[linux-2.6-block.git] / fs / btrfs / transaction.c

/*
 * 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.
 */

#include <linux/module.h>
#include <linux/fs.h>
#include <linux/sched.h>
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"

static int total_trans = 0;
extern struct kmem_cache *btrfs_trans_handle_cachep;
extern struct kmem_cache *btrfs_transaction_cachep;

static struct workqueue_struct *trans_wq;

#define BTRFS_ROOT_TRANS_TAG 0

static void put_transaction(struct btrfs_transaction *transaction)
{
	WARN_ON(transaction->use_count == 0);
	transaction->use_count--;
	if (transaction->use_count == 0) {
		WARN_ON(total_trans == 0);
		total_trans--;
		list_del_init(&transaction->list);
		memset(transaction, 0, sizeof(*transaction));
		kmem_cache_free(btrfs_transaction_cachep, transaction);
	}
}

static int join_transaction(struct btrfs_root *root)
{
	struct btrfs_transaction *cur_trans;
	cur_trans = root->fs_info->running_transaction;
	if (!cur_trans) {
		cur_trans = kmem_cache_alloc(btrfs_transaction_cachep,
					     GFP_NOFS);
		total_trans++;
		BUG_ON(!cur_trans);
		root->fs_info->generation++;
		root->fs_info->running_transaction = cur_trans;
		cur_trans->num_writers = 0;
		cur_trans->transid = root->fs_info->generation;
		init_waitqueue_head(&cur_trans->writer_wait);
		init_waitqueue_head(&cur_trans->commit_wait);
		cur_trans->in_commit = 0;
		cur_trans->use_count = 1;
		cur_trans->commit_done = 0;
		cur_trans->start_time = get_seconds();
		list_add_tail(&cur_trans->list, &root->fs_info->trans_list);
		init_bit_radix(&cur_trans->dirty_pages);
	}
	cur_trans->num_writers++;
	return 0;
}

struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
						   int num_blocks)
{
	struct btrfs_trans_handle *h =
		kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS);
	int ret;
	u64 running_trans_id;

	mutex_lock(&root->fs_info->trans_mutex);
	ret = join_transaction(root);
	BUG_ON(ret);
	running_trans_id = root->fs_info->running_transaction->transid;

	if (root != root->fs_info->tree_root && root->last_trans <
	    running_trans_id) {
		radix_tree_tag_set(&root->fs_info->fs_roots_radix,
				   (unsigned long)root->root_key.objectid,
				   BTRFS_ROOT_TRANS_TAG);
		root->commit_root = root->node;
		get_bh(root->node);
	}
	root->last_trans = running_trans_id;
	h->transid = running_trans_id;
	h->transaction = root->fs_info->running_transaction;
	h->blocks_reserved = num_blocks;
	h->blocks_used = 0;
	h->block_group = NULL;
	root->fs_info->running_transaction->use_count++;
	mutex_unlock(&root->fs_info->trans_mutex);
	return h;
}

int btrfs_end_transaction(struct btrfs_trans_handle *trans,
			  struct btrfs_root *root)
{
	struct btrfs_transaction *cur_trans;

	mutex_lock(&root->fs_info->trans_mutex);
	cur_trans = root->fs_info->running_transaction;
	WARN_ON(cur_trans->num_writers < 1);
	if (waitqueue_active(&cur_trans->writer_wait))
		wake_up(&cur_trans->writer_wait);
	cur_trans->num_writers--;
	put_transaction(cur_trans);
	mutex_unlock(&root->fs_info->trans_mutex);
	memset(trans, 0, sizeof(*trans));
	kmem_cache_free(btrfs_trans_handle_cachep, trans);
	return 0;
}


int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root)
{
	unsigned long gang[16];
	int ret;
	int i;
	int err;
	int werr = 0;
	struct page *page;
	struct radix_tree_root *dirty_pages;
	struct inode *btree_inode = root->fs_info->btree_inode;

	if (!trans || !trans->transaction) {
		return filemap_write_and_wait(btree_inode->i_mapping);
	}
	dirty_pages = &trans->transaction->dirty_pages;
	while(1) {
		ret = find_first_radix_bit(dirty_pages, gang,
					   0, ARRAY_SIZE(gang));
		if (!ret)
			break;
		for (i = 0; i < ret; i++) {
			/* FIXME EIO */
			clear_radix_bit(dirty_pages, gang[i]);
			page = find_lock_page(btree_inode->i_mapping,
					      gang[i]);
			if (!page)
				continue;
			err = write_one_page(page, 0);
			if (err)
				werr = err;
			page_cache_release(page);
		}
	}
	err = filemap_fdatawait(btree_inode->i_mapping);
	if (err)
		werr = err;
	return werr;
}

int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
			    struct btrfs_root *root)
{
	int ret;
	u64 old_extent_block;
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct btrfs_root *tree_root = fs_info->tree_root;
	struct btrfs_root *extent_root = fs_info->extent_root;

	btrfs_write_dirty_block_groups(trans, extent_root);
	while(1) {
		old_extent_block = btrfs_root_blocknr(&extent_root->root_item);
		if (old_extent_block == bh_blocknr(extent_root->node))
			break;
		btrfs_set_root_blocknr(&extent_root->root_item,
				       bh_blocknr(extent_root->node));
		ret = btrfs_update_root(trans, tree_root,
					&extent_root->root_key,
					&extent_root->root_item);
		BUG_ON(ret);
		btrfs_write_dirty_block_groups(trans, extent_root);
	}
	return 0;
}

static int wait_for_commit(struct btrfs_root *root,
			   struct btrfs_transaction *commit)
{
	DEFINE_WAIT(wait);
	while(!commit->commit_done) {
		prepare_to_wait(&commit->commit_wait, &wait,
				TASK_UNINTERRUPTIBLE);
		if (commit->commit_done)
			break;
		mutex_unlock(&root->fs_info->trans_mutex);
		schedule();
		mutex_lock(&root->fs_info->trans_mutex);
	}
	finish_wait(&commit->commit_wait, &wait);
	return 0;
}

struct dirty_root {
	struct list_head list;
	struct btrfs_key snap_key;
	struct buffer_head *commit_root;
	struct btrfs_root *root;
};

static int add_dirty_roots(struct btrfs_trans_handle *trans,
			   struct radix_tree_root *radix,
			   struct list_head *list)
{
	struct dirty_root *dirty;
	struct btrfs_root *gang[8];
	struct btrfs_root *root;
	int i;
	int ret;
	int err = 0;

	while(1) {
		ret = radix_tree_gang_lookup_tag(radix, (void **)gang, 0,
						 ARRAY_SIZE(gang),
						 BTRFS_ROOT_TRANS_TAG);
		if (ret == 0)
			break;
		for (i = 0; i < ret; i++) {
			root = gang[i];
			radix_tree_tag_clear(radix,
				     (unsigned long)root->root_key.objectid,
				     BTRFS_ROOT_TRANS_TAG);
			if (root->commit_root == root->node) {
				WARN_ON(bh_blocknr(root->node) !=
					btrfs_root_blocknr(&root->root_item));
				brelse(root->commit_root);
				root->commit_root = NULL;
				continue;
			}
			dirty = kmalloc(sizeof(*dirty), GFP_NOFS);
			BUG_ON(!dirty);
			memcpy(&dirty->snap_key, &root->root_key,
			       sizeof(root->root_key));
			dirty->commit_root = root->commit_root;
			root->commit_root = NULL;
			dirty->root = root;
			root->root_key.offset = root->fs_info->generation;
			btrfs_set_root_blocknr(&root->root_item,
					       bh_blocknr(root->node));
			err = btrfs_insert_root(trans, root->fs_info->tree_root,
						&root->root_key,
						&root->root_item);
			if (err)
				break;
			list_add(&dirty->list, list);
		}
	}
	return err;
}

static int drop_dirty_roots(struct btrfs_root *tree_root,
			    struct list_head *list)
{
	struct dirty_root *dirty;
	struct btrfs_trans_handle *trans;
	int ret = 0;
	while(!list_empty(list)) {
		mutex_lock(&tree_root->fs_info->fs_mutex);
		dirty = list_entry(list->next, struct dirty_root, list);
		list_del_init(&dirty->list);
		trans = btrfs_start_transaction(tree_root, 1);
		ret = btrfs_drop_snapshot(trans, dirty->root,
					  dirty->commit_root);
		BUG_ON(ret);

		ret = btrfs_del_root(trans, tree_root, &dirty->snap_key);
		if (ret)
			break;
		ret = btrfs_end_transaction(trans, tree_root);
		BUG_ON(ret);
		kfree(dirty);
		mutex_unlock(&tree_root->fs_info->fs_mutex);
		btrfs_btree_balance_dirty(tree_root);
	}
	return ret;
}

int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
			     struct btrfs_root *root)
{
	int ret = 0;
	struct btrfs_transaction *cur_trans;
	struct btrfs_transaction *prev_trans = NULL;
	struct list_head dirty_fs_roots;
	DEFINE_WAIT(wait);

	INIT_LIST_HEAD(&dirty_fs_roots);

	mutex_lock(&root->fs_info->trans_mutex);
	if (trans->transaction->in_commit) {
		cur_trans = trans->transaction;
		trans->transaction->use_count++;
		btrfs_end_transaction(trans, root);
		ret = wait_for_commit(root, cur_trans);
		BUG_ON(ret);
		put_transaction(cur_trans);
		mutex_unlock(&root->fs_info->trans_mutex);
		return 0;
	}
	cur_trans = trans->transaction;
	trans->transaction->in_commit = 1;
	while (trans->transaction->num_writers > 1) {
		WARN_ON(cur_trans != trans->transaction);
		prepare_to_wait(&trans->transaction->writer_wait, &wait,
				TASK_UNINTERRUPTIBLE);
		if (trans->transaction->num_writers <= 1)
			break;
		mutex_unlock(&root->fs_info->trans_mutex);
		schedule();
		mutex_lock(&root->fs_info->trans_mutex);
		finish_wait(&trans->transaction->writer_wait, &wait);
	}
	finish_wait(&trans->transaction->writer_wait, &wait);
	WARN_ON(cur_trans != trans->transaction);
	ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix,
			      &dirty_fs_roots);
	BUG_ON(ret);

	ret = btrfs_commit_tree_roots(trans, root);
	BUG_ON(ret);

	cur_trans = root->fs_info->running_transaction;
	root->fs_info->running_transaction = NULL;
	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)
			prev_trans = NULL;
		else
			prev_trans->use_count++;
	}
	mutex_unlock(&root->fs_info->trans_mutex);
	mutex_unlock(&root->fs_info->fs_mutex);
	ret = btrfs_write_and_wait_transaction(trans, root);
	if (prev_trans) {
		mutex_lock(&root->fs_info->trans_mutex);
		wait_for_commit(root, prev_trans);
		put_transaction(prev_trans);
		mutex_unlock(&root->fs_info->trans_mutex);
	}
	btrfs_set_super_generation(root->fs_info->disk_super,
				   cur_trans->transid);
	BUG_ON(ret);
	write_ctree_super(trans, root);

	mutex_lock(&root->fs_info->fs_mutex);
	btrfs_finish_extent_commit(trans, root);
	mutex_lock(&root->fs_info->trans_mutex);
	cur_trans->commit_done = 1;
	wake_up(&cur_trans->commit_wait);
	put_transaction(cur_trans);
	put_transaction(cur_trans);
	if (root->fs_info->closing)
		list_splice_init(&root->fs_info->dead_roots, &dirty_fs_roots);
	else
		list_splice_init(&dirty_fs_roots, &root->fs_info->dead_roots);
	mutex_unlock(&root->fs_info->trans_mutex);
	kmem_cache_free(btrfs_trans_handle_cachep, trans);

	if (root->fs_info->closing) {
		mutex_unlock(&root->fs_info->fs_mutex);
		drop_dirty_roots(root->fs_info->tree_root, &dirty_fs_roots);
		mutex_lock(&root->fs_info->fs_mutex);
	}
	return ret;
}

void btrfs_transaction_cleaner(struct work_struct *work)
{
	struct btrfs_fs_info *fs_info = container_of(work,
						     struct btrfs_fs_info,
						     trans_work.work);

	struct btrfs_root *root = fs_info->tree_root;
	struct btrfs_transaction *cur;
	struct btrfs_trans_handle *trans;
	struct list_head dirty_roots;
	unsigned long now;
	unsigned long delay = HZ * 30;
	int ret;

	INIT_LIST_HEAD(&dirty_roots);
	mutex_lock(&root->fs_info->fs_mutex);
	mutex_lock(&root->fs_info->trans_mutex);
	cur = root->fs_info->running_transaction;
	if (!cur) {
		mutex_unlock(&root->fs_info->trans_mutex);
		goto out;
	}
	now = get_seconds();
	if (now < cur->start_time || now - cur->start_time < 30) {
		mutex_unlock(&root->fs_info->trans_mutex);
		delay = HZ * 5;
		goto out;
	}
	mutex_unlock(&root->fs_info->trans_mutex);
	trans = btrfs_start_transaction(root, 1);
	ret = btrfs_commit_transaction(trans, root);
out:
	mutex_unlock(&root->fs_info->fs_mutex);

	mutex_lock(&root->fs_info->trans_mutex);
	list_splice_init(&root->fs_info->dead_roots, &dirty_roots);
	mutex_unlock(&root->fs_info->trans_mutex);

	if (!list_empty(&dirty_roots)) {
		drop_dirty_roots(root, &dirty_roots);
	}
	btrfs_transaction_queue_work(root, delay);
}

void btrfs_transaction_queue_work(struct btrfs_root *root, int delay)
{
	queue_delayed_work(trans_wq, &root->fs_info->trans_work, delay);
}

void btrfs_transaction_flush_work(struct btrfs_root *root)
{
	cancel_rearming_delayed_workqueue(trans_wq, &root->fs_info->trans_work);
	flush_workqueue(trans_wq);
}

void __init btrfs_init_transaction_sys(void)
{
	trans_wq = create_workqueue("btrfs");
}

void __exit btrfs_exit_transaction_sys(void)
{
	destroy_workqueue(trans_wq);
}
Commit	Line	Data
6cbd5570 CM	1	/*
	2	* Copyright (C) 2007 Oracle. All rights reserved.
	3	*
	4	* This program is free software; you can redistribute it and/or
	5	* modify it under the terms of the GNU General Public
	6	* License v2 as published by the Free Software Foundation.
	7	*
	8	* This program is distributed in the hope that it will be useful,
	9	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	10	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	11	* General Public License for more details.
	12	*
	13	* You should have received a copy of the GNU General Public
	14	* License along with this program; if not, write to the
	15	* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
	16	* Boston, MA 021110-1307, USA.
	17	*/
	18
79154b1b CM	19	#include <linux/module.h>
79154b1b CM	20	#include <linux/fs.h>
34088780	21	#include <linux/sched.h>
79154b1b CM	22	#include "ctree.h"
	23	#include "disk-io.h"
	24	#include "transaction.h"
	25
78fae27e	26	static int total_trans = 0;
2c90e5d6 CM	27	extern struct kmem_cache *btrfs_trans_handle_cachep;
	28	extern struct kmem_cache *btrfs_transaction_cachep;
	29
08607c1b CM	30	static struct workqueue_struct *trans_wq;
08607c1b CM	31
0f7d52f4 CM	32	#define BTRFS_ROOT_TRANS_TAG 0
0f7d52f4 CM	33
79154b1b CM	34	static void put_transaction(struct btrfs_transaction *transaction)
79154b1b CM	35	{
2c90e5d6	36	WARN_ON(transaction->use_count == 0);
79154b1b	37	transaction->use_count--;
78fae27e CM	38	if (transaction->use_count == 0) {
	39	WARN_ON(total_trans == 0);
	40	total_trans--;
8fd17795	41	list_del_init(&transaction->list);
2c90e5d6 CM	42	memset(transaction, 0, sizeof(*transaction));
2c90e5d6 CM	43	kmem_cache_free(btrfs_transaction_cachep, transaction);
78fae27e	44	}
79154b1b CM	45	}
	46
	47	static int join_transaction(struct btrfs_root *root)
	48	{
	49	struct btrfs_transaction *cur_trans;
	50	cur_trans = root->fs_info->running_transaction;
	51	if (!cur_trans) {
2c90e5d6 CM	52	cur_trans = kmem_cache_alloc(btrfs_transaction_cachep,
2c90e5d6 CM	53	GFP_NOFS);
78fae27e	54	total_trans++;
79154b1b	55	BUG_ON(!cur_trans);
0f7d52f4	56	root->fs_info->generation++;
79154b1b CM	57	root->fs_info->running_transaction = cur_trans;
79154b1b CM	58	cur_trans->num_writers = 0;
0f7d52f4	59	cur_trans->transid = root->fs_info->generation;
79154b1b CM	60	init_waitqueue_head(&cur_trans->writer_wait);
	61	init_waitqueue_head(&cur_trans->commit_wait);
	62	cur_trans->in_commit = 0;
d5719762	63	cur_trans->use_count = 1;
79154b1b	64	cur_trans->commit_done = 0;
08607c1b	65	cur_trans->start_time = get_seconds();
8fd17795	66	list_add_tail(&cur_trans->list, &root->fs_info->trans_list);
7c4452b9	67	init_bit_radix(&cur_trans->dirty_pages);
79154b1b CM	68	}
	69	cur_trans->num_writers++;
	70	return 0;
	71	}
	72
	73	struct btrfs_trans_handle btrfs_start_transaction(struct btrfs_root root,
	74	int num_blocks)
	75	{
2c90e5d6 CM	76	struct btrfs_trans_handle *h =
2c90e5d6 CM	77	kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS);
79154b1b	78	int ret;
0f7d52f4	79	u64 running_trans_id;
79154b1b CM	80
	81	mutex_lock(&root->fs_info->trans_mutex);
	82	ret = join_transaction(root);
	83	BUG_ON(ret);
0f7d52f4 CM	84	running_trans_id = root->fs_info->running_transaction->transid;
	85
	86	if (root != root->fs_info->tree_root && root->last_trans <
	87	running_trans_id) {
	88	radix_tree_tag_set(&root->fs_info->fs_roots_radix,
2619ba1f CM	89	(unsigned long)root->root_key.objectid,
2619ba1f CM	90	BTRFS_ROOT_TRANS_TAG);
0f7d52f4 CM	91	root->commit_root = root->node;
	92	get_bh(root->node);
	93	}
	94	root->last_trans = running_trans_id;
	95	h->transid = running_trans_id;
79154b1b CM	96	h->transaction = root->fs_info->running_transaction;
	97	h->blocks_reserved = num_blocks;
	98	h->blocks_used = 0;
31f3c99b	99	h->block_group = NULL;
79154b1b CM	100	root->fs_info->running_transaction->use_count++;
	101	mutex_unlock(&root->fs_info->trans_mutex);
	102	return h;
	103	}
	104
	105	int btrfs_end_transaction(struct btrfs_trans_handle *trans,
	106	struct btrfs_root *root)
	107	{
	108	struct btrfs_transaction *cur_trans;
d6e4a428	109
79154b1b CM	110	mutex_lock(&root->fs_info->trans_mutex);
79154b1b CM	111	cur_trans = root->fs_info->running_transaction;
d5719762	112	WARN_ON(cur_trans->num_writers < 1);
79154b1b CM	113	if (waitqueue_active(&cur_trans->writer_wait))
	114	wake_up(&cur_trans->writer_wait);
	115	cur_trans->num_writers--;
	116	put_transaction(cur_trans);
	117	mutex_unlock(&root->fs_info->trans_mutex);
d6025579	118	memset(trans, 0, sizeof(*trans));
2c90e5d6	119	kmem_cache_free(btrfs_trans_handle_cachep, trans);
79154b1b CM	120	return 0;
	121	}
	122
	123
	124	int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
	125	struct btrfs_root *root)
	126	{
7c4452b9 CM	127	unsigned long gang[16];
	128	int ret;
	129	int i;
	130	int err;
	131	int werr = 0;
	132	struct page *page;
	133	struct radix_tree_root *dirty_pages;
	134	struct inode *btree_inode = root->fs_info->btree_inode;
	135
	136	if (!trans \|\| !trans->transaction) {
	137	return filemap_write_and_wait(btree_inode->i_mapping);
	138	}
	139	dirty_pages = &trans->transaction->dirty_pages;
	140	while(1) {
e37c9e69 CM	141	ret = find_first_radix_bit(dirty_pages, gang,
e37c9e69 CM	142	0, ARRAY_SIZE(gang));
7c4452b9 CM	143	if (!ret)
	144	break;
	145	for (i = 0; i < ret; i++) {
	146	/* FIXME EIO */
	147	clear_radix_bit(dirty_pages, gang[i]);
	148	page = find_lock_page(btree_inode->i_mapping,
	149	gang[i]);
	150	if (!page)
	151	continue;
	152	err = write_one_page(page, 0);
	153	if (err)
	154	werr = err;
	155	page_cache_release(page);
	156	}
	157	}
	158	err = filemap_fdatawait(btree_inode->i_mapping);
	159	if (err)
	160	werr = err;
	161	return werr;
79154b1b CM	162	}
	163
	164	int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
	165	struct btrfs_root *root)
	166	{
	167	int ret;
	168	u64 old_extent_block;
	169	struct btrfs_fs_info *fs_info = root->fs_info;
	170	struct btrfs_root *tree_root = fs_info->tree_root;
	171	struct btrfs_root *extent_root = fs_info->extent_root;
79154b1b	172
9078a3e1	173	btrfs_write_dirty_block_groups(trans, extent_root);
79154b1b CM	174	while(1) {
79154b1b CM	175	old_extent_block = btrfs_root_blocknr(&extent_root->root_item);
7eccb903	176	if (old_extent_block == bh_blocknr(extent_root->node))
79154b1b CM	177	break;
79154b1b CM	178	btrfs_set_root_blocknr(&extent_root->root_item,
7eccb903	179	bh_blocknr(extent_root->node));
79154b1b CM	180	ret = btrfs_update_root(trans, tree_root,
	181	&extent_root->root_key,
	182	&extent_root->root_item);
	183	BUG_ON(ret);
9078a3e1	184	btrfs_write_dirty_block_groups(trans, extent_root);
79154b1b CM	185	}
	186	return 0;
	187	}
	188
	189	static int wait_for_commit(struct btrfs_root *root,
	190	struct btrfs_transaction *commit)
	191	{
	192	DEFINE_WAIT(wait);
79154b1b CM	193	while(!commit->commit_done) {
	194	prepare_to_wait(&commit->commit_wait, &wait,
	195	TASK_UNINTERRUPTIBLE);
	196	if (commit->commit_done)
	197	break;
	198	mutex_unlock(&root->fs_info->trans_mutex);
	199	schedule();
	200	mutex_lock(&root->fs_info->trans_mutex);
	201	}
	202	finish_wait(&commit->commit_wait, &wait);
	203	return 0;
	204	}
	205
0f7d52f4 CM	206	struct dirty_root {
	207	struct list_head list;
	208	struct btrfs_key snap_key;
	209	struct buffer_head *commit_root;
	210	struct btrfs_root *root;
	211	};
	212
35b7e476 CM	213	static int add_dirty_roots(struct btrfs_trans_handle *trans,
	214	struct radix_tree_root *radix,
	215	struct list_head *list)
0f7d52f4 CM	216	{
	217	struct dirty_root *dirty;
	218	struct btrfs_root *gang[8];
	219	struct btrfs_root *root;
	220	int i;
	221	int ret;
54aa1f4d CM	222	int err = 0;
54aa1f4d CM	223
0f7d52f4 CM	224	while(1) {
	225	ret = radix_tree_gang_lookup_tag(radix, (void **)gang, 0,
	226	ARRAY_SIZE(gang),
	227	BTRFS_ROOT_TRANS_TAG);
	228	if (ret == 0)
	229	break;
	230	for (i = 0; i < ret; i++) {
	231	root = gang[i];
2619ba1f CM	232	radix_tree_tag_clear(radix,
	233	(unsigned long)root->root_key.objectid,
	234	BTRFS_ROOT_TRANS_TAG);
0f7d52f4	235	if (root->commit_root == root->node) {
7eccb903	236	WARN_ON(bh_blocknr(root->node) !=
0f7d52f4 CM	237	btrfs_root_blocknr(&root->root_item));
	238	brelse(root->commit_root);
	239	root->commit_root = NULL;
	240	continue;
	241	}
	242	dirty = kmalloc(sizeof(*dirty), GFP_NOFS);
	243	BUG_ON(!dirty);
	244	memcpy(&dirty->snap_key, &root->root_key,
	245	sizeof(root->root_key));
	246	dirty->commit_root = root->commit_root;
	247	root->commit_root = NULL;
	248	dirty->root = root;
0f7d52f4 CM	249	root->root_key.offset = root->fs_info->generation;
0f7d52f4 CM	250	btrfs_set_root_blocknr(&root->root_item,
7eccb903	251	bh_blocknr(root->node));
0f7d52f4 CM	252	err = btrfs_insert_root(trans, root->fs_info->tree_root,
	253	&root->root_key,
	254	&root->root_item);
54aa1f4d CM	255	if (err)
54aa1f4d CM	256	break;
0f7d52f4 CM	257	list_add(&dirty->list, list);
	258	}
	259	}
54aa1f4d	260	return err;
0f7d52f4 CM	261	}
0f7d52f4 CM	262
35b7e476 CM	263	static int drop_dirty_roots(struct btrfs_root *tree_root,
35b7e476 CM	264	struct list_head *list)
0f7d52f4 CM	265	{
	266	struct dirty_root *dirty;
	267	struct btrfs_trans_handle *trans;
54aa1f4d	268	int ret = 0;
0f7d52f4	269	while(!list_empty(list)) {
facda1e7	270	mutex_lock(&tree_root->fs_info->fs_mutex);
0f7d52f4 CM	271	dirty = list_entry(list->next, struct dirty_root, list);
	272	list_del_init(&dirty->list);
	273	trans = btrfs_start_transaction(tree_root, 1);
0f7d52f4 CM	274	ret = btrfs_drop_snapshot(trans, dirty->root,
	275	dirty->commit_root);
	276	BUG_ON(ret);
	277
0f7d52f4	278	ret = btrfs_del_root(trans, tree_root, &dirty->snap_key);
54aa1f4d CM	279	if (ret)
54aa1f4d CM	280	break;
0f7d52f4 CM	281	ret = btrfs_end_transaction(trans, tree_root);
	282	BUG_ON(ret);
	283	kfree(dirty);
facda1e7	284	mutex_unlock(&tree_root->fs_info->fs_mutex);
8c2383c3	285	btrfs_btree_balance_dirty(tree_root);
0f7d52f4	286	}
54aa1f4d	287	return ret;
0f7d52f4 CM	288	}
0f7d52f4 CM	289
79154b1b CM	290	int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
	291	struct btrfs_root *root)
	292	{
	293	int ret = 0;
79154b1b	294	struct btrfs_transaction *cur_trans;
8fd17795	295	struct btrfs_transaction *prev_trans = NULL;
0f7d52f4	296	struct list_head dirty_fs_roots;
79154b1b CM	297	DEFINE_WAIT(wait);
79154b1b CM	298
0f7d52f4	299	INIT_LIST_HEAD(&dirty_fs_roots);
d6e4a428	300
79154b1b CM	301	mutex_lock(&root->fs_info->trans_mutex);
	302	if (trans->transaction->in_commit) {
	303	cur_trans = trans->transaction;
	304	trans->transaction->use_count++;
	305	btrfs_end_transaction(trans, root);
	306	ret = wait_for_commit(root, cur_trans);
	307	BUG_ON(ret);
	308	put_transaction(cur_trans);
	309	mutex_unlock(&root->fs_info->trans_mutex);
	310	return 0;
	311	}
2c90e5d6 CM	312	cur_trans = trans->transaction;
2c90e5d6 CM	313	trans->transaction->in_commit = 1;
79154b1b	314	while (trans->transaction->num_writers > 1) {
2c90e5d6	315	WARN_ON(cur_trans != trans->transaction);
79154b1b CM	316	prepare_to_wait(&trans->transaction->writer_wait, &wait,
	317	TASK_UNINTERRUPTIBLE);
	318	if (trans->transaction->num_writers <= 1)
	319	break;
	320	mutex_unlock(&root->fs_info->trans_mutex);
	321	schedule();
	322	mutex_lock(&root->fs_info->trans_mutex);
2c90e5d6	323	finish_wait(&trans->transaction->writer_wait, &wait);
79154b1b CM	324	}
79154b1b CM	325	finish_wait(&trans->transaction->writer_wait, &wait);
2c90e5d6	326	WARN_ON(cur_trans != trans->transaction);
54aa1f4d CM	327	ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix,
	328	&dirty_fs_roots);
	329	BUG_ON(ret);
	330
79154b1b CM	331	ret = btrfs_commit_tree_roots(trans, root);
79154b1b CM	332	BUG_ON(ret);
54aa1f4d	333
78fae27e CM	334	cur_trans = root->fs_info->running_transaction;
78fae27e CM	335	root->fs_info->running_transaction = NULL;
8fd17795 CM	336	if (cur_trans->list.prev != &root->fs_info->trans_list) {
	337	prev_trans = list_entry(cur_trans->list.prev,
	338	struct btrfs_transaction, list);
	339	if (prev_trans->commit_done)
	340	prev_trans = NULL;
	341	else
	342	prev_trans->use_count++;
	343	}
78fae27e	344	mutex_unlock(&root->fs_info->trans_mutex);
8fd17795	345	mutex_unlock(&root->fs_info->fs_mutex);
79154b1b	346	ret = btrfs_write_and_wait_transaction(trans, root);
8fd17795 CM	347	if (prev_trans) {
	348	mutex_lock(&root->fs_info->trans_mutex);
	349	wait_for_commit(root, prev_trans);
	350	put_transaction(prev_trans);
	351	mutex_unlock(&root->fs_info->trans_mutex);
	352	}
	353	btrfs_set_super_generation(root->fs_info->disk_super,
	354	cur_trans->transid);
79154b1b	355	BUG_ON(ret);
79154b1b	356	write_ctree_super(trans, root);
8fd17795 CM	357
8fd17795 CM	358	mutex_lock(&root->fs_info->fs_mutex);
78fae27e CM	359	btrfs_finish_extent_commit(trans, root);
78fae27e CM	360	mutex_lock(&root->fs_info->trans_mutex);
2c90e5d6 CM	361	cur_trans->commit_done = 1;
2c90e5d6 CM	362	wake_up(&cur_trans->commit_wait);
78fae27e	363	put_transaction(cur_trans);
79154b1b	364	put_transaction(cur_trans);
facda1e7 CM	365	if (root->fs_info->closing)
	366	list_splice_init(&root->fs_info->dead_roots, &dirty_fs_roots);
	367	else
	368	list_splice_init(&dirty_fs_roots, &root->fs_info->dead_roots);
78fae27e	369	mutex_unlock(&root->fs_info->trans_mutex);
2c90e5d6	370	kmem_cache_free(btrfs_trans_handle_cachep, trans);
79154b1b	371
facda1e7 CM	372	if (root->fs_info->closing) {
	373	mutex_unlock(&root->fs_info->fs_mutex);
	374	drop_dirty_roots(root->fs_info->tree_root, &dirty_fs_roots);
	375	mutex_lock(&root->fs_info->fs_mutex);
	376	}
79154b1b CM	377	return ret;
	378	}
	379
08607c1b CM	380	void btrfs_transaction_cleaner(struct work_struct *work)
	381	{
	382	struct btrfs_fs_info *fs_info = container_of(work,
	383	struct btrfs_fs_info,
	384	trans_work.work);
	385
	386	struct btrfs_root *root = fs_info->tree_root;
	387	struct btrfs_transaction *cur;
	388	struct btrfs_trans_handle *trans;
facda1e7	389	struct list_head dirty_roots;
08607c1b CM	390	unsigned long now;
	391	unsigned long delay = HZ * 30;
	392	int ret;
	393
facda1e7	394	INIT_LIST_HEAD(&dirty_roots);
08607c1b CM	395	mutex_lock(&root->fs_info->fs_mutex);
	396	mutex_lock(&root->fs_info->trans_mutex);
	397	cur = root->fs_info->running_transaction;
	398	if (!cur) {
	399	mutex_unlock(&root->fs_info->trans_mutex);
	400	goto out;
	401	}
	402	now = get_seconds();
	403	if (now < cur->start_time \|\| now - cur->start_time < 30) {
	404	mutex_unlock(&root->fs_info->trans_mutex);
	405	delay = HZ * 5;
	406	goto out;
	407	}
	408	mutex_unlock(&root->fs_info->trans_mutex);
08607c1b CM	409	trans = btrfs_start_transaction(root, 1);
	410	ret = btrfs_commit_transaction(trans, root);
	411	out:
	412	mutex_unlock(&root->fs_info->fs_mutex);
ad693af6 CM	413
	414	mutex_lock(&root->fs_info->trans_mutex);
	415	list_splice_init(&root->fs_info->dead_roots, &dirty_roots);
	416	mutex_unlock(&root->fs_info->trans_mutex);
	417
	418	if (!list_empty(&dirty_roots)) {
	419	drop_dirty_roots(root, &dirty_roots);
	420	}
08607c1b CM	421	btrfs_transaction_queue_work(root, delay);
	422	}
	423
	424	void btrfs_transaction_queue_work(struct btrfs_root *root, int delay)
	425	{
	426	queue_delayed_work(trans_wq, &root->fs_info->trans_work, delay);
	427	}
	428
	429	void btrfs_transaction_flush_work(struct btrfs_root *root)
	430	{
	431	cancel_rearming_delayed_workqueue(trans_wq, &root->fs_info->trans_work);
	432	flush_workqueue(trans_wq);
	433	}
	434
	435	void __init btrfs_init_transaction_sys(void)
	436	{
	437	trans_wq = create_workqueue("btrfs");
	438	}
	439
	440	void __exit btrfs_exit_transaction_sys(void)
	441	{
	442	destroy_workqueue(trans_wq);
	443	}
	444