[deliverable/linux.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 <linux/writeback.h>
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
#include "btrfs_inode.h"

u64 bh_blocknr(struct buffer_head *bh)
{
	return bh->b_blocknr;
}

static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
{
	struct btrfs_node *node = btrfs_buffer_node(buf);
	if (bh_blocknr(buf) != btrfs_header_blocknr(&node->header)) {
		printk(KERN_CRIT "bh_blocknr(buf) is %llu, header is %llu\n",
		       (unsigned long long)bh_blocknr(buf),
		       (unsigned long long)btrfs_header_blocknr(&node->header));
		return 1;
	}
	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_blocknr(bh) == blocknr) {
			ret = bh;
			get_bh(bh);
			goto out_unlock;
		}
		bh = bh->b_this_page;
	} while (bh != head);
out_unlock:
	unlock_page(page);
	page_cache_release(page);
	return ret;
}

int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
			     u64 logical)
{
	if (logical == 0) {
		bh->b_bdev = NULL;
		bh->b_blocknr = 0;
		set_buffer_mapped(bh);
	} else {
		map_bh(bh, root->fs_info->sb, logical);
	}
	return 0;
}

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;
	int err;
	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)) {
			err = btrfs_map_bh_to_logical(root, bh, first_block);
			BUG_ON(err);
		}
		if (bh_blocknr(bh) == 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 int btree_get_block(struct inode *inode, sector_t iblock,
			   struct buffer_head *bh, int create)
{
	int err;
	struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root;
	err = btrfs_map_bh_to_logical(root, bh, iblock);
	return err;
}

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("digest failed\n");
	}
	return ret;
}
static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
			   int verify)
{
	char result[BTRFS_CRC32_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_CRC32_SIZE)) {
			printk("btrfs: %s checksum verify failed on %llu\n",
			       root->fs_info->sb->s_id,
			       (unsigned long long)bh_blocknr(bh));
			return 1;
		}
	} else {
		node = btrfs_buffer_node(bh);
		memcpy(node->header.csum, result, BTRFS_CRC32_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,
};

int readahead_tree_block(struct btrfs_root *root, u64 blocknr)
{
	struct buffer_head *bh = NULL;
	int ret = 0;

	bh = btrfs_find_create_tree_block(root, blocknr);
	if (!bh)
		return 0;
	if (buffer_uptodate(bh)) {
		ret = 1;
		goto done;
	}
	if (test_set_buffer_locked(bh)) {
		ret = 1;
		goto done;
	}
	if (!buffer_uptodate(bh)) {
		get_bh(bh);
		bh->b_end_io = end_buffer_read_sync;
		submit_bh(READ, bh);
	} else {
		unlock_buffer(bh);
		ret = 1;
	}
done:
	brelse(bh);
	return ret;
}

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;
	} else {
		unlock_buffer(bh);
	}
uptodate:
	if (!buffer_checked(bh)) {
		csum_tree_block(root, bh, 1);
		set_buffer_checked(bh);
	}
	if (check_tree_block(root, bh))
		goto fail;
	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;
	root->highest_inode = 0;
	root->last_inode_alloc = 0;
	memset(&root->root_key, 0, sizeof(root->root_key));
	memset(&root->root_item, 0, sizeof(root->root_item));
	root->root_key.objectid = objectid;
	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;
	u64 highest_inode;
	int ret = 0;

	root = radix_tree_lookup(&fs_info->fs_roots_radix,
				 (unsigned long)location->objectid);
	if (root)
		return root;
	root = kmalloc(sizeof(*root), GFP_NOFS);
	if (!root)
		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) {
			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) {
		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:
	root->ref_cows = 1;
	ret = radix_tree_insert(&fs_info->fs_roots_radix,
				(unsigned long)root->root_key.objectid,
				root);
	if (ret) {
		brelse(root->node);
		kfree(root);
		return ERR_PTR(ret);
	}
	ret = btrfs_find_highest_inode(root, &highest_inode);
	if (ret == 0) {
		root->highest_inode = highest_inode;
		root->last_inode_alloc = highest_inode;
	}
	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_fs_info *fs_info = kmalloc(sizeof(*fs_info),
						GFP_NOFS);
	int ret;
	int err = -EIO;
	struct btrfs_super_block *disk_super;

	if (!extent_root || !tree_root || !fs_info) {
		err = -ENOMEM;
		goto fail;
	}
	init_bit_radix(&fs_info->pinned_radix);
	init_bit_radix(&fs_info->pending_del_radix);
	init_bit_radix(&fs_info->extent_map_radix);
	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
	INIT_RADIX_TREE(&fs_info->block_group_radix, GFP_KERNEL);
	INIT_RADIX_TREE(&fs_info->block_group_data_radix, GFP_KERNEL);
	INIT_LIST_HEAD(&fs_info->trans_list);
	INIT_LIST_HEAD(&fs_info->dead_roots);
	sb_set_blocksize(sb, 4096);
	fs_info->running_transaction = NULL;
	fs_info->tree_root = tree_root;
	fs_info->extent_root = extent_root;
	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;
	fs_info->do_barriers = 1;
	fs_info->extent_tree_insert_nr = 0;
	fs_info->extent_tree_prealloc_nr = 0;
	fs_info->closing = 0;

	INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
	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("crc32c", 0, CRYPTO_ALG_ASYNC);
	spin_lock_init(&fs_info->hash_lock);

	if (!fs_info->hash_tfm || IS_ERR(fs_info->hash_tfm)) {
		printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
		err = -ENOMEM;
		goto fail_iput;
	}
	mutex_init(&fs_info->trans_mutex);
	mutex_init(&fs_info->fs_mutex);

	__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)
		goto fail_iput;
	disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;

	if (!btrfs_super_root(disk_super))
		goto fail_sb_buffer;

	i_size_write(fs_info->btree_inode,
		     btrfs_super_total_blocks(disk_super) <<
		     fs_info->btree_inode->i_blkbits);

	fs_info->disk_super = disk_super;

	if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
		    sizeof(disk_super->magic))) {
		printk("btrfs: valid FS not found on %s\n", sb->s_id);
		goto fail_sb_buffer;
	}
	tree_root->node = read_tree_block(tree_root,
					  btrfs_super_root(disk_super));
	if (!tree_root->node)
		goto fail_sb_buffer;

	mutex_lock(&fs_info->fs_mutex);
	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
	if (ret) {
		mutex_unlock(&fs_info->fs_mutex);
		goto fail_tree_root;
	}

	btrfs_read_block_groups(extent_root);

	fs_info->generation = btrfs_super_generation(disk_super) + 1;
	mutex_unlock(&fs_info->fs_mutex);
	return tree_root;

fail_tree_root:
	btrfs_block_release(tree_root, tree_root->node);
fail_sb_buffer:
	btrfs_block_release(tree_root, fs_info->sb_buffer);
fail_iput:
	iput(fs_info->btree_inode);
fail:
	kfree(extent_root);
	kfree(tree_root);
	kfree(fs_info);
	return ERR_PTR(err);
}

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

	btrfs_set_super_root(root->fs_info->disk_super,
			     bh_blocknr(root->fs_info->tree_root->node));
	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);
	if (root->fs_info->do_barriers)
		ret = submit_bh(WRITE_BARRIER, bh);
	else
		ret = submit_bh(WRITE, bh);
	if (ret == -EOPNOTSUPP) {
		set_buffer_uptodate(bh);
		root->fs_info->do_barriers = 0;
		ret = submit_bh(WRITE, bh);
	}
	wait_on_buffer(bh);
	if (!buffer_uptodate(bh)) {
		WARN_ON(1);
		return -EIO;
	}
	return 0;
}

static int free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
{
	radix_tree_delete(&fs_info->fs_roots_radix,
			  (unsigned long)root->root_key.objectid);
	if (root->inode)
		iput(root->inode);
	if (root->node)
		brelse(root->node);
	if (root->commit_root)
		brelse(root->commit_root);
	kfree(root);
	return 0;
}

static 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++)
			free_fs_root(fs_info, 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;

	fs_info->closing = 1;
	btrfs_transaction_flush_work(root);
	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->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);

	btrfs_free_block_groups(root->fs_info);
	del_fs_roots(fs_info);
	kfree(fs_info->extent_root);
	kfree(fs_info->tree_root);
	return 0;
}

void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
{
	brelse(buf);
}

void btrfs_btree_balance_dirty(struct btrfs_root *root)
{
	balance_dirty_pages_ratelimited(root->fs_info->btree_inode->i_mapping);
}
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>
35b7e476	8	#include <linux/writeback.h>
eb60ceac CM	9	#include "ctree.h"
eb60ceac CM	10	#include "disk-io.h"
e089f05c	11	#include "transaction.h"
0f7d52f4	12	#include "btrfs_inode.h"
eb60ceac	13
7eccb903 CM	14	u64 bh_blocknr(struct buffer_head *bh)
7eccb903 CM	15	{
0cf6c620	16	return bh->b_blocknr;
7eccb903 CM	17	}
7eccb903 CM	18
e20d96d6	19	static int check_tree_block(struct btrfs_root root, struct buffer_head buf)
eb60ceac	20	{
e20d96d6	21	struct btrfs_node *node = btrfs_buffer_node(buf);
7eccb903	22	if (bh_blocknr(buf) != btrfs_header_blocknr(&node->header)) {
5af3981c CM	23	printk(KERN_CRIT "bh_blocknr(buf) is %llu, header is %llu\n",
	24	(unsigned long long)bh_blocknr(buf),
	25	(unsigned long long)btrfs_header_blocknr(&node->header));
39279cc3	26	return 1;
d98237b3	27	}
9a8dd150	28	return 0;
eb60ceac CM	29	}
eb60ceac CM	30
d98237b3 CM	31	struct buffer_head btrfs_find_tree_block(struct btrfs_root root, u64 blocknr)
	32	{
	33	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
	34	int blockbits = root->fs_info->sb->s_blocksize_bits;
	35	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
	36	struct page *page;
	37	struct buffer_head *bh;
	38	struct buffer_head *head;
	39	struct buffer_head *ret = NULL;
	40
2c90e5d6	41
d98237b3 CM	42	page = find_lock_page(mapping, index);
	43	if (!page)
	44	return NULL;
	45
	46	if (!page_has_buffers(page))
	47	goto out_unlock;
	48
	49	head = page_buffers(page);
	50	bh = head;
	51	do {
7eccb903	52	if (buffer_mapped(bh) && bh_blocknr(bh) == blocknr) {
d98237b3 CM	53	ret = bh;
	54	get_bh(bh);
	55	goto out_unlock;
	56	}
	57	bh = bh->b_this_page;
	58	} while (bh != head);
	59	out_unlock:
	60	unlock_page(page);
	61	page_cache_release(page);
	62	return ret;
	63	}
	64
8352d8a4	65	int btrfs_map_bh_to_logical(struct btrfs_root root, struct buffer_head bh,
7eccb903 CM	66	u64 logical)
7eccb903 CM	67	{
236454df CM	68	if (logical == 0) {
	69	bh->b_bdev = NULL;
	70	bh->b_blocknr = 0;
	71	set_buffer_mapped(bh);
0cf6c620 CM	72	} else {
0cf6c620 CM	73	map_bh(bh, root->fs_info->sb, logical);
7eccb903	74	}
0cf6c620	75	return 0;
7eccb903 CM	76	}
7eccb903 CM	77
d98237b3 CM	78	struct buffer_head btrfs_find_create_tree_block(struct btrfs_root root,
	79	u64 blocknr)
	80	{
	81	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
	82	int blockbits = root->fs_info->sb->s_blocksize_bits;
	83	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
	84	struct page *page;
	85	struct buffer_head *bh;
	86	struct buffer_head *head;
	87	struct buffer_head *ret = NULL;
7eccb903	88	int err;
d98237b3	89	u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
22b0ebda	90
d98237b3 CM	91	page = grab_cache_page(mapping, index);
	92	if (!page)
	93	return NULL;
	94
d98237b3 CM	95	if (!page_has_buffers(page))
	96	create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
	97	head = page_buffers(page);
	98	bh = head;
	99	do {
	100	if (!buffer_mapped(bh)) {
8352d8a4	101	err = btrfs_map_bh_to_logical(root, bh, first_block);
7eccb903	102	BUG_ON(err);
d98237b3	103	}
7eccb903	104	if (bh_blocknr(bh) == blocknr) {
d98237b3 CM	105	ret = bh;
	106	get_bh(bh);
	107	goto out_unlock;
	108	}
	109	bh = bh->b_this_page;
	110	first_block++;
	111	} while (bh != head);
	112	out_unlock:
	113	unlock_page(page);
22b0ebda CM	114	if (ret)
22b0ebda CM	115	touch_buffer(ret);
d98237b3 CM	116	page_cache_release(page);
	117	return ret;
	118	}
	119
d98237b3 CM	120	static int btree_get_block(struct inode *inode, sector_t iblock,
	121	struct buffer_head *bh, int create)
	122	{
7eccb903 CM	123	int err;
7eccb903 CM	124	struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root;
8352d8a4	125	err = btrfs_map_bh_to_logical(root, bh, iblock);
7eccb903	126	return err;
d98237b3 CM	127	}
d98237b3 CM	128
f254e52c CM	129	int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
f254e52c CM	130	char *result)
87cbda5c	131	{
87cbda5c CM	132	struct scatterlist sg;
	133	struct crypto_hash *tfm = root->fs_info->hash_tfm;
	134	struct hash_desc desc;
	135	int ret;
87cbda5c CM	136
	137	desc.tfm = tfm;
	138	desc.flags = 0;
f254e52c	139	sg_init_one(&sg, data, len);
87cbda5c	140	spin_lock(&root->fs_info->hash_lock);
22b0ebda	141	ret = crypto_hash_digest(&desc, &sg, 1, result);
87cbda5c CM	142	spin_unlock(&root->fs_info->hash_lock);
87cbda5c CM	143	if (ret) {
509659cd	144	printk("digest failed\n");
87cbda5c	145	}
f254e52c CM	146	return ret;
	147	}
	148	static int csum_tree_block(struct btrfs_root root, struct buffer_head bh,
	149	int verify)
	150	{
509659cd	151	char result[BTRFS_CRC32_SIZE];
f254e52c CM	152	int ret;
	153	struct btrfs_node *node;
	154
	155	ret = btrfs_csum_data(root, bh->b_data + BTRFS_CSUM_SIZE,
	156	bh->b_size - BTRFS_CSUM_SIZE, result);
	157	if (ret)
	158	return ret;
87cbda5c	159	if (verify) {
509659cd	160	if (memcmp(bh->b_data, result, BTRFS_CRC32_SIZE)) {
5af3981c CM	161	printk("btrfs: %s checksum verify failed on %llu\n",
	162	root->fs_info->sb->s_id,
	163	(unsigned long long)bh_blocknr(bh));
f254e52c CM	164	return 1;
	165	}
	166	} else {
	167	node = btrfs_buffer_node(bh);
509659cd	168	memcpy(node->header.csum, result, BTRFS_CRC32_SIZE);
f254e52c	169	}
87cbda5c CM	170	return 0;
	171	}
	172
d98237b3	173	static int btree_writepage(struct page page, struct writeback_control wbc)
ed2ff2cb	174	{
87cbda5c	175	struct buffer_head *bh;
0f7d52f4	176	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
87cbda5c	177	struct buffer_head *head;
87cbda5c CM	178	if (!page_has_buffers(page)) {
	179	create_empty_buffers(page, root->fs_info->sb->s_blocksize,
	180	(1 << BH_Dirty)\|(1 << BH_Uptodate));
	181	}
	182	head = page_buffers(page);
	183	bh = head;
	184	do {
	185	if (buffer_dirty(bh))
	186	csum_tree_block(root, bh, 0);
	187	bh = bh->b_this_page;
	188	} while (bh != head);
d98237b3	189	return block_write_full_page(page, btree_get_block, wbc);
ed2ff2cb CM	190	}
ed2ff2cb CM	191
d98237b3	192	static int btree_readpage(struct file * file, struct page * page)
eb60ceac	193	{
d98237b3	194	return block_read_full_page(page, btree_get_block);
eb60ceac CM	195	}
eb60ceac CM	196
d98237b3 CM	197	static struct address_space_operations btree_aops = {
	198	.readpage = btree_readpage,
	199	.writepage = btree_writepage,
	200	.sync_page = block_sync_page,
	201	};
	202
090d1875 CM	203	int readahead_tree_block(struct btrfs_root *root, u64 blocknr)
	204	{
	205	struct buffer_head *bh = NULL;
de428b63	206	int ret = 0;
090d1875 CM	207
	208	bh = btrfs_find_create_tree_block(root, blocknr);
	209	if (!bh)
	210	return 0;
de428b63 CM	211	if (buffer_uptodate(bh)) {
de428b63 CM	212	ret = 1;
090d1875	213	goto done;
de428b63 CM	214	}
	215	if (test_set_buffer_locked(bh)) {
	216	ret = 1;
090d1875	217	goto done;
de428b63	218	}
090d1875 CM	219	if (!buffer_uptodate(bh)) {
	220	get_bh(bh);
	221	bh->b_end_io = end_buffer_read_sync;
	222	submit_bh(READ, bh);
	223	} else {
	224	unlock_buffer(bh);
de428b63	225	ret = 1;
090d1875 CM	226	}
	227	done:
	228	brelse(bh);
de428b63	229	return ret;
090d1875 CM	230	}
090d1875 CM	231
e20d96d6	232	struct buffer_head read_tree_block(struct btrfs_root root, u64 blocknr)
eb60ceac	233	{
d98237b3	234	struct buffer_head *bh = NULL;
eb60ceac	235
d98237b3 CM	236	bh = btrfs_find_create_tree_block(root, blocknr);
	237	if (!bh)
	238	return bh;
9d64272c CM	239	if (buffer_uptodate(bh))
9d64272c CM	240	goto uptodate;
d98237b3 CM	241	lock_buffer(bh);
	242	if (!buffer_uptodate(bh)) {
	243	get_bh(bh);
	244	bh->b_end_io = end_buffer_read_sync;
	245	submit_bh(READ, bh);
	246	wait_on_buffer(bh);
	247	if (!buffer_uptodate(bh))
	248	goto fail;
	249	} else {
	250	unlock_buffer(bh);
	251	}
9d64272c	252	uptodate:
090d1875 CM	253	if (!buffer_checked(bh)) {
	254	csum_tree_block(root, bh, 1);
	255	set_buffer_checked(bh);
	256	}
d98237b3	257	if (check_tree_block(root, bh))
39279cc3	258	goto fail;
d98237b3 CM	259	return bh;
	260	fail:
	261	brelse(bh);
	262	return NULL;
eb60ceac CM	263	}
eb60ceac CM	264
e089f05c	265	int dirty_tree_block(struct btrfs_trans_handle trans, struct btrfs_root root,
e20d96d6	266	struct buffer_head *buf)
ed2ff2cb	267	{
d6025579	268	WARN_ON(atomic_read(&buf->b_count) == 0);
e20d96d6	269	mark_buffer_dirty(buf);
ed2ff2cb CM	270	return 0;
	271	}
	272
e089f05c	273	int clean_tree_block(struct btrfs_trans_handle trans, struct btrfs_root root,
e20d96d6	274	struct buffer_head *buf)
ed2ff2cb	275	{
d6025579	276	WARN_ON(atomic_read(&buf->b_count) == 0);
e20d96d6	277	clear_buffer_dirty(buf);
ed2ff2cb CM	278	return 0;
	279	}
	280
2c90e5d6	281	static int __setup_root(int blocksize,
9f5fae2f CM	282	struct btrfs_root *root,
9f5fae2f CM	283	struct btrfs_fs_info *fs_info,
e20d96d6	284	u64 objectid)
d97e63b6	285	{
cfaa7295	286	root->node = NULL;
0f7d52f4	287	root->inode = NULL;
a28ec197	288	root->commit_root = NULL;
2c90e5d6	289	root->blocksize = blocksize;
123abc88	290	root->ref_cows = 0;
9f5fae2f	291	root->fs_info = fs_info;
0f7d52f4 CM	292	root->objectid = objectid;
0f7d52f4 CM	293	root->last_trans = 0;
1b05da2e CM	294	root->highest_inode = 0;
1b05da2e CM	295	root->last_inode_alloc = 0;
3768f368 CM	296	memset(&root->root_key, 0, sizeof(root->root_key));
3768f368 CM	297	memset(&root->root_item, 0, sizeof(root->root_item));
4d775673	298	root->root_key.objectid = objectid;
3768f368 CM	299	return 0;
	300	}
	301
2c90e5d6	302	static int find_and_setup_root(int blocksize,
9f5fae2f CM	303	struct btrfs_root *tree_root,
	304	struct btrfs_fs_info *fs_info,
	305	u64 objectid,
e20d96d6	306	struct btrfs_root *root)
3768f368 CM	307	{
	308	int ret;
	309
2c90e5d6	310	__setup_root(blocksize, root, fs_info, objectid);
3768f368 CM	311	ret = btrfs_find_last_root(tree_root, objectid,
	312	&root->root_item, &root->root_key);
	313	BUG_ON(ret);
	314
	315	root->node = read_tree_block(root,
	316	btrfs_root_blocknr(&root->root_item));
3768f368	317	BUG_ON(!root->node);
d97e63b6 CM	318	return 0;
	319	}
	320
0f7d52f4 CM	321	struct btrfs_root btrfs_read_fs_root(struct btrfs_fs_info fs_info,
	322	struct btrfs_key *location)
	323	{
	324	struct btrfs_root *root;
	325	struct btrfs_root *tree_root = fs_info->tree_root;
	326	struct btrfs_path *path;
	327	struct btrfs_leaf *l;
1b05da2e	328	u64 highest_inode;
0f7d52f4 CM	329	int ret = 0;
0f7d52f4 CM	330
2619ba1f CM	331	root = radix_tree_lookup(&fs_info->fs_roots_radix,
2619ba1f CM	332	(unsigned long)location->objectid);
0cf6c620	333	if (root)
2619ba1f	334	return root;
0f7d52f4	335	root = kmalloc(sizeof(*root), GFP_NOFS);
0cf6c620	336	if (!root)
0f7d52f4	337	return ERR_PTR(-ENOMEM);
0f7d52f4 CM	338	if (location->offset == (u64)-1) {
	339	ret = find_and_setup_root(fs_info->sb->s_blocksize,
	340	fs_info->tree_root, fs_info,
	341	location->objectid, root);
	342	if (ret) {
0f7d52f4 CM	343	kfree(root);
	344	return ERR_PTR(ret);
	345	}
	346	goto insert;
	347	}
	348
	349	__setup_root(fs_info->sb->s_blocksize, root, fs_info,
	350	location->objectid);
	351
	352	path = btrfs_alloc_path();
	353	BUG_ON(!path);
	354	ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
	355	if (ret != 0) {
0f7d52f4 CM	356	if (ret > 0)
	357	ret = -ENOENT;
	358	goto out;
	359	}
	360	l = btrfs_buffer_leaf(path->nodes[0]);
	361	memcpy(&root->root_item,
	362	btrfs_item_ptr(l, path->slots[0], struct btrfs_root_item),
	363	sizeof(root->root_item));
	364	memcpy(&root->root_key, location, sizeof(*location));
	365	ret = 0;
	366	out:
	367	btrfs_release_path(root, path);
	368	btrfs_free_path(path);
	369	if (ret) {
	370	kfree(root);
	371	return ERR_PTR(ret);
	372	}
	373	root->node = read_tree_block(root,
	374	btrfs_root_blocknr(&root->root_item));
	375	BUG_ON(!root->node);
	376	insert:
0f7d52f4	377	root->ref_cows = 1;
2619ba1f CM	378	ret = radix_tree_insert(&fs_info->fs_roots_radix,
2619ba1f CM	379	(unsigned long)root->root_key.objectid,
0f7d52f4 CM	380	root);
0f7d52f4 CM	381	if (ret) {
0f7d52f4 CM	382	brelse(root->node);
	383	kfree(root);
	384	return ERR_PTR(ret);
	385	}
1b05da2e CM	386	ret = btrfs_find_highest_inode(root, &highest_inode);
	387	if (ret == 0) {
	388	root->highest_inode = highest_inode;
	389	root->last_inode_alloc = highest_inode;
1b05da2e	390	}
0f7d52f4 CM	391	return root;
	392	}
	393
2c90e5d6	394	struct btrfs_root open_ctree(struct super_block sb)
2e635a27	395	{
e20d96d6 CM	396	struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
	397	GFP_NOFS);
	398	struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
	399	GFP_NOFS);
e20d96d6 CM	400	struct btrfs_fs_info fs_info = kmalloc(sizeof(fs_info),
e20d96d6 CM	401	GFP_NOFS);
eb60ceac	402	int ret;
39279cc3	403	int err = -EIO;
2c90e5d6	404	struct btrfs_super_block *disk_super;
eb60ceac	405
39279cc3 CM	406	if (!extent_root \|\| !tree_root \|\| !fs_info) {
	407	err = -ENOMEM;
	408	goto fail;
	409	}
8ef97622 CM	410	init_bit_radix(&fs_info->pinned_radix);
8ef97622 CM	411	init_bit_radix(&fs_info->pending_del_radix);
e37c9e69	412	init_bit_radix(&fs_info->extent_map_radix);
0f7d52f4	413	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
9078a3e1	414	INIT_RADIX_TREE(&fs_info->block_group_radix, GFP_KERNEL);
be744175	415	INIT_RADIX_TREE(&fs_info->block_group_data_radix, GFP_KERNEL);
8fd17795	416	INIT_LIST_HEAD(&fs_info->trans_list);
facda1e7	417	INIT_LIST_HEAD(&fs_info->dead_roots);
2c90e5d6	418	sb_set_blocksize(sb, 4096);
9f5fae2f	419	fs_info->running_transaction = NULL;
9f5fae2f CM	420	fs_info->tree_root = tree_root;
9f5fae2f CM	421	fs_info->extent_root = extent_root;
e20d96d6	422	fs_info->sb = sb;
d98237b3 CM	423	fs_info->btree_inode = new_inode(sb);
d98237b3 CM	424	fs_info->btree_inode->i_ino = 1;
2c90e5d6	425	fs_info->btree_inode->i_nlink = 1;
d98237b3 CM	426	fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
d98237b3 CM	427	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
e66f709b	428	fs_info->do_barriers = 1;
f2458e1d CM	429	fs_info->extent_tree_insert_nr = 0;
f2458e1d CM	430	fs_info->extent_tree_prealloc_nr = 0;
facda1e7 CM	431	fs_info->closing = 0;
facda1e7 CM	432
08607c1b	433	INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
0f7d52f4 CM	434	BTRFS_I(fs_info->btree_inode)->root = tree_root;
	435	memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
	436	sizeof(struct btrfs_key));
22b0ebda	437	insert_inode_hash(fs_info->btree_inode);
d98237b3	438	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
509659cd	439	fs_info->hash_tfm = crypto_alloc_hash("crc32c", 0, CRYPTO_ALG_ASYNC);
30ae8467	440	spin_lock_init(&fs_info->hash_lock);
39279cc3	441
30ae8467	442	if (!fs_info->hash_tfm \|\| IS_ERR(fs_info->hash_tfm)) {
39279cc3 CM	443	printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
	444	err = -ENOMEM;
	445	goto fail_iput;
87cbda5c	446	}
79154b1b	447	mutex_init(&fs_info->trans_mutex);
d561c025	448	mutex_init(&fs_info->fs_mutex);
3768f368	449
2c90e5d6 CM	450	__setup_root(sb->s_blocksize, tree_root,
2c90e5d6 CM	451	fs_info, BTRFS_ROOT_TREE_OBJECTID);
7eccb903	452
2c90e5d6 CM	453	fs_info->sb_buffer = read_tree_block(tree_root,
	454	BTRFS_SUPER_INFO_OFFSET /
	455	sb->s_blocksize);
d98237b3	456
0f7d52f4	457	if (!fs_info->sb_buffer)
39279cc3	458	goto fail_iput;
d98237b3	459	disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
39279cc3	460
0f7d52f4	461	if (!btrfs_super_root(disk_super))
39279cc3	462	goto fail_sb_buffer;
0f7d52f4	463
8352d8a4 CM	464	i_size_write(fs_info->btree_inode,
	465	btrfs_super_total_blocks(disk_super) <<
	466	fs_info->btree_inode->i_blkbits);
	467
d98237b3	468	fs_info->disk_super = disk_super;
39279cc3 CM	469
	470	if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
	471	sizeof(disk_super->magic))) {
	472	printk("btrfs: valid FS not found on %s\n", sb->s_id);
	473	goto fail_sb_buffer;
	474	}
e20d96d6 CM	475	tree_root->node = read_tree_block(tree_root,
e20d96d6 CM	476	btrfs_super_root(disk_super));
39279cc3 CM	477	if (!tree_root->node)
39279cc3 CM	478	goto fail_sb_buffer;
3768f368	479
2c90e5d6 CM	480	mutex_lock(&fs_info->fs_mutex);
2c90e5d6 CM	481	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
e20d96d6	482	BTRFS_EXTENT_TREE_OBJECTID, extent_root);
39279cc3 CM	483	if (ret) {
	484	mutex_unlock(&fs_info->fs_mutex);
	485	goto fail_tree_root;
	486	}
3768f368	487
9078a3e1 CM	488	btrfs_read_block_groups(extent_root);
9078a3e1 CM	489
0f7d52f4	490	fs_info->generation = btrfs_super_generation(disk_super) + 1;
5be6f7f1	491	mutex_unlock(&fs_info->fs_mutex);
0f7d52f4	492	return tree_root;
39279cc3 CM	493
	494	fail_tree_root:
	495	btrfs_block_release(tree_root, tree_root->node);
	496	fail_sb_buffer:
	497	btrfs_block_release(tree_root, fs_info->sb_buffer);
	498	fail_iput:
	499	iput(fs_info->btree_inode);
	500	fail:
	501	kfree(extent_root);
	502	kfree(tree_root);
	503	kfree(fs_info);
	504	return ERR_PTR(err);
eb60ceac CM	505	}
eb60ceac CM	506
e089f05c	507	int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
79154b1b	508	*root)
eb60ceac	509	{
e66f709b	510	int ret;
d5719762	511	struct buffer_head *bh = root->fs_info->sb_buffer;
2c90e5d6	512
d5719762	513	btrfs_set_super_root(root->fs_info->disk_super,
7eccb903	514	bh_blocknr(root->fs_info->tree_root->node));
d5719762	515	lock_buffer(bh);
2c90e5d6	516	WARN_ON(atomic_read(&bh->b_count) < 1);
d5719762	517	clear_buffer_dirty(bh);
87cbda5c	518	csum_tree_block(root, bh, 0);
d5719762 CM	519	bh->b_end_io = end_buffer_write_sync;
d5719762 CM	520	get_bh(bh);
e66f709b CM	521	if (root->fs_info->do_barriers)
	522	ret = submit_bh(WRITE_BARRIER, bh);
	523	else
	524	ret = submit_bh(WRITE, bh);
	525	if (ret == -EOPNOTSUPP) {
	526	set_buffer_uptodate(bh);
	527	root->fs_info->do_barriers = 0;
	528	ret = submit_bh(WRITE, bh);
	529	}
d5719762 CM	530	wait_on_buffer(bh);
	531	if (!buffer_uptodate(bh)) {
	532	WARN_ON(1);
	533	return -EIO;
cfaa7295 CM	534	}
	535	return 0;
	536	}
	537
2619ba1f CM	538	static int free_fs_root(struct btrfs_fs_info fs_info, struct btrfs_root root)
	539	{
	540	radix_tree_delete(&fs_info->fs_roots_radix,
	541	(unsigned long)root->root_key.objectid);
	542	if (root->inode)
	543	iput(root->inode);
	544	if (root->node)
	545	brelse(root->node);
	546	if (root->commit_root)
	547	brelse(root->commit_root);
	548	kfree(root);
	549	return 0;
	550	}
	551
35b7e476	552	static int del_fs_roots(struct btrfs_fs_info *fs_info)
0f7d52f4 CM	553	{
	554	int ret;
	555	struct btrfs_root *gang[8];
	556	int i;
	557
	558	while(1) {
	559	ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
	560	(void **)gang, 0,
	561	ARRAY_SIZE(gang));
	562	if (!ret)
	563	break;
2619ba1f CM	564	for (i = 0; i < ret; i++)
2619ba1f CM	565	free_fs_root(fs_info, gang[i]);
0f7d52f4 CM	566	}
	567	return 0;
	568	}
b4100d64	569
e20d96d6	570	int close_ctree(struct btrfs_root *root)
cfaa7295	571	{
3768f368	572	int ret;
e089f05c	573	struct btrfs_trans_handle *trans;
0f7d52f4	574	struct btrfs_fs_info *fs_info = root->fs_info;
e089f05c	575
facda1e7	576	fs_info->closing = 1;
08607c1b	577	btrfs_transaction_flush_work(root);
0f7d52f4	578	mutex_lock(&fs_info->fs_mutex);
79154b1b CM	579	trans = btrfs_start_transaction(root, 1);
	580	btrfs_commit_transaction(trans, root);
	581	/* run commit again to drop the original snapshot */
	582	trans = btrfs_start_transaction(root, 1);
	583	btrfs_commit_transaction(trans, root);
	584	ret = btrfs_write_and_wait_transaction(NULL, root);
3768f368	585	BUG_ON(ret);
79154b1b	586	write_ctree_super(NULL, root);
0f7d52f4 CM	587	mutex_unlock(&fs_info->fs_mutex);
	588
	589	if (fs_info->extent_root->node)
	590	btrfs_block_release(fs_info->extent_root,
	591	fs_info->extent_root->node);
0f7d52f4 CM	592	if (fs_info->tree_root->node)
	593	btrfs_block_release(fs_info->tree_root,
	594	fs_info->tree_root->node);
	595	btrfs_block_release(root, fs_info->sb_buffer);
	596	crypto_free_hash(fs_info->hash_tfm);
	597	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
	598	iput(fs_info->btree_inode);
7eccb903	599
9078a3e1	600	btrfs_free_block_groups(root->fs_info);
0f7d52f4 CM	601	del_fs_roots(fs_info);
0f7d52f4 CM	602	kfree(fs_info->extent_root);
0f7d52f4	603	kfree(fs_info->tree_root);
eb60ceac CM	604	return 0;
	605	}
	606
e20d96d6	607	void btrfs_block_release(struct btrfs_root root, struct buffer_head buf)
eb60ceac	608	{
7cfcc17e	609	brelse(buf);
eb60ceac CM	610	}
eb60ceac CM	611
35b7e476 CM	612	void btrfs_btree_balance_dirty(struct btrfs_root *root)
	613	{
	614	balance_dirty_pages_ratelimited(root->fs_info->btree_inode->i_mapping);
	615	}