[mirror_ubuntu-artful-kernel.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.
 *
 * Written by Koji Sato.
 */

#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))) {
		nilfs_msg(dat->i_sb, KERN_CRIT,
			  "%s: invalid vblocknr = %llu, [%llu, %llu)",
			  __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 int 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) {
		nilfs_msg(sb, KERN_ERR, "too large DAT entry size: %zu bytes",
			  entry_size);
		return -EINVAL;
	} else if (entry_size < NILFS_MIN_DAT_ENTRY_SIZE) {
		nilfs_msg(sb, KERN_ERR, "too small DAT entry size: %zu bytes",
			  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	*
4b420ab4	16	* Written by Koji Sato.
a17564f5 KS	17	*/
	18
	19	#include <linux/types.h>
	20	#include <linux/buffer_head.h>
	21	#include <linux/string.h>
	22	#include <linux/errno.h>
	23	#include "nilfs.h"
	24	#include "mdt.h"
	25	#include "alloc.h"
	26	#include "dat.h"
	27
	28
	29	#define NILFS_CNO_MIN ((__u64)1)
	30	#define NILFS_CNO_MAX (~(__u64)0)
	31
f5974c8f VD	32	/**
	33	* struct nilfs_dat_info - on-memory private data of DAT file
	34	* @mi: on-memory private data of metadata file
	35	* @palloc_cache: persistent object allocator cache of DAT file
	36	* @shadow: shadow map of DAT file
	37	*/
8908b2f7 RK	38	struct nilfs_dat_info {
	39	struct nilfs_mdt_info mi;
	40	struct nilfs_palloc_cache palloc_cache;
c1c1d709	41	struct nilfs_shadow_map shadow;
8908b2f7 RK	42	};
	43
	44	static inline struct nilfs_dat_info NILFS_DAT_I(struct inode dat)
	45	{
	46	return (struct nilfs_dat_info *)NILFS_MDT(dat);
	47	}
	48
a17564f5 KS	49	static int nilfs_dat_prepare_entry(struct inode *dat,
	50	struct nilfs_palloc_req *req, int create)
	51	{
	52	return nilfs_palloc_get_entry_block(dat, req->pr_entry_nr,
	53	create, &req->pr_entry_bh);
	54	}
	55
	56	static void nilfs_dat_commit_entry(struct inode *dat,
	57	struct nilfs_palloc_req *req)
	58	{
5fc7b141	59	mark_buffer_dirty(req->pr_entry_bh);
a17564f5 KS	60	nilfs_mdt_mark_dirty(dat);
	61	brelse(req->pr_entry_bh);
	62	}
	63
	64	static void nilfs_dat_abort_entry(struct inode *dat,
	65	struct nilfs_palloc_req *req)
	66	{
	67	brelse(req->pr_entry_bh);
	68	}
	69
	70	int nilfs_dat_prepare_alloc(struct inode dat, struct nilfs_palloc_req req)
	71	{
	72	int ret;
	73
	74	ret = nilfs_palloc_prepare_alloc_entry(dat, req);
	75	if (ret < 0)
	76	return ret;
	77
	78	ret = nilfs_dat_prepare_entry(dat, req, 1);
	79	if (ret < 0)
	80	nilfs_palloc_abort_alloc_entry(dat, req);
	81
	82	return ret;
	83	}
	84
	85	void nilfs_dat_commit_alloc(struct inode dat, struct nilfs_palloc_req req)
	86	{
	87	struct nilfs_dat_entry *entry;
	88	void *kaddr;
	89
7b9c0976	90	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	91	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
	92	req->pr_entry_bh, kaddr);
	93	entry->de_start = cpu_to_le64(NILFS_CNO_MIN);
	94	entry->de_end = cpu_to_le64(NILFS_CNO_MAX);
	95	entry->de_blocknr = cpu_to_le64(0);
7b9c0976	96	kunmap_atomic(kaddr);
a17564f5 KS	97
	98	nilfs_palloc_commit_alloc_entry(dat, req);
	99	nilfs_dat_commit_entry(dat, req);
	100	}
	101
	102	void nilfs_dat_abort_alloc(struct inode dat, struct nilfs_palloc_req req)
	103	{
	104	nilfs_dat_abort_entry(dat, req);
	105	nilfs_palloc_abort_alloc_entry(dat, req);
	106	}
	107
abc0b50b JS	108	static void nilfs_dat_commit_free(struct inode *dat,
abc0b50b JS	109	struct nilfs_palloc_req *req)
a17564f5 KS	110	{
	111	struct nilfs_dat_entry *entry;
	112	void *kaddr;
	113
7b9c0976	114	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	115	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
	116	req->pr_entry_bh, kaddr);
	117	entry->de_start = cpu_to_le64(NILFS_CNO_MIN);
	118	entry->de_end = cpu_to_le64(NILFS_CNO_MIN);
	119	entry->de_blocknr = cpu_to_le64(0);
7b9c0976	120	kunmap_atomic(kaddr);
a17564f5 KS	121
	122	nilfs_dat_commit_entry(dat, req);
	123	nilfs_palloc_commit_free_entry(dat, req);
	124	}
	125
a17564f5 KS	126	int nilfs_dat_prepare_start(struct inode dat, struct nilfs_palloc_req req)
	127	{
	128	int ret;
	129
	130	ret = nilfs_dat_prepare_entry(dat, req, 0);
1f5abe7e	131	WARN_ON(ret == -ENOENT);
a17564f5 KS	132	return ret;
	133	}
	134
	135	void nilfs_dat_commit_start(struct inode dat, struct nilfs_palloc_req req,
	136	sector_t blocknr)
	137	{
	138	struct nilfs_dat_entry *entry;
	139	void *kaddr;
	140
7b9c0976	141	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	142	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
	143	req->pr_entry_bh, kaddr);
	144	entry->de_start = cpu_to_le64(nilfs_mdt_cno(dat));
a17564f5	145	entry->de_blocknr = cpu_to_le64(blocknr);
7b9c0976	146	kunmap_atomic(kaddr);
a17564f5 KS	147
	148	nilfs_dat_commit_entry(dat, req);
	149	}
	150
a17564f5 KS	151	int nilfs_dat_prepare_end(struct inode dat, struct nilfs_palloc_req req)
	152	{
	153	struct nilfs_dat_entry *entry;
a17564f5 KS	154	sector_t blocknr;
	155	void *kaddr;
	156	int ret;
	157
	158	ret = nilfs_dat_prepare_entry(dat, req, 0);
	159	if (ret < 0) {
1f5abe7e	160	WARN_ON(ret == -ENOENT);
a17564f5 KS	161	return ret;
	162	}
	163
7b9c0976	164	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	165	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
a17564f5 KS	166	req->pr_entry_bh, kaddr);
a17564f5	167	blocknr = le64_to_cpu(entry->de_blocknr);
7b9c0976	168	kunmap_atomic(kaddr);
a17564f5 KS	169
	170	if (blocknr == 0) {
	171	ret = nilfs_palloc_prepare_free_entry(dat, req);
	172	if (ret < 0) {
	173	nilfs_dat_abort_entry(dat, req);
	174	return ret;
	175	}
	176	}
	177
	178	return 0;
	179	}
	180
	181	void nilfs_dat_commit_end(struct inode dat, struct nilfs_palloc_req req,
	182	int dead)
	183	{
	184	struct nilfs_dat_entry *entry;
	185	__u64 start, end;
	186	sector_t blocknr;
	187	void *kaddr;
	188
7b9c0976	189	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	190	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
	191	req->pr_entry_bh, kaddr);
	192	end = start = le64_to_cpu(entry->de_start);
	193	if (!dead) {
	194	end = nilfs_mdt_cno(dat);
1f5abe7e	195	WARN_ON(start > end);
a17564f5 KS	196	}
	197	entry->de_end = cpu_to_le64(end);
	198	blocknr = le64_to_cpu(entry->de_blocknr);
7b9c0976	199	kunmap_atomic(kaddr);
a17564f5 KS	200
	201	if (blocknr == 0)
	202	nilfs_dat_commit_free(dat, req);
	203	else
	204	nilfs_dat_commit_entry(dat, req);
	205	}
	206
	207	void nilfs_dat_abort_end(struct inode dat, struct nilfs_palloc_req req)
	208	{
	209	struct nilfs_dat_entry *entry;
	210	__u64 start;
	211	sector_t blocknr;
	212	void *kaddr;
	213
7b9c0976	214	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	215	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
	216	req->pr_entry_bh, kaddr);
	217	start = le64_to_cpu(entry->de_start);
	218	blocknr = le64_to_cpu(entry->de_blocknr);
7b9c0976	219	kunmap_atomic(kaddr);
a17564f5 KS	220
	221	if (start == nilfs_mdt_cno(dat) && blocknr == 0)
	222	nilfs_palloc_abort_free_entry(dat, req);
	223	nilfs_dat_abort_entry(dat, req);
	224	}
	225
bd8169ef RK	226	int nilfs_dat_prepare_update(struct inode *dat,
	227	struct nilfs_palloc_req *oldreq,
	228	struct nilfs_palloc_req *newreq)
	229	{
	230	int ret;
	231
	232	ret = nilfs_dat_prepare_end(dat, oldreq);
	233	if (!ret) {
	234	ret = nilfs_dat_prepare_alloc(dat, newreq);
	235	if (ret < 0)
	236	nilfs_dat_abort_end(dat, oldreq);
	237	}
	238	return ret;
	239	}
	240
	241	void nilfs_dat_commit_update(struct inode *dat,
	242	struct nilfs_palloc_req *oldreq,
	243	struct nilfs_palloc_req *newreq, int dead)
	244	{
	245	nilfs_dat_commit_end(dat, oldreq, dead);
	246	nilfs_dat_commit_alloc(dat, newreq);
	247	}
	248
	249	void nilfs_dat_abort_update(struct inode *dat,
	250	struct nilfs_palloc_req *oldreq,
	251	struct nilfs_palloc_req *newreq)
	252	{
	253	nilfs_dat_abort_end(dat, oldreq);
	254	nilfs_dat_abort_alloc(dat, newreq);
	255	}
	256
a17564f5 KS	257	/**
	258	* nilfs_dat_mark_dirty -
	259	* @dat: DAT file inode
	260	* @vblocknr: virtual block number
	261	*
	262	* Description:
	263	*
	264	* Return Value: On success, 0 is returned. On error, one of the following
	265	* negative error codes is returned.
	266	*
	267	* %-EIO - I/O error.
	268	*
	269	* %-ENOMEM - Insufficient amount of memory available.
	270	*/
	271	int nilfs_dat_mark_dirty(struct inode *dat, __u64 vblocknr)
	272	{
	273	struct nilfs_palloc_req req;
	274	int ret;
	275
	276	req.pr_entry_nr = vblocknr;
	277	ret = nilfs_dat_prepare_entry(dat, &req, 0);
	278	if (ret == 0)
	279	nilfs_dat_commit_entry(dat, &req);
	280	return ret;
	281	}
	282
	283	/**
	284	* nilfs_dat_freev - free virtual block numbers
	285	* @dat: DAT file inode
	286	* @vblocknrs: array of virtual block numbers
	287	* @nitems: number of virtual block numbers
	288	*
	289	* Description: nilfs_dat_freev() frees the virtual block numbers specified by
	290	* @vblocknrs and @nitems.
	291	*
	292	* Return Value: On success, 0 is returned. On error, one of the following
7a65004b	293	* negative error codes is returned.
a17564f5 KS	294	*
	295	* %-EIO - I/O error.
	296	*
	297	* %-ENOMEM - Insufficient amount of memory available.
	298	*
	299	* %-ENOENT - The virtual block number have not been allocated.
	300	*/
	301	int nilfs_dat_freev(struct inode dat, __u64 vblocknrs, size_t nitems)
	302	{
	303	return nilfs_palloc_freev(dat, vblocknrs, nitems);
	304	}
	305
	306	/**
	307	* nilfs_dat_move - change a block number
	308	* @dat: DAT file inode
	309	* @vblocknr: virtual block number
	310	* @blocknr: block number
	311	*
	312	* Description: nilfs_dat_move() changes the block number associated with
	313	* @vblocknr to @blocknr.
	314	*
	315	* Return Value: On success, 0 is returned. On error, one of the following
	316	* negative error codes is returned.
	317	*
	318	* %-EIO - I/O error.
	319	*
	320	* %-ENOMEM - Insufficient amount of memory available.
	321	*/
	322	int nilfs_dat_move(struct inode *dat, __u64 vblocknr, sector_t blocknr)
	323	{
	324	struct buffer_head *entry_bh;
	325	struct nilfs_dat_entry *entry;
	326	void *kaddr;
	327	int ret;
	328
	329	ret = nilfs_palloc_get_entry_block(dat, vblocknr, 0, &entry_bh);
	330	if (ret < 0)
	331	return ret;
c1c1d709 RK	332
	333	/*
	334	* The given disk block number (blocknr) is not yet written to
	335	* the device at this point.
	336	*
	337	* To prevent nilfs_dat_translate() from returning the
f6c26ec5	338	* uncommitted block number, this makes a copy of the entry
c1c1d709 RK	339	* buffer and redirects nilfs_dat_translate() to the copy.
	340	*/
	341	if (!buffer_nilfs_redirected(entry_bh)) {
	342	ret = nilfs_mdt_freeze_buffer(dat, entry_bh);
	343	if (ret) {
	344	brelse(entry_bh);
	345	return ret;
	346	}
	347	}
	348
7b9c0976	349	kaddr = kmap_atomic(entry_bh->b_page);
a17564f5	350	entry = nilfs_palloc_block_get_entry(dat, vblocknr, entry_bh, kaddr);
1f5abe7e	351	if (unlikely(entry->de_blocknr == cpu_to_le64(0))) {
feee880f RK	352	nilfs_msg(dat->i_sb, KERN_CRIT,
	353	"%s: invalid vblocknr = %llu, [%llu, %llu)",
	354	__func__, (unsigned long long)vblocknr,
	355	(unsigned long long)le64_to_cpu(entry->de_start),
	356	(unsigned long long)le64_to_cpu(entry->de_end));
7b9c0976	357	kunmap_atomic(kaddr);
1f5abe7e RK	358	brelse(entry_bh);
1f5abe7e RK	359	return -EINVAL;
a17564f5	360	}
1f5abe7e	361	WARN_ON(blocknr == 0);
a17564f5	362	entry->de_blocknr = cpu_to_le64(blocknr);
7b9c0976	363	kunmap_atomic(kaddr);
a17564f5	364
5fc7b141	365	mark_buffer_dirty(entry_bh);
a17564f5 KS	366	nilfs_mdt_mark_dirty(dat);
	367
	368	brelse(entry_bh);
	369
	370	return 0;
	371	}
	372
	373	/**
	374	* nilfs_dat_translate - translate a virtual block number to a block number
	375	* @dat: DAT file inode
	376	* @vblocknr: virtual block number
	377	* @blocknrp: pointer to a block number
	378	*
	379	* Description: nilfs_dat_translate() maps the virtual block number @vblocknr
	380	* to the corresponding block number.
	381	*
	382	* Return Value: On success, 0 is returned and the block number associated
	383	* with @vblocknr is stored in the place pointed by @blocknrp. On error, one
	384	* of the following negative error codes is returned.
	385	*
	386	* %-EIO - I/O error.
	387	*
	388	* %-ENOMEM - Insufficient amount of memory available.
	389	*
	390	* %-ENOENT - A block number associated with @vblocknr does not exist.
	391	*/
	392	int nilfs_dat_translate(struct inode dat, __u64 vblocknr, sector_t blocknrp)
	393	{
c1c1d709	394	struct buffer_head entry_bh, bh;
a17564f5 KS	395	struct nilfs_dat_entry *entry;
	396	sector_t blocknr;
	397	void *kaddr;
	398	int ret;
	399
	400	ret = nilfs_palloc_get_entry_block(dat, vblocknr, 0, &entry_bh);
	401	if (ret < 0)
	402	return ret;
	403
c1c1d709 RK	404	if (!nilfs_doing_gc() && buffer_nilfs_redirected(entry_bh)) {
	405	bh = nilfs_mdt_get_frozen_buffer(dat, entry_bh);
	406	if (bh) {
	407	WARN_ON(!buffer_uptodate(bh));
	408	brelse(entry_bh);
	409	entry_bh = bh;
	410	}
	411	}
	412
7b9c0976	413	kaddr = kmap_atomic(entry_bh->b_page);
a17564f5 KS	414	entry = nilfs_palloc_block_get_entry(dat, vblocknr, entry_bh, kaddr);
	415	blocknr = le64_to_cpu(entry->de_blocknr);
	416	if (blocknr == 0) {
	417	ret = -ENOENT;
	418	goto out;
	419	}
086d1764	420	*blocknrp = blocknr;
a17564f5 KS	421
a17564f5 KS	422	out:
7b9c0976	423	kunmap_atomic(kaddr);
a17564f5 KS	424	brelse(entry_bh);
	425	return ret;
	426	}
	427
0c6c44cb	428	ssize_t nilfs_dat_get_vinfo(struct inode dat, void buf, unsigned int visz,
a17564f5 KS	429	size_t nvi)
	430	{
	431	struct buffer_head *entry_bh;
	432	struct nilfs_dat_entry *entry;
003ff182	433	struct nilfs_vinfo *vinfo = buf;
a17564f5 KS	434	__u64 first, last;
	435	void *kaddr;
	436	unsigned long entries_per_block = NILFS_MDT(dat)->mi_entries_per_block;
	437	int i, j, n, ret;
	438
	439	for (i = 0; i < nvi; i += n) {
003ff182	440	ret = nilfs_palloc_get_entry_block(dat, vinfo->vi_vblocknr,
a17564f5 KS	441	0, &entry_bh);
	442	if (ret < 0)
	443	return ret;
7b9c0976	444	kaddr = kmap_atomic(entry_bh->b_page);
a17564f5	445	/* last virtual block number in this block */
003ff182	446	first = vinfo->vi_vblocknr;
a17564f5 KS	447	do_div(first, entries_per_block);
	448	first *= entries_per_block;
	449	last = first + entries_per_block - 1;
	450	for (j = i, n = 0;
003ff182 RK	451	j < nvi && vinfo->vi_vblocknr >= first &&
	452	vinfo->vi_vblocknr <= last;
	453	j++, n++, vinfo = (void *)vinfo + visz) {
a17564f5	454	entry = nilfs_palloc_block_get_entry(
003ff182 RK	455	dat, vinfo->vi_vblocknr, entry_bh, kaddr);
	456	vinfo->vi_start = le64_to_cpu(entry->de_start);
	457	vinfo->vi_end = le64_to_cpu(entry->de_end);
	458	vinfo->vi_blocknr = le64_to_cpu(entry->de_blocknr);
a17564f5	459	}
7b9c0976	460	kunmap_atomic(kaddr);
a17564f5 KS	461	brelse(entry_bh);
	462	}
	463
	464	return nvi;
	465	}
79739565	466
8707df38	467	/**
f1e89c86 RK	468	* nilfs_dat_read - read or get dat inode
f1e89c86 RK	469	* @sb: super block instance
79739565	470	* @entry_size: size of a dat entry
f1e89c86 RK	471	* @raw_inode: on-disk dat inode
f1e89c86 RK	472	* @inodep: buffer to store the inode
79739565	473	*/
f1e89c86 RK	474	int nilfs_dat_read(struct super_block *sb, size_t entry_size,
f1e89c86 RK	475	struct nilfs_inode raw_inode, struct inode *inodep)
79739565 RK	476	{
	477	static struct lock_class_key dat_lock_key;
	478	struct inode *dat;
8908b2f7	479	struct nilfs_dat_info *di;
79739565 RK	480	int err;
79739565 RK	481
0ec060d1	482	if (entry_size > sb->s_blocksize) {
feee880f RK	483	nilfs_msg(sb, KERN_ERR, "too large DAT entry size: %zu bytes",
feee880f RK	484	entry_size);
0ec060d1 RK	485	return -EINVAL;
0ec060d1 RK	486	} else if (entry_size < NILFS_MIN_DAT_ENTRY_SIZE) {
feee880f RK	487	nilfs_msg(sb, KERN_ERR, "too small DAT entry size: %zu bytes",
feee880f RK	488	entry_size);
0ec060d1 RK	489	return -EINVAL;
	490	}
	491
f1e89c86 RK	492	dat = nilfs_iget_locked(sb, NULL, NILFS_DAT_INO);
	493	if (unlikely(!dat))
	494	return -ENOMEM;
	495	if (!(dat->i_state & I_NEW))
	496	goto out;
8908b2f7	497
f1e89c86 RK	498	err = nilfs_mdt_init(dat, NILFS_MDT_GFP, sizeof(*di));
	499	if (err)
	500	goto failed;
	501
	502	err = nilfs_palloc_init_blockgroup(dat, entry_size);
	503	if (err)
	504	goto failed;
	505
	506	di = NILFS_DAT_I(dat);
	507	lockdep_set_class(&di->mi.mi_sem, &dat_lock_key);
	508	nilfs_palloc_setup_cache(dat, &di->palloc_cache);
	509	nilfs_mdt_setup_shadow_map(dat, &di->shadow);
	510
	511	err = nilfs_read_inode_common(dat, raw_inode);
	512	if (err)
	513	goto failed;
	514
	515	unlock_new_inode(dat);
	516	out:
	517	*inodep = dat;
	518	return 0;
	519	failed:
	520	iget_failed(dat);
	521	return err;
79739565	522	}