[PATCH 05/16] f2fs: add checkpoint operations

[PATCH 05/16] f2fs: add checkpoint operations

[김재극]

This adds functions required by the checkpoint operations.

Basically, f2fs adopts a roll-back model with checkpoint blocks written in the
CP area. The checkpoint procedure includes as follows.

- write_checkpoint()
1. block_operations() freezes VFS calls.
2. submit cached bios.
3. flush_nat_entries() writes NAT pages updated by dirty NAT entries.
4. flush_sit_entries() writes SIT pages updated by dirty SIT entries.
5. do_checkpoint() writes,
  - checkpoint block (#0)
  - orphan inode blocks
  - summary blocks made by active logs
  - checkpoint block (copy of #0)
6. unblock_opeations()

In order to provide an address space for meta pages, f2fs_sb_info has a special
inode, namely meta_inode. This patch also adds the address space operations for
meta_inode.

Signed-off-by: Chul Lee <chur.lee@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
---
 fs/f2fs/checkpoint.c |  791 ++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 791 insertions(+)
 create mode 100644 fs/f2fs/checkpoint.c

diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
new file mode 100644
index 0000000..2186b82
--- /dev/null
+++ b/fs/f2fs/checkpoint.c
@@ -0,0 +1,791 @@
+/**
+ * fs/f2fs/checkpoint.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/bio.h>
+#include <linux/mpage.h>
+#include <linux/writeback.h>
+#include <linux/blkdev.h>
+#include <linux/f2fs_fs.h>
+#include <linux/pagevec.h>
+#include <linux/swap.h>
+
+#include "f2fs.h"
+#include "node.h"
+#include "segment.h"
+
+static struct kmem_cache *orphan_entry_slab;
+static struct kmem_cache *inode_entry_slab;
+
+/**
+ * We guarantee no failure on the returned page.
+ */
+struct page *grab_meta_page(struct f2fs_sb_info *sbi, pgoff_t index)
+{
+	struct address_space *mapping = sbi->meta_inode->i_mapping;
+	struct page *page = NULL;
+repeat:
+	page = grab_cache_page(mapping, index);
+	if (!page) {
+		cond_resched();
+		goto repeat;
+	}
+
+	/* We wait writeback only inside grab_meta_page() */
+	wait_on_page_writeback(page);
+	SetPageUptodate(page);
+	return page;
+}
+
+/**
+ * We guarantee no failure on the returned page.
+ */
+struct page *get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index)
+{
+	struct address_space *mapping = sbi->meta_inode->i_mapping;
+	struct page *page;
+repeat:
+	page = grab_cache_page(mapping, index);
+	if (!page) {
+		cond_resched();
+		goto repeat;
+	}
+	if (f2fs_readpage(sbi, page, index, READ_SYNC)) {
+		f2fs_put_page(page, 1);
+		goto repeat;
+	}
+	mark_page_accessed(page);
+
+	/* We do not allow returning an errorneous page */
+	return page;
+}
+
+static int f2fs_write_meta_page(struct page *page,
+				struct writeback_control *wbc)
+{
+	struct inode *inode = page->mapping->host;
+	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
+	int err;
+
+	wait_on_page_writeback(page);
+
+	err = write_meta_page(sbi, page, wbc);
+	if (err) {
+		wbc->pages_skipped++;
+		set_page_dirty(page);
+	}
+
+	dec_page_count(sbi, F2FS_DIRTY_META);
+
+	/* In this case, we should not unlock this page */
+	if (err != AOP_WRITEPAGE_ACTIVATE)
+		unlock_page(page);
+	return err;
+}
+
+static int f2fs_write_meta_pages(struct address_space *mapping,
+				struct writeback_control *wbc)
+{
+	struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb);
+	struct block_device *bdev = sbi->sb->s_bdev;
+	long written;
+
+	if (wbc->for_kupdate)
+		return 0;
+
+	if (get_pages(sbi, F2FS_DIRTY_META) == 0)
+		return 0;
+
+	/* if mounting is failed, skip writing node pages */
+	mutex_lock(&sbi->cp_mutex);
+	written = sync_meta_pages(sbi, META, bio_get_nr_vecs(bdev));
+	mutex_unlock(&sbi->cp_mutex);
+	wbc->nr_to_write -= written;
+	return 0;
+}
+
+long sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type,
+						long nr_to_write)
+{
+	struct address_space *mapping = sbi->meta_inode->i_mapping;
+	pgoff_t index = 0, end = LONG_MAX;
+	struct pagevec pvec;
+	long nwritten = 0;
+	struct writeback_control wbc = {
+		.for_reclaim = 0,
+	};
+
+	pagevec_init(&pvec, 0);
+
+	while (index <= end) {
+		int i, nr_pages;
+		nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
+				PAGECACHE_TAG_DIRTY,
+				min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1);
+		if (nr_pages == 0)
+			break;
+
+		for (i = 0; i < nr_pages; i++) {
+			struct page *page = pvec.pages[i];
+			lock_page(page);
+			BUG_ON(page->mapping != mapping);
+			BUG_ON(!PageDirty(page));
+			clear_page_dirty_for_io(page);
+			f2fs_write_meta_page(page, &wbc);
+			if (nwritten++ >= nr_to_write)
+				break;
+		}
+		pagevec_release(&pvec);
+		cond_resched();
+	}
+
+	if (nwritten)
+		f2fs_submit_bio(sbi, type, nr_to_write == LONG_MAX);
+
+	return nwritten;
+}
+
+static int f2fs_set_meta_page_dirty(struct page *page)
+{
+	struct address_space *mapping = page->mapping;
+	struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb);
+
+	SetPageUptodate(page);
+	if (!PageDirty(page)) {
+		__set_page_dirty_nobuffers(page);
+		inc_page_count(sbi, F2FS_DIRTY_META);
+		F2FS_SET_SB_DIRT(sbi);
+		return 1;
+	}
+	return 0;
+}
+
+const struct address_space_operations f2fs_meta_aops = {
+	.writepage	= f2fs_write_meta_page,
+	.writepages	= f2fs_write_meta_pages,
+	.set_page_dirty	= f2fs_set_meta_page_dirty,
+};
+
+int check_orphan_space(struct f2fs_sb_info *sbi)
+{
+	unsigned int max_orphans;
+	int err = 0;
+
+	/*
+	 * considering 512 blocks in a segment 5 blocks are needed for cp
+	 * and log segment summaries. Remaining blocks are used to keep
+	 * orphan entries with the limitation one reserved segment
+	 * for cp pack we can have max 1020*507 orphan entries
+	 */
+	max_orphans = (sbi->blocks_per_seg - 5) * F2FS_ORPHANS_PER_BLOCK;
+	mutex_lock(&sbi->orphan_inode_mutex);
+	if (sbi->n_orphans >= max_orphans)
+		err = -ENOSPC;
+	mutex_unlock(&sbi->orphan_inode_mutex);
+	return err;
+}
+
+void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
+{
+	struct list_head *head, *this;
+	struct orphan_inode_entry *new = NULL, *orphan = NULL;
+
+	mutex_lock(&sbi->orphan_inode_mutex);
+	head = &sbi->orphan_inode_list;
+	list_for_each(this, head) {
+		orphan = list_entry(this, struct orphan_inode_entry, list);
+		if (orphan->ino == ino)
+			goto out;
+		if (orphan->ino > ino)
+			break;
+		orphan = NULL;
+	}
+retry:
+	new = kmem_cache_alloc(orphan_entry_slab, GFP_NOFS);
+	if (!new) {
+		cond_resched();
+		goto retry;
+	}
+	new->ino = ino;
+	INIT_LIST_HEAD(&new->list);
+
+	/* add new_oentry into list which is sorted by inode number */
+	if (orphan) {
+		struct orphan_inode_entry *prev;
+
+		/* get previous entry */
+		prev = list_entry(orphan->list.prev, typeof(*prev), list);
+		if (&prev->list != head)
+			/* insert new orphan inode entry */
+			list_add(&new->list, &prev->list);
+		else
+			list_add(&new->list, head);
+	} else {
+		list_add_tail(&new->list, head);
+	}
+	sbi->n_orphans++;
+out:
+	mutex_unlock(&sbi->orphan_inode_mutex);
+}
+
+void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
+{
+	struct list_head *this, *next, *head;
+	struct orphan_inode_entry *orphan;
+
+	mutex_lock(&sbi->orphan_inode_mutex);
+	head = &sbi->orphan_inode_list;
+	list_for_each_safe(this, next, head) {
+		orphan = list_entry(this, struct orphan_inode_entry, list);
+		if (orphan->ino == ino) {
+			list_del(&orphan->list);
+			kmem_cache_free(orphan_entry_slab, orphan);
+			sbi->n_orphans--;
+			break;
+		}
+	}
+	mutex_unlock(&sbi->orphan_inode_mutex);
+}
+
+static void recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
+{
+	struct inode *inode = f2fs_iget(sbi->sb, ino);
+	BUG_ON(IS_ERR(inode));
+	clear_nlink(inode);
+
+	/* truncate all the data during iput */
+	iput(inode);
+}
+
+int recover_orphan_inodes(struct f2fs_sb_info *sbi)
+{
+	block_t start_blk, orphan_blkaddr, i, j;
+
+	if (!(F2FS_CKPT(sbi)->ckpt_flags & CP_ORPHAN_PRESENT_FLAG))
+		return 0;
+
+	sbi->por_doing = 1;
+	start_blk = __start_cp_addr(sbi) + 1;
+	orphan_blkaddr = __start_sum_addr(sbi) - 1;
+
+	for (i = 0; i < orphan_blkaddr; i++) {
+		struct page *page = get_meta_page(sbi, start_blk + i);
+		struct f2fs_orphan_block *orphan_blk;
+
+		orphan_blk = (struct f2fs_orphan_block *)page_address(page);
+		for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) {
+			nid_t ino = le32_to_cpu(orphan_blk->ino[j]);
+			recover_orphan_inode(sbi, ino);
+		}
+		f2fs_put_page(page, 1);
+	}
+	/* clear Orphan Flag */
+	F2FS_CKPT(sbi)->ckpt_flags &= (~CP_ORPHAN_PRESENT_FLAG);
+	sbi->por_doing = 0;
+	return 0;
+}
+
+static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk)
+{
+	struct list_head *head, *this, *next;
+	struct f2fs_orphan_block *orphan_blk = NULL;
+	struct page *page = NULL;
+	unsigned int nentries = 0;
+	unsigned short index = 1;
+	unsigned short orphan_blocks;
+
+	orphan_blocks = (unsigned short)((sbi->n_orphans +
+		(F2FS_ORPHANS_PER_BLOCK - 1)) / F2FS_ORPHANS_PER_BLOCK);
+
+	mutex_lock(&sbi->orphan_inode_mutex);
+	head = &sbi->orphan_inode_list;
+
+	/* loop for each orphan inode entry and write them in Jornal block */
+	list_for_each_safe(this, next, head) {
+		struct orphan_inode_entry *orphan;
+
+		orphan = list_entry(this, struct orphan_inode_entry, list);
+
+		if (nentries == F2FS_ORPHANS_PER_BLOCK) {
+			/*
+			 * an orphan block is full of 1020 entries,
+			 * then we need to flush current orphan blocks
+			 * and bring another one in memory
+			 */
+			orphan_blk->blk_addr = cpu_to_le16(index);
+			orphan_blk->blk_count = cpu_to_le16(orphan_blocks);
+			orphan_blk->entry_count = cpu_to_le32(nentries);
+			set_page_dirty(page);
+			f2fs_put_page(page, 1);
+			index++;
+			start_blk++;
+			nentries = 0;
+			page = NULL;
+		}
+		if (page)
+			goto page_exist;
+
+		page = grab_meta_page(sbi, start_blk);
+		orphan_blk = (struct f2fs_orphan_block *)page_address(page);
+		memset(orphan_blk, 0, sizeof(*orphan_blk));
+page_exist:
+		orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino);
+	}
+	if (!page)
+		goto end;
+
+	orphan_blk->blk_addr = cpu_to_le16(index);
+	orphan_blk->blk_count = cpu_to_le16(orphan_blocks);
+	orphan_blk->entry_count = cpu_to_le32(nentries);
+	set_page_dirty(page);
+	f2fs_put_page(page, 1);
+end:
+	mutex_unlock(&sbi->orphan_inode_mutex);
+}
+
+static struct page *validate_checkpoint(struct f2fs_sb_info *sbi,
+				block_t cp_addr, unsigned long long *version)
+{
+	struct page *cp_page_1, *cp_page_2 = NULL;
+	unsigned long blk_size = sbi->blocksize;
+	struct f2fs_checkpoint *cp_block;
+	unsigned long long cur_version = 0, pre_version = 0;
+	unsigned int crc = 0;
+	size_t crc_offset;
+
+	/* Read the 1st cp block in this CP pack */
+	cp_page_1 = get_meta_page(sbi, cp_addr);
+
+	/* get the version number */
+	cp_block = (struct f2fs_checkpoint *)page_address(cp_page_1);
+	crc_offset = le32_to_cpu(cp_block->checksum_offset);
+	if (crc_offset >= blk_size)
+		goto invalid_cp1;
+
+	crc = *(unsigned int *)((unsigned char *)cp_block + crc_offset);
+	if (!f2fs_crc_valid(crc, cp_block, crc_offset))
+		goto invalid_cp1;
+
+	pre_version = le64_to_cpu(cp_block->checkpoint_ver);
+
+	/* Read the 2nd cp block in this CP pack */
+	cp_addr += le64_to_cpu(cp_block->cp_pack_total_block_count) - 1;
+	cp_page_2 = get_meta_page(sbi, cp_addr);
+
+	cp_block = (struct f2fs_checkpoint *)page_address(cp_page_2);
+	crc_offset = le32_to_cpu(cp_block->checksum_offset);
+	if (crc_offset >= blk_size)
+		goto invalid_cp2;
+
+	crc = *(unsigned int *)((unsigned char *)cp_block + crc_offset);
+	if (!f2fs_crc_valid(crc, cp_block, crc_offset))
+		goto invalid_cp2;
+
+	cur_version = le64_to_cpu(cp_block->checkpoint_ver);
+
+	if (cur_version == pre_version) {
+		*version = cur_version;
+		f2fs_put_page(cp_page_2, 1);
+		return cp_page_1;
+	}
+invalid_cp2:
+	f2fs_put_page(cp_page_2, 1);
+invalid_cp1:
+	f2fs_put_page(cp_page_1, 1);
+	return NULL;
+}
+
+int get_valid_checkpoint(struct f2fs_sb_info *sbi)
+{
+	struct f2fs_checkpoint *cp_block;
+	struct f2fs_super_block *fsb = sbi->raw_super;
+	struct page *cp1, *cp2, *cur_page;
+	unsigned long blk_size = sbi->blocksize;
+	unsigned long long cp1_version = 0, cp2_version = 0;
+	unsigned long long cp_start_blk_no;
+
+	sbi->ckpt = kzalloc(blk_size, GFP_KERNEL);
+	if (!sbi->ckpt)
+		return -ENOMEM;
+	/*
+	 * Finding out valid cp block involves read both
+	 * sets( cp pack1 and cp pack 2)
+	 */
+	cp_start_blk_no = le64_to_cpu(fsb->start_segment_checkpoint);
+	cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version);
+
+	/* The second checkpoint pack should start at the next segment */
+	cp_start_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg);
+	cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version);
+
+	if (cp1 && cp2) {
+		if (ver_after(cp2_version, cp1_version))
+			cur_page = cp2;
+		else
+			cur_page = cp1;
+	} else if (cp1) {
+		cur_page = cp1;
+	} else if (cp2) {
+		cur_page = cp2;
+	} else {
+		goto fail_no_cp;
+	}
+
+	cp_block = (struct f2fs_checkpoint *)page_address(cur_page);
+	memcpy(sbi->ckpt, cp_block, blk_size);
+
+	f2fs_put_page(cp1, 1);
+	f2fs_put_page(cp2, 1);
+	return 0;
+
+fail_no_cp:
+	kfree(sbi->ckpt);
+	return -EINVAL;
+}
+
+void set_dirty_dir_page(struct inode *inode, struct page *page)
+{
+	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
+	struct list_head *head = &sbi->dir_inode_list;
+	struct dir_inode_entry *new;
+	struct list_head *this;
+
+	if (!S_ISDIR(inode->i_mode))
+		return;
+
+	new = kmem_cache_alloc(inode_entry_slab, __GFP_HIGH | __GFP_NOFAIL);
+	new->inode = inode;
+	INIT_LIST_HEAD(&new->list);
+
+	spin_lock(&sbi->dir_inode_lock);
+	list_for_each(this, head) {
+		struct dir_inode_entry *entry;
+		entry = list_entry(this, struct dir_inode_entry, list);
+		if (entry->inode == inode) {
+			kmem_cache_free(inode_entry_slab, new);
+			goto out;
+		}
+	}
+	list_add_tail(&new->list, head);
+	sbi->n_dirty_dirs++;
+
+	BUG_ON(!S_ISDIR(inode->i_mode));
+out:
+	inc_page_count(sbi, F2FS_DIRTY_DENTS);
+	inode_inc_dirty_dents(inode);
+	SetPagePrivate(page);
+
+	spin_unlock(&sbi->dir_inode_lock);
+}
+
+void remove_dirty_dir_inode(struct inode *inode)
+{
+	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
+	struct list_head *head = &sbi->dir_inode_list;
+	struct list_head *this;
+
+	if (!S_ISDIR(inode->i_mode))
+		return;
+
+	spin_lock(&sbi->dir_inode_lock);
+	if (atomic_read(&F2FS_I(inode)->dirty_dents))
+		goto out;
+
+	list_for_each(this, head) {
+		struct dir_inode_entry *entry;
+		entry = list_entry(this, struct dir_inode_entry, list);
+		if (entry->inode == inode) {
+			list_del(&entry->list);
+			kmem_cache_free(inode_entry_slab, entry);
+			sbi->n_dirty_dirs--;
+			break;
+		}
+	}
+out:
+	spin_unlock(&sbi->dir_inode_lock);
+}
+
+void sync_dirty_dir_inodes(struct f2fs_sb_info *sbi)
+{
+	struct list_head *head = &sbi->dir_inode_list;
+	struct dir_inode_entry *entry;
+	struct inode *inode;
+retry:
+	spin_lock(&sbi->dir_inode_lock);
+	if (list_empty(head)) {
+		spin_unlock(&sbi->dir_inode_lock);
+		return;
+	}
+	entry = list_entry(head->next, struct dir_inode_entry, list);
+	inode = igrab(entry->inode);
+	spin_unlock(&sbi->dir_inode_lock);
+	if (inode) {
+		filemap_flush(inode->i_mapping);
+		iput(inode);
+	} else {
+		/*
+		 * We should submit bio, since it exists several
+		 * wribacking dentry pages in the freeing inode.
+		 */
+		f2fs_submit_bio(sbi, DATA, true);
+	}
+	goto retry;
+}
+
+/**
+ * Freeze all the FS-operations for checkpoint.
+ */
+void block_operations(struct f2fs_sb_info *sbi)
+{
+	int t;
+	struct writeback_control wbc = {
+		.sync_mode = WB_SYNC_ALL,
+		.nr_to_write = LONG_MAX,
+		.for_reclaim = 0,
+	};
+
+	/* Stop renaming operation */
+	mutex_lock_op(sbi, RENAME);
+	mutex_lock_op(sbi, DENTRY_OPS);
+
+retry_dents:
+	/* write all the dirty dentry pages */
+	sync_dirty_dir_inodes(sbi);
+
+	mutex_lock_op(sbi, DATA_WRITE);
+	if (get_pages(sbi, F2FS_DIRTY_DENTS)) {
+		mutex_unlock_op(sbi, DATA_WRITE);
+		goto retry_dents;
+	}
+
+	/* block all the operations */
+	for (t = DATA_NEW; t <= NODE_TRUNC; t++)
+		mutex_lock_op(sbi, t);
+
+	mutex_lock(&sbi->write_inode);
+
+	/*
+	 * POR: we should ensure that there is no dirty node pages
+	 * until finishing nat/sit flush.
+	 */
+retry:
+	sync_node_pages(sbi, 0, &wbc);
+
+	mutex_lock_op(sbi, NODE_WRITE);
+
+	if (get_pages(sbi, F2FS_DIRTY_NODES)) {
+		mutex_unlock_op(sbi, NODE_WRITE);
+		goto retry;
+	}
+	mutex_unlock(&sbi->write_inode);
+}
+
+static void unblock_operations(struct f2fs_sb_info *sbi)
+{
+	int t;
+	for (t = NODE_WRITE; t >= RENAME; t--)
+		mutex_unlock_op(sbi, t);
+}
+
+static void do_checkpoint(struct f2fs_sb_info *sbi, bool is_umount)
+{
+	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
+	nid_t last_nid = 0;
+	int nat_upd_blkoff[3];
+	block_t start_blk;
+	struct page *cp_page;
+	unsigned int data_sum_blocks, orphan_blocks;
+	void *kaddr;
+	__u32 crc32 = 0;
+	int i;
+
+	/* Flush all the NAT/SIT pages */
+	while (get_pages(sbi, F2FS_DIRTY_META))
+		sync_meta_pages(sbi, META, LONG_MAX);
+
+	next_free_nid(sbi, &last_nid);
+
+	/*
+	 * modify checkpoint
+	 * version number is already updated
+	 */
+	ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi));
+	ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi));
+	ckpt->free_segment_count = cpu_to_le32(free_segments(sbi));
+	for (i = 0; i < 3; i++) {
+		ckpt->cur_node_segno[i] =
+			cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE));
+		ckpt->cur_node_blkoff[i] =
+			cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE));
+		nat_upd_blkoff[i] = NM_I(sbi)->nat_upd_blkoff[i];
+		ckpt->nat_upd_blkoff[i] = cpu_to_le16(nat_upd_blkoff[i]);
+		ckpt->alloc_type[i + CURSEG_HOT_NODE] =
+				curseg_alloc_type(sbi, i + CURSEG_HOT_NODE);
+	}
+	for (i = 0; i < 3; i++) {
+		ckpt->cur_data_segno[i] =
+			cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA));
+		ckpt->cur_data_blkoff[i] =
+			cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA));
+		ckpt->alloc_type[i + CURSEG_HOT_DATA] =
+				curseg_alloc_type(sbi, i + CURSEG_HOT_DATA);
+	}
+
+	ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi));
+	ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi));
+	ckpt->next_free_nid = cpu_to_le32(last_nid);
+
+	/* 2 cp  + n data seg summary + orphan inode blocks */
+	data_sum_blocks = npages_for_summary_flush(sbi);
+	if (data_sum_blocks < 3)
+		ckpt->ckpt_flags |= CP_COMPACT_SUM_FLAG;
+	else
+		ckpt->ckpt_flags &= (~CP_COMPACT_SUM_FLAG);
+
+	orphan_blocks = (sbi->n_orphans + F2FS_ORPHANS_PER_BLOCK - 1)
+					/ F2FS_ORPHANS_PER_BLOCK;
+	ckpt->cp_pack_start_sum = 1 + orphan_blocks;
+	ckpt->cp_pack_total_block_count = 2 + data_sum_blocks + orphan_blocks;
+
+	if (is_umount) {
+		ckpt->ckpt_flags |= CP_UMOUNT_FLAG;
+		ckpt->cp_pack_total_block_count += NR_CURSEG_NODE_TYPE;
+	} else {
+		ckpt->ckpt_flags &= (~CP_UMOUNT_FLAG);
+	}
+
+	if (sbi->n_orphans)
+		ckpt->ckpt_flags |= CP_ORPHAN_PRESENT_FLAG;
+	else
+		ckpt->ckpt_flags &= (~CP_ORPHAN_PRESENT_FLAG);
+
+	/* update SIT/NAT bitmap */
+	get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP));
+	get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP));
+
+	crc32 = f2fs_crc32(ckpt, le32_to_cpu(ckpt->checksum_offset));
+	*(__u32 *)((unsigned char *)ckpt +
+				le32_to_cpu(ckpt->checksum_offset))
+				= cpu_to_le32(crc32);
+
+	start_blk = __start_cp_addr(sbi);
+
+	/* write out checkpoint buffer at block 0 */
+	cp_page = grab_meta_page(sbi, start_blk++);
+	kaddr = page_address(cp_page);
+	memcpy(kaddr, ckpt, (1 << sbi->log_blocksize));
+	set_page_dirty(cp_page);
+	f2fs_put_page(cp_page, 1);
+
+	if (sbi->n_orphans) {
+		write_orphan_inodes(sbi, start_blk);
+		start_blk += orphan_blocks;
+	}
+
+	write_data_summaries(sbi, start_blk);
+	start_blk += data_sum_blocks;
+	if (is_umount) {
+		write_node_summaries(sbi, start_blk);
+		start_blk += NR_CURSEG_NODE_TYPE;
+	}
+
+	/* writeout checkpoint block */
+	cp_page = grab_meta_page(sbi, start_blk);
+	kaddr = page_address(cp_page);
+	memcpy(kaddr, ckpt, (1 << sbi->log_blocksize));
+	set_page_dirty(cp_page);
+	f2fs_put_page(cp_page, 1);
+
+	/* wait for previous submitted node/meta pages writeback */
+	while (get_pages(sbi, F2FS_WRITEBACK))
+		congestion_wait(BLK_RW_ASYNC, HZ / 50);
+
+	filemap_fdatawait_range(sbi->node_inode->i_mapping, 0, LONG_MAX);
+	filemap_fdatawait_range(sbi->meta_inode->i_mapping, 0, LONG_MAX);
+
+	/* update user_block_counts */
+	sbi->last_valid_block_count = sbi->total_valid_block_count;
+	sbi->alloc_valid_block_count = 0;
+
+	/* Here, we only have one bio having CP pack */
+	if (sbi->ckpt->ckpt_flags & CP_ERROR_FLAG)
+		sbi->sb->s_flags |= MS_RDONLY;
+	else
+		sync_meta_pages(sbi, META_FLUSH, LONG_MAX);
+
+	clear_prefree_segments(sbi);
+	F2FS_RESET_SB_DIRT(sbi);
+}
+
+/**
+ * We guarantee that this checkpoint procedure should not fail.
+ */
+void write_checkpoint(struct f2fs_sb_info *sbi, bool blocked, bool is_umount)
+{
+	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
+	unsigned long long ckpt_ver;
+
+	if (!blocked) {
+		mutex_lock(&sbi->cp_mutex);
+		block_operations(sbi);
+	}
+
+	f2fs_submit_bio(sbi, DATA, true);
+	f2fs_submit_bio(sbi, NODE, true);
+	f2fs_submit_bio(sbi, META, true);
+
+	/*
+	 * update checkpoint pack index
+	 * Increase the version number so that
+	 * SIT entries and seg summaries are written at correct place
+	 */
+	ckpt_ver = le64_to_cpu(ckpt->checkpoint_ver);
+	ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver);
+
+	/* write cached NAT/SIT entries to NAT/SIT area */
+	flush_nat_entries(sbi);
+	flush_sit_entries(sbi);
+
+	reset_victim_segmap(sbi);
+
+	/* unlock all the fs_lock[] in do_checkpoint() */
+	do_checkpoint(sbi, is_umount);
+
+	unblock_operations(sbi);
+	mutex_unlock(&sbi->cp_mutex);
+}
+
+void init_orphan_info(struct f2fs_sb_info *sbi)
+{
+	mutex_init(&sbi->orphan_inode_mutex);
+	INIT_LIST_HEAD(&sbi->orphan_inode_list);
+	sbi->n_orphans = 0;
+}
+
+int create_checkpoint_caches(void)
+{
+	orphan_entry_slab = f2fs_kmem_cache_create("f2fs_orphan_entry",
+			sizeof(struct orphan_inode_entry), NULL);
+	if (unlikely(!orphan_entry_slab))
+		return -ENOMEM;
+	inode_entry_slab = f2fs_kmem_cache_create("f2fs_dirty_dir_entry",
+			sizeof(struct dir_inode_entry), NULL);
+	if (unlikely(!inode_entry_slab)) {
+		kmem_cache_destroy(orphan_entry_slab);
+		return -ENOMEM;
+	}
+	return 0;
+}
+
+void destroy_checkpoint_caches(void)
+{
+	kmem_cache_destroy(orphan_entry_slab);
+	kmem_cache_destroy(inode_entry_slab);
+}
-- 
1.7.9.5




---
Jaegeuk Kim
Samsung

Re: [PATCH 05/16] f2fs: add checkpoint operations

[NeilBrown]

[-- Attachment #1: Type: text/plain, Size: 3195 bytes --]

On Fri, 05 Oct 2012 20:59:29 +0900 김재극 <jaegeuk.kim@samsung.com> wrote:

> +static void do_checkpoint(struct f2fs_sb_info *sbi, bool is_umount)
> +{
> +	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
> +	nid_t last_nid = 0;
> +	int nat_upd_blkoff[3];
> +	block_t start_blk;
> +	struct page *cp_page;
> +	unsigned int data_sum_blocks, orphan_blocks;
> +	void *kaddr;
> +	__u32 crc32 = 0;
> +	int i;
> +
> +	/* Flush all the NAT/SIT pages */
> +	while (get_pages(sbi, F2FS_DIRTY_META))
> +		sync_meta_pages(sbi, META, LONG_MAX);
> +
> +	next_free_nid(sbi, &last_nid);
> +
> +	/*
> +	 * modify checkpoint
> +	 * version number is already updated
> +	 */
> +	ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi));
> +	ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi));
> +	ckpt->free_segment_count = cpu_to_le32(free_segments(sbi));
> +	for (i = 0; i < 3; i++) {
> +		ckpt->cur_node_segno[i] =
> +			cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE));
> +		ckpt->cur_node_blkoff[i] =
> +			cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE));
> +		nat_upd_blkoff[i] = NM_I(sbi)->nat_upd_blkoff[i];
> +		ckpt->nat_upd_blkoff[i] = cpu_to_le16(nat_upd_blkoff[i]);
> +		ckpt->alloc_type[i + CURSEG_HOT_NODE] =
> +				curseg_alloc_type(sbi, i + CURSEG_HOT_NODE);
> +	}
> +	for (i = 0; i < 3; i++) {
> +		ckpt->cur_data_segno[i] =
> +			cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA));
> +		ckpt->cur_data_blkoff[i] =
> +			cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA));
> +		ckpt->alloc_type[i + CURSEG_HOT_DATA] =
> +				curseg_alloc_type(sbi, i + CURSEG_HOT_DATA);
> +	}
> +
> +	ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi));
> +	ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi));
> +	ckpt->next_free_nid = cpu_to_le32(last_nid);
> +
> +	/* 2 cp  + n data seg summary + orphan inode blocks */
> +	data_sum_blocks = npages_for_summary_flush(sbi);
> +	if (data_sum_blocks < 3)
> +		ckpt->ckpt_flags |= CP_COMPACT_SUM_FLAG;
> +	else
> +		ckpt->ckpt_flags &= (~CP_COMPACT_SUM_FLAG);
> +
> +	orphan_blocks = (sbi->n_orphans + F2FS_ORPHANS_PER_BLOCK - 1)
> +					/ F2FS_ORPHANS_PER_BLOCK;
> +	ckpt->cp_pack_start_sum = 1 + orphan_blocks;
> +	ckpt->cp_pack_total_block_count = 2 + data_sum_blocks + orphan_blocks;

This looks a bit weird to me, though I might be misunderstanding something.

data_sum_blocks is either 1, 2, or 3.
"3" actually means "at least 3".

If it is 3, you choose not to set CP_COMPACT_SUM_FLAG.  In that case the NAT
and SIT journal entries go into SSA blocks, not into the checkpoint at all.
So in that case, zero blocks of the checkpoint are used for journalling.  Yet
you still add data_sum_blocks (==3) to the cp_pack_total_block_count (and
later to the start block).
Is that really what you want to do?  Leave 3 empty blocks?

I would suggest changing npages_for_summary_flush to return 0 if the number
of blocks needed would be more than three, and set CP_COMPACT_SUM_FLAG only
when data_sum_blocks > 0.

I don't know if you would need to make a corresponding change to the recovery
code, I haven't fully examined that yet.

Regards,
NeilBrown

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]

Re: [PATCH 05/16] f2fs: add checkpoint operations

[Jaegeuk Kim]

2012-10-11 (목), 09:24 +1100, NeilBrown:
> On Fri, 05 Oct 2012 20:59:29 +0900 김재극 <jaegeuk.kim@samsung.com> wrote:
> 
> > +static void do_checkpoint(struct f2fs_sb_info *sbi, bool is_umount)
> > +{
> > +	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
> > +	nid_t last_nid = 0;
> > +	int nat_upd_blkoff[3];
> > +	block_t start_blk;
> > +	struct page *cp_page;
> > +	unsigned int data_sum_blocks, orphan_blocks;
> > +	void *kaddr;
> > +	__u32 crc32 = 0;
> > +	int i;
> > +
> > +	/* Flush all the NAT/SIT pages */
> > +	while (get_pages(sbi, F2FS_DIRTY_META))
> > +		sync_meta_pages(sbi, META, LONG_MAX);
> > +
> > +	next_free_nid(sbi, &last_nid);
> > +
> > +	/*
> > +	 * modify checkpoint
> > +	 * version number is already updated
> > +	 */
> > +	ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi));
> > +	ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi));
> > +	ckpt->free_segment_count = cpu_to_le32(free_segments(sbi));
> > +	for (i = 0; i < 3; i++) {
> > +		ckpt->cur_node_segno[i] =
> > +			cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE));
> > +		ckpt->cur_node_blkoff[i] =
> > +			cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE));
> > +		nat_upd_blkoff[i] = NM_I(sbi)->nat_upd_blkoff[i];
> > +		ckpt->nat_upd_blkoff[i] = cpu_to_le16(nat_upd_blkoff[i]);
> > +		ckpt->alloc_type[i + CURSEG_HOT_NODE] =
> > +				curseg_alloc_type(sbi, i + CURSEG_HOT_NODE);
> > +	}
> > +	for (i = 0; i < 3; i++) {
> > +		ckpt->cur_data_segno[i] =
> > +			cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA));
> > +		ckpt->cur_data_blkoff[i] =
> > +			cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA));
> > +		ckpt->alloc_type[i + CURSEG_HOT_DATA] =
> > +				curseg_alloc_type(sbi, i + CURSEG_HOT_DATA);
> > +	}
> > +
> > +	ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi));
> > +	ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi));
> > +	ckpt->next_free_nid = cpu_to_le32(last_nid);
> > +
> > +	/* 2 cp  + n data seg summary + orphan inode blocks */
> > +	data_sum_blocks = npages_for_summary_flush(sbi);
> > +	if (data_sum_blocks < 3)
> > +		ckpt->ckpt_flags |= CP_COMPACT_SUM_FLAG;
> > +	else
> > +		ckpt->ckpt_flags &= (~CP_COMPACT_SUM_FLAG);
> > +
> > +	orphan_blocks = (sbi->n_orphans + F2FS_ORPHANS_PER_BLOCK - 1)
> > +					/ F2FS_ORPHANS_PER_BLOCK;
> > +	ckpt->cp_pack_start_sum = 1 + orphan_blocks;
> > +	ckpt->cp_pack_total_block_count = 2 + data_sum_blocks + orphan_blocks;
> 
> This looks a bit weird to me, though I might be misunderstanding something.
> 
> data_sum_blocks is either 1, 2, or 3.
> "3" actually means "at least 3".
> 
> If it is 3, you choose not to set CP_COMPACT_SUM_FLAG.  In that case the NAT
> and SIT journal entries go into SSA blocks, not into the checkpoint at all.
> So in that case, zero blocks of the checkpoint are used for journalling.  Yet
> you still add data_sum_blocks (==3) to the cp_pack_total_block_count (and
> later to the start block).
> Is that really what you want to do?  Leave 3 empty blocks?
> 
> I would suggest changing npages_for_summary_flush to return 0 if the number
> of blocks needed would be more than three, and set CP_COMPACT_SUM_FLAG only
> when data_sum_blocks > 0.
> 
> I don't know if you would need to make a corresponding change to the recovery
> code, I haven't fully examined that yet.

Ok, let me explain about CP_COMPACT_SUM_FLAG.
Let's assume that there are some journal entries and data summaries.
Note that this scenario is not from the umount procedure.

Basically f2fs writes three data summary blocks for current active logs
inside the checkpoint pack.
And NAT and SIT journal entries are stored in hot and cold data summary
blocks.
So, if the CP_COMPACT_SUM_FLAG is not set, f2fs writes the checkpoint
pack like this.

[CP 0]
[Orphan blocks]
[Hot sum block w/ NAT journal]
[Warm sum block]
[Cold sum block w/ SIT journal]
[CP 0']

But, if the CP_COMPACT_SUM_FLAG is set, the checkpoint pack consists of
1 or 2 summary blocks as follows.

[CP 0]
[Orphan blocks]
[summary entries w/ NAT and SIT journal]
[CP 0']

or,

[CP 0]
[Orphan blocks]
[summary entries]
[summary entries w/ NAT and SIT journal]
[CP 0']

So, I think it needs no change.
Any idea?
Thanks,

> 
> Regards,
> NeilBrown

-- 
Jaegeuk Kim
Samsung

Re: [PATCH 05/16] f2fs: add checkpoint operations

[NeilBrown]

[-- Attachment #1: Type: text/plain, Size: 4909 bytes --]

On Sat, 13 Oct 2012 00:49:06 +0900 Jaegeuk Kim <jaegeuk.kim@gmail.com> wrote:

> 2012-10-11 (목), 09:24 +1100, NeilBrown:
> > On Fri, 05 Oct 2012 20:59:29 +0900 김재극 <jaegeuk.kim@samsung.com> wrote:
> > 
> > > +static void do_checkpoint(struct f2fs_sb_info *sbi, bool is_umount)
> > > +{
> > > +	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
> > > +	nid_t last_nid = 0;
> > > +	int nat_upd_blkoff[3];
> > > +	block_t start_blk;
> > > +	struct page *cp_page;
> > > +	unsigned int data_sum_blocks, orphan_blocks;
> > > +	void *kaddr;
> > > +	__u32 crc32 = 0;
> > > +	int i;
> > > +
> > > +	/* Flush all the NAT/SIT pages */
> > > +	while (get_pages(sbi, F2FS_DIRTY_META))
> > > +		sync_meta_pages(sbi, META, LONG_MAX);
> > > +
> > > +	next_free_nid(sbi, &last_nid);
> > > +
> > > +	/*
> > > +	 * modify checkpoint
> > > +	 * version number is already updated
> > > +	 */
> > > +	ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi));
> > > +	ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi));
> > > +	ckpt->free_segment_count = cpu_to_le32(free_segments(sbi));
> > > +	for (i = 0; i < 3; i++) {
> > > +		ckpt->cur_node_segno[i] =
> > > +			cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE));
> > > +		ckpt->cur_node_blkoff[i] =
> > > +			cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE));
> > > +		nat_upd_blkoff[i] = NM_I(sbi)->nat_upd_blkoff[i];
> > > +		ckpt->nat_upd_blkoff[i] = cpu_to_le16(nat_upd_blkoff[i]);
> > > +		ckpt->alloc_type[i + CURSEG_HOT_NODE] =
> > > +				curseg_alloc_type(sbi, i + CURSEG_HOT_NODE);
> > > +	}
> > > +	for (i = 0; i < 3; i++) {
> > > +		ckpt->cur_data_segno[i] =
> > > +			cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA));
> > > +		ckpt->cur_data_blkoff[i] =
> > > +			cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA));
> > > +		ckpt->alloc_type[i + CURSEG_HOT_DATA] =
> > > +				curseg_alloc_type(sbi, i + CURSEG_HOT_DATA);
> > > +	}
> > > +
> > > +	ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi));
> > > +	ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi));
> > > +	ckpt->next_free_nid = cpu_to_le32(last_nid);
> > > +
> > > +	/* 2 cp  + n data seg summary + orphan inode blocks */
> > > +	data_sum_blocks = npages_for_summary_flush(sbi);
> > > +	if (data_sum_blocks < 3)
> > > +		ckpt->ckpt_flags |= CP_COMPACT_SUM_FLAG;
> > > +	else
> > > +		ckpt->ckpt_flags &= (~CP_COMPACT_SUM_FLAG);
> > > +
> > > +	orphan_blocks = (sbi->n_orphans + F2FS_ORPHANS_PER_BLOCK - 1)
> > > +					/ F2FS_ORPHANS_PER_BLOCK;
> > > +	ckpt->cp_pack_start_sum = 1 + orphan_blocks;
> > > +	ckpt->cp_pack_total_block_count = 2 + data_sum_blocks + orphan_blocks;
> > 
> > This looks a bit weird to me, though I might be misunderstanding something.
> > 
> > data_sum_blocks is either 1, 2, or 3.
> > "3" actually means "at least 3".
> > 
> > If it is 3, you choose not to set CP_COMPACT_SUM_FLAG.  In that case the NAT
> > and SIT journal entries go into SSA blocks, not into the checkpoint at all.
> > So in that case, zero blocks of the checkpoint are used for journalling.  Yet
> > you still add data_sum_blocks (==3) to the cp_pack_total_block_count (and
> > later to the start block).
> > Is that really what you want to do?  Leave 3 empty blocks?
> > 
> > I would suggest changing npages_for_summary_flush to return 0 if the number
> > of blocks needed would be more than three, and set CP_COMPACT_SUM_FLAG only
> > when data_sum_blocks > 0.
> > 
> > I don't know if you would need to make a corresponding change to the recovery
> > code, I haven't fully examined that yet.
> 
> Ok, let me explain about CP_COMPACT_SUM_FLAG.
> Let's assume that there are some journal entries and data summaries.
> Note that this scenario is not from the umount procedure.
> 
> Basically f2fs writes three data summary blocks for current active logs
> inside the checkpoint pack.
> And NAT and SIT journal entries are stored in hot and cold data summary
> blocks.
> So, if the CP_COMPACT_SUM_FLAG is not set, f2fs writes the checkpoint
> pack like this.
> 
> [CP 0]
> [Orphan blocks]
> [Hot sum block w/ NAT journal]
> [Warm sum block]
> [Cold sum block w/ SIT journal]
> [CP 0']
> 
> But, if the CP_COMPACT_SUM_FLAG is set, the checkpoint pack consists of
> 1 or 2 summary blocks as follows.
> 
> [CP 0]
> [Orphan blocks]
> [summary entries w/ NAT and SIT journal]
> [CP 0']
> 
> or,
> 
> [CP 0]
> [Orphan blocks]
> [summary entries]
> [summary entries w/ NAT and SIT journal]
> [CP 0']
> 
> So, I think it needs no change.
> Any idea?
> Thanks,

I see.  I missed the fact that the current data summary blocks are always
written to the checkpoint area - I assumed they were being written back to
the SSA.

So it makes sense now and you are right - no change needed.

Thanks,
NeilBrown


> 
> > 
> > Regards,
> > NeilBrown
> 


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]