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

#include <linux/module.h>
#include <linux/fs.h>
#include <linux/blkdev.h>
#include <linux/crypto.h>
#include <linux/scatterlist.h>
#include <linux/swap.h>
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"

static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
{
	struct btrfs_node *node = btrfs_buffer_node(buf);
	if (buf->b_blocknr != btrfs_header_blocknr(&node->header)) {
		BUG();
	}
	if (root->node && btrfs_header_parentid(&node->header) !=
	    btrfs_header_parentid(btrfs_buffer_header(root->node))) {
		BUG();
	}
	return 0;
}

struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr)
{
	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
	int blockbits = root->fs_info->sb->s_blocksize_bits;
	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
	struct page *page;
	struct buffer_head *bh;
	struct buffer_head *head;
	struct buffer_head *ret = NULL;


	page = find_lock_page(mapping, index);
	if (!page)
		return NULL;

	if (!page_has_buffers(page))
		goto out_unlock;

	head = page_buffers(page);
	bh = head;
	do {
		if (buffer_mapped(bh) && bh->b_blocknr == blocknr) {
			ret = bh;
			get_bh(bh);
			goto out_unlock;
		}
		bh = bh->b_this_page;
	} while (bh != head);
out_unlock:
	unlock_page(page);
	if (ret) {
		touch_buffer(ret);
	}
	page_cache_release(page);
	return ret;
}

struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
						 u64 blocknr)
{
	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
	int blockbits = root->fs_info->sb->s_blocksize_bits;
	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
	struct page *page;
	struct buffer_head *bh;
	struct buffer_head *head;
	struct buffer_head *ret = NULL;
	u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);

	page = grab_cache_page(mapping, index);
	if (!page)
		return NULL;

	if (!page_has_buffers(page))
		create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
	head = page_buffers(page);
	bh = head;
	do {
		if (!buffer_mapped(bh)) {
			bh->b_bdev = root->fs_info->sb->s_bdev;
			bh->b_blocknr = first_block;
			set_buffer_mapped(bh);
		}
		if (bh->b_blocknr == blocknr) {
			ret = bh;
			get_bh(bh);
			goto out_unlock;
		}
		bh = bh->b_this_page;
		first_block++;
	} while (bh != head);
out_unlock:
	unlock_page(page);
	if (ret)
		touch_buffer(ret);
	page_cache_release(page);
	return ret;
}

static sector_t max_block(struct block_device *bdev)
{
	sector_t retval = ~((sector_t)0);
	loff_t sz = i_size_read(bdev->bd_inode);

	if (sz) {
		unsigned int size = block_size(bdev);
		unsigned int sizebits = blksize_bits(size);
		retval = (sz >> sizebits);
	}
	return retval;
}

static int btree_get_block(struct inode *inode, sector_t iblock,
			   struct buffer_head *bh, int create)
{
	if (iblock >= max_block(inode->i_sb->s_bdev)) {
		if (create)
			return -EIO;

		/*
		 * for reads, we're just trying to fill a partial page.
		 * return a hole, they will have to call get_block again
		 * before they can fill it, and they will get -EIO at that
		 * time
		 */
		return 0;
	}
	bh->b_bdev = inode->i_sb->s_bdev;
	bh->b_blocknr = iblock;
	set_buffer_mapped(bh);
	return 0;
}

int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
		    char *result)
{
	struct scatterlist sg;
	struct crypto_hash *tfm = root->fs_info->hash_tfm;
	struct hash_desc desc;
	int ret;

	desc.tfm = tfm;
	desc.flags = 0;
	sg_init_one(&sg, data, len);
	spin_lock(&root->fs_info->hash_lock);
	ret = crypto_hash_digest(&desc, &sg, 1, result);
	spin_unlock(&root->fs_info->hash_lock);
	if (ret) {
		printk("sha256 digest failed\n");
	}
	return ret;
}
static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
			   int verify)
{
	char result[BTRFS_CSUM_SIZE];
	int ret;
	struct btrfs_node *node;

	ret = btrfs_csum_data(root, bh->b_data + BTRFS_CSUM_SIZE,
			      bh->b_size - BTRFS_CSUM_SIZE, result);
	if (ret)
		return ret;
	if (verify) {
		if (memcmp(bh->b_data, result, BTRFS_CSUM_SIZE)) {
			printk("checksum verify failed on %lu\n",
			       bh->b_blocknr);
			return 1;
		}
	} else {
		node = btrfs_buffer_node(bh);
		memcpy(node->header.csum, result, BTRFS_CSUM_SIZE);
	}
	return 0;
}

static int btree_writepage(struct page *page, struct writeback_control *wbc)
{
	struct buffer_head *bh;
	struct btrfs_root *root = btrfs_sb(page->mapping->host->i_sb);
	struct buffer_head *head;
	if (!page_has_buffers(page)) {
		create_empty_buffers(page, root->fs_info->sb->s_blocksize,
					(1 << BH_Dirty)|(1 << BH_Uptodate));
	}
	head = page_buffers(page);
	bh = head;
	do {
		if (buffer_dirty(bh))
			csum_tree_block(root, bh, 0);
		bh = bh->b_this_page;
	} while (bh != head);
	return block_write_full_page(page, btree_get_block, wbc);
}

static int btree_readpage(struct file * file, struct page * page)
{
	return block_read_full_page(page, btree_get_block);
}

static struct address_space_operations btree_aops = {
	.readpage	= btree_readpage,
	.writepage	= btree_writepage,
	.sync_page	= block_sync_page,
};

struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
{
	struct buffer_head *bh = NULL;

	bh = btrfs_find_create_tree_block(root, blocknr);
	if (!bh)
		return bh;
	if (buffer_uptodate(bh))
		goto uptodate;
	lock_buffer(bh);
	if (!buffer_uptodate(bh)) {
		get_bh(bh);
		bh->b_end_io = end_buffer_read_sync;
		submit_bh(READ, bh);
		wait_on_buffer(bh);
		if (!buffer_uptodate(bh))
			goto fail;
		csum_tree_block(root, bh, 1);
	} else {
		unlock_buffer(bh);
	}
uptodate:
	if (check_tree_block(root, bh))
		BUG();
	return bh;
fail:
	brelse(bh);
	return NULL;
}

int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
		     struct buffer_head *buf)
{
	WARN_ON(atomic_read(&buf->b_count) == 0);
	mark_buffer_dirty(buf);
	return 0;
}

int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
		     struct buffer_head *buf)
{
	WARN_ON(atomic_read(&buf->b_count) == 0);
	clear_buffer_dirty(buf);
	return 0;
}

static int __setup_root(int blocksize,
			struct btrfs_root *root,
			struct btrfs_fs_info *fs_info,
			u64 objectid)
{
	root->node = NULL;
	root->commit_root = NULL;
	root->blocksize = blocksize;
	root->ref_cows = 0;
	root->fs_info = fs_info;
	memset(&root->root_key, 0, sizeof(root->root_key));
	memset(&root->root_item, 0, sizeof(root->root_item));
	return 0;
}

static int find_and_setup_root(int blocksize,
			       struct btrfs_root *tree_root,
			       struct btrfs_fs_info *fs_info,
			       u64 objectid,
			       struct btrfs_root *root)
{
	int ret;

	__setup_root(blocksize, root, fs_info, objectid);
	ret = btrfs_find_last_root(tree_root, objectid,
				   &root->root_item, &root->root_key);
	BUG_ON(ret);

	root->node = read_tree_block(root,
				     btrfs_root_blocknr(&root->root_item));
	BUG_ON(!root->node);
	return 0;
}

struct btrfs_root *open_ctree(struct super_block *sb)
{
	struct btrfs_root *root = kmalloc(sizeof(struct btrfs_root),
					  GFP_NOFS);
	struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
						 GFP_NOFS);
	struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
					       GFP_NOFS);
	struct btrfs_root *inode_root = kmalloc(sizeof(struct btrfs_root),
						GFP_NOFS);
	struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
						GFP_NOFS);
	int ret;
	struct btrfs_super_block *disk_super;

	init_bit_radix(&fs_info->pinned_radix);
	init_bit_radix(&fs_info->pending_del_radix);
	sb_set_blocksize(sb, 4096);
	fs_info->running_transaction = NULL;
	fs_info->fs_root = root;
	fs_info->tree_root = tree_root;
	fs_info->extent_root = extent_root;
	fs_info->inode_root = inode_root;
	fs_info->last_inode_alloc = 0;
	fs_info->highest_inode = 0;
	fs_info->sb = sb;
	fs_info->btree_inode = new_inode(sb);
	fs_info->btree_inode->i_ino = 1;
	fs_info->btree_inode->i_nlink = 1;
	fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
	insert_inode_hash(fs_info->btree_inode);
	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
	fs_info->hash_tfm = crypto_alloc_hash("sha256", 0, CRYPTO_ALG_ASYNC);
	spin_lock_init(&fs_info->hash_lock);
	if (!fs_info->hash_tfm || IS_ERR(fs_info->hash_tfm)) {
		printk("failed to allocate sha256 hash\n");
		return NULL;
	}
	mutex_init(&fs_info->trans_mutex);
	mutex_init(&fs_info->fs_mutex);
	memset(&fs_info->current_insert, 0, sizeof(fs_info->current_insert));
	memset(&fs_info->last_insert, 0, sizeof(fs_info->last_insert));

	__setup_root(sb->s_blocksize, tree_root,
		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
	fs_info->sb_buffer = read_tree_block(tree_root,
					     BTRFS_SUPER_INFO_OFFSET /
					     sb->s_blocksize);

	if (!fs_info->sb_buffer) {
		return NULL;
	}
	disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
	if (!btrfs_super_root(disk_super)) {
		return NULL;
	}
	fs_info->disk_super = disk_super;
	tree_root->node = read_tree_block(tree_root,
					  btrfs_super_root(disk_super));
	BUG_ON(!tree_root->node);

	mutex_lock(&fs_info->fs_mutex);
	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
	BUG_ON(ret);

	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
				  BTRFS_INODE_MAP_OBJECTID, inode_root);
	BUG_ON(ret);

	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
				  BTRFS_FS_TREE_OBJECTID, root);
	BUG_ON(ret);
	root->commit_root = root->node;
	get_bh(root->node);
	root->ref_cows = 1;
	root->fs_info->generation = root->root_key.offset + 1;
	ret = btrfs_find_highest_inode(root, &root->fs_info->last_inode_alloc);
	if (ret == 0)
		fs_info->highest_inode = fs_info->last_inode_alloc;
	memset(&fs_info->kobj, 0, sizeof(fs_info->kobj));
	kobj_set_kset_s(fs_info, btrfs_subsys);
	kobject_set_name(&fs_info->kobj, "%s", sb->s_id);
	kobject_register(&fs_info->kobj);
	mutex_unlock(&fs_info->fs_mutex);
	return root;
}

int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
		      *root)
{
	struct buffer_head *bh = root->fs_info->sb_buffer;

	btrfs_set_super_root(root->fs_info->disk_super,
			     root->fs_info->tree_root->node->b_blocknr);
	lock_buffer(bh);
	WARN_ON(atomic_read(&bh->b_count) < 1);
	clear_buffer_dirty(bh);
	csum_tree_block(root, bh, 0);
	bh->b_end_io = end_buffer_write_sync;
	get_bh(bh);
	submit_bh(WRITE, bh);
	wait_on_buffer(bh);
	if (!buffer_uptodate(bh)) {
		WARN_ON(1);
		return -EIO;
	}
	return 0;
}

int close_ctree(struct btrfs_root *root)
{
	int ret;
	struct btrfs_trans_handle *trans;

	mutex_lock(&root->fs_info->fs_mutex);
	trans = btrfs_start_transaction(root, 1);
	btrfs_commit_transaction(trans, root);
	/* run commit again to  drop the original snapshot */
	trans = btrfs_start_transaction(root, 1);
	btrfs_commit_transaction(trans, root);
	ret = btrfs_write_and_wait_transaction(NULL, root);
	BUG_ON(ret);
	write_ctree_super(NULL, root);
	mutex_unlock(&root->fs_info->fs_mutex);

	if (root->node)
		btrfs_block_release(root, root->node);
	if (root->fs_info->extent_root->node)
		btrfs_block_release(root->fs_info->extent_root,
				    root->fs_info->extent_root->node);
	if (root->fs_info->inode_root->node)
		btrfs_block_release(root->fs_info->inode_root,
				    root->fs_info->inode_root->node);
	if (root->fs_info->tree_root->node)
		btrfs_block_release(root->fs_info->tree_root,
				    root->fs_info->tree_root->node);
	btrfs_block_release(root, root->commit_root);
	btrfs_block_release(root, root->fs_info->sb_buffer);
	crypto_free_hash(root->fs_info->hash_tfm);
	truncate_inode_pages(root->fs_info->btree_inode->i_mapping, 0);
	iput(root->fs_info->btree_inode);
	kfree(root->fs_info->extent_root);
	kfree(root->fs_info->inode_root);
	kfree(root->fs_info->tree_root);
	kobject_unregister(&root->fs_info->kobj);
	kfree(root);
	return 0;
}

void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
{
	brelse(buf);
}
Commit	Line	Data
e20d96d6 CM	1	#include <linux/module.h>
e20d96d6 CM	2	#include <linux/fs.h>
d98237b3	3	#include <linux/blkdev.h>
87cbda5c CM	4	#include <linux/crypto.h>
87cbda5c CM	5	#include <linux/scatterlist.h>
22b0ebda	6	#include <linux/swap.h>
eb60ceac CM	7	#include "ctree.h"
eb60ceac CM	8	#include "disk-io.h"
e089f05c	9	#include "transaction.h"
eb60ceac	10
e20d96d6	11	static int check_tree_block(struct btrfs_root root, struct buffer_head buf)
eb60ceac	12	{
e20d96d6	13	struct btrfs_node *node = btrfs_buffer_node(buf);
d98237b3	14	if (buf->b_blocknr != btrfs_header_blocknr(&node->header)) {
9a8dd150	15	BUG();
d98237b3	16	}
e20d96d6	17	if (root->node && btrfs_header_parentid(&node->header) !=
df2ce34c	18	btrfs_header_parentid(btrfs_buffer_header(root->node))) {
7f5c1516	19	BUG();
df2ce34c	20	}
9a8dd150	21	return 0;
eb60ceac CM	22	}
eb60ceac CM	23
d98237b3 CM	24	struct buffer_head btrfs_find_tree_block(struct btrfs_root root, u64 blocknr)
	25	{
	26	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
	27	int blockbits = root->fs_info->sb->s_blocksize_bits;
	28	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
	29	struct page *page;
	30	struct buffer_head *bh;
	31	struct buffer_head *head;
	32	struct buffer_head *ret = NULL;
	33
2c90e5d6	34
d98237b3 CM	35	page = find_lock_page(mapping, index);
	36	if (!page)
	37	return NULL;
	38
	39	if (!page_has_buffers(page))
	40	goto out_unlock;
	41
	42	head = page_buffers(page);
	43	bh = head;
	44	do {
	45	if (buffer_mapped(bh) && bh->b_blocknr == blocknr) {
	46	ret = bh;
	47	get_bh(bh);
	48	goto out_unlock;
	49	}
	50	bh = bh->b_this_page;
	51	} while (bh != head);
	52	out_unlock:
	53	unlock_page(page);
d6025579	54	if (ret) {
22b0ebda	55	touch_buffer(ret);
d6025579	56	}
d98237b3 CM	57	page_cache_release(page);
	58	return ret;
	59	}
	60
	61	struct buffer_head btrfs_find_create_tree_block(struct btrfs_root root,
	62	u64 blocknr)
	63	{
	64	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
	65	int blockbits = root->fs_info->sb->s_blocksize_bits;
	66	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
	67	struct page *page;
	68	struct buffer_head *bh;
	69	struct buffer_head *head;
	70	struct buffer_head *ret = NULL;
	71	u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
22b0ebda	72
d98237b3 CM	73	page = grab_cache_page(mapping, index);
	74	if (!page)
	75	return NULL;
	76
d98237b3 CM	77	if (!page_has_buffers(page))
	78	create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
	79	head = page_buffers(page);
	80	bh = head;
	81	do {
	82	if (!buffer_mapped(bh)) {
	83	bh->b_bdev = root->fs_info->sb->s_bdev;
	84	bh->b_blocknr = first_block;
	85	set_buffer_mapped(bh);
	86	}
	87	if (bh->b_blocknr == blocknr) {
	88	ret = bh;
	89	get_bh(bh);
	90	goto out_unlock;
	91	}
	92	bh = bh->b_this_page;
	93	first_block++;
	94	} while (bh != head);
	95	out_unlock:
	96	unlock_page(page);
22b0ebda CM	97	if (ret)
22b0ebda CM	98	touch_buffer(ret);
d98237b3 CM	99	page_cache_release(page);
	100	return ret;
	101	}
	102
	103	static sector_t max_block(struct block_device *bdev)
	104	{
	105	sector_t retval = ~((sector_t)0);
	106	loff_t sz = i_size_read(bdev->bd_inode);
	107
	108	if (sz) {
	109	unsigned int size = block_size(bdev);
	110	unsigned int sizebits = blksize_bits(size);
	111	retval = (sz >> sizebits);
	112	}
	113	return retval;
	114	}
	115
	116	static int btree_get_block(struct inode *inode, sector_t iblock,
	117	struct buffer_head *bh, int create)
	118	{
	119	if (iblock >= max_block(inode->i_sb->s_bdev)) {
	120	if (create)
	121	return -EIO;
	122
	123	/*
	124	* for reads, we're just trying to fill a partial page.
	125	* return a hole, they will have to call get_block again
	126	* before they can fill it, and they will get -EIO at that
	127	* time
	128	*/
	129	return 0;
	130	}
	131	bh->b_bdev = inode->i_sb->s_bdev;
	132	bh->b_blocknr = iblock;
	133	set_buffer_mapped(bh);
	134	return 0;
	135	}
	136
f254e52c CM	137	int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
f254e52c CM	138	char *result)
87cbda5c	139	{
87cbda5c CM	140	struct scatterlist sg;
	141	struct crypto_hash *tfm = root->fs_info->hash_tfm;
	142	struct hash_desc desc;
	143	int ret;
87cbda5c CM	144
	145	desc.tfm = tfm;
	146	desc.flags = 0;
f254e52c	147	sg_init_one(&sg, data, len);
87cbda5c	148	spin_lock(&root->fs_info->hash_lock);
22b0ebda	149	ret = crypto_hash_digest(&desc, &sg, 1, result);
87cbda5c CM	150	spin_unlock(&root->fs_info->hash_lock);
	151	if (ret) {
	152	printk("sha256 digest failed\n");
	153	}
f254e52c CM	154	return ret;
	155	}
	156	static int csum_tree_block(struct btrfs_root root, struct buffer_head bh,
	157	int verify)
	158	{
	159	char result[BTRFS_CSUM_SIZE];
	160	int ret;
	161	struct btrfs_node *node;
	162
	163	ret = btrfs_csum_data(root, bh->b_data + BTRFS_CSUM_SIZE,
	164	bh->b_size - BTRFS_CSUM_SIZE, result);
	165	if (ret)
	166	return ret;
87cbda5c	167	if (verify) {
f254e52c CM	168	if (memcmp(bh->b_data, result, BTRFS_CSUM_SIZE)) {
	169	printk("checksum verify failed on %lu\n",
	170	bh->b_blocknr);
	171	return 1;
	172	}
	173	} else {
	174	node = btrfs_buffer_node(bh);
22b0ebda	175	memcpy(node->header.csum, result, BTRFS_CSUM_SIZE);
f254e52c	176	}
87cbda5c CM	177	return 0;
	178	}
	179
d98237b3	180	static int btree_writepage(struct page page, struct writeback_control wbc)
ed2ff2cb	181	{
87cbda5c CM	182	struct buffer_head *bh;
	183	struct btrfs_root *root = btrfs_sb(page->mapping->host->i_sb);
	184	struct buffer_head *head;
87cbda5c CM	185	if (!page_has_buffers(page)) {
	186	create_empty_buffers(page, root->fs_info->sb->s_blocksize,
	187	(1 << BH_Dirty)\|(1 << BH_Uptodate));
	188	}
	189	head = page_buffers(page);
	190	bh = head;
	191	do {
	192	if (buffer_dirty(bh))
	193	csum_tree_block(root, bh, 0);
	194	bh = bh->b_this_page;
	195	} while (bh != head);
d98237b3	196	return block_write_full_page(page, btree_get_block, wbc);
ed2ff2cb CM	197	}
ed2ff2cb CM	198
d98237b3	199	static int btree_readpage(struct file * file, struct page * page)
eb60ceac	200	{
d98237b3	201	return block_read_full_page(page, btree_get_block);
eb60ceac CM	202	}
eb60ceac CM	203
d98237b3 CM	204	static struct address_space_operations btree_aops = {
	205	.readpage = btree_readpage,
	206	.writepage = btree_writepage,
	207	.sync_page = block_sync_page,
	208	};
	209
e20d96d6	210	struct buffer_head read_tree_block(struct btrfs_root root, u64 blocknr)
eb60ceac	211	{
d98237b3	212	struct buffer_head *bh = NULL;
eb60ceac	213
d98237b3 CM	214	bh = btrfs_find_create_tree_block(root, blocknr);
	215	if (!bh)
	216	return bh;
9d64272c CM	217	if (buffer_uptodate(bh))
9d64272c CM	218	goto uptodate;
d98237b3 CM	219	lock_buffer(bh);
	220	if (!buffer_uptodate(bh)) {
	221	get_bh(bh);
	222	bh->b_end_io = end_buffer_read_sync;
	223	submit_bh(READ, bh);
	224	wait_on_buffer(bh);
	225	if (!buffer_uptodate(bh))
	226	goto fail;
87cbda5c	227	csum_tree_block(root, bh, 1);
d98237b3 CM	228	} else {
	229	unlock_buffer(bh);
	230	}
9d64272c	231	uptodate:
d98237b3	232	if (check_tree_block(root, bh))
cfaa7295	233	BUG();
d98237b3 CM	234	return bh;
	235	fail:
	236	brelse(bh);
	237	return NULL;
eb60ceac CM	238	}
eb60ceac CM	239
e089f05c	240	int dirty_tree_block(struct btrfs_trans_handle trans, struct btrfs_root root,
e20d96d6	241	struct buffer_head *buf)
ed2ff2cb	242	{
d6025579	243	WARN_ON(atomic_read(&buf->b_count) == 0);
e20d96d6	244	mark_buffer_dirty(buf);
ed2ff2cb CM	245	return 0;
	246	}
	247
e089f05c	248	int clean_tree_block(struct btrfs_trans_handle trans, struct btrfs_root root,
e20d96d6	249	struct buffer_head *buf)
ed2ff2cb	250	{
d6025579	251	WARN_ON(atomic_read(&buf->b_count) == 0);
e20d96d6	252	clear_buffer_dirty(buf);
ed2ff2cb CM	253	return 0;
	254	}
	255
2c90e5d6	256	static int __setup_root(int blocksize,
9f5fae2f CM	257	struct btrfs_root *root,
9f5fae2f CM	258	struct btrfs_fs_info *fs_info,
e20d96d6	259	u64 objectid)
d97e63b6	260	{
cfaa7295	261	root->node = NULL;
a28ec197	262	root->commit_root = NULL;
2c90e5d6	263	root->blocksize = blocksize;
123abc88	264	root->ref_cows = 0;
9f5fae2f	265	root->fs_info = fs_info;
3768f368 CM	266	memset(&root->root_key, 0, sizeof(root->root_key));
	267	memset(&root->root_item, 0, sizeof(root->root_item));
	268	return 0;
	269	}
	270
2c90e5d6	271	static int find_and_setup_root(int blocksize,
9f5fae2f CM	272	struct btrfs_root *tree_root,
	273	struct btrfs_fs_info *fs_info,
	274	u64 objectid,
e20d96d6	275	struct btrfs_root *root)
3768f368 CM	276	{
	277	int ret;
	278
2c90e5d6	279	__setup_root(blocksize, root, fs_info, objectid);
3768f368 CM	280	ret = btrfs_find_last_root(tree_root, objectid,
	281	&root->root_item, &root->root_key);
	282	BUG_ON(ret);
	283
	284	root->node = read_tree_block(root,
	285	btrfs_root_blocknr(&root->root_item));
3768f368	286	BUG_ON(!root->node);
d97e63b6 CM	287	return 0;
	288	}
	289
2c90e5d6	290	struct btrfs_root open_ctree(struct super_block sb)
2e635a27	291	{
e20d96d6 CM	292	struct btrfs_root *root = kmalloc(sizeof(struct btrfs_root),
	293	GFP_NOFS);
	294	struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
	295	GFP_NOFS);
	296	struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
	297	GFP_NOFS);
	298	struct btrfs_root *inode_root = kmalloc(sizeof(struct btrfs_root),
	299	GFP_NOFS);
	300	struct btrfs_fs_info fs_info = kmalloc(sizeof(fs_info),
	301	GFP_NOFS);
eb60ceac	302	int ret;
2c90e5d6	303	struct btrfs_super_block *disk_super;
eb60ceac	304
8ef97622 CM	305	init_bit_radix(&fs_info->pinned_radix);
8ef97622 CM	306	init_bit_radix(&fs_info->pending_del_radix);
2c90e5d6	307	sb_set_blocksize(sb, 4096);
9f5fae2f CM	308	fs_info->running_transaction = NULL;
	309	fs_info->fs_root = root;
	310	fs_info->tree_root = tree_root;
	311	fs_info->extent_root = extent_root;
	312	fs_info->inode_root = inode_root;
	313	fs_info->last_inode_alloc = 0;
5be6f7f1	314	fs_info->highest_inode = 0;
e20d96d6	315	fs_info->sb = sb;
d98237b3 CM	316	fs_info->btree_inode = new_inode(sb);
d98237b3 CM	317	fs_info->btree_inode->i_ino = 1;
2c90e5d6	318	fs_info->btree_inode->i_nlink = 1;
d98237b3 CM	319	fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
d98237b3 CM	320	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
22b0ebda	321	insert_inode_hash(fs_info->btree_inode);
d98237b3	322	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
87cbda5c	323	fs_info->hash_tfm = crypto_alloc_hash("sha256", 0, CRYPTO_ALG_ASYNC);
30ae8467	324	spin_lock_init(&fs_info->hash_lock);
30ae8467	325	if (!fs_info->hash_tfm \|\| IS_ERR(fs_info->hash_tfm)) {
87cbda5c CM	326	printk("failed to allocate sha256 hash\n");
	327	return NULL;
	328	}
79154b1b	329	mutex_init(&fs_info->trans_mutex);
d561c025	330	mutex_init(&fs_info->fs_mutex);
9f5fae2f CM	331	memset(&fs_info->current_insert, 0, sizeof(fs_info->current_insert));
9f5fae2f CM	332	memset(&fs_info->last_insert, 0, sizeof(fs_info->last_insert));
3768f368	333
2c90e5d6 CM	334	__setup_root(sb->s_blocksize, tree_root,
	335	fs_info, BTRFS_ROOT_TREE_OBJECTID);
	336	fs_info->sb_buffer = read_tree_block(tree_root,
	337	BTRFS_SUPER_INFO_OFFSET /
	338	sb->s_blocksize);
d98237b3	339
87cbda5c	340	if (!fs_info->sb_buffer) {
d98237b3	341	return NULL;
87cbda5c	342	}
d98237b3	343	disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
2c90e5d6 CM	344	if (!btrfs_super_root(disk_super)) {
	345	return NULL;
	346	}
d98237b3	347	fs_info->disk_super = disk_super;
e20d96d6 CM	348	tree_root->node = read_tree_block(tree_root,
e20d96d6 CM	349	btrfs_super_root(disk_super));
3768f368 CM	350	BUG_ON(!tree_root->node);
3768f368 CM	351
2c90e5d6 CM	352	mutex_lock(&fs_info->fs_mutex);
2c90e5d6 CM	353	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
e20d96d6	354	BTRFS_EXTENT_TREE_OBJECTID, extent_root);
3768f368 CM	355	BUG_ON(ret);
3768f368 CM	356
2c90e5d6	357	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
e20d96d6	358	BTRFS_INODE_MAP_OBJECTID, inode_root);
9f5fae2f CM	359	BUG_ON(ret);
9f5fae2f CM	360
2c90e5d6	361	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
e20d96d6	362	BTRFS_FS_TREE_OBJECTID, root);
3768f368	363	BUG_ON(ret);
a28ec197	364	root->commit_root = root->node;
e20d96d6	365	get_bh(root->node);
3768f368	366	root->ref_cows = 1;
293ffd5f	367	root->fs_info->generation = root->root_key.offset + 1;
5be6f7f1 CM	368	ret = btrfs_find_highest_inode(root, &root->fs_info->last_inode_alloc);
	369	if (ret == 0)
	370	fs_info->highest_inode = fs_info->last_inode_alloc;
d6e4a428 CM	371	memset(&fs_info->kobj, 0, sizeof(fs_info->kobj));
	372	kobj_set_kset_s(fs_info, btrfs_subsys);
	373	kobject_set_name(&fs_info->kobj, "%s", sb->s_id);
	374	kobject_register(&fs_info->kobj);
5be6f7f1	375	mutex_unlock(&fs_info->fs_mutex);
eb60ceac CM	376	return root;
	377	}
	378
e089f05c	379	int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
79154b1b	380	*root)
eb60ceac	381	{
d5719762	382	struct buffer_head *bh = root->fs_info->sb_buffer;
2c90e5d6	383
d5719762 CM	384	btrfs_set_super_root(root->fs_info->disk_super,
	385	root->fs_info->tree_root->node->b_blocknr);
	386	lock_buffer(bh);
2c90e5d6	387	WARN_ON(atomic_read(&bh->b_count) < 1);
d5719762	388	clear_buffer_dirty(bh);
87cbda5c	389	csum_tree_block(root, bh, 0);
d5719762 CM	390	bh->b_end_io = end_buffer_write_sync;
	391	get_bh(bh);
	392	submit_bh(WRITE, bh);
	393	wait_on_buffer(bh);
	394	if (!buffer_uptodate(bh)) {
	395	WARN_ON(1);
	396	return -EIO;
cfaa7295 CM	397	}
	398	return 0;
	399	}
	400
e20d96d6	401	int close_ctree(struct btrfs_root *root)
cfaa7295	402	{
3768f368	403	int ret;
e089f05c CM	404	struct btrfs_trans_handle *trans;
e089f05c CM	405
2c90e5d6	406	mutex_lock(&root->fs_info->fs_mutex);
79154b1b CM	407	trans = btrfs_start_transaction(root, 1);
	408	btrfs_commit_transaction(trans, root);
	409	/* run commit again to drop the original snapshot */
	410	trans = btrfs_start_transaction(root, 1);
	411	btrfs_commit_transaction(trans, root);
	412	ret = btrfs_write_and_wait_transaction(NULL, root);
3768f368	413	BUG_ON(ret);
79154b1b	414	write_ctree_super(NULL, root);
2c90e5d6	415	mutex_unlock(&root->fs_info->fs_mutex);
ed2ff2cb	416
cfaa7295	417	if (root->node)
234b63a0	418	btrfs_block_release(root, root->node);
9f5fae2f CM	419	if (root->fs_info->extent_root->node)
	420	btrfs_block_release(root->fs_info->extent_root,
	421	root->fs_info->extent_root->node);
	422	if (root->fs_info->inode_root->node)
	423	btrfs_block_release(root->fs_info->inode_root,
	424	root->fs_info->inode_root->node);
	425	if (root->fs_info->tree_root->node)
	426	btrfs_block_release(root->fs_info->tree_root,
	427	root->fs_info->tree_root->node);
234b63a0	428	btrfs_block_release(root, root->commit_root);
e20d96d6	429	btrfs_block_release(root, root->fs_info->sb_buffer);
87cbda5c	430	crypto_free_hash(root->fs_info->hash_tfm);
7cfcc17e CM	431	truncate_inode_pages(root->fs_info->btree_inode->i_mapping, 0);
7cfcc17e CM	432	iput(root->fs_info->btree_inode);
e20d96d6 CM	433	kfree(root->fs_info->extent_root);
	434	kfree(root->fs_info->inode_root);
	435	kfree(root->fs_info->tree_root);
d6e4a428	436	kobject_unregister(&root->fs_info->kobj);
e20d96d6	437	kfree(root);
eb60ceac CM	438	return 0;
	439	}
	440
e20d96d6	441	void btrfs_block_release(struct btrfs_root root, struct buffer_head buf)
eb60ceac	442	{
7cfcc17e	443	brelse(buf);
eb60ceac CM	444	}
eb60ceac CM	445