[net-next-2.6.git] / fs / ext4 / migrate.c

/*
 * Copyright IBM Corporation, 2007
 * Author Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of version 2.1 of the GNU Lesser General Public License
 * as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it would be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
 *
 */

#include <linux/module.h>
#include "ext4_jbd2.h"
#include "ext4_extents.h"

/*
 * The contiguous blocks details which can be
 * represented by a single extent
 */
struct list_blocks_struct {
	ext4_lblk_t first_block, last_block;
	ext4_fsblk_t first_pblock, last_pblock;
};

static int finish_range(handle_t *handle, struct inode *inode,
				struct list_blocks_struct *lb)

{
	int retval = 0, needed;
	struct ext4_extent newext;
	struct ext4_ext_path *path;
	if (lb->first_pblock == 0)
		return 0;

	/* Add the extent to temp inode*/
	newext.ee_block = cpu_to_le32(lb->first_block);
	newext.ee_len   = cpu_to_le16(lb->last_block - lb->first_block + 1);
	ext4_ext_store_pblock(&newext, lb->first_pblock);
	path = ext4_ext_find_extent(inode, lb->first_block, NULL);

	if (IS_ERR(path)) {
		retval = PTR_ERR(path);
		path = NULL;
		goto err_out;
	}

	/*
	 * Calculate the credit needed to inserting this extent
	 * Since we are doing this in loop we may accumalate extra
	 * credit. But below we try to not accumalate too much
	 * of them by restarting the journal.
	 */
	needed = ext4_ext_calc_credits_for_single_extent(inode,
		    lb->last_block - lb->first_block + 1, path);

	/*
	 * Make sure the credit we accumalated is not really high
	 */
	if (needed && ext4_handle_has_enough_credits(handle,
						EXT4_RESERVE_TRANS_BLOCKS)) {
		retval = ext4_journal_restart(handle, needed);
		if (retval)
			goto err_out;
	} else if (needed) {
		retval = ext4_journal_extend(handle, needed);
		if (retval) {
			/*
			 * IF not able to extend the journal restart the journal
			 */
			retval = ext4_journal_restart(handle, needed);
			if (retval)
				goto err_out;
		}
	}
	retval = ext4_ext_insert_extent(handle, inode, path, &newext, 0);
err_out:
	if (path) {
		ext4_ext_drop_refs(path);
		kfree(path);
	}
	lb->first_pblock = 0;
	return retval;
}

static int update_extent_range(handle_t *handle, struct inode *inode,
				ext4_fsblk_t pblock, ext4_lblk_t blk_num,
				struct list_blocks_struct *lb)
{
	int retval;
	/*
	 * See if we can add on to the existing range (if it exists)
	 */
	if (lb->first_pblock &&
		(lb->last_pblock+1 == pblock) &&
		(lb->last_block+1 == blk_num)) {
		lb->last_pblock = pblock;
		lb->last_block = blk_num;
		return 0;
	}
	/*
	 * Start a new range.
	 */
	retval = finish_range(handle, inode, lb);
	lb->first_pblock = lb->last_pblock = pblock;
	lb->first_block = lb->last_block = blk_num;

	return retval;
}

static int update_ind_extent_range(handle_t *handle, struct inode *inode,
				   ext4_fsblk_t pblock, ext4_lblk_t *blk_nump,
				   struct list_blocks_struct *lb)
{
	struct buffer_head *bh;
	__le32 *i_data;
	int i, retval = 0;
	ext4_lblk_t blk_count = *blk_nump;
	unsigned long max_entries = inode->i_sb->s_blocksize >> 2;

	if (!pblock) {
		/* Only update the file block number */
		*blk_nump += max_entries;
		return 0;
	}

	bh = sb_bread(inode->i_sb, pblock);
	if (!bh)
		return -EIO;

	i_data = (__le32 *)bh->b_data;
	for (i = 0; i < max_entries; i++, blk_count++) {
		if (i_data[i]) {
			retval = update_extent_range(handle, inode,
						le32_to_cpu(i_data[i]),
						blk_count, lb);
			if (retval)
				break;
		}
	}

	/* Update the file block number */
	*blk_nump = blk_count;
	put_bh(bh);
	return retval;

}

static int update_dind_extent_range(handle_t *handle, struct inode *inode,
				    ext4_fsblk_t pblock, ext4_lblk_t *blk_nump,
				    struct list_blocks_struct *lb)
{
	struct buffer_head *bh;
	__le32 *i_data;
	int i, retval = 0;
	ext4_lblk_t blk_count = *blk_nump;
	unsigned long max_entries = inode->i_sb->s_blocksize >> 2;

	if (!pblock) {
		/* Only update the file block number */
		*blk_nump += max_entries * max_entries;
		return 0;
	}
	bh = sb_bread(inode->i_sb, pblock);
	if (!bh)
		return -EIO;

	i_data = (__le32 *)bh->b_data;
	for (i = 0; i < max_entries; i++) {
		if (i_data[i]) {
			retval = update_ind_extent_range(handle, inode,
						le32_to_cpu(i_data[i]),
						&blk_count, lb);
			if (retval)
				break;
		} else {
			/* Only update the file block number */
			blk_count += max_entries;
		}
	}

	/* Update the file block number */
	*blk_nump = blk_count;
	put_bh(bh);
	return retval;

}

static int update_tind_extent_range(handle_t *handle, struct inode *inode,
				     ext4_fsblk_t pblock, ext4_lblk_t *blk_nump,
				     struct list_blocks_struct *lb)
{
	struct buffer_head *bh;
	__le32 *i_data;
	int i, retval = 0;
	ext4_lblk_t blk_count = *blk_nump;
	unsigned long max_entries = inode->i_sb->s_blocksize >> 2;

	if (!pblock) {
		/* Only update the file block number */
		*blk_nump += max_entries * max_entries * max_entries;
		return 0;
	}
	bh = sb_bread(inode->i_sb, pblock);
	if (!bh)
		return -EIO;

	i_data = (__le32 *)bh->b_data;
	for (i = 0; i < max_entries; i++) {
		if (i_data[i]) {
			retval = update_dind_extent_range(handle, inode,
						le32_to_cpu(i_data[i]),
						&blk_count, lb);
			if (retval)
				break;
		} else
			/* Only update the file block number */
			blk_count += max_entries * max_entries;
	}
	/* Update the file block number */
	*blk_nump = blk_count;
	put_bh(bh);
	return retval;

}

static int extend_credit_for_blkdel(handle_t *handle, struct inode *inode)
{
	int retval = 0, needed;

	if (ext4_handle_has_enough_credits(handle, EXT4_RESERVE_TRANS_BLOCKS+1))
		return 0;
	/*
	 * We are freeing a blocks. During this we touch
	 * superblock, group descriptor and block bitmap.
	 * So allocate a credit of 3. We may update
	 * quota (user and group).
	 */
	needed = 3 + 2*EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb);

	if (ext4_journal_extend(handle, needed) != 0)
		retval = ext4_journal_restart(handle, needed);

	return retval;
}

static int free_dind_blocks(handle_t *handle,
				struct inode *inode, __le32 i_data)
{
	int i;
	__le32 *tmp_idata;
	struct buffer_head *bh;
	unsigned long max_entries = inode->i_sb->s_blocksize >> 2;

	bh = sb_bread(inode->i_sb, le32_to_cpu(i_data));
	if (!bh)
		return -EIO;

	tmp_idata = (__le32 *)bh->b_data;
	for (i = 0; i < max_entries; i++) {
		if (tmp_idata[i]) {
			extend_credit_for_blkdel(handle, inode);
			ext4_free_blocks(handle, inode,
					le32_to_cpu(tmp_idata[i]), 1, 1);
		}
	}
	put_bh(bh);
	extend_credit_for_blkdel(handle, inode);
	ext4_free_blocks(handle, inode, le32_to_cpu(i_data), 1, 1);
	return 0;
}

static int free_tind_blocks(handle_t *handle,
				struct inode *inode, __le32 i_data)
{
	int i, retval = 0;
	__le32 *tmp_idata;
	struct buffer_head *bh;
	unsigned long max_entries = inode->i_sb->s_blocksize >> 2;

	bh = sb_bread(inode->i_sb, le32_to_cpu(i_data));
	if (!bh)
		return -EIO;

	tmp_idata = (__le32 *)bh->b_data;
	for (i = 0; i < max_entries; i++) {
		if (tmp_idata[i]) {
			retval = free_dind_blocks(handle,
					inode, tmp_idata[i]);
			if (retval) {
				put_bh(bh);
				return retval;
			}
		}
	}
	put_bh(bh);
	extend_credit_for_blkdel(handle, inode);
	ext4_free_blocks(handle, inode, le32_to_cpu(i_data), 1, 1);
	return 0;
}

static int free_ind_block(handle_t *handle, struct inode *inode, __le32 *i_data)
{
	int retval;

	/* ei->i_data[EXT4_IND_BLOCK] */
	if (i_data[0]) {
		extend_credit_for_blkdel(handle, inode);
		ext4_free_blocks(handle, inode,
				le32_to_cpu(i_data[0]), 1, 1);
	}

	/* ei->i_data[EXT4_DIND_BLOCK] */
	if (i_data[1]) {
		retval = free_dind_blocks(handle, inode, i_data[1]);
		if (retval)
			return retval;
	}

	/* ei->i_data[EXT4_TIND_BLOCK] */
	if (i_data[2]) {
		retval = free_tind_blocks(handle, inode, i_data[2]);
		if (retval)
			return retval;
	}
	return 0;
}

static int ext4_ext_swap_inode_data(handle_t *handle, struct inode *inode,
						struct inode *tmp_inode)
{
	int retval;
	__le32	i_data[3];
	struct ext4_inode_info *ei = EXT4_I(inode);
	struct ext4_inode_info *tmp_ei = EXT4_I(tmp_inode);

	/*
	 * One credit accounted for writing the
	 * i_data field of the original inode
	 */
	retval = ext4_journal_extend(handle, 1);
	if (retval) {
		retval = ext4_journal_restart(handle, 1);
		if (retval)
			goto err_out;
	}

	i_data[0] = ei->i_data[EXT4_IND_BLOCK];
	i_data[1] = ei->i_data[EXT4_DIND_BLOCK];
	i_data[2] = ei->i_data[EXT4_TIND_BLOCK];

	down_write(&EXT4_I(inode)->i_data_sem);
	/*
	 * if EXT4_STATE_EXT_MIGRATE is cleared a block allocation
	 * happened after we started the migrate. We need to
	 * fail the migrate
	 */
	if (!(EXT4_I(inode)->i_state & EXT4_STATE_EXT_MIGRATE)) {
		retval = -EAGAIN;
		up_write(&EXT4_I(inode)->i_data_sem);
		goto err_out;
	} else
		EXT4_I(inode)->i_state &= ~EXT4_STATE_EXT_MIGRATE;
	/*
	 * We have the extent map build with the tmp inode.
	 * Now copy the i_data across
	 */
	ei->i_flags |= EXT4_EXTENTS_FL;
	memcpy(ei->i_data, tmp_ei->i_data, sizeof(ei->i_data));

	/*
	 * Update i_blocks with the new blocks that got
	 * allocated while adding extents for extent index
	 * blocks.
	 *
	 * While converting to extents we need not
	 * update the orignal inode i_blocks for extent blocks
	 * via quota APIs. The quota update happened via tmp_inode already.
	 */
	spin_lock(&inode->i_lock);
	inode->i_blocks += tmp_inode->i_blocks;
	spin_unlock(&inode->i_lock);
	up_write(&EXT4_I(inode)->i_data_sem);

	/*
	 * We mark the inode dirty after, because we decrement the
	 * i_blocks when freeing the indirect meta-data blocks
	 */
	retval = free_ind_block(handle, inode, i_data);
	ext4_mark_inode_dirty(handle, inode);

err_out:
	return retval;
}

static int free_ext_idx(handle_t *handle, struct inode *inode,
					struct ext4_extent_idx *ix)
{
	int i, retval = 0;
	ext4_fsblk_t block;
	struct buffer_head *bh;
	struct ext4_extent_header *eh;

	block = idx_pblock(ix);
	bh = sb_bread(inode->i_sb, block);
	if (!bh)
		return -EIO;

	eh = (struct ext4_extent_header *)bh->b_data;
	if (eh->eh_depth != 0) {
		ix = EXT_FIRST_INDEX(eh);
		for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ix++) {
			retval = free_ext_idx(handle, inode, ix);
			if (retval)
				break;
		}
	}
	put_bh(bh);
	extend_credit_for_blkdel(handle, inode);
	ext4_free_blocks(handle, inode, block, 1, 1);
	return retval;
}

/*
 * Free the extent meta data blocks only
 */
static int free_ext_block(handle_t *handle, struct inode *inode)
{
	int i, retval = 0;
	struct ext4_inode_info *ei = EXT4_I(inode);
	struct ext4_extent_header *eh = (struct ext4_extent_header *)ei->i_data;
	struct ext4_extent_idx *ix;
	if (eh->eh_depth == 0)
		/*
		 * No extra blocks allocated for extent meta data
		 */
		return 0;
	ix = EXT_FIRST_INDEX(eh);
	for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ix++) {
		retval = free_ext_idx(handle, inode, ix);
		if (retval)
			return retval;
	}
	return retval;

}

int ext4_ext_migrate(struct inode *inode)
{
	handle_t *handle;
	int retval = 0, i;
	__le32 *i_data;
	ext4_lblk_t blk_count = 0;
	struct ext4_inode_info *ei;
	struct inode *tmp_inode = NULL;
	struct list_blocks_struct lb;
	unsigned long max_entries;
	__u32 goal;

	/*
	 * If the filesystem does not support extents, or the inode
	 * already is extent-based, error out.
	 */
	if (!EXT4_HAS_INCOMPAT_FEATURE(inode->i_sb,
				       EXT4_FEATURE_INCOMPAT_EXTENTS) ||
	    (EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL))
		return -EINVAL;

	if (S_ISLNK(inode->i_mode) && inode->i_blocks == 0)
		/*
		 * don't migrate fast symlink
		 */
		return retval;

	handle = ext4_journal_start(inode,
					EXT4_DATA_TRANS_BLOCKS(inode->i_sb) +
					EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 +
					2 * EXT4_QUOTA_INIT_BLOCKS(inode->i_sb)
					+ 1);
	if (IS_ERR(handle)) {
		retval = PTR_ERR(handle);
		return retval;
	}
	goal = (((inode->i_ino - 1) / EXT4_INODES_PER_GROUP(inode->i_sb)) *
		EXT4_INODES_PER_GROUP(inode->i_sb)) + 1;
	tmp_inode = ext4_new_inode(handle, inode->i_sb->s_root->d_inode,
				   S_IFREG, 0, goal);
	if (IS_ERR(tmp_inode)) {
		retval = -ENOMEM;
		ext4_journal_stop(handle);
		return retval;
	}
	i_size_write(tmp_inode, i_size_read(inode));
	/*
	 * We don't want the inode to be reclaimed
	 * if we got interrupted in between. We have
	 * this tmp inode carrying reference to the
	 * data blocks of the original file. We set
	 * the i_nlink to zero at the last stage after
	 * switching the original file to extent format
	 */
	tmp_inode->i_nlink = 1;

	ext4_ext_tree_init(handle, tmp_inode);
	ext4_orphan_add(handle, tmp_inode);
	ext4_journal_stop(handle);

	/*
	 * start with one credit accounted for
	 * superblock modification.
	 *
	 * For the tmp_inode we already have commited the
	 * trascation that created the inode. Later as and
	 * when we add extents we extent the journal
	 */
	/*
	 * Even though we take i_mutex we can still cause block
	 * allocation via mmap write to holes. If we have allocated
	 * new blocks we fail migrate.  New block allocation will
	 * clear EXT4_STATE_EXT_MIGRATE flag.  The flag is updated
	 * with i_data_sem held to prevent racing with block
	 * allocation.
	 */
	down_read((&EXT4_I(inode)->i_data_sem));
	EXT4_I(inode)->i_state |= EXT4_STATE_EXT_MIGRATE;
	up_read((&EXT4_I(inode)->i_data_sem));

	handle = ext4_journal_start(inode, 1);

	ei = EXT4_I(inode);
	i_data = ei->i_data;
	memset(&lb, 0, sizeof(lb));

	/* 32 bit block address 4 bytes */
	max_entries = inode->i_sb->s_blocksize >> 2;
	for (i = 0; i < EXT4_NDIR_BLOCKS; i++, blk_count++) {
		if (i_data[i]) {
			retval = update_extent_range(handle, tmp_inode,
						le32_to_cpu(i_data[i]),
						blk_count, &lb);
			if (retval)
				goto err_out;
		}
	}
	if (i_data[EXT4_IND_BLOCK]) {
		retval = update_ind_extent_range(handle, tmp_inode,
					le32_to_cpu(i_data[EXT4_IND_BLOCK]),
					&blk_count, &lb);
			if (retval)
				goto err_out;
	} else
		blk_count +=  max_entries;
	if (i_data[EXT4_DIND_BLOCK]) {
		retval = update_dind_extent_range(handle, tmp_inode,
					le32_to_cpu(i_data[EXT4_DIND_BLOCK]),
					&blk_count, &lb);
			if (retval)
				goto err_out;
	} else
		blk_count += max_entries * max_entries;
	if (i_data[EXT4_TIND_BLOCK]) {
		retval = update_tind_extent_range(handle, tmp_inode,
					le32_to_cpu(i_data[EXT4_TIND_BLOCK]),
					&blk_count, &lb);
			if (retval)
				goto err_out;
	}
	/*
	 * Build the last extent
	 */
	retval = finish_range(handle, tmp_inode, &lb);
err_out:
	if (retval)
		/*
		 * Failure case delete the extent information with the
		 * tmp_inode
		 */
		free_ext_block(handle, tmp_inode);
	else {
		retval = ext4_ext_swap_inode_data(handle, inode, tmp_inode);
		if (retval)
			/*
			 * if we fail to swap inode data free the extent
			 * details of the tmp inode
			 */
			free_ext_block(handle, tmp_inode);
	}

	/* We mark the tmp_inode dirty via ext4_ext_tree_init. */
	if (ext4_journal_extend(handle, 1) != 0)
		ext4_journal_restart(handle, 1);

	/*
	 * Mark the tmp_inode as of size zero
	 */
	i_size_write(tmp_inode, 0);

	/*
	 * set the  i_blocks count to zero
	 * so that the ext4_delete_inode does the
	 * right job
	 *
	 * We don't need to take the i_lock because
	 * the inode is not visible to user space.
	 */
	tmp_inode->i_blocks = 0;

	/* Reset the extent details */
	ext4_ext_tree_init(handle, tmp_inode);

	/*
	 * Set the i_nlink to zero so that
	 * generic_drop_inode really deletes the
	 * inode
	 */
	tmp_inode->i_nlink = 0;

	ext4_journal_stop(handle);
	unlock_new_inode(tmp_inode);
	iput(tmp_inode);

	return retval;
}
Commit	Line	Data
c14c6fd5 AK	1	/*
	2	* Copyright IBM Corporation, 2007
	3	* Author Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
	4	*
	5	* This program is free software; you can redistribute it and/or modify it
	6	* under the terms of version 2.1 of the GNU Lesser General Public License
	7	* as published by the Free Software Foundation.
	8	*
	9	* This program is distributed in the hope that it would be useful, but
	10	* WITHOUT ANY WARRANTY; without even the implied warranty of
	11	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
	12	*
	13	*/
	14
	15	#include <linux/module.h>
3dcf5451 CH	16	#include "ext4_jbd2.h"
3dcf5451 CH	17	#include "ext4_extents.h"
c14c6fd5 AK	18
	19	/*
	20	* The contiguous blocks details which can be
	21	* represented by a single extent
	22	*/
	23	struct list_blocks_struct {
	24	ext4_lblk_t first_block, last_block;
	25	ext4_fsblk_t first_pblock, last_pblock;
	26	};
	27
	28	static int finish_range(handle_t handle, struct inode inode,
	29	struct list_blocks_struct *lb)
	30
	31	{
	32	int retval = 0, needed;
	33	struct ext4_extent newext;
	34	struct ext4_ext_path *path;
	35	if (lb->first_pblock == 0)
	36	return 0;
	37
	38	/* Add the extent to temp inode*/
	39	newext.ee_block = cpu_to_le32(lb->first_block);
	40	newext.ee_len = cpu_to_le16(lb->last_block - lb->first_block + 1);
	41	ext4_ext_store_pblock(&newext, lb->first_pblock);
	42	path = ext4_ext_find_extent(inode, lb->first_block, NULL);
	43
	44	if (IS_ERR(path)) {
	45	retval = PTR_ERR(path);
b35905c1	46	path = NULL;
c14c6fd5 AK	47	goto err_out;
	48	}
	49
	50	/*
	51	* Calculate the credit needed to inserting this extent
	52	* Since we are doing this in loop we may accumalate extra
	53	* credit. But below we try to not accumalate too much
	54	* of them by restarting the journal.
	55	*/
ee12b630 MC	56	needed = ext4_ext_calc_credits_for_single_extent(inode,
ee12b630 MC	57	lb->last_block - lb->first_block + 1, path);
c14c6fd5 AK	58
	59	/*
	60	* Make sure the credit we accumalated is not really high
	61	*/
0390131b FM	62	if (needed && ext4_handle_has_enough_credits(handle,
0390131b FM	63	EXT4_RESERVE_TRANS_BLOCKS)) {
c14c6fd5 AK	64	retval = ext4_journal_restart(handle, needed);
	65	if (retval)
	66	goto err_out;
8009f9fb	67	} else if (needed) {
c14c6fd5	68	retval = ext4_journal_extend(handle, needed);
8009f9fb	69	if (retval) {
c14c6fd5 AK	70	/*
	71	* IF not able to extend the journal restart the journal
	72	*/
	73	retval = ext4_journal_restart(handle, needed);
	74	if (retval)
	75	goto err_out;
	76	}
	77	}
0031462b	78	retval = ext4_ext_insert_extent(handle, inode, path, &newext, 0);
c14c6fd5	79	err_out:
b35905c1 AK	80	if (path) {
	81	ext4_ext_drop_refs(path);
	82	kfree(path);
	83	}
c14c6fd5 AK	84	lb->first_pblock = 0;
	85	return retval;
	86	}
	87
	88	static int update_extent_range(handle_t handle, struct inode inode,
	89	ext4_fsblk_t pblock, ext4_lblk_t blk_num,
	90	struct list_blocks_struct *lb)
	91	{
	92	int retval;
	93	/*
	94	* See if we can add on to the existing range (if it exists)
	95	*/
	96	if (lb->first_pblock &&
	97	(lb->last_pblock+1 == pblock) &&
	98	(lb->last_block+1 == blk_num)) {
	99	lb->last_pblock = pblock;
	100	lb->last_block = blk_num;
	101	return 0;
	102	}
	103	/*
	104	* Start a new range.
	105	*/
	106	retval = finish_range(handle, inode, lb);
	107	lb->first_pblock = lb->last_pblock = pblock;
	108	lb->first_block = lb->last_block = blk_num;
	109
	110	return retval;
	111	}
	112
	113	static int update_ind_extent_range(handle_t handle, struct inode inode,
	114	ext4_fsblk_t pblock, ext4_lblk_t *blk_nump,
	115	struct list_blocks_struct *lb)
	116	{
	117	struct buffer_head *bh;
	118	__le32 *i_data;
	119	int i, retval = 0;
	120	ext4_lblk_t blk_count = *blk_nump;
	121	unsigned long max_entries = inode->i_sb->s_blocksize >> 2;
	122
	123	if (!pblock) {
	124	/* Only update the file block number */
	125	*blk_nump += max_entries;
	126	return 0;
	127	}
	128
	129	bh = sb_bread(inode->i_sb, pblock);
	130	if (!bh)
	131	return -EIO;
	132
	133	i_data = (__le32 *)bh->b_data;
	134	for (i = 0; i < max_entries; i++, blk_count++) {
	135	if (i_data[i]) {
	136	retval = update_extent_range(handle, inode,
	137	le32_to_cpu(i_data[i]),
	138	blk_count, lb);
	139	if (retval)
	140	break;
	141	}
	142	}
	143
	144	/* Update the file block number */
	145	*blk_nump = blk_count;
	146	put_bh(bh);
	147	return retval;
148
149	}
150
151	static int update_dind_extent_range(handle_t handle, struct inode inode,
152	ext4_fsblk_t pblock, ext4_lblk_t *blk_nump,
153	struct list_blocks_struct *lb)
154	{
155	struct buffer_head *bh;
156	__le32 *i_data;
157	int i, retval = 0;
158	ext4_lblk_t blk_count = *blk_nump;
159	unsigned long max_entries = inode->i_sb->s_blocksize >> 2;
160
161	if (!pblock) {
162	/* Only update the file block number */
163	blk_nump += max_entries max_entries;
164	return 0;
165	}
166	bh = sb_bread(inode->i_sb, pblock);
167	if (!bh)
168	return -EIO;
169
170	i_data = (__le32 *)bh->b_data;
171	for (i = 0; i < max_entries; i++) {
172	if (i_data[i]) {
173	retval = update_ind_extent_range(handle, inode,
174	le32_to_cpu(i_data[i]),
175	&blk_count, lb);
176	if (retval)
177	break;
178	} else {
179	/* Only update the file block number */
180	blk_count += max_entries;
181	}
182	}
183
184	/* Update the file block number */
185	*blk_nump = blk_count;
186	put_bh(bh);
187	return retval;
188
189	}
190
191	static int update_tind_extent_range(handle_t handle, struct inode inode,
192	ext4_fsblk_t pblock, ext4_lblk_t *blk_nump,
193	struct list_blocks_struct *lb)
194	{
195	struct buffer_head *bh;
196	__le32 *i_data;
197	int i, retval = 0;
198	ext4_lblk_t blk_count = *blk_nump;
199	unsigned long max_entries = inode->i_sb->s_blocksize >> 2;
200
201	if (!pblock) {
202	/* Only update the file block number */
203	blk_nump += max_entries max_entries * max_entries;
204	return 0;
205	}
206	bh = sb_bread(inode->i_sb, pblock);
207	if (!bh)
208	return -EIO;
209
210	i_data = (__le32 *)bh->b_data;
211	for (i = 0; i < max_entries; i++) {
212	if (i_data[i]) {
213	retval = update_dind_extent_range(handle, inode,
214	le32_to_cpu(i_data[i]),
215	&blk_count, lb);
216	if (retval)
217	break;
218	} else
219	/* Only update the file block number */
220	blk_count += max_entries * max_entries;
221	}
222	/* Update the file block number */
223	*blk_nump = blk_count;
224	put_bh(bh);
225	return retval;
226
227	}
228
8009f9fb AK	229	static int extend_credit_for_blkdel(handle_t handle, struct inode inode)
	230	{
	231	int retval = 0, needed;
	232
0390131b	233	if (ext4_handle_has_enough_credits(handle, EXT4_RESERVE_TRANS_BLOCKS+1))
8009f9fb AK	234	return 0;
	235	/*
	236	* We are freeing a blocks. During this we touch
	237	* superblock, group descriptor and block bitmap.
	238	* So allocate a credit of 3. We may update
	239	* quota (user and group).
	240	*/
	241	needed = 3 + 2*EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb);
	242
	243	if (ext4_journal_extend(handle, needed) != 0)
	244	retval = ext4_journal_restart(handle, needed);
	245
	246	return retval;
	247	}
	248
c14c6fd5 AK	249	static int free_dind_blocks(handle_t *handle,
	250	struct inode *inode, __le32 i_data)
	251	{
	252	int i;
	253	__le32 *tmp_idata;
	254	struct buffer_head *bh;
	255	unsigned long max_entries = inode->i_sb->s_blocksize >> 2;
	256
	257	bh = sb_bread(inode->i_sb, le32_to_cpu(i_data));
	258	if (!bh)
	259	return -EIO;
	260
	261	tmp_idata = (__le32 *)bh->b_data;
	262	for (i = 0; i < max_entries; i++) {
8009f9fb AK	263	if (tmp_idata[i]) {
8009f9fb AK	264	extend_credit_for_blkdel(handle, inode);
c14c6fd5	265	ext4_free_blocks(handle, inode,
c9de560d	266	le32_to_cpu(tmp_idata[i]), 1, 1);
8009f9fb	267	}
c14c6fd5 AK	268	}
c14c6fd5 AK	269	put_bh(bh);
8009f9fb	270	extend_credit_for_blkdel(handle, inode);
c9de560d	271	ext4_free_blocks(handle, inode, le32_to_cpu(i_data), 1, 1);
c14c6fd5 AK	272	return 0;
	273	}
	274
	275	static int free_tind_blocks(handle_t *handle,
	276	struct inode *inode, __le32 i_data)
	277	{
	278	int i, retval = 0;
	279	__le32 *tmp_idata;
	280	struct buffer_head *bh;
	281	unsigned long max_entries = inode->i_sb->s_blocksize >> 2;
	282
	283	bh = sb_bread(inode->i_sb, le32_to_cpu(i_data));
	284	if (!bh)
	285	return -EIO;
	286
	287	tmp_idata = (__le32 *)bh->b_data;
	288	for (i = 0; i < max_entries; i++) {
	289	if (tmp_idata[i]) {
	290	retval = free_dind_blocks(handle,
	291	inode, tmp_idata[i]);
	292	if (retval) {
	293	put_bh(bh);
	294	return retval;
	295	}
	296	}
	297	}
	298	put_bh(bh);
8009f9fb	299	extend_credit_for_blkdel(handle, inode);
c9de560d	300	ext4_free_blocks(handle, inode, le32_to_cpu(i_data), 1, 1);
c14c6fd5 AK	301	return 0;
	302	}
	303
8009f9fb	304	static int free_ind_block(handle_t handle, struct inode inode, __le32 *i_data)
c14c6fd5 AK	305	{
c14c6fd5 AK	306	int retval;
c14c6fd5	307
8009f9fb AK	308	/* ei->i_data[EXT4_IND_BLOCK] */
	309	if (i_data[0]) {
	310	extend_credit_for_blkdel(handle, inode);
c14c6fd5	311	ext4_free_blocks(handle, inode,
8009f9fb AK	312	le32_to_cpu(i_data[0]), 1, 1);
8009f9fb AK	313	}
c14c6fd5	314
8009f9fb AK	315	/* ei->i_data[EXT4_DIND_BLOCK] */
	316	if (i_data[1]) {
	317	retval = free_dind_blocks(handle, inode, i_data[1]);
c14c6fd5 AK	318	if (retval)
	319	return retval;
	320	}
	321
8009f9fb AK	322	/* ei->i_data[EXT4_TIND_BLOCK] */
	323	if (i_data[2]) {
	324	retval = free_tind_blocks(handle, inode, i_data[2]);
c14c6fd5 AK	325	if (retval)
	326	return retval;
	327	}
	328	return 0;
	329	}
	330
	331	static int ext4_ext_swap_inode_data(handle_t handle, struct inode inode,
267e4db9	332	struct inode *tmp_inode)
c14c6fd5	333	{
8009f9fb AK	334	int retval;
8009f9fb AK	335	__le32 i_data[3];
c14c6fd5 AK	336	struct ext4_inode_info *ei = EXT4_I(inode);
	337	struct ext4_inode_info *tmp_ei = EXT4_I(tmp_inode);
	338
c14c6fd5 AK	339	/*
	340	* One credit accounted for writing the
	341	* i_data field of the original inode
	342	*/
	343	retval = ext4_journal_extend(handle, 1);
267e4db9	344	if (retval) {
c14c6fd5 AK	345	retval = ext4_journal_restart(handle, 1);
	346	if (retval)
	347	goto err_out;
	348	}
	349
8009f9fb AK	350	i_data[0] = ei->i_data[EXT4_IND_BLOCK];
	351	i_data[1] = ei->i_data[EXT4_DIND_BLOCK];
	352	i_data[2] = ei->i_data[EXT4_TIND_BLOCK];
	353
	354	down_write(&EXT4_I(inode)->i_data_sem);
267e4db9	355	/*
1b9c12f4	356	* if EXT4_STATE_EXT_MIGRATE is cleared a block allocation
267e4db9 AK	357	* happened after we started the migrate. We need to
	358	* fail the migrate
	359	*/
1b9c12f4	360	if (!(EXT4_I(inode)->i_state & EXT4_STATE_EXT_MIGRATE)) {
267e4db9 AK	361	retval = -EAGAIN;
	362	up_write(&EXT4_I(inode)->i_data_sem);
	363	goto err_out;
	364	} else
1b9c12f4	365	EXT4_I(inode)->i_state &= ~EXT4_STATE_EXT_MIGRATE;
c14c6fd5 AK	366	/*
	367	* We have the extent map build with the tmp inode.
	368	* Now copy the i_data across
	369	*/
	370	ei->i_flags \|= EXT4_EXTENTS_FL;
	371	memcpy(ei->i_data, tmp_ei->i_data, sizeof(ei->i_data));
	372
	373	/*
	374	* Update i_blocks with the new blocks that got
	375	* allocated while adding extents for extent index
	376	* blocks.
	377	*
	378	* While converting to extents we need not
	379	* update the orignal inode i_blocks for extent blocks
	380	* via quota APIs. The quota update happened via tmp_inode already.
	381	*/
	382	spin_lock(&inode->i_lock);
	383	inode->i_blocks += tmp_inode->i_blocks;
	384	spin_unlock(&inode->i_lock);
8009f9fb	385	up_write(&EXT4_I(inode)->i_data_sem);
c14c6fd5	386
8009f9fb AK	387	/*
	388	* We mark the inode dirty after, because we decrement the
	389	* i_blocks when freeing the indirect meta-data blocks
	390	*/
	391	retval = free_ind_block(handle, inode, i_data);
c14c6fd5	392	ext4_mark_inode_dirty(handle, inode);
8009f9fb	393
c14c6fd5 AK	394	err_out:
	395	return retval;
	396	}
	397
	398	static int free_ext_idx(handle_t handle, struct inode inode,
	399	struct ext4_extent_idx *ix)
	400	{
	401	int i, retval = 0;
	402	ext4_fsblk_t block;
	403	struct buffer_head *bh;
	404	struct ext4_extent_header *eh;
	405
	406	block = idx_pblock(ix);
	407	bh = sb_bread(inode->i_sb, block);
	408	if (!bh)
	409	return -EIO;
	410
	411	eh = (struct ext4_extent_header *)bh->b_data;
	412	if (eh->eh_depth != 0) {
	413	ix = EXT_FIRST_INDEX(eh);
	414	for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ix++) {
	415	retval = free_ext_idx(handle, inode, ix);
	416	if (retval)
	417	break;
	418	}
	419	}
	420	put_bh(bh);
8009f9fb	421	extend_credit_for_blkdel(handle, inode);
c9de560d	422	ext4_free_blocks(handle, inode, block, 1, 1);
c14c6fd5 AK	423	return retval;
	424	}
	425
	426	/*
	427	* Free the extent meta data blocks only
	428	*/
	429	static int free_ext_block(handle_t handle, struct inode inode)
	430	{
	431	int i, retval = 0;
	432	struct ext4_inode_info *ei = EXT4_I(inode);
	433	struct ext4_extent_header eh = (struct ext4_extent_header )ei->i_data;
	434	struct ext4_extent_idx *ix;
	435	if (eh->eh_depth == 0)
	436	/*
	437	* No extra blocks allocated for extent meta data
	438	*/
	439	return 0;
	440	ix = EXT_FIRST_INDEX(eh);
	441	for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ix++) {
	442	retval = free_ext_idx(handle, inode, ix);
	443	if (retval)
	444	return retval;
	445	}
	446	return retval;
	447
	448	}
	449
2a43a878	450	int ext4_ext_migrate(struct inode *inode)
c14c6fd5 AK	451	{
	452	handle_t *handle;
	453	int retval = 0, i;
	454	__le32 *i_data;
	455	ext4_lblk_t blk_count = 0;
	456	struct ext4_inode_info *ei;
	457	struct inode *tmp_inode = NULL;
	458	struct list_blocks_struct lb;
	459	unsigned long max_entries;
11013911	460	__u32 goal;
c14c6fd5	461
83982b6f TT	462	/*
	463	* If the filesystem does not support extents, or the inode
	464	* already is extent-based, error out.
	465	*/
	466	if (!EXT4_HAS_INCOMPAT_FEATURE(inode->i_sb,
	467	EXT4_FEATURE_INCOMPAT_EXTENTS) \|\|
	468	(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL))
c14c6fd5 AK	469	return -EINVAL;
c14c6fd5 AK	470
b8356c46 VC	471	if (S_ISLNK(inode->i_mode) && inode->i_blocks == 0)
	472	/*
	473	* don't migrate fast symlink
	474	*/
	475	return retval;
	476
c14c6fd5 AK	477	handle = ext4_journal_start(inode,
	478	EXT4_DATA_TRANS_BLOCKS(inode->i_sb) +
	479	EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 +
	480	2 * EXT4_QUOTA_INIT_BLOCKS(inode->i_sb)
	481	+ 1);
	482	if (IS_ERR(handle)) {
	483	retval = PTR_ERR(handle);
09054264	484	return retval;
c14c6fd5	485	}
11013911 AD	486	goal = (((inode->i_ino - 1) / EXT4_INODES_PER_GROUP(inode->i_sb)) *
11013911 AD	487	EXT4_INODES_PER_GROUP(inode->i_sb)) + 1;
f157a4aa	488	tmp_inode = ext4_new_inode(handle, inode->i_sb->s_root->d_inode,
11013911	489	S_IFREG, 0, goal);
c14c6fd5 AK	490	if (IS_ERR(tmp_inode)) {
	491	retval = -ENOMEM;
	492	ext4_journal_stop(handle);
09054264	493	return retval;
c14c6fd5 AK	494	}
	495	i_size_write(tmp_inode, i_size_read(inode));
	496	/*
	497	* We don't want the inode to be reclaimed
	498	* if we got interrupted in between. We have
	499	* this tmp inode carrying reference to the
	500	* data blocks of the original file. We set
	501	* the i_nlink to zero at the last stage after
	502	* switching the original file to extent format
	503	*/
	504	tmp_inode->i_nlink = 1;
	505
	506	ext4_ext_tree_init(handle, tmp_inode);
	507	ext4_orphan_add(handle, tmp_inode);
	508	ext4_journal_stop(handle);
	509
c14c6fd5 AK	510	/*
	511	* start with one credit accounted for
	512	* superblock modification.
	513	*
	514	* For the tmp_inode we already have commited the
	515	* trascation that created the inode. Later as and
	516	* when we add extents we extent the journal
	517	*/
267e4db9	518	/*
1b9c12f4 TT	519	* Even though we take i_mutex we can still cause block
	520	* allocation via mmap write to holes. If we have allocated
	521	* new blocks we fail migrate. New block allocation will
	522	* clear EXT4_STATE_EXT_MIGRATE flag. The flag is updated
	523	* with i_data_sem held to prevent racing with block
	524	* allocation.
267e4db9 AK	525	*/
267e4db9 AK	526	down_read((&EXT4_I(inode)->i_data_sem));
1b9c12f4	527	EXT4_I(inode)->i_state \|= EXT4_STATE_EXT_MIGRATE;
267e4db9 AK	528	up_read((&EXT4_I(inode)->i_data_sem));
267e4db9 AK	529
c14c6fd5	530	handle = ext4_journal_start(inode, 1);
8009f9fb AK	531
	532	ei = EXT4_I(inode);
	533	i_data = ei->i_data;
	534	memset(&lb, 0, sizeof(lb));
	535
	536	/* 32 bit block address 4 bytes */
	537	max_entries = inode->i_sb->s_blocksize >> 2;
c14c6fd5 AK	538	for (i = 0; i < EXT4_NDIR_BLOCKS; i++, blk_count++) {
	539	if (i_data[i]) {
	540	retval = update_extent_range(handle, tmp_inode,
	541	le32_to_cpu(i_data[i]),
	542	blk_count, &lb);
	543	if (retval)
	544	goto err_out;
	545	}
	546	}
	547	if (i_data[EXT4_IND_BLOCK]) {
	548	retval = update_ind_extent_range(handle, tmp_inode,
	549	le32_to_cpu(i_data[EXT4_IND_BLOCK]),
	550	&blk_count, &lb);
	551	if (retval)
	552	goto err_out;
	553	} else
	554	blk_count += max_entries;
	555	if (i_data[EXT4_DIND_BLOCK]) {
	556	retval = update_dind_extent_range(handle, tmp_inode,
	557	le32_to_cpu(i_data[EXT4_DIND_BLOCK]),
	558	&blk_count, &lb);
	559	if (retval)
	560	goto err_out;
	561	} else
	562	blk_count += max_entries * max_entries;
	563	if (i_data[EXT4_TIND_BLOCK]) {
	564	retval = update_tind_extent_range(handle, tmp_inode,
	565	le32_to_cpu(i_data[EXT4_TIND_BLOCK]),
	566	&blk_count, &lb);
	567	if (retval)
	568	goto err_out;
	569	}
	570	/*
	571	* Build the last extent
	572	*/
	573	retval = finish_range(handle, tmp_inode, &lb);
	574	err_out:
c14c6fd5 AK	575	if (retval)
	576	/*
	577	* Failure case delete the extent information with the
	578	* tmp_inode
	579	*/
	580	free_ext_block(handle, tmp_inode);
267e4db9 AK	581	else {
	582	retval = ext4_ext_swap_inode_data(handle, inode, tmp_inode);
	583	if (retval)
	584	/*
	585	* if we fail to swap inode data free the extent
	586	* details of the tmp inode
	587	*/
	588	free_ext_block(handle, tmp_inode);
	589	}
8009f9fb AK	590
	591	/* We mark the tmp_inode dirty via ext4_ext_tree_init. */
	592	if (ext4_journal_extend(handle, 1) != 0)
	593	ext4_journal_restart(handle, 1);
c14c6fd5 AK	594
	595	/*
	596	* Mark the tmp_inode as of size zero
	597	*/
	598	i_size_write(tmp_inode, 0);
	599
	600	/*
	601	* set the i_blocks count to zero
	602	* so that the ext4_delete_inode does the
	603	* right job
	604	*
	605	* We don't need to take the i_lock because
	606	* the inode is not visible to user space.
	607	*/
	608	tmp_inode->i_blocks = 0;
	609
	610	/* Reset the extent details */
	611	ext4_ext_tree_init(handle, tmp_inode);
	612
	613	/*
	614	* Set the i_nlink to zero so that
	615	* generic_drop_inode really deletes the
	616	* inode
	617	*/
	618	tmp_inode->i_nlink = 0;
	619
	620	ext4_journal_stop(handle);
a8526e84	621	unlock_new_inode(tmp_inode);
09054264	622	iput(tmp_inode);
c14c6fd5 AK	623
	624	return retval;
	625	}