[linux-2.6-block.git] / fs / f2fs / gc.c

/*
 * fs/f2fs/gc.c
 *
 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
 *             http://www.samsung.com/
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 */
#include <linux/fs.h>
#include <linux/module.h>
#include <linux/backing-dev.h>
#include <linux/proc_fs.h>
#include <linux/init.h>
#include <linux/f2fs_fs.h>
#include <linux/kthread.h>
#include <linux/delay.h>
#include <linux/freezer.h>
#include <linux/blkdev.h>

#include "f2fs.h"
#include "node.h"
#include "segment.h"
#include "gc.h"

static struct kmem_cache *winode_slab;

static int gc_thread_func(void *data)
{
	struct f2fs_sb_info *sbi = data;
	wait_queue_head_t *wq = &sbi->gc_thread->gc_wait_queue_head;
	long wait_ms;

	wait_ms = GC_THREAD_MIN_SLEEP_TIME;

	do {
		if (try_to_freeze())
			continue;
		else
			wait_event_interruptible_timeout(*wq,
						kthread_should_stop(),
						msecs_to_jiffies(wait_ms));
		if (kthread_should_stop())
			break;

		if (sbi->sb->s_writers.frozen >= SB_FREEZE_WRITE) {
			wait_ms = GC_THREAD_MAX_SLEEP_TIME;
			continue;
		}

		/*
		 * [GC triggering condition]
		 * 0. GC is not conducted currently.
		 * 1. There are enough dirty segments.
		 * 2. IO subsystem is idle by checking the # of writeback pages.
		 * 3. IO subsystem is idle by checking the # of requests in
		 *    bdev's request list.
		 *
		 * Note) We have to avoid triggering GCs too much frequently.
		 * Because it is possible that some segments can be
		 * invalidated soon after by user update or deletion.
		 * So, I'd like to wait some time to collect dirty segments.
		 */
		if (!mutex_trylock(&sbi->gc_mutex))
			continue;

		if (!is_idle(sbi)) {
			wait_ms = increase_sleep_time(wait_ms);
			mutex_unlock(&sbi->gc_mutex);
			continue;
		}

		if (has_enough_invalid_blocks(sbi))
			wait_ms = decrease_sleep_time(wait_ms);
		else
			wait_ms = increase_sleep_time(wait_ms);

		sbi->bg_gc++;

		/* if return value is not zero, no victim was selected */
		if (f2fs_gc(sbi))
			wait_ms = GC_THREAD_NOGC_SLEEP_TIME;
		else if (wait_ms == GC_THREAD_NOGC_SLEEP_TIME)
			wait_ms = GC_THREAD_MAX_SLEEP_TIME;

	} while (!kthread_should_stop());
	return 0;
}

int start_gc_thread(struct f2fs_sb_info *sbi)
{
	struct f2fs_gc_kthread *gc_th;
	dev_t dev = sbi->sb->s_bdev->bd_dev;

	if (!test_opt(sbi, BG_GC))
		return 0;
	gc_th = kmalloc(sizeof(struct f2fs_gc_kthread), GFP_KERNEL);
	if (!gc_th)
		return -ENOMEM;

	sbi->gc_thread = gc_th;
	init_waitqueue_head(&sbi->gc_thread->gc_wait_queue_head);
	sbi->gc_thread->f2fs_gc_task = kthread_run(gc_thread_func, sbi,
			"f2fs_gc-%u:%u", MAJOR(dev), MINOR(dev));
	if (IS_ERR(gc_th->f2fs_gc_task)) {
		kfree(gc_th);
		sbi->gc_thread = NULL;
		return -ENOMEM;
	}
	return 0;
}

void stop_gc_thread(struct f2fs_sb_info *sbi)
{
	struct f2fs_gc_kthread *gc_th = sbi->gc_thread;
	if (!gc_th)
		return;
	kthread_stop(gc_th->f2fs_gc_task);
	kfree(gc_th);
	sbi->gc_thread = NULL;
}

static int select_gc_type(int gc_type)
{
	return (gc_type == BG_GC) ? GC_CB : GC_GREEDY;
}

static void select_policy(struct f2fs_sb_info *sbi, int gc_type,
			int type, struct victim_sel_policy *p)
{
	struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);

	if (p->alloc_mode) {
		p->gc_mode = GC_GREEDY;
		p->dirty_segmap = dirty_i->dirty_segmap[type];
		p->ofs_unit = 1;
	} else {
		p->gc_mode = select_gc_type(gc_type);
		p->dirty_segmap = dirty_i->dirty_segmap[DIRTY];
		p->ofs_unit = sbi->segs_per_sec;
	}
	p->offset = sbi->last_victim[p->gc_mode];
}

static unsigned int get_max_cost(struct f2fs_sb_info *sbi,
				struct victim_sel_policy *p)
{
	if (p->gc_mode == GC_GREEDY)
		return (1 << sbi->log_blocks_per_seg) * p->ofs_unit;
	else if (p->gc_mode == GC_CB)
		return UINT_MAX;
	else /* No other gc_mode */
		return 0;
}

static unsigned int check_bg_victims(struct f2fs_sb_info *sbi)
{
	struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
	unsigned int segno;

	/*
	 * If the gc_type is FG_GC, we can select victim segments
	 * selected by background GC before.
	 * Those segments guarantee they have small valid blocks.
	 */
	segno = find_next_bit(dirty_i->victim_segmap[BG_GC],
						TOTAL_SEGS(sbi), 0);
	if (segno < TOTAL_SEGS(sbi)) {
		clear_bit(segno, dirty_i->victim_segmap[BG_GC]);
		return segno;
	}
	return NULL_SEGNO;
}

static unsigned int get_cb_cost(struct f2fs_sb_info *sbi, unsigned int segno)
{
	struct sit_info *sit_i = SIT_I(sbi);
	unsigned int secno = GET_SECNO(sbi, segno);
	unsigned int start = secno * sbi->segs_per_sec;
	unsigned long long mtime = 0;
	unsigned int vblocks;
	unsigned char age = 0;
	unsigned char u;
	unsigned int i;

	for (i = 0; i < sbi->segs_per_sec; i++)
		mtime += get_seg_entry(sbi, start + i)->mtime;
	vblocks = get_valid_blocks(sbi, segno, sbi->segs_per_sec);

	mtime = div_u64(mtime, sbi->segs_per_sec);
	vblocks = div_u64(vblocks, sbi->segs_per_sec);

	u = (vblocks * 100) >> sbi->log_blocks_per_seg;

	/* Handle if the system time is changed by user */
	if (mtime < sit_i->min_mtime)
		sit_i->min_mtime = mtime;
	if (mtime > sit_i->max_mtime)
		sit_i->max_mtime = mtime;
	if (sit_i->max_mtime != sit_i->min_mtime)
		age = 100 - div64_u64(100 * (mtime - sit_i->min_mtime),
				sit_i->max_mtime - sit_i->min_mtime);

	return UINT_MAX - ((100 * (100 - u) * age) / (100 + u));
}

static unsigned int get_gc_cost(struct f2fs_sb_info *sbi, unsigned int segno,
					struct victim_sel_policy *p)
{
	if (p->alloc_mode == SSR)
		return get_seg_entry(sbi, segno)->ckpt_valid_blocks;

	/* alloc_mode == LFS */
	if (p->gc_mode == GC_GREEDY)
		return get_valid_blocks(sbi, segno, sbi->segs_per_sec);
	else
		return get_cb_cost(sbi, segno);
}

/*
 * This function is called from two pathes.
 * One is garbage collection and the other is SSR segment selection.
 * When it is called during GC, it just gets a victim segment
 * and it does not remove it from dirty seglist.
 * When it is called from SSR segment selection, it finds a segment
 * which has minimum valid blocks and removes it from dirty seglist.
 */
static int get_victim_by_default(struct f2fs_sb_info *sbi,
		unsigned int *result, int gc_type, int type, char alloc_mode)
{
	struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
	struct victim_sel_policy p;
	unsigned int segno;
	int nsearched = 0;

	p.alloc_mode = alloc_mode;
	select_policy(sbi, gc_type, type, &p);

	p.min_segno = NULL_SEGNO;
	p.min_cost = get_max_cost(sbi, &p);

	mutex_lock(&dirty_i->seglist_lock);

	if (p.alloc_mode == LFS && gc_type == FG_GC) {
		p.min_segno = check_bg_victims(sbi);
		if (p.min_segno != NULL_SEGNO)
			goto got_it;
	}

	while (1) {
		unsigned long cost;

		segno = find_next_bit(p.dirty_segmap,
						TOTAL_SEGS(sbi), p.offset);
		if (segno >= TOTAL_SEGS(sbi)) {
			if (sbi->last_victim[p.gc_mode]) {
				sbi->last_victim[p.gc_mode] = 0;
				p.offset = 0;
				continue;
			}
			break;
		}
		p.offset = ((segno / p.ofs_unit) * p.ofs_unit) + p.ofs_unit;

		if (test_bit(segno, dirty_i->victim_segmap[FG_GC]))
			continue;
		if (gc_type == BG_GC &&
				test_bit(segno, dirty_i->victim_segmap[BG_GC]))
			continue;
		if (IS_CURSEC(sbi, GET_SECNO(sbi, segno)))
			continue;

		cost = get_gc_cost(sbi, segno, &p);

		if (p.min_cost > cost) {
			p.min_segno = segno;
			p.min_cost = cost;
		}

		if (cost == get_max_cost(sbi, &p))
			continue;

		if (nsearched++ >= MAX_VICTIM_SEARCH) {
			sbi->last_victim[p.gc_mode] = segno;
			break;
		}
	}
got_it:
	if (p.min_segno != NULL_SEGNO) {
		*result = (p.min_segno / p.ofs_unit) * p.ofs_unit;
		if (p.alloc_mode == LFS) {
			int i;
			for (i = 0; i < p.ofs_unit; i++)
				set_bit(*result + i,
					dirty_i->victim_segmap[gc_type]);
		}
	}
	mutex_unlock(&dirty_i->seglist_lock);

	return (p.min_segno == NULL_SEGNO) ? 0 : 1;
}

static const struct victim_selection default_v_ops = {
	.get_victim = get_victim_by_default,
};

static struct inode *find_gc_inode(nid_t ino, struct list_head *ilist)
{
	struct list_head *this;
	struct inode_entry *ie;

	list_for_each(this, ilist) {
		ie = list_entry(this, struct inode_entry, list);
		if (ie->inode->i_ino == ino)
			return ie->inode;
	}
	return NULL;
}

static void add_gc_inode(struct inode *inode, struct list_head *ilist)
{
	struct list_head *this;
	struct inode_entry *new_ie, *ie;

	list_for_each(this, ilist) {
		ie = list_entry(this, struct inode_entry, list);
		if (ie->inode == inode) {
			iput(inode);
			return;
		}
	}
repeat:
	new_ie = kmem_cache_alloc(winode_slab, GFP_NOFS);
	if (!new_ie) {
		cond_resched();
		goto repeat;
	}
	new_ie->inode = inode;
	list_add_tail(&new_ie->list, ilist);
}

static void put_gc_inode(struct list_head *ilist)
{
	struct inode_entry *ie, *next_ie;
	list_for_each_entry_safe(ie, next_ie, ilist, list) {
		iput(ie->inode);
		list_del(&ie->list);
		kmem_cache_free(winode_slab, ie);
	}
}

static int check_valid_map(struct f2fs_sb_info *sbi,
				unsigned int segno, int offset)
{
	struct sit_info *sit_i = SIT_I(sbi);
	struct seg_entry *sentry;
	int ret;

	mutex_lock(&sit_i->sentry_lock);
	sentry = get_seg_entry(sbi, segno);
	ret = f2fs_test_bit(offset, sentry->cur_valid_map);
	mutex_unlock(&sit_i->sentry_lock);
	return ret;
}

/*
 * This function compares node address got in summary with that in NAT.
 * On validity, copy that node with cold status, otherwise (invalid node)
 * ignore that.
 */
static void gc_node_segment(struct f2fs_sb_info *sbi,
		struct f2fs_summary *sum, unsigned int segno, int gc_type)
{
	bool initial = true;
	struct f2fs_summary *entry;
	int off;

next_step:
	entry = sum;
	for (off = 0; off < sbi->blocks_per_seg; off++, entry++) {
		nid_t nid = le32_to_cpu(entry->nid);
		struct page *node_page;

		/* stop BG_GC if there is not enough free sections. */
		if (gc_type == BG_GC && has_not_enough_free_secs(sbi, 0))
			return;

		if (check_valid_map(sbi, segno, off) == 0)
			continue;

		if (initial) {
			ra_node_page(sbi, nid);
			continue;
		}
		node_page = get_node_page(sbi, nid);
		if (IS_ERR(node_page))
			continue;

		/* set page dirty and write it */
		if (!PageWriteback(node_page))
			set_page_dirty(node_page);
		f2fs_put_page(node_page, 1);
		stat_inc_node_blk_count(sbi, 1);
	}
	if (initial) {
		initial = false;
		goto next_step;
	}

	if (gc_type == FG_GC) {
		struct writeback_control wbc = {
			.sync_mode = WB_SYNC_ALL,
			.nr_to_write = LONG_MAX,
			.for_reclaim = 0,
		};
		sync_node_pages(sbi, 0, &wbc);
	}
}

/*
 * Calculate start block index indicating the given node offset.
 * Be careful, caller should give this node offset only indicating direct node
 * blocks. If any node offsets, which point the other types of node blocks such
 * as indirect or double indirect node blocks, are given, it must be a caller's
 * bug.
 */
block_t start_bidx_of_node(unsigned int node_ofs)
{
	unsigned int indirect_blks = 2 * NIDS_PER_BLOCK + 4;
	unsigned int bidx;

	if (node_ofs == 0)
		return 0;

	if (node_ofs <= 2) {
		bidx = node_ofs - 1;
	} else if (node_ofs <= indirect_blks) {
		int dec = (node_ofs - 4) / (NIDS_PER_BLOCK + 1);
		bidx = node_ofs - 2 - dec;
	} else {
		int dec = (node_ofs - indirect_blks - 3) / (NIDS_PER_BLOCK + 1);
		bidx = node_ofs - 5 - dec;
	}
	return bidx * ADDRS_PER_BLOCK + ADDRS_PER_INODE;
}

static int check_dnode(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
		struct node_info *dni, block_t blkaddr, unsigned int *nofs)
{
	struct page *node_page;
	nid_t nid;
	unsigned int ofs_in_node;
	block_t source_blkaddr;

	nid = le32_to_cpu(sum->nid);
	ofs_in_node = le16_to_cpu(sum->ofs_in_node);

	node_page = get_node_page(sbi, nid);
	if (IS_ERR(node_page))
		return 0;

	get_node_info(sbi, nid, dni);

	if (sum->version != dni->version) {
		f2fs_put_page(node_page, 1);
		return 0;
	}

	*nofs = ofs_of_node(node_page);
	source_blkaddr = datablock_addr(node_page, ofs_in_node);
	f2fs_put_page(node_page, 1);

	if (source_blkaddr != blkaddr)
		return 0;
	return 1;
}

static void move_data_page(struct inode *inode, struct page *page, int gc_type)
{
	if (page->mapping != inode->i_mapping)
		goto out;

	if (inode != page->mapping->host)
		goto out;

	if (PageWriteback(page))
		goto out;

	if (gc_type == BG_GC) {
		set_page_dirty(page);
		set_cold_data(page);
	} else {
		struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
		mutex_lock_op(sbi, DATA_WRITE);
		if (clear_page_dirty_for_io(page) &&
			S_ISDIR(inode->i_mode)) {
			dec_page_count(sbi, F2FS_DIRTY_DENTS);
			inode_dec_dirty_dents(inode);
		}
		set_cold_data(page);
		do_write_data_page(page);
		mutex_unlock_op(sbi, DATA_WRITE);
		clear_cold_data(page);
	}
out:
	f2fs_put_page(page, 1);
}

/*
 * This function tries to get parent node of victim data block, and identifies
 * data block validity. If the block is valid, copy that with cold status and
 * modify parent node.
 * If the parent node is not valid or the data block address is different,
 * the victim data block is ignored.
 */
static void gc_data_segment(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
		struct list_head *ilist, unsigned int segno, int gc_type)
{
	struct super_block *sb = sbi->sb;
	struct f2fs_summary *entry;
	block_t start_addr;
	int off;
	int phase = 0;

	start_addr = START_BLOCK(sbi, segno);

next_step:
	entry = sum;
	for (off = 0; off < sbi->blocks_per_seg; off++, entry++) {
		struct page *data_page;
		struct inode *inode;
		struct node_info dni; /* dnode info for the data */
		unsigned int ofs_in_node, nofs;
		block_t start_bidx;

		/* stop BG_GC if there is not enough free sections. */
		if (gc_type == BG_GC && has_not_enough_free_secs(sbi, 0))
			return;

		if (check_valid_map(sbi, segno, off) == 0)
			continue;

		if (phase == 0) {
			ra_node_page(sbi, le32_to_cpu(entry->nid));
			continue;
		}

		/* Get an inode by ino with checking validity */
		if (check_dnode(sbi, entry, &dni, start_addr + off, &nofs) == 0)
			continue;

		if (phase == 1) {
			ra_node_page(sbi, dni.ino);
			continue;
		}

		start_bidx = start_bidx_of_node(nofs);
		ofs_in_node = le16_to_cpu(entry->ofs_in_node);

		if (phase == 2) {
			inode = f2fs_iget(sb, dni.ino);
			if (IS_ERR(inode))
				continue;

			data_page = find_data_page(inode,
					start_bidx + ofs_in_node);
			if (IS_ERR(data_page))
				goto next_iput;

			f2fs_put_page(data_page, 0);
			add_gc_inode(inode, ilist);
		} else {
			inode = find_gc_inode(dni.ino, ilist);
			if (inode) {
				data_page = get_lock_data_page(inode,
						start_bidx + ofs_in_node);
				if (IS_ERR(data_page))
					continue;
				move_data_page(inode, data_page, gc_type);
				stat_inc_data_blk_count(sbi, 1);
			}
		}
		continue;
next_iput:
		iput(inode);
	}
	if (++phase < 4)
		goto next_step;

	if (gc_type == FG_GC)
		f2fs_submit_bio(sbi, DATA, true);
}

static int __get_victim(struct f2fs_sb_info *sbi, unsigned int *victim,
						int gc_type, int type)
{
	struct sit_info *sit_i = SIT_I(sbi);
	int ret;
	mutex_lock(&sit_i->sentry_lock);
	ret = DIRTY_I(sbi)->v_ops->get_victim(sbi, victim, gc_type, type, LFS);
	mutex_unlock(&sit_i->sentry_lock);
	return ret;
}

static void do_garbage_collect(struct f2fs_sb_info *sbi, unsigned int segno,
				struct list_head *ilist, int gc_type)
{
	struct page *sum_page;
	struct f2fs_summary_block *sum;

	/* read segment summary of victim */
	sum_page = get_sum_page(sbi, segno);
	if (IS_ERR(sum_page))
		return;

	/*
	 * CP needs to lock sum_page. In this time, we don't need
	 * to lock this page, because this summary page is not gone anywhere.
	 * Also, this page is not gonna be updated before GC is done.
	 */
	unlock_page(sum_page);
	sum = page_address(sum_page);

	switch (GET_SUM_TYPE((&sum->footer))) {
	case SUM_TYPE_NODE:
		gc_node_segment(sbi, sum->entries, segno, gc_type);
		break;
	case SUM_TYPE_DATA:
		gc_data_segment(sbi, sum->entries, ilist, segno, gc_type);
		break;
	}
	stat_inc_seg_count(sbi, GET_SUM_TYPE((&sum->footer)));
	stat_inc_call_count(sbi->stat_info);

	f2fs_put_page(sum_page, 0);
}

int f2fs_gc(struct f2fs_sb_info *sbi)
{
	struct list_head ilist;
	unsigned int segno, i;
	int gc_type = BG_GC;
	int nfree = 0;
	int ret = -1;

	INIT_LIST_HEAD(&ilist);
gc_more:
	if (!(sbi->sb->s_flags & MS_ACTIVE))
		goto stop;

	if (gc_type == BG_GC && has_not_enough_free_secs(sbi, nfree))
		gc_type = FG_GC;

	if (!__get_victim(sbi, &segno, gc_type, NO_CHECK_TYPE))
		goto stop;
	ret = 0;

	for (i = 0; i < sbi->segs_per_sec; i++)
		do_garbage_collect(sbi, segno + i, &ilist, gc_type);

	if (gc_type == FG_GC &&
			get_valid_blocks(sbi, segno, sbi->segs_per_sec) == 0)
		nfree++;

	if (has_not_enough_free_secs(sbi, nfree))
		goto gc_more;

	if (gc_type == FG_GC)
		write_checkpoint(sbi, false);
stop:
	mutex_unlock(&sbi->gc_mutex);

	put_gc_inode(&ilist);
	return ret;
}

void build_gc_manager(struct f2fs_sb_info *sbi)
{
	DIRTY_I(sbi)->v_ops = &default_v_ops;
}

int __init create_gc_caches(void)
{
	winode_slab = f2fs_kmem_cache_create("f2fs_gc_inodes",
			sizeof(struct inode_entry), NULL);
	if (!winode_slab)
		return -ENOMEM;
	return 0;
}

void destroy_gc_caches(void)
{
	kmem_cache_destroy(winode_slab);
}
Commit	Line	Data
0a8165d7	1	/*
7bc09003 JK	2	* fs/f2fs/gc.c
	3	*
	4	* Copyright (c) 2012 Samsung Electronics Co., Ltd.
	5	* http://www.samsung.com/
	6	*
	7	* This program is free software; you can redistribute it and/or modify
	8	* it under the terms of the GNU General Public License version 2 as
	9	* published by the Free Software Foundation.
	10	*/
	11	#include <linux/fs.h>
	12	#include <linux/module.h>
	13	#include <linux/backing-dev.h>
	14	#include <linux/proc_fs.h>
	15	#include <linux/init.h>
	16	#include <linux/f2fs_fs.h>
	17	#include <linux/kthread.h>
	18	#include <linux/delay.h>
	19	#include <linux/freezer.h>
	20	#include <linux/blkdev.h>
	21
	22	#include "f2fs.h"
	23	#include "node.h"
	24	#include "segment.h"
	25	#include "gc.h"
	26
	27	static struct kmem_cache *winode_slab;
	28
	29	static int gc_thread_func(void *data)
	30	{
	31	struct f2fs_sb_info *sbi = data;
	32	wait_queue_head_t *wq = &sbi->gc_thread->gc_wait_queue_head;
	33	long wait_ms;
	34
	35	wait_ms = GC_THREAD_MIN_SLEEP_TIME;
	36
	37	do {
	38	if (try_to_freeze())
	39	continue;
	40	else
	41	wait_event_interruptible_timeout(*wq,
	42	kthread_should_stop(),
	43	msecs_to_jiffies(wait_ms));
	44	if (kthread_should_stop())
	45	break;
	46
d6212a5f CL	47	if (sbi->sb->s_writers.frozen >= SB_FREEZE_WRITE) {
	48	wait_ms = GC_THREAD_MAX_SLEEP_TIME;
	49	continue;
	50	}
	51
7bc09003 JK	52	/*
	53	* [GC triggering condition]
	54	* 0. GC is not conducted currently.
	55	* 1. There are enough dirty segments.
	56	* 2. IO subsystem is idle by checking the # of writeback pages.
	57	* 3. IO subsystem is idle by checking the # of requests in
	58	* bdev's request list.
	59	*
	60	* Note) We have to avoid triggering GCs too much frequently.
	61	* Because it is possible that some segments can be
	62	* invalidated soon after by user update or deletion.
	63	* So, I'd like to wait some time to collect dirty segments.
	64	*/
	65	if (!mutex_trylock(&sbi->gc_mutex))
	66	continue;
	67
	68	if (!is_idle(sbi)) {
	69	wait_ms = increase_sleep_time(wait_ms);
	70	mutex_unlock(&sbi->gc_mutex);
	71	continue;
	72	}
	73
	74	if (has_enough_invalid_blocks(sbi))
	75	wait_ms = decrease_sleep_time(wait_ms);
	76	else
	77	wait_ms = increase_sleep_time(wait_ms);
	78
	79	sbi->bg_gc++;
	80
43727527 JK	81	/* if return value is not zero, no victim was selected */
43727527 JK	82	if (f2fs_gc(sbi))
7bc09003 JK	83	wait_ms = GC_THREAD_NOGC_SLEEP_TIME;
	84	else if (wait_ms == GC_THREAD_NOGC_SLEEP_TIME)
	85	wait_ms = GC_THREAD_MAX_SLEEP_TIME;
	86
	87	} while (!kthread_should_stop());
	88	return 0;
	89	}
	90
	91	int start_gc_thread(struct f2fs_sb_info *sbi)
	92	{
1042d60f	93	struct f2fs_gc_kthread *gc_th;
ec7b1f2d	94	dev_t dev = sbi->sb->s_bdev->bd_dev;
7bc09003	95
48600e44 CL	96	if (!test_opt(sbi, BG_GC))
48600e44 CL	97	return 0;
7bc09003 JK	98	gc_th = kmalloc(sizeof(struct f2fs_gc_kthread), GFP_KERNEL);
	99	if (!gc_th)
	100	return -ENOMEM;
	101
	102	sbi->gc_thread = gc_th;
	103	init_waitqueue_head(&sbi->gc_thread->gc_wait_queue_head);
	104	sbi->gc_thread->f2fs_gc_task = kthread_run(gc_thread_func, sbi,
ec7b1f2d	105	"f2fs_gc-%u:%u", MAJOR(dev), MINOR(dev));
7bc09003 JK	106	if (IS_ERR(gc_th->f2fs_gc_task)) {
7bc09003 JK	107	kfree(gc_th);
25718423	108	sbi->gc_thread = NULL;
7bc09003 JK	109	return -ENOMEM;
	110	}
	111	return 0;
	112	}
	113
	114	void stop_gc_thread(struct f2fs_sb_info *sbi)
	115	{
	116	struct f2fs_gc_kthread *gc_th = sbi->gc_thread;
	117	if (!gc_th)
	118	return;
	119	kthread_stop(gc_th->f2fs_gc_task);
	120	kfree(gc_th);
	121	sbi->gc_thread = NULL;
	122	}
	123
	124	static int select_gc_type(int gc_type)
	125	{
	126	return (gc_type == BG_GC) ? GC_CB : GC_GREEDY;
	127	}
	128
	129	static void select_policy(struct f2fs_sb_info *sbi, int gc_type,
	130	int type, struct victim_sel_policy *p)
	131	{
	132	struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
	133
	134	if (p->alloc_mode) {
	135	p->gc_mode = GC_GREEDY;
	136	p->dirty_segmap = dirty_i->dirty_segmap[type];
	137	p->ofs_unit = 1;
	138	} else {
	139	p->gc_mode = select_gc_type(gc_type);
	140	p->dirty_segmap = dirty_i->dirty_segmap[DIRTY];
	141	p->ofs_unit = sbi->segs_per_sec;
	142	}
	143	p->offset = sbi->last_victim[p->gc_mode];
	144	}
	145
	146	static unsigned int get_max_cost(struct f2fs_sb_info *sbi,
	147	struct victim_sel_policy *p)
	148	{
	149	if (p->gc_mode == GC_GREEDY)
	150	return (1 << sbi->log_blocks_per_seg) * p->ofs_unit;
	151	else if (p->gc_mode == GC_CB)
	152	return UINT_MAX;
	153	else /* No other gc_mode */
	154	return 0;
	155	}
	156
	157	static unsigned int check_bg_victims(struct f2fs_sb_info *sbi)
	158	{
	159	struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
	160	unsigned int segno;
	161
	162	/*
	163	* If the gc_type is FG_GC, we can select victim segments
	164	* selected by background GC before.
	165	* Those segments guarantee they have small valid blocks.
	166	*/
	167	segno = find_next_bit(dirty_i->victim_segmap[BG_GC],
	168	TOTAL_SEGS(sbi), 0);
	169	if (segno < TOTAL_SEGS(sbi)) {
	170	clear_bit(segno, dirty_i->victim_segmap[BG_GC]);
	171	return segno;
	172	}
173	return NULL_SEGNO;
174	}
175
176	static unsigned int get_cb_cost(struct f2fs_sb_info *sbi, unsigned int segno)
177	{
178	struct sit_info *sit_i = SIT_I(sbi);
179	unsigned int secno = GET_SECNO(sbi, segno);
180	unsigned int start = secno * sbi->segs_per_sec;
181	unsigned long long mtime = 0;
182	unsigned int vblocks;
183	unsigned char age = 0;
184	unsigned char u;
185	unsigned int i;
186
187	for (i = 0; i < sbi->segs_per_sec; i++)
188	mtime += get_seg_entry(sbi, start + i)->mtime;
189	vblocks = get_valid_blocks(sbi, segno, sbi->segs_per_sec);
190
191	mtime = div_u64(mtime, sbi->segs_per_sec);
192	vblocks = div_u64(vblocks, sbi->segs_per_sec);
193
194	u = (vblocks * 100) >> sbi->log_blocks_per_seg;
195
196	/* Handle if the system time is changed by user */
197	if (mtime < sit_i->min_mtime)
198	sit_i->min_mtime = mtime;
199	if (mtime > sit_i->max_mtime)
200	sit_i->max_mtime = mtime;
201	if (sit_i->max_mtime != sit_i->min_mtime)
202	age = 100 - div64_u64(100 * (mtime - sit_i->min_mtime),
203	sit_i->max_mtime - sit_i->min_mtime);
204
205	return UINT_MAX - ((100 * (100 - u) * age) / (100 + u));
206	}
207
208	static unsigned int get_gc_cost(struct f2fs_sb_info *sbi, unsigned int segno,
209	struct victim_sel_policy *p)
210	{
211	if (p->alloc_mode == SSR)
212	return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
213
214	/* alloc_mode == LFS */
215	if (p->gc_mode == GC_GREEDY)
216	return get_valid_blocks(sbi, segno, sbi->segs_per_sec);
217	else
218	return get_cb_cost(sbi, segno);
219	}
220
0a8165d7	221	/*
7bc09003 JK	222	* This function is called from two pathes.
	223	* One is garbage collection and the other is SSR segment selection.
	224	* When it is called during GC, it just gets a victim segment
	225	* and it does not remove it from dirty seglist.
	226	* When it is called from SSR segment selection, it finds a segment
	227	* which has minimum valid blocks and removes it from dirty seglist.
	228	*/
	229	static int get_victim_by_default(struct f2fs_sb_info *sbi,
	230	unsigned int *result, int gc_type, int type, char alloc_mode)
	231	{
	232	struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
	233	struct victim_sel_policy p;
	234	unsigned int segno;
	235	int nsearched = 0;
	236
	237	p.alloc_mode = alloc_mode;
	238	select_policy(sbi, gc_type, type, &p);
	239
	240	p.min_segno = NULL_SEGNO;
	241	p.min_cost = get_max_cost(sbi, &p);
	242
	243	mutex_lock(&dirty_i->seglist_lock);
	244
	245	if (p.alloc_mode == LFS && gc_type == FG_GC) {
	246	p.min_segno = check_bg_victims(sbi);
	247	if (p.min_segno != NULL_SEGNO)
	248	goto got_it;
	249	}
	250
	251	while (1) {
	252	unsigned long cost;
	253
	254	segno = find_next_bit(p.dirty_segmap,
	255	TOTAL_SEGS(sbi), p.offset);
	256	if (segno >= TOTAL_SEGS(sbi)) {
	257	if (sbi->last_victim[p.gc_mode]) {
	258	sbi->last_victim[p.gc_mode] = 0;
	259	p.offset = 0;
	260	continue;
	261	}
	262	break;
	263	}
	264	p.offset = ((segno / p.ofs_unit) * p.ofs_unit) + p.ofs_unit;
	265
	266	if (test_bit(segno, dirty_i->victim_segmap[FG_GC]))
	267	continue;
	268	if (gc_type == BG_GC &&
	269	test_bit(segno, dirty_i->victim_segmap[BG_GC]))
	270	continue;
	271	if (IS_CURSEC(sbi, GET_SECNO(sbi, segno)))
	272	continue;
	273
	274	cost = get_gc_cost(sbi, segno, &p);
	275
	276	if (p.min_cost > cost) {
	277	p.min_segno = segno;
	278	p.min_cost = cost;
	279	}
	280
	281	if (cost == get_max_cost(sbi, &p))
	282	continue;
	283
	284	if (nsearched++ >= MAX_VICTIM_SEARCH) {
	285	sbi->last_victim[p.gc_mode] = segno;
286	break;
287	}
288	}
289	got_it:
290	if (p.min_segno != NULL_SEGNO) {
291	result = (p.min_segno / p.ofs_unit) p.ofs_unit;
292	if (p.alloc_mode == LFS) {
293	int i;
294	for (i = 0; i < p.ofs_unit; i++)
295	set_bit(*result + i,
296	dirty_i->victim_segmap[gc_type]);
297	}
298	}
299	mutex_unlock(&dirty_i->seglist_lock);
300
301	return (p.min_segno == NULL_SEGNO) ? 0 : 1;
302	}
303
304	static const struct victim_selection default_v_ops = {
305	.get_victim = get_victim_by_default,
306	};
307
308	static struct inode find_gc_inode(nid_t ino, struct list_head ilist)
309	{
310	struct list_head *this;
311	struct inode_entry *ie;
312
313	list_for_each(this, ilist) {
314	ie = list_entry(this, struct inode_entry, list);
315	if (ie->inode->i_ino == ino)
316	return ie->inode;
317	}
318	return NULL;
319	}
320
321	static void add_gc_inode(struct inode inode, struct list_head ilist)
322	{
323	struct list_head *this;
324	struct inode_entry new_ie, ie;
325
326	list_for_each(this, ilist) {
327	ie = list_entry(this, struct inode_entry, list);
328	if (ie->inode == inode) {
329	iput(inode);
330	return;
331	}
332	}
333	repeat:
334	new_ie = kmem_cache_alloc(winode_slab, GFP_NOFS);
335	if (!new_ie) {
336	cond_resched();
337	goto repeat;
338	}
339	new_ie->inode = inode;
340	list_add_tail(&new_ie->list, ilist);
341	}
342
343	static void put_gc_inode(struct list_head *ilist)
344	{
345	struct inode_entry ie, next_ie;
346	list_for_each_entry_safe(ie, next_ie, ilist, list) {
347	iput(ie->inode);
348	list_del(&ie->list);
349	kmem_cache_free(winode_slab, ie);
350	}
351	}
352
353	static int check_valid_map(struct f2fs_sb_info *sbi,
354	unsigned int segno, int offset)
355	{
356	struct sit_info *sit_i = SIT_I(sbi);
357	struct seg_entry *sentry;
358	int ret;
359
360	mutex_lock(&sit_i->sentry_lock);
361	sentry = get_seg_entry(sbi, segno);
362	ret = f2fs_test_bit(offset, sentry->cur_valid_map);
363	mutex_unlock(&sit_i->sentry_lock);
43727527	364	return ret;
7bc09003 JK	365	}
7bc09003 JK	366
0a8165d7	367	/*
7bc09003 JK	368	* This function compares node address got in summary with that in NAT.
	369	* On validity, copy that node with cold status, otherwise (invalid node)
	370	* ignore that.
	371	*/
43727527	372	static void gc_node_segment(struct f2fs_sb_info *sbi,
7bc09003 JK	373	struct f2fs_summary *sum, unsigned int segno, int gc_type)
	374	{
	375	bool initial = true;
	376	struct f2fs_summary *entry;
	377	int off;
	378
	379	next_step:
	380	entry = sum;
	381	for (off = 0; off < sbi->blocks_per_seg; off++, entry++) {
	382	nid_t nid = le32_to_cpu(entry->nid);
	383	struct page *node_page;
7bc09003	384
43727527 JK	385	/* stop BG_GC if there is not enough free sections. */
	386	if (gc_type == BG_GC && has_not_enough_free_secs(sbi, 0))
	387	return;
7bc09003	388
43727527	389	if (check_valid_map(sbi, segno, off) == 0)
7bc09003 JK	390	continue;
	391
	392	if (initial) {
	393	ra_node_page(sbi, nid);
	394	continue;
	395	}
	396	node_page = get_node_page(sbi, nid);
	397	if (IS_ERR(node_page))
	398	continue;
	399
	400	/* set page dirty and write it */
	401	if (!PageWriteback(node_page))
	402	set_page_dirty(node_page);
	403	f2fs_put_page(node_page, 1);
	404	stat_inc_node_blk_count(sbi, 1);
	405	}
	406	if (initial) {
	407	initial = false;
	408	goto next_step;
	409	}
	410
	411	if (gc_type == FG_GC) {
	412	struct writeback_control wbc = {
	413	.sync_mode = WB_SYNC_ALL,
	414	.nr_to_write = LONG_MAX,
	415	.for_reclaim = 0,
	416	};
	417	sync_node_pages(sbi, 0, &wbc);
	418	}
7bc09003 JK	419	}
7bc09003 JK	420
0a8165d7	421	/*
9af45ef5 JK	422	* Calculate start block index indicating the given node offset.
	423	* Be careful, caller should give this node offset only indicating direct node
	424	* blocks. If any node offsets, which point the other types of node blocks such
	425	* as indirect or double indirect node blocks, are given, it must be a caller's
	426	* bug.
7bc09003 JK	427	*/
	428	block_t start_bidx_of_node(unsigned int node_ofs)
	429	{
ce19a5d4 JK	430	unsigned int indirect_blks = 2 * NIDS_PER_BLOCK + 4;
ce19a5d4 JK	431	unsigned int bidx;
7bc09003	432
ce19a5d4 JK	433	if (node_ofs == 0)
ce19a5d4 JK	434	return 0;
7bc09003	435
ce19a5d4	436	if (node_ofs <= 2) {
7bc09003 JK	437	bidx = node_ofs - 1;
7bc09003 JK	438	} else if (node_ofs <= indirect_blks) {
ce19a5d4	439	int dec = (node_ofs - 4) / (NIDS_PER_BLOCK + 1);
7bc09003 JK	440	bidx = node_ofs - 2 - dec;
7bc09003 JK	441	} else {
ce19a5d4	442	int dec = (node_ofs - indirect_blks - 3) / (NIDS_PER_BLOCK + 1);
7bc09003 JK	443	bidx = node_ofs - 5 - dec;
7bc09003 JK	444	}
ce19a5d4	445	return bidx * ADDRS_PER_BLOCK + ADDRS_PER_INODE;
7bc09003 JK	446	}
	447
	448	static int check_dnode(struct f2fs_sb_info sbi, struct f2fs_summary sum,
	449	struct node_info dni, block_t blkaddr, unsigned int nofs)
	450	{
	451	struct page *node_page;
	452	nid_t nid;
	453	unsigned int ofs_in_node;
	454	block_t source_blkaddr;
	455
	456	nid = le32_to_cpu(sum->nid);
	457	ofs_in_node = le16_to_cpu(sum->ofs_in_node);
	458
	459	node_page = get_node_page(sbi, nid);
	460	if (IS_ERR(node_page))
43727527	461	return 0;
7bc09003 JK	462
	463	get_node_info(sbi, nid, dni);
	464
	465	if (sum->version != dni->version) {
	466	f2fs_put_page(node_page, 1);
43727527	467	return 0;
7bc09003 JK	468	}
	469
	470	*nofs = ofs_of_node(node_page);
	471	source_blkaddr = datablock_addr(node_page, ofs_in_node);
	472	f2fs_put_page(node_page, 1);
	473
	474	if (source_blkaddr != blkaddr)
43727527 JK	475	return 0;
43727527 JK	476	return 1;
7bc09003 JK	477	}
	478
	479	static void move_data_page(struct inode inode, struct page page, int gc_type)
	480	{
	481	if (page->mapping != inode->i_mapping)
	482	goto out;
	483
	484	if (inode != page->mapping->host)
	485	goto out;
	486
	487	if (PageWriteback(page))
	488	goto out;
	489
	490	if (gc_type == BG_GC) {
	491	set_page_dirty(page);
	492	set_cold_data(page);
	493	} else {
	494	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
	495	mutex_lock_op(sbi, DATA_WRITE);
	496	if (clear_page_dirty_for_io(page) &&
	497	S_ISDIR(inode->i_mode)) {
	498	dec_page_count(sbi, F2FS_DIRTY_DENTS);
	499	inode_dec_dirty_dents(inode);
	500	}
	501	set_cold_data(page);
	502	do_write_data_page(page);
	503	mutex_unlock_op(sbi, DATA_WRITE);
	504	clear_cold_data(page);
	505	}
	506	out:
	507	f2fs_put_page(page, 1);
	508	}
	509
0a8165d7	510	/*
7bc09003 JK	511	* This function tries to get parent node of victim data block, and identifies
	512	* data block validity. If the block is valid, copy that with cold status and
	513	* modify parent node.
	514	* If the parent node is not valid or the data block address is different,
	515	* the victim data block is ignored.
	516	*/
43727527	517	static void gc_data_segment(struct f2fs_sb_info sbi, struct f2fs_summary sum,
7bc09003 JK	518	struct list_head *ilist, unsigned int segno, int gc_type)
	519	{
	520	struct super_block *sb = sbi->sb;
	521	struct f2fs_summary *entry;
	522	block_t start_addr;
43727527	523	int off;
7bc09003 JK	524	int phase = 0;
	525
	526	start_addr = START_BLOCK(sbi, segno);
	527
	528	next_step:
	529	entry = sum;
	530	for (off = 0; off < sbi->blocks_per_seg; off++, entry++) {
	531	struct page *data_page;
	532	struct inode *inode;
	533	struct node_info dni; /* dnode info for the data */
	534	unsigned int ofs_in_node, nofs;
	535	block_t start_bidx;
	536
43727527 JK	537	/* stop BG_GC if there is not enough free sections. */
	538	if (gc_type == BG_GC && has_not_enough_free_secs(sbi, 0))
	539	return;
7bc09003	540
43727527	541	if (check_valid_map(sbi, segno, off) == 0)
7bc09003 JK	542	continue;
	543
	544	if (phase == 0) {
	545	ra_node_page(sbi, le32_to_cpu(entry->nid));
	546	continue;
	547	}
	548
	549	/* Get an inode by ino with checking validity */
43727527	550	if (check_dnode(sbi, entry, &dni, start_addr + off, &nofs) == 0)
7bc09003 JK	551	continue;
	552
	553	if (phase == 1) {
	554	ra_node_page(sbi, dni.ino);
	555	continue;
	556	}
	557
	558	start_bidx = start_bidx_of_node(nofs);
	559	ofs_in_node = le16_to_cpu(entry->ofs_in_node);
	560
	561	if (phase == 2) {
d4686d56	562	inode = f2fs_iget(sb, dni.ino);
7bc09003 JK	563	if (IS_ERR(inode))
	564	continue;
	565
	566	data_page = find_data_page(inode,
	567	start_bidx + ofs_in_node);
	568	if (IS_ERR(data_page))
	569	goto next_iput;
	570
	571	f2fs_put_page(data_page, 0);
	572	add_gc_inode(inode, ilist);
	573	} else {
	574	inode = find_gc_inode(dni.ino, ilist);
	575	if (inode) {
	576	data_page = get_lock_data_page(inode,
	577	start_bidx + ofs_in_node);
	578	if (IS_ERR(data_page))
	579	continue;
	580	move_data_page(inode, data_page, gc_type);
	581	stat_inc_data_blk_count(sbi, 1);
	582	}
	583	}
	584	continue;
	585	next_iput:
	586	iput(inode);
	587	}
	588	if (++phase < 4)
	589	goto next_step;
43727527	590
7bc09003 JK	591	if (gc_type == FG_GC)
7bc09003 JK	592	f2fs_submit_bio(sbi, DATA, true);
7bc09003 JK	593	}
	594
	595	static int __get_victim(struct f2fs_sb_info sbi, unsigned int victim,
	596	int gc_type, int type)
	597	{
	598	struct sit_info *sit_i = SIT_I(sbi);
	599	int ret;
	600	mutex_lock(&sit_i->sentry_lock);
	601	ret = DIRTY_I(sbi)->v_ops->get_victim(sbi, victim, gc_type, type, LFS);
	602	mutex_unlock(&sit_i->sentry_lock);
	603	return ret;
	604	}
	605
43727527	606	static void do_garbage_collect(struct f2fs_sb_info *sbi, unsigned int segno,
7bc09003 JK	607	struct list_head *ilist, int gc_type)
	608	{
	609	struct page *sum_page;
	610	struct f2fs_summary_block *sum;
7bc09003 JK	611
	612	/* read segment summary of victim */
	613	sum_page = get_sum_page(sbi, segno);
	614	if (IS_ERR(sum_page))
43727527	615	return;
7bc09003 JK	616
	617	/*
	618	* CP needs to lock sum_page. In this time, we don't need
	619	* to lock this page, because this summary page is not gone anywhere.
	620	* Also, this page is not gonna be updated before GC is done.
	621	*/
	622	unlock_page(sum_page);
	623	sum = page_address(sum_page);
	624
	625	switch (GET_SUM_TYPE((&sum->footer))) {
	626	case SUM_TYPE_NODE:
43727527	627	gc_node_segment(sbi, sum->entries, segno, gc_type);
7bc09003 JK	628	break;
7bc09003 JK	629	case SUM_TYPE_DATA:
43727527	630	gc_data_segment(sbi, sum->entries, ilist, segno, gc_type);
7bc09003 JK	631	break;
	632	}
	633	stat_inc_seg_count(sbi, GET_SUM_TYPE((&sum->footer)));
	634	stat_inc_call_count(sbi->stat_info);
	635
	636	f2fs_put_page(sum_page, 0);
7bc09003 JK	637	}
7bc09003 JK	638
408e9375	639	int f2fs_gc(struct f2fs_sb_info *sbi)
7bc09003	640	{
7bc09003	641	struct list_head ilist;
408e9375	642	unsigned int segno, i;
7bc09003	643	int gc_type = BG_GC;
43727527 JK	644	int nfree = 0;
43727527 JK	645	int ret = -1;
7bc09003 JK	646
	647	INIT_LIST_HEAD(&ilist);
	648	gc_more:
408e9375 JK	649	if (!(sbi->sb->s_flags & MS_ACTIVE))
408e9375 JK	650	goto stop;
7bc09003	651
43727527	652	if (gc_type == BG_GC && has_not_enough_free_secs(sbi, nfree))
408e9375	653	gc_type = FG_GC;
7bc09003	654
408e9375 JK	655	if (!__get_victim(sbi, &segno, gc_type, NO_CHECK_TYPE))
408e9375 JK	656	goto stop;
43727527	657	ret = 0;
7bc09003	658
43727527 JK	659	for (i = 0; i < sbi->segs_per_sec; i++)
	660	do_garbage_collect(sbi, segno + i, &ilist, gc_type);
	661
	662	if (gc_type == FG_GC &&
	663	get_valid_blocks(sbi, segno, sbi->segs_per_sec) == 0)
	664	nfree++;
	665
	666	if (has_not_enough_free_secs(sbi, nfree))
	667	goto gc_more;
	668
	669	if (gc_type == FG_GC)
	670	write_checkpoint(sbi, false);
408e9375	671	stop:
7bc09003 JK	672	mutex_unlock(&sbi->gc_mutex);
	673
	674	put_gc_inode(&ilist);
43727527	675	return ret;
7bc09003 JK	676	}
	677
	678	void build_gc_manager(struct f2fs_sb_info *sbi)
	679	{
	680	DIRTY_I(sbi)->v_ops = &default_v_ops;
	681	}
	682
6e6093a8	683	int __init create_gc_caches(void)
7bc09003 JK	684	{
	685	winode_slab = f2fs_kmem_cache_create("f2fs_gc_inodes",
	686	sizeof(struct inode_entry), NULL);
	687	if (!winode_slab)
	688	return -ENOMEM;
	689	return 0;
	690	}
	691
	692	void destroy_gc_caches(void)
	693	{
	694	kmem_cache_destroy(winode_slab);
	695	}