[mirror_ubuntu-jammy-kernel.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 "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;
	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);
			BUG_ON(err);
			list_add(&dirty->list, list);
		}
	}
	return 0;
}

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;
	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);
		BUG_ON(ret);
		ret = btrfs_end_transaction(trans, tree_root);
		BUG_ON(ret);
		kfree(dirty);
		mutex_unlock(&tree_root->fs_info->fs_mutex);
	}
	return 0;
}

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