transaction.c 48.0 KB
Newer Older
C
Chris Mason 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/*
 * Copyright (C) 2007 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

C
Chris Mason 已提交
19
#include <linux/fs.h>
20
#include <linux/slab.h>
C
Chris Mason 已提交
21
#include <linux/sched.h>
22
#include <linux/writeback.h>
23
#include <linux/pagemap.h>
24
#include <linux/blkdev.h>
25
#include <linux/uuid.h>
C
Chris Mason 已提交
26 27 28
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
29
#include "locking.h"
30
#include "tree-log.h"
31
#include "inode-map.h"
32
#include "volumes.h"
33
#include "dev-replace.h"
C
Chris Mason 已提交
34

35 36
#define BTRFS_ROOT_TRANS_TAG 0

37
void put_transaction(struct btrfs_transaction *transaction)
C
Chris Mason 已提交
38
{
39 40
	WARN_ON(atomic_read(&transaction->use_count) == 0);
	if (atomic_dec_and_test(&transaction->use_count)) {
J
Josef Bacik 已提交
41
		BUG_ON(!list_empty(&transaction->list));
42
		WARN_ON(transaction->delayed_refs.root.rb_node);
C
Chris Mason 已提交
43
		kmem_cache_free(btrfs_transaction_cachep, transaction);
C
Chris Mason 已提交
44
	}
C
Chris Mason 已提交
45 46
}

J
Josef Bacik 已提交
47 48 49 50 51 52
static noinline void switch_commit_root(struct btrfs_root *root)
{
	free_extent_buffer(root->commit_root);
	root->commit_root = btrfs_root_node(root);
}

C
Chris Mason 已提交
53 54 55
/*
 * either allocate a new transaction or hop into the existing one
 */
56
static noinline int join_transaction(struct btrfs_root *root, int type)
C
Chris Mason 已提交
57 58
{
	struct btrfs_transaction *cur_trans;
59
	struct btrfs_fs_info *fs_info = root->fs_info;
J
Josef Bacik 已提交
60

61
	spin_lock(&fs_info->trans_lock);
62
loop:
63
	/* The file system has been taken offline. No new transactions. */
64
	if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
65
		spin_unlock(&fs_info->trans_lock);
66 67 68
		return -EROFS;
	}

69
	if (fs_info->trans_no_join) {
70 71 72 73 74 75 76
		/* 
		 * If we are JOIN_NOLOCK we're already committing a current
		 * transaction, we just need a handle to deal with something
		 * when committing the transaction, such as inode cache and
		 * space cache. It is a special case.
		 */
		if (type != TRANS_JOIN_NOLOCK) {
77
			spin_unlock(&fs_info->trans_lock);
J
Josef Bacik 已提交
78 79 80 81
			return -EBUSY;
		}
	}

82
	cur_trans = fs_info->running_transaction;
J
Josef Bacik 已提交
83
	if (cur_trans) {
84
		if (cur_trans->aborted) {
85
			spin_unlock(&fs_info->trans_lock);
86
			return cur_trans->aborted;
87
		}
J
Josef Bacik 已提交
88
		atomic_inc(&cur_trans->use_count);
89
		atomic_inc(&cur_trans->num_writers);
90
		cur_trans->num_joined++;
91
		spin_unlock(&fs_info->trans_lock);
J
Josef Bacik 已提交
92
		return 0;
C
Chris Mason 已提交
93
	}
94
	spin_unlock(&fs_info->trans_lock);
J
Josef Bacik 已提交
95

96 97 98 99 100 101 102
	/*
	 * If we are ATTACH, we just want to catch the current transaction,
	 * and commit it. If there is no transaction, just return ENOENT.
	 */
	if (type == TRANS_ATTACH)
		return -ENOENT;

J
Josef Bacik 已提交
103 104 105
	cur_trans = kmem_cache_alloc(btrfs_transaction_cachep, GFP_NOFS);
	if (!cur_trans)
		return -ENOMEM;
106

107 108
	spin_lock(&fs_info->trans_lock);
	if (fs_info->running_transaction) {
109 110 111 112
		/*
		 * someone started a transaction after we unlocked.  Make sure
		 * to redo the trans_no_join checks above
		 */
J
Josef Bacik 已提交
113
		kmem_cache_free(btrfs_transaction_cachep, cur_trans);
114
		cur_trans = fs_info->running_transaction;
115
		goto loop;
116
	} else if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
117
		spin_unlock(&fs_info->trans_lock);
118 119
		kmem_cache_free(btrfs_transaction_cachep, cur_trans);
		return -EROFS;
C
Chris Mason 已提交
120
	}
121

J
Josef Bacik 已提交
122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141
	atomic_set(&cur_trans->num_writers, 1);
	cur_trans->num_joined = 0;
	init_waitqueue_head(&cur_trans->writer_wait);
	init_waitqueue_head(&cur_trans->commit_wait);
	cur_trans->in_commit = 0;
	cur_trans->blocked = 0;
	/*
	 * One for this trans handle, one so it will live on until we
	 * commit the transaction.
	 */
	atomic_set(&cur_trans->use_count, 2);
	cur_trans->commit_done = 0;
	cur_trans->start_time = get_seconds();

	cur_trans->delayed_refs.root = RB_ROOT;
	cur_trans->delayed_refs.num_entries = 0;
	cur_trans->delayed_refs.num_heads_ready = 0;
	cur_trans->delayed_refs.num_heads = 0;
	cur_trans->delayed_refs.flushing = 0;
	cur_trans->delayed_refs.run_delayed_start = 0;
142 143 144 145 146 147

	/*
	 * although the tree mod log is per file system and not per transaction,
	 * the log must never go across transaction boundaries.
	 */
	smp_mb();
J
Julia Lawall 已提交
148 149
	if (!list_empty(&fs_info->tree_mod_seq_list))
		WARN(1, KERN_ERR "btrfs: tree_mod_seq_list not empty when "
150
			"creating a fresh transaction\n");
J
Julia Lawall 已提交
151 152
	if (!RB_EMPTY_ROOT(&fs_info->tree_mod_log))
		WARN(1, KERN_ERR "btrfs: tree_mod_log rb tree not empty when "
153 154 155
			"creating a fresh transaction\n");
	atomic_set(&fs_info->tree_mod_seq, 0);

J
Josef Bacik 已提交
156 157 158 159
	spin_lock_init(&cur_trans->commit_lock);
	spin_lock_init(&cur_trans->delayed_refs.lock);

	INIT_LIST_HEAD(&cur_trans->pending_snapshots);
160
	list_add_tail(&cur_trans->list, &fs_info->trans_list);
J
Josef Bacik 已提交
161
	extent_io_tree_init(&cur_trans->dirty_pages,
162 163 164 165
			     fs_info->btree_inode->i_mapping);
	fs_info->generation++;
	cur_trans->transid = fs_info->generation;
	fs_info->running_transaction = cur_trans;
166
	cur_trans->aborted = 0;
167
	spin_unlock(&fs_info->trans_lock);
168

C
Chris Mason 已提交
169 170 171
	return 0;
}

C
Chris Mason 已提交
172
/*
C
Chris Mason 已提交
173 174 175 176
 * this does all the record keeping required to make sure that a reference
 * counted root is properly recorded in a given transaction.  This is required
 * to make sure the old root from before we joined the transaction is deleted
 * when the transaction commits
C
Chris Mason 已提交
177
 */
C
Chris Mason 已提交
178
static int record_root_in_trans(struct btrfs_trans_handle *trans,
J
Josef Bacik 已提交
179
			       struct btrfs_root *root)
180
{
181
	if (root->ref_cows && root->last_trans < trans->transid) {
182
		WARN_ON(root == root->fs_info->extent_root);
183 184
		WARN_ON(root->commit_root != root->node);

C
Chris Mason 已提交
185 186 187 188 189 190 191 192 193 194 195 196
		/*
		 * see below for in_trans_setup usage rules
		 * we have the reloc mutex held now, so there
		 * is only one writer in this function
		 */
		root->in_trans_setup = 1;

		/* make sure readers find in_trans_setup before
		 * they find our root->last_trans update
		 */
		smp_wmb();

J
Josef Bacik 已提交
197 198 199 200 201
		spin_lock(&root->fs_info->fs_roots_radix_lock);
		if (root->last_trans == trans->transid) {
			spin_unlock(&root->fs_info->fs_roots_radix_lock);
			return 0;
		}
202 203 204
		radix_tree_tag_set(&root->fs_info->fs_roots_radix,
			   (unsigned long)root->root_key.objectid,
			   BTRFS_ROOT_TRANS_TAG);
J
Josef Bacik 已提交
205
		spin_unlock(&root->fs_info->fs_roots_radix_lock);
C
Chris Mason 已提交
206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226
		root->last_trans = trans->transid;

		/* this is pretty tricky.  We don't want to
		 * take the relocation lock in btrfs_record_root_in_trans
		 * unless we're really doing the first setup for this root in
		 * this transaction.
		 *
		 * Normally we'd use root->last_trans as a flag to decide
		 * if we want to take the expensive mutex.
		 *
		 * But, we have to set root->last_trans before we
		 * init the relocation root, otherwise, we trip over warnings
		 * in ctree.c.  The solution used here is to flag ourselves
		 * with root->in_trans_setup.  When this is 1, we're still
		 * fixing up the reloc trees and everyone must wait.
		 *
		 * When this is zero, they can trust root->last_trans and fly
		 * through btrfs_record_root_in_trans without having to take the
		 * lock.  smp_wmb() makes sure that all the writes above are
		 * done before we pop in the zero below
		 */
227
		btrfs_init_reloc_root(trans, root);
C
Chris Mason 已提交
228 229
		smp_wmb();
		root->in_trans_setup = 0;
230 231 232
	}
	return 0;
}
233

C
Chris Mason 已提交
234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256

int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans,
			       struct btrfs_root *root)
{
	if (!root->ref_cows)
		return 0;

	/*
	 * see record_root_in_trans for comments about in_trans_setup usage
	 * and barriers
	 */
	smp_rmb();
	if (root->last_trans == trans->transid &&
	    !root->in_trans_setup)
		return 0;

	mutex_lock(&root->fs_info->reloc_mutex);
	record_root_in_trans(trans, root);
	mutex_unlock(&root->fs_info->reloc_mutex);

	return 0;
}

C
Chris Mason 已提交
257 258 259 260
/* wait for commit against the current transaction to become unblocked
 * when this is done, it is safe to start a new transaction, but the current
 * transaction might not be fully on disk.
 */
C
Chris Mason 已提交
261
static void wait_current_trans(struct btrfs_root *root)
C
Chris Mason 已提交
262
{
263
	struct btrfs_transaction *cur_trans;
C
Chris Mason 已提交
264

J
Josef Bacik 已提交
265
	spin_lock(&root->fs_info->trans_lock);
266
	cur_trans = root->fs_info->running_transaction;
C
Chris Mason 已提交
267
	if (cur_trans && cur_trans->blocked) {
268
		atomic_inc(&cur_trans->use_count);
J
Josef Bacik 已提交
269
		spin_unlock(&root->fs_info->trans_lock);
L
Li Zefan 已提交
270 271 272

		wait_event(root->fs_info->transaction_wait,
			   !cur_trans->blocked);
273
		put_transaction(cur_trans);
J
Josef Bacik 已提交
274 275
	} else {
		spin_unlock(&root->fs_info->trans_lock);
276
	}
C
Chris Mason 已提交
277 278
}

279 280
static int may_wait_transaction(struct btrfs_root *root, int type)
{
J
Josef Bacik 已提交
281 282 283 284 285 286 287 288
	if (root->fs_info->log_root_recovering)
		return 0;

	if (type == TRANS_USERSPACE)
		return 1;

	if (type == TRANS_START &&
	    !atomic_read(&root->fs_info->open_ioctl_trans))
289
		return 1;
J
Josef Bacik 已提交
290

291 292 293
	return 0;
}

M
Miao Xie 已提交
294 295 296
static struct btrfs_trans_handle *
start_transaction(struct btrfs_root *root, u64 num_items, int type,
		  enum btrfs_reserve_flush_enum flush)
C
Chris Mason 已提交
297
{
298 299
	struct btrfs_trans_handle *h;
	struct btrfs_transaction *cur_trans;
300
	u64 num_bytes = 0;
C
Chris Mason 已提交
301
	int ret;
302
	u64 qgroup_reserved = 0;
L
liubo 已提交
303

304
	if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state))
L
liubo 已提交
305
		return ERR_PTR(-EROFS);
306 307 308 309 310

	if (current->journal_info) {
		WARN_ON(type != TRANS_JOIN && type != TRANS_JOIN_NOLOCK);
		h = current->journal_info;
		h->use_count++;
311
		WARN_ON(h->use_count > 2);
312 313 314 315
		h->orig_rsv = h->block_rsv;
		h->block_rsv = NULL;
		goto got_it;
	}
316 317 318 319 320 321

	/*
	 * Do the reservation before we join the transaction so we can do all
	 * the appropriate flushing if need be.
	 */
	if (num_items > 0 && root != root->fs_info->chunk_root) {
322 323 324 325 326 327 328 329
		if (root->fs_info->quota_enabled &&
		    is_fstree(root->root_key.objectid)) {
			qgroup_reserved = num_items * root->leafsize;
			ret = btrfs_qgroup_reserve(root, qgroup_reserved);
			if (ret)
				return ERR_PTR(ret);
		}

330
		num_bytes = btrfs_calc_trans_metadata_size(root, num_items);
M
Miao Xie 已提交
331 332 333
		ret = btrfs_block_rsv_add(root,
					  &root->fs_info->trans_block_rsv,
					  num_bytes, flush);
334
		if (ret)
335
			goto reserve_fail;
336
	}
337 338
again:
	h = kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS);
339 340 341 342
	if (!h) {
		ret = -ENOMEM;
		goto alloc_fail;
	}
C
Chris Mason 已提交
343

344 345 346 347 348 349
	/*
	 * If we are JOIN_NOLOCK we're already committing a transaction and
	 * waiting on this guy, so we don't need to do the sb_start_intwrite
	 * because we're already holding a ref.  We need this because we could
	 * have raced in and did an fsync() on a file which can kick a commit
	 * and then we deadlock with somebody doing a freeze.
350 351 352
	 *
	 * If we are ATTACH, it means we just want to catch the current
	 * transaction and commit it, so we needn't do sb_start_intwrite(). 
353
	 */
354
	if (type < TRANS_JOIN_NOLOCK)
355
		sb_start_intwrite(root->fs_info->sb);
356

357
	if (may_wait_transaction(root, type))
C
Chris Mason 已提交
358
		wait_current_trans(root);
359

J
Josef Bacik 已提交
360
	do {
361
		ret = join_transaction(root, type);
J
Josef Bacik 已提交
362 363 364 365
		if (ret == -EBUSY)
			wait_current_trans(root);
	} while (ret == -EBUSY);

T
Tsutomu Itoh 已提交
366
	if (ret < 0) {
367 368
		/* We must get the transaction if we are JOIN_NOLOCK. */
		BUG_ON(type == TRANS_JOIN_NOLOCK);
369
		goto join_fail;
T
Tsutomu Itoh 已提交
370
	}
371

372 373 374 375
	cur_trans = root->fs_info->running_transaction;

	h->transid = cur_trans->transid;
	h->transaction = cur_trans;
C
Chris Mason 已提交
376
	h->blocks_used = 0;
377
	h->bytes_reserved = 0;
378
	h->root = root;
379
	h->delayed_ref_updates = 0;
380
	h->use_count = 1;
381
	h->adding_csums = 0;
382
	h->block_rsv = NULL;
383
	h->orig_rsv = NULL;
384
	h->aborted = 0;
385
	h->qgroup_reserved = qgroup_reserved;
386
	h->delayed_ref_elem.seq = 0;
387
	h->type = type;
388
	h->allocating_chunk = false;
389
	INIT_LIST_HEAD(&h->qgroup_ref_list);
390
	INIT_LIST_HEAD(&h->new_bgs);
391

392 393 394 395 396 397
	smp_mb();
	if (cur_trans->blocked && may_wait_transaction(root, type)) {
		btrfs_commit_transaction(h, root);
		goto again;
	}

398
	if (num_bytes) {
J
Josef Bacik 已提交
399
		trace_btrfs_space_reservation(root->fs_info, "transaction",
400
					      h->transid, num_bytes, 1);
401 402
		h->block_rsv = &root->fs_info->trans_block_rsv;
		h->bytes_reserved = num_bytes;
403
	}
J
Josef Bacik 已提交
404

405
got_it:
J
Josef Bacik 已提交
406
	btrfs_record_root_in_trans(h, root);
407 408 409

	if (!current->journal_info && type != TRANS_USERSPACE)
		current->journal_info = h;
C
Chris Mason 已提交
410
	return h;
411 412 413 414 415 416 417 418 419 420 421 422 423

join_fail:
	if (type < TRANS_JOIN_NOLOCK)
		sb_end_intwrite(root->fs_info->sb);
	kmem_cache_free(btrfs_trans_handle_cachep, h);
alloc_fail:
	if (num_bytes)
		btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv,
					num_bytes);
reserve_fail:
	if (qgroup_reserved)
		btrfs_qgroup_free(root, qgroup_reserved);
	return ERR_PTR(ret);
C
Chris Mason 已提交
424 425
}

426
struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
427
						   int num_items)
428
{
M
Miao Xie 已提交
429 430
	return start_transaction(root, num_items, TRANS_START,
				 BTRFS_RESERVE_FLUSH_ALL);
431
}
432

M
Miao Xie 已提交
433
struct btrfs_trans_handle *btrfs_start_transaction_lflush(
434 435
					struct btrfs_root *root, int num_items)
{
M
Miao Xie 已提交
436 437
	return start_transaction(root, num_items, TRANS_START,
				 BTRFS_RESERVE_FLUSH_LIMIT);
438 439
}

440
struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root)
441
{
442
	return start_transaction(root, 0, TRANS_JOIN, 0);
443 444
}

445
struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root)
446
{
447
	return start_transaction(root, 0, TRANS_JOIN_NOLOCK, 0);
448 449
}

450
struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *root)
451
{
452
	return start_transaction(root, 0, TRANS_USERSPACE, 0);
453 454
}

455
struct btrfs_trans_handle *btrfs_attach_transaction(struct btrfs_root *root)
456
{
457
	return start_transaction(root, 0, TRANS_ATTACH, 0);
458 459
}

C
Chris Mason 已提交
460
/* wait for a transaction commit to be fully complete */
461
static noinline void wait_for_commit(struct btrfs_root *root,
462 463
				    struct btrfs_transaction *commit)
{
L
Li Zefan 已提交
464
	wait_event(commit->commit_wait, commit->commit_done);
465 466
}

467 468 469
int btrfs_wait_for_commit(struct btrfs_root *root, u64 transid)
{
	struct btrfs_transaction *cur_trans = NULL, *t;
470
	int ret = 0;
471 472 473

	if (transid) {
		if (transid <= root->fs_info->last_trans_committed)
J
Josef Bacik 已提交
474
			goto out;
475

476
		ret = -EINVAL;
477
		/* find specified transaction */
J
Josef Bacik 已提交
478
		spin_lock(&root->fs_info->trans_lock);
479 480 481
		list_for_each_entry(t, &root->fs_info->trans_list, list) {
			if (t->transid == transid) {
				cur_trans = t;
J
Josef Bacik 已提交
482
				atomic_inc(&cur_trans->use_count);
483
				ret = 0;
484 485
				break;
			}
486 487
			if (t->transid > transid) {
				ret = 0;
488
				break;
489
			}
490
		}
J
Josef Bacik 已提交
491
		spin_unlock(&root->fs_info->trans_lock);
492
		/* The specified transaction doesn't exist */
493
		if (!cur_trans)
494
			goto out;
495 496
	} else {
		/* find newest transaction that is committing | committed */
J
Josef Bacik 已提交
497
		spin_lock(&root->fs_info->trans_lock);
498 499 500 501
		list_for_each_entry_reverse(t, &root->fs_info->trans_list,
					    list) {
			if (t->in_commit) {
				if (t->commit_done)
502
					break;
503
				cur_trans = t;
J
Josef Bacik 已提交
504
				atomic_inc(&cur_trans->use_count);
505 506 507
				break;
			}
		}
J
Josef Bacik 已提交
508
		spin_unlock(&root->fs_info->trans_lock);
509
		if (!cur_trans)
J
Josef Bacik 已提交
510
			goto out;  /* nothing committing|committed */
511 512 513 514
	}

	wait_for_commit(root, cur_trans);
	put_transaction(cur_trans);
J
Josef Bacik 已提交
515
out:
516 517 518
	return ret;
}

C
Chris Mason 已提交
519 520
void btrfs_throttle(struct btrfs_root *root)
{
J
Josef Bacik 已提交
521
	if (!atomic_read(&root->fs_info->open_ioctl_trans))
522
		wait_current_trans(root);
C
Chris Mason 已提交
523 524
}

525 526 527 528
static int should_end_transaction(struct btrfs_trans_handle *trans,
				  struct btrfs_root *root)
{
	int ret;
529 530

	ret = btrfs_block_rsv_check(root, &root->fs_info->global_block_rsv, 5);
531 532 533 534 535 536 537 538
	return ret ? 1 : 0;
}

int btrfs_should_end_transaction(struct btrfs_trans_handle *trans,
				 struct btrfs_root *root)
{
	struct btrfs_transaction *cur_trans = trans->transaction;
	int updates;
539
	int err;
540

J
Josef Bacik 已提交
541
	smp_mb();
542 543 544 545 546
	if (cur_trans->blocked || cur_trans->delayed_refs.flushing)
		return 1;

	updates = trans->delayed_ref_updates;
	trans->delayed_ref_updates = 0;
547 548 549 550 551
	if (updates) {
		err = btrfs_run_delayed_refs(trans, root, updates);
		if (err) /* Error code will also eval true */
			return err;
	}
552 553 554 555

	return should_end_transaction(trans, root);
}

556
static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
557
			  struct btrfs_root *root, int throttle)
C
Chris Mason 已提交
558
{
559
	struct btrfs_transaction *cur_trans = trans->transaction;
560
	struct btrfs_fs_info *info = root->fs_info;
561
	int count = 0;
562
	int lock = (trans->type != TRANS_JOIN_NOLOCK);
563
	int err = 0;
564

565 566 567 568 569
	if (--trans->use_count) {
		trans->block_rsv = trans->orig_rsv;
		return 0;
	}

570 571 572 573 574
	/*
	 * do the qgroup accounting as early as possible
	 */
	err = btrfs_delayed_refs_qgroup_accounting(trans, info);

575
	btrfs_trans_release_metadata(trans, root);
576
	trans->block_rsv = NULL;
577 578 579 580 581
	/*
	 * the same root has to be passed to start_transaction and
	 * end_transaction. Subvolume quota depends on this.
	 */
	WARN_ON(trans->root != root);
582 583 584 585 586 587

	if (trans->qgroup_reserved) {
		btrfs_qgroup_free(root, trans->qgroup_reserved);
		trans->qgroup_reserved = 0;
	}

588 589 590
	if (!list_empty(&trans->new_bgs))
		btrfs_create_pending_block_groups(trans, root);

591
	while (count < 2) {
592 593 594 595 596 597 598 599 600 601
		unsigned long cur = trans->delayed_ref_updates;
		trans->delayed_ref_updates = 0;
		if (cur &&
		    trans->transaction->delayed_refs.num_heads_ready > 64) {
			trans->delayed_ref_updates = 0;
			btrfs_run_delayed_refs(trans, root, cur);
		} else {
			break;
		}
		count++;
602
	}
603 604
	btrfs_trans_release_metadata(trans, root);
	trans->block_rsv = NULL;
605

606 607 608
	if (!list_empty(&trans->new_bgs))
		btrfs_create_pending_block_groups(trans, root);

J
Josef Bacik 已提交
609 610
	if (lock && !atomic_read(&root->fs_info->open_ioctl_trans) &&
	    should_end_transaction(trans, root)) {
611
		trans->transaction->blocked = 1;
J
Josef Bacik 已提交
612 613
		smp_wmb();
	}
614

615
	if (lock && cur_trans->blocked && !cur_trans->in_commit) {
616 617 618 619 620 621 622
		if (throttle) {
			/*
			 * We may race with somebody else here so end up having
			 * to call end_transaction on ourselves again, so inc
			 * our use_count.
			 */
			trans->use_count++;
623
			return btrfs_commit_transaction(trans, root);
624
		} else {
625
			wake_up_process(info->transaction_kthread);
626
		}
627 628
	}

629
	if (trans->type < TRANS_JOIN_NOLOCK)
630
		sb_end_intwrite(root->fs_info->sb);
631

632
	WARN_ON(cur_trans != info->running_transaction);
633 634
	WARN_ON(atomic_read(&cur_trans->num_writers) < 1);
	atomic_dec(&cur_trans->num_writers);
635

636
	smp_mb();
C
Chris Mason 已提交
637 638 639
	if (waitqueue_active(&cur_trans->writer_wait))
		wake_up(&cur_trans->writer_wait);
	put_transaction(cur_trans);
J
Josef Bacik 已提交
640 641 642

	if (current->journal_info == trans)
		current->journal_info = NULL;
643

Y
Yan, Zheng 已提交
644 645 646
	if (throttle)
		btrfs_run_delayed_iputs(root);

647
	if (trans->aborted ||
648
	    test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state))
649
		err = -EIO;
650
	assert_qgroups_uptodate(trans);
651

652 653
	kmem_cache_free(btrfs_trans_handle_cachep, trans);
	return err;
C
Chris Mason 已提交
654 655
}

656 657 658
int btrfs_end_transaction(struct btrfs_trans_handle *trans,
			  struct btrfs_root *root)
{
659 660
	int ret;

661
	ret = __btrfs_end_transaction(trans, root, 0);
662 663 664
	if (ret)
		return ret;
	return 0;
665 666 667 668 669
}

int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
				   struct btrfs_root *root)
{
670 671
	int ret;

672
	ret = __btrfs_end_transaction(trans, root, 1);
673 674 675 676 677 678 679 680
	if (ret)
		return ret;
	return 0;
}

int btrfs_end_transaction_dmeta(struct btrfs_trans_handle *trans,
				struct btrfs_root *root)
{
681
	return __btrfs_end_transaction(trans, root, 1);
682 683
}

C
Chris Mason 已提交
684 685 686
/*
 * when btree blocks are allocated, they have some corresponding bits set for
 * them in one of two extent_io trees.  This is used to make sure all of
687
 * those extents are sent to disk but does not wait on them
C
Chris Mason 已提交
688
 */
689
int btrfs_write_marked_extents(struct btrfs_root *root,
690
			       struct extent_io_tree *dirty_pages, int mark)
C
Chris Mason 已提交
691
{
692
	int err = 0;
693
	int werr = 0;
J
Josef Bacik 已提交
694
	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
695
	struct extent_state *cached_state = NULL;
696
	u64 start = 0;
697
	u64 end;
698

J
Josef Bacik 已提交
699
	while (!find_first_extent_bit(dirty_pages, start, &start, &end,
700 701 702 703
				      mark, &cached_state)) {
		convert_extent_bit(dirty_pages, start, end, EXTENT_NEED_WAIT,
				   mark, &cached_state, GFP_NOFS);
		cached_state = NULL;
J
Josef Bacik 已提交
704 705 706 707 708
		err = filemap_fdatawrite_range(mapping, start, end);
		if (err)
			werr = err;
		cond_resched();
		start = end + 1;
709
	}
710 711 712 713 714 715 716 717 718 719 720 721
	if (err)
		werr = err;
	return werr;
}

/*
 * when btree blocks are allocated, they have some corresponding bits set for
 * them in one of two extent_io trees.  This is used to make sure all of
 * those extents are on disk for transaction or log commit.  We wait
 * on all the pages and clear them from the dirty pages state tree
 */
int btrfs_wait_marked_extents(struct btrfs_root *root,
722
			      struct extent_io_tree *dirty_pages, int mark)
723 724 725
{
	int err = 0;
	int werr = 0;
J
Josef Bacik 已提交
726
	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
727
	struct extent_state *cached_state = NULL;
728 729
	u64 start = 0;
	u64 end;
730

J
Josef Bacik 已提交
731
	while (!find_first_extent_bit(dirty_pages, start, &start, &end,
732 733 734
				      EXTENT_NEED_WAIT, &cached_state)) {
		clear_extent_bit(dirty_pages, start, end, EXTENT_NEED_WAIT,
				 0, 0, &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
735 736 737 738 739
		err = filemap_fdatawait_range(mapping, start, end);
		if (err)
			werr = err;
		cond_resched();
		start = end + 1;
740
	}
741 742 743
	if (err)
		werr = err;
	return werr;
C
Chris Mason 已提交
744 745
}

746 747 748 749 750 751
/*
 * when btree blocks are allocated, they have some corresponding bits set for
 * them in one of two extent_io trees.  This is used to make sure all of
 * those extents are on disk for transaction or log commit
 */
int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
752
				struct extent_io_tree *dirty_pages, int mark)
753 754 755 756
{
	int ret;
	int ret2;

757 758
	ret = btrfs_write_marked_extents(root, dirty_pages, mark);
	ret2 = btrfs_wait_marked_extents(root, dirty_pages, mark);
759 760 761 762 763 764

	if (ret)
		return ret;
	if (ret2)
		return ret2;
	return 0;
765 766
}

767 768 769 770 771 772 773 774 775
int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root)
{
	if (!trans || !trans->transaction) {
		struct inode *btree_inode;
		btree_inode = root->fs_info->btree_inode;
		return filemap_write_and_wait(btree_inode->i_mapping);
	}
	return btrfs_write_and_wait_marked_extents(root,
776 777
					   &trans->transaction->dirty_pages,
					   EXTENT_DIRTY);
778 779
}

C
Chris Mason 已提交
780 781 782 783 784 785 786 787 788 789
/*
 * this is used to update the root pointer in the tree of tree roots.
 *
 * But, in the case of the extent allocation tree, updating the root
 * pointer may allocate blocks which may change the root of the extent
 * allocation tree.
 *
 * So, this loops and repeats and makes sure the cowonly root didn't
 * change while the root pointer was being updated in the metadata.
 */
790 791
static int update_cowonly_root(struct btrfs_trans_handle *trans,
			       struct btrfs_root *root)
C
Chris Mason 已提交
792 793
{
	int ret;
794
	u64 old_root_bytenr;
795
	u64 old_root_used;
796
	struct btrfs_root *tree_root = root->fs_info->tree_root;
C
Chris Mason 已提交
797

798
	old_root_used = btrfs_root_used(&root->root_item);
799
	btrfs_write_dirty_block_groups(trans, root);
800

C
Chris Mason 已提交
801
	while (1) {
802
		old_root_bytenr = btrfs_root_bytenr(&root->root_item);
803 804
		if (old_root_bytenr == root->node->start &&
		    old_root_used == btrfs_root_used(&root->root_item))
C
Chris Mason 已提交
805
			break;
806

807
		btrfs_set_root_node(&root->root_item, root->node);
C
Chris Mason 已提交
808
		ret = btrfs_update_root(trans, tree_root,
809 810
					&root->root_key,
					&root->root_item);
811 812
		if (ret)
			return ret;
813

814
		old_root_used = btrfs_root_used(&root->root_item);
815
		ret = btrfs_write_dirty_block_groups(trans, root);
816 817
		if (ret)
			return ret;
818
	}
819 820 821 822

	if (root != root->fs_info->extent_root)
		switch_commit_root(root);

823 824 825
	return 0;
}

C
Chris Mason 已提交
826 827
/*
 * update all the cowonly tree roots on disk
828 829 830 831
 *
 * The error handling in this function may not be obvious. Any of the
 * failures will cause the file system to go offline. We still need
 * to clean up the delayed refs.
C
Chris Mason 已提交
832
 */
833 834
static noinline int commit_cowonly_roots(struct btrfs_trans_handle *trans,
					 struct btrfs_root *root)
835 836 837
{
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct list_head *next;
838
	struct extent_buffer *eb;
839
	int ret;
840

841
	ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
842 843
	if (ret)
		return ret;
844

845
	eb = btrfs_lock_root_node(fs_info->tree_root);
846 847
	ret = btrfs_cow_block(trans, fs_info->tree_root, eb, NULL,
			      0, &eb);
848 849
	btrfs_tree_unlock(eb);
	free_extent_buffer(eb);
850

851 852 853
	if (ret)
		return ret;

854
	ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
855 856
	if (ret)
		return ret;
857

858
	ret = btrfs_run_dev_stats(trans, root->fs_info);
859 860 861
	WARN_ON(ret);
	ret = btrfs_run_dev_replace(trans, root->fs_info);
	WARN_ON(ret);
862

863 864 865 866 867 868 869
	ret = btrfs_run_qgroups(trans, root->fs_info);
	BUG_ON(ret);

	/* run_qgroups might have added some more refs */
	ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
	BUG_ON(ret);

C
Chris Mason 已提交
870
	while (!list_empty(&fs_info->dirty_cowonly_roots)) {
871 872 873
		next = fs_info->dirty_cowonly_roots.next;
		list_del_init(next);
		root = list_entry(next, struct btrfs_root, dirty_list);
874

875 876 877
		ret = update_cowonly_root(trans, root);
		if (ret)
			return ret;
C
Chris Mason 已提交
878
	}
879 880 881 882 883

	down_write(&fs_info->extent_commit_sem);
	switch_commit_root(fs_info->extent_root);
	up_write(&fs_info->extent_commit_sem);

884 885
	btrfs_after_dev_replace_commit(fs_info);

C
Chris Mason 已提交
886 887 888
	return 0;
}

C
Chris Mason 已提交
889 890 891 892 893
/*
 * dead roots are old snapshots that need to be deleted.  This allocates
 * a dirty root struct and adds it into the list of dead roots that need to
 * be deleted
 */
894
int btrfs_add_dead_root(struct btrfs_root *root)
895
{
J
Josef Bacik 已提交
896
	spin_lock(&root->fs_info->trans_lock);
897
	list_add(&root->root_list, &root->fs_info->dead_roots);
J
Josef Bacik 已提交
898
	spin_unlock(&root->fs_info->trans_lock);
899 900 901
	return 0;
}

C
Chris Mason 已提交
902
/*
903
 * update all the cowonly tree roots on disk
C
Chris Mason 已提交
904
 */
905 906
static noinline int commit_fs_roots(struct btrfs_trans_handle *trans,
				    struct btrfs_root *root)
907 908
{
	struct btrfs_root *gang[8];
909
	struct btrfs_fs_info *fs_info = root->fs_info;
910 911
	int i;
	int ret;
912 913
	int err = 0;

J
Josef Bacik 已提交
914
	spin_lock(&fs_info->fs_roots_radix_lock);
C
Chris Mason 已提交
915
	while (1) {
916 917
		ret = radix_tree_gang_lookup_tag(&fs_info->fs_roots_radix,
						 (void **)gang, 0,
918 919 920 921 922 923
						 ARRAY_SIZE(gang),
						 BTRFS_ROOT_TRANS_TAG);
		if (ret == 0)
			break;
		for (i = 0; i < ret; i++) {
			root = gang[i];
924 925 926
			radix_tree_tag_clear(&fs_info->fs_roots_radix,
					(unsigned long)root->root_key.objectid,
					BTRFS_ROOT_TRANS_TAG);
J
Josef Bacik 已提交
927
			spin_unlock(&fs_info->fs_roots_radix_lock);
Y
Yan Zheng 已提交
928

929
			btrfs_free_log(trans, root);
930
			btrfs_update_reloc_root(trans, root);
931
			btrfs_orphan_commit_root(trans, root);
932

933 934
			btrfs_save_ino_cache(root, trans);

935 936 937 938
			/* see comments in should_cow_block() */
			root->force_cow = 0;
			smp_wmb();

939
			if (root->commit_root != root->node) {
940
				mutex_lock(&root->fs_commit_mutex);
J
Josef Bacik 已提交
941
				switch_commit_root(root);
942 943 944
				btrfs_unpin_free_ino(root);
				mutex_unlock(&root->fs_commit_mutex);

945 946 947
				btrfs_set_root_node(&root->root_item,
						    root->node);
			}
948 949

			err = btrfs_update_root(trans, fs_info->tree_root,
950 951
						&root->root_key,
						&root->root_item);
J
Josef Bacik 已提交
952
			spin_lock(&fs_info->fs_roots_radix_lock);
953 954
			if (err)
				break;
955 956
		}
	}
J
Josef Bacik 已提交
957
	spin_unlock(&fs_info->fs_roots_radix_lock);
958
	return err;
959 960
}

C
Chris Mason 已提交
961
/*
962 963
 * defrag a given btree.
 * Every leaf in the btree is read and defragged.
C
Chris Mason 已提交
964
 */
965
int btrfs_defrag_root(struct btrfs_root *root)
966 967 968
{
	struct btrfs_fs_info *info = root->fs_info;
	struct btrfs_trans_handle *trans;
969
	int ret;
970

971
	if (xchg(&root->defrag_running, 1))
972
		return 0;
973

974
	while (1) {
975 976 977 978
		trans = btrfs_start_transaction(root, 0);
		if (IS_ERR(trans))
			return PTR_ERR(trans);

979
		ret = btrfs_defrag_leaves(trans, root);
980

981
		btrfs_end_transaction(trans, root);
982
		btrfs_btree_balance_dirty(info->tree_root);
983 984
		cond_resched();

985
		if (btrfs_fs_closing(root->fs_info) || ret != -EAGAIN)
986
			break;
987 988 989 990 991 992

		if (btrfs_defrag_cancelled(root->fs_info)) {
			printk(KERN_DEBUG "btrfs: defrag_root cancelled\n");
			ret = -EAGAIN;
			break;
		}
993 994
	}
	root->defrag_running = 0;
995
	return ret;
996 997
}

C
Chris Mason 已提交
998 999 1000 1001
/*
 * new snapshots need to be created at a very specific time in the
 * transaction commit.  This does the actual creation
 */
1002
static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
1003 1004 1005 1006
				   struct btrfs_fs_info *fs_info,
				   struct btrfs_pending_snapshot *pending)
{
	struct btrfs_key key;
1007
	struct btrfs_root_item *new_root_item;
1008 1009
	struct btrfs_root *tree_root = fs_info->tree_root;
	struct btrfs_root *root = pending->root;
1010
	struct btrfs_root *parent_root;
L
Liu Bo 已提交
1011
	struct btrfs_block_rsv *rsv;
1012
	struct inode *parent_inode;
1013 1014
	struct btrfs_path *path;
	struct btrfs_dir_item *dir_item;
1015
	struct dentry *parent;
1016
	struct dentry *dentry;
1017
	struct extent_buffer *tmp;
1018
	struct extent_buffer *old;
1019
	struct timespec cur_time = CURRENT_TIME;
1020
	int ret;
1021
	u64 to_reserve = 0;
1022
	u64 index = 0;
1023
	u64 objectid;
L
Li Zefan 已提交
1024
	u64 root_flags;
1025
	uuid_le new_uuid;
1026

1027 1028 1029 1030 1031 1032
	path = btrfs_alloc_path();
	if (!path) {
		ret = pending->error = -ENOMEM;
		goto path_alloc_fail;
	}

1033 1034
	new_root_item = kmalloc(sizeof(*new_root_item), GFP_NOFS);
	if (!new_root_item) {
1035
		ret = pending->error = -ENOMEM;
1036
		goto root_item_alloc_fail;
1037
	}
1038

1039
	ret = btrfs_find_free_objectid(tree_root, &objectid);
1040 1041
	if (ret) {
		pending->error = ret;
1042
		goto no_free_objectid;
1043
	}
1044

1045
	btrfs_reloc_pre_snapshot(trans, pending, &to_reserve);
1046 1047

	if (to_reserve > 0) {
M
Miao Xie 已提交
1048 1049 1050
		ret = btrfs_block_rsv_add(root, &pending->block_rsv,
					  to_reserve,
					  BTRFS_RESERVE_NO_FLUSH);
1051 1052
		if (ret) {
			pending->error = ret;
1053
			goto no_free_objectid;
1054 1055 1056
		}
	}

A
Arne Jansen 已提交
1057 1058 1059 1060
	ret = btrfs_qgroup_inherit(trans, fs_info, root->root_key.objectid,
				   objectid, pending->inherit);
	if (ret) {
		pending->error = ret;
1061
		goto no_free_objectid;
A
Arne Jansen 已提交
1062 1063
	}

1064
	key.objectid = objectid;
1065 1066
	key.offset = (u64)-1;
	key.type = BTRFS_ROOT_ITEM_KEY;
1067

1068
	rsv = trans->block_rsv;
1069
	trans->block_rsv = &pending->block_rsv;
1070

1071
	dentry = pending->dentry;
1072 1073
	parent = dget_parent(dentry);
	parent_inode = parent->d_inode;
1074
	parent_root = BTRFS_I(parent_inode)->root;
C
Chris Mason 已提交
1075
	record_root_in_trans(trans, parent_root);
1076

1077 1078 1079
	/*
	 * insert the directory item
	 */
1080
	ret = btrfs_set_inode_index(parent_inode, &index);
1081
	BUG_ON(ret); /* -ENOMEM */
1082 1083 1084 1085 1086 1087 1088

	/* check if there is a file/dir which has the same name. */
	dir_item = btrfs_lookup_dir_item(NULL, parent_root, path,
					 btrfs_ino(parent_inode),
					 dentry->d_name.name,
					 dentry->d_name.len, 0);
	if (dir_item != NULL && !IS_ERR(dir_item)) {
1089 1090
		pending->error = -EEXIST;
		goto fail;
1091 1092
	} else if (IS_ERR(dir_item)) {
		ret = PTR_ERR(dir_item);
1093 1094
		btrfs_abort_transaction(trans, root, ret);
		goto fail;
1095
	}
1096
	btrfs_release_path(path);
1097

1098 1099 1100 1101 1102 1103 1104
	/*
	 * pull in the delayed directory update
	 * and the delayed inode item
	 * otherwise we corrupt the FS during
	 * snapshot
	 */
	ret = btrfs_run_delayed_items(trans, root);
1105 1106 1107 1108
	if (ret) {	/* Transaction aborted */
		btrfs_abort_transaction(trans, root, ret);
		goto fail;
	}
1109

C
Chris Mason 已提交
1110
	record_root_in_trans(trans, root);
1111 1112
	btrfs_set_root_last_snapshot(&root->root_item, trans->transid);
	memcpy(new_root_item, &root->root_item, sizeof(*new_root_item));
1113
	btrfs_check_and_init_root_item(new_root_item);
1114

L
Li Zefan 已提交
1115 1116 1117 1118 1119 1120 1121
	root_flags = btrfs_root_flags(new_root_item);
	if (pending->readonly)
		root_flags |= BTRFS_ROOT_SUBVOL_RDONLY;
	else
		root_flags &= ~BTRFS_ROOT_SUBVOL_RDONLY;
	btrfs_set_root_flags(new_root_item, root_flags);

1122 1123 1124 1125 1126 1127 1128
	btrfs_set_root_generation_v2(new_root_item,
			trans->transid);
	uuid_le_gen(&new_uuid);
	memcpy(new_root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
	memcpy(new_root_item->parent_uuid, root->root_item.uuid,
			BTRFS_UUID_SIZE);
	new_root_item->otime.sec = cpu_to_le64(cur_time.tv_sec);
1129
	new_root_item->otime.nsec = cpu_to_le32(cur_time.tv_nsec);
1130 1131 1132 1133 1134 1135
	btrfs_set_root_otransid(new_root_item, trans->transid);
	memset(&new_root_item->stime, 0, sizeof(new_root_item->stime));
	memset(&new_root_item->rtime, 0, sizeof(new_root_item->rtime));
	btrfs_set_root_stransid(new_root_item, 0);
	btrfs_set_root_rtransid(new_root_item, 0);

1136
	old = btrfs_lock_root_node(root);
1137
	ret = btrfs_cow_block(trans, root, old, NULL, 0, &old);
1138 1139 1140
	if (ret) {
		btrfs_tree_unlock(old);
		free_extent_buffer(old);
1141 1142
		btrfs_abort_transaction(trans, root, ret);
		goto fail;
1143
	}
1144

1145 1146
	btrfs_set_lock_blocking(old);

1147
	ret = btrfs_copy_root(trans, root, old, &tmp, objectid);
1148
	/* clean up in any case */
1149 1150
	btrfs_tree_unlock(old);
	free_extent_buffer(old);
1151 1152 1153 1154
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto fail;
	}
1155

1156 1157 1158 1159
	/* see comments in should_cow_block() */
	root->force_cow = 1;
	smp_wmb();

1160
	btrfs_set_root_node(new_root_item, tmp);
1161 1162 1163
	/* record when the snapshot was created in key.offset */
	key.offset = trans->transid;
	ret = btrfs_insert_root(trans, tree_root, &key, new_root_item);
1164 1165
	btrfs_tree_unlock(tmp);
	free_extent_buffer(tmp);
1166 1167 1168 1169
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto fail;
	}
1170

1171 1172 1173 1174
	/*
	 * insert root back/forward references
	 */
	ret = btrfs_add_root_ref(trans, tree_root, objectid,
1175
				 parent_root->root_key.objectid,
L
Li Zefan 已提交
1176
				 btrfs_ino(parent_inode), index,
1177
				 dentry->d_name.name, dentry->d_name.len);
1178 1179 1180 1181
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto fail;
	}
1182

1183 1184
	key.offset = (u64)-1;
	pending->snap = btrfs_read_fs_root_no_name(root->fs_info, &key);
1185 1186
	if (IS_ERR(pending->snap)) {
		ret = PTR_ERR(pending->snap);
1187 1188
		btrfs_abort_transaction(trans, root, ret);
		goto fail;
1189
	}
1190

1191
	ret = btrfs_reloc_post_snapshot(trans, pending);
1192 1193 1194 1195
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto fail;
	}
1196 1197

	ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
1198 1199 1200 1201
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto fail;
	}
1202 1203 1204 1205 1206 1207

	ret = btrfs_insert_dir_item(trans, parent_root,
				    dentry->d_name.name, dentry->d_name.len,
				    parent_inode, &key,
				    BTRFS_FT_DIR, index);
	/* We have check then name at the beginning, so it is impossible. */
C
Chris Mason 已提交
1208
	BUG_ON(ret == -EEXIST || ret == -EOVERFLOW);
1209 1210 1211 1212
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto fail;
	}
1213 1214 1215 1216

	btrfs_i_size_write(parent_inode, parent_inode->i_size +
					 dentry->d_name.len * 2);
	parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1217
	ret = btrfs_update_inode_fallback(trans, parent_root, parent_inode);
1218
	if (ret)
1219
		btrfs_abort_transaction(trans, root, ret);
1220
fail:
1221
	dput(parent);
L
Liu Bo 已提交
1222
	trans->block_rsv = rsv;
1223 1224 1225
no_free_objectid:
	kfree(new_root_item);
root_item_alloc_fail:
1226 1227
	btrfs_free_path(path);
path_alloc_fail:
1228
	btrfs_block_rsv_release(root, &pending->block_rsv, (u64)-1);
1229
	return ret;
1230 1231
}

C
Chris Mason 已提交
1232 1233 1234
/*
 * create all the snapshots we've scheduled for creation
 */
1235 1236
static noinline int create_pending_snapshots(struct btrfs_trans_handle *trans,
					     struct btrfs_fs_info *fs_info)
1237 1238 1239 1240
{
	struct btrfs_pending_snapshot *pending;
	struct list_head *head = &trans->transaction->pending_snapshots;

1241 1242
	list_for_each_entry(pending, head, list)
		create_pending_snapshot(trans, fs_info, pending);
1243 1244 1245
	return 0;
}

1246 1247 1248 1249 1250
static void update_super_roots(struct btrfs_root *root)
{
	struct btrfs_root_item *root_item;
	struct btrfs_super_block *super;

1251
	super = root->fs_info->super_copy;
1252 1253 1254 1255 1256 1257 1258 1259 1260 1261

	root_item = &root->fs_info->chunk_root->root_item;
	super->chunk_root = root_item->bytenr;
	super->chunk_root_generation = root_item->generation;
	super->chunk_root_level = root_item->level;

	root_item = &root->fs_info->tree_root->root_item;
	super->root = root_item->bytenr;
	super->generation = root_item->generation;
	super->root_level = root_item->level;
1262
	if (btrfs_test_opt(root, SPACE_CACHE))
1263
		super->cache_generation = root_item->generation;
1264 1265
}

1266 1267 1268
int btrfs_transaction_in_commit(struct btrfs_fs_info *info)
{
	int ret = 0;
J
Josef Bacik 已提交
1269
	spin_lock(&info->trans_lock);
1270 1271
	if (info->running_transaction)
		ret = info->running_transaction->in_commit;
J
Josef Bacik 已提交
1272
	spin_unlock(&info->trans_lock);
1273 1274 1275
	return ret;
}

1276 1277 1278
int btrfs_transaction_blocked(struct btrfs_fs_info *info)
{
	int ret = 0;
J
Josef Bacik 已提交
1279
	spin_lock(&info->trans_lock);
1280 1281
	if (info->running_transaction)
		ret = info->running_transaction->blocked;
J
Josef Bacik 已提交
1282
	spin_unlock(&info->trans_lock);
1283 1284 1285
	return ret;
}

S
Sage Weil 已提交
1286 1287 1288 1289 1290 1291 1292
/*
 * wait for the current transaction commit to start and block subsequent
 * transaction joins
 */
static void wait_current_trans_commit_start(struct btrfs_root *root,
					    struct btrfs_transaction *trans)
{
L
Li Zefan 已提交
1293
	wait_event(root->fs_info->transaction_blocked_wait, trans->in_commit);
S
Sage Weil 已提交
1294 1295 1296 1297 1298 1299 1300 1301 1302
}

/*
 * wait for the current transaction to start and then become unblocked.
 * caller holds ref.
 */
static void wait_current_trans_commit_start_and_unblock(struct btrfs_root *root,
					 struct btrfs_transaction *trans)
{
L
Li Zefan 已提交
1303 1304
	wait_event(root->fs_info->transaction_wait,
		   trans->commit_done || (trans->in_commit && !trans->blocked));
S
Sage Weil 已提交
1305 1306 1307 1308 1309 1310 1311 1312 1313
}

/*
 * commit transactions asynchronously. once btrfs_commit_transaction_async
 * returns, any subsequent transaction will not be allowed to join.
 */
struct btrfs_async_commit {
	struct btrfs_trans_handle *newtrans;
	struct btrfs_root *root;
1314
	struct work_struct work;
S
Sage Weil 已提交
1315 1316 1317 1318 1319
};

static void do_async_commit(struct work_struct *work)
{
	struct btrfs_async_commit *ac =
1320
		container_of(work, struct btrfs_async_commit, work);
S
Sage Weil 已提交
1321

1322 1323 1324 1325
	/*
	 * We've got freeze protection passed with the transaction.
	 * Tell lockdep about it.
	 */
1326 1327 1328 1329
	if (ac->newtrans->type < TRANS_JOIN_NOLOCK)
		rwsem_acquire_read(
		     &ac->root->fs_info->sb->s_writers.lock_map[SB_FREEZE_FS-1],
		     0, 1, _THIS_IP_);
1330

1331 1332
	current->journal_info = ac->newtrans;

S
Sage Weil 已提交
1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344
	btrfs_commit_transaction(ac->newtrans, ac->root);
	kfree(ac);
}

int btrfs_commit_transaction_async(struct btrfs_trans_handle *trans,
				   struct btrfs_root *root,
				   int wait_for_unblock)
{
	struct btrfs_async_commit *ac;
	struct btrfs_transaction *cur_trans;

	ac = kmalloc(sizeof(*ac), GFP_NOFS);
T
Tsutomu Itoh 已提交
1345 1346
	if (!ac)
		return -ENOMEM;
S
Sage Weil 已提交
1347

1348
	INIT_WORK(&ac->work, do_async_commit);
S
Sage Weil 已提交
1349
	ac->root = root;
1350
	ac->newtrans = btrfs_join_transaction(root);
1351 1352 1353 1354 1355
	if (IS_ERR(ac->newtrans)) {
		int err = PTR_ERR(ac->newtrans);
		kfree(ac);
		return err;
	}
S
Sage Weil 已提交
1356 1357 1358

	/* take transaction reference */
	cur_trans = trans->transaction;
1359
	atomic_inc(&cur_trans->use_count);
S
Sage Weil 已提交
1360 1361

	btrfs_end_transaction(trans, root);
1362 1363 1364 1365 1366

	/*
	 * Tell lockdep we've released the freeze rwsem, since the
	 * async commit thread will be the one to unlock it.
	 */
1367 1368 1369 1370
	if (trans->type < TRANS_JOIN_NOLOCK)
		rwsem_release(
			&root->fs_info->sb->s_writers.lock_map[SB_FREEZE_FS-1],
			1, _THIS_IP_);
1371

1372
	schedule_work(&ac->work);
S
Sage Weil 已提交
1373 1374 1375 1376 1377 1378 1379

	/* wait for transaction to start and unblock */
	if (wait_for_unblock)
		wait_current_trans_commit_start_and_unblock(root, cur_trans);
	else
		wait_current_trans_commit_start(root, cur_trans);

1380 1381 1382 1383
	if (current->journal_info == trans)
		current->journal_info = NULL;

	put_transaction(cur_trans);
S
Sage Weil 已提交
1384 1385 1386
	return 0;
}

1387 1388

static void cleanup_transaction(struct btrfs_trans_handle *trans,
1389
				struct btrfs_root *root, int err)
1390 1391 1392 1393 1394
{
	struct btrfs_transaction *cur_trans = trans->transaction;

	WARN_ON(trans->use_count > 1);

1395 1396
	btrfs_abort_transaction(trans, root, err);

1397 1398
	spin_lock(&root->fs_info->trans_lock);
	list_del_init(&cur_trans->list);
1399 1400 1401 1402
	if (cur_trans == root->fs_info->running_transaction) {
		root->fs_info->running_transaction = NULL;
		root->fs_info->trans_no_join = 0;
	}
1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419
	spin_unlock(&root->fs_info->trans_lock);

	btrfs_cleanup_one_transaction(trans->transaction, root);

	put_transaction(cur_trans);
	put_transaction(cur_trans);

	trace_btrfs_transaction_commit(root);

	btrfs_scrub_continue(root);

	if (current->journal_info == trans)
		current->journal_info = NULL;

	kmem_cache_free(btrfs_trans_handle_cachep, trans);
}

1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434
static int btrfs_flush_all_pending_stuffs(struct btrfs_trans_handle *trans,
					  struct btrfs_root *root)
{
	int flush_on_commit = btrfs_test_opt(root, FLUSHONCOMMIT);
	int snap_pending = 0;
	int ret;

	if (!flush_on_commit) {
		spin_lock(&root->fs_info->trans_lock);
		if (!list_empty(&trans->transaction->pending_snapshots))
			snap_pending = 1;
		spin_unlock(&root->fs_info->trans_lock);
	}

	if (flush_on_commit || snap_pending) {
1435 1436 1437
		ret = btrfs_start_delalloc_inodes(root, 1);
		if (ret)
			return ret;
1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458
		btrfs_wait_ordered_extents(root, 1);
	}

	ret = btrfs_run_delayed_items(trans, root);
	if (ret)
		return ret;

	/*
	 * running the delayed items may have added new refs. account
	 * them now so that they hinder processing of more delayed refs
	 * as little as possible.
	 */
	btrfs_delayed_refs_qgroup_accounting(trans, root->fs_info);

	/*
	 * rename don't use btrfs_join_transaction, so, once we
	 * set the transaction to blocked above, we aren't going
	 * to get any new ordered operations.  We can safely run
	 * it here and no for sure that nothing new will be added
	 * to the list
	 */
1459
	ret = btrfs_run_ordered_operations(root, 1);
1460

1461
	return ret;
1462 1463
}

S
Sage Weil 已提交
1464 1465 1466 1467 1468 1469 1470
/*
 * btrfs_transaction state sequence:
 *    in_commit = 0, blocked = 0  (initial)
 *    in_commit = 1, blocked = 1
 *    blocked = 0
 *    commit_done = 1
 */
C
Chris Mason 已提交
1471 1472 1473
int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
			     struct btrfs_root *root)
{
1474
	unsigned long joined = 0;
1475
	struct btrfs_transaction *cur_trans = trans->transaction;
C
Chris Mason 已提交
1476
	struct btrfs_transaction *prev_trans = NULL;
C
Chris Mason 已提交
1477
	DEFINE_WAIT(wait);
1478
	int ret;
1479 1480
	int should_grow = 0;
	unsigned long now = get_seconds();
C
Chris Mason 已提交
1481

1482 1483 1484
	ret = btrfs_run_ordered_operations(root, 0);
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
1485 1486
		btrfs_end_transaction(trans, root);
		return ret;
1487
	}
1488

1489 1490
	/* Stop the commit early if ->aborted is set */
	if (unlikely(ACCESS_ONCE(cur_trans->aborted))) {
1491
		ret = cur_trans->aborted;
1492 1493
		btrfs_end_transaction(trans, root);
		return ret;
1494
	}
1495

1496 1497 1498 1499
	/* make a pass through all the delayed refs we have so far
	 * any runnings procs may add more while we are here
	 */
	ret = btrfs_run_delayed_refs(trans, root, 0);
1500 1501 1502 1503
	if (ret) {
		btrfs_end_transaction(trans, root);
		return ret;
	}
1504

1505 1506 1507
	btrfs_trans_release_metadata(trans, root);
	trans->block_rsv = NULL;

1508
	cur_trans = trans->transaction;
1509

1510 1511 1512 1513
	/*
	 * set the flushing flag so procs in this transaction have to
	 * start sending their work down.
	 */
1514
	cur_trans->delayed_refs.flushing = 1;
1515

1516 1517 1518
	if (!list_empty(&trans->new_bgs))
		btrfs_create_pending_block_groups(trans, root);

1519
	ret = btrfs_run_delayed_refs(trans, root, 0);
1520 1521 1522 1523
	if (ret) {
		btrfs_end_transaction(trans, root);
		return ret;
	}
1524

J
Josef Bacik 已提交
1525
	spin_lock(&cur_trans->commit_lock);
1526
	if (cur_trans->in_commit) {
J
Josef Bacik 已提交
1527
		spin_unlock(&cur_trans->commit_lock);
1528
		atomic_inc(&cur_trans->use_count);
1529
		ret = btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
1530

1531
		wait_for_commit(root, cur_trans);
1532

C
Chris Mason 已提交
1533
		put_transaction(cur_trans);
1534

1535
		return ret;
C
Chris Mason 已提交
1536
	}
1537

C
Chris Mason 已提交
1538
	trans->transaction->in_commit = 1;
1539
	trans->transaction->blocked = 1;
J
Josef Bacik 已提交
1540
	spin_unlock(&cur_trans->commit_lock);
S
Sage Weil 已提交
1541 1542
	wake_up(&root->fs_info->transaction_blocked_wait);

J
Josef Bacik 已提交
1543
	spin_lock(&root->fs_info->trans_lock);
C
Chris Mason 已提交
1544 1545 1546 1547
	if (cur_trans->list.prev != &root->fs_info->trans_list) {
		prev_trans = list_entry(cur_trans->list.prev,
					struct btrfs_transaction, list);
		if (!prev_trans->commit_done) {
1548
			atomic_inc(&prev_trans->use_count);
J
Josef Bacik 已提交
1549
			spin_unlock(&root->fs_info->trans_lock);
C
Chris Mason 已提交
1550 1551 1552

			wait_for_commit(root, prev_trans);

1553
			put_transaction(prev_trans);
J
Josef Bacik 已提交
1554 1555
		} else {
			spin_unlock(&root->fs_info->trans_lock);
C
Chris Mason 已提交
1556
		}
J
Josef Bacik 已提交
1557 1558
	} else {
		spin_unlock(&root->fs_info->trans_lock);
C
Chris Mason 已提交
1559
	}
1560

1561 1562
	if (!btrfs_test_opt(root, SSD) &&
	    (now < cur_trans->start_time || now - cur_trans->start_time < 1))
1563 1564
		should_grow = 1;

1565 1566
	do {
		joined = cur_trans->num_joined;
1567

C
Chris Mason 已提交
1568
		WARN_ON(cur_trans != trans->transaction);
1569

1570
		ret = btrfs_flush_all_pending_stuffs(trans, root);
1571 1572
		if (ret)
			goto cleanup_transaction;
1573

1574 1575 1576
		prepare_to_wait(&cur_trans->writer_wait, &wait,
				TASK_UNINTERRUPTIBLE);

1577
		if (atomic_read(&cur_trans->num_writers) > 1)
1578 1579 1580
			schedule_timeout(MAX_SCHEDULE_TIMEOUT);
		else if (should_grow)
			schedule_timeout(1);
1581 1582

		finish_wait(&cur_trans->writer_wait, &wait);
1583
	} while (atomic_read(&cur_trans->num_writers) > 1 ||
1584
		 (should_grow && cur_trans->num_joined != joined));
1585

1586 1587 1588 1589
	ret = btrfs_flush_all_pending_stuffs(trans, root);
	if (ret)
		goto cleanup_transaction;

1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600
	/*
	 * Ok now we need to make sure to block out any other joins while we
	 * commit the transaction.  We could have started a join before setting
	 * no_join so make sure to wait for num_writers to == 1 again.
	 */
	spin_lock(&root->fs_info->trans_lock);
	root->fs_info->trans_no_join = 1;
	spin_unlock(&root->fs_info->trans_lock);
	wait_event(cur_trans->writer_wait,
		   atomic_read(&cur_trans->num_writers) == 1);

1601 1602 1603 1604 1605
	/* ->aborted might be set after the previous check, so check it */
	if (unlikely(ACCESS_ONCE(cur_trans->aborted))) {
		ret = cur_trans->aborted;
		goto cleanup_transaction;
	}
C
Chris Mason 已提交
1606 1607 1608 1609 1610 1611 1612
	/*
	 * the reloc mutex makes sure that we stop
	 * the balancing code from coming in and moving
	 * extents around in the middle of the commit
	 */
	mutex_lock(&root->fs_info->reloc_mutex);

1613 1614 1615 1616 1617 1618
	/*
	 * We needn't worry about the delayed items because we will
	 * deal with them in create_pending_snapshot(), which is the
	 * core function of the snapshot creation.
	 */
	ret = create_pending_snapshots(trans, root->fs_info);
1619 1620 1621 1622
	if (ret) {
		mutex_unlock(&root->fs_info->reloc_mutex);
		goto cleanup_transaction;
	}
1623

1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634
	/*
	 * We insert the dir indexes of the snapshots and update the inode
	 * of the snapshots' parents after the snapshot creation, so there
	 * are some delayed items which are not dealt with. Now deal with
	 * them.
	 *
	 * We needn't worry that this operation will corrupt the snapshots,
	 * because all the tree which are snapshoted will be forced to COW
	 * the nodes and leaves.
	 */
	ret = btrfs_run_delayed_items(trans, root);
1635 1636 1637 1638
	if (ret) {
		mutex_unlock(&root->fs_info->reloc_mutex);
		goto cleanup_transaction;
	}
1639

1640
	ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
1641 1642 1643 1644
	if (ret) {
		mutex_unlock(&root->fs_info->reloc_mutex);
		goto cleanup_transaction;
	}
1645

1646 1647 1648 1649 1650 1651
	/*
	 * make sure none of the code above managed to slip in a
	 * delayed item
	 */
	btrfs_assert_delayed_root_empty(root);

C
Chris Mason 已提交
1652
	WARN_ON(cur_trans != trans->transaction);
C
Chris Mason 已提交
1653

A
Arne Jansen 已提交
1654
	btrfs_scrub_pause(root);
1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669
	/* btrfs_commit_tree_roots is responsible for getting the
	 * various roots consistent with each other.  Every pointer
	 * in the tree of tree roots has to point to the most up to date
	 * root for every subvolume and other tree.  So, we have to keep
	 * the tree logging code from jumping in and changing any
	 * of the trees.
	 *
	 * At this point in the commit, there can't be any tree-log
	 * writers, but a little lower down we drop the trans mutex
	 * and let new people in.  By holding the tree_log_mutex
	 * from now until after the super is written, we avoid races
	 * with the tree-log code.
	 */
	mutex_lock(&root->fs_info->tree_log_mutex);

1670
	ret = commit_fs_roots(trans, root);
1671 1672
	if (ret) {
		mutex_unlock(&root->fs_info->tree_log_mutex);
1673
		mutex_unlock(&root->fs_info->reloc_mutex);
1674 1675
		goto cleanup_transaction;
	}
1676

1677
	/* commit_fs_roots gets rid of all the tree log roots, it is now
1678 1679 1680 1681
	 * safe to free the root of tree log roots
	 */
	btrfs_free_log_root_tree(trans, root->fs_info);

1682
	ret = commit_cowonly_roots(trans, root);
1683 1684
	if (ret) {
		mutex_unlock(&root->fs_info->tree_log_mutex);
1685
		mutex_unlock(&root->fs_info->reloc_mutex);
1686 1687
		goto cleanup_transaction;
	}
1688

1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699
	/*
	 * The tasks which save the space cache and inode cache may also
	 * update ->aborted, check it.
	 */
	if (unlikely(ACCESS_ONCE(cur_trans->aborted))) {
		ret = cur_trans->aborted;
		mutex_unlock(&root->fs_info->tree_log_mutex);
		mutex_unlock(&root->fs_info->reloc_mutex);
		goto cleanup_transaction;
	}

1700 1701
	btrfs_prepare_extent_commit(trans, root);

C
Chris Mason 已提交
1702
	cur_trans = root->fs_info->running_transaction;
1703 1704 1705

	btrfs_set_root_node(&root->fs_info->tree_root->root_item,
			    root->fs_info->tree_root->node);
J
Josef Bacik 已提交
1706
	switch_commit_root(root->fs_info->tree_root);
1707 1708 1709

	btrfs_set_root_node(&root->fs_info->chunk_root->root_item,
			    root->fs_info->chunk_root->node);
J
Josef Bacik 已提交
1710
	switch_commit_root(root->fs_info->chunk_root);
1711

1712
	assert_qgroups_uptodate(trans);
1713
	update_super_roots(root);
1714 1715

	if (!root->fs_info->log_root_recovering) {
1716 1717
		btrfs_set_super_log_root(root->fs_info->super_copy, 0);
		btrfs_set_super_log_root_level(root->fs_info->super_copy, 0);
1718 1719
	}

1720 1721
	memcpy(root->fs_info->super_for_commit, root->fs_info->super_copy,
	       sizeof(*root->fs_info->super_copy));
C
Chris Mason 已提交
1722

1723
	trans->transaction->blocked = 0;
J
Josef Bacik 已提交
1724 1725 1726 1727
	spin_lock(&root->fs_info->trans_lock);
	root->fs_info->running_transaction = NULL;
	root->fs_info->trans_no_join = 0;
	spin_unlock(&root->fs_info->trans_lock);
C
Chris Mason 已提交
1728
	mutex_unlock(&root->fs_info->reloc_mutex);
1729

1730
	wake_up(&root->fs_info->transaction_wait);
1731

C
Chris Mason 已提交
1732
	ret = btrfs_write_and_wait_transaction(trans, root);
1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744
	if (ret) {
		btrfs_error(root->fs_info, ret,
			    "Error while writing out transaction.");
		mutex_unlock(&root->fs_info->tree_log_mutex);
		goto cleanup_transaction;
	}

	ret = write_ctree_super(trans, root, 0);
	if (ret) {
		mutex_unlock(&root->fs_info->tree_log_mutex);
		goto cleanup_transaction;
	}
1745

1746 1747 1748 1749 1750 1751
	/*
	 * the super is written, we can safely allow the tree-loggers
	 * to go about their business
	 */
	mutex_unlock(&root->fs_info->tree_log_mutex);

1752
	btrfs_finish_extent_commit(trans, root);
1753

C
Chris Mason 已提交
1754
	cur_trans->commit_done = 1;
1755

1756
	root->fs_info->last_trans_committed = cur_trans->transid;
J
Josef Bacik 已提交
1757

C
Chris Mason 已提交
1758
	wake_up(&cur_trans->commit_wait);
1759

J
Josef Bacik 已提交
1760
	spin_lock(&root->fs_info->trans_lock);
1761
	list_del_init(&cur_trans->list);
J
Josef Bacik 已提交
1762 1763
	spin_unlock(&root->fs_info->trans_lock);

C
Chris Mason 已提交
1764
	put_transaction(cur_trans);
C
Chris Mason 已提交
1765
	put_transaction(cur_trans);
1766

1767 1768
	if (trans->type < TRANS_JOIN_NOLOCK)
		sb_end_intwrite(root->fs_info->sb);
1769

1770 1771
	trace_btrfs_transaction_commit(root);

A
Arne Jansen 已提交
1772 1773
	btrfs_scrub_continue(root);

J
Josef Bacik 已提交
1774 1775 1776
	if (current->journal_info == trans)
		current->journal_info = NULL;

C
Chris Mason 已提交
1777
	kmem_cache_free(btrfs_trans_handle_cachep, trans);
Y
Yan, Zheng 已提交
1778 1779 1780 1781

	if (current != root->fs_info->transaction_kthread)
		btrfs_run_delayed_iputs(root);

C
Chris Mason 已提交
1782
	return ret;
1783 1784

cleanup_transaction:
1785 1786
	btrfs_trans_release_metadata(trans, root);
	trans->block_rsv = NULL;
1787 1788 1789 1790
	btrfs_printk(root->fs_info, "Skipping commit of aborted transaction.\n");
//	WARN_ON(1);
	if (current->journal_info == trans)
		current->journal_info = NULL;
1791
	cleanup_transaction(trans, root, ret);
1792 1793

	return ret;
C
Chris Mason 已提交
1794 1795
}

C
Chris Mason 已提交
1796 1797 1798
/*
 * interface function to delete all the snapshots we have scheduled for deletion
 */
1799 1800
int btrfs_clean_old_snapshots(struct btrfs_root *root)
{
1801 1802 1803
	LIST_HEAD(list);
	struct btrfs_fs_info *fs_info = root->fs_info;

J
Josef Bacik 已提交
1804
	spin_lock(&fs_info->trans_lock);
1805
	list_splice_init(&fs_info->dead_roots, &list);
J
Josef Bacik 已提交
1806
	spin_unlock(&fs_info->trans_lock);
1807

1808
	while (!list_empty(&list)) {
1809 1810
		int ret;

1811
		root = list_entry(list.next, struct btrfs_root, root_list);
1812 1813
		list_del(&root->root_list);

1814 1815
		btrfs_kill_all_delayed_nodes(root);

1816 1817
		if (btrfs_header_backref_rev(root->node) <
		    BTRFS_MIXED_BACKREF_REV)
1818
			ret = btrfs_drop_snapshot(root, NULL, 0, 0);
1819
		else
1820 1821
			ret =btrfs_drop_snapshot(root, NULL, 1, 0);
		BUG_ON(ret < 0);
1822 1823 1824
	}
	return 0;
}