f2fs mkfs 格式化过程系列 1

本文以f2fs-tools 1.9为例，描述一下f2fs格式化过程做了那些事，都往f2fs mete data里面写了什么东西。

f2fs metedata包含sb, cp(2 segments), sit, nat, ssa,

先从cp开始，看一下格式化过程往这两个check point里面写了哪些数据。

static int f2fs_write_check_point_pack(void)
{
	struct f2fs_summary_block *sum = NULL;
	struct f2fs_journal *journal;
	u_int32_t blk_size_bytes;
	u_int32_t nat_bits_bytes, nat_bits_blocks;
	unsigned char *nat_bits = NULL, *empty_nat_bits;
	u_int64_t cp_seg_blk = 0;
	u_int32_t crc = 0, flags;
	unsigned int i;
	char *cp_payload = NULL;
	char *sum_compact, *sum_compact_p;
	struct f2fs_summary *sum_entry;
	int ret = -1;

	cp = calloc(F2FS_BLKSIZE, 1);
	if (cp == NULL) {
		MSG(1, "\tError: Calloc Failed for f2fs_checkpoint!!!\n");
		return ret;
	}

	sum = calloc(F2FS_BLKSIZE, 1);
	if (sum == NULL) {
		MSG(1, "\tError: Calloc Failed for summay_node!!!\n");
		goto free_cp;
	}

	sum_compact = calloc(F2FS_BLKSIZE, 1);
	if (sum_compact == NULL) {
		MSG(1, "\tError: Calloc Failed for summay buffer!!!\n");
		goto free_sum;
	}
	sum_compact_p = sum_compact;

    //共有多少个node block, 用byte表示
    //为什么<<5呢(segment_count_nat * 32)?
    //segment_count_nat是两份nat, 一份就是segment_count_nat/2
    //共有block数为segment_count_nat * 512 /2,所以用byte再除以8
    //最终结果就是get_sb(segment_count_nat) << 5;
	nat_bits_bytes = get_sb(segment_count_nat) << 5;
	//用block表示
	nat_bits_blocks = F2FS_BYTES_TO_BLK((nat_bits_bytes << 1) + 8 +
						F2FS_BLKSIZE - 1);
	nat_bits = calloc(F2FS_BLKSIZE, nat_bits_blocks);
	if (nat_bits == NULL) {
		MSG(1, "\tError: Calloc Failed for nat bits buffer!!!\n");
		goto free_sum_compact;
	}

	cp_payload = calloc(F2FS_BLKSIZE, 1);
	if (cp_payload == NULL) {
		MSG(1, "\tError: Calloc Failed for cp_payload!!!\n");
		goto free_nat_bits;
	}

	/* 1. cp page 1 of checkpoint pack 1 */
	//将current node segment与current data segment赋值
	//c.cur_seg是在prepare_super_block赋值的
	//cur_node_segno[0] = 0；cur_node_segno[1] = 1；cur_node_segno[2] = 2；
	//cur_data_segno[0] = 3;
	cp->checkpoint_ver = cpu_to_le64(rand() | 0x1);
	set_cp(cur_node_segno[0], c.cur_seg[CURSEG_HOT_NODE]);
	set_cp(cur_node_segno[1], c.cur_seg[CURSEG_WARM_NODE]);
	set_cp(cur_node_segno[2], c.cur_seg[CURSEG_COLD_NODE]);
	set_cp(cur_data_segno[0], c.cur_seg[CURSEG_HOT_DATA]);
	set_cp(cur_data_segno[1], c.cur_seg[CURSEG_WARM_DATA]);
	set_cp(cur_data_segno[2], c.cur_seg[CURSEG_COLD_DATA]);
	for (i = 3; i < MAX_ACTIVE_NODE_LOGS; i++) {
		set_cp(cur_node_segno[i], 0xffffffff);
		set_cp(cur_data_segno[i], 0xffffffff);
	}

    //current node segment只写了一个root inode，所以cur_node_blkoff为1
    //current data segment只写了一个root dentry,所以cur_data_blkoff为1
	set_cp(cur_node_blkoff[0], 1);
	set_cp(cur_data_blkoff[0], 1);
	//只有两个有效的block，root inode和root dentry(data block)
	set_cp(valid_block_count, 2);

	//计算reserved segmetn，供gc用
	set_cp(rsvd_segment_count, c.reserved_segments);
	set_cp(overprov_segment_count, (get_sb(segment_count_main) -
			get_cp(rsvd_segment_count)) *
			c.overprovision / 100);
	set_cp(overprov_segment_count, get_cp(overprov_segment_count) +
			get_cp(rsvd_segment_count));

	MSG(0, "Info: Overprovision ratio = %.3lf%%\n", c.overprovision);
	MSG(0, "Info: Overprovision segments = %u (GC reserved = %u)\n",
					get_cp(overprov_segment_count),
					c.reserved_segments);

	/* main segments - reserved segments - (node + data segments) */
	//此处有一点疑虑，为什么-6,是减去hot/warm/cold的node与data么?
	set_cp(free_segment_count, get_sb(segment_count_main) - 6);
	set_cp(user_block_count, ((get_cp(free_segment_count) + 6 -
			get_cp(overprov_segment_count)) * c.blks_per_seg));
	
	/* cp page (2), data summaries (1), node summaries (3) */
	//两个cp page block,一个data summary block, 3个node summary blocks
	set_cp(cp_pack_total_block_count, 6 + get_sb(cp_payload));
	flags = CP_UMOUNT_FLAG | CP_COMPACT_SUM_FLAG;
	if (get_cp(cp_pack_total_block_count) <=
			(1 << get_sb(log_blocks_per_seg)) - nat_bits_blocks)
		flags |= CP_NAT_BITS_FLAG;

	if (c.trimmed)
		flags |= CP_TRIMMED_FLAG;

	set_cp(ckpt_flags, flags);
	set_cp(cp_pack_start_sum, 1 + get_sb(cp_payload));
	set_cp(valid_node_count, 1);
	set_cp(valid_inode_count, 1);
	set_cp(next_free_nid, get_sb(root_ino) + 1);
	set_cp(sit_ver_bitmap_bytesize, ((get_sb(segment_count_sit) / 2) <<
			get_sb(log_blocks_per_seg)) / 8);

	set_cp(nat_ver_bitmap_bytesize, ((get_sb(segment_count_nat) / 2) <<
			 get_sb(log_blocks_per_seg)) / 8);

	set_cp(checksum_offset, CHECKSUM_OFFSET);

	crc = f2fs_cal_crc32(F2FS_SUPER_MAGIC, cp, CHECKSUM_OFFSET);
	*((__le32 *)((unsigned char *)cp + CHECKSUM_OFFSET)) =
							cpu_to_le32(crc);

	blk_size_bytes = 1 << get_sb(log_blocksize);

	if (blk_size_bytes != F2FS_BLKSIZE) {
		MSG(1, "\tError: Wrong block size %d / %d!!!\n",
					blk_size_bytes, F2FS_BLKSIZE);
		goto free_cp_payload;
	}
    //cp写入的地址
	cp_seg_blk = get_sb(segment0_blkaddr);

	DBG(1, "\tWriting main segments, cp at offset 0x%08"PRIx64"\n",
						cp_seg_blk);
	if (dev_write_block(cp, cp_seg_blk)) {
		MSG(1, "\tError: While writing the cp to disk!!!\n");
		goto free_cp_payload;
	}
    //一个cp page已经够了，不需要扩展cp page，所以cp_payload值为0
    //如果需要扩展cp_payload，则需要把扩展的部分也写回flash
	for (i = 0; i < get_sb(cp_payload); i++) {
		cp_seg_blk++;
		if (dev_fill_block(cp_payload, cp_seg_blk)) {
			MSG(1, "\tError: While zeroing out the sit bitmap area "
					"on disk!!!\n");
			goto free_cp_payload;
		}
	}

	/* Prepare and write Segment summary for HOT/WARM/COLD DATA
	 *
	 * The structure of compact summary
	 * +-------------------+
	 * | nat_journal       |
	 * +-------------------+
	 * | sit_journal       |
	 * +-------------------+
	 * | hot data summary  |
	 * +-------------------+
	 * | warm data summary |
	 * +-------------------+
	 * | cold data summary |
	 * +-------------------+
	*/
	memset(sum, 0, sizeof(struct f2fs_summary_block));
	SET_SUM_TYPE((&sum->footer), SUM_TYPE_DATA);

    //初始化summary, node entry相关的信息，第一个是root inode info
	journal = &sum->journal;
	journal->n_nats = cpu_to_le16(1);//只有一个inode
	journal->nat_j.entries[0].nid = sb->root_ino;//node id
	journal->nat_j.entries[0].ne.version = 0;
	journal->nat_j.entries[0].ne.ino = sb->root_ino;
	//root inode block address
	journal->nat_j.entries[0].ne.block_addr = cpu_to_le32(
			get_sb(main_blkaddr) +
			get_cp(cur_node_segno[0]) * c.blks_per_seg);

    //将journal nats拷贝到sum_compact_p
	memcpy(sum_compact_p, &journal->n_nats, SUM_JOURNAL_SIZE);
	sum_compact_p += SUM_JOURNAL_SIZE;

	memset(sum, 0, sizeof(struct f2fs_summary_block));
	/* inode sit for root */
	journal->n_sits = cpu_to_le16(6);//segment info，hot/warm/code segment共6个
	journal->sit_j.entries[0].segno = cp->cur_node_segno[0];//hot node no
	//segment 中valid block counts,后10位用来表示valid block count
	//此时只有一个valid block(inode block)
	journal->sit_j.entries[0].se.vblocks =
				cpu_to_le16((CURSEG_HOT_NODE << 10) | 1);
	//将valid_map的第一位设置为1
	f2fs_set_bit(0, (char *)journal->sit_j.entries[0].se.valid_map);

	//设置current warm segment info
	journal->sit_j.entries[1].segno = cp->cur_node_segno[1];//segment number
	journal->sit_j.entries[1].se.vblocks =
				cpu_to_le16((CURSEG_WARM_NODE << 10));//valid blocks
	journal->sit_j.entries[2].segno = cp->cur_node_segno[2];//cold segment number
	journal->sit_j.entries[2].se.vblocks =
				cpu_to_le16((CURSEG_COLD_NODE << 10));//valid blocks

	/* data sit for root */
	journal->sit_j.entries[3].segno = cp->cur_data_segno[0];//hot data segment

	//hot data segment中的valid block,目前只有一个valid block(root dentry block)
	journal->sit_j.entries[3].se.vblocks =
				cpu_to_le16((CURSEG_HOT_DATA << 10) | 1);
	f2fs_set_bit(0, (char *)journal->sit_j.entries[3].se.valid_map);
	journal->sit_j.entries[4].segno = cp->cur_data_segno[1];
	journal->sit_j.entries[4].se.vblocks =
				cpu_to_le16((CURSEG_WARM_DATA << 10));
	journal->sit_j.entries[5].segno = cp->cur_data_segno[2];
	journal->sit_j.entries[5].se.vblocks =
				cpu_to_le16((CURSEG_COLD_DATA << 10));

	memcpy(sum_compact_p, &journal->n_sits, SUM_JOURNAL_SIZE);
	sum_compact_p += SUM_JOURNAL_SIZE;

	/* hot data summary */
	sum_entry = (struct f2fs_summary *)sum_compact_p;
	sum_entry->nid = sb->root_ino;
	sum_entry->ofs_in_node = 0;
	/* warm data summary, nothing to do */
	/* cold data summary, nothing to do */

    //将hot data summary 信息写入flash, warm/cold data不需要操作
	cp_seg_blk++;
	DBG(1, "\tWriting Segment summary for HOT/WARM/COLD_DATA, at offset 0x%08"PRIx64"\n",
			cp_seg_blk);
	if (dev_write_block(sum_compact, cp_seg_blk)) {
		MSG(1, "\tError: While writing the sum_blk to disk!!!\n");
		goto free_cp_payload;
	}

	/* Prepare and write Segment summary for HOT_NODE */
	memset(sum, 0, sizeof(struct f2fs_summary_block));
	SET_SUM_TYPE((&sum->footer), SUM_TYPE_NODE);

	sum->entries[0].nid = sb->root_ino;
	sum->entries[0].ofs_in_node = 0;

	cp_seg_blk++;
	DBG(1, "\tWriting Segment summary for HOT_NODE, at offset 0x%08"PRIx64"\n",
			cp_seg_blk);
	if (dev_write_block(sum, cp_seg_blk)) {
		MSG(1, "\tError: While writing the sum_blk to disk!!!\n");
		goto free_cp_payload;
	}

	/* Fill segment summary for WARM_NODE to zero. */
	memset(sum, 0, sizeof(struct f2fs_summary_block));
	SET_SUM_TYPE((&sum->footer), SUM_TYPE_NODE);

	cp_seg_blk++;
	DBG(1, "\tWriting Segment summary for WARM_NODE, at offset 0x%08"PRIx64"\n",
			cp_seg_blk);
	if (dev_write_block(sum, cp_seg_blk)) {
		MSG(1, "\tError: While writing the sum_blk to disk!!!\n");
		goto free_cp_payload;
	}

	/* Fill segment summary for COLD_NODE to zero. */
	memset(sum, 0, sizeof(struct f2fs_summary_block));
	SET_SUM_TYPE((&sum->footer), SUM_TYPE_NODE);
	cp_seg_blk++;
	DBG(1, "\tWriting Segment summary for COLD_NODE, at offset 0x%08"PRIx64"\n",
			cp_seg_blk);
	if (dev_write_block(sum, cp_seg_blk)) {
		MSG(1, "\tError: While writing the sum_blk to disk!!!\n");
		goto free_cp_payload;
	}
	//以上将hot/cold/warm node summary写入

	/* cp page2 *///cp checkpoint 1, cp page2,两个cp checkpoint，每个checkpoint包含两个cp page
	cp_seg_blk++;
	DBG(1, "\tWriting cp page2, at offset 0x%08"PRIx64"\n", cp_seg_blk);
	if (dev_write_block(cp, cp_seg_blk)) {
		MSG(1, "\tError: While writing the cp to disk!!!\n");
		goto free_cp_payload;
	}

	/* write NAT bits, if possible */
	if (flags & CP_NAT_BITS_FLAG) {
		uint32_t i;

		*(__le64 *)nat_bits = get_cp_crc(cp);
		empty_nat_bits = nat_bits + 8 + nat_bits_bytes;
		memset(empty_nat_bits, 0xff, nat_bits_bytes);
		test_and_clear_bit_le(0, empty_nat_bits);

		/* write the last blocks in cp pack */
		cp_seg_blk = get_sb(segment0_blkaddr) + (1 <<
				get_sb(log_blocks_per_seg)) - nat_bits_blocks;

		DBG(1, "\tWriting NAT bits pages, at offset 0x%08"PRIx64"\n",
					cp_seg_blk);

		for (i = 0; i < nat_bits_blocks; i++) {
			if (dev_write_block(nat_bits + i *
						F2FS_BLKSIZE, cp_seg_blk + i)) {
				MSG(1, "\tError: write NAT bits to disk!!!\n");
				goto free_cp_payload;
			}
		}
	}

	/* cp page 1 of check point pack 2
	 * Initiatialize other checkpoint pack with version zero
	 */
	cp->checkpoint_ver = 0;

	crc = f2fs_cal_crc32(F2FS_SUPER_MAGIC, cp, CHECKSUM_OFFSET);
	*((__le32 *)((unsigned char *)cp + CHECKSUM_OFFSET)) =
							cpu_to_le32(crc);
	cp_seg_blk = get_sb(segment0_blkaddr) + c.blks_per_seg;
	DBG(1, "\tWriting cp page 1 of checkpoint pack 2, at offset 0x%08"PRIx64"\n",
				cp_seg_blk);
	if (dev_write_block(cp, cp_seg_blk)) {
		MSG(1, "\tError: While writing the cp to disk!!!\n");
		goto free_cp_payload;
	}

	for (i = 0; i < get_sb(cp_payload); i++) {
		cp_seg_blk++;
		if (dev_fill_block(cp_payload, cp_seg_blk)) {
			MSG(1, "\tError: While zeroing out the sit bitmap area "
					"on disk!!!\n");
			goto free_cp_payload;
		}
	}

	/* cp page 2 of check point pack 2 */
	cp_seg_blk += (le32_to_cpu(cp->cp_pack_total_block_count) -
					get_sb(cp_payload) - 1);
	DBG(1, "\tWriting cp page 2 of checkpoint pack 2, at offset 0x%08"PRIx64"\n",
				cp_seg_blk);
	if (dev_write_block(cp, cp_seg_blk)) {
		MSG(1, "\tError: While writing the cp to disk!!!\n");
		goto free_cp_payload;
	}

	ret = 0;

free_cp_payload:
	free(cp_payload);
free_nat_bits:
	free(nat_bits);
free_sum_compact:
	free(sum_compact);
free_sum:
	free(sum);
free_cp:
	free(cp);
	return ret;
}