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

// SPDX-License-Identifier: GPL-2.0+
/*
 * NILFS disk address translation.
 *
 * Copyright (C) 2006-2008 Nippon Telegraph and Telephone Corporation.
 *
 * 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)
{
	int ret;

	ret = nilfs_palloc_get_entry_block(dat, req->pr_entry_nr,
					   create, &req->pr_entry_bh);
	if (unlikely(ret == -ENOENT)) {
		nilfs_err(dat->i_sb,
			  "DAT doesn't have a block to manage vblocknr = %llu",
			  (unsigned long long)req->pr_entry_nr);
		/*
		 * Return internal code -EINVAL to notify bmap layer of
		 * metadata corruption.
		 */
		ret = -EINVAL;
	}
	return ret;
}

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);

	if (unlikely(req->pr_desc_bh == NULL || req->pr_bitmap_bh == NULL)) {
		nilfs_error(dat->i_sb,
			    "state inconsistency probably due to duplicate use of vblocknr = %llu",
			    (unsigned long long)req->pr_entry_nr);
		return;
	}
	nilfs_palloc_commit_free_entry(dat, req);
}

int nilfs_dat_prepare_start(struct inode *dat, struct nilfs_palloc_req *req)
{
	return nilfs_dat_prepare_entry(dat, req, 0);
}

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;
	__u64 start;
	sector_t blocknr;
	void *kaddr;
	int ret;

	ret = nilfs_dat_prepare_entry(dat, req, 0);
	if (ret < 0)
		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);
	start = le64_to_cpu(entry->de_start);
	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;
		}
	}
	if (unlikely(start > nilfs_mdt_cno(dat))) {
		nilfs_err(dat->i_sb,
			  "vblocknr = %llu has abnormal lifetime: start cno (= %llu) > current cno (= %llu)",
			  (unsigned long long)req->pr_entry_nr,
			  (unsigned long long)start,
			  (unsigned long long)nilfs_mdt_cno(dat));
		nilfs_dat_abort_entry(dat, req);
		return -EINVAL;
	}

	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_crit(dat->i_sb,
			   "%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_err(sb, "too large DAT entry size: %zu bytes",
			  entry_size);
		return -EINVAL;
	} else if (entry_size < NILFS_MIN_DAT_ENTRY_SIZE) {
		nilfs_err(sb, "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);
	err = nilfs_mdt_setup_shadow_map(dat, &di->shadow);
	if (err)
		goto failed;

	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
ae98043f	1	// SPDX-License-Identifier: GPL-2.0+
a17564f5	2	/*
94ee1d91	3	* NILFS disk address translation.
a17564f5 KS	4	*
	5	* Copyright (C) 2006-2008 Nippon Telegraph and Telephone Corporation.
	6	*
4b420ab4	7	* Written by Koji Sato.
a17564f5 KS	8	*/
	9
	10	#include <linux/types.h>
	11	#include <linux/buffer_head.h>
	12	#include <linux/string.h>
	13	#include <linux/errno.h>
	14	#include "nilfs.h"
	15	#include "mdt.h"
	16	#include "alloc.h"
	17	#include "dat.h"
	18
	19
	20	#define NILFS_CNO_MIN ((__u64)1)
	21	#define NILFS_CNO_MAX (~(__u64)0)
	22
f5974c8f VD	23	/**
	24	* struct nilfs_dat_info - on-memory private data of DAT file
	25	* @mi: on-memory private data of metadata file
	26	* @palloc_cache: persistent object allocator cache of DAT file
	27	* @shadow: shadow map of DAT file
	28	*/
8908b2f7 RK	29	struct nilfs_dat_info {
	30	struct nilfs_mdt_info mi;
	31	struct nilfs_palloc_cache palloc_cache;
c1c1d709	32	struct nilfs_shadow_map shadow;
8908b2f7 RK	33	};
	34
	35	static inline struct nilfs_dat_info NILFS_DAT_I(struct inode dat)
	36	{
	37	return (struct nilfs_dat_info *)NILFS_MDT(dat);
	38	}
	39
a17564f5 KS	40	static int nilfs_dat_prepare_entry(struct inode *dat,
	41	struct nilfs_palloc_req *req, int create)
	42	{
5124a0a5 RK	43	int ret;
	44
	45	ret = nilfs_palloc_get_entry_block(dat, req->pr_entry_nr,
	46	create, &req->pr_entry_bh);
	47	if (unlikely(ret == -ENOENT)) {
	48	nilfs_err(dat->i_sb,
	49	"DAT doesn't have a block to manage vblocknr = %llu",
	50	(unsigned long long)req->pr_entry_nr);
	51	/*
	52	* Return internal code -EINVAL to notify bmap layer of
	53	* metadata corruption.
	54	*/
	55	ret = -EINVAL;
	56	}
	57	return ret;
a17564f5 KS	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
a17564f5 KS	126	nilfs_dat_commit_entry(dat, req);
f0a0ccda Z	127
	128	if (unlikely(req->pr_desc_bh == NULL \|\| req->pr_bitmap_bh == NULL)) {
	129	nilfs_error(dat->i_sb,
	130	"state inconsistency probably due to duplicate use of vblocknr = %llu",
	131	(unsigned long long)req->pr_entry_nr);
	132	return;
	133	}
a17564f5 KS	134	nilfs_palloc_commit_free_entry(dat, req);
	135	}
	136
a17564f5 KS	137	int nilfs_dat_prepare_start(struct inode dat, struct nilfs_palloc_req req)
a17564f5 KS	138	{
5124a0a5	139	return nilfs_dat_prepare_entry(dat, req, 0);
a17564f5 KS	140	}
	141
	142	void nilfs_dat_commit_start(struct inode dat, struct nilfs_palloc_req req,
	143	sector_t blocknr)
	144	{
	145	struct nilfs_dat_entry *entry;
	146	void *kaddr;
	147
7b9c0976	148	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	149	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
	150	req->pr_entry_bh, kaddr);
	151	entry->de_start = cpu_to_le64(nilfs_mdt_cno(dat));
a17564f5	152	entry->de_blocknr = cpu_to_le64(blocknr);
7b9c0976	153	kunmap_atomic(kaddr);
a17564f5 KS	154
	155	nilfs_dat_commit_entry(dat, req);
	156	}
	157
a17564f5 KS	158	int nilfs_dat_prepare_end(struct inode dat, struct nilfs_palloc_req req)
	159	{
	160	struct nilfs_dat_entry *entry;
602ce7b8	161	__u64 start;
a17564f5 KS	162	sector_t blocknr;
	163	void *kaddr;
	164	int ret;
	165
	166	ret = nilfs_dat_prepare_entry(dat, req, 0);
5124a0a5	167	if (ret < 0)
a17564f5	168	return ret;
a17564f5	169
7b9c0976	170	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	171	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
a17564f5 KS	172	req->pr_entry_bh, kaddr);
602ce7b8	173	start = le64_to_cpu(entry->de_start);
a17564f5	174	blocknr = le64_to_cpu(entry->de_blocknr);
7b9c0976	175	kunmap_atomic(kaddr);
a17564f5 KS	176
	177	if (blocknr == 0) {
	178	ret = nilfs_palloc_prepare_free_entry(dat, req);
	179	if (ret < 0) {
	180	nilfs_dat_abort_entry(dat, req);
	181	return ret;
	182	}
	183	}
602ce7b8 RK	184	if (unlikely(start > nilfs_mdt_cno(dat))) {
	185	nilfs_err(dat->i_sb,
	186	"vblocknr = %llu has abnormal lifetime: start cno (= %llu) > current cno (= %llu)",
	187	(unsigned long long)req->pr_entry_nr,
	188	(unsigned long long)start,
	189	(unsigned long long)nilfs_mdt_cno(dat));
	190	nilfs_dat_abort_entry(dat, req);
	191	return -EINVAL;
	192	}
a17564f5 KS	193
	194	return 0;
	195	}
	196
	197	void nilfs_dat_commit_end(struct inode dat, struct nilfs_palloc_req req,
	198	int dead)
	199	{
	200	struct nilfs_dat_entry *entry;
	201	__u64 start, end;
	202	sector_t blocknr;
	203	void *kaddr;
	204
7b9c0976	205	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	206	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
	207	req->pr_entry_bh, kaddr);
	208	end = start = le64_to_cpu(entry->de_start);
	209	if (!dead) {
	210	end = nilfs_mdt_cno(dat);
1f5abe7e	211	WARN_ON(start > end);
a17564f5 KS	212	}
	213	entry->de_end = cpu_to_le64(end);
	214	blocknr = le64_to_cpu(entry->de_blocknr);
7b9c0976	215	kunmap_atomic(kaddr);
a17564f5 KS	216
	217	if (blocknr == 0)
	218	nilfs_dat_commit_free(dat, req);
	219	else
	220	nilfs_dat_commit_entry(dat, req);
	221	}
	222
	223	void nilfs_dat_abort_end(struct inode dat, struct nilfs_palloc_req req)
	224	{
	225	struct nilfs_dat_entry *entry;
	226	__u64 start;
	227	sector_t blocknr;
	228	void *kaddr;
	229
7b9c0976	230	kaddr = kmap_atomic(req->pr_entry_bh->b_page);
a17564f5 KS	231	entry = nilfs_palloc_block_get_entry(dat, req->pr_entry_nr,
	232	req->pr_entry_bh, kaddr);
	233	start = le64_to_cpu(entry->de_start);
	234	blocknr = le64_to_cpu(entry->de_blocknr);
7b9c0976	235	kunmap_atomic(kaddr);
a17564f5 KS	236
	237	if (start == nilfs_mdt_cno(dat) && blocknr == 0)
	238	nilfs_palloc_abort_free_entry(dat, req);
	239	nilfs_dat_abort_entry(dat, req);
	240	}
	241
bd8169ef RK	242	int nilfs_dat_prepare_update(struct inode *dat,
	243	struct nilfs_palloc_req *oldreq,
	244	struct nilfs_palloc_req *newreq)
	245	{
	246	int ret;
	247
	248	ret = nilfs_dat_prepare_end(dat, oldreq);
	249	if (!ret) {
	250	ret = nilfs_dat_prepare_alloc(dat, newreq);
	251	if (ret < 0)
	252	nilfs_dat_abort_end(dat, oldreq);
	253	}
	254	return ret;
	255	}
	256
	257	void nilfs_dat_commit_update(struct inode *dat,
	258	struct nilfs_palloc_req *oldreq,
	259	struct nilfs_palloc_req *newreq, int dead)
	260	{
	261	nilfs_dat_commit_end(dat, oldreq, dead);
	262	nilfs_dat_commit_alloc(dat, newreq);
	263	}
	264
	265	void nilfs_dat_abort_update(struct inode *dat,
	266	struct nilfs_palloc_req *oldreq,
	267	struct nilfs_palloc_req *newreq)
	268	{
	269	nilfs_dat_abort_end(dat, oldreq);
	270	nilfs_dat_abort_alloc(dat, newreq);
	271	}
	272
a17564f5 KS	273	/**
	274	* nilfs_dat_mark_dirty -
	275	* @dat: DAT file inode
	276	* @vblocknr: virtual block number
	277	*
	278	* Description:
	279	*
	280	* Return Value: On success, 0 is returned. On error, one of the following
	281	* negative error codes is returned.
	282	*
	283	* %-EIO - I/O error.
	284	*
	285	* %-ENOMEM - Insufficient amount of memory available.
	286	*/
	287	int nilfs_dat_mark_dirty(struct inode *dat, __u64 vblocknr)
	288	{
	289	struct nilfs_palloc_req req;
	290	int ret;
	291
	292	req.pr_entry_nr = vblocknr;
	293	ret = nilfs_dat_prepare_entry(dat, &req, 0);
	294	if (ret == 0)
	295	nilfs_dat_commit_entry(dat, &req);
	296	return ret;
	297	}
	298
	299	/**
	300	* nilfs_dat_freev - free virtual block numbers
	301	* @dat: DAT file inode
	302	* @vblocknrs: array of virtual block numbers
	303	* @nitems: number of virtual block numbers
	304	*
	305	* Description: nilfs_dat_freev() frees the virtual block numbers specified by
	306	* @vblocknrs and @nitems.
	307	*
	308	* Return Value: On success, 0 is returned. On error, one of the following
7a65004b	309	* negative error codes is returned.
a17564f5 KS	310	*
	311	* %-EIO - I/O error.
	312	*
	313	* %-ENOMEM - Insufficient amount of memory available.
	314	*
	315	* %-ENOENT - The virtual block number have not been allocated.
	316	*/
	317	int nilfs_dat_freev(struct inode dat, __u64 vblocknrs, size_t nitems)
	318	{
	319	return nilfs_palloc_freev(dat, vblocknrs, nitems);
	320	}
	321
	322	/**
	323	* nilfs_dat_move - change a block number
	324	* @dat: DAT file inode
	325	* @vblocknr: virtual block number
	326	* @blocknr: block number
	327	*
	328	* Description: nilfs_dat_move() changes the block number associated with
	329	* @vblocknr to @blocknr.
	330	*
	331	* Return Value: On success, 0 is returned. On error, one of the following
	332	* negative error codes is returned.
	333	*
	334	* %-EIO - I/O error.
	335	*
	336	* %-ENOMEM - Insufficient amount of memory available.
	337	*/
	338	int nilfs_dat_move(struct inode *dat, __u64 vblocknr, sector_t blocknr)
	339	{
	340	struct buffer_head *entry_bh;
	341	struct nilfs_dat_entry *entry;
	342	void *kaddr;
	343	int ret;
	344
	345	ret = nilfs_palloc_get_entry_block(dat, vblocknr, 0, &entry_bh);
	346	if (ret < 0)
	347	return ret;
c1c1d709 RK	348
	349	/*
	350	* The given disk block number (blocknr) is not yet written to
	351	* the device at this point.
	352	*
	353	* To prevent nilfs_dat_translate() from returning the
f6c26ec5	354	* uncommitted block number, this makes a copy of the entry
c1c1d709 RK	355	* buffer and redirects nilfs_dat_translate() to the copy.
	356	*/
	357	if (!buffer_nilfs_redirected(entry_bh)) {
	358	ret = nilfs_mdt_freeze_buffer(dat, entry_bh);
	359	if (ret) {
	360	brelse(entry_bh);
	361	return ret;
	362	}
	363	}
	364
7b9c0976	365	kaddr = kmap_atomic(entry_bh->b_page);
a17564f5	366	entry = nilfs_palloc_block_get_entry(dat, vblocknr, entry_bh, kaddr);
1f5abe7e	367	if (unlikely(entry->de_blocknr == cpu_to_le64(0))) {
a1d0747a JP	368	nilfs_crit(dat->i_sb,
	369	"%s: invalid vblocknr = %llu, [%llu, %llu)",
	370	__func__, (unsigned long long)vblocknr,
	371	(unsigned long long)le64_to_cpu(entry->de_start),
	372	(unsigned long long)le64_to_cpu(entry->de_end));
7b9c0976	373	kunmap_atomic(kaddr);
1f5abe7e RK	374	brelse(entry_bh);
1f5abe7e RK	375	return -EINVAL;
a17564f5	376	}
1f5abe7e	377	WARN_ON(blocknr == 0);
a17564f5	378	entry->de_blocknr = cpu_to_le64(blocknr);
7b9c0976	379	kunmap_atomic(kaddr);
a17564f5	380
5fc7b141	381	mark_buffer_dirty(entry_bh);
a17564f5 KS	382	nilfs_mdt_mark_dirty(dat);
	383
	384	brelse(entry_bh);
	385
	386	return 0;
	387	}
	388
	389	/**
	390	* nilfs_dat_translate - translate a virtual block number to a block number
	391	* @dat: DAT file inode
	392	* @vblocknr: virtual block number
	393	* @blocknrp: pointer to a block number
	394	*
	395	* Description: nilfs_dat_translate() maps the virtual block number @vblocknr
	396	* to the corresponding block number.
	397	*
	398	* Return Value: On success, 0 is returned and the block number associated
	399	* with @vblocknr is stored in the place pointed by @blocknrp. On error, one
	400	* of the following negative error codes is returned.
	401	*
	402	* %-EIO - I/O error.
	403	*
	404	* %-ENOMEM - Insufficient amount of memory available.
	405	*
	406	* %-ENOENT - A block number associated with @vblocknr does not exist.
	407	*/
	408	int nilfs_dat_translate(struct inode dat, __u64 vblocknr, sector_t blocknrp)
	409	{
c1c1d709	410	struct buffer_head entry_bh, bh;
a17564f5 KS	411	struct nilfs_dat_entry *entry;
	412	sector_t blocknr;
	413	void *kaddr;
	414	int ret;
	415
	416	ret = nilfs_palloc_get_entry_block(dat, vblocknr, 0, &entry_bh);
	417	if (ret < 0)
	418	return ret;
	419
c1c1d709 RK	420	if (!nilfs_doing_gc() && buffer_nilfs_redirected(entry_bh)) {
	421	bh = nilfs_mdt_get_frozen_buffer(dat, entry_bh);
	422	if (bh) {
	423	WARN_ON(!buffer_uptodate(bh));
	424	brelse(entry_bh);
	425	entry_bh = bh;
	426	}
	427	}
	428
7b9c0976	429	kaddr = kmap_atomic(entry_bh->b_page);
a17564f5 KS	430	entry = nilfs_palloc_block_get_entry(dat, vblocknr, entry_bh, kaddr);
	431	blocknr = le64_to_cpu(entry->de_blocknr);
	432	if (blocknr == 0) {
	433	ret = -ENOENT;
	434	goto out;
	435	}
086d1764	436	*blocknrp = blocknr;
a17564f5 KS	437
a17564f5 KS	438	out:
7b9c0976	439	kunmap_atomic(kaddr);
a17564f5 KS	440	brelse(entry_bh);
	441	return ret;
	442	}
	443
0c6c44cb	444	ssize_t nilfs_dat_get_vinfo(struct inode dat, void buf, unsigned int visz,
a17564f5 KS	445	size_t nvi)
	446	{
	447	struct buffer_head *entry_bh;
	448	struct nilfs_dat_entry *entry;
003ff182	449	struct nilfs_vinfo *vinfo = buf;
a17564f5 KS	450	__u64 first, last;
	451	void *kaddr;
	452	unsigned long entries_per_block = NILFS_MDT(dat)->mi_entries_per_block;
	453	int i, j, n, ret;
	454
	455	for (i = 0; i < nvi; i += n) {
003ff182	456	ret = nilfs_palloc_get_entry_block(dat, vinfo->vi_vblocknr,
a17564f5 KS	457	0, &entry_bh);
	458	if (ret < 0)
	459	return ret;
7b9c0976	460	kaddr = kmap_atomic(entry_bh->b_page);
a17564f5	461	/* last virtual block number in this block */
003ff182	462	first = vinfo->vi_vblocknr;
a17564f5 KS	463	do_div(first, entries_per_block);
	464	first *= entries_per_block;
	465	last = first + entries_per_block - 1;
	466	for (j = i, n = 0;
003ff182 RK	467	j < nvi && vinfo->vi_vblocknr >= first &&
	468	vinfo->vi_vblocknr <= last;
	469	j++, n++, vinfo = (void *)vinfo + visz) {
a17564f5	470	entry = nilfs_palloc_block_get_entry(
003ff182 RK	471	dat, vinfo->vi_vblocknr, entry_bh, kaddr);
	472	vinfo->vi_start = le64_to_cpu(entry->de_start);
	473	vinfo->vi_end = le64_to_cpu(entry->de_end);
	474	vinfo->vi_blocknr = le64_to_cpu(entry->de_blocknr);
a17564f5	475	}
7b9c0976	476	kunmap_atomic(kaddr);
a17564f5 KS	477	brelse(entry_bh);
	478	}
	479
	480	return nvi;
	481	}
79739565	482
8707df38	483	/**
f1e89c86 RK	484	* nilfs_dat_read - read or get dat inode
f1e89c86 RK	485	* @sb: super block instance
79739565	486	* @entry_size: size of a dat entry
f1e89c86 RK	487	* @raw_inode: on-disk dat inode
f1e89c86 RK	488	* @inodep: buffer to store the inode
79739565	489	*/
f1e89c86 RK	490	int nilfs_dat_read(struct super_block *sb, size_t entry_size,
f1e89c86 RK	491	struct nilfs_inode raw_inode, struct inode *inodep)
79739565 RK	492	{
	493	static struct lock_class_key dat_lock_key;
	494	struct inode *dat;
8908b2f7	495	struct nilfs_dat_info *di;
79739565 RK	496	int err;
79739565 RK	497
0ec060d1	498	if (entry_size > sb->s_blocksize) {
a1d0747a	499	nilfs_err(sb, "too large DAT entry size: %zu bytes",
feee880f	500	entry_size);
0ec060d1 RK	501	return -EINVAL;
0ec060d1 RK	502	} else if (entry_size < NILFS_MIN_DAT_ENTRY_SIZE) {
a1d0747a	503	nilfs_err(sb, "too small DAT entry size: %zu bytes",
feee880f	504	entry_size);
0ec060d1 RK	505	return -EINVAL;
	506	}
	507
f1e89c86 RK	508	dat = nilfs_iget_locked(sb, NULL, NILFS_DAT_INO);
	509	if (unlikely(!dat))
	510	return -ENOMEM;
	511	if (!(dat->i_state & I_NEW))
	512	goto out;
8908b2f7	513
f1e89c86 RK	514	err = nilfs_mdt_init(dat, NILFS_MDT_GFP, sizeof(*di));
	515	if (err)
	516	goto failed;
	517
	518	err = nilfs_palloc_init_blockgroup(dat, entry_size);
	519	if (err)
	520	goto failed;
	521
	522	di = NILFS_DAT_I(dat);
	523	lockdep_set_class(&di->mi.mi_sem, &dat_lock_key);
	524	nilfs_palloc_setup_cache(dat, &di->palloc_cache);
6e211930 RK	525	err = nilfs_mdt_setup_shadow_map(dat, &di->shadow);
	526	if (err)
	527	goto failed;
f1e89c86 RK	528
	529	err = nilfs_read_inode_common(dat, raw_inode);
	530	if (err)
	531	goto failed;
	532
	533	unlock_new_inode(dat);
	534	out:
	535	*inodep = dat;
	536	return 0;
	537	failed:
	538	iget_failed(dat);
	539	return err;
79739565	540	}