[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 <linux/radix-tree.h>
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
#include "btrfs_inode.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();
	}
	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_I(page->mapping->host)->root;
	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->inode = NULL;
	root->commit_root = NULL;
	root->blocksize = blocksize;
	root->ref_cows = 0;
	root->fs_info = fs_info;
	root->objectid = objectid;
	root->last_trans = 0;
	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 *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
				      struct btrfs_key *location)
{
	struct btrfs_root *root;
	struct btrfs_root *tree_root = fs_info->tree_root;
	struct btrfs_path *path;
	struct btrfs_leaf *l;
	int ret = 0;

printk("read_fs_root looking for %Lu %Lu %u\n", location->objectid, location->offset, location->flags);
	root = kmalloc(sizeof(*root), GFP_NOFS);
	if (!root) {
printk("failed1\n");
		return ERR_PTR(-ENOMEM);
	}
	if (location->offset == (u64)-1) {
		ret = find_and_setup_root(fs_info->sb->s_blocksize,
					  fs_info->tree_root, fs_info,
					  location->objectid, root);
		if (ret) {
printk("failed2\n");
			kfree(root);
			return ERR_PTR(ret);
		}
		goto insert;
	}

	__setup_root(fs_info->sb->s_blocksize, root, fs_info,
		     location->objectid);

	path = btrfs_alloc_path();
	BUG_ON(!path);
	ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
	if (ret != 0) {
printk("internal search_slot gives us %d\n", ret);
		if (ret > 0)
			ret = -ENOENT;
		goto out;
	}
	l = btrfs_buffer_leaf(path->nodes[0]);
	memcpy(&root->root_item,
	       btrfs_item_ptr(l, path->slots[0], struct btrfs_root_item),
	       sizeof(root->root_item));
	memcpy(&root->root_key, location, sizeof(*location));
	ret = 0;
out:
	btrfs_release_path(root, path);
	btrfs_free_path(path);
	if (ret) {
		kfree(root);
		return ERR_PTR(ret);
	}
	root->node = read_tree_block(root,
				     btrfs_root_blocknr(&root->root_item));
	BUG_ON(!root->node);
insert:
printk("inserting %p\n", root);
	root->ref_cows = 1;
	ret = radix_tree_insert(&fs_info->fs_roots_radix, (unsigned long)root,
				root);
	if (ret) {
printk("radix_tree_insert gives us %d\n", ret);
		brelse(root->node);
		kfree(root);
		return ERR_PTR(ret);
	}
printk("all worked\n");
	return root;
}

struct btrfs_root *open_ctree(struct super_block *sb)
{
	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);
	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
	sb_set_blocksize(sb, 4096);
	fs_info->running_transaction = NULL;
	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;
	BTRFS_I(fs_info->btree_inode)->root = tree_root;
	memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
	       sizeof(struct btrfs_key));
	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);

	fs_info->generation = btrfs_super_generation(disk_super) + 1;
	ret = btrfs_find_highest_inode(tree_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 tree_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 del_fs_roots(struct btrfs_fs_info *fs_info)
{
	int ret;
	struct btrfs_root *gang[8];
	int i;

	while(1) {
		ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
					     (void **)gang, 0,
					     ARRAY_SIZE(gang));
		if (!ret)
			break;
		for (i = 0; i < ret; i++) {
			radix_tree_delete(&fs_info->fs_roots_radix,
					  (unsigned long)gang[i]);
			if (gang[i]->inode)
				iput(gang[i]->inode);
			else
				printk("no inode for root %p\n", gang[i]);
			if (gang[i]->node)
				brelse(gang[i]->node);
			if (gang[i]->commit_root)
				brelse(gang[i]->commit_root);
			kfree(gang[i]);
		}
	}
	return 0;
}

int close_ctree(struct btrfs_root *root)
{
	int ret;
	struct btrfs_trans_handle *trans;
	struct btrfs_fs_info *fs_info = root->fs_info;

	mutex_lock(&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(&fs_info->fs_mutex);

	if (fs_info->extent_root->node)
		btrfs_block_release(fs_info->extent_root,
				    fs_info->extent_root->node);
	if (fs_info->inode_root->node)
		btrfs_block_release(fs_info->inode_root,
				    fs_info->inode_root->node);
	if (fs_info->tree_root->node)
		btrfs_block_release(fs_info->tree_root,
				    fs_info->tree_root->node);
	btrfs_block_release(root, fs_info->sb_buffer);
	crypto_free_hash(fs_info->hash_tfm);
	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
	iput(fs_info->btree_inode);
	del_fs_roots(fs_info);
	kfree(fs_info->extent_root);
	kfree(fs_info->inode_root);
	kfree(fs_info->tree_root);
	kobject_unregister(&fs_info->kobj);
	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>
0f7d52f4	7	#include <linux/radix-tree.h>
eb60ceac CM	8	#include "ctree.h"
eb60ceac CM	9	#include "disk-io.h"
e089f05c	10	#include "transaction.h"
0f7d52f4	11	#include "btrfs_inode.h"
eb60ceac	12
e20d96d6	13	static int check_tree_block(struct btrfs_root root, struct buffer_head buf)
eb60ceac	14	{
e20d96d6	15	struct btrfs_node *node = btrfs_buffer_node(buf);
d98237b3	16	if (buf->b_blocknr != btrfs_header_blocknr(&node->header)) {
9a8dd150	17	BUG();
d98237b3	18	}
9a8dd150	19	return 0;
eb60ceac CM	20	}
eb60ceac CM	21
d98237b3 CM	22	struct buffer_head btrfs_find_tree_block(struct btrfs_root root, u64 blocknr)
	23	{
	24	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
	25	int blockbits = root->fs_info->sb->s_blocksize_bits;
	26	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
	27	struct page *page;
	28	struct buffer_head *bh;
	29	struct buffer_head *head;
	30	struct buffer_head *ret = NULL;
	31
2c90e5d6	32
d98237b3 CM	33	page = find_lock_page(mapping, index);
	34	if (!page)
	35	return NULL;
	36
	37	if (!page_has_buffers(page))
	38	goto out_unlock;
	39
	40	head = page_buffers(page);
	41	bh = head;
	42	do {
	43	if (buffer_mapped(bh) && bh->b_blocknr == blocknr) {
	44	ret = bh;
	45	get_bh(bh);
	46	goto out_unlock;
	47	}
	48	bh = bh->b_this_page;
	49	} while (bh != head);
	50	out_unlock:
	51	unlock_page(page);
d6025579	52	if (ret) {
22b0ebda	53	touch_buffer(ret);
d6025579	54	}
d98237b3 CM	55	page_cache_release(page);
	56	return ret;
	57	}
	58
	59	struct buffer_head btrfs_find_create_tree_block(struct btrfs_root root,
	60	u64 blocknr)
	61	{
	62	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
	63	int blockbits = root->fs_info->sb->s_blocksize_bits;
	64	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
	65	struct page *page;
	66	struct buffer_head *bh;
	67	struct buffer_head *head;
	68	struct buffer_head *ret = NULL;
	69	u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
22b0ebda	70
d98237b3 CM	71	page = grab_cache_page(mapping, index);
	72	if (!page)
	73	return NULL;
	74
d98237b3 CM	75	if (!page_has_buffers(page))
	76	create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
	77	head = page_buffers(page);
	78	bh = head;
	79	do {
	80	if (!buffer_mapped(bh)) {
	81	bh->b_bdev = root->fs_info->sb->s_bdev;
	82	bh->b_blocknr = first_block;
	83	set_buffer_mapped(bh);
	84	}
	85	if (bh->b_blocknr == blocknr) {
	86	ret = bh;
	87	get_bh(bh);
	88	goto out_unlock;
	89	}
	90	bh = bh->b_this_page;
	91	first_block++;
	92	} while (bh != head);
	93	out_unlock:
	94	unlock_page(page);
22b0ebda CM	95	if (ret)
22b0ebda CM	96	touch_buffer(ret);
d98237b3 CM	97	page_cache_release(page);
	98	return ret;
	99	}
	100
	101	static sector_t max_block(struct block_device *bdev)
	102	{
	103	sector_t retval = ~((sector_t)0);
	104	loff_t sz = i_size_read(bdev->bd_inode);
	105
	106	if (sz) {
	107	unsigned int size = block_size(bdev);
	108	unsigned int sizebits = blksize_bits(size);
	109	retval = (sz >> sizebits);
	110	}
	111	return retval;
	112	}
	113
	114	static int btree_get_block(struct inode *inode, sector_t iblock,
	115	struct buffer_head *bh, int create)
	116	{
	117	if (iblock >= max_block(inode->i_sb->s_bdev)) {
	118	if (create)
	119	return -EIO;
	120
	121	/*
	122	* for reads, we're just trying to fill a partial page.
	123	* return a hole, they will have to call get_block again
	124	* before they can fill it, and they will get -EIO at that
	125	* time
	126	*/
	127	return 0;
	128	}
	129	bh->b_bdev = inode->i_sb->s_bdev;
	130	bh->b_blocknr = iblock;
	131	set_buffer_mapped(bh);
	132	return 0;
	133	}
	134
f254e52c CM	135	int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
f254e52c CM	136	char *result)
87cbda5c	137	{
87cbda5c CM	138	struct scatterlist sg;
	139	struct crypto_hash *tfm = root->fs_info->hash_tfm;
	140	struct hash_desc desc;
	141	int ret;
87cbda5c CM	142
	143	desc.tfm = tfm;
	144	desc.flags = 0;
f254e52c	145	sg_init_one(&sg, data, len);
87cbda5c	146	spin_lock(&root->fs_info->hash_lock);
22b0ebda	147	ret = crypto_hash_digest(&desc, &sg, 1, result);
87cbda5c CM	148	spin_unlock(&root->fs_info->hash_lock);
	149	if (ret) {
	150	printk("sha256 digest failed\n");
	151	}
f254e52c CM	152	return ret;
	153	}
	154	static int csum_tree_block(struct btrfs_root root, struct buffer_head bh,
	155	int verify)
	156	{
	157	char result[BTRFS_CSUM_SIZE];
	158	int ret;
	159	struct btrfs_node *node;
	160
	161	ret = btrfs_csum_data(root, bh->b_data + BTRFS_CSUM_SIZE,
	162	bh->b_size - BTRFS_CSUM_SIZE, result);
	163	if (ret)
	164	return ret;
87cbda5c	165	if (verify) {
f254e52c CM	166	if (memcmp(bh->b_data, result, BTRFS_CSUM_SIZE)) {
	167	printk("checksum verify failed on %lu\n",
	168	bh->b_blocknr);
	169	return 1;
	170	}
	171	} else {
	172	node = btrfs_buffer_node(bh);
22b0ebda	173	memcpy(node->header.csum, result, BTRFS_CSUM_SIZE);
f254e52c	174	}
87cbda5c CM	175	return 0;
	176	}
	177
d98237b3	178	static int btree_writepage(struct page page, struct writeback_control wbc)
ed2ff2cb	179	{
87cbda5c	180	struct buffer_head *bh;
0f7d52f4	181	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
87cbda5c	182	struct buffer_head *head;
87cbda5c CM	183	if (!page_has_buffers(page)) {
	184	create_empty_buffers(page, root->fs_info->sb->s_blocksize,
	185	(1 << BH_Dirty)\|(1 << BH_Uptodate));
	186	}
	187	head = page_buffers(page);
	188	bh = head;
	189	do {
	190	if (buffer_dirty(bh))
	191	csum_tree_block(root, bh, 0);
	192	bh = bh->b_this_page;
	193	} while (bh != head);
d98237b3	194	return block_write_full_page(page, btree_get_block, wbc);
ed2ff2cb CM	195	}
ed2ff2cb CM	196
d98237b3	197	static int btree_readpage(struct file * file, struct page * page)
eb60ceac	198	{
d98237b3	199	return block_read_full_page(page, btree_get_block);
eb60ceac CM	200	}
eb60ceac CM	201
d98237b3 CM	202	static struct address_space_operations btree_aops = {
	203	.readpage = btree_readpage,
	204	.writepage = btree_writepage,
	205	.sync_page = block_sync_page,
	206	};
	207
e20d96d6	208	struct buffer_head read_tree_block(struct btrfs_root root, u64 blocknr)
eb60ceac	209	{
d98237b3	210	struct buffer_head *bh = NULL;
eb60ceac	211
d98237b3 CM	212	bh = btrfs_find_create_tree_block(root, blocknr);
	213	if (!bh)
	214	return bh;
9d64272c CM	215	if (buffer_uptodate(bh))
9d64272c CM	216	goto uptodate;
d98237b3 CM	217	lock_buffer(bh);
	218	if (!buffer_uptodate(bh)) {
	219	get_bh(bh);
	220	bh->b_end_io = end_buffer_read_sync;
	221	submit_bh(READ, bh);
	222	wait_on_buffer(bh);
	223	if (!buffer_uptodate(bh))
	224	goto fail;
87cbda5c	225	csum_tree_block(root, bh, 1);
d98237b3 CM	226	} else {
	227	unlock_buffer(bh);
	228	}
9d64272c	229	uptodate:
d98237b3	230	if (check_tree_block(root, bh))
cfaa7295	231	BUG();
d98237b3 CM	232	return bh;
	233	fail:
	234	brelse(bh);
	235	return NULL;
eb60ceac CM	236	}
eb60ceac CM	237
e089f05c	238	int dirty_tree_block(struct btrfs_trans_handle trans, struct btrfs_root root,
e20d96d6	239	struct buffer_head *buf)
ed2ff2cb	240	{
d6025579	241	WARN_ON(atomic_read(&buf->b_count) == 0);
e20d96d6	242	mark_buffer_dirty(buf);
ed2ff2cb CM	243	return 0;
	244	}
	245
e089f05c	246	int clean_tree_block(struct btrfs_trans_handle trans, struct btrfs_root root,
e20d96d6	247	struct buffer_head *buf)
ed2ff2cb	248	{
d6025579	249	WARN_ON(atomic_read(&buf->b_count) == 0);
e20d96d6	250	clear_buffer_dirty(buf);
ed2ff2cb CM	251	return 0;
	252	}
	253
2c90e5d6	254	static int __setup_root(int blocksize,
9f5fae2f CM	255	struct btrfs_root *root,
9f5fae2f CM	256	struct btrfs_fs_info *fs_info,
e20d96d6	257	u64 objectid)
d97e63b6	258	{
cfaa7295	259	root->node = NULL;
0f7d52f4	260	root->inode = NULL;
a28ec197	261	root->commit_root = NULL;
2c90e5d6	262	root->blocksize = blocksize;
123abc88	263	root->ref_cows = 0;
9f5fae2f	264	root->fs_info = fs_info;
0f7d52f4 CM	265	root->objectid = objectid;
0f7d52f4 CM	266	root->last_trans = 0;
3768f368 CM	267	memset(&root->root_key, 0, sizeof(root->root_key));
	268	memset(&root->root_item, 0, sizeof(root->root_item));
	269	return 0;
	270	}
	271
2c90e5d6	272	static int find_and_setup_root(int blocksize,
9f5fae2f CM	273	struct btrfs_root *tree_root,
	274	struct btrfs_fs_info *fs_info,
	275	u64 objectid,
e20d96d6	276	struct btrfs_root *root)
3768f368 CM	277	{
	278	int ret;
	279
2c90e5d6	280	__setup_root(blocksize, root, fs_info, objectid);
3768f368 CM	281	ret = btrfs_find_last_root(tree_root, objectid,
	282	&root->root_item, &root->root_key);
	283	BUG_ON(ret);
	284
	285	root->node = read_tree_block(root,
	286	btrfs_root_blocknr(&root->root_item));
3768f368	287	BUG_ON(!root->node);
d97e63b6 CM	288	return 0;
	289	}
	290
0f7d52f4 CM	291	struct btrfs_root btrfs_read_fs_root(struct btrfs_fs_info fs_info,
	292	struct btrfs_key *location)
	293	{
	294	struct btrfs_root *root;
	295	struct btrfs_root *tree_root = fs_info->tree_root;
	296	struct btrfs_path *path;
	297	struct btrfs_leaf *l;
	298	int ret = 0;
	299
	300	printk("read_fs_root looking for %Lu %Lu %u\n", location->objectid, location->offset, location->flags);
	301	root = kmalloc(sizeof(*root), GFP_NOFS);
	302	if (!root) {
	303	printk("failed1\n");
	304	return ERR_PTR(-ENOMEM);
	305	}
	306	if (location->offset == (u64)-1) {
	307	ret = find_and_setup_root(fs_info->sb->s_blocksize,
	308	fs_info->tree_root, fs_info,
	309	location->objectid, root);
	310	if (ret) {
	311	printk("failed2\n");
	312	kfree(root);
	313	return ERR_PTR(ret);
	314	}
	315	goto insert;
	316	}
	317
	318	__setup_root(fs_info->sb->s_blocksize, root, fs_info,
	319	location->objectid);
	320
	321	path = btrfs_alloc_path();
	322	BUG_ON(!path);
	323	ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
	324	if (ret != 0) {
	325	printk("internal search_slot gives us %d\n", ret);
	326	if (ret > 0)
	327	ret = -ENOENT;
	328	goto out;
	329	}
	330	l = btrfs_buffer_leaf(path->nodes[0]);
	331	memcpy(&root->root_item,
	332	btrfs_item_ptr(l, path->slots[0], struct btrfs_root_item),
	333	sizeof(root->root_item));
	334	memcpy(&root->root_key, location, sizeof(*location));
	335	ret = 0;
	336	out:
	337	btrfs_release_path(root, path);
	338	btrfs_free_path(path);
	339	if (ret) {
	340	kfree(root);
	341	return ERR_PTR(ret);
	342	}
	343	root->node = read_tree_block(root,
	344	btrfs_root_blocknr(&root->root_item));
	345	BUG_ON(!root->node);
	346	insert:
	347	printk("inserting %p\n", root);
	348	root->ref_cows = 1;
	349	ret = radix_tree_insert(&fs_info->fs_roots_radix, (unsigned long)root,
	350	root);
	351	if (ret) {
	352	printk("radix_tree_insert gives us %d\n", ret);
	353	brelse(root->node);
	354	kfree(root);
355	return ERR_PTR(ret);
356	}
357	printk("all worked\n");
358	return root;
359	}
360
2c90e5d6	361	struct btrfs_root open_ctree(struct super_block sb)
2e635a27	362	{
e20d96d6 CM	363	struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
	364	GFP_NOFS);
	365	struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
	366	GFP_NOFS);
	367	struct btrfs_root *inode_root = kmalloc(sizeof(struct btrfs_root),
	368	GFP_NOFS);
	369	struct btrfs_fs_info fs_info = kmalloc(sizeof(fs_info),
	370	GFP_NOFS);
eb60ceac	371	int ret;
2c90e5d6	372	struct btrfs_super_block *disk_super;
eb60ceac	373
8ef97622 CM	374	init_bit_radix(&fs_info->pinned_radix);
8ef97622 CM	375	init_bit_radix(&fs_info->pending_del_radix);
0f7d52f4	376	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
2c90e5d6	377	sb_set_blocksize(sb, 4096);
9f5fae2f	378	fs_info->running_transaction = NULL;
9f5fae2f CM	379	fs_info->tree_root = tree_root;
	380	fs_info->extent_root = extent_root;
	381	fs_info->inode_root = inode_root;
	382	fs_info->last_inode_alloc = 0;
5be6f7f1	383	fs_info->highest_inode = 0;
e20d96d6	384	fs_info->sb = sb;
d98237b3 CM	385	fs_info->btree_inode = new_inode(sb);
d98237b3 CM	386	fs_info->btree_inode->i_ino = 1;
2c90e5d6	387	fs_info->btree_inode->i_nlink = 1;
d98237b3 CM	388	fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
d98237b3 CM	389	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
0f7d52f4 CM	390	BTRFS_I(fs_info->btree_inode)->root = tree_root;
	391	memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
	392	sizeof(struct btrfs_key));
22b0ebda	393	insert_inode_hash(fs_info->btree_inode);
d98237b3	394	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
87cbda5c	395	fs_info->hash_tfm = crypto_alloc_hash("sha256", 0, CRYPTO_ALG_ASYNC);
30ae8467	396	spin_lock_init(&fs_info->hash_lock);
30ae8467	397	if (!fs_info->hash_tfm \|\| IS_ERR(fs_info->hash_tfm)) {
87cbda5c CM	398	printk("failed to allocate sha256 hash\n");
	399	return NULL;
	400	}
79154b1b	401	mutex_init(&fs_info->trans_mutex);
d561c025	402	mutex_init(&fs_info->fs_mutex);
9f5fae2f CM	403	memset(&fs_info->current_insert, 0, sizeof(fs_info->current_insert));
9f5fae2f CM	404	memset(&fs_info->last_insert, 0, sizeof(fs_info->last_insert));
3768f368	405
2c90e5d6 CM	406	__setup_root(sb->s_blocksize, tree_root,
	407	fs_info, BTRFS_ROOT_TREE_OBJECTID);
	408	fs_info->sb_buffer = read_tree_block(tree_root,
	409	BTRFS_SUPER_INFO_OFFSET /
	410	sb->s_blocksize);
d98237b3	411
0f7d52f4	412	if (!fs_info->sb_buffer)
d98237b3	413	return NULL;
d98237b3	414	disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
0f7d52f4	415	if (!btrfs_super_root(disk_super))
2c90e5d6	416	return NULL;
0f7d52f4	417
d98237b3	418	fs_info->disk_super = disk_super;
e20d96d6 CM	419	tree_root->node = read_tree_block(tree_root,
e20d96d6 CM	420	btrfs_super_root(disk_super));
3768f368 CM	421	BUG_ON(!tree_root->node);
3768f368 CM	422
2c90e5d6 CM	423	mutex_lock(&fs_info->fs_mutex);
2c90e5d6 CM	424	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
e20d96d6	425	BTRFS_EXTENT_TREE_OBJECTID, extent_root);
3768f368 CM	426	BUG_ON(ret);
3768f368 CM	427
2c90e5d6	428	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
e20d96d6	429	BTRFS_INODE_MAP_OBJECTID, inode_root);
9f5fae2f CM	430	BUG_ON(ret);
9f5fae2f CM	431
0f7d52f4 CM	432	fs_info->generation = btrfs_super_generation(disk_super) + 1;
0f7d52f4 CM	433	ret = btrfs_find_highest_inode(tree_root, &fs_info->last_inode_alloc);
5be6f7f1 CM	434	if (ret == 0)
5be6f7f1 CM	435	fs_info->highest_inode = fs_info->last_inode_alloc;
d6e4a428 CM	436	memset(&fs_info->kobj, 0, sizeof(fs_info->kobj));
	437	kobj_set_kset_s(fs_info, btrfs_subsys);
	438	kobject_set_name(&fs_info->kobj, "%s", sb->s_id);
	439	kobject_register(&fs_info->kobj);
5be6f7f1	440	mutex_unlock(&fs_info->fs_mutex);
0f7d52f4	441	return tree_root;
eb60ceac CM	442	}
eb60ceac CM	443
e089f05c	444	int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
79154b1b	445	*root)
eb60ceac	446	{
d5719762	447	struct buffer_head *bh = root->fs_info->sb_buffer;
2c90e5d6	448
d5719762 CM	449	btrfs_set_super_root(root->fs_info->disk_super,
	450	root->fs_info->tree_root->node->b_blocknr);
	451	lock_buffer(bh);
2c90e5d6	452	WARN_ON(atomic_read(&bh->b_count) < 1);
d5719762	453	clear_buffer_dirty(bh);
87cbda5c	454	csum_tree_block(root, bh, 0);
d5719762 CM	455	bh->b_end_io = end_buffer_write_sync;
	456	get_bh(bh);
	457	submit_bh(WRITE, bh);
	458	wait_on_buffer(bh);
	459	if (!buffer_uptodate(bh)) {
	460	WARN_ON(1);
	461	return -EIO;
cfaa7295 CM	462	}
	463	return 0;
	464	}
	465
0f7d52f4 CM	466	int del_fs_roots(struct btrfs_fs_info *fs_info)
	467	{
	468	int ret;
	469	struct btrfs_root *gang[8];
	470	int i;
	471
	472	while(1) {
	473	ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
	474	(void **)gang, 0,
	475	ARRAY_SIZE(gang));
	476	if (!ret)
	477	break;
	478	for (i = 0; i < ret; i++) {
	479	radix_tree_delete(&fs_info->fs_roots_radix,
	480	(unsigned long)gang[i]);
	481	if (gang[i]->inode)
	482	iput(gang[i]->inode);
	483	else
	484	printk("no inode for root %p\n", gang[i]);
	485	if (gang[i]->node)
	486	brelse(gang[i]->node);
	487	if (gang[i]->commit_root)
	488	brelse(gang[i]->commit_root);
	489	kfree(gang[i]);
	490	}
	491	}
	492	return 0;
	493	}
	494
e20d96d6	495	int close_ctree(struct btrfs_root *root)
cfaa7295	496	{
3768f368	497	int ret;
e089f05c	498	struct btrfs_trans_handle *trans;
0f7d52f4	499	struct btrfs_fs_info *fs_info = root->fs_info;
e089f05c	500
0f7d52f4	501	mutex_lock(&fs_info->fs_mutex);
79154b1b CM	502	trans = btrfs_start_transaction(root, 1);
	503	btrfs_commit_transaction(trans, root);
	504	/* run commit again to drop the original snapshot */
	505	trans = btrfs_start_transaction(root, 1);
	506	btrfs_commit_transaction(trans, root);
	507	ret = btrfs_write_and_wait_transaction(NULL, root);
3768f368	508	BUG_ON(ret);
79154b1b	509	write_ctree_super(NULL, root);
0f7d52f4 CM	510	mutex_unlock(&fs_info->fs_mutex);
	511
	512	if (fs_info->extent_root->node)
	513	btrfs_block_release(fs_info->extent_root,
	514	fs_info->extent_root->node);
	515	if (fs_info->inode_root->node)
	516	btrfs_block_release(fs_info->inode_root,
	517	fs_info->inode_root->node);
	518	if (fs_info->tree_root->node)
	519	btrfs_block_release(fs_info->tree_root,
	520	fs_info->tree_root->node);
	521	btrfs_block_release(root, fs_info->sb_buffer);
	522	crypto_free_hash(fs_info->hash_tfm);
	523	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
	524	iput(fs_info->btree_inode);
	525	del_fs_roots(fs_info);
	526	kfree(fs_info->extent_root);
	527	kfree(fs_info->inode_root);
	528	kfree(fs_info->tree_root);
	529	kobject_unregister(&fs_info->kobj);
eb60ceac CM	530	return 0;
	531	}
	532
e20d96d6	533	void btrfs_block_release(struct btrfs_root root, struct buffer_head buf)
eb60ceac	534	{
7cfcc17e	535	brelse(buf);
eb60ceac CM	536	}
eb60ceac CM	537