
	if (IS_ERR(root))
		return root;
1087 1088 1089

	set_anon_super(&root->anon_super, NULL);

C
Chris Mason 已提交
1090 1091
	ret = radix_tree_insert(&fs_info->fs_roots_radix,
				(unsigned long)root->root_key.objectid,
1092 1093
				root);
	if (ret) {
1094
		free_extent_buffer(root->node);
1095 1096 1097
		kfree(root);
		return ERR_PTR(ret);
	}
Y
Yan Zheng 已提交
1098 1099 1100 1101 1102 1103
	if (!(fs_info->sb->s_flags & MS_RDONLY)) {
		ret = btrfs_find_dead_roots(fs_info->tree_root,
					    root->root_key.objectid, root);
		BUG_ON(ret);
		btrfs_orphan_cleanup(root);
	}
1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116
	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;
1117

1118 1119 1120
	if (root->in_sysfs)
		return root;

1121 1122
	ret = btrfs_set_root_name(root, name, namelen);
	if (ret) {
1123
		free_extent_buffer(root->node);
1124 1125 1126
		kfree(root);
		return ERR_PTR(ret);
	}
1127
#if 0
1128 1129
	ret = btrfs_sysfs_add_root(root);
	if (ret) {
1130
		free_extent_buffer(root->node);
1131 1132 1133 1134
		kfree(root->name);
		kfree(root);
		return ERR_PTR(ret);
	}
1135
#endif
1136
	root->in_sysfs = 1;
1137 1138
	return root;
}
1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156
#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
1157 1158 1159 1160 1161 1162 1163 1164

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;
1165
#if 0
1166
	if ((bdi_bits & (1 << BDI_write_congested)) &&
1167
	    btrfs_congested_async(info, 0))
1168
		return 1;
1169
#endif
C
Chris Mason 已提交
1170 1171
	list_for_each(cur, &info->fs_devices->devices) {
		device = list_entry(cur, struct btrfs_device, dev_list);
1172 1173
		if (!device->bdev)
			continue;
C
Chris Mason 已提交
1174 1175 1176 1177 1178 1179 1180 1181 1182
		bdi = blk_get_backing_dev_info(device->bdev);
		if (bdi && bdi_congested(bdi, bdi_bits)) {
			ret = 1;
			break;
		}
	}
	return ret;
}

1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202
/*
 * 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);
		}
	}
}

1203
static void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
C
Chris Mason 已提交
1204
{
1205
	struct inode *inode;
1206 1207
	struct extent_map_tree *em_tree;
	struct extent_map *em;
1208
	struct address_space *mapping;
1209 1210
	u64 offset;

1211
	/* the generic O_DIRECT read code does this */
1212
	if (1 || !page) {
1213 1214 1215 1216
		__unplug_io_fn(bdi, page);
		return;
	}

1217 1218 1219 1220 1221 1222 1223 1224 1225 1226
	/*
	 * 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;
1227 1228 1229 1230 1231 1232 1233 1234 1235 1236

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

1237
	offset = page_offset(page);
C
Chris Mason 已提交
1238

1239 1240 1241 1242
	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);
1243 1244
	if (!em) {
		__unplug_io_fn(bdi, page);
1245
		return;
1246
	}
1247

1248 1249 1250 1251 1252
	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
		free_extent_map(em);
		__unplug_io_fn(bdi, page);
		return;
	}
1253 1254 1255 1256
	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
1257 1258 1259 1260 1261
}

static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
{
	bdi_init(bdi);
1262
	bdi->ra_pages	= default_backing_dev_info.ra_pages;
C
Chris Mason 已提交
1263 1264 1265 1266 1267 1268 1269 1270 1271
	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;
}

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 1309 1310
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;
}

1311 1312 1313 1314 1315
/*
 * 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)
1316 1317
{
	struct bio *bio;
1318 1319
	struct end_io_wq *end_io_wq;
	struct btrfs_fs_info *fs_info;
1320 1321
	int error;

1322 1323 1324
	end_io_wq = container_of(work, struct end_io_wq, work);
	bio = end_io_wq->bio;
	fs_info = end_io_wq->info;
1325

1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340
	/* 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);
1341 1342
}

1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388
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);

1389 1390 1391 1392
		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
1393

1394 1395 1396 1397 1398 1399
		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
1400

1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426
		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;
}

1427
struct btrfs_root *open_ctree(struct super_block *sb,
1428 1429
			      struct btrfs_fs_devices *fs_devices,
			      char *options)
1430
{
1431 1432 1433 1434
	u32 sectorsize;
	u32 nodesize;
	u32 leafsize;
	u32 blocksize;
1435
	u32 stripesize;
1436
	u64 generation;
1437
	u64 features;
1438
	struct btrfs_key location;
1439
	struct buffer_head *bh;
1440
	struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
C
Chris Mason 已提交
1441
						 GFP_NOFS);
1442
	struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root),
C
Chris Mason 已提交
1443
					       GFP_NOFS);
1444
	struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
C
Chris Mason 已提交
1445
						GFP_NOFS);
1446
	struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
1447
						GFP_NOFS);
1448
	struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
1449
					      GFP_NOFS);
1450 1451
	struct btrfs_root *log_tree_root;

1452
	int ret;
1453
	int err = -EINVAL;
1454

C
Chris Mason 已提交
1455
	struct btrfs_super_block *disk_super;
1456

1457 1458
	if (!extent_root || !tree_root || !fs_info ||
	    !chunk_root || !dev_root) {
C
Chris Mason 已提交
1459 1460 1461
		err = -ENOMEM;
		goto fail;
	}
1462
	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
C
Chris Mason 已提交
1463
	INIT_LIST_HEAD(&fs_info->trans_list);
1464
	INIT_LIST_HEAD(&fs_info->dead_roots);
1465
	INIT_LIST_HEAD(&fs_info->hashers);
1466
	INIT_LIST_HEAD(&fs_info->delalloc_inodes);
1467
	spin_lock_init(&fs_info->hash_lock);
1468
	spin_lock_init(&fs_info->delalloc_lock);
1469
	spin_lock_init(&fs_info->new_trans_lock);
Y
Yan Zheng 已提交
1470
	spin_lock_init(&fs_info->ref_cache_lock);
1471

1472
	init_completion(&fs_info->kobj_unregister);
1473 1474
	fs_info->tree_root = tree_root;
	fs_info->extent_root = extent_root;
1475 1476
	fs_info->chunk_root = chunk_root;
	fs_info->dev_root = dev_root;
1477
	fs_info->fs_devices = fs_devices;
1478
	INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
1479
	INIT_LIST_HEAD(&fs_info->space_info);
1480
	btrfs_mapping_init(&fs_info->mapping_tree);
1481
	atomic_set(&fs_info->nr_async_submits, 0);
1482
	atomic_set(&fs_info->async_delalloc_pages, 0);
1483
	atomic_set(&fs_info->async_submit_draining, 0);
1484
	atomic_set(&fs_info->nr_async_bios, 0);
1485
	atomic_set(&fs_info->throttles, 0);
1486
	atomic_set(&fs_info->throttle_gen, 0);
C
Chris Mason 已提交
1487
	fs_info->sb = sb;
1488
	fs_info->max_extent = (u64)-1;
1489
	fs_info->max_inline = 8192 * 1024;
C
Chris Mason 已提交
1490
	setup_bdi(fs_info, &fs_info->bdi);
1491 1492
	fs_info->btree_inode = new_inode(sb);
	fs_info->btree_inode->i_ino = 1;
C
Chris Mason 已提交
1493
	fs_info->btree_inode->i_nlink = 1;
C
Chris Mason 已提交
1494

1495
	fs_info->thread_pool_size = min(num_online_cpus() + 2, 8);
1496

1497 1498 1499
	INIT_LIST_HEAD(&fs_info->ordered_extents);
	spin_lock_init(&fs_info->ordered_extent_lock);

1500 1501 1502
	sb->s_blocksize = 4096;
	sb->s_blocksize_bits = blksize_bits(4096);

1503 1504 1505 1506 1507 1508
	/*
	 * 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;
1509
	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
C
Chris Mason 已提交
1510 1511
	fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;

1512
	extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
1513 1514
			     fs_info->btree_inode->i_mapping,
			     GFP_NOFS);
1515 1516 1517 1518
	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;
1519

J
Josef Bacik 已提交
1520 1521 1522
	spin_lock_init(&fs_info->block_group_cache_lock);
	fs_info->block_group_cache_tree.rb_node = NULL;

1523
	extent_io_tree_init(&fs_info->pinned_extents,
1524
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
1525
	extent_io_tree_init(&fs_info->pending_del,
1526
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
1527
	extent_io_tree_init(&fs_info->extent_ins,
1528
			     fs_info->btree_inode->i_mapping, GFP_NOFS);
1529
	fs_info->do_barriers = 1;
1530

Z
Zheng Yan 已提交
1531 1532
	INIT_LIST_HEAD(&fs_info->dead_reloc_roots);
	btrfs_leaf_ref_tree_init(&fs_info->reloc_ref_tree);
Z
Zheng Yan 已提交
1533 1534
	btrfs_leaf_ref_tree_init(&fs_info->shared_ref_tree);

1535 1536 1537
	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
1538
	insert_inode_hash(fs_info->btree_inode);
C
Chris Mason 已提交
1539

C
Chris Mason 已提交
1540
	mutex_init(&fs_info->trans_mutex);
1541
	mutex_init(&fs_info->tree_log_mutex);
1542
	mutex_init(&fs_info->drop_mutex);
1543 1544
	mutex_init(&fs_info->extent_ins_mutex);
	mutex_init(&fs_info->pinned_mutex);
1545
	mutex_init(&fs_info->chunk_mutex);
1546 1547
	mutex_init(&fs_info->transaction_kthread_mutex);
	mutex_init(&fs_info->cleaner_mutex);
1548
	mutex_init(&fs_info->volume_mutex);
Z
Zheng Yan 已提交
1549
	mutex_init(&fs_info->tree_reloc_mutex);
1550
	init_waitqueue_head(&fs_info->transaction_throttle);
1551
	init_waitqueue_head(&fs_info->transaction_wait);
1552
	init_waitqueue_head(&fs_info->async_submit_wait);
1553 1554 1555 1556
	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;
1557

1558 1559 1560 1561 1562 1563 1564 1565
#if 0
	ret = add_hasher(fs_info, "crc32c");
	if (ret) {
		printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
		err = -ENOMEM;
		goto fail_iput;
	}
#endif
1566
	__setup_root(4096, 4096, 4096, 4096, tree_root,
C
Chris Mason 已提交
1567
		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
1568

1569

1570 1571 1572
	bh = __bread(fs_devices->latest_bdev,
		     BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
	if (!bh)
C
Chris Mason 已提交
1573 1574
		goto fail_iput;

1575 1576
	memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy));
	brelse(bh);
1577

1578
	memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE);
1579

1580
	disk_super = &fs_info->super_copy;
1581
	if (!btrfs_super_root(disk_super))
1582
		goto fail_iput;
1583

Y
Yan Zheng 已提交
1584 1585 1586
	ret = btrfs_parse_options(tree_root, options);
	if (ret) {
		err = ret;
1587
		goto fail_iput;
Y
Yan Zheng 已提交
1588
	}
1589

1590 1591 1592 1593 1594 1595 1596
	features = btrfs_super_incompat_flags(disk_super) &
		~BTRFS_FEATURE_INCOMPAT_SUPP;
	if (features) {
		printk(KERN_ERR "BTRFS: couldn't mount because of "
		       "unsupported optional features (%Lx).\n",
		       features);
		err = -EINVAL;
1597
		goto fail_iput;
1598 1599 1600 1601 1602 1603 1604 1605 1606
	}

	features = btrfs_super_compat_ro_flags(disk_super) &
		~BTRFS_FEATURE_COMPAT_RO_SUPP;
	if (!(sb->s_flags & MS_RDONLY) && features) {
		printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
		       "unsupported option features (%Lx).\n",
		       features);
		err = -EINVAL;
1607
		goto fail_iput;
1608 1609
	}

1610 1611 1612 1613 1614
	/*
	 * 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.
	 */
1615 1616
	btrfs_init_workers(&fs_info->workers, "worker",
			   fs_info->thread_pool_size);
C
Chris Mason 已提交
1617

1618 1619 1620
	btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
			   fs_info->thread_pool_size);

1621
	btrfs_init_workers(&fs_info->submit_workers, "submit",
1622 1623
			   min_t(u64, fs_devices->num_devices,
			   fs_info->thread_pool_size));
1624 1625 1626 1627 1628 1629

	/* 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;
1630

1631
	fs_info->workers.idle_thresh = 16;
C
Chris Mason 已提交
1632
	fs_info->workers.ordered = 1;
1633

1634 1635 1636
	fs_info->delalloc_workers.idle_thresh = 2;
	fs_info->delalloc_workers.ordered = 1;

1637 1638 1639 1640
	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",
1641
			   fs_info->thread_pool_size);
1642 1643 1644 1645 1646 1647

	/*
	 * endios are largely parallel and should have a very
	 * low idle thresh
	 */
	fs_info->endio_workers.idle_thresh = 4;
1648
	fs_info->endio_write_workers.idle_thresh = 64;
1649

1650
	btrfs_start_workers(&fs_info->workers, 1);
1651
	btrfs_start_workers(&fs_info->submit_workers, 1);
1652
	btrfs_start_workers(&fs_info->delalloc_workers, 1);
1653
	btrfs_start_workers(&fs_info->fixup_workers, 1);
1654
	btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size);
1655 1656
	btrfs_start_workers(&fs_info->endio_write_workers,
			    fs_info->thread_pool_size);
1657

1658
	fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
C
Chris Mason 已提交
1659 1660
	fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
				    4 * 1024 * 1024 / PAGE_CACHE_SIZE);
1661

1662 1663 1664
	nodesize = btrfs_super_nodesize(disk_super);
	leafsize = btrfs_super_leafsize(disk_super);
	sectorsize = btrfs_super_sectorsize(disk_super);
1665
	stripesize = btrfs_super_stripesize(disk_super);
1666 1667 1668
	tree_root->nodesize = nodesize;
	tree_root->leafsize = leafsize;
	tree_root->sectorsize = sectorsize;
1669
	tree_root->stripesize = stripesize;
1670 1671 1672

	sb->s_blocksize = sectorsize;
	sb->s_blocksize_bits = blksize_bits(sectorsize);
1673

C
Chris Mason 已提交
1674 1675 1676 1677 1678
	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;
	}
1679

1680
	mutex_lock(&fs_info->chunk_mutex);
1681
	ret = btrfs_read_sys_array(tree_root);
1682
	mutex_unlock(&fs_info->chunk_mutex);
1683 1684 1685 1686 1687
	if (ret) {
		printk("btrfs: failed to read the system array on %s\n",
		       sb->s_id);
		goto fail_sys_array;
	}
1688 1689 1690

	blocksize = btrfs_level_size(tree_root,
				     btrfs_super_chunk_root_level(disk_super));
1691
	generation = btrfs_super_chunk_root_generation(disk_super);
1692 1693 1694 1695 1696 1697

	__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),
1698
					   blocksize, generation);
1699 1700
	BUG_ON(!chunk_root->node);

1701 1702 1703 1704
	read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
	         (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
		 BTRFS_UUID_SIZE);

1705
	mutex_lock(&fs_info->chunk_mutex);
1706
	ret = btrfs_read_chunk_tree(chunk_root);
1707
	mutex_unlock(&fs_info->chunk_mutex);
Y
Yan Zheng 已提交
1708 1709 1710 1711
	if (ret) {
		printk("btrfs: failed to read chunk tree on %s\n", sb->s_id);
		goto fail_chunk_root;
	}
1712

1713 1714
	btrfs_close_extra_devices(fs_devices);

1715 1716
	blocksize = btrfs_level_size(tree_root,
				     btrfs_super_root_level(disk_super));
1717
	generation = btrfs_super_generation(disk_super);
1718

C
Chris Mason 已提交
1719
	tree_root->node = read_tree_block(tree_root,
1720
					  btrfs_super_root(disk_super),
1721
					  blocksize, generation);
C
Chris Mason 已提交
1722
	if (!tree_root->node)
Y
Yan Zheng 已提交
1723
		goto fail_chunk_root;
1724

1725 1726

	ret = find_and_setup_root(tree_root, fs_info,
C
Chris Mason 已提交
1727
				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
1728
	if (ret)
C
Chris Mason 已提交
1729
		goto fail_tree_root;
1730 1731 1732 1733 1734 1735 1736 1737
	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;
1738

C
Chris Mason 已提交
1739 1740
	btrfs_read_block_groups(extent_root);

Y
Yan Zheng 已提交
1741 1742
	fs_info->generation = generation + 1;
	fs_info->last_trans_committed = generation;
C
Chris Mason 已提交
1743 1744 1745
	fs_info->data_alloc_profile = (u64)-1;
	fs_info->metadata_alloc_profile = (u64)-1;
	fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
1746 1747 1748 1749 1750 1751 1752 1753 1754
	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)
1755
		goto fail_cleaner;
1756

1757 1758 1759
	if (btrfs_super_log_root(disk_super) != 0) {
		u64 bytenr = btrfs_super_log_root(disk_super);

1760 1761 1762 1763 1764
		if (fs_devices->rw_devices == 0) {
			printk("Btrfs log replay required on RO media\n");
			err = -EIO;
			goto fail_trans_kthread;
		}
1765 1766 1767
		blocksize =
		     btrfs_level_size(tree_root,
				      btrfs_super_log_root_level(disk_super));
C
Chris Mason 已提交
1768

1769 1770 1771 1772 1773 1774 1775
		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,
1776 1777
						      blocksize,
						      generation + 1);
1778 1779
		ret = btrfs_recover_log_trees(log_tree_root);
		BUG_ON(ret);
1780 1781 1782 1783 1784

		if (sb->s_flags & MS_RDONLY) {
			ret =  btrfs_commit_super(tree_root);
			BUG_ON(ret);
		}
1785
	}
Z
Zheng Yan 已提交
1786

1787 1788 1789 1790
	if (!(sb->s_flags & MS_RDONLY)) {
		ret = btrfs_cleanup_reloc_trees(tree_root);
		BUG_ON(ret);
	}
Z
Zheng Yan 已提交
1791

1792 1793 1794 1795 1796 1797
	location.objectid = BTRFS_FS_TREE_OBJECTID;
	location.type = BTRFS_ROOT_ITEM_KEY;
	location.offset = (u64)-1;

	fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
	if (!fs_info->fs_root)
1798
		goto fail_trans_kthread;
1799
	return tree_root;
C
Chris Mason 已提交
1800

1801 1802
fail_trans_kthread:
	kthread_stop(fs_info->transaction_kthread);
1803
fail_cleaner:
1804
	kthread_stop(fs_info->cleaner_kthread);
1805 1806 1807 1808 1809 1810 1811 1812

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

1813 1814
fail_extent_root:
	free_extent_buffer(extent_root->node);
C
Chris Mason 已提交
1815
fail_tree_root:
1816
	free_extent_buffer(tree_root->node);
Y
Yan Zheng 已提交
1817 1818
fail_chunk_root:
	free_extent_buffer(chunk_root->node);
1819
fail_sys_array:
1820
	free_extent_buffer(dev_root->node);
C
Chris Mason 已提交
1821
fail_sb_buffer:
1822
	btrfs_stop_workers(&fs_info->fixup_workers);
1823
	btrfs_stop_workers(&fs_info->delalloc_workers);
1824 1825
	btrfs_stop_workers(&fs_info->workers);
	btrfs_stop_workers(&fs_info->endio_workers);
1826
	btrfs_stop_workers(&fs_info->endio_write_workers);
1827
	btrfs_stop_workers(&fs_info->submit_workers);
1828
fail_iput:
1829
	invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
1830
	iput(fs_info->btree_inode);
C
Chris Mason 已提交
1831
fail:
1832
	btrfs_close_devices(fs_info->fs_devices);
1833 1834
	btrfs_mapping_tree_free(&fs_info->mapping_tree);

C
Chris Mason 已提交
1835 1836
	kfree(extent_root);
	kfree(tree_root);
1837
	bdi_destroy(&fs_info->bdi);
C
Chris Mason 已提交
1838
	kfree(fs_info);
1839 1840
	kfree(chunk_root);
	kfree(dev_root);
C
Chris Mason 已提交
1841
	return ERR_PTR(err);
1842 1843
}

1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855
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));
		}
1856 1857 1858
		/* note, we dont' set_buffer_write_io_error because we have
		 * our own ways of dealing with the IO errors
		 */
1859 1860 1861 1862 1863 1864
		clear_buffer_uptodate(bh);
	}
	unlock_buffer(bh);
	put_bh(bh);
}

1865
static int write_all_supers(struct btrfs_root *root)
1866 1867 1868 1869
{
	struct list_head *cur;
	struct list_head *head = &root->fs_info->fs_devices->devices;
	struct btrfs_device *dev;
1870
	struct btrfs_super_block *sb;
1871 1872 1873 1874
	struct btrfs_dev_item *dev_item;
	struct buffer_head *bh;
	int ret;
	int do_barriers;
1875 1876
	int max_errors;
	int total_errors = 0;
1877 1878
	u32 crc;
	u64 flags;
1879

1880
	max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
1881 1882
	do_barriers = !btrfs_test_opt(root, NOBARRIER);

1883 1884
	sb = &root->fs_info->super_for_commit;
	dev_item = &sb->dev_item;
1885 1886
	list_for_each(cur, head) {
		dev = list_entry(cur, struct btrfs_device, dev_list);
1887 1888 1889 1890
		if (!dev->bdev) {
			total_errors++;
			continue;
		}
Y
Yan Zheng 已提交
1891
		if (!dev->in_fs_metadata || !dev->writeable)
1892 1893
			continue;

Y
Yan Zheng 已提交
1894
		btrfs_set_stack_device_generation(dev_item, 0);
1895 1896 1897 1898 1899 1900 1901 1902
		btrfs_set_stack_device_type(dev_item, dev->type);
		btrfs_set_stack_device_id(dev_item, dev->devid);
		btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
		btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
		btrfs_set_stack_device_io_align(dev_item, dev->io_align);
		btrfs_set_stack_device_io_width(dev_item, dev->io_width);
		btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
		memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
Y
Yan Zheng 已提交
1903
		memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
1904 1905 1906 1907 1908 1909 1910 1911 1912 1913
		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,
1914 1915
			      BTRFS_SUPER_INFO_SIZE);

1916
		memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937
		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);
		}
1938 1939
		if (ret)
			total_errors++;
1940
	}
1941 1942 1943 1944 1945
	if (total_errors > max_errors) {
		printk("btrfs: %d errors while writing supers\n", total_errors);
		BUG();
	}
	total_errors = 0;
1946 1947 1948

	list_for_each(cur, head) {
		dev = list_entry(cur, struct btrfs_device, dev_list);
1949 1950
		if (!dev->bdev)
			continue;
Y
Yan Zheng 已提交
1951
		if (!dev->in_fs_metadata || !dev->writeable)
1952 1953
			continue;

1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967
		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);
1968 1969
				if (!buffer_uptodate(bh))
					total_errors++;
1970
			} else {
1971
				total_errors++;
1972 1973 1974 1975 1976 1977
			}

		}
		dev->pending_io = NULL;
		brelse(bh);
	}
1978 1979 1980 1981
	if (total_errors > max_errors) {
		printk("btrfs: %d errors while writing supers\n", total_errors);
		BUG();
	}
1982 1983 1984
	return 0;
}

1985
int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
C
Chris Mason 已提交
1986
		      *root)
1987
{
1988
	int ret;
1989

1990
	ret = write_all_supers(root);
1991
	return ret;
C
Chris Mason 已提交
1992 1993
}

1994
int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
C
Chris Mason 已提交
1995 1996 1997
{
	radix_tree_delete(&fs_info->fs_roots_radix,
			  (unsigned long)root->root_key.objectid);
1998 1999 2000 2001
	if (root->anon_super.s_dev) {
		down_write(&root->anon_super.s_umount);
		kill_anon_super(&root->anon_super);
	}
2002
#if 0
2003 2004
	if (root->in_sysfs)
		btrfs_sysfs_del_root(root);
2005
#endif
C
Chris Mason 已提交
2006
	if (root->node)
2007
		free_extent_buffer(root->node);
C
Chris Mason 已提交
2008
	if (root->commit_root)
2009
		free_extent_buffer(root->commit_root);
2010 2011
	if (root->name)
		kfree(root->name);
C
Chris Mason 已提交
2012 2013 2014 2015
	kfree(root);
	return 0;
}

C
Chris Mason 已提交
2016
static int del_fs_roots(struct btrfs_fs_info *fs_info)
2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027
{
	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
2028
		for (i = 0; i < ret; i++)
2029
			btrfs_free_fs_root(fs_info, gang[i]);
2030 2031 2032
	}
	return 0;
}
2033

Y
Yan Zheng 已提交
2034
int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
C
Chris Mason 已提交
2035
{
Y
Yan Zheng 已提交
2036 2037 2038
	u64 root_objectid = 0;
	struct btrfs_root *gang[8];
	int i;
2039
	int ret;
2040

Y
Yan Zheng 已提交
2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057
	while (1) {
		ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
					     (void **)gang, root_objectid,
					     ARRAY_SIZE(gang));
		if (!ret)
			break;
		for (i = 0; i < ret; i++) {
			root_objectid = gang[i]->root_key.objectid;
			ret = btrfs_find_dead_roots(fs_info->tree_root,
						    root_objectid, gang[i]);
			BUG_ON(ret);
			btrfs_orphan_cleanup(gang[i]);
		}
		root_objectid++;
	}
	return 0;
}
2058

Y
Yan Zheng 已提交
2059 2060 2061 2062
int btrfs_commit_super(struct btrfs_root *root)
{
	struct btrfs_trans_handle *trans;
	int ret;
2063

Y
Yan Zheng 已提交
2064
	mutex_lock(&root->fs_info->cleaner_mutex);
2065
	btrfs_clean_old_snapshots(root);
Y
Yan Zheng 已提交
2066
	mutex_unlock(&root->fs_info->cleaner_mutex);
C
Chris Mason 已提交
2067
	trans = btrfs_start_transaction(root, 1);
2068
	ret = btrfs_commit_transaction(trans, root);
Y
Yan Zheng 已提交
2069 2070
	BUG_ON(ret);
	/* run commit again to drop the original snapshot */
C
Chris Mason 已提交
2071 2072 2073
	trans = btrfs_start_transaction(root, 1);
	btrfs_commit_transaction(trans, root);
	ret = btrfs_write_and_wait_transaction(NULL, root);
2074
	BUG_ON(ret);
2075

Y
Yan Zheng 已提交
2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096
	ret = write_ctree_super(NULL, root);
	return ret;
}

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

	fs_info->closing = 1;
	smp_mb();

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

	if (!(fs_info->sb->s_flags & MS_RDONLY)) {
		ret =  btrfs_commit_super(root);
		if (ret) {
			printk("btrfs: commit super returns %d\n", ret);
		}
	}
2097

C
Chris Mason 已提交
2098 2099 2100 2101
	if (fs_info->delalloc_bytes) {
		printk("btrfs: at unmount delalloc count %Lu\n",
		       fs_info->delalloc_bytes);
	}
Y
Yan Zheng 已提交
2102 2103 2104 2105
	if (fs_info->total_ref_cache_size) {
		printk("btrfs: at umount reference cache size %Lu\n",
			fs_info->total_ref_cache_size);
	}
2106

2107
	if (fs_info->extent_root->node)
2108
		free_extent_buffer(fs_info->extent_root->node);
2109

2110
	if (fs_info->tree_root->node)
2111
		free_extent_buffer(fs_info->tree_root->node);
2112

2113 2114 2115 2116 2117 2118
	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
2119
	btrfs_free_block_groups(root->fs_info);
2120

Y
Yan Zheng 已提交
2121
	del_fs_roots(fs_info);
2122

Y
Yan Zheng 已提交
2123
	iput(fs_info->btree_inode);
2124

2125
	btrfs_stop_workers(&fs_info->fixup_workers);
2126
	btrfs_stop_workers(&fs_info->delalloc_workers);
2127 2128
	btrfs_stop_workers(&fs_info->workers);
	btrfs_stop_workers(&fs_info->endio_workers);
2129
	btrfs_stop_workers(&fs_info->endio_write_workers);
2130
	btrfs_stop_workers(&fs_info->submit_workers);
2131

2132 2133 2134 2135 2136 2137 2138 2139 2140 2141
#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
2142
	btrfs_close_devices(fs_info->fs_devices);
2143
	btrfs_mapping_tree_free(&fs_info->mapping_tree);
2144

C
Chris Mason 已提交
2145
	bdi_destroy(&fs_info->bdi);
2146

2147 2148
	kfree(fs_info->extent_root);
	kfree(fs_info->tree_root);
2149 2150
	kfree(fs_info->chunk_root);
	kfree(fs_info->dev_root);
2151 2152 2153
	return 0;
}

2154
int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
2155
{
2156
	int ret;
2157
	struct inode *btree_inode = buf->first_page->mapping->host;
2158 2159 2160 2161 2162 2163 2164 2165

	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;
2166 2167 2168
}

int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
C
Chris Mason 已提交
2169
{
2170
	struct inode *btree_inode = buf->first_page->mapping->host;
2171
	return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
2172 2173
					  buf);
}
2174

2175 2176
void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
{
2177
	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
2178 2179
	u64 transid = btrfs_header_generation(buf);
	struct inode *btree_inode = root->fs_info->btree_inode;
2180

2181
	WARN_ON(!btrfs_tree_locked(buf));
C
Chris Mason 已提交
2182 2183
	if (transid != root->fs_info->generation) {
		printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
2184
			(unsigned long long)buf->start,
C
Chris Mason 已提交
2185 2186 2187
			transid, root->fs_info->generation);
		WARN_ON(1);
	}
2188
	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf);
2189 2190
}

2191
void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
C
Chris Mason 已提交
2192
{
2193 2194 2195 2196
	/*
	 * looks as though older kernels can get into trouble with
	 * this code, they end up stuck in balance_dirty_pages forever
	 */
2197 2198 2199
	struct extent_io_tree *tree;
	u64 num_dirty;
	u64 start = 0;
2200
	unsigned long thresh = 32 * 1024 * 1024;
2201 2202
	tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;

2203
	if (current_is_pdflush() || current->flags & PF_MEMALLOC)
2204 2205 2206 2207 2208 2209
		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
2210
				   root->fs_info->btree_inode->i_mapping, 1);
2211
	}
2212
	return;
C
Chris Mason 已提交
2213
}
2214

2215
int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
2216
{
2217
	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
2218
	int ret;
2219
	ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
2220 2221 2222 2223
	if (ret == 0) {
		buf->flags |= EXTENT_UPTODATE;
	}
	return ret;
2224
}
2225

C
Chris Mason 已提交
2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253
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;
}

2254
static struct extent_io_ops btree_extent_io_ops = {
C
Chris Mason 已提交
2255
	.write_cache_pages_lock_hook = btree_lock_page_hook,
2256
	.readpage_end_io_hook = btree_readpage_end_io_hook,
2257
	.submit_bio_hook = btree_submit_bio_hook,
2258 2259
	/* note we're sharing with inode.c for the merge bio hook */
	.merge_bio_hook = btrfs_merge_bio_hook,
2260
};