[linux-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"

#define PATTERN 0xDEADBEEFUL
static inline void check_pattern(struct buffer_head *buf)
{
	if (buf->b_private != (void *)PATTERN)
		WARN_ON(1);
}

static inline void set_pattern(struct buffer_head *buf)
{
	buf->b_private = (void *)PATTERN;
}

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);
		check_pattern(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);
			set_pattern(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;

	return 0;
	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)
{
#if 0
	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);
#endif
	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;
	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);
		set_pattern(bh);
	} else {
		unlock_buffer(bh);
	}
	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(struct btrfs_super_block *super,
			struct btrfs_root *root,
			struct btrfs_fs_info *fs_info,
			u64 objectid)
{
	root->node = NULL;
	root->commit_root = NULL;
	root->blocksize = btrfs_super_blocksize(super);
	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(struct btrfs_super_block *super,
			       struct btrfs_root *tree_root,
			       struct btrfs_fs_info *fs_info,
			       u64 objectid,
			       struct btrfs_root *root)
{
	int ret;

	__setup_root(super, 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 buffer_head *sb_buffer,
			      struct btrfs_super_block *disk_super)
{
	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;

	if (!btrfs_super_root(disk_super)) {
		return NULL;
	}
	init_bit_radix(&fs_info->pinned_radix);
	init_bit_radix(&fs_info->pending_del_radix);
	sb_set_blocksize(sb, sb_buffer->b_size);
	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->last_inode_alloc_dirid = 0;
	fs_info->disk_super = disk_super;
	fs_info->sb = sb;
	fs_info->btree_inode = new_inode(sb);
	fs_info->btree_inode->i_ino = 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(disk_super, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID);

	fs_info->sb_buffer = read_tree_block(tree_root, sb_buffer->b_blocknr);

	if (!fs_info->sb_buffer) {
printk("failed2\n");
		return NULL;
	}
	brelse(sb_buffer);
	sb_buffer = NULL;
	disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
	fs_info->disk_super = disk_super;

	tree_root->node = read_tree_block(tree_root,
					  btrfs_super_root(disk_super));
	BUG_ON(!tree_root->node);

	ret = find_and_setup_root(disk_super, tree_root, fs_info,
				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
	BUG_ON(ret);

	ret = find_and_setup_root(disk_super, tree_root, fs_info,
				  BTRFS_INODE_MAP_OBJECTID, inode_root);
	BUG_ON(ret);

	ret = find_and_setup_root(disk_super, 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;
	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);
	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;

	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);

	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);
	kfree(root->fs_info);
	kfree(root);
	return 0;
}

void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
{
	check_pattern(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
d6025579 CM	11	#define PATTERN 0xDEADBEEFUL
	12	static inline void check_pattern(struct buffer_head *buf)
	13	{
	14	if (buf->b_private != (void *)PATTERN)
	15	WARN_ON(1);
	16	}
	17
	18	static inline void set_pattern(struct buffer_head *buf)
	19	{
	20	buf->b_private = (void *)PATTERN;
	21	}
d98237b3	22
e20d96d6	23	static int check_tree_block(struct btrfs_root root, struct buffer_head buf)
eb60ceac	24	{
e20d96d6	25	struct btrfs_node *node = btrfs_buffer_node(buf);
d98237b3	26	if (buf->b_blocknr != btrfs_header_blocknr(&node->header)) {
9a8dd150	27	BUG();
d98237b3	28	}
e20d96d6	29	if (root->node && btrfs_header_parentid(&node->header) !=
df2ce34c	30	btrfs_header_parentid(btrfs_buffer_header(root->node))) {
7f5c1516	31	BUG();
df2ce34c	32	}
9a8dd150	33	return 0;
eb60ceac CM	34	}
eb60ceac CM	35
d98237b3 CM	36	struct buffer_head btrfs_find_tree_block(struct btrfs_root root, u64 blocknr)
	37	{
	38	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
	39	int blockbits = root->fs_info->sb->s_blocksize_bits;
	40	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
	41	struct page *page;
	42	struct buffer_head *bh;
	43	struct buffer_head *head;
	44	struct buffer_head *ret = NULL;
	45
	46	page = find_lock_page(mapping, index);
	47	if (!page)
	48	return NULL;
	49
	50	if (!page_has_buffers(page))
	51	goto out_unlock;
	52
	53	head = page_buffers(page);
	54	bh = head;
	55	do {
	56	if (buffer_mapped(bh) && bh->b_blocknr == blocknr) {
	57	ret = bh;
	58	get_bh(bh);
	59	goto out_unlock;
	60	}
	61	bh = bh->b_this_page;
	62	} while (bh != head);
	63	out_unlock:
	64	unlock_page(page);
d6025579	65	if (ret) {
22b0ebda	66	touch_buffer(ret);
d6025579 CM	67	check_pattern(ret);
d6025579 CM	68	}
d98237b3 CM	69	page_cache_release(page);
	70	return ret;
	71	}
	72
	73	struct buffer_head btrfs_find_create_tree_block(struct btrfs_root root,
	74	u64 blocknr)
	75	{
	76	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
	77	int blockbits = root->fs_info->sb->s_blocksize_bits;
	78	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
	79	struct page *page;
	80	struct buffer_head *bh;
	81	struct buffer_head *head;
	82	struct buffer_head *ret = NULL;
	83	u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
22b0ebda	84
d98237b3 CM	85	page = grab_cache_page(mapping, index);
	86	if (!page)
	87	return NULL;
	88
d98237b3 CM	89	if (!page_has_buffers(page))
	90	create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
	91	head = page_buffers(page);
	92	bh = head;
	93	do {
	94	if (!buffer_mapped(bh)) {
	95	bh->b_bdev = root->fs_info->sb->s_bdev;
	96	bh->b_blocknr = first_block;
	97	set_buffer_mapped(bh);
d6025579	98	set_pattern(bh);
d98237b3 CM	99	}
	100	if (bh->b_blocknr == blocknr) {
	101	ret = bh;
	102	get_bh(bh);
	103	goto out_unlock;
	104	}
	105	bh = bh->b_this_page;
	106	first_block++;
	107	} while (bh != head);
	108	out_unlock:
	109	unlock_page(page);
22b0ebda CM	110	if (ret)
22b0ebda CM	111	touch_buffer(ret);
d98237b3 CM	112	page_cache_release(page);
	113	return ret;
	114	}
	115
	116	static sector_t max_block(struct block_device *bdev)
	117	{
	118	sector_t retval = ~((sector_t)0);
	119	loff_t sz = i_size_read(bdev->bd_inode);
	120
	121	if (sz) {
	122	unsigned int size = block_size(bdev);
	123	unsigned int sizebits = blksize_bits(size);
	124	retval = (sz >> sizebits);
	125	}
	126	return retval;
	127	}
	128
	129	static int btree_get_block(struct inode *inode, sector_t iblock,
	130	struct buffer_head *bh, int create)
	131	{
	132	if (iblock >= max_block(inode->i_sb->s_bdev)) {
	133	if (create)
	134	return -EIO;
	135
	136	/*
	137	* for reads, we're just trying to fill a partial page.
	138	* return a hole, they will have to call get_block again
	139	* before they can fill it, and they will get -EIO at that
	140	* time
	141	*/
	142	return 0;
	143	}
	144	bh->b_bdev = inode->i_sb->s_bdev;
	145	bh->b_blocknr = iblock;
	146	set_buffer_mapped(bh);
	147	return 0;
	148	}
	149
f254e52c CM	150	int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
f254e52c CM	151	char *result)
87cbda5c	152	{
87cbda5c CM	153	struct scatterlist sg;
	154	struct crypto_hash *tfm = root->fs_info->hash_tfm;
	155	struct hash_desc desc;
	156	int ret;
87cbda5c CM	157
	158	desc.tfm = tfm;
	159	desc.flags = 0;
f254e52c	160	sg_init_one(&sg, data, len);
87cbda5c	161	spin_lock(&root->fs_info->hash_lock);
22b0ebda	162	ret = crypto_hash_digest(&desc, &sg, 1, result);
87cbda5c CM	163	spin_unlock(&root->fs_info->hash_lock);
	164	if (ret) {
	165	printk("sha256 digest failed\n");
	166	}
f254e52c CM	167	return ret;
	168	}
	169	static int csum_tree_block(struct btrfs_root root, struct buffer_head bh,
	170	int verify)
	171	{
	172	char result[BTRFS_CSUM_SIZE];
	173	int ret;
	174	struct btrfs_node *node;
	175
22b0ebda	176	return 0;
f254e52c CM	177	ret = btrfs_csum_data(root, bh->b_data + BTRFS_CSUM_SIZE,
	178	bh->b_size - BTRFS_CSUM_SIZE, result);
	179	if (ret)
	180	return ret;
87cbda5c	181	if (verify) {
f254e52c CM	182	if (memcmp(bh->b_data, result, BTRFS_CSUM_SIZE)) {
	183	printk("checksum verify failed on %lu\n",
	184	bh->b_blocknr);
	185	return 1;
	186	}
	187	} else {
	188	node = btrfs_buffer_node(bh);
22b0ebda	189	memcpy(node->header.csum, result, BTRFS_CSUM_SIZE);
f254e52c	190	}
87cbda5c CM	191	return 0;
	192	}
	193
d98237b3	194	static int btree_writepage(struct page page, struct writeback_control wbc)
ed2ff2cb	195	{
22b0ebda	196	#if 0
87cbda5c CM	197	struct buffer_head *bh;
	198	struct btrfs_root *root = btrfs_sb(page->mapping->host->i_sb);
	199	struct buffer_head *head;
87cbda5c CM	200	if (!page_has_buffers(page)) {
	201	create_empty_buffers(page, root->fs_info->sb->s_blocksize,
	202	(1 << BH_Dirty)\|(1 << BH_Uptodate));
	203	}
	204	head = page_buffers(page);
	205	bh = head;
	206	do {
	207	if (buffer_dirty(bh))
	208	csum_tree_block(root, bh, 0);
	209	bh = bh->b_this_page;
	210	} while (bh != head);
22b0ebda	211	#endif
d98237b3	212	return block_write_full_page(page, btree_get_block, wbc);
ed2ff2cb CM	213	}
ed2ff2cb CM	214
d98237b3	215	static int btree_readpage(struct file * file, struct page * page)
eb60ceac	216	{
d98237b3	217	return block_read_full_page(page, btree_get_block);
eb60ceac CM	218	}
eb60ceac CM	219
d98237b3 CM	220	static struct address_space_operations btree_aops = {
	221	.readpage = btree_readpage,
	222	.writepage = btree_writepage,
	223	.sync_page = block_sync_page,
	224	};
	225
e20d96d6	226	struct buffer_head read_tree_block(struct btrfs_root root, u64 blocknr)
eb60ceac	227	{
d98237b3	228	struct buffer_head *bh = NULL;
eb60ceac	229
d98237b3 CM	230	bh = btrfs_find_create_tree_block(root, blocknr);
	231	if (!bh)
	232	return bh;
	233	lock_buffer(bh);
	234	if (!buffer_uptodate(bh)) {
	235	get_bh(bh);
	236	bh->b_end_io = end_buffer_read_sync;
	237	submit_bh(READ, bh);
	238	wait_on_buffer(bh);
	239	if (!buffer_uptodate(bh))
	240	goto fail;
87cbda5c	241	csum_tree_block(root, bh, 1);
d6025579	242	set_pattern(bh);
d98237b3 CM	243	} else {
	244	unlock_buffer(bh);
	245	}
	246	if (check_tree_block(root, bh))
cfaa7295	247	BUG();
d98237b3 CM	248	return bh;
	249	fail:
	250	brelse(bh);
	251	return NULL;
	252
eb60ceac CM	253	}
eb60ceac CM	254
e089f05c	255	int dirty_tree_block(struct btrfs_trans_handle trans, struct btrfs_root root,
e20d96d6	256	struct buffer_head *buf)
ed2ff2cb	257	{
d6025579	258	WARN_ON(atomic_read(&buf->b_count) == 0);
e20d96d6	259	mark_buffer_dirty(buf);
ed2ff2cb CM	260	return 0;
	261	}
	262
e089f05c	263	int clean_tree_block(struct btrfs_trans_handle trans, struct btrfs_root root,
e20d96d6	264	struct buffer_head *buf)
ed2ff2cb	265	{
d6025579	266	WARN_ON(atomic_read(&buf->b_count) == 0);
e20d96d6	267	clear_buffer_dirty(buf);
ed2ff2cb CM	268	return 0;
	269	}
	270
123abc88	271	static int __setup_root(struct btrfs_super_block *super,
9f5fae2f CM	272	struct btrfs_root *root,
9f5fae2f CM	273	struct btrfs_fs_info *fs_info,
e20d96d6	274	u64 objectid)
d97e63b6	275	{
cfaa7295	276	root->node = NULL;
a28ec197	277	root->commit_root = NULL;
123abc88 CM	278	root->blocksize = btrfs_super_blocksize(super);
123abc88 CM	279	root->ref_cows = 0;
9f5fae2f	280	root->fs_info = fs_info;
3768f368 CM	281	memset(&root->root_key, 0, sizeof(root->root_key));
	282	memset(&root->root_item, 0, sizeof(root->root_item));
	283	return 0;
	284	}
	285
123abc88	286	static int find_and_setup_root(struct btrfs_super_block *super,
9f5fae2f CM	287	struct btrfs_root *tree_root,
	288	struct btrfs_fs_info *fs_info,
	289	u64 objectid,
e20d96d6	290	struct btrfs_root *root)
3768f368 CM	291	{
	292	int ret;
	293
e20d96d6	294	__setup_root(super, root, fs_info, objectid);
3768f368 CM	295	ret = btrfs_find_last_root(tree_root, objectid,
	296	&root->root_item, &root->root_key);
	297	BUG_ON(ret);
	298
	299	root->node = read_tree_block(root,
	300	btrfs_root_blocknr(&root->root_item));
3768f368	301	BUG_ON(!root->node);
d97e63b6 CM	302	return 0;
	303	}
	304
e20d96d6 CM	305	struct btrfs_root open_ctree(struct super_block sb,
	306	struct buffer_head *sb_buffer,
	307	struct btrfs_super_block *disk_super)
2e635a27	308	{
e20d96d6 CM	309	struct btrfs_root *root = kmalloc(sizeof(struct btrfs_root),
	310	GFP_NOFS);
	311	struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
	312	GFP_NOFS);
	313	struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
	314	GFP_NOFS);
	315	struct btrfs_root *inode_root = kmalloc(sizeof(struct btrfs_root),
	316	GFP_NOFS);
	317	struct btrfs_fs_info fs_info = kmalloc(sizeof(fs_info),
	318	GFP_NOFS);
eb60ceac CM	319	int ret;
eb60ceac CM	320
87cbda5c	321	if (!btrfs_super_root(disk_super)) {
e20d96d6	322	return NULL;
87cbda5c	323	}
8ef97622 CM	324	init_bit_radix(&fs_info->pinned_radix);
8ef97622 CM	325	init_bit_radix(&fs_info->pending_del_radix);
d98237b3	326	sb_set_blocksize(sb, sb_buffer->b_size);
9f5fae2f CM	327	fs_info->running_transaction = NULL;
	328	fs_info->fs_root = root;
	329	fs_info->tree_root = tree_root;
	330	fs_info->extent_root = extent_root;
	331	fs_info->inode_root = inode_root;
	332	fs_info->last_inode_alloc = 0;
	333	fs_info->last_inode_alloc_dirid = 0;
e20d96d6	334	fs_info->disk_super = disk_super;
e20d96d6	335	fs_info->sb = sb;
d98237b3 CM	336	fs_info->btree_inode = new_inode(sb);
	337	fs_info->btree_inode->i_ino = 1;
	338	fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
	339	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
22b0ebda CM	340	insert_inode_hash(fs_info->btree_inode);
22b0ebda CM	341
d98237b3	342	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
87cbda5c	343	fs_info->hash_tfm = crypto_alloc_hash("sha256", 0, CRYPTO_ALG_ASYNC);
30ae8467 CM	344	spin_lock_init(&fs_info->hash_lock);
	345
	346	if (!fs_info->hash_tfm \|\| IS_ERR(fs_info->hash_tfm)) {
87cbda5c CM	347	printk("failed to allocate sha256 hash\n");
	348	return NULL;
	349	}
d98237b3	350
79154b1b	351	mutex_init(&fs_info->trans_mutex);
d561c025	352	mutex_init(&fs_info->fs_mutex);
9f5fae2f CM	353	memset(&fs_info->current_insert, 0, sizeof(fs_info->current_insert));
9f5fae2f CM	354	memset(&fs_info->last_insert, 0, sizeof(fs_info->last_insert));
3768f368	355
e20d96d6	356	__setup_root(disk_super, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
d98237b3 CM	357
	358	fs_info->sb_buffer = read_tree_block(tree_root, sb_buffer->b_blocknr);
	359
87cbda5c CM	360	if (!fs_info->sb_buffer) {
87cbda5c CM	361	printk("failed2\n");
d98237b3	362	return NULL;
87cbda5c	363	}
d98237b3 CM	364	brelse(sb_buffer);
	365	sb_buffer = NULL;
	366	disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
	367	fs_info->disk_super = disk_super;
	368
e20d96d6 CM	369	tree_root->node = read_tree_block(tree_root,
e20d96d6 CM	370	btrfs_super_root(disk_super));
3768f368 CM	371	BUG_ON(!tree_root->node);
3768f368 CM	372
e20d96d6 CM	373	ret = find_and_setup_root(disk_super, tree_root, fs_info,
e20d96d6 CM	374	BTRFS_EXTENT_TREE_OBJECTID, extent_root);
3768f368 CM	375	BUG_ON(ret);
3768f368 CM	376
e20d96d6 CM	377	ret = find_and_setup_root(disk_super, tree_root, fs_info,
e20d96d6 CM	378	BTRFS_INODE_MAP_OBJECTID, inode_root);
9f5fae2f CM	379	BUG_ON(ret);
9f5fae2f CM	380
e20d96d6 CM	381	ret = find_and_setup_root(disk_super, tree_root, fs_info,
e20d96d6 CM	382	BTRFS_FS_TREE_OBJECTID, root);
3768f368	383	BUG_ON(ret);
a28ec197	384	root->commit_root = root->node;
e20d96d6	385	get_bh(root->node);
3768f368	386	root->ref_cows = 1;
293ffd5f	387	root->fs_info->generation = root->root_key.offset + 1;
eb60ceac CM	388	return root;
	389	}
	390
e089f05c	391	int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
79154b1b	392	*root)
eb60ceac	393	{
d5719762 CM	394	struct buffer_head *bh = root->fs_info->sb_buffer;
	395	btrfs_set_super_root(root->fs_info->disk_super,
	396	root->fs_info->tree_root->node->b_blocknr);
	397	lock_buffer(bh);
	398	clear_buffer_dirty(bh);
87cbda5c	399	csum_tree_block(root, bh, 0);
d5719762 CM	400	bh->b_end_io = end_buffer_write_sync;
	401	get_bh(bh);
	402	submit_bh(WRITE, bh);
	403	wait_on_buffer(bh);
	404	if (!buffer_uptodate(bh)) {
	405	WARN_ON(1);
	406	return -EIO;
cfaa7295 CM	407	}
	408	return 0;
	409	}
	410
e20d96d6	411	int close_ctree(struct btrfs_root *root)
cfaa7295	412	{
3768f368	413	int ret;
e089f05c CM	414	struct btrfs_trans_handle *trans;
e089f05c CM	415
79154b1b CM	416	trans = btrfs_start_transaction(root, 1);
	417	btrfs_commit_transaction(trans, root);
	418	/* run commit again to drop the original snapshot */
	419	trans = btrfs_start_transaction(root, 1);
	420	btrfs_commit_transaction(trans, root);
	421	ret = btrfs_write_and_wait_transaction(NULL, root);
3768f368	422	BUG_ON(ret);
79154b1b	423	write_ctree_super(NULL, root);
ed2ff2cb	424
cfaa7295	425	if (root->node)
234b63a0	426	btrfs_block_release(root, root->node);
9f5fae2f CM	427	if (root->fs_info->extent_root->node)
	428	btrfs_block_release(root->fs_info->extent_root,
	429	root->fs_info->extent_root->node);
	430	if (root->fs_info->inode_root->node)
	431	btrfs_block_release(root->fs_info->inode_root,
	432	root->fs_info->inode_root->node);
	433	if (root->fs_info->tree_root->node)
	434	btrfs_block_release(root->fs_info->tree_root,
	435	root->fs_info->tree_root->node);
234b63a0	436	btrfs_block_release(root, root->commit_root);
e20d96d6	437	btrfs_block_release(root, root->fs_info->sb_buffer);
87cbda5c	438	crypto_free_hash(root->fs_info->hash_tfm);
30ae8467	439	truncate_inode_pages(root->fs_info->btree_inode->i_mapping, 0);
d98237b3	440	iput(root->fs_info->btree_inode);
e20d96d6 CM	441	kfree(root->fs_info->extent_root);
	442	kfree(root->fs_info->inode_root);
	443	kfree(root->fs_info->tree_root);
	444	kfree(root->fs_info);
	445	kfree(root);
eb60ceac CM	446	return 0;
	447	}
	448
e20d96d6	449	void btrfs_block_release(struct btrfs_root root, struct buffer_head buf)
eb60ceac	450	{
d6025579	451	check_pattern(buf);
e20d96d6	452	brelse(buf);
eb60ceac CM	453	}
eb60ceac CM	454