[linux.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 <[email protected]>
 *
 */

#include <linux/buffer_head.h>
#include <linux/writeback.h>
#include <linux/crc32.h>
#include <linux/backing-dev.h>
#include <linux/slab.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);

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);
	segbuf->sb_super_root = NULL;

	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, __u64 cno)
{
	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;
	segbuf->sb_sum.cno = cno;
	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;
	raw_sum->ss_cno      = cpu_to_le64(segbuf->sb_sum.cno);
}

/*
 * CRC calculation routines
 */
static 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);
}

static 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_segbuf_fill_in_super_root_crc(struct nilfs_segment_buffer *segbuf,
				    u32 seed)
{
	struct nilfs_super_root *raw_sr;
	u32 crc;

	raw_sr = (struct nilfs_super_root *)segbuf->sb_super_root->b_data;
	crc = crc32_le(seed,
		       (unsigned char *)raw_sr + sizeof(raw_sr->sr_sum),
		       NILFS_SR_BYTES - sizeof(raw_sr->sr_sum));
	raw_sr->sr_sum = 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);
	segbuf->sb_super_root = NULL;
}

/*
 * 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, ret = 0;

	list_for_each_entry(segbuf, logs, sb_list) {
		err = nilfs_segbuf_wait(segbuf);
		if (err && !ret)
			ret = err;
	}
	return ret;
}

/**
 * nilfs_add_checksums_on_logs - add checksums on the logs
 * @logs: list of segment buffers storing target logs
 * @seed: checksum seed value
 */
void nilfs_add_checksums_on_logs(struct list_head *logs, u32 seed)
{
	struct nilfs_segment_buffer *segbuf;

	list_for_each_entry(segbuf, logs, sb_list) {
		if (segbuf->sb_super_root)
			nilfs_segbuf_fill_in_super_root_crc(segbuf, seed);
		nilfs_segbuf_fill_in_segsum_crc(segbuf, seed);
		nilfs_segbuf_fill_in_data_crc(segbuf, seed);
	}
}

/*
 * 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 <[email protected]>
	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>
5a0e3ad6	28	#include <linux/slab.h>
64b5a32e RK	29	#include "page.h"
64b5a32e RK	30	#include "segbuf.h"
64b5a32e RK	31
64b5a32e RK	32
9c965bac RK	33	struct nilfs_write_info {
	34	struct the_nilfs *nilfs;
	35	struct bio *bio;
55480a06	36	int start, end; /* The region to be submitted */
9c965bac RK	37	int rest_blocks;
	38	int max_pages;
	39	int nr_vecs;
	40	sector_t blocknr;
	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
64b5a32e RK	47	struct nilfs_segment_buffer nilfs_segbuf_new(struct super_block sb)
	48	{
	49	struct nilfs_segment_buffer *segbuf;
	50
	51	segbuf = kmem_cache_alloc(nilfs_segbuf_cachep, GFP_NOFS);
	52	if (unlikely(!segbuf))
	53	return NULL;
	54
	55	segbuf->sb_super = sb;
	56	INIT_LIST_HEAD(&segbuf->sb_list);
	57	INIT_LIST_HEAD(&segbuf->sb_segsum_buffers);
	58	INIT_LIST_HEAD(&segbuf->sb_payload_buffers);
1e2b68bf	59	segbuf->sb_super_root = NULL;
9284ad2a RK	60
	61	init_completion(&segbuf->sb_bio_event);
	62	atomic_set(&segbuf->sb_err, 0);
	63	segbuf->sb_nbio = 0;
	64
64b5a32e RK	65	return segbuf;
	66	}
	67
	68	void nilfs_segbuf_free(struct nilfs_segment_buffer *segbuf)
	69	{
64b5a32e RK	70	kmem_cache_free(nilfs_segbuf_cachep, segbuf);
	71	}
	72
cece5520	73	void nilfs_segbuf_map(struct nilfs_segment_buffer *segbuf, __u64 segnum,
64b5a32e RK	74	unsigned long offset, struct the_nilfs *nilfs)
64b5a32e RK	75	{
64b5a32e RK	76	segbuf->sb_segnum = segnum;
	77	nilfs_get_segment_range(nilfs, segnum, &segbuf->sb_fseg_start,
	78	&segbuf->sb_fseg_end);
	79
	80	segbuf->sb_pseg_start = segbuf->sb_fseg_start + offset;
	81	segbuf->sb_rest_blocks =
	82	segbuf->sb_fseg_end - segbuf->sb_pseg_start + 1;
64b5a32e RK	83	}
64b5a32e RK	84
a694291a RK	85	/**
	86	* nilfs_segbuf_map_cont - map a new log behind a given log
	87	* @segbuf: new segment buffer
	88	* @prev: segment buffer containing a log to be continued
	89	*/
	90	void nilfs_segbuf_map_cont(struct nilfs_segment_buffer *segbuf,
	91	struct nilfs_segment_buffer *prev)
	92	{
	93	segbuf->sb_segnum = prev->sb_segnum;
	94	segbuf->sb_fseg_start = prev->sb_fseg_start;
	95	segbuf->sb_fseg_end = prev->sb_fseg_end;
	96	segbuf->sb_pseg_start = prev->sb_pseg_start + prev->sb_sum.nblocks;
	97	segbuf->sb_rest_blocks =
	98	segbuf->sb_fseg_end - segbuf->sb_pseg_start + 1;
	99	}
	100
64b5a32e RK	101	void nilfs_segbuf_set_next_segnum(struct nilfs_segment_buffer *segbuf,
	102	__u64 nextnum, struct the_nilfs *nilfs)
	103	{
	104	segbuf->sb_nextnum = nextnum;
	105	segbuf->sb_sum.next = nilfs_get_segment_start_blocknr(nilfs, nextnum);
	106	}
	107
	108	int nilfs_segbuf_extend_segsum(struct nilfs_segment_buffer *segbuf)
	109	{
	110	struct buffer_head *bh;
	111
	112	bh = sb_getblk(segbuf->sb_super,
	113	segbuf->sb_pseg_start + segbuf->sb_sum.nsumblk);
	114	if (unlikely(!bh))
	115	return -ENOMEM;
	116
	117	nilfs_segbuf_add_segsum_buffer(segbuf, bh);
	118	return 0;
	119	}
	120
	121	int nilfs_segbuf_extend_payload(struct nilfs_segment_buffer *segbuf,
	122	struct buffer_head **bhp)
	123	{
	124	struct buffer_head *bh;
	125
	126	bh = sb_getblk(segbuf->sb_super,
	127	segbuf->sb_pseg_start + segbuf->sb_sum.nblocks);
	128	if (unlikely(!bh))
	129	return -ENOMEM;
	130
	131	nilfs_segbuf_add_payload_buffer(segbuf, bh);
	132	*bhp = bh;
	133	return 0;
	134	}
	135
	136	int nilfs_segbuf_reset(struct nilfs_segment_buffer *segbuf, unsigned flags,
50614bcf	137	time_t ctime, __u64 cno)
64b5a32e RK	138	{
	139	int err;
	140
	141	segbuf->sb_sum.nblocks = segbuf->sb_sum.nsumblk = 0;
	142	err = nilfs_segbuf_extend_segsum(segbuf);
	143	if (unlikely(err))
	144	return err;
	145
	146	segbuf->sb_sum.flags = flags;
	147	segbuf->sb_sum.sumbytes = sizeof(struct nilfs_segment_summary);
	148	segbuf->sb_sum.nfinfo = segbuf->sb_sum.nfileblk = 0;
	149	segbuf->sb_sum.ctime = ctime;
50614bcf	150	segbuf->sb_sum.cno = cno;
64b5a32e RK	151	return 0;
	152	}
	153
	154	/*
7a65004b	155	* Setup segment summary
64b5a32e RK	156	*/
	157	void nilfs_segbuf_fill_in_segsum(struct nilfs_segment_buffer *segbuf)
	158	{
	159	struct nilfs_segment_summary *raw_sum;
	160	struct buffer_head *bh_sum;
	161
	162	bh_sum = list_entry(segbuf->sb_segsum_buffers.next,
	163	struct buffer_head, b_assoc_buffers);
	164	raw_sum = (struct nilfs_segment_summary *)bh_sum->b_data;
	165
	166	raw_sum->ss_magic = cpu_to_le32(NILFS_SEGSUM_MAGIC);
	167	raw_sum->ss_bytes = cpu_to_le16(sizeof(*raw_sum));
	168	raw_sum->ss_flags = cpu_to_le16(segbuf->sb_sum.flags);
	169	raw_sum->ss_seq = cpu_to_le64(segbuf->sb_sum.seg_seq);
	170	raw_sum->ss_create = cpu_to_le64(segbuf->sb_sum.ctime);
	171	raw_sum->ss_next = cpu_to_le64(segbuf->sb_sum.next);
	172	raw_sum->ss_nblocks = cpu_to_le32(segbuf->sb_sum.nblocks);
	173	raw_sum->ss_nfinfo = cpu_to_le32(segbuf->sb_sum.nfinfo);
	174	raw_sum->ss_sumbytes = cpu_to_le32(segbuf->sb_sum.sumbytes);
	175	raw_sum->ss_pad = 0;
50614bcf	176	raw_sum->ss_cno = cpu_to_le64(segbuf->sb_sum.cno);
64b5a32e RK	177	}
	178
	179	/*
	180	* CRC calculation routines
	181	*/
aaed1d5b RK	182	static void
aaed1d5b RK	183	nilfs_segbuf_fill_in_segsum_crc(struct nilfs_segment_buffer *segbuf, u32 seed)
64b5a32e RK	184	{
	185	struct buffer_head *bh;
	186	struct nilfs_segment_summary *raw_sum;
	187	unsigned long size, bytes = segbuf->sb_sum.sumbytes;
	188	u32 crc;
	189
	190	bh = list_entry(segbuf->sb_segsum_buffers.next, struct buffer_head,
	191	b_assoc_buffers);
	192
	193	raw_sum = (struct nilfs_segment_summary *)bh->b_data;
	194	size = min_t(unsigned long, bytes, bh->b_size);
	195	crc = crc32_le(seed,
	196	(unsigned char *)raw_sum +
	197	sizeof(raw_sum->ss_datasum) + sizeof(raw_sum->ss_sumsum),
	198	size - (sizeof(raw_sum->ss_datasum) +
	199	sizeof(raw_sum->ss_sumsum)));
	200
	201	list_for_each_entry_continue(bh, &segbuf->sb_segsum_buffers,
	202	b_assoc_buffers) {
	203	bytes -= size;
	204	size = min_t(unsigned long, bytes, bh->b_size);
	205	crc = crc32_le(crc, bh->b_data, size);
	206	}
	207	raw_sum->ss_sumsum = cpu_to_le32(crc);
	208	}
	209
aaed1d5b RK	210	static void nilfs_segbuf_fill_in_data_crc(struct nilfs_segment_buffer *segbuf,
aaed1d5b RK	211	u32 seed)
64b5a32e RK	212	{
	213	struct buffer_head *bh;
	214	struct nilfs_segment_summary *raw_sum;
	215	void *kaddr;
	216	u32 crc;
	217
	218	bh = list_entry(segbuf->sb_segsum_buffers.next, struct buffer_head,
	219	b_assoc_buffers);
	220	raw_sum = (struct nilfs_segment_summary *)bh->b_data;
	221	crc = crc32_le(seed,
	222	(unsigned char *)raw_sum + sizeof(raw_sum->ss_datasum),
	223	bh->b_size - sizeof(raw_sum->ss_datasum));
	224
	225	list_for_each_entry_continue(bh, &segbuf->sb_segsum_buffers,
	226	b_assoc_buffers) {
	227	crc = crc32_le(crc, bh->b_data, bh->b_size);
	228	}
	229	list_for_each_entry(bh, &segbuf->sb_payload_buffers, b_assoc_buffers) {
	230	kaddr = kmap_atomic(bh->b_page, KM_USER0);
	231	crc = crc32_le(crc, kaddr + bh_offset(bh), bh->b_size);
	232	kunmap_atomic(kaddr, KM_USER0);
	233	}
	234	raw_sum->ss_datasum = cpu_to_le32(crc);
	235	}
	236
aaed1d5b RK	237	static void
	238	nilfs_segbuf_fill_in_super_root_crc(struct nilfs_segment_buffer *segbuf,
	239	u32 seed)
	240	{
	241	struct nilfs_super_root *raw_sr;
	242	u32 crc;
	243
	244	raw_sr = (struct nilfs_super_root *)segbuf->sb_super_root->b_data;
	245	crc = crc32_le(seed,
	246	(unsigned char *)raw_sr + sizeof(raw_sr->sr_sum),
	247	NILFS_SR_BYTES - sizeof(raw_sr->sr_sum));
	248	raw_sr->sr_sum = cpu_to_le32(crc);
	249	}
	250
e29df395	251	static void nilfs_release_buffers(struct list_head *list)
64b5a32e RK	252	{
	253	struct buffer_head bh, n;
	254
	255	list_for_each_entry_safe(bh, n, list, b_assoc_buffers) {
	256	list_del_init(&bh->b_assoc_buffers);
	257	if (buffer_nilfs_allocated(bh)) {
	258	struct page *clone_page = bh->b_page;
	259
	260	/* remove clone page */
	261	brelse(bh);
	262	page_cache_release(clone_page); /* for each bh */
	263	if (page_count(clone_page) <= 2) {
	264	lock_page(clone_page);
	265	nilfs_free_private_page(clone_page);
	266	}
	267	continue;
	268	}
	269	brelse(bh);
	270	}
	271	}
	272
e29df395 RK	273	static void nilfs_segbuf_clear(struct nilfs_segment_buffer *segbuf)
	274	{
	275	nilfs_release_buffers(&segbuf->sb_segsum_buffers);
	276	nilfs_release_buffers(&segbuf->sb_payload_buffers);
1e2b68bf	277	segbuf->sb_super_root = NULL;
e29df395 RK	278	}
	279
	280	/*
	281	* Iterators for segment buffers
	282	*/
	283	void nilfs_clear_logs(struct list_head *logs)
	284	{
	285	struct nilfs_segment_buffer *segbuf;
	286
	287	list_for_each_entry(segbuf, logs, sb_list)
	288	nilfs_segbuf_clear(segbuf);
	289	}
	290
	291	void nilfs_truncate_logs(struct list_head *logs,
	292	struct nilfs_segment_buffer *last)
	293	{
	294	struct nilfs_segment_buffer n, segbuf;
	295
	296	segbuf = list_prepare_entry(last, logs, sb_list);
	297	list_for_each_entry_safe_continue(segbuf, n, logs, sb_list) {
	298	list_del_init(&segbuf->sb_list);
	299	nilfs_segbuf_clear(segbuf);
	300	nilfs_segbuf_free(segbuf);
	301	}
	302	}
	303
d1c6b72a RK	304	int nilfs_write_logs(struct list_head logs, struct the_nilfs nilfs)
	305	{
	306	struct nilfs_segment_buffer *segbuf;
	307	int ret = 0;
	308
	309	list_for_each_entry(segbuf, logs, sb_list) {
	310	ret = nilfs_segbuf_write(segbuf, nilfs);
	311	if (ret)
	312	break;
	313	}
	314	return ret;
	315	}
	316
e29df395 RK	317	int nilfs_wait_on_logs(struct list_head *logs)
	318	{
	319	struct nilfs_segment_buffer *segbuf;
d067633b	320	int err, ret = 0;
e29df395 RK	321
	322	list_for_each_entry(segbuf, logs, sb_list) {
	323	err = nilfs_segbuf_wait(segbuf);
d067633b RK	324	if (err && !ret)
d067633b RK	325	ret = err;
e29df395	326	}
d067633b	327	return ret;
e29df395 RK	328	}
e29df395 RK	329
aaed1d5b RK	330	/**
	331	* nilfs_add_checksums_on_logs - add checksums on the logs
	332	* @logs: list of segment buffers storing target logs
	333	* @seed: checksum seed value
	334	*/
	335	void nilfs_add_checksums_on_logs(struct list_head *logs, u32 seed)
	336	{
	337	struct nilfs_segment_buffer *segbuf;
	338
	339	list_for_each_entry(segbuf, logs, sb_list) {
	340	if (segbuf->sb_super_root)
	341	nilfs_segbuf_fill_in_super_root_crc(segbuf, seed);
	342	nilfs_segbuf_fill_in_segsum_crc(segbuf, seed);
	343	nilfs_segbuf_fill_in_data_crc(segbuf, seed);
	344	}
	345	}
	346
64b5a32e RK	347	/*
	348	* BIO operations
	349	*/
	350	static void nilfs_end_bio_write(struct bio *bio, int err)
	351	{
	352	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
9284ad2a	353	struct nilfs_segment_buffer *segbuf = bio->bi_private;
64b5a32e RK	354
	355	if (err == -EOPNOTSUPP) {
	356	set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
	357	bio_put(bio);
	358	/* to be detected by submit_seg_bio() */
	359	}
	360
	361	if (!uptodate)
9284ad2a	362	atomic_inc(&segbuf->sb_err);
64b5a32e RK	363
64b5a32e RK	364	bio_put(bio);
9284ad2a	365	complete(&segbuf->sb_bio_event);
64b5a32e RK	366	}
64b5a32e RK	367
9284ad2a RK	368	static int nilfs_segbuf_submit_bio(struct nilfs_segment_buffer *segbuf,
9284ad2a RK	369	struct nilfs_write_info *wi, int mode)
64b5a32e RK	370	{
	371	struct bio *bio = wi->bio;
	372	int err;
	373
9c965bac	374	if (segbuf->sb_nbio > 0 && bdi_write_congested(wi->nilfs->ns_bdi)) {
9284ad2a RK	375	wait_for_completion(&segbuf->sb_bio_event);
	376	segbuf->sb_nbio--;
	377	if (unlikely(atomic_read(&segbuf->sb_err))) {
64b5a32e RK	378	bio_put(bio);
	379	err = -EIO;
	380	goto failed;
	381	}
	382	}
	383
	384	bio->bi_end_io = nilfs_end_bio_write;
9284ad2a	385	bio->bi_private = segbuf;
64b5a32e RK	386	bio_get(bio);
	387	submit_bio(mode, bio);
	388	if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
	389	bio_put(bio);
	390	err = -EOPNOTSUPP;
	391	goto failed;
	392	}
9284ad2a	393	segbuf->sb_nbio++;
64b5a32e RK	394	bio_put(bio);
	395
	396	wi->bio = NULL;
	397	wi->rest_blocks -= wi->end - wi->start;
	398	wi->nr_vecs = min(wi->max_pages, wi->rest_blocks);
	399	wi->start = wi->end;
	400	return 0;
	401
	402	failed:
	403	wi->bio = NULL;
	404	return err;
	405	}
	406
	407	/**
9c965bac RK	408	* nilfs_alloc_seg_bio - allocate a new bio for writing log
	409	* @nilfs: nilfs object
	410	* @start: start block number of the bio
64b5a32e RK	411	* @nr_vecs: request size of page vector.
64b5a32e RK	412	*
64b5a32e RK	413	* Return Value: On success, pointer to the struct bio is returned.
	414	* On error, NULL is returned.
	415	*/
9c965bac	416	static struct bio nilfs_alloc_seg_bio(struct the_nilfs nilfs, sector_t start,
64b5a32e RK	417	int nr_vecs)
	418	{
	419	struct bio *bio;
	420
c1b353f0	421	bio = bio_alloc(GFP_NOIO, nr_vecs);
64b5a32e RK	422	if (bio == NULL) {
64b5a32e RK	423	while (!bio && (nr_vecs >>= 1))
c1b353f0	424	bio = bio_alloc(GFP_NOIO, nr_vecs);
64b5a32e RK	425	}
64b5a32e RK	426	if (likely(bio)) {
9c965bac RK	427	bio->bi_bdev = nilfs->ns_bdev;
9c965bac RK	428	bio->bi_sector = start << (nilfs->ns_blocksize_bits - 9);
64b5a32e RK	429	}
	430	return bio;
	431	}
	432
9c965bac RK	433	static void nilfs_segbuf_prepare_write(struct nilfs_segment_buffer *segbuf,
9c965bac RK	434	struct nilfs_write_info *wi)
64b5a32e RK	435	{
	436	wi->bio = NULL;
	437	wi->rest_blocks = segbuf->sb_sum.nblocks;
9c965bac	438	wi->max_pages = bio_get_nr_vecs(wi->nilfs->ns_bdev);
64b5a32e RK	439	wi->nr_vecs = min(wi->max_pages, wi->rest_blocks);
64b5a32e RK	440	wi->start = wi->end = 0;
64b5a32e	441	wi->blocknr = segbuf->sb_pseg_start;
64b5a32e RK	442	}
64b5a32e RK	443
9284ad2a RK	444	static int nilfs_segbuf_submit_bh(struct nilfs_segment_buffer *segbuf,
	445	struct nilfs_write_info *wi,
	446	struct buffer_head *bh, int mode)
64b5a32e RK	447	{
	448	int len, err;
	449
	450	BUG_ON(wi->nr_vecs <= 0);
	451	repeat:
	452	if (!wi->bio) {
9c965bac	453	wi->bio = nilfs_alloc_seg_bio(wi->nilfs, wi->blocknr + wi->end,
64b5a32e RK	454	wi->nr_vecs);
	455	if (unlikely(!wi->bio))
	456	return -ENOMEM;
	457	}
	458
	459	len = bio_add_page(wi->bio, bh->b_page, bh->b_size, bh_offset(bh));
	460	if (len == bh->b_size) {
	461	wi->end++;
	462	return 0;
	463	}
	464	/* bio is FULL */
9284ad2a	465	err = nilfs_segbuf_submit_bio(segbuf, wi, mode);
64b5a32e RK	466	/* never submit current bh */
	467	if (likely(!err))
	468	goto repeat;
	469	return err;
	470	}
	471
9c965bac RK	472	/**
	473	* nilfs_segbuf_write - submit write requests of a log
	474	* @segbuf: buffer storing a log to be written
	475	* @nilfs: nilfs object
	476	*
	477	* Return Value: On Success, 0 is returned. On Error, one of the following
	478	* negative error code is returned.
	479	*
	480	* %-EIO - I/O error
	481	*
	482	* %-ENOMEM - Insufficient memory available.
	483	*/
6c477d44 RK	484	static int nilfs_segbuf_write(struct nilfs_segment_buffer *segbuf,
6c477d44 RK	485	struct the_nilfs *nilfs)
64b5a32e	486	{
9c965bac	487	struct nilfs_write_info wi;
64b5a32e	488	struct buffer_head *bh;
5f1586d0	489	int res = 0, rw = WRITE;
64b5a32e	490
9c965bac RK	491	wi.nilfs = nilfs;
	492	nilfs_segbuf_prepare_write(segbuf, &wi);
	493
64b5a32e	494	list_for_each_entry(bh, &segbuf->sb_segsum_buffers, b_assoc_buffers) {
9c965bac	495	res = nilfs_segbuf_submit_bh(segbuf, &wi, bh, rw);
64b5a32e RK	496	if (unlikely(res))
	497	goto failed_bio;
	498	}
	499
	500	list_for_each_entry(bh, &segbuf->sb_payload_buffers, b_assoc_buffers) {
9c965bac	501	res = nilfs_segbuf_submit_bh(segbuf, &wi, bh, rw);
64b5a32e RK	502	if (unlikely(res))
	503	goto failed_bio;
	504	}
	505
9c965bac	506	if (wi.bio) {
64b5a32e RK	507	/*
	508	* Last BIO is always sent through the following
	509	* submission.
	510	*/
30bda0b8	511	rw \|= (1 << BIO_RW_SYNCIO) \| (1 << BIO_RW_UNPLUG);
9c965bac	512	res = nilfs_segbuf_submit_bio(segbuf, &wi, rw);
64b5a32e RK	513	}
64b5a32e RK	514
64b5a32e	515	failed_bio:
5f1586d0	516	return res;
64b5a32e RK	517	}
	518
	519	/**
	520	* nilfs_segbuf_wait - wait for completion of requested BIOs
9284ad2a	521	* @segbuf: segment buffer
64b5a32e RK	522	*
	523	* Return Value: On Success, 0 is returned. On Error, one of the following
	524	* negative error code is returned.
	525	*
	526	* %-EIO - I/O error
	527	*/
6c477d44	528	static int nilfs_segbuf_wait(struct nilfs_segment_buffer *segbuf)
64b5a32e RK	529	{
	530	int err = 0;
	531
9284ad2a	532	if (!segbuf->sb_nbio)
64b5a32e RK	533	return 0;
	534
	535	do {
9284ad2a RK	536	wait_for_completion(&segbuf->sb_bio_event);
9284ad2a RK	537	} while (--segbuf->sb_nbio > 0);
64b5a32e	538
9284ad2a	539	if (unlikely(atomic_read(&segbuf->sb_err) > 0)) {
64b5a32e RK	540	printk(KERN_ERR "NILFS: IO error writing segment\n");
64b5a32e RK	541	err = -EIO;
64b5a32e RK	542	}
	543	return err;
	544	}