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

/*
 * segbuf.c - NILFS segment buffer
 *
 * Copyright (C) 2005-2008 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 Ryusuke Konishi <ryusuke@osrg.net>
 *
 */

#include <linux/buffer_head.h>
#include <linux/writeback.h>
#include <linux/crc32.h>
#include <linux/backing-dev.h>
#include "page.h"
#include "segbuf.h"


struct nilfs_write_info {
	struct the_nilfs       *nilfs;
	struct bio	       *bio;
	int			start, end; /* The region to be submitted */
	int			rest_blocks;
	int			max_pages;
	int			nr_vecs;
	sector_t		blocknr;
};


static int nilfs_segbuf_write(struct nilfs_segment_buffer *segbuf,
			      struct the_nilfs *nilfs);
static int nilfs_segbuf_wait(struct nilfs_segment_buffer *segbuf);


static struct kmem_cache *nilfs_segbuf_cachep;

static void nilfs_segbuf_init_once(void *obj)
{
	memset(obj, 0, sizeof(struct nilfs_segment_buffer));
}

int __init nilfs_init_segbuf_cache(void)
{
	nilfs_segbuf_cachep =
		kmem_cache_create("nilfs2_segbuf_cache",
				  sizeof(struct nilfs_segment_buffer),
				  0, SLAB_RECLAIM_ACCOUNT,
				  nilfs_segbuf_init_once);

	return (nilfs_segbuf_cachep == NULL) ? -ENOMEM : 0;
}

void nilfs_destroy_segbuf_cache(void)
{
	kmem_cache_destroy(nilfs_segbuf_cachep);
}

struct nilfs_segment_buffer *nilfs_segbuf_new(struct super_block *sb)
{
	struct nilfs_segment_buffer *segbuf;

	segbuf = kmem_cache_alloc(nilfs_segbuf_cachep, GFP_NOFS);
	if (unlikely(!segbuf))
		return NULL;

	segbuf->sb_super = sb;
	INIT_LIST_HEAD(&segbuf->sb_list);
	INIT_LIST_HEAD(&segbuf->sb_segsum_buffers);
	INIT_LIST_HEAD(&segbuf->sb_payload_buffers);

	init_completion(&segbuf->sb_bio_event);
	atomic_set(&segbuf->sb_err, 0);
	segbuf->sb_nbio = 0;

	return segbuf;
}

void nilfs_segbuf_free(struct nilfs_segment_buffer *segbuf)
{
	kmem_cache_free(nilfs_segbuf_cachep, segbuf);
}

void nilfs_segbuf_map(struct nilfs_segment_buffer *segbuf, __u64 segnum,
		     unsigned long offset, struct the_nilfs *nilfs)
{
	segbuf->sb_segnum = segnum;
	nilfs_get_segment_range(nilfs, segnum, &segbuf->sb_fseg_start,
				&segbuf->sb_fseg_end);

	segbuf->sb_pseg_start = segbuf->sb_fseg_start + offset;
	segbuf->sb_rest_blocks =
		segbuf->sb_fseg_end - segbuf->sb_pseg_start + 1;
}

/**
 * nilfs_segbuf_map_cont - map a new log behind a given log
 * @segbuf: new segment buffer
 * @prev: segment buffer containing a log to be continued
 */
void nilfs_segbuf_map_cont(struct nilfs_segment_buffer *segbuf,
			   struct nilfs_segment_buffer *prev)
{
	segbuf->sb_segnum = prev->sb_segnum;
	segbuf->sb_fseg_start = prev->sb_fseg_start;
	segbuf->sb_fseg_end = prev->sb_fseg_end;
	segbuf->sb_pseg_start = prev->sb_pseg_start + prev->sb_sum.nblocks;
	segbuf->sb_rest_blocks =
		segbuf->sb_fseg_end - segbuf->sb_pseg_start + 1;
}

void nilfs_segbuf_set_next_segnum(struct nilfs_segment_buffer *segbuf,
				  __u64 nextnum, struct the_nilfs *nilfs)
{
	segbuf->sb_nextnum = nextnum;
	segbuf->sb_sum.next = nilfs_get_segment_start_blocknr(nilfs, nextnum);
}

int nilfs_segbuf_extend_segsum(struct nilfs_segment_buffer *segbuf)
{
	struct buffer_head *bh;

	bh = sb_getblk(segbuf->sb_super,
		       segbuf->sb_pseg_start + segbuf->sb_sum.nsumblk);
	if (unlikely(!bh))
		return -ENOMEM;

	nilfs_segbuf_add_segsum_buffer(segbuf, bh);
	return 0;
}

int nilfs_segbuf_extend_payload(struct nilfs_segment_buffer *segbuf,
				struct buffer_head **bhp)
{
	struct buffer_head *bh;

	bh = sb_getblk(segbuf->sb_super,
		       segbuf->sb_pseg_start + segbuf->sb_sum.nblocks);
	if (unlikely(!bh))
		return -ENOMEM;

	nilfs_segbuf_add_payload_buffer(segbuf, bh);
	*bhp = bh;
	return 0;
}

int nilfs_segbuf_reset(struct nilfs_segment_buffer *segbuf, unsigned flags,
		       time_t ctime)
{
	int err;

	segbuf->sb_sum.nblocks = segbuf->sb_sum.nsumblk = 0;
	err = nilfs_segbuf_extend_segsum(segbuf);
	if (unlikely(err))
		return err;

	segbuf->sb_sum.flags = flags;
	segbuf->sb_sum.sumbytes = sizeof(struct nilfs_segment_summary);
	segbuf->sb_sum.nfinfo = segbuf->sb_sum.nfileblk = 0;
	segbuf->sb_sum.ctime = ctime;
	return 0;
}

/*
 * Setup segment summary
 */
void nilfs_segbuf_fill_in_segsum(struct nilfs_segment_buffer *segbuf)
{
	struct nilfs_segment_summary *raw_sum;
	struct buffer_head *bh_sum;

	bh_sum = list_entry(segbuf->sb_segsum_buffers.next,
			    struct buffer_head, b_assoc_buffers);
	raw_sum = (struct nilfs_segment_summary *)bh_sum->b_data;

	raw_sum->ss_magic    = cpu_to_le32(NILFS_SEGSUM_MAGIC);
	raw_sum->ss_bytes    = cpu_to_le16(sizeof(*raw_sum));
	raw_sum->ss_flags    = cpu_to_le16(segbuf->sb_sum.flags);
	raw_sum->ss_seq      = cpu_to_le64(segbuf->sb_sum.seg_seq);
	raw_sum->ss_create   = cpu_to_le64(segbuf->sb_sum.ctime);
	raw_sum->ss_next     = cpu_to_le64(segbuf->sb_sum.next);
	raw_sum->ss_nblocks  = cpu_to_le32(segbuf->sb_sum.nblocks);
	raw_sum->ss_nfinfo   = cpu_to_le32(segbuf->sb_sum.nfinfo);
	raw_sum->ss_sumbytes = cpu_to_le32(segbuf->sb_sum.sumbytes);
	raw_sum->ss_pad      = 0;
}

/*
 * CRC calculation routines
 */
void nilfs_segbuf_fill_in_segsum_crc(struct nilfs_segment_buffer *segbuf,
				     u32 seed)
{
	struct buffer_head *bh;
	struct nilfs_segment_summary *raw_sum;
	unsigned long size, bytes = segbuf->sb_sum.sumbytes;
	u32 crc;

	bh = list_entry(segbuf->sb_segsum_buffers.next, struct buffer_head,
			b_assoc_buffers);

	raw_sum = (struct nilfs_segment_summary *)bh->b_data;
	size = min_t(unsigned long, bytes, bh->b_size);
	crc = crc32_le(seed,
		       (unsigned char *)raw_sum +
		       sizeof(raw_sum->ss_datasum) + sizeof(raw_sum->ss_sumsum),
		       size - (sizeof(raw_sum->ss_datasum) +
			       sizeof(raw_sum->ss_sumsum)));

	list_for_each_entry_continue(bh, &segbuf->sb_segsum_buffers,
				     b_assoc_buffers) {
		bytes -= size;
		size = min_t(unsigned long, bytes, bh->b_size);
		crc = crc32_le(crc, bh->b_data, size);
	}
	raw_sum->ss_sumsum = cpu_to_le32(crc);
}

void nilfs_segbuf_fill_in_data_crc(struct nilfs_segment_buffer *segbuf,
				   u32 seed)
{
	struct buffer_head *bh;
	struct nilfs_segment_summary *raw_sum;
	void *kaddr;
	u32 crc;

	bh = list_entry(segbuf->sb_segsum_buffers.next, struct buffer_head,
			b_assoc_buffers);
	raw_sum = (struct nilfs_segment_summary *)bh->b_data;
	crc = crc32_le(seed,
		       (unsigned char *)raw_sum + sizeof(raw_sum->ss_datasum),
		       bh->b_size - sizeof(raw_sum->ss_datasum));

	list_for_each_entry_continue(bh, &segbuf->sb_segsum_buffers,
				     b_assoc_buffers) {
		crc = crc32_le(crc, bh->b_data, bh->b_size);
	}
	list_for_each_entry(bh, &segbuf->sb_payload_buffers, b_assoc_buffers) {
		kaddr = kmap_atomic(bh->b_page, KM_USER0);
		crc = crc32_le(crc, kaddr + bh_offset(bh), bh->b_size);
		kunmap_atomic(kaddr, KM_USER0);
	}
	raw_sum->ss_datasum = cpu_to_le32(crc);
}

static void nilfs_release_buffers(struct list_head *list)
{
	struct buffer_head *bh, *n;

	list_for_each_entry_safe(bh, n, list, b_assoc_buffers) {
		list_del_init(&bh->b_assoc_buffers);
		if (buffer_nilfs_allocated(bh)) {
			struct page *clone_page = bh->b_page;

			/* remove clone page */
			brelse(bh);
			page_cache_release(clone_page); /* for each bh */
			if (page_count(clone_page) <= 2) {
				lock_page(clone_page);
				nilfs_free_private_page(clone_page);
			}
			continue;
		}
		brelse(bh);
	}
}

static void nilfs_segbuf_clear(struct nilfs_segment_buffer *segbuf)
{
	nilfs_release_buffers(&segbuf->sb_segsum_buffers);
	nilfs_release_buffers(&segbuf->sb_payload_buffers);
}

/*
 * Iterators for segment buffers
 */
void nilfs_clear_logs(struct list_head *logs)
{
	struct nilfs_segment_buffer *segbuf;

	list_for_each_entry(segbuf, logs, sb_list)
		nilfs_segbuf_clear(segbuf);
}

void nilfs_truncate_logs(struct list_head *logs,
			 struct nilfs_segment_buffer *last)
{
	struct nilfs_segment_buffer *n, *segbuf;

	segbuf = list_prepare_entry(last, logs, sb_list);
	list_for_each_entry_safe_continue(segbuf, n, logs, sb_list) {
		list_del_init(&segbuf->sb_list);
		nilfs_segbuf_clear(segbuf);
		nilfs_segbuf_free(segbuf);
	}
}

int nilfs_write_logs(struct list_head *logs, struct the_nilfs *nilfs)
{
	struct nilfs_segment_buffer *segbuf;
	int ret = 0;

	list_for_each_entry(segbuf, logs, sb_list) {
		ret = nilfs_segbuf_write(segbuf, nilfs);
		if (ret)
			break;
	}
	return ret;
}

int nilfs_wait_on_logs(struct list_head *logs)
{
	struct nilfs_segment_buffer *segbuf;
	int err;

	list_for_each_entry(segbuf, logs, sb_list) {
		err = nilfs_segbuf_wait(segbuf);
		if (err)
			return err;
	}
	return 0;
}

/*
 * BIO operations
 */
static void nilfs_end_bio_write(struct bio *bio, int err)
{
	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	struct nilfs_segment_buffer *segbuf = bio->bi_private;

	if (err == -EOPNOTSUPP) {
		set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
		bio_put(bio);
		/* to be detected by submit_seg_bio() */
	}

	if (!uptodate)
		atomic_inc(&segbuf->sb_err);

	bio_put(bio);
	complete(&segbuf->sb_bio_event);
}

static int nilfs_segbuf_submit_bio(struct nilfs_segment_buffer *segbuf,
				   struct nilfs_write_info *wi, int mode)
{
	struct bio *bio = wi->bio;
	int err;

	if (segbuf->sb_nbio > 0 && bdi_write_congested(wi->nilfs->ns_bdi)) {
		wait_for_completion(&segbuf->sb_bio_event);
		segbuf->sb_nbio--;
		if (unlikely(atomic_read(&segbuf->sb_err))) {
			bio_put(bio);
			err = -EIO;
			goto failed;
		}
	}

	bio->bi_end_io = nilfs_end_bio_write;
	bio->bi_private = segbuf;
	bio_get(bio);
	submit_bio(mode, bio);
	if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
		bio_put(bio);
		err = -EOPNOTSUPP;
		goto failed;
	}
	segbuf->sb_nbio++;
	bio_put(bio);

	wi->bio = NULL;
	wi->rest_blocks -= wi->end - wi->start;
	wi->nr_vecs = min(wi->max_pages, wi->rest_blocks);
	wi->start = wi->end;
	return 0;

 failed:
	wi->bio = NULL;
	return err;
}

/**
 * nilfs_alloc_seg_bio - allocate a new bio for writing log
 * @nilfs: nilfs object
 * @start: start block number of the bio
 * @nr_vecs: request size of page vector.
 *
 * Return Value: On success, pointer to the struct bio is returned.
 * On error, NULL is returned.
 */
static struct bio *nilfs_alloc_seg_bio(struct the_nilfs *nilfs, sector_t start,
				       int nr_vecs)
{
	struct bio *bio;

	bio = bio_alloc(GFP_NOIO, nr_vecs);
	if (bio == NULL) {
		while (!bio && (nr_vecs >>= 1))
			bio = bio_alloc(GFP_NOIO, nr_vecs);
	}
	if (likely(bio)) {
		bio->bi_bdev = nilfs->ns_bdev;
		bio->bi_sector = start << (nilfs->ns_blocksize_bits - 9);
	}
	return bio;
}

static void nilfs_segbuf_prepare_write(struct nilfs_segment_buffer *segbuf,
				       struct nilfs_write_info *wi)
{
	wi->bio = NULL;
	wi->rest_blocks = segbuf->sb_sum.nblocks;
	wi->max_pages = bio_get_nr_vecs(wi->nilfs->ns_bdev);
	wi->nr_vecs = min(wi->max_pages, wi->rest_blocks);
	wi->start = wi->end = 0;
	wi->blocknr = segbuf->sb_pseg_start;
}

static int nilfs_segbuf_submit_bh(struct nilfs_segment_buffer *segbuf,
				  struct nilfs_write_info *wi,
				  struct buffer_head *bh, int mode)
{
	int len, err;

	BUG_ON(wi->nr_vecs <= 0);
 repeat:
	if (!wi->bio) {
		wi->bio = nilfs_alloc_seg_bio(wi->nilfs, wi->blocknr + wi->end,
					      wi->nr_vecs);
		if (unlikely(!wi->bio))
			return -ENOMEM;
	}

	len = bio_add_page(wi->bio, bh->b_page, bh->b_size, bh_offset(bh));
	if (len == bh->b_size) {
		wi->end++;
		return 0;
	}
	/* bio is FULL */
	err = nilfs_segbuf_submit_bio(segbuf, wi, mode);
	/* never submit current bh */
	if (likely(!err))
		goto repeat;
	return err;
}

/**
 * nilfs_segbuf_write - submit write requests of a log
 * @segbuf: buffer storing a log to be written
 * @nilfs: nilfs object
 *
 * Return Value: On Success, 0 is returned. On Error, one of the following
 * negative error code is returned.
 *
 * %-EIO - I/O error
 *
 * %-ENOMEM - Insufficient memory available.
 */
static int nilfs_segbuf_write(struct nilfs_segment_buffer *segbuf,
			      struct the_nilfs *nilfs)
{
	struct nilfs_write_info wi;
	struct buffer_head *bh;
	int res = 0, rw = WRITE;

	wi.nilfs = nilfs;
	nilfs_segbuf_prepare_write(segbuf, &wi);

	list_for_each_entry(bh, &segbuf->sb_segsum_buffers, b_assoc_buffers) {
		res = nilfs_segbuf_submit_bh(segbuf, &wi, bh, rw);
		if (unlikely(res))
			goto failed_bio;
	}

	list_for_each_entry(bh, &segbuf->sb_payload_buffers, b_assoc_buffers) {
		res = nilfs_segbuf_submit_bh(segbuf, &wi, bh, rw);
		if (unlikely(res))
			goto failed_bio;
	}

	if (wi.bio) {
		/*
		 * Last BIO is always sent through the following
		 * submission.
		 */
		rw |= (1 << BIO_RW_SYNCIO) | (1 << BIO_RW_UNPLUG);
		res = nilfs_segbuf_submit_bio(segbuf, &wi, rw);
	}

 failed_bio:
	return res;
}

/**
 * nilfs_segbuf_wait - wait for completion of requested BIOs
 * @segbuf: segment buffer
 *
 * Return Value: On Success, 0 is returned. On Error, one of the following
 * negative error code is returned.
 *
 * %-EIO - I/O error
 */
static int nilfs_segbuf_wait(struct nilfs_segment_buffer *segbuf)
{
	int err = 0;

	if (!segbuf->sb_nbio)
		return 0;

	do {
		wait_for_completion(&segbuf->sb_bio_event);
	} while (--segbuf->sb_nbio > 0);

	if (unlikely(atomic_read(&segbuf->sb_err) > 0)) {
		printk(KERN_ERR "NILFS: IO error writing segment\n");
		err = -EIO;
	}
	return err;
}
Commit	Line	Data
64b5a32e RK	1	/*
	2	* segbuf.c - NILFS segment buffer
	3	*
	4	* Copyright (C) 2005-2008 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 Ryusuke Konishi <ryusuke@osrg.net>
	21	*
	22	*/
	23
	24	#include <linux/buffer_head.h>
	25	#include <linux/writeback.h>
	26	#include <linux/crc32.h>
9c965bac	27	#include <linux/backing-dev.h>
64b5a32e RK	28	#include "page.h"
64b5a32e RK	29	#include "segbuf.h"
64b5a32e RK	30
64b5a32e RK	31
9c965bac RK	32	struct nilfs_write_info {
	33	struct the_nilfs *nilfs;
	34	struct bio *bio;
55480a06	35	int start, end; /* The region to be submitted */
9c965bac RK	36	int rest_blocks;
	37	int max_pages;
	38	int nr_vecs;
	39	sector_t blocknr;
	40	};
	41
	42
d1c6b72a RK	43	static int nilfs_segbuf_write(struct nilfs_segment_buffer *segbuf,
	44	struct the_nilfs *nilfs);
	45	static int nilfs_segbuf_wait(struct nilfs_segment_buffer *segbuf);
	46
	47
64b5a32e RK	48	static struct kmem_cache *nilfs_segbuf_cachep;
	49
	50	static void nilfs_segbuf_init_once(void *obj)
	51	{
	52	memset(obj, 0, sizeof(struct nilfs_segment_buffer));
	53	}
	54
	55	int __init nilfs_init_segbuf_cache(void)
	56	{
	57	nilfs_segbuf_cachep =
	58	kmem_cache_create("nilfs2_segbuf_cache",
	59	sizeof(struct nilfs_segment_buffer),
	60	0, SLAB_RECLAIM_ACCOUNT,
	61	nilfs_segbuf_init_once);
	62
	63	return (nilfs_segbuf_cachep == NULL) ? -ENOMEM : 0;
	64	}
	65
	66	void nilfs_destroy_segbuf_cache(void)
	67	{
	68	kmem_cache_destroy(nilfs_segbuf_cachep);
	69	}
	70
	71	struct nilfs_segment_buffer nilfs_segbuf_new(struct super_block sb)
	72	{
	73	struct nilfs_segment_buffer *segbuf;
	74
	75	segbuf = kmem_cache_alloc(nilfs_segbuf_cachep, GFP_NOFS);
	76	if (unlikely(!segbuf))
	77	return NULL;
	78
	79	segbuf->sb_super = sb;
	80	INIT_LIST_HEAD(&segbuf->sb_list);
	81	INIT_LIST_HEAD(&segbuf->sb_segsum_buffers);
	82	INIT_LIST_HEAD(&segbuf->sb_payload_buffers);
9284ad2a RK	83
	84	init_completion(&segbuf->sb_bio_event);
	85	atomic_set(&segbuf->sb_err, 0);
	86	segbuf->sb_nbio = 0;
	87
64b5a32e RK	88	return segbuf;
	89	}
	90
	91	void nilfs_segbuf_free(struct nilfs_segment_buffer *segbuf)
	92	{
64b5a32e RK	93	kmem_cache_free(nilfs_segbuf_cachep, segbuf);
	94	}
	95
cece5520	96	void nilfs_segbuf_map(struct nilfs_segment_buffer *segbuf, __u64 segnum,
64b5a32e RK	97	unsigned long offset, struct the_nilfs *nilfs)
64b5a32e RK	98	{
64b5a32e RK	99	segbuf->sb_segnum = segnum;
	100	nilfs_get_segment_range(nilfs, segnum, &segbuf->sb_fseg_start,
	101	&segbuf->sb_fseg_end);
	102
	103	segbuf->sb_pseg_start = segbuf->sb_fseg_start + offset;
	104	segbuf->sb_rest_blocks =
	105	segbuf->sb_fseg_end - segbuf->sb_pseg_start + 1;
64b5a32e RK	106	}
64b5a32e RK	107
a694291a RK	108	/**
	109	* nilfs_segbuf_map_cont - map a new log behind a given log
	110	* @segbuf: new segment buffer
	111	* @prev: segment buffer containing a log to be continued
	112	*/
	113	void nilfs_segbuf_map_cont(struct nilfs_segment_buffer *segbuf,
	114	struct nilfs_segment_buffer *prev)
	115	{
	116	segbuf->sb_segnum = prev->sb_segnum;
	117	segbuf->sb_fseg_start = prev->sb_fseg_start;
	118	segbuf->sb_fseg_end = prev->sb_fseg_end;
	119	segbuf->sb_pseg_start = prev->sb_pseg_start + prev->sb_sum.nblocks;
	120	segbuf->sb_rest_blocks =
	121	segbuf->sb_fseg_end - segbuf->sb_pseg_start + 1;
	122	}
	123
64b5a32e RK	124	void nilfs_segbuf_set_next_segnum(struct nilfs_segment_buffer *segbuf,
	125	__u64 nextnum, struct the_nilfs *nilfs)
	126	{
	127	segbuf->sb_nextnum = nextnum;
	128	segbuf->sb_sum.next = nilfs_get_segment_start_blocknr(nilfs, nextnum);
	129	}
	130
	131	int nilfs_segbuf_extend_segsum(struct nilfs_segment_buffer *segbuf)
	132	{
	133	struct buffer_head *bh;
	134
	135	bh = sb_getblk(segbuf->sb_super,
	136	segbuf->sb_pseg_start + segbuf->sb_sum.nsumblk);
	137	if (unlikely(!bh))
	138	return -ENOMEM;
	139
	140	nilfs_segbuf_add_segsum_buffer(segbuf, bh);
	141	return 0;
	142	}
	143
	144	int nilfs_segbuf_extend_payload(struct nilfs_segment_buffer *segbuf,
	145	struct buffer_head **bhp)
	146	{
	147	struct buffer_head *bh;
	148
	149	bh = sb_getblk(segbuf->sb_super,
	150	segbuf->sb_pseg_start + segbuf->sb_sum.nblocks);
	151	if (unlikely(!bh))
	152	return -ENOMEM;
	153
	154	nilfs_segbuf_add_payload_buffer(segbuf, bh);
	155	*bhp = bh;
	156	return 0;
	157	}
	158
	159	int nilfs_segbuf_reset(struct nilfs_segment_buffer *segbuf, unsigned flags,
	160	time_t ctime)
	161	{
	162	int err;
	163
	164	segbuf->sb_sum.nblocks = segbuf->sb_sum.nsumblk = 0;
	165	err = nilfs_segbuf_extend_segsum(segbuf);
	166	if (unlikely(err))
	167	return err;
	168
	169	segbuf->sb_sum.flags = flags;
	170	segbuf->sb_sum.sumbytes = sizeof(struct nilfs_segment_summary);
	171	segbuf->sb_sum.nfinfo = segbuf->sb_sum.nfileblk = 0;
	172	segbuf->sb_sum.ctime = ctime;
64b5a32e RK	173	return 0;
	174	}
	175
	176	/*
7a65004b	177	* Setup segment summary
64b5a32e RK	178	*/
	179	void nilfs_segbuf_fill_in_segsum(struct nilfs_segment_buffer *segbuf)
	180	{
	181	struct nilfs_segment_summary *raw_sum;
	182	struct buffer_head *bh_sum;
	183
	184	bh_sum = list_entry(segbuf->sb_segsum_buffers.next,
	185	struct buffer_head, b_assoc_buffers);
	186	raw_sum = (struct nilfs_segment_summary *)bh_sum->b_data;
	187
	188	raw_sum->ss_magic = cpu_to_le32(NILFS_SEGSUM_MAGIC);
	189	raw_sum->ss_bytes = cpu_to_le16(sizeof(*raw_sum));
	190	raw_sum->ss_flags = cpu_to_le16(segbuf->sb_sum.flags);
	191	raw_sum->ss_seq = cpu_to_le64(segbuf->sb_sum.seg_seq);
	192	raw_sum->ss_create = cpu_to_le64(segbuf->sb_sum.ctime);
	193	raw_sum->ss_next = cpu_to_le64(segbuf->sb_sum.next);
	194	raw_sum->ss_nblocks = cpu_to_le32(segbuf->sb_sum.nblocks);
	195	raw_sum->ss_nfinfo = cpu_to_le32(segbuf->sb_sum.nfinfo);
	196	raw_sum->ss_sumbytes = cpu_to_le32(segbuf->sb_sum.sumbytes);
	197	raw_sum->ss_pad = 0;
	198	}
	199
	200	/*
	201	* CRC calculation routines
	202	*/
	203	void nilfs_segbuf_fill_in_segsum_crc(struct nilfs_segment_buffer *segbuf,
	204	u32 seed)
	205	{
	206	struct buffer_head *bh;
	207	struct nilfs_segment_summary *raw_sum;
	208	unsigned long size, bytes = segbuf->sb_sum.sumbytes;
	209	u32 crc;
	210
	211	bh = list_entry(segbuf->sb_segsum_buffers.next, struct buffer_head,
	212	b_assoc_buffers);
	213
	214	raw_sum = (struct nilfs_segment_summary *)bh->b_data;
	215	size = min_t(unsigned long, bytes, bh->b_size);
	216	crc = crc32_le(seed,
	217	(unsigned char *)raw_sum +
	218	sizeof(raw_sum->ss_datasum) + sizeof(raw_sum->ss_sumsum),
	219	size - (sizeof(raw_sum->ss_datasum) +
	220	sizeof(raw_sum->ss_sumsum)));
	221
	222	list_for_each_entry_continue(bh, &segbuf->sb_segsum_buffers,
	223	b_assoc_buffers) {
	224	bytes -= size;
	225	size = min_t(unsigned long, bytes, bh->b_size);
	226	crc = crc32_le(crc, bh->b_data, size);
	227	}
	228	raw_sum->ss_sumsum = cpu_to_le32(crc);
	229	}
	230
	231	void nilfs_segbuf_fill_in_data_crc(struct nilfs_segment_buffer *segbuf,
	232	u32 seed)
	233	{
	234	struct buffer_head *bh;
	235	struct nilfs_segment_summary *raw_sum;
	236	void *kaddr;
	237	u32 crc;
	238
	239	bh = list_entry(segbuf->sb_segsum_buffers.next, struct buffer_head,
	240	b_assoc_buffers);
	241	raw_sum = (struct nilfs_segment_summary *)bh->b_data;
242	crc = crc32_le(seed,
243	(unsigned char *)raw_sum + sizeof(raw_sum->ss_datasum),
244	bh->b_size - sizeof(raw_sum->ss_datasum));
245
246	list_for_each_entry_continue(bh, &segbuf->sb_segsum_buffers,
247	b_assoc_buffers) {
248	crc = crc32_le(crc, bh->b_data, bh->b_size);
249	}
250	list_for_each_entry(bh, &segbuf->sb_payload_buffers, b_assoc_buffers) {
251	kaddr = kmap_atomic(bh->b_page, KM_USER0);
252	crc = crc32_le(crc, kaddr + bh_offset(bh), bh->b_size);
253	kunmap_atomic(kaddr, KM_USER0);
254	}
255	raw_sum->ss_datasum = cpu_to_le32(crc);
256	}
257
e29df395	258	static void nilfs_release_buffers(struct list_head *list)
64b5a32e RK	259	{
	260	struct buffer_head bh, n;
	261
	262	list_for_each_entry_safe(bh, n, list, b_assoc_buffers) {
	263	list_del_init(&bh->b_assoc_buffers);
	264	if (buffer_nilfs_allocated(bh)) {
	265	struct page *clone_page = bh->b_page;
	266
	267	/* remove clone page */
	268	brelse(bh);
	269	page_cache_release(clone_page); /* for each bh */
	270	if (page_count(clone_page) <= 2) {
	271	lock_page(clone_page);
	272	nilfs_free_private_page(clone_page);
	273	}
	274	continue;
	275	}
	276	brelse(bh);
	277	}
	278	}
	279
e29df395 RK	280	static void nilfs_segbuf_clear(struct nilfs_segment_buffer *segbuf)
	281	{
	282	nilfs_release_buffers(&segbuf->sb_segsum_buffers);
	283	nilfs_release_buffers(&segbuf->sb_payload_buffers);
	284	}
	285
	286	/*
	287	* Iterators for segment buffers
	288	*/
	289	void nilfs_clear_logs(struct list_head *logs)
	290	{
	291	struct nilfs_segment_buffer *segbuf;
	292
	293	list_for_each_entry(segbuf, logs, sb_list)
	294	nilfs_segbuf_clear(segbuf);
	295	}
	296
	297	void nilfs_truncate_logs(struct list_head *logs,
	298	struct nilfs_segment_buffer *last)
	299	{
	300	struct nilfs_segment_buffer n, segbuf;
	301
	302	segbuf = list_prepare_entry(last, logs, sb_list);
	303	list_for_each_entry_safe_continue(segbuf, n, logs, sb_list) {
	304	list_del_init(&segbuf->sb_list);
	305	nilfs_segbuf_clear(segbuf);
	306	nilfs_segbuf_free(segbuf);
	307	}
	308	}
	309
d1c6b72a RK	310	int nilfs_write_logs(struct list_head logs, struct the_nilfs nilfs)
	311	{
	312	struct nilfs_segment_buffer *segbuf;
	313	int ret = 0;
	314
	315	list_for_each_entry(segbuf, logs, sb_list) {
	316	ret = nilfs_segbuf_write(segbuf, nilfs);
	317	if (ret)
	318	break;
	319	}
	320	return ret;
	321	}
	322
e29df395 RK	323	int nilfs_wait_on_logs(struct list_head *logs)
	324	{
	325	struct nilfs_segment_buffer *segbuf;
	326	int err;
	327
	328	list_for_each_entry(segbuf, logs, sb_list) {
	329	err = nilfs_segbuf_wait(segbuf);
	330	if (err)
	331	return err;
	332	}
	333	return 0;
	334	}
	335
64b5a32e RK	336	/*
	337	* BIO operations
	338	*/
	339	static void nilfs_end_bio_write(struct bio *bio, int err)
	340	{
	341	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
9284ad2a	342	struct nilfs_segment_buffer *segbuf = bio->bi_private;
64b5a32e RK	343
	344	if (err == -EOPNOTSUPP) {
	345	set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
	346	bio_put(bio);
	347	/* to be detected by submit_seg_bio() */
	348	}
	349
	350	if (!uptodate)
9284ad2a	351	atomic_inc(&segbuf->sb_err);
64b5a32e RK	352
64b5a32e RK	353	bio_put(bio);
9284ad2a	354	complete(&segbuf->sb_bio_event);
64b5a32e RK	355	}
64b5a32e RK	356
9284ad2a RK	357	static int nilfs_segbuf_submit_bio(struct nilfs_segment_buffer *segbuf,
9284ad2a RK	358	struct nilfs_write_info *wi, int mode)
64b5a32e RK	359	{
	360	struct bio *bio = wi->bio;
	361	int err;
	362
9c965bac	363	if (segbuf->sb_nbio > 0 && bdi_write_congested(wi->nilfs->ns_bdi)) {
9284ad2a RK	364	wait_for_completion(&segbuf->sb_bio_event);
	365	segbuf->sb_nbio--;
	366	if (unlikely(atomic_read(&segbuf->sb_err))) {
64b5a32e RK	367	bio_put(bio);
	368	err = -EIO;
	369	goto failed;
	370	}
	371	}
	372
	373	bio->bi_end_io = nilfs_end_bio_write;
9284ad2a	374	bio->bi_private = segbuf;
64b5a32e RK	375	bio_get(bio);
	376	submit_bio(mode, bio);
	377	if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
	378	bio_put(bio);
	379	err = -EOPNOTSUPP;
	380	goto failed;
	381	}
9284ad2a	382	segbuf->sb_nbio++;
64b5a32e RK	383	bio_put(bio);
	384
	385	wi->bio = NULL;
	386	wi->rest_blocks -= wi->end - wi->start;
	387	wi->nr_vecs = min(wi->max_pages, wi->rest_blocks);
	388	wi->start = wi->end;
	389	return 0;
	390
	391	failed:
	392	wi->bio = NULL;
	393	return err;
	394	}
	395
	396	/**
9c965bac RK	397	* nilfs_alloc_seg_bio - allocate a new bio for writing log
	398	* @nilfs: nilfs object
	399	* @start: start block number of the bio
64b5a32e RK	400	* @nr_vecs: request size of page vector.
64b5a32e RK	401	*
64b5a32e RK	402	* Return Value: On success, pointer to the struct bio is returned.
	403	* On error, NULL is returned.
	404	*/
9c965bac	405	static struct bio nilfs_alloc_seg_bio(struct the_nilfs nilfs, sector_t start,
64b5a32e RK	406	int nr_vecs)
	407	{
	408	struct bio *bio;
	409
c1b353f0	410	bio = bio_alloc(GFP_NOIO, nr_vecs);
64b5a32e RK	411	if (bio == NULL) {
64b5a32e RK	412	while (!bio && (nr_vecs >>= 1))
c1b353f0	413	bio = bio_alloc(GFP_NOIO, nr_vecs);
64b5a32e RK	414	}
64b5a32e RK	415	if (likely(bio)) {
9c965bac RK	416	bio->bi_bdev = nilfs->ns_bdev;
9c965bac RK	417	bio->bi_sector = start << (nilfs->ns_blocksize_bits - 9);
64b5a32e RK	418	}
	419	return bio;
	420	}
	421
9c965bac RK	422	static void nilfs_segbuf_prepare_write(struct nilfs_segment_buffer *segbuf,
9c965bac RK	423	struct nilfs_write_info *wi)
64b5a32e RK	424	{
	425	wi->bio = NULL;
	426	wi->rest_blocks = segbuf->sb_sum.nblocks;
9c965bac	427	wi->max_pages = bio_get_nr_vecs(wi->nilfs->ns_bdev);
64b5a32e RK	428	wi->nr_vecs = min(wi->max_pages, wi->rest_blocks);
64b5a32e RK	429	wi->start = wi->end = 0;
64b5a32e	430	wi->blocknr = segbuf->sb_pseg_start;
64b5a32e RK	431	}
64b5a32e RK	432
9284ad2a RK	433	static int nilfs_segbuf_submit_bh(struct nilfs_segment_buffer *segbuf,
	434	struct nilfs_write_info *wi,
	435	struct buffer_head *bh, int mode)
64b5a32e RK	436	{
	437	int len, err;
	438
	439	BUG_ON(wi->nr_vecs <= 0);
	440	repeat:
	441	if (!wi->bio) {
9c965bac	442	wi->bio = nilfs_alloc_seg_bio(wi->nilfs, wi->blocknr + wi->end,
64b5a32e RK	443	wi->nr_vecs);
	444	if (unlikely(!wi->bio))
	445	return -ENOMEM;
	446	}
	447
	448	len = bio_add_page(wi->bio, bh->b_page, bh->b_size, bh_offset(bh));
	449	if (len == bh->b_size) {
	450	wi->end++;
	451	return 0;
	452	}
	453	/* bio is FULL */
9284ad2a	454	err = nilfs_segbuf_submit_bio(segbuf, wi, mode);
64b5a32e RK	455	/* never submit current bh */
	456	if (likely(!err))
	457	goto repeat;
	458	return err;
	459	}
	460
9c965bac RK	461	/**
	462	* nilfs_segbuf_write - submit write requests of a log
	463	* @segbuf: buffer storing a log to be written
	464	* @nilfs: nilfs object
	465	*
	466	* Return Value: On Success, 0 is returned. On Error, one of the following
	467	* negative error code is returned.
	468	*
	469	* %-EIO - I/O error
	470	*
	471	* %-ENOMEM - Insufficient memory available.
	472	*/
6c477d44 RK	473	static int nilfs_segbuf_write(struct nilfs_segment_buffer *segbuf,
6c477d44 RK	474	struct the_nilfs *nilfs)
64b5a32e	475	{
9c965bac	476	struct nilfs_write_info wi;
64b5a32e	477	struct buffer_head *bh;
5f1586d0	478	int res = 0, rw = WRITE;
64b5a32e	479
9c965bac RK	480	wi.nilfs = nilfs;
	481	nilfs_segbuf_prepare_write(segbuf, &wi);
	482
64b5a32e	483	list_for_each_entry(bh, &segbuf->sb_segsum_buffers, b_assoc_buffers) {
9c965bac	484	res = nilfs_segbuf_submit_bh(segbuf, &wi, bh, rw);
64b5a32e RK	485	if (unlikely(res))
	486	goto failed_bio;
	487	}
	488
	489	list_for_each_entry(bh, &segbuf->sb_payload_buffers, b_assoc_buffers) {
9c965bac	490	res = nilfs_segbuf_submit_bh(segbuf, &wi, bh, rw);
64b5a32e RK	491	if (unlikely(res))
	492	goto failed_bio;
	493	}
	494
9c965bac	495	if (wi.bio) {
64b5a32e RK	496	/*
	497	* Last BIO is always sent through the following
	498	* submission.
	499	*/
30bda0b8	500	rw \|= (1 << BIO_RW_SYNCIO) \| (1 << BIO_RW_UNPLUG);
9c965bac	501	res = nilfs_segbuf_submit_bio(segbuf, &wi, rw);
64b5a32e RK	502	}
64b5a32e RK	503
64b5a32e	504	failed_bio:
5f1586d0	505	return res;
64b5a32e RK	506	}
	507
	508	/**
	509	* nilfs_segbuf_wait - wait for completion of requested BIOs
9284ad2a	510	* @segbuf: segment buffer
64b5a32e RK	511	*
	512	* Return Value: On Success, 0 is returned. On Error, one of the following
	513	* negative error code is returned.
	514	*
	515	* %-EIO - I/O error
	516	*/
6c477d44	517	static int nilfs_segbuf_wait(struct nilfs_segment_buffer *segbuf)
64b5a32e RK	518	{
	519	int err = 0;
	520
9284ad2a	521	if (!segbuf->sb_nbio)
64b5a32e RK	522	return 0;
	523
	524	do {
9284ad2a RK	525	wait_for_completion(&segbuf->sb_bio_event);
9284ad2a RK	526	} while (--segbuf->sb_nbio > 0);
64b5a32e	527
9284ad2a	528	if (unlikely(atomic_read(&segbuf->sb_err) > 0)) {
64b5a32e RK	529	printk(KERN_ERR "NILFS: IO error writing segment\n");
64b5a32e RK	530	err = -EIO;
64b5a32e RK	531	}
	532	return err;
	533	}