[linux-2.6-block.git] / fs / nilfs2 / dat.c

/*
 * dat.c - NILFS disk address translation.
 *
 * Copyright (C) 2006-2008 Nippon Telegraph and Telephone Corporation.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 *
 * Written by Koji Sato <koji@osrg.net>.
 */

#include <linux/types.h>
#include <linux/buffer_head.h>
#include <linux/string.h>
#include <linux/errno.h>
#include "nilfs.h"
#include "mdt.h"
#include "alloc.h"
#include "dat.h"


#define NILFS_CNO_MIN	((__u64)1)
#define NILFS_CNO_MAX	(~(__u64)0)

/**
 * struct nilfs_dat_info - on-memory private data of DAT file
 * @mi: on-memory private data of metadata file
 * @palloc_cache: persistent object allocator cache of DAT file
 * @shadow: shadow map of DAT file
 */
struct nilfs_dat_info {
	struct nilfs_mdt_info mi;
	struct nilfs_palloc_cache palloc_cache;
	struct nilfs_shadow_map shadow;
};

static inline struct nilfs_dat_info *NILFS_DAT_I(struct inode *dat)
{
	return (struct nilfs_dat_info *)NILFS_MDT(dat);
}

static int nilfs_dat_prepare_entry(struct inode *dat,
				   struct nilfs_palloc_req *req, int create)
{
	return nilfs_palloc_get_entry_block(dat, req->pr_entry_nr,
					    create, &req->pr_entry_bh);
}

static void nilfs_dat_commit_entry(struct inode *dat,
				   struct nilfs_palloc_req *req)
{
	mark_buffer_dirty(req->pr_entry_bh);
	nilfs_mdt_mark_dirty(dat);
	brelse(req->pr_entry_bh);
}

static void nilfs_dat_abort_entry(struct inode *dat,
				  struct nilfs_palloc_req *req)
{
	brelse(req->pr_entry_bh);
}

int nilfs_dat_prepare_alloc(struct inode *dat, struct nilfs_palloc_req *req)
{
	int ret;

	ret = nilfs_palloc_prepare_alloc_entry(dat, req);
	if (ret < 0)
		return ret;

	ret = nilfs_dat_prepare_entry(dat, req, 1);
	if (ret < 0)
		nilfs_palloc_abort_alloc_entry(dat, req);

	return ret;
}

void nilfs_dat_commit_alloc(struct inode *dat, struct nilfs_palloc_req *req)
{
	struct nilfs_dat_entry *entry;
	void *kaddr;

	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
					     req->pr_entry_bh, kaddr);
	entry->de_start = cpu_to_le64(NILFS_CNO_MIN);
	entry->de_end = cpu_to_le64(NILFS_CNO_MAX);
	entry->de_blocknr = cpu_to_le64(0);
	kunmap_atomic(kaddr);

	nilfs_palloc_commit_alloc_entry(dat, req);
	nilfs_dat_commit_entry(dat, req);
}

void nilfs_dat_abort_alloc(struct inode *dat, struct nilfs_palloc_req *req)
{
	nilfs_dat_abort_entry(dat, req);
	nilfs_palloc_abort_alloc_entry(dat, req);
}

static void nilfs_dat_commit_free(struct inode *dat,
				  struct nilfs_palloc_req *req)
{
	struct nilfs_dat_entry *entry;
	void *kaddr;

	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
					     req->pr_entry_bh, kaddr);
	entry->de_start = cpu_to_le64(NILFS_CNO_MIN);
	entry->de_end = cpu_to_le64(NILFS_CNO_MIN);
	entry->de_blocknr = cpu_to_le64(0);
	kunmap_atomic(kaddr);

	nilfs_dat_commit_entry(dat, req);
	nilfs_palloc_commit_free_entry(dat, req);
}

int nilfs_dat_prepare_start(struct inode *dat, struct nilfs_palloc_req *req)
{
	int ret;

	ret = nilfs_dat_prepare_entry(dat, req, 0);
	WARN_ON(ret == -ENOENT);
	return ret;
}

void nilfs_dat_commit_start(struct inode *dat, struct nilfs_palloc_req *req,
			    sector_t blocknr)
{
	struct nilfs_dat_entry *entry;
	void *kaddr;

	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
					     req->pr_entry_bh, kaddr);
	entry->de_start = cpu_to_le64(nilfs_mdt_cno(dat));
	entry->de_blocknr = cpu_to_le64(blocknr);
	kunmap_atomic(kaddr);

	nilfs_dat_commit_entry(dat, req);
}

int nilfs_dat_prepare_end(struct inode *dat, struct nilfs_palloc_req *req)
{
	struct nilfs_dat_entry *entry;
	sector_t blocknr;
	void *kaddr;
	int ret;

	ret = nilfs_dat_prepare_entry(dat, req, 0);
	if (ret < 0) {
		WARN_ON(ret == -ENOENT);
		return ret;
	}

	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
					     req->pr_entry_bh, kaddr);
	blocknr = le64_to_cpu(entry->de_blocknr);
	kunmap_atomic(kaddr);

	if (blocknr == 0) {
		ret = nilfs_palloc_prepare_free_entry(dat, req);
		if (ret < 0) {
			nilfs_dat_abort_entry(dat, req);
			return ret;
		}
	}

	return 0;
}

void nilfs_dat_commit_end(struct inode *dat, struct nilfs_palloc_req *req,
			  int dead)
{
	struct nilfs_dat_entry *entry;
	__u64 start, end;
	sector_t blocknr;
	void *kaddr;

	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
					     req->pr_entry_bh, kaddr);
	end = start = le64_to_cpu(entry->de_start);
	if (!dead) {
		end = nilfs_mdt_cno(dat);
		WARN_ON(start > end);
	}
	entry->de_end = cpu_to_le64(end);
	blocknr = le64_to_cpu(entry->de_blocknr);
	kunmap_atomic(kaddr);

	if (blocknr == 0)
		nilfs_dat_commit_free(dat, req);
	else
		nilfs_dat_commit_entry(dat, req);
}

void nilfs_dat_abort_end(struct inode *dat, struct nilfs_palloc_req *req)
{
	struct nilfs_dat_entry *entry;
	__u64 start;
	sector_t blocknr;
	void *kaddr;

	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
					     req->pr_entry_bh, kaddr);
	start = le64_to_cpu(entry->de_start);
	blocknr = le64_to_cpu(entry->de_blocknr);
	kunmap_atomic(kaddr);

	if (start == nilfs_mdt_cno(dat) && blocknr == 0)
		nilfs_palloc_abort_free_entry(dat, req);
	nilfs_dat_abort_entry(dat, req);
}

int nilfs_dat_prepare_update(struct inode *dat,
			     struct nilfs_palloc_req *oldreq,
			     struct nilfs_palloc_req *newreq)
{
	int ret;

	ret = nilfs_dat_prepare_end(dat, oldreq);
	if (!ret) {
		ret = nilfs_dat_prepare_alloc(dat, newreq);
		if (ret < 0)
			nilfs_dat_abort_end(dat, oldreq);
	}
	return ret;
}

void nilfs_dat_commit_update(struct inode *dat,
			     struct nilfs_palloc_req *oldreq,
			     struct nilfs_palloc_req *newreq, int dead)
{
	nilfs_dat_commit_end(dat, oldreq, dead);
	nilfs_dat_commit_alloc(dat, newreq);
}

void nilfs_dat_abort_update(struct inode *dat,
			    struct nilfs_palloc_req *oldreq,
			    struct nilfs_palloc_req *newreq)
{
	nilfs_dat_abort_end(dat, oldreq);
	nilfs_dat_abort_alloc(dat, newreq);
}

/**
 * nilfs_dat_mark_dirty -
 * @dat: DAT file inode
 * @vblocknr: virtual block number
 *
 * Description:
 *
 * Return Value: On success, 0 is returned. On error, one of the following
 * negative error codes is returned.
 *
 * %-EIO - I/O error.
 *
 * %-ENOMEM - Insufficient amount of memory available.
 */
int nilfs_dat_mark_dirty(struct inode *dat, __u64 vblocknr)
{
	struct nilfs_palloc_req req;
	int ret;

	req.pr_entry_nr = vblocknr;
	ret = nilfs_dat_prepare_entry(dat, &req, 0);
	if (ret == 0)
		nilfs_dat_commit_entry(dat, &req);
	return ret;
}

/**
 * nilfs_dat_freev - free virtual block numbers
 * @dat: DAT file inode
 * @vblocknrs: array of virtual block numbers
 * @nitems: number of virtual block numbers
 *
 * Description: nilfs_dat_freev() frees the virtual block numbers specified by
 * @vblocknrs and @nitems.
 *
 * Return Value: On success, 0 is returned. On error, one of the following
 * negative error codes is returned.
 *
 * %-EIO - I/O error.
 *
 * %-ENOMEM - Insufficient amount of memory available.
 *
 * %-ENOENT - The virtual block number have not been allocated.
 */
int nilfs_dat_freev(struct inode *dat, __u64 *vblocknrs, size_t nitems)
{
	return nilfs_palloc_freev(dat, vblocknrs, nitems);
}

/**
 * nilfs_dat_move - change a block number
 * @dat: DAT file inode
 * @vblocknr: virtual block number
 * @blocknr: block number
 *
 * Description: nilfs_dat_move() changes the block number associated with
 * @vblocknr to @blocknr.
 *
 * Return Value: On success, 0 is returned. On error, one of the following
 * negative error codes is returned.
 *
 * %-EIO - I/O error.
 *
 * %-ENOMEM - Insufficient amount of memory available.
 */
int nilfs_dat_move(struct inode *dat, __u64 vblocknr, sector_t blocknr)
{
	struct buffer_head *entry_bh;
	struct nilfs_dat_entry *entry;
	void *kaddr;
	int ret;

	ret = nilfs_palloc_get_entry_block(dat, vblocknr, 0, &entry_bh);
	if (ret < 0)
		return ret;

	/*
	 * The given disk block number (blocknr) is not yet written to
	 * the device at this point.
	 *
	 * To prevent nilfs_dat_translate() from returning the
	 * uncommitted block number, this makes a copy of the entry
	 * buffer and redirects nilfs_dat_translate() to the copy.
	 */
	if (!buffer_nilfs_redirected(entry_bh)) {
		ret = nilfs_mdt_freeze_buffer(dat, entry_bh);
		if (ret) {
			brelse(entry_bh);
			return ret;
		}
	}

	kaddr = kmap_atomic(entry_bh->b_page);
	entry = nilfs_palloc_block_get_entry(dat, vblocknr, entry_bh, kaddr);
	if (unlikely(entry->de_blocknr == cpu_to_le64(0))) {
		printk(KERN_CRIT "%s: vbn = %llu, [%llu, %llu)\n", __func__,
		       (unsigned long long)vblocknr,
		       (unsigned long long)le64_to_cpu(entry->de_start),
		       (unsigned long long)le64_to_cpu(entry->de_end));
		kunmap_atomic(kaddr);
		brelse(entry_bh);
		return -EINVAL;
	}
	WARN_ON(blocknr == 0);
	entry->de_blocknr = cpu_to_le64(blocknr);
	kunmap_atomic(kaddr);

	mark_buffer_dirty(entry_bh);
	nilfs_mdt_mark_dirty(dat);

	brelse(entry_bh);

	return 0;
}

/**
 * nilfs_dat_translate - translate a virtual block number to a block number
 * @dat: DAT file inode
 * @vblocknr: virtual block number
 * @blocknrp: pointer to a block number
 *
 * Description: nilfs_dat_translate() maps the virtual block number @vblocknr
 * to the corresponding block number.
 *
 * Return Value: On success, 0 is returned and the block number associated
 * with @vblocknr is stored in the place pointed by @blocknrp. On error, one
 * of the following negative error codes is returned.
 *
 * %-EIO - I/O error.
 *
 * %-ENOMEM - Insufficient amount of memory available.
 *
 * %-ENOENT - A block number associated with @vblocknr does not exist.
 */
int nilfs_dat_translate(struct inode *dat, __u64 vblocknr, sector_t *blocknrp)
{
	struct buffer_head *entry_bh, *bh;
	struct nilfs_dat_entry *entry;
	sector_t blocknr;
	void *kaddr;
	int ret;

	ret = nilfs_palloc_get_entry_block(dat, vblocknr, 0, &entry_bh);
	if (ret < 0)
		return ret;

	if (!nilfs_doing_gc() && buffer_nilfs_redirected(entry_bh)) {
		bh = nilfs_mdt_get_frozen_buffer(dat, entry_bh);
		if (bh) {
			WARN_ON(!buffer_uptodate(bh));
			brelse(entry_bh);
			entry_bh = bh;
		}
	}

	kaddr = kmap_atomic(entry_bh->b_page);
	entry = nilfs_palloc_block_get_entry(dat, vblocknr, entry_bh, kaddr);
	blocknr = le64_to_cpu(entry->de_blocknr);
	if (blocknr == 0) {
		ret = -ENOENT;
		goto out;
	}
	*blocknrp = blocknr;

 out:
	kunmap_atomic(kaddr);
	brelse(entry_bh);
	return ret;
}

ssize_t nilfs_dat_get_vinfo(struct inode *dat, void *buf, unsigned visz,
			    size_t nvi)
{
	struct buffer_head *entry_bh;
	struct nilfs_dat_entry *entry;
	struct nilfs_vinfo *vinfo = buf;
	__u64 first, last;
	void *kaddr;
	unsigned long entries_per_block = NILFS_MDT(dat)->mi_entries_per_block;
	int i, j, n, ret;

	for (i = 0; i < nvi; i += n) {
		ret = nilfs_palloc_get_entry_block(dat, vinfo->vi_vblocknr,
						   0, &entry_bh);
		if (ret < 0)
			return ret;
		kaddr = kmap_atomic(entry_bh->b_page);
		/* last virtual block number in this block */
		first = vinfo->vi_vblocknr;
		do_div(first, entries_per_block);
		first *= entries_per_block;
		last = first + entries_per_block - 1;
		for (j = i, n = 0;
		     j < nvi && vinfo->vi_vblocknr >= first &&
			     vinfo->vi_vblocknr <= last;
		     j++, n++, vinfo = (void *)vinfo + visz) {
			entry = nilfs_palloc_block_get_entry(
				dat, vinfo->vi_vblocknr, entry_bh, kaddr);
			vinfo->vi_start = le64_to_cpu(entry->de_start);
			vinfo->vi_end = le64_to_cpu(entry->de_end);
			vinfo->vi_blocknr = le64_to_cpu(entry->de_blocknr);
		}
		kunmap_atomic(kaddr);
		brelse(entry_bh);
	}

	return nvi;
}

/**
 * nilfs_dat_read - read or get dat inode
 * @sb: super block instance
 * @entry_size: size of a dat entry
 * @raw_inode: on-disk dat inode
 * @inodep: buffer to store the inode
 */
int nilfs_dat_read(struct super_block *sb, size_t entry_size,
		   struct nilfs_inode *raw_inode, struct inode **inodep)
{
	static struct lock_class_key dat_lock_key;
	struct inode *dat;
	struct nilfs_dat_info *di;
	int err;

	if (entry_size > sb->s_blocksize) {
		printk(KERN_ERR
		       "NILFS: too large DAT entry size: %zu bytes.\n",
		       entry_size);
		return -EINVAL;
	} else if (entry_size < NILFS_MIN_DAT_ENTRY_SIZE) {
		printk(KERN_ERR
		       "NILFS: too small DAT entry size: %zu bytes.\n",
		       entry_size);
		return -EINVAL;
	}

	dat = nilfs_iget_locked(sb, NULL, NILFS_DAT_INO);
	if (unlikely(!dat))
		return -ENOMEM;
	if (!(dat->i_state & I_NEW))
		goto out;

	err = nilfs_mdt_init(dat, NILFS_MDT_GFP, sizeof(*di));
	if (err)
		goto failed;

	err = nilfs_palloc_init_blockgroup(dat, entry_size);
	if (err)
		goto failed;

	di = NILFS_DAT_I(dat);
	lockdep_set_class(&di->mi.mi_sem, &dat_lock_key);
	nilfs_palloc_setup_cache(dat, &di->palloc_cache);
	nilfs_mdt_setup_shadow_map(dat, &di->shadow);

	err = nilfs_read_inode_common(dat, raw_inode);
	if (err)
		goto failed;

	unlock_new_inode(dat);
 out:
	*inodep = dat;
	return 0;
 failed:
	iget_failed(dat);
	return err;
}
Commit	Line	Data
a17564f5 KS	1	/*
	2	* dat.c - NILFS disk address translation.
	3	*
	4	* Copyright (C) 2006-2008 Nippon Telegraph and Telephone Corporation.
	5	*
	6	* This program is free software; you can redistribute it and/or modify
	7	* it under the terms of the GNU General Public License as published by
	8	* the Free Software Foundation; either version 2 of the License, or
	9	* (at your option) any later version.
	10	*
	11	* This program is distributed in the hope that it will be useful,
	12	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	13	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	14	* GNU General Public License for more details.
	15	*
	16	* You should have received a copy of the GNU General Public License
	17	* along with this program; if not, write to the Free Software
	18	* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
	19	*
	20	* Written by Koji Sato <koji@osrg.net>.
	21	*/
	22
	23	#include <linux/types.h>
	24	#include <linux/buffer_head.h>
	25	#include <linux/string.h>
	26	#include <linux/errno.h>
	27	#include "nilfs.h"
	28	#include "mdt.h"
	29	#include "alloc.h"
	30	#include "dat.h"
	31
	32
	33	#define NILFS_CNO_MIN ((__u64)1)
	34	#define NILFS_CNO_MAX (~(__u64)0)
	35
f5974c8f VD	36	/**
	37	* struct nilfs_dat_info - on-memory private data of DAT file
	38	* @mi: on-memory private data of metadata file
	39	* @palloc_cache: persistent object allocator cache of DAT file
	40	* @shadow: shadow map of DAT file
	41	*/
8908b2f7 RK	42	struct nilfs_dat_info {
	43	struct nilfs_mdt_info mi;
	44	struct nilfs_palloc_cache palloc_cache;
c1c1d709	45	struct nilfs_shadow_map shadow;
8908b2f7 RK	46	};
	47
	48	static inline struct nilfs_dat_info NILFS_DAT_I(struct inode dat)
	49	{
	50	return (struct nilfs_dat_info *)NILFS_MDT(dat);
	51	}
	52
a17564f5 KS	53	static int nilfs_dat_prepare_entry(struct inode *dat,
	54	struct nilfs_palloc_req *req, int create)
	55	{
	56	return nilfs_palloc_get_entry_block(dat, req->pr_entry_nr,
	57	create, &req->pr_entry_bh);
	58	}
	59
	60	static void nilfs_dat_commit_entry(struct inode *dat,
	61	struct nilfs_palloc_req *req)
	62	{
5fc7b141	63	mark_buffer_dirty(req->pr_entry_bh);
a17564f5 KS	64	nilfs_mdt_mark_dirty(dat);
	65	brelse(req->pr_entry_bh);
	66	}
	67
	68	static void nilfs_dat_abort_entry(struct inode *dat,
	69	struct nilfs_palloc_req *req)
	70	{
	71	brelse(req->pr_entry_bh);
	72	}
	73
	74	int nilfs_dat_prepare_alloc(struct inode dat, struct nilfs_palloc_req req)
	75	{
	76	int ret;
	77
	78	ret = nilfs_palloc_prepare_alloc_entry(dat, req);
	79	if (ret < 0)
	80	return ret;
	81
	82	ret = nilfs_dat_prepare_entry(dat, req, 1);
	83	if (ret < 0)
	84	nilfs_palloc_abort_alloc_entry(dat, req);
	85
	86	return ret;
	87	}
	88
	89	void nilfs_dat_commit_alloc(struct inode dat, struct nilfs_palloc_req req)
	90	{
	91	struct nilfs_dat_entry *entry;
	92	void *kaddr;
	93
7b9c0976	94	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	95	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
	96	req->pr_entry_bh, kaddr);
	97	entry->de_start = cpu_to_le64(NILFS_CNO_MIN);
	98	entry->de_end = cpu_to_le64(NILFS_CNO_MAX);
	99	entry->de_blocknr = cpu_to_le64(0);
7b9c0976	100	kunmap_atomic(kaddr);
a17564f5 KS	101
	102	nilfs_palloc_commit_alloc_entry(dat, req);
	103	nilfs_dat_commit_entry(dat, req);
	104	}
	105
	106	void nilfs_dat_abort_alloc(struct inode dat, struct nilfs_palloc_req req)
	107	{
	108	nilfs_dat_abort_entry(dat, req);
	109	nilfs_palloc_abort_alloc_entry(dat, req);
	110	}
	111
abc0b50b JS	112	static void nilfs_dat_commit_free(struct inode *dat,
abc0b50b JS	113	struct nilfs_palloc_req *req)
a17564f5 KS	114	{
	115	struct nilfs_dat_entry *entry;
	116	void *kaddr;
	117
7b9c0976	118	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	119	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
	120	req->pr_entry_bh, kaddr);
	121	entry->de_start = cpu_to_le64(NILFS_CNO_MIN);
	122	entry->de_end = cpu_to_le64(NILFS_CNO_MIN);
	123	entry->de_blocknr = cpu_to_le64(0);
7b9c0976	124	kunmap_atomic(kaddr);
a17564f5 KS	125
	126	nilfs_dat_commit_entry(dat, req);
	127	nilfs_palloc_commit_free_entry(dat, req);
	128	}
	129
a17564f5 KS	130	int nilfs_dat_prepare_start(struct inode dat, struct nilfs_palloc_req req)
	131	{
	132	int ret;
	133
	134	ret = nilfs_dat_prepare_entry(dat, req, 0);
1f5abe7e	135	WARN_ON(ret == -ENOENT);
a17564f5 KS	136	return ret;
	137	}
	138
	139	void nilfs_dat_commit_start(struct inode dat, struct nilfs_palloc_req req,
	140	sector_t blocknr)
	141	{
	142	struct nilfs_dat_entry *entry;
	143	void *kaddr;
	144
7b9c0976	145	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	146	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
	147	req->pr_entry_bh, kaddr);
	148	entry->de_start = cpu_to_le64(nilfs_mdt_cno(dat));
a17564f5	149	entry->de_blocknr = cpu_to_le64(blocknr);
7b9c0976	150	kunmap_atomic(kaddr);
a17564f5 KS	151
	152	nilfs_dat_commit_entry(dat, req);
	153	}
	154
a17564f5 KS	155	int nilfs_dat_prepare_end(struct inode dat, struct nilfs_palloc_req req)
	156	{
	157	struct nilfs_dat_entry *entry;
a17564f5 KS	158	sector_t blocknr;
	159	void *kaddr;
	160	int ret;
	161
	162	ret = nilfs_dat_prepare_entry(dat, req, 0);
	163	if (ret < 0) {
1f5abe7e	164	WARN_ON(ret == -ENOENT);
a17564f5 KS	165	return ret;
	166	}
	167
7b9c0976	168	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	169	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
a17564f5 KS	170	req->pr_entry_bh, kaddr);
a17564f5	171	blocknr = le64_to_cpu(entry->de_blocknr);
7b9c0976	172	kunmap_atomic(kaddr);
a17564f5 KS	173
	174	if (blocknr == 0) {
	175	ret = nilfs_palloc_prepare_free_entry(dat, req);
	176	if (ret < 0) {
	177	nilfs_dat_abort_entry(dat, req);
	178	return ret;
	179	}
	180	}
	181
	182	return 0;
	183	}
	184
	185	void nilfs_dat_commit_end(struct inode dat, struct nilfs_palloc_req req,
	186	int dead)
	187	{
	188	struct nilfs_dat_entry *entry;
	189	__u64 start, end;
	190	sector_t blocknr;
	191	void *kaddr;
	192
7b9c0976	193	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	194	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
	195	req->pr_entry_bh, kaddr);
	196	end = start = le64_to_cpu(entry->de_start);
	197	if (!dead) {
	198	end = nilfs_mdt_cno(dat);
1f5abe7e	199	WARN_ON(start > end);
a17564f5 KS	200	}
	201	entry->de_end = cpu_to_le64(end);
	202	blocknr = le64_to_cpu(entry->de_blocknr);
7b9c0976	203	kunmap_atomic(kaddr);
a17564f5 KS	204
	205	if (blocknr == 0)
	206	nilfs_dat_commit_free(dat, req);
	207	else
	208	nilfs_dat_commit_entry(dat, req);
	209	}
	210
	211	void nilfs_dat_abort_end(struct inode dat, struct nilfs_palloc_req req)
	212	{
	213	struct nilfs_dat_entry *entry;
	214	__u64 start;
	215	sector_t blocknr;
	216	void *kaddr;
	217
7b9c0976	218	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	219	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
	220	req->pr_entry_bh, kaddr);
	221	start = le64_to_cpu(entry->de_start);
	222	blocknr = le64_to_cpu(entry->de_blocknr);
7b9c0976	223	kunmap_atomic(kaddr);
a17564f5 KS	224
	225	if (start == nilfs_mdt_cno(dat) && blocknr == 0)
	226	nilfs_palloc_abort_free_entry(dat, req);
	227	nilfs_dat_abort_entry(dat, req);
	228	}
	229
bd8169ef RK	230	int nilfs_dat_prepare_update(struct inode *dat,
	231	struct nilfs_palloc_req *oldreq,
	232	struct nilfs_palloc_req *newreq)
	233	{
	234	int ret;
	235
	236	ret = nilfs_dat_prepare_end(dat, oldreq);
	237	if (!ret) {
	238	ret = nilfs_dat_prepare_alloc(dat, newreq);
	239	if (ret < 0)
	240	nilfs_dat_abort_end(dat, oldreq);
	241	}
	242	return ret;
	243	}
	244
	245	void nilfs_dat_commit_update(struct inode *dat,
	246	struct nilfs_palloc_req *oldreq,
	247	struct nilfs_palloc_req *newreq, int dead)
	248	{
	249	nilfs_dat_commit_end(dat, oldreq, dead);
	250	nilfs_dat_commit_alloc(dat, newreq);
	251	}
	252
	253	void nilfs_dat_abort_update(struct inode *dat,
	254	struct nilfs_palloc_req *oldreq,
	255	struct nilfs_palloc_req *newreq)
	256	{
	257	nilfs_dat_abort_end(dat, oldreq);
	258	nilfs_dat_abort_alloc(dat, newreq);
	259	}
	260
a17564f5 KS	261	/**
	262	* nilfs_dat_mark_dirty -
	263	* @dat: DAT file inode
	264	* @vblocknr: virtual block number
	265	*
	266	* Description:
	267	*
	268	* Return Value: On success, 0 is returned. On error, one of the following
	269	* negative error codes is returned.
	270	*
	271	* %-EIO - I/O error.
	272	*
	273	* %-ENOMEM - Insufficient amount of memory available.
	274	*/
	275	int nilfs_dat_mark_dirty(struct inode *dat, __u64 vblocknr)
	276	{
	277	struct nilfs_palloc_req req;
	278	int ret;
	279
	280	req.pr_entry_nr = vblocknr;
	281	ret = nilfs_dat_prepare_entry(dat, &req, 0);
	282	if (ret == 0)
	283	nilfs_dat_commit_entry(dat, &req);
	284	return ret;
	285	}
	286
	287	/**
	288	* nilfs_dat_freev - free virtual block numbers
	289	* @dat: DAT file inode
	290	* @vblocknrs: array of virtual block numbers
	291	* @nitems: number of virtual block numbers
	292	*
	293	* Description: nilfs_dat_freev() frees the virtual block numbers specified by
	294	* @vblocknrs and @nitems.
	295	*
	296	* Return Value: On success, 0 is returned. On error, one of the following
7a65004b	297	* negative error codes is returned.
a17564f5 KS	298	*
	299	* %-EIO - I/O error.
	300	*
	301	* %-ENOMEM - Insufficient amount of memory available.
	302	*
	303	* %-ENOENT - The virtual block number have not been allocated.
	304	*/
	305	int nilfs_dat_freev(struct inode dat, __u64 vblocknrs, size_t nitems)
	306	{
	307	return nilfs_palloc_freev(dat, vblocknrs, nitems);
	308	}
	309
	310	/**
	311	* nilfs_dat_move - change a block number
	312	* @dat: DAT file inode
	313	* @vblocknr: virtual block number
	314	* @blocknr: block number
	315	*
	316	* Description: nilfs_dat_move() changes the block number associated with
	317	* @vblocknr to @blocknr.
	318	*
	319	* Return Value: On success, 0 is returned. On error, one of the following
	320	* negative error codes is returned.
	321	*
	322	* %-EIO - I/O error.
	323	*
	324	* %-ENOMEM - Insufficient amount of memory available.
	325	*/
	326	int nilfs_dat_move(struct inode *dat, __u64 vblocknr, sector_t blocknr)
	327	{
	328	struct buffer_head *entry_bh;
	329	struct nilfs_dat_entry *entry;
	330	void *kaddr;
	331	int ret;
	332
	333	ret = nilfs_palloc_get_entry_block(dat, vblocknr, 0, &entry_bh);
	334	if (ret < 0)
	335	return ret;
c1c1d709 RK	336
	337	/*
	338	* The given disk block number (blocknr) is not yet written to
	339	* the device at this point.
	340	*
	341	* To prevent nilfs_dat_translate() from returning the
f6c26ec5	342	* uncommitted block number, this makes a copy of the entry
c1c1d709 RK	343	* buffer and redirects nilfs_dat_translate() to the copy.
	344	*/
	345	if (!buffer_nilfs_redirected(entry_bh)) {
	346	ret = nilfs_mdt_freeze_buffer(dat, entry_bh);
	347	if (ret) {
	348	brelse(entry_bh);
	349	return ret;
	350	}
	351	}
	352
7b9c0976	353	kaddr = kmap_atomic(entry_bh->b_page);
a17564f5	354	entry = nilfs_palloc_block_get_entry(dat, vblocknr, entry_bh, kaddr);
1f5abe7e	355	if (unlikely(entry->de_blocknr == cpu_to_le64(0))) {
a17564f5 KS	356	printk(KERN_CRIT "%s: vbn = %llu, [%llu, %llu)\n", __func__,
	357	(unsigned long long)vblocknr,
	358	(unsigned long long)le64_to_cpu(entry->de_start),
	359	(unsigned long long)le64_to_cpu(entry->de_end));
7b9c0976	360	kunmap_atomic(kaddr);
1f5abe7e RK	361	brelse(entry_bh);
1f5abe7e RK	362	return -EINVAL;
a17564f5	363	}
1f5abe7e	364	WARN_ON(blocknr == 0);
a17564f5	365	entry->de_blocknr = cpu_to_le64(blocknr);
7b9c0976	366	kunmap_atomic(kaddr);
a17564f5	367
5fc7b141	368	mark_buffer_dirty(entry_bh);
a17564f5 KS	369	nilfs_mdt_mark_dirty(dat);
	370
	371	brelse(entry_bh);
	372
	373	return 0;
	374	}
	375
	376	/**
	377	* nilfs_dat_translate - translate a virtual block number to a block number
	378	* @dat: DAT file inode
	379	* @vblocknr: virtual block number
	380	* @blocknrp: pointer to a block number
	381	*
	382	* Description: nilfs_dat_translate() maps the virtual block number @vblocknr
	383	* to the corresponding block number.
	384	*
	385	* Return Value: On success, 0 is returned and the block number associated
	386	* with @vblocknr is stored in the place pointed by @blocknrp. On error, one
	387	* of the following negative error codes is returned.
	388	*
	389	* %-EIO - I/O error.
	390	*
	391	* %-ENOMEM - Insufficient amount of memory available.
	392	*
	393	* %-ENOENT - A block number associated with @vblocknr does not exist.
	394	*/
	395	int nilfs_dat_translate(struct inode dat, __u64 vblocknr, sector_t blocknrp)
	396	{
c1c1d709	397	struct buffer_head entry_bh, bh;
a17564f5 KS	398	struct nilfs_dat_entry *entry;
	399	sector_t blocknr;
	400	void *kaddr;
	401	int ret;
	402
	403	ret = nilfs_palloc_get_entry_block(dat, vblocknr, 0, &entry_bh);
	404	if (ret < 0)
	405	return ret;
	406
c1c1d709 RK	407	if (!nilfs_doing_gc() && buffer_nilfs_redirected(entry_bh)) {
	408	bh = nilfs_mdt_get_frozen_buffer(dat, entry_bh);
	409	if (bh) {
	410	WARN_ON(!buffer_uptodate(bh));
	411	brelse(entry_bh);
	412	entry_bh = bh;
	413	}
	414	}
	415
7b9c0976	416	kaddr = kmap_atomic(entry_bh->b_page);
a17564f5 KS	417	entry = nilfs_palloc_block_get_entry(dat, vblocknr, entry_bh, kaddr);
	418	blocknr = le64_to_cpu(entry->de_blocknr);
	419	if (blocknr == 0) {
	420	ret = -ENOENT;
	421	goto out;
	422	}
086d1764	423	*blocknrp = blocknr;
a17564f5 KS	424
a17564f5 KS	425	out:
7b9c0976	426	kunmap_atomic(kaddr);
a17564f5 KS	427	brelse(entry_bh);
	428	return ret;
	429	}
	430
003ff182	431	ssize_t nilfs_dat_get_vinfo(struct inode dat, void buf, unsigned visz,
a17564f5 KS	432	size_t nvi)
	433	{
	434	struct buffer_head *entry_bh;
	435	struct nilfs_dat_entry *entry;
003ff182	436	struct nilfs_vinfo *vinfo = buf;
a17564f5 KS	437	__u64 first, last;
	438	void *kaddr;
	439	unsigned long entries_per_block = NILFS_MDT(dat)->mi_entries_per_block;
	440	int i, j, n, ret;
	441
	442	for (i = 0; i < nvi; i += n) {
003ff182	443	ret = nilfs_palloc_get_entry_block(dat, vinfo->vi_vblocknr,
a17564f5 KS	444	0, &entry_bh);
	445	if (ret < 0)
	446	return ret;
7b9c0976	447	kaddr = kmap_atomic(entry_bh->b_page);
a17564f5	448	/* last virtual block number in this block */
003ff182	449	first = vinfo->vi_vblocknr;
a17564f5 KS	450	do_div(first, entries_per_block);
	451	first *= entries_per_block;
	452	last = first + entries_per_block - 1;
	453	for (j = i, n = 0;
003ff182 RK	454	j < nvi && vinfo->vi_vblocknr >= first &&
	455	vinfo->vi_vblocknr <= last;
	456	j++, n++, vinfo = (void *)vinfo + visz) {
a17564f5	457	entry = nilfs_palloc_block_get_entry(
003ff182 RK	458	dat, vinfo->vi_vblocknr, entry_bh, kaddr);
	459	vinfo->vi_start = le64_to_cpu(entry->de_start);
	460	vinfo->vi_end = le64_to_cpu(entry->de_end);
	461	vinfo->vi_blocknr = le64_to_cpu(entry->de_blocknr);
a17564f5	462	}
7b9c0976	463	kunmap_atomic(kaddr);
a17564f5 KS	464	brelse(entry_bh);
	465	}
	466
	467	return nvi;
	468	}
79739565	469
8707df38	470	/**
f1e89c86 RK	471	* nilfs_dat_read - read or get dat inode
f1e89c86 RK	472	* @sb: super block instance
79739565	473	* @entry_size: size of a dat entry
f1e89c86 RK	474	* @raw_inode: on-disk dat inode
f1e89c86 RK	475	* @inodep: buffer to store the inode
79739565	476	*/
f1e89c86 RK	477	int nilfs_dat_read(struct super_block *sb, size_t entry_size,
f1e89c86 RK	478	struct nilfs_inode raw_inode, struct inode *inodep)
79739565 RK	479	{
	480	static struct lock_class_key dat_lock_key;
	481	struct inode *dat;
8908b2f7	482	struct nilfs_dat_info *di;
79739565 RK	483	int err;
79739565 RK	484
0ec060d1 RK	485	if (entry_size > sb->s_blocksize) {
	486	printk(KERN_ERR
	487	"NILFS: too large DAT entry size: %zu bytes.\n",
	488	entry_size);
	489	return -EINVAL;
	490	} else if (entry_size < NILFS_MIN_DAT_ENTRY_SIZE) {
	491	printk(KERN_ERR
	492	"NILFS: too small DAT entry size: %zu bytes.\n",
	493	entry_size);
	494	return -EINVAL;
	495	}
	496
f1e89c86 RK	497	dat = nilfs_iget_locked(sb, NULL, NILFS_DAT_INO);
	498	if (unlikely(!dat))
	499	return -ENOMEM;
	500	if (!(dat->i_state & I_NEW))
	501	goto out;
8908b2f7	502
f1e89c86 RK	503	err = nilfs_mdt_init(dat, NILFS_MDT_GFP, sizeof(*di));
	504	if (err)
	505	goto failed;
	506
	507	err = nilfs_palloc_init_blockgroup(dat, entry_size);
	508	if (err)
	509	goto failed;
	510
	511	di = NILFS_DAT_I(dat);
	512	lockdep_set_class(&di->mi.mi_sem, &dat_lock_key);
	513	nilfs_palloc_setup_cache(dat, &di->palloc_cache);
	514	nilfs_mdt_setup_shadow_map(dat, &di->shadow);
	515
	516	err = nilfs_read_inode_common(dat, raw_inode);
	517	if (err)
	518	goto failed;
	519
	520	unlock_new_inode(dat);
	521	out:
	522	*inodep = dat;
	523	return 0;
	524	failed:
	525	iget_failed(dat);
	526	return err;
79739565	527	}