check-integrity.c

			block->orig_bio_bh_end_io.bh = bh->b_end_io;
			block->next_in_same_bio = NULL;
			bh->b_private = block;
			bh->b_end_io = btrfsic_bh_end_io;
		} else {
			block->is_iodone = 1;
			block->orig_bio_bh_private = NULL;
			block->orig_bio_bh_end_io.bio = NULL;
			block->next_in_same_bio = NULL;
		}
		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
			printk(KERN_INFO
			       "New written %c-block @%llu (%s/%llu/%d)\n",
			       is_metadata ? 'M' : 'D',
			       (unsigned long long)block->logical_bytenr,
			       block->dev_state->name,
			       (unsigned long long)block->dev_bytenr,
			       block->mirror_num);
		list_add(&block->all_blocks_node, &state->all_blocks_list);
		btrfsic_block_hashtable_add(block, &state->block_hashtable);

		if (is_metadata) {
			ret = btrfsic_process_metablock(state, block,
							&block_ctx,
							(struct btrfs_header *)
							block_ctx.data, 0, 0);
			if (ret)
				printk(KERN_INFO
				       "btrfsic: process_metablock(root @%llu)"
				       " failed!\n",
				       (unsigned long long)dev_bytenr);
		}
		btrfsic_release_block_ctx(&block_ctx);
	}
}

static void btrfsic_bio_end_io(struct bio *bp, int bio_error_status)
{
	struct btrfsic_block *block = (struct btrfsic_block *)bp->bi_private;
	int iodone_w_error;

	/* mutex is not held! This is not save if IO is not yet completed
	 * on umount */
	iodone_w_error = 0;
	if (bio_error_status)
		iodone_w_error = 1;

	BUG_ON(NULL == block);
	bp->bi_private = block->orig_bio_bh_private;
	bp->bi_end_io = block->orig_bio_bh_end_io.bio;

	do {
		struct btrfsic_block *next_block;
		struct btrfsic_dev_state *const dev_state = block->dev_state;

		if ((dev_state->state->print_mask &
		     BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
			printk(KERN_INFO
			       "bio_end_io(err=%d) for %c @%llu (%s/%llu/%d)\n",
			       bio_error_status,
			       btrfsic_get_block_type(dev_state->state, block),
			       (unsigned long long)block->logical_bytenr,
			       dev_state->name,
			       (unsigned long long)block->dev_bytenr,
			       block->mirror_num);
		next_block = block->next_in_same_bio;
		block->iodone_w_error = iodone_w_error;
		if (block->submit_bio_bh_rw & REQ_FLUSH) {
			dev_state->last_flush_gen++;
			if ((dev_state->state->print_mask &
			     BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
				printk(KERN_INFO
				       "bio_end_io() new %s flush_gen=%llu\n",
				       dev_state->name,
				       (unsigned long long)
				       dev_state->last_flush_gen);
		}
		if (block->submit_bio_bh_rw & REQ_FUA)
			block->flush_gen = 0; /* FUA completed means block is
					       * on disk */
		block->is_iodone = 1; /* for FLUSH, this releases the block */
		block = next_block;
	} while (NULL != block);

	bp->bi_end_io(bp, bio_error_status);
}

static void btrfsic_bh_end_io(struct buffer_head *bh, int uptodate)
{
	struct btrfsic_block *block = (struct btrfsic_block *)bh->b_private;
	int iodone_w_error = !uptodate;
	struct btrfsic_dev_state *dev_state;

	BUG_ON(NULL == block);
	dev_state = block->dev_state;
	if ((dev_state->state->print_mask & BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
		printk(KERN_INFO
		       "bh_end_io(error=%d) for %c @%llu (%s/%llu/%d)\n",
		       iodone_w_error,
		       btrfsic_get_block_type(dev_state->state, block),
		       (unsigned long long)block->logical_bytenr,
		       block->dev_state->name,
		       (unsigned long long)block->dev_bytenr,
		       block->mirror_num);

	block->iodone_w_error = iodone_w_error;
	if (block->submit_bio_bh_rw & REQ_FLUSH) {
		dev_state->last_flush_gen++;
		if ((dev_state->state->print_mask &
		     BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
			printk(KERN_INFO
			       "bh_end_io() new %s flush_gen=%llu\n",
			       dev_state->name,
			       (unsigned long long)dev_state->last_flush_gen);
	}
	if (block->submit_bio_bh_rw & REQ_FUA)
		block->flush_gen = 0; /* FUA completed means block is on disk */

	bh->b_private = block->orig_bio_bh_private;
	bh->b_end_io = block->orig_bio_bh_end_io.bh;
	block->is_iodone = 1; /* for FLUSH, this releases the block */
	bh->b_end_io(bh, uptodate);
}

static int btrfsic_process_written_superblock(
		struct btrfsic_state *state,
		struct btrfsic_block *const superblock,
		struct btrfs_super_block *const super_hdr)
{
	int pass;

	superblock->generation = btrfs_super_generation(super_hdr);
	if (!(superblock->generation > state->max_superblock_generation ||
	      0 == state->max_superblock_generation)) {
		if (state->print_mask & BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE)
			printk(KERN_INFO
			       "btrfsic: superblock @%llu (%s/%llu/%d)"
			       " with old gen %llu <= %llu\n",
			       (unsigned long long)superblock->logical_bytenr,
			       superblock->dev_state->name,
			       (unsigned long long)superblock->dev_bytenr,
			       superblock->mirror_num,
			       (unsigned long long)
			       btrfs_super_generation(super_hdr),
			       (unsigned long long)
			       state->max_superblock_generation);
	} else {
		if (state->print_mask & BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE)
			printk(KERN_INFO
			       "btrfsic: got new superblock @%llu (%s/%llu/%d)"
			       " with new gen %llu > %llu\n",
			       (unsigned long long)superblock->logical_bytenr,
			       superblock->dev_state->name,
			       (unsigned long long)superblock->dev_bytenr,
			       superblock->mirror_num,
			       (unsigned long long)
			       btrfs_super_generation(super_hdr),
			       (unsigned long long)
			       state->max_superblock_generation);

		state->max_superblock_generation =
		    btrfs_super_generation(super_hdr);
		state->latest_superblock = superblock;
	}

	for (pass = 0; pass < 3; pass++) {
		int ret;
		u64 next_bytenr;
		struct btrfsic_block *next_block;
		struct btrfsic_block_data_ctx tmp_next_block_ctx;
		struct btrfsic_block_link *l;
		int num_copies;
		int mirror_num;
		const char *additional_string = NULL;
		struct btrfs_disk_key tmp_disk_key;

		tmp_disk_key.type = BTRFS_ROOT_ITEM_KEY;
		tmp_disk_key.offset = 0;

		switch (pass) {
		case 0:
			tmp_disk_key.objectid =
			    cpu_to_le64(BTRFS_ROOT_TREE_OBJECTID);
			additional_string = "root ";
			next_bytenr = btrfs_super_root(super_hdr);
			if (state->print_mask &
			    BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
				printk(KERN_INFO "root@%llu\n",
				       (unsigned long long)next_bytenr);
			break;
		case 1:
			tmp_disk_key.objectid =
			    cpu_to_le64(BTRFS_CHUNK_TREE_OBJECTID);
			additional_string = "chunk ";
			next_bytenr = btrfs_super_chunk_root(super_hdr);
			if (state->print_mask &
			    BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
				printk(KERN_INFO "chunk@%llu\n",
				       (unsigned long long)next_bytenr);
			break;
		case 2:
			tmp_disk_key.objectid =
			    cpu_to_le64(BTRFS_TREE_LOG_OBJECTID);
			additional_string = "log ";
			next_bytenr = btrfs_super_log_root(super_hdr);
			if (0 == next_bytenr)
				continue;
			if (state->print_mask &
			    BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
				printk(KERN_INFO "log@%llu\n",
				       (unsigned long long)next_bytenr);
			break;
		}

		num_copies =
		    btrfs_num_copies(&state->root->fs_info->mapping_tree,
				     next_bytenr, PAGE_SIZE);
		if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
			printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
			       (unsigned long long)next_bytenr, num_copies);
		for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
			int was_created;

			if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
				printk(KERN_INFO
				       "btrfsic_process_written_superblock("
				       "mirror_num=%d)\n", mirror_num);
			ret = btrfsic_map_block(state, next_bytenr, PAGE_SIZE,
						&tmp_next_block_ctx,
						mirror_num);
			if (ret) {
				printk(KERN_INFO
				       "btrfsic: btrfsic_map_block(@%llu,"
				       " mirror=%d) failed!\n",
				       (unsigned long long)next_bytenr,
				       mirror_num);
				return -1;
			}

			next_block = btrfsic_block_lookup_or_add(
					state,
					&tmp_next_block_ctx,
					additional_string,
					1, 0, 1,
					mirror_num,
					&was_created);
			if (NULL == next_block) {
				printk(KERN_INFO
				       "btrfsic: error, kmalloc failed!\n");
				btrfsic_release_block_ctx(&tmp_next_block_ctx);
				return -1;
			}

			next_block->disk_key = tmp_disk_key;
			if (was_created)
				next_block->generation =
				    BTRFSIC_GENERATION_UNKNOWN;
			l = btrfsic_block_link_lookup_or_add(
					state,
					&tmp_next_block_ctx,
					next_block,
					superblock,
					BTRFSIC_GENERATION_UNKNOWN);
			btrfsic_release_block_ctx(&tmp_next_block_ctx);
			if (NULL == l)
				return -1;
		}
	}

	if (-1 == btrfsic_check_all_ref_blocks(state, superblock, 0)) {
		WARN_ON(1);
		btrfsic_dump_tree(state);
	}

	return 0;
}

static int btrfsic_check_all_ref_blocks(struct btrfsic_state *state,
					struct btrfsic_block *const block,
					int recursion_level)
{
	struct list_head *elem_ref_to;
	int ret = 0;

	if (recursion_level >= 3 + BTRFS_MAX_LEVEL) {
		/*
		 * Note that this situation can happen and does not
		 * indicate an error in regular cases. It happens
		 * when disk blocks are freed and later reused.
		 * The check-integrity module is not aware of any
		 * block free operations, it just recognizes block
		 * write operations. Therefore it keeps the linkage
		 * information for a block until a block is
		 * rewritten. This can temporarily cause incorrect
		 * and even circular linkage informations. This
		 * causes no harm unless such blocks are referenced
		 * by the most recent super block.
		 */
		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
			printk(KERN_INFO
			       "btrfsic: abort cyclic linkage (case 1).\n");

		return ret;
	}

	/*
	 * This algorithm is recursive because the amount of used stack
	 * space is very small and the max recursion depth is limited.
	 */
	list_for_each(elem_ref_to, &block->ref_to_list) {
		const struct btrfsic_block_link *const l =
		    list_entry(elem_ref_to, struct btrfsic_block_link,
			       node_ref_to);

		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
			printk(KERN_INFO
			       "rl=%d, %c @%llu (%s/%llu/%d)"
			       " %u* refers to %c @%llu (%s/%llu/%d)\n",
			       recursion_level,
			       btrfsic_get_block_type(state, block),
			       (unsigned long long)block->logical_bytenr,
			       block->dev_state->name,
			       (unsigned long long)block->dev_bytenr,
			       block->mirror_num,
			       l->ref_cnt,
			       btrfsic_get_block_type(state, l->block_ref_to),
			       (unsigned long long)
			       l->block_ref_to->logical_bytenr,
			       l->block_ref_to->dev_state->name,
			       (unsigned long long)l->block_ref_to->dev_bytenr,
			       l->block_ref_to->mirror_num);
		if (l->block_ref_to->never_written) {
			printk(KERN_INFO "btrfs: attempt to write superblock"
			       " which references block %c @%llu (%s/%llu/%d)"
			       " which is never written!\n",
			       btrfsic_get_block_type(state, l->block_ref_to),
			       (unsigned long long)
			       l->block_ref_to->logical_bytenr,
			       l->block_ref_to->dev_state->name,
			       (unsigned long long)l->block_ref_to->dev_bytenr,
			       l->block_ref_to->mirror_num);
			ret = -1;
		} else if (!l->block_ref_to->is_iodone) {
			printk(KERN_INFO "btrfs: attempt to write superblock"
			       " which references block %c @%llu (%s/%llu/%d)"
			       " which is not yet iodone!\n",
			       btrfsic_get_block_type(state, l->block_ref_to),
			       (unsigned long long)
			       l->block_ref_to->logical_bytenr,
			       l->block_ref_to->dev_state->name,
			       (unsigned long long)l->block_ref_to->dev_bytenr,
			       l->block_ref_to->mirror_num);
			ret = -1;
		} else if (l->parent_generation !=
			   l->block_ref_to->generation &&
			   BTRFSIC_GENERATION_UNKNOWN !=
			   l->parent_generation &&
			   BTRFSIC_GENERATION_UNKNOWN !=
			   l->block_ref_to->generation) {
			printk(KERN_INFO "btrfs: attempt to write superblock"
			       " which references block %c @%llu (%s/%llu/%d)"
			       " with generation %llu !="
			       " parent generation %llu!\n",
			       btrfsic_get_block_type(state, l->block_ref_to),
			       (unsigned long long)
			       l->block_ref_to->logical_bytenr,
			       l->block_ref_to->dev_state->name,
			       (unsigned long long)l->block_ref_to->dev_bytenr,
			       l->block_ref_to->mirror_num,
			       (unsigned long long)l->block_ref_to->generation,
			       (unsigned long long)l->parent_generation);
			ret = -1;
		} else if (l->block_ref_to->flush_gen >
			   l->block_ref_to->dev_state->last_flush_gen) {
			printk(KERN_INFO "btrfs: attempt to write superblock"
			       " which references block %c @%llu (%s/%llu/%d)"
			       " which is not flushed out of disk's write cache"
			       " (block flush_gen=%llu,"
			       " dev->flush_gen=%llu)!\n",
			       btrfsic_get_block_type(state, l->block_ref_to),
			       (unsigned long long)
			       l->block_ref_to->logical_bytenr,
			       l->block_ref_to->dev_state->name,
			       (unsigned long long)l->block_ref_to->dev_bytenr,
			       l->block_ref_to->mirror_num,
			       (unsigned long long)block->flush_gen,
			       (unsigned long long)
			       l->block_ref_to->dev_state->last_flush_gen);
			ret = -1;
		} else if (-1 == btrfsic_check_all_ref_blocks(state,
							      l->block_ref_to,
							      recursion_level +
							      1)) {
			ret = -1;
		}
	}

	return ret;
}

static int btrfsic_is_block_ref_by_superblock(
		const struct btrfsic_state *state,
		const struct btrfsic_block *block,
		int recursion_level)
{
	struct list_head *elem_ref_from;

	if (recursion_level >= 3 + BTRFS_MAX_LEVEL) {
		/* refer to comment at "abort cyclic linkage (case 1)" */
		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
			printk(KERN_INFO
			       "btrfsic: abort cyclic linkage (case 2).\n");

		return 0;
	}

	/*
	 * This algorithm is recursive because the amount of used stack space
	 * is very small and the max recursion depth is limited.
	 */
	list_for_each(elem_ref_from, &block->ref_from_list) {
		const struct btrfsic_block_link *const l =
		    list_entry(elem_ref_from, struct btrfsic_block_link,
			       node_ref_from);

		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
			printk(KERN_INFO
			       "rl=%d, %c @%llu (%s/%llu/%d)"
			       " is ref %u* from %c @%llu (%s/%llu/%d)\n",
			       recursion_level,
			       btrfsic_get_block_type(state, block),
			       (unsigned long long)block->logical_bytenr,
			       block->dev_state->name,
			       (unsigned long long)block->dev_bytenr,
			       block->mirror_num,
			       l->ref_cnt,
			       btrfsic_get_block_type(state, l->block_ref_from),
			       (unsigned long long)
			       l->block_ref_from->logical_bytenr,
			       l->block_ref_from->dev_state->name,
			       (unsigned long long)
			       l->block_ref_from->dev_bytenr,
			       l->block_ref_from->mirror_num);
		if (l->block_ref_from->is_superblock &&
		    state->latest_superblock->dev_bytenr ==
		    l->block_ref_from->dev_bytenr &&
		    state->latest_superblock->dev_state->bdev ==
		    l->block_ref_from->dev_state->bdev)
			return 1;
		else if (btrfsic_is_block_ref_by_superblock(state,
							    l->block_ref_from,
							    recursion_level +
							    1))
			return 1;
	}

	return 0;
}

static void btrfsic_print_add_link(const struct btrfsic_state *state,
				   const struct btrfsic_block_link *l)
{
	printk(KERN_INFO
	       "Add %u* link from %c @%llu (%s/%llu/%d)"
	       " to %c @%llu (%s/%llu/%d).\n",
	       l->ref_cnt,
	       btrfsic_get_block_type(state, l->block_ref_from),
	       (unsigned long long)l->block_ref_from->logical_bytenr,
	       l->block_ref_from->dev_state->name,
	       (unsigned long long)l->block_ref_from->dev_bytenr,
	       l->block_ref_from->mirror_num,
	       btrfsic_get_block_type(state, l->block_ref_to),
	       (unsigned long long)l->block_ref_to->logical_bytenr,
	       l->block_ref_to->dev_state->name,
	       (unsigned long long)l->block_ref_to->dev_bytenr,
	       l->block_ref_to->mirror_num);
}

static void btrfsic_print_rem_link(const struct btrfsic_state *state,
				   const struct btrfsic_block_link *l)
{
	printk(KERN_INFO
	       "Rem %u* link from %c @%llu (%s/%llu/%d)"
	       " to %c @%llu (%s/%llu/%d).\n",
	       l->ref_cnt,
	       btrfsic_get_block_type(state, l->block_ref_from),
	       (unsigned long long)l->block_ref_from->logical_bytenr,
	       l->block_ref_from->dev_state->name,
	       (unsigned long long)l->block_ref_from->dev_bytenr,
	       l->block_ref_from->mirror_num,
	       btrfsic_get_block_type(state, l->block_ref_to),
	       (unsigned long long)l->block_ref_to->logical_bytenr,
	       l->block_ref_to->dev_state->name,
	       (unsigned long long)l->block_ref_to->dev_bytenr,
	       l->block_ref_to->mirror_num);
}

static char btrfsic_get_block_type(const struct btrfsic_state *state,
				   const struct btrfsic_block *block)
{
	if (block->is_superblock &&
	    state->latest_superblock->dev_bytenr == block->dev_bytenr &&
	    state->latest_superblock->dev_state->bdev == block->dev_state->bdev)
		return 'S';
	else if (block->is_superblock)
		return 's';
	else if (block->is_metadata)
		return 'M';
	else
		return 'D';
}

static void btrfsic_dump_tree(const struct btrfsic_state *state)
{
	btrfsic_dump_tree_sub(state, state->latest_superblock, 0);
}

static void btrfsic_dump_tree_sub(const struct btrfsic_state *state,
				  const struct btrfsic_block *block,
				  int indent_level)
{
	struct list_head *elem_ref_to;
	int indent_add;
	static char buf[80];
	int cursor_position;

	/*
	 * Should better fill an on-stack buffer with a complete line and
	 * dump it at once when it is time to print a newline character.
	 */

	/*
	 * This algorithm is recursive because the amount of used stack space
	 * is very small and the max recursion depth is limited.
	 */
	indent_add = sprintf(buf, "%c-%llu(%s/%llu/%d)",
			     btrfsic_get_block_type(state, block),
			     (unsigned long long)block->logical_bytenr,
			     block->dev_state->name,
			     (unsigned long long)block->dev_bytenr,
			     block->mirror_num);
	if (indent_level + indent_add > BTRFSIC_TREE_DUMP_MAX_INDENT_LEVEL) {
		printk("[...]\n");
		return;
	}
	printk(buf);
	indent_level += indent_add;
	if (list_empty(&block->ref_to_list)) {
		printk("\n");
		return;
	}
	if (block->mirror_num > 1 &&
	    !(state->print_mask & BTRFSIC_PRINT_MASK_TREE_WITH_ALL_MIRRORS)) {
		printk(" [...]\n");
		return;
	}

	cursor_position = indent_level;
	list_for_each(elem_ref_to, &block->ref_to_list) {
		const struct btrfsic_block_link *const l =
		    list_entry(elem_ref_to, struct btrfsic_block_link,
			       node_ref_to);

		while (cursor_position < indent_level) {
			printk(" ");
			cursor_position++;
		}
		if (l->ref_cnt > 1)
			indent_add = sprintf(buf, " %d*--> ", l->ref_cnt);
		else
			indent_add = sprintf(buf, " --> ");
		if (indent_level + indent_add >
		    BTRFSIC_TREE_DUMP_MAX_INDENT_LEVEL) {
			printk("[...]\n");
			cursor_position = 0;
			continue;
		}

		printk(buf);

		btrfsic_dump_tree_sub(state, l->block_ref_to,
				      indent_level + indent_add);
		cursor_position = 0;
	}
}

static struct btrfsic_block_link *btrfsic_block_link_lookup_or_add(
		struct btrfsic_state *state,
		struct btrfsic_block_data_ctx *next_block_ctx,
		struct btrfsic_block *next_block,
		struct btrfsic_block *from_block,
		u64 parent_generation)
{
	struct btrfsic_block_link *l;

	l = btrfsic_block_link_hashtable_lookup(next_block_ctx->dev->bdev,
						next_block_ctx->dev_bytenr,
						from_block->dev_state->bdev,
						from_block->dev_bytenr,
						&state->block_link_hashtable);
	if (NULL == l) {
		l = btrfsic_block_link_alloc();
		if (NULL == l) {
			printk(KERN_INFO
			       "btrfsic: error, kmalloc" " failed!\n");
			return NULL;
		}

		l->block_ref_to = next_block;
		l->block_ref_from = from_block;
		l->ref_cnt = 1;
		l->parent_generation = parent_generation;

		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
			btrfsic_print_add_link(state, l);

		list_add(&l->node_ref_to, &from_block->ref_to_list);
		list_add(&l->node_ref_from, &next_block->ref_from_list);

		btrfsic_block_link_hashtable_add(l,
						 &state->block_link_hashtable);
	} else {
		l->ref_cnt++;
		l->parent_generation = parent_generation;
		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
			btrfsic_print_add_link(state, l);
	}

	return l;
}

static struct btrfsic_block *btrfsic_block_lookup_or_add(
		struct btrfsic_state *state,
		struct btrfsic_block_data_ctx *block_ctx,
		const char *additional_string,
		int is_metadata,
		int is_iodone,
		int never_written,
		int mirror_num,
		int *was_created)
{
	struct btrfsic_block *block;

	block = btrfsic_block_hashtable_lookup(block_ctx->dev->bdev,
					       block_ctx->dev_bytenr,
					       &state->block_hashtable);
	if (NULL == block) {
		struct btrfsic_dev_state *dev_state;

		block = btrfsic_block_alloc();
		if (NULL == block) {
			printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
			return NULL;
		}
		dev_state = btrfsic_dev_state_lookup(block_ctx->dev->bdev);
		if (NULL == dev_state) {
			printk(KERN_INFO
			       "btrfsic: error, lookup dev_state failed!\n");
			btrfsic_block_free(block);
			return NULL;
		}
		block->dev_state = dev_state;
		block->dev_bytenr = block_ctx->dev_bytenr;
		block->logical_bytenr = block_ctx->start;
		block->is_metadata = is_metadata;
		block->is_iodone = is_iodone;
		block->never_written = never_written;
		block->mirror_num = mirror_num;
		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
			printk(KERN_INFO
			       "New %s%c-block @%llu (%s/%llu/%d)\n",
			       additional_string,
			       btrfsic_get_block_type(state, block),
			       (unsigned long long)block->logical_bytenr,
			       dev_state->name,
			       (unsigned long long)block->dev_bytenr,
			       mirror_num);
		list_add(&block->all_blocks_node, &state->all_blocks_list);
		btrfsic_block_hashtable_add(block, &state->block_hashtable);
		if (NULL != was_created)
			*was_created = 1;
	} else {
		if (NULL != was_created)
			*was_created = 0;
	}

	return block;
}

static void btrfsic_cmp_log_and_dev_bytenr(struct btrfsic_state *state,
					   u64 bytenr,
					   struct btrfsic_dev_state *dev_state,
					   u64 dev_bytenr, char *data)
{
	int num_copies;
	int mirror_num;
	int ret;
	struct btrfsic_block_data_ctx block_ctx;
	int match = 0;

	num_copies = btrfs_num_copies(&state->root->fs_info->mapping_tree,
				      bytenr, PAGE_SIZE);

	for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
		ret = btrfsic_map_block(state, bytenr, PAGE_SIZE,
					&block_ctx, mirror_num);
		if (ret) {
			printk(KERN_INFO "btrfsic:"
			       " btrfsic_map_block(logical @%llu,"
			       " mirror %d) failed!\n",
			       (unsigned long long)bytenr, mirror_num);
			continue;
		}

		if (dev_state->bdev == block_ctx.dev->bdev &&
		    dev_bytenr == block_ctx.dev_bytenr) {
			match++;
			btrfsic_release_block_ctx(&block_ctx);
			break;
		}
		btrfsic_release_block_ctx(&block_ctx);
	}

	if (!match) {
		printk(KERN_INFO "btrfs: attempt to write M-block which contains logical bytenr that doesn't map to dev+physical bytenr of submit_bio,"
		       " buffer->log_bytenr=%llu, submit_bio(bdev=%s,"
		       " phys_bytenr=%llu)!\n",
		       (unsigned long long)bytenr, dev_state->name,
		       (unsigned long long)dev_bytenr);
		for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
			ret = btrfsic_map_block(state, bytenr, PAGE_SIZE,
						&block_ctx, mirror_num);
			if (ret)
				continue;

			printk(KERN_INFO "Read logical bytenr @%llu maps to"
			       " (%s/%llu/%d)\n",
			       (unsigned long long)bytenr,
			       block_ctx.dev->name,
			       (unsigned long long)block_ctx.dev_bytenr,
			       mirror_num);
		}
		WARN_ON(1);
	}
}

static struct btrfsic_dev_state *btrfsic_dev_state_lookup(
		struct block_device *bdev)
{
	struct btrfsic_dev_state *ds;

	ds = btrfsic_dev_state_hashtable_lookup(bdev,
						&btrfsic_dev_state_hashtable);
	return ds;
}

int btrfsic_submit_bh(int rw, struct buffer_head *bh)
{
	struct btrfsic_dev_state *dev_state;

	if (!btrfsic_is_initialized)
		return submit_bh(rw, bh);

	mutex_lock(&btrfsic_mutex);
	/* since btrfsic_submit_bh() might also be called before
	 * btrfsic_mount(), this might return NULL */
	dev_state = btrfsic_dev_state_lookup(bh->b_bdev);

	/* Only called to write the superblock (incl. FLUSH/FUA) */
	if (NULL != dev_state &&
	    (rw & WRITE) && bh->b_size > 0) {
		u64 dev_bytenr;

		dev_bytenr = 4096 * bh->b_blocknr;
		if (dev_state->state->print_mask &
		    BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
			printk(KERN_INFO
			       "submit_bh(rw=0x%x, blocknr=%lu (bytenr %llu),"
			       " size=%lu, data=%p, bdev=%p)\n",
			       rw, (unsigned long)bh->b_blocknr,
			       (unsigned long long)dev_bytenr,
			       (unsigned long)bh->b_size, bh->b_data,
			       bh->b_bdev);
		btrfsic_process_written_block(dev_state, dev_bytenr,
					      bh->b_data, bh->b_size, NULL,
					      NULL, bh, rw);
	} else if (NULL != dev_state && (rw & REQ_FLUSH)) {
		if (dev_state->state->print_mask &
		    BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
			printk(KERN_INFO
			       "submit_bh(rw=0x%x) FLUSH, bdev=%p)\n",
			       rw, bh->b_bdev);
		if (!dev_state->dummy_block_for_bio_bh_flush.is_iodone) {
			if ((dev_state->state->print_mask &
			     (BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH |
			      BTRFSIC_PRINT_MASK_VERBOSE)))
				printk(KERN_INFO
				       "btrfsic_submit_bh(%s) with FLUSH"
				       " but dummy block already in use"
				       " (ignored)!\n",
				       dev_state->name);
		} else {
			struct btrfsic_block *const block =
				&dev_state->dummy_block_for_bio_bh_flush;

			block->is_iodone = 0;
			block->never_written = 0;
			block->iodone_w_error = 0;
			block->flush_gen = dev_state->last_flush_gen + 1;
			block->submit_bio_bh_rw = rw;
			block->orig_bio_bh_private = bh->b_private;
			block->orig_bio_bh_end_io.bh = bh->b_end_io;
			block->next_in_same_bio = NULL;
			bh->b_private = block;
			bh->b_end_io = btrfsic_bh_end_io;
		}
	}
	mutex_unlock(&btrfsic_mutex);
	return submit_bh(rw, bh);
}

void btrfsic_submit_bio(int rw, struct bio *bio)
{
	struct btrfsic_dev_state *dev_state;

	if (!btrfsic_is_initialized) {
		submit_bio(rw, bio);
		return;
	}

	mutex_lock(&btrfsic_mutex);
	/* since btrfsic_submit_bio() is also called before
	 * btrfsic_mount(), this might return NULL */
	dev_state = btrfsic_dev_state_lookup(bio->bi_bdev);
	if (NULL != dev_state &&
	    (rw & WRITE) && NULL != bio->bi_io_vec) {
		unsigned int i;
		u64 dev_bytenr;
		int bio_is_patched;

		dev_bytenr = 512 * bio->bi_sector;
		bio_is_patched = 0;
		if (dev_state->state->print_mask &
		    BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
			printk(KERN_INFO
			       "submit_bio(rw=0x%x, bi_vcnt=%u,"
			       " bi_sector=%lu (bytenr %llu), bi_bdev=%p)\n",
			       rw, bio->bi_vcnt, (unsigned long)bio->bi_sector,
			       (unsigned long long)dev_bytenr,
			       bio->bi_bdev);

		for (i = 0; i < bio->bi_vcnt; i++) {
			u8 *mapped_data;

			mapped_data = kmap(bio->bi_io_vec[i].bv_page);
			if ((BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH |
			     BTRFSIC_PRINT_MASK_VERBOSE) ==
			    (dev_state->state->print_mask &
			     (BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH |
			      BTRFSIC_PRINT_MASK_VERBOSE)))
				printk(KERN_INFO
				       "#%u: page=%p, mapped=%p, len=%u,"
				       " offset=%u\n",
				       i, bio->bi_io_vec[i].bv_page,
				       mapped_data,
				       bio->bi_io_vec[i].bv_len,
				       bio->bi_io_vec[i].bv_offset);
			btrfsic_process_written_block(dev_state, dev_bytenr,
						      mapped_data,
						      bio->bi_io_vec[i].bv_len,
						      bio, &bio_is_patched,
						      NULL, rw);
			kunmap(bio->bi_io_vec[i].bv_page);
			dev_bytenr += bio->bi_io_vec[i].bv_len;
		}
	} else if (NULL != dev_state && (rw & REQ_FLUSH)) {
		if (dev_state->state->print_mask &
		    BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
			printk(KERN_INFO
			       "submit_bio(rw=0x%x) FLUSH, bdev=%p)\n",
			       rw, bio->bi_bdev);
		if (!dev_state->dummy_block_for_bio_bh_flush.is_iodone) {
			if ((dev_state->state->print_mask &
			     (BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH |
			      BTRFSIC_PRINT_MASK_VERBOSE)))
				printk(KERN_INFO
				       "btrfsic_submit_bio(%s) with FLUSH"
				       " but dummy block already in use"
				       " (ignored)!\n",
				       dev_state->name);
		} else {
			struct btrfsic_block *const block =
				&dev_state->dummy_block_for_bio_bh_flush;

			block->is_iodone = 0;
			block->never_written = 0;
			block->iodone_w_error = 0;
			block->flush_gen = dev_state->last_flush_gen + 1;
			block->submit_bio_bh_rw = rw;
			block->orig_bio_bh_private = bio->bi_private;
			block->orig_bio_bh_end_io.bio = bio->bi_end_io;
			block->next_in_same_bio = NULL;
			bio->bi_private = block;
			bio->bi_end_io = btrfsic_bio_end_io;
		}
	}
	mutex_unlock(&btrfsic_mutex);

	submit_bio(rw, bio);
}

int btrfsic_mount(struct btrfs_root *root,
		  struct btrfs_fs_devices *fs_devices,
		  int including_extent_data, u32 print_mask)
{
	int ret;
	struct btrfsic_state *state;
	struct list_head *dev_head = &fs_devices->devices;
	struct btrfs_device *device;

	state = kzalloc(sizeof(*state), GFP_NOFS);
	if (NULL == state) {
		printk(KERN_INFO "btrfs check-integrity: kmalloc() failed!\n");
		return -1;
	}

	if (!btrfsic_is_initialized) {
		mutex_init(&btrfsic_mutex);
		btrfsic_dev_state_hashtable_init(&btrfsic_dev_state_hashtable);
		btrfsic_is_initialized = 1;
	}
	mutex_lock(&btrfsic_mutex);
	state->root = root;
	state->print_mask = print_mask;
	state->include_extent_data = including_extent_data;
	state->csum_size = 0;
	INIT_LIST_HEAD(&state->all_blocks_list);
	btrfsic_block_hashtable_init(&state->block_hashtable);
	btrfsic_block_link_hashtable_init(&state->block_link_hashtable);
	state->max_superblock_generation = 0;
	state->latest_superblock = NULL;

	list_for_each_entry(device, dev_head, dev_list) {
		struct btrfsic_dev_state *ds;
		char *p;

		if (!device->bdev || !device->name)
			continue;

		ds = btrfsic_dev_state_alloc();
		if (NULL == ds) {
			printk(KERN_INFO
			       "btrfs check-integrity: kmalloc() failed!\n");
			mutex_unlock(&btrfsic_mutex);
			return -1;
		}
		ds->bdev = device->bdev;
		ds->state = state;
		bdevname(ds->bdev, ds->name);
		ds->name[BDEVNAME_SIZE - 1] = '\0';
		for (p = ds->name; *p != '\0'; p++);
		while (p > ds->name && *p != '/')
			p--;
		if (*p == '/')
			p++;
		strlcpy(ds->name, p, sizeof(ds->name));
		btrfsic_dev_state_hashtable_add(ds,
						&btrfsic_dev_state_hashtable);
	}

	ret = btrfsic_process_superblock(state, fs_devices);
	if (0 != ret) {
		mutex_unlock(&btrfsic_mutex);
		btrfsic_unmount(root, fs_devices);
		return ret;
	}

	if (state->print_mask & BTRFSIC_PRINT_MASK_INITIAL_DATABASE)
		btrfsic_dump_database(state);
	if (state->print_mask & BTRFSIC_PRINT_MASK_INITIAL_TREE)
		btrfsic_dump_tree(state);

	mutex_unlock(&btrfsic_mutex);
	return 0;
}

void btrfsic_unmount(struct btrfs_root *root,
		     struct btrfs_fs_devices *fs_devices)
{
	struct list_head *elem_all;
	struct list_head *tmp_all;
	struct btrfsic_state *state;
	struct list_head *dev_head = &fs_devices->devices;
	struct btrfs_device *device;

	if (!btrfsic_is_initialized)
		return;

	mutex_lock(&btrfsic_mutex);