path: root/fs/ext4/extents.c

// SPDX-License-Identifier: GPL-2.0
/*
 * Copyright (c) 2003-2006, Cluster File Systems, Inc, info@clusterfs.com
 * Written by Alex Tomas <alex@clusterfs.com>
 *
 * Architecture independence:
 *   Copyright (c) 2005, Bull S.A.
 *   Written by Pierre Peiffer <pierre.peiffer@bull.net>
 */

/*
 * Extents support for EXT4
 *
 * TODO:
 *   - ext4*_error() should be used in some situations
 *   - analyze all BUG()/BUG_ON(), use -EIO where appropriate
 *   - smart tree reduction
 */

#include <linux/fs.h>
#include <linux/time.h>
#include <linux/jbd2.h>
#include <linux/highuid.h>
#include <linux/pagemap.h>
#include <linux/quotaops.h>
#include <linux/string.h>
#include <linux/slab.h>
#include <linux/uaccess.h>
#include <linux/fiemap.h>
#include <linux/backing-dev.h>
#include <linux/iomap.h>
#include "ext4_jbd2.h"
#include "ext4_extents.h"
#include "xattr.h"

#include <trace/events/ext4.h>

/*
 * used by extent splitting.
 */
#define EXT4_EXT_MAY_ZEROOUT	0x1  /* safe to zeroout if split fails \
					due to ENOSPC */
#define EXT4_EXT_MARK_UNWRIT1	0x2  /* mark first half unwritten */
#define EXT4_EXT_MARK_UNWRIT2	0x4  /* mark second half unwritten */

#define EXT4_EXT_DATA_VALID1	0x8  /* first half contains valid data */
#define EXT4_EXT_DATA_VALID2	0x10 /* second half contains valid data */

static __le32 ext4_extent_block_csum(struct inode *inode,
				     struct ext4_extent_header *eh)
{
	struct ext4_inode_info *ei = EXT4_I(inode);
	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
	__u32 csum;

	csum = ext4_chksum(sbi, ei->i_csum_seed, (__u8 *)eh,
			   EXT4_EXTENT_TAIL_OFFSET(eh));
	return cpu_to_le32(csum);
}

static int ext4_extent_block_csum_verify(struct inode *inode,
					 struct ext4_extent_header *eh)
{
	struct ext4_extent_tail *et;

	if (!ext4_has_metadata_csum(inode->i_sb))
		return 1;

	et = find_ext4_extent_tail(eh);
	if (et->et_checksum != ext4_extent_block_csum(inode, eh))
		return 0;
	return 1;
}

static void ext4_extent_block_csum_set(struct inode *inode,
				       struct ext4_extent_header *eh)
{
	struct ext4_extent_tail *et;

	if (!ext4_has_metadata_csum(inode->i_sb))
		return;

	et = find_ext4_extent_tail(eh);
	et->et_checksum = ext4_extent_block_csum(inode, eh);
}

static int ext4_split_extent_at(handle_t *handle,
			     struct inode *inode,
			     struct ext4_ext_path **ppath,
			     ext4_lblk_t split,
			     int split_flag,
			     int flags);

static int ext4_ext_trunc_restart_fn(struct inode *inode, int *dropped)
{
	/*
	 * Drop i_data_sem to avoid deadlock with ext4_map_blocks.  At this
	 * moment, get_block can be called only for blocks inside i_size since
	 * page cache has been already dropped and writes are blocked by
	 * i_mutex. So we can safely drop the i_data_sem here.
	 */
	BUG_ON(EXT4_JOURNAL(inode) == NULL);
	ext4_discard_preallocations(inode);
	up_write(&EXT4_I(inode)->i_data_sem);
	*dropped = 1;
	return 0;
}

/*
 * Make sure 'handle' has at least 'check_cred' credits. If not, restart
 * transaction with 'restart_cred' credits. The function drops i_data_sem
 * when restarting transaction and gets it after transaction is restarted.
 *
 * The function returns 0 on success, 1 if transaction had to be restarted,
 * and < 0 in case of fatal error.
 */
int ext4_datasem_ensure_credits(handle_t *handle, struct inode *inode,
				int check_cred, int restart_cred,
				int revoke_cred)
{
	int ret;
	int dropped = 0;

	ret = ext4_journal_ensure_credits_fn(handle, check_cred, restart_cred,
		revoke_cred, ext4_ext_trunc_restart_fn(inode, &dropped));
	if (dropped)
		down_write(&EXT4_I(inode)->i_data_sem);
	return ret;
}

/*
 * could return:
 *  - EROFS
 *  - ENOMEM
 */
static int ext4_ext_get_access(handle_t *handle, struct inode *inode,
				struct ext4_ext_path *path)
{
	if (path->p_bh) {
		/* path points to block */
		BUFFER_TRACE(path->p_bh, "get_write_access");
		return ext4_journal_get_write_access(handle, path->p_bh);
	}
	/* path points to leaf/index in inode body */
	/* we use in-core data, no need to protect them */
	return 0;
}

/*
 * could return:
 *  - EROFS
 *  - ENOMEM
 *  - EIO
 */
static int __ext4_ext_dirty(const char *where, unsigned int line,
			    handle_t *handle, struct inode *inode,
			    struct ext4_ext_path *path)
{
	int err;

	WARN_ON(!rwsem_is_locked(&EXT4_I(inode)->i_data_sem));
	if (path->p_bh) {
		ext4_extent_block_csum_set(inode, ext_block_hdr(path->p_bh));
		/* path points to block */
		err = __ext4_handle_dirty_metadata(where, line, handle,
						   inode, path->p_bh);
	} else {
		/* path points to leaf/index in inode body */
		err = ext4_mark_inode_dirty(handle, inode);
	}
	return err;
}

#define ext4_ext_dirty(handle, inode, path) \
		__ext4_ext_dirty(__func__, __LINE__, (handle), (inode), (path))

static ext4_fsblk_t ext4_ext_find_goal(struct inode *inode,
			      struct ext4_ext_path *path,
			      ext4_lblk_t block)
{
	if (path) {
		int depth = path->p_depth;
		struct ext4_extent *ex;

		/*
		 * Try to predict block placement assuming that we are
		 * filling in a file which will eventually be
		 * non-sparse --- i.e., in the case of libbfd writing
		 * an ELF object sections out-of-order but in a way
		 * the eventually results in a contiguous object or
		 * executable file, or some database extending a table
		 * space file.  However, this is actually somewhat
		 * non-ideal if we are writing a sparse file such as
		 * qemu or KVM writing a raw image file that is going
		 * to stay fairly sparse, since it will end up
		 * fragmenting the file system's free space.  Maybe we
		 * should have some hueristics or some way to allow
		 * userspace to pass a hint to file system,
		 * especially if the latter case turns out to be
		 * common.
		 */
		ex = path[depth].p_ext;
		if (ex) {
			ext4_fsblk_t ext_pblk = ext4_ext_pblock(ex);
			ext4_lblk_t ext_block = le32_to_cpu(ex->ee_block);

			if (block > ext_block)
				return ext_pblk + (block - ext_block);
			else
				return ext_pblk - (ext_block - block);
		}

		/* it looks like index is empty;
		 * try to find starting block from index itself */
		if (path[depth].p_bh)
			return path[depth].p_bh->b_blocknr;
	}

	/* OK. use inode's group */
	return ext4_inode_to_goal_block(inode);
}

/*
 * Allocation for a meta data block
 */
static ext4_fsblk_t
ext4_ext_new_meta_block(handle_t *handle, struct inode *inode,
			struct ext4_ext_path *path,
			struct ext4_extent *ex, int *err, unsigned int flags)
{
	ext4_fsblk_t goal, newblock;

	goal = ext4_ext_find_goal(inode, path, le32_to_cpu(ex->ee_block));
	newblock = ext4_new_meta_blocks(handle, inode, goal, flags,
					NULL, err);
	return newblock;
}

static inline int ext4_ext_space_block(struct inode *inode, int check)
{
	int size;

	size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
			/ sizeof(struct ext4_extent);
#ifdef AGGRESSIVE_TEST
	if (!check && size > 6)
		size = 6;
#endif
	return size;
}

static inline int ext4_ext_space_block_idx(struct inode *inode, int check)
{
	int size;

	size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
			/ sizeof(struct ext4_extent_idx);
#ifdef AGGRESSIVE_TEST
	if (!check && size > 5)
		size = 5;
#endif
	return size;
}

static inline int ext4_ext_space_root(struct inode *inode, int check)
{
	int size;

	size = sizeof(EXT4_I(inode)->i_data);
	size -= sizeof(struct ext4_extent_header);
	size /= sizeof(struct ext4_extent);
#ifdef AGGRESSIVE_TEST
	if (!check && size > 3)
		size = 3;
#endif
	return size;
}

static inline int ext4_ext_space_root_idx(struct inode *inode, int check)
{
	int size;

	size = sizeof(EXT4_I(inode)->i_data);
	size -= sizeof(struct ext4_extent_header);
	size /= sizeof(struct ext4_extent_idx);
#ifdef AGGRESSIVE_TEST
	if (!check && size > 4)
		size = 4;
#endif
	return size;
}

static inline int
ext4_force_split_extent_at(handle_t *handle, struct inode *inode,
			   struct ext4_ext_path **ppath, ext4_lblk_t lblk,
			   int nofail)
{
	struct ext4_ext_path *path = *ppath;
	int unwritten = ext4_ext_is_unwritten(path[path->p_depth].p_ext);
	int flags = EXT4_EX_NOCACHE | EXT4_GET_BLOCKS_PRE_IO;

	if (nofail)
		flags |= EXT4_GET_BLOCKS_METADATA_NOFAIL | EXT4_EX_NOFAIL;

	return ext4_split_extent_at(handle, inode, ppath, lblk, unwritten ?
			EXT4_EXT_MARK_UNWRIT1|EXT4_EXT_MARK_UNWRIT2 : 0,
			flags);
}

static int
ext4_ext_max_entries(struct inode *inode, int depth)
{
	int max;

	if (depth == ext_depth(inode)) {
		if (depth == 0)
			max = ext4_ext_space_root(inode, 1);
		else
			max = ext4_ext_space_root_idx(inode, 1);
	} else {
		if (depth == 0)
			max = ext4_ext_space_block(inode, 1);
		else
			max = ext4_ext_space_block_idx(inode, 1);
	}

	return max;
}

static int ext4_valid_extent(struct inode *inode, struct ext4_extent *ext)
{
	ext4_fsblk_t block = ext4_ext_pblock(ext);
	int len = ext4_ext_get_actual_len(ext);
	ext4_lblk_t lblock = le32_to_cpu(ext->ee_block);

	/*
	 * We allow neither:
	 *  - zero length
	 *  - overflow/wrap-around
	 */
	if (lblock + len <= lblock)
		return 0;
	return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, len);
}

static int ext4_valid_extent_idx(struct inode *inode,
				struct ext4_extent_idx *ext_idx)
{
	ext4_fsblk_t block = ext4_idx_pblock(ext_idx);

	return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, 1);
}

static int ext4_valid_extent_entries(struct inode *inode,
				     struct ext4_extent_header *eh,
				     ext4_fsblk_t *pblk, int depth)
{
	unsigned short entries;
	if (eh->eh_entries == 0)
		return 1;

	entries = le16_to_cpu(eh->eh_entries);

	if (depth == 0) {
		/* leaf entries */
		struct ext4_extent *ext = EXT_FIRST_EXTENT(eh);
		ext4_lblk_t lblock = 0;
		ext4_lblk_t prev = 0;
		int len = 0;
		while (entries) {
			if (!ext4_valid_extent(inode, ext))
				return 0;

			/* Check for overlapping extents */
			lblock = le32_to_cpu(ext->ee_block);
			len = ext4_ext_get_actual_len(ext);
			if ((lblock <= prev) && prev) {
				*pblk = ext4_ext_pblock(ext);
				return 0;
			}
			ext++;
			entries--;
			prev = lblock + len - 1;
		}
	} else {
		struct ext4_extent_idx *ext_idx = EXT_FIRST_INDEX(eh);
		while (entries) {
			if (!ext4_valid_extent_idx(inode, ext_idx))
				return 0;
			ext_idx++;
			entries--;
		}
	}
	return 1;
}

static int __ext4_ext_check(const char *function, unsigned int line,
			    struct inode *inode, struct ext4_extent_header *eh,
			    int depth, ext4_fsblk_t pblk)
{
	const char *error_msg;
	int max = 0, err = -EFSCORRUPTED;

	if (unlikely(eh->eh_magic != EXT4_EXT_MAGIC)) {
		error_msg = "invalid magic";
		goto corrupted;
	}
	if (unlikely(le16_to_cpu(eh->eh_depth) != depth)) {
		error_msg = "unexpected eh_depth";
		goto corrupted;
	}
	if (unlikely(eh->eh_max == 0)) {
		error_msg = "invalid eh_max";
		goto corrupted;
	}
	max = ext4_ext_max_entries(inode, depth);
	if (unlikely(le16_to_cpu(eh->eh_max) > max)) {
		error_msg = "too large eh_max";
		goto corrupted;
	}
	if (unlikely(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max))) {
		error_msg = "invalid eh_entries";
		goto corrupted;
	}
	if (!ext4_valid_extent_entries(inode, eh, &pblk, depth)) {
		error_msg = "invalid extent entries";
		goto corrupted;
	}
	if (unlikely(depth > 32)) {
		error_msg = "too large eh_depth";
		goto corrupted;
	}
	/* Verify checksum on non-root extent tree nodes */
	if (ext_depth(inode) != depth &&
	    !ext4_extent_block_csum_verify(inode, eh)) {
		error_msg = "extent tree corrupted";
		err = -EFSBADCRC;
		goto corrupted;
	}
	return 0;

corrupted:
	ext4_error_inode_err(inode, function, line, 0, -err,
			     "pblk %llu bad header/extent: %s - magic %x, "
			     "entries %u, max %u(%u), depth %u(%u)",
			     (unsigned long long) pblk, error_msg,
			     le16_to_cpu(eh->eh_magic),
			     le16_to_cpu(eh->eh_entries),
			     le16_to_cpu(eh->eh_max),
			     max, le16_to_cpu(eh->eh_depth), depth);
	return err;
}

#define ext4_ext_check(inode, eh, depth, pblk)			\
	__ext4_ext_check(__func__, __LINE__, (inode), (eh), (depth), (pblk))

int ext4_ext_check_inode(struct inode *inode)
{
	return ext4_ext_check(inode, ext_inode_hdr(inode), ext_depth(inode), 0);
}

static void ext4_cache_extents(struct inode *inode,
			       struct ext4_extent_header *eh)
{
	struct ext4_extent *ex = EXT_FIRST_EXTENT(eh);
	ext4_lblk_t prev = 0;
	int i;

	for (i = le16_to_cpu(eh->eh_entries); i > 0; i--, ex++) {
		unsigned int status = EXTENT_STATUS_WRITTEN;
		ext4_lblk_t lblk = le32_to_cpu(ex->ee_block);
		int len = ext4_ext_get_actual_len(ex);

		if (prev && (prev != lblk))
			ext4_es_cache_extent(inode, prev, lblk - prev, ~0,
					     EXTENT_STATUS_HOLE);

		if (ext4_ext_is_unwritten(ex))
			status = EXTENT_STATUS_UNWRITTEN;
		ext4_es_cache_extent(inode, lblk, len,
				     ext4_ext_pblock(ex), status);
		prev = lblk + len;
	}
}

static struct buffer_head *
__read_extent_tree_block(const char *function, unsigned int line,
			 struct inode *inode, ext4_fsblk_t pblk, int depth,
			 int flags)
{
	struct buffer_head		*bh;
	int				err;
	gfp_t				gfp_flags = __GFP_MOVABLE | GFP_NOFS;

	if (flags & EXT4_EX_NOFAIL)
		gfp_flags |= __GFP_NOFAIL;

	bh = sb_getblk_gfp(inode->i_sb, pblk, gfp_flags);
	if (unlikely(!bh))
		return ERR_PTR(-ENOMEM);

	if (!bh_uptodate_or_lock(bh)) {
		trace_ext4_ext_load_extent(inode, pblk, _RET_IP_);
		err = bh_submit_read(bh);
		if (err < 0)
			goto errout;
	}
	if (buffer_verified(bh) && !(flags & EXT4_EX_FORCE_CACHE))
		return bh;
	if (!ext4_has_feature_journal(inode->i_sb) ||
	    (inode->i_ino !=
	     le32_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_journal_inum))) {
		err = __ext4_ext_check(function, line, inode,
				       ext_block_hdr(bh), depth, pblk);
		if (err)
			goto errout;
	}
	set_buffer_verified(bh);
	/*
	 * If this is a leaf block, cache all of its entries
	 */
	if (!(flags & EXT4_EX_NOCACHE) && depth == 0) {
		struct ext4_extent_header *eh = ext_block_hdr(bh);
		ext4_cache_extents(inode, eh);
	}
	return bh;
errout:
	put_bh(bh);
	return ERR_PTR(err);

}

#define read_extent_tree_block(inode, pblk, depth, flags)		\
	__read_extent_tree_block(__func__, __LINE__, (inode), (pblk),   \
				 (depth), (flags))

/*
 * This function is called to cache a file's extent information in the
 * extent status tree
 */
int ext4_ext_precache(struct inode *inode)
{
	struct ext4_inode_info *ei = EXT4_I(inode);
	struct ext4_ext_path *path = NULL;
	struct buffer_head *bh;
	int i = 0, depth, ret = 0;

	if (!ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
		return 0;	/* not an extent-mapped inode */

	down_read(&ei->i_data_sem);
	depth = ext_depth(inode);

	/* Don't cache anything if there are no external extent blocks */
	if (!depth) {
		up_read(&ei->i_data_sem);
		return ret;
	}

	path = kcalloc(depth + 1, sizeof(struct ext4_ext_path),
		       GFP_NOFS);
	if (path == NULL) {
		up_read(&ei->i_data_sem);
		return -ENOMEM;
	}

	path[0].p_hdr = ext_inode_hdr(inode);
	ret = ext4_ext_check(inode, path[0].p_hdr, depth, 0);
	if (ret)
		goto out;
	path[0].p_idx = EXT_FIRST_INDEX(path[0].p_hdr);
	while (i >= 0) {
		/*
		 * If this is a leaf block or we've reached the end of
		 * the index block, go up
		 */
		if ((i == depth) ||
		    path[i].p_idx > EXT_LAST_INDEX(path[i].p_hdr)) {
			brelse(path[i].p_bh);
			path[i].p_bh = NULL;
			i--;
			continue;
		}
		bh = read_extent_tree_block(inode,
					    ext4_idx_pblock(path[i].p_idx++),
					    depth - i - 1,
					    EXT4_EX_FORCE_CACHE);
		if (IS_ERR(bh)) {
			ret = PTR_ERR(bh);
			break;
		}
		i++;
		path[i].p_bh = bh;
		path[i].p_hdr = ext_block_hdr(bh);
		path[i].p_idx = EXT_FIRST_INDEX(path[i].p_hdr);
	}
	ext4_set_inode_state(inode, EXT4_STATE_EXT_PRECACHED);
out:
	up_read(&ei->i_data_sem);
	ext4_ext_drop_refs(path);
	kfree(path);
	return ret;
}

#ifdef EXT_DEBUG
static void ext4_ext_show_path(struct inode *inode, struct ext4_ext_path *path)
{
	int k, l = path->p_depth;

	ext_debug(inode, "path:");
	for (k = 0; k <= l; k++, path++) {
		if (path->p_idx) {
			ext_debug(inode, "  %d->%llu",
				  le32_to_cpu(path->p_idx->ei_block),
				  ext4_idx_pblock(path->p_idx));
		} else if (path->p_ext) {
			ext_debug(inode, "  %d:[%d]%d:%llu ",
				  le32_to_cpu(path->p_ext->ee_block),
				  ext4_ext_is_unwritten(path->p_ext),
				  ext4_ext_get_actual_len(path->p_ext),
				  ext4_ext_pblock(path->p_ext));
		} else
			ext_debug(inode, "  []");
	}
	ext_debug(inode, "\n");
}

static void ext4_ext_show_leaf(struct inode *inode, struct ext4_ext_path *path)
{
	int depth = ext_depth(inode);
	struct ext4_extent_header *eh;
	struct ext4_extent *ex;
	int i;

	if (!path)
		return;

	eh = path[depth].p_hdr;
	ex = EXT_FIRST_EXTENT(eh);

	ext_debug(inode, "Displaying leaf extents\n");

	for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ex++) {
		ext_debug(inode, "%d:[%d]%d:%llu ", le32_to_cpu(ex->ee_block),
			  ext4_ext_is_unwritten(ex),
			  ext4_ext_get_actual_len(ex), ext4_ext_pblock(ex));
	}
	ext_debug(inode, "\n");
}

static void ext4_ext_show_move(struct inode *inode, struct ext4_ext_path *path,
			ext4_fsblk_t newblock, int level)
{
	int depth = ext_depth(inode);
	struct ext4_extent *ex;

	if (depth != level) {
		struct ext4_extent_idx *idx;
		idx = path[level].p_idx;
		while (idx <= EXT_MAX_INDEX(path[level].p_hdr)) {
			ext_debug(inode, "%d: move %d:%llu in new index %llu\n",
				  level, le32_to_cpu(idx->ei_block),
				  ext4_idx_pblock(idx), newblock);
			idx++;
		}

		return;
	}

	ex = path[depth].p_ext;
	while (ex <= EXT_MAX_EXTENT(path[depth].p_hdr)) {
		ext_debug(inode, "move %d:%llu:[%d]%d in new leaf %llu\n",
				le32_to_cpu(ex->ee_block),
				ext4_ext_pblock(ex),
				ext4_ext_is_unwritten(ex),
				ext4_ext_get_actual_len(ex),
				newblock);
		ex++;
	}
}

#else
#define ext4_ext_show_path(inode, path)
#define ext4_ext_show_leaf(inode, path)
#define ext4_ext_show_move(inode, path, newblock, level)
#endif

void ext4_ext_drop_refs(struct ext4_ext_path *path)
{
	int depth, i;

	if (!path)
		return;
	depth = path->p_depth;
	for (i = 0; i <= depth; i++, path++) {
		if (path->p_bh) {
			brelse(path->p_bh);
			path->p_bh = NULL;
		}
	}
}

/*
 * ext4_ext_binsearch_idx:
 * binary search for the closest index of the given block
 * the header must be checked before calling this
 */
static void
ext4_ext_binsearch_idx(struct inode *inode,
			struct ext4_ext_path *path, ext4_lblk_t block)
{
	struct ext4_extent_header *eh = path->p_hdr;
	struct ext4_extent_idx *r, *l, *m;


	ext_debug(inode, "binsearch for %u(idx):  ", block);

	l = EXT_FIRST_INDEX(eh) + 1;
	r = EXT_LAST_INDEX(eh);
	while (l <= r) {
		m = l + (r - l) / 2;
		if (block < le32_to_cpu(m->ei_block))
			r = m - 1;
		else
			l = m + 1;
		ext_debug(inode, "%p(%u):%p(%u):%p(%u) ", l,
			  le32_to_cpu(l->ei_block), m, le32_to_cpu(m->ei_block),
			  r, le32_to_cpu(r->ei_block));
	}

	path->p_idx = l - 1;
	ext_debug(inode, "  -> %u->%lld ", le32_to_cpu(path->p_idx->ei_block),
		  ext4_idx_pblock(path->p_idx));

#ifdef CHECK_BINSEARCH
	{
		struct ext4_extent_idx *chix, *ix;
		int k;

		chix = ix = EXT_FIRST_INDEX(eh);
		for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ix++) {
			if (k != 0 && le32_to_cpu(ix->ei_block) <=
			    le32_to_cpu(ix[-1].ei_block)) {
				printk(KERN_DEBUG "k=%d, ix=0x%p, "
				       "first=0x%p\n", k,
				       ix, EXT_FIRST_INDEX(eh));
				printk(KERN_DEBUG "%u <= %u\n",
				       le32_to_cpu(ix->ei_block),
				       le32_to_cpu(ix[-1].ei_block));
			}
			BUG_ON(k && le32_to_cpu(ix->ei_block)
					   <= le32_to_cpu(ix[-1].ei_block));
			if (block < le32_to_cpu(ix->ei_block))
				break;
			chix = ix;
		}
		BUG_ON(chix != path->p_idx);
	}
#endif

}

/*
 * ext4_ext_binsearch:
 * binary search for closest extent of the given block
 * the header must be checked before calling this
 */
static void
ext4_ext_binsearch(struct inode *inode,
		struct ext4_ext_path *path, ext4_lblk_t block)
{
	struct ext4_extent_header *eh = path->p_hdr;
	struct ext4_extent *r, *l, *m;

	if (eh->eh_entries == 0) {
		/*
		 * this leaf is empty:
		 * we get such a leaf in split/add case
		 */
		return;
	}

	ext_debug(inode, "binsearch for %u:  ", block);

	l = EXT_FIRST_EXTENT(eh) + 1;
	r = EXT_LAST_EXTENT(eh);

	while (l <= r) {
		m = l + (r - l) / 2;
		if (block < le32_to_cpu(m->ee_block))
			r = m - 1;
		else
			l = m + 1;
		ext_debug(inode, "%p(%u):%p(%u):%p(%u) ", l,
			  le32_to_cpu(l->ee_block), m, le32_to_cpu(m->ee_block),
			  r, le32_to_cpu(r->ee_block));
	}

	path->p_ext = l - 1;
	ext_debug(inode, "  -> %d:%llu:[%d]%d ",
			le32_to_cpu(path->p_ext->ee_block),
			ext4_ext_pblock(path->p_ext),
			ext4_ext_is_unwritten(path->p_ext),
			ext4_ext_get_actual_len(path->p_ext));

#ifdef CHECK_BINSEARCH
	{
		struct ext4_extent *chex, *ex;
		int k;

		chex = ex = EXT_FIRST_EXTENT(eh);
		for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ex++) {
			BUG_ON(k && le32_to_cpu(ex->ee_block)
					  <= le32_to_cpu(ex[-1].ee_block));
			if (block < le32_to_cpu(ex->ee_block))
				break;
			chex = ex;
		}
		BUG_ON(chex != path->p_ext);
	}
#endif

}

void ext4_ext_tree_init(handle_t *handle, struct inode *inode)
{
	struct ext4_extent_header *eh;

	eh = ext_inode_hdr(inode);
	eh->eh_depth = 0;
	eh->eh_entries = 0;
	eh->eh_magic = EXT4_EXT_MAGIC;
	eh->eh_max = cpu_to_le16(ext4_ext_space_root(inode, 0));
	ext4_mark_inode_dirty(handle, inode);
}

struct ext4_ext_path *
ext4_find_extent(struct inode *inode, ext4_lblk_t block,
		 struct ext4_ext_path **orig_path, int flags)
{
	struct ext4_extent_header *eh;
	struct buffer_head *bh;
	struct ext4_ext_path *path = orig_path ? *orig_path : NULL;
	short int depth, i, ppos = 0;
	int ret;
	gfp_t gfp_flags = GFP_NOFS;

	if (flags & EXT4_EX_NOFAIL)
		gfp_flags |= __GFP_NOFAIL;

	eh = ext_inode_hdr(inode);
	depth = ext_depth(inode);
	if (depth < 0 || depth > EXT4_MAX_EXTENT_DEPTH) {
		EXT4_ERROR_INODE(inode, "inode has invalid extent depth: %d",
				 depth);
		ret = -EFSCORRUPTED;
		goto err;
	}

	if (path) {
		ext4_ext_drop_refs(path);
		if (depth > path[0].p_maxdepth) {
			kfree(path);
			*orig_path = path = NULL;
		}
	}
	if (!path) {
		/* account possible depth increase */
		path = kcalloc(depth + 2, sizeof(struct ext4_ext_path),
				gfp_flags);
		if (unlikely(!path))
			return ERR_PTR(-ENOMEM);
		path[0].p_maxdepth = depth + 1;
	}
	path[0].p_hdr = eh;
	path[0].p_bh = NULL;

	i = depth;
	if (!(flags & EXT4_EX_NOCACHE) && depth == 0)
		ext4_cache_extents(inode, eh);
	/* walk through the tree */
	while (i) {
		ext_debug(inode, "depth %d: num %d, max %d\n",
			  ppos, le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max));

		ext4_ext_binsearch_idx(inode, path + ppos, block);
		path[ppos].p_block = ext4_idx_pblock(path[ppos].p_idx);
		path[ppos].p_depth = i;
		path[ppos].p_ext = NULL;

		bh = read_extent_tree_block(inode, path[ppos].p_block, --i,
					    flags);
		if (IS_ERR(bh)) {
			ret = PTR_ERR(bh);
			goto err;
		}

		eh = ext_block_hdr(bh);
		ppos++;
		path[ppos].p_bh = bh;
		path[ppos].p_hdr = eh;
	}

	path[ppos].p_depth = i;
	path[ppos].p_ext = NULL;
	path[ppos].p_idx = NULL;

	/* find extent */
	ext4_ext_binsearch(inode, path + ppos, block);
	/* if not an empty leaf */
	if (path[ppos].p_ext)
		path[ppos].p_block = ext4_ext_pblock(path[ppos].p_ext);

	ext4_ext_show_path(inode, path);

	return path;

err:
	ext4_ext_drop_refs(path);
	kfree(path);
	if (orig_path)
		*orig_path = NULL;
	return ERR_PTR(ret);
}

/*
 * ext4_ext_insert_index:
 * insert new index [@logical;@ptr] into the block at @curp;
 * check where to insert: before @curp or after @curp
 */
static int ext4_ext_insert_index(handle_t *handle, struct inode *inode,
				 struct ext4_ext_path *curp,
				 int logical, ext4_fsblk_t ptr)
{
	struct ext4_extent_idx *ix;
	int len, err;

	err = ext4_ext_get_access(handle, inode, curp);
	if (err)
		return err;

	if (unlikely(logical == le32_to_cpu(curp->p_idx->ei_block))) {
		EXT4_ERROR_INODE(inode,
				 "logical %d == ei_block %d!",
				 logical, le32_to_cpu(curp->p_idx->ei_block));
		return -EFSCORRUPTED;
	}

	if (unlikely(le16_to_cpu(curp->p_hdr->eh_entries)
			     >= le16_to_cpu(curp->p_hdr->eh_max))) {
		EXT4_ERROR_INODE(inode,
				 "eh_entries %d >= eh_max %d!",
				 le16_to_cpu(curp->p_hdr->eh_entries),
				 le16_to_cpu(curp->p_hdr->eh_max));
		return -EFSCORRUPTED;
	}

	if (logical > le32_to_cpu(curp->p_idx->ei_block)) {
		/* insert after */
		ext_debug(inode, "insert new index %d after: %llu\n",
			  logical, ptr);
		ix = curp->p_idx + 1;
	} else {
		/* insert before */
		ext_debug(inode, "insert new index %d before: %llu\n",
			  logical, ptr);
		ix = curp->p_idx;
	}

	len = EXT_LAST_INDEX(curp->p_hdr) - ix + 1;
	BUG_ON(len < 0);
	if (len > 0) {
		ext_debug(inode, "insert new index %d: "
				"move %d indices from 0x%p to 0x%p\n",
				logical, len, ix, ix + 1);
		memmove(ix + 1, ix, len * sizeof(struct ext4_extent_idx));
	}

	if (unlikely(ix > EXT_MAX_INDEX(curp->p_hdr))) {
		EXT4_ERROR_INODE(inode, "ix > EXT_MAX_INDEX!");
		return -EFSCORRUPTED;
	}

	ix->ei_block = cpu_to_le32(logical);
	ext4_idx_store_pblock(ix, ptr);
	le16_add_cpu(&curp->p_hdr->eh_entries, 1);

	if (unlikely(ix > EXT_LAST_INDEX(curp->p_hdr))) {
		EXT4_ERROR_INODE(inode, "ix > EXT_LAST_INDEX!");
		return -EFSCORRUPTED;
	}

	err = ext4_ext_dirty(handle, inode, curp);
	ext4_std_error(inode->i_sb, err);

	return err;
}

/*
 * ext4_ext_split:
 * inserts new subtree into the path, using free index entry
 * at depth @at:
 * - allocates all needed blocks (new leaf and all intermediate index blocks)
 * - makes decision where to split
 * - moves remaining extents and index entries (right to the split point)
 *   into the newly allocated blocks
 * - initializes subtree
 */
static int ext4_ext_split(handle_t *handle, struct inode *inode,
			  unsigned int flags,
			  struct ext4_ext_path *path,
			  struct ext4_extent *newext, int at)
{
	struct buffer_head *bh = NULL;
	int depth = ext_depth(inode);
	struct ext4_extent_header *neh;
	struct ext4_extent_idx *fidx;
	int i = at, k, m, a;
	ext4_fsblk_t newblock, oldblock;
	__le32 border;
	ext4_fsblk_t *ablocks = NULL; /* array of allocated blocks */
	gfp_t gfp_flags = GFP_NOFS;
	int err = 0;
	size_t ext_size = 0;

	if (flags & EXT4_EX_NOFAIL)
		gfp_flags |= __GFP_NOFAIL;

	/* make decision: where to split? */
	/* FIXME: now decision is simplest: at current extent */

	/* if current leaf will be split, then we should use
	 * border from split point */
	if (unlikely(path[depth].p_ext > EXT_MAX_EXTENT(path[depth].p_hdr))) {
		EXT4_ERROR_INODE(inode, "p_ext > EXT_MAX_EXTENT!");
		return -EFSCORRUPTED;
	}
	if (path[depth].p_ext != EXT_MAX_EXTENT(path[depth].p_hdr)) {
		border = path[depth].p_ext[1].ee_block;
		ext_debug(inode, "leaf will be split."
				" next leaf starts at %d\n",
				  le32_to_cpu(border));
	} else {
		border = newext->ee_block;
		ext_debug(inode, "leaf will be added."
				" next leaf starts at %d\n",
				le32_to_cpu(border));
	}

	/*
	 * If error occurs, then we break processing
	 * and mark filesystem read-only. index won't
	 * be inserted and tree will be in consistent
	 * state. Next mount will repair buffers too.
	 */

	/*
	 * Get array to track all allocated blocks.
	 * We need this to handle errors and free blocks
	 * upon them.
	 */
	ablocks = kcalloc(depth, sizeof(ext4_fsblk_t), gfp_flags);
	if (!ablocks)
		return -ENOMEM;

	/* allocate all needed blocks */
	ext_debug(inode, "allocate %d blocks for indexes/leaf\n", depth - at);
	for (a = 0; a < depth - at; a++) {
		newblock = ext4_ext_new_meta_block(handle, inode, path,
						   newext, &err, flags);
		if (newblock == 0)
			goto cleanup;
		ablocks[a] = newblock;
	}

	/* initialize new leaf */
	newblock = ablocks[--a];
	if (unlikely(newblock == 0)) {
		EXT4_ERROR_INODE(inode, "newblock == 0!");
		err = -EFSCORRUPTED;
		goto cleanup;
	}
	bh = sb_getblk_gfp(inode->i_sb, newblock, __GFP_MOVABLE | GFP_NOFS);
	if (unlikely(!bh)) {
		err = -ENOMEM;
		goto cleanup;
	}
	lock_buffer(bh);

	err = ext4_journal_get_create_access(handle, bh);
	if (err)
		goto cleanup;

	neh = ext_block_hdr(bh);
	neh->eh_entries = 0;
	neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode, 0));
	neh->eh_magic = EXT4_EXT_MAGIC;
	neh->eh_depth = 0;

	/* move remainder of path[depth] to the new leaf */
	if (unlikely(path[depth].p_hdr->eh_entries !=
		     path[depth].p_hdr->eh_max)) {
		EXT4_ERROR_INODE(inode, "eh_entries %d != eh_max %d!",
				 path[depth].p_hdr->eh_entries,
				 path[depth].p_hdr->eh_max);
		err = -EFSCORRUPTED;
		goto cleanup;
	}
	/* start copy from next extent */
	m = EXT_MAX_EXTENT(path[depth].p_hdr) - path[depth].p_ext++;
	ext4_ext_show_move(inode, path, newblock, depth);
	if (m) {
		struct ext4_extent *ex;
		ex = EXT_FIRST_EXTENT(neh);
		memmove(ex, path[depth].p_ext, sizeof(struct ext4_extent) * m);
		le16_add_cpu(&neh->eh_entries, m);
	}

	/* zero out unused area in the extent block */
	ext_size = sizeof(struct ext4_extent_header) +
		sizeof(struct ext4_extent) * le16_to_cpu(neh->eh_entries);
	memset(bh->b_data + ext_size, 0, inode->i_sb->s_blocksize - ext_size);
	ext4_extent_block_csum_set(inode, neh);
	set_buffer_uptodate(bh);
	unlock_buffer(bh);

	err = ext4_handle_dirty_metadata(handle, inode, bh);
	if (err)
		goto cleanup;
	brelse(bh);
	bh = NULL;

	/* correct old leaf */
	if (m) {
		err = ext4_ext_get_access(handle, inode, path + depth);
		if (err)
			goto cleanup;
		le16_add_cpu(&path[depth].p_hdr->eh_entries, -m);
		err = ext4_ext_dirty(handle, inode, path + depth);
		if (err)
			goto cleanup;

	}

	/* create intermediate indexes */
	k = depth - at - 1;
	if (unlikely(k < 0)) {
		EXT4_ERROR_INODE(inode, "k %d < 0!", k);
		err = -EFSCORRUPTED;
		goto cleanup;
	}
	if (k)
		ext_debug(inode, "create %d intermediate indices\n", k);
	/* insert new index into current index block */
	/* current depth stored in i var */
	i = depth - 1;
	while (k--) {
		oldblock = newblock;
		newblock = ablocks[--a];
		bh = sb_getblk(inode->i_sb, newblock);
		if (unlikely(!bh)) {
			err = -ENOMEM;
			goto cleanup;
		}
		lock_buffer(bh);

		err = ext4_journal_get_create_access(handle, bh);
		if (err)
			goto cleanup;

		neh = ext_block_hdr(bh);
		neh->eh_entries = cpu_to_le16(1);
		neh->eh_magic = EXT4_EXT_MAGIC;
		neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode, 0));
		neh->eh_depth = cpu_to_le16(depth - i);
		fidx = EXT_FIRST_INDEX(neh);
		fidx->ei_block = border;
		ext4_idx_store_pblock(fidx, oldblock);

		ext_debug(inode, "int.index at %d (block %llu): %u -> %llu\n",
				i, newblock, le32_to_cpu(border), oldblock);

		/* move remainder of path[i] to the new index block */
		if (unlikely(EXT_MAX_INDEX(path[i].p_hdr) !=
					EXT_LAST_INDEX(path[i].p_hdr))) {
			EXT4_ERROR_INODE(inode,
					 "EXT_MAX_INDEX != EXT_LAST_INDEX ee_block %d!",
					 le32_to_cpu(path[i].p_ext->ee_block));
			err = -EFSCORRUPTED;
			goto cleanup;
		}
		/* start copy indexes */
		m = EXT_MAX_INDEX(path[i].p_hdr) - path[i].p_idx++;
		ext_debug(inode, "cur 0x%p, last 0x%p\n", path[i].p_idx,
				EXT_MAX_INDEX(path[i].p_hdr));
		ext4_ext_show_move(inode, path, newblock, i);
		if (m) {
			memmove(++fidx, path[i].p_idx,
				sizeof(struct ext4_extent_idx) * m);
			le16_add_cpu(&neh->eh_entries, m);
		}
		/* zero out unused area in the extent block */
		ext_size = sizeof(struct ext4_extent_header) +
		   (sizeof(struct ext4_extent) * le16_to_cpu(neh->eh_entries));
		memset(bh->b_data + ext_size, 0,
			inode->i_sb->s_blocksize - ext_size);
		ext4_extent_block_csum_set(inode, neh);
		set_buffer_uptodate(bh);
		unlock_buffer(bh);

		err = ext4_handle_dirty_metadata(handle, inode, bh);
		if (err)
			goto cleanup;
		brelse(bh);
		bh = NULL;

		/* correct old index */
		if (m) {
			err = ext4_ext_get_access(handle, inode, path + i);
			if (err)
				goto cleanup;
			le16_add_cpu(&path[i].p_hdr->eh_entries, -m);
			err = ext4_ext_dirty(handle, inode, path + i);
			if (err)
				goto cleanup;
		}

		i--;
	}

	/* insert new index */
	err = ext4_ext_insert_index(handle, inode, path + at,
				    le32_to_cpu(border), newblock);

cleanup:
	if (bh) {
		if (buffer_locked(bh))
			unlock_buffer(bh);
		brelse(bh);
	}

	if (err) {
		/* free all allocated blocks in error case */
		for (i = 0; i < depth; i++) {
			if (!ablocks[i])
				continue;
			ext4_free_blocks(handle, inode, NULL, ablocks[i], 1,
					 EXT4_FREE_BLOCKS_METADATA);
		}
	}
	kfree(ablocks);

	return err;
}

/*
 * ext4_ext_grow_indepth:
 * implements tree growing procedure:
 * - allocates new block
 * - moves top-level data (index block or leaf) into the new block
 * - initializes new top-level, creating index that points to the
 *   just created block
 */
static int ext4_ext_grow_indepth(handle_t *handle, struct inode *inode,
				 unsigned int flags)
{
	struct ext4_extent_header *neh;
	struct buffer_head *bh;
	ext4_fsblk_t newblock, goal = 0;
	struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
	int err = 0;
	size_t ext_size = 0;

	/* Try to prepend new index to old one */
	if (ext_depth(inode))
		goal = ext4_idx_pblock(EXT_FIRST_INDEX(ext_inode_hdr(inode)));
	if (goal > le32_to_cpu(es->s_first_data_block)) {
		flags |= EXT4_MB_HINT_TRY_GOAL;
		goal--;
	} else
		goal = ext4_inode_to_goal_block(inode);
	newblock = ext4_new_meta_blocks(handle, inode, goal, flags,
					NULL, &err);
	if (newblock == 0)
		return err;

	bh = sb_getblk_gfp(inode->i_sb, newblock, __GFP_MOVABLE | GFP_NOFS);
	if (unlikely(!bh))
		return -ENOMEM;
	lock_buffer(bh);

	err = ext4_journal_get_create_access(handle, bh);
	if (err) {
		unlock_buffer(bh);
		goto out;
	}

	ext_size = sizeof(EXT4_I(inode)->i_data);
	/* move top-level index/leaf into new block */
	memmove(bh->b_data, EXT4_I(inode)->i_data, ext_size);
	/* zero out unused area in the extent block */
	memset(bh->b_data + ext_size, 0, inode->i_sb->s_blocksize - ext_size);

	/* set size of new block */
	neh = ext_block_hdr(bh);
	/* old root could have indexes or leaves
	 * so calculate e_max right way */
	if (ext_depth(inode))
		neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode, 0));
	else
		neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode, 0));
	neh->eh_magic = EXT4_EXT_MAGIC;
	ext4_extent_block_csum_set(inode, neh);
	set_buffer_uptodate(bh);
	unlock_buffer(bh);

	err = ext4_handle_dirty_metadata(handle, inode, bh);
	if (err)
		goto out;

	/* Update top-level index: num,max,pointer */
	neh = ext_inode_hdr(inode);
	neh->eh_entries = cpu_to_le16(1);
	ext4_idx_store_pblock(EXT_FIRST_INDEX(neh), newblock);
	if (neh->eh_depth == 0) {
		/* Root extent block becomes index block */
		neh->eh_max = cpu_to_le16(ext4_ext_space_root_idx(inode, 0));
		EXT_FIRST_INDEX(neh)->ei_block =
			EXT_FIRST_EXTENT(neh)->ee_block;
	}
	ext_debug(inode, "new root: num %d(%d), lblock %d, ptr %llu\n",
		  le16_to_cpu(neh->eh_entries), le16_to_cpu(neh->eh_max),
		  le32_to_cpu(EXT_FIRST_INDEX(neh)->ei_block),
		  ext4_idx_pblock(EXT_FIRST_INDEX(neh)));

	le16_add_cpu(&neh->eh_depth, 1);
	err = ext4_mark_inode_dirty(handle, inode);
out:
	brelse(bh);

	return err;
}

/*
 * ext4_ext_create_new_leaf:
 * finds empty index and adds new leaf.
 * if no free index is found, then it requests in-depth growing.
 */
static int ext4_ext_create_new_leaf(handle_t *handle, struct inode *inode,
				    unsigned int mb_flags,
				    unsigned int gb_flags,
				    struct ext4_ext_path **ppath,
				    struct ext4_extent *newext)
{
	struct ext4_ext_path *path = *ppath;
	struct ext4_ext_path *curp;
	int depth, i, err = 0;

repeat:
	i = depth = ext_depth(inode);

	/* walk up to the tree and look for free index entry */
	curp = path + depth;
	while (i > 0 && !EXT_HAS_FREE_INDEX(curp)) {
		i--;
		curp--;
	}

	/* we use already allocated block for index block,
	 * so subsequent data blocks should be contiguous */
	if (EXT_HAS_FREE_INDEX(curp)) {
		/* if we found index with free entry, then use that
		 * entry: create all needed subtree and add new leaf */
		err = ext4_ext_split(handle, inode, mb_flags, path, newext, i);
		if (err)
			goto out;

		/* refill path */
		path = ext4_find_extent(inode,
				    (ext4_lblk_t)le32_to_cpu(newext->ee_block),
				    ppath, gb_flags);
		if (IS_ERR(path))
			err = PTR_ERR(path);
	} else {
		/* tree is full, time to grow in depth */
		err = ext4_ext_grow_indepth(handle, inode, mb_flags);
		if (err)
			goto out;

		/* refill path */
		path = ext4_find_extent(inode,
				   (ext4_lblk_t)le32_to_cpu(newext->ee_block),
				    ppath, gb_flags);
		if (IS_ERR(path)) {
			err = PTR_ERR(path);
			goto out;
		}

		/*
		 * only first (depth 0 -> 1) produces free space;
		 * in all other cases we have to split the grown tree
		 */
		depth = ext_depth(inode);
		if (path[depth].p_hdr->eh_entries == path[depth].p_hdr->eh_max) {
			/* now we need to split */
			goto repeat;
		}
	}

out:
	return err;
}

/*
 * search the closest allocated block to the left for *logical
 * and returns it at @logical + it's physical address at @phys
 * if *logical is the smallest allocated block, the function
 * returns 0 at @phys
 * return value contains 0 (success) or error code
 */
static int ext4_ext_search_left(struct inode *inode,
				struct ext4_ext_path *path,
				ext4_lblk_t *logical, ext4_fsblk_t *phys)
{
	struct ext4_extent_idx *ix;
	struct ext4_extent *ex;
	int depth, ee_len;

	if (unlikely(path == NULL)) {
		EXT4_ERROR_INODE(inode, "path == NULL *logical %d!", *logical);
		return -EFSCORRUPTED;
	}
	depth = path->p_depth;
	*phys = 0;

	if (depth == 0 && path->p_ext == NULL)
		return 0;

	/* usually extent in the path covers blocks smaller
	 * then *logical, but it can be that extent is the
	 * first one in the file */

	ex = path[depth].p_ext;
	ee_len = ext4_ext_get_actual_len(ex);
	if (*logical < le32_to_cpu(ex->ee_block)) {
		if (unlikely(EXT_FIRST_EXTENT(path[depth].p_hdr) != ex)) {
			EXT4_ERROR_INODE(inode,
					 "EXT_FIRST_EXTENT != ex *logical %d ee_block %d!",
					 *logical, le32_to_cpu(ex->ee_block));
			return -EFSCORRUPTED;
		}
		while (--depth >= 0) {
			ix = path[depth].p_idx;
			if (unlikely(ix != EXT_FIRST_INDEX(path[depth].p_hdr))) {
				EXT4_ERROR_INODE(inode,
				  "ix (%d) != EXT_FIRST_INDEX (%d) (depth %d)!"