Merge branch 'master' of github.com:jthornber/thin-provisioning-tools

Makefile.in

@@ -118,6 +118,8 @@ SOURCE=\
 	thin-provisioning/xml_format.cc
 
 DEVTOOLS_SOURCE=\
+	thin-provisioning/thin_journal.cc \
+	thin-provisioning/thin_journal_check.cc \
 	thin-provisioning/thin_ll_dump.cc \
 	thin-provisioning/thin_ll_restore.cc \
 	thin-provisioning/thin_show_duplicates.cc \

bin/thin_journal_check

@@ -0,0 +1 @@
+pdata_tools

ft-lib/bcache.c

@@ -8,7 +8,6 @@
 #include <stdbool.h>
 #include <stdlib.h>
 #include <stdio.h>
-#include <stdint.h>
 #include <libaio.h>
 #include <unistd.h>
 #include <linux/fs.h>

functional-tests/thin-functional-tests.scm

@@ -126,6 +126,13 @@
       (run-fail-rcv (_ stderr) (thin-check xml)
         (assert-matches ".*This looks like XML.  thin_check only checks the binary metadata format." stderr))))
 
+  (define-scenario (thin-check info-fields)
+    "Outputs info fields"
+    (with-valid-metadata (md)
+      (run-ok-rcv (stdout stderr) (thin-check md)
+        (assert-matches ".*TRANSACTION_ID=[0-9]+.*" stdout)
+	(assert-matches ".*METADATA_FREE_BLOCKS=[0-9]+.*" stdout))))
+
   ;;;-----------------------------------------------------------
   ;;; thin_restore scenarios
   ;;;-----------------------------------------------------------

thin-provisioning/commands.cc

@@ -25,6 +25,7 @@ thin_provisioning::register_thin_commands(base::application &app)
 	app.add_cmd(command::ptr(new thin_generate_metadata_cmd()));
 	app.add_cmd(command::ptr(new thin_show_duplicates_cmd()));
 	app.add_cmd(command::ptr(new thin_show_metadata_cmd()));
+	app.add_cmd(command::ptr(new thin_journal_cmd()));
 #endif
 }
 

thin-provisioning/commands.h

@@ -116,6 +116,13 @@ namespace thin_provisioning {
 		virtual void usage(std::ostream &out) const;
 		virtual int run(int argc, char **argv);
 	};
+
+	class thin_journal_cmd : public base::command {
+	public:
+		thin_journal_cmd();
+		virtual void usage(std::ostream &out) const;
+		virtual int run(int argc, char **argv);
+	};
 #endif
 
 	void register_thin_commands(base::application &app);

thin-provisioning/thin_check.cc

@@ -195,6 +195,13 @@ namespace {
 		return fs.override_mapping_root ? *fs.override_mapping_root : sb.data_mapping_root_;
 	}
 
+	void print_info(superblock_detail::superblock const &sb,
+                        transaction_manager::ptr tm)
+	{
+		cout << "TRANSACTION_ID=" << sb.trans_id_ << "\n";
+		cout << "METADATA_FREE_BLOCKS=" << tm->get_sm()->get_nr_free() << "\n";
+	}
+
 	error_state metadata_check(string const &path, flags fs) {
 		nested_output out(cerr, 2);
 		if (fs.quiet)
@@ -228,6 +235,9 @@ namespace {
 		transaction_manager::ptr tm =
 			open_tm(bm, superblock_detail::SUPERBLOCK_LOCATION);
 
+		if (!fs.quiet)
+			print_info(sb, tm);
+
 		if (fs.check_device_tree) {
 			out << "examining devices tree" << end_message();
 			{

thin-provisioning/thin_journal.cc

@@ -0,0 +1,366 @@
+// Copyright (C) 2018 Red Hat, Inc. All rights reserved.
+//
+// This file is part of the thin-provisioning-tools source.
+//
+// thin-provisioning-tools is free software: you can redistribute it
+// and/or modify it under the terms of the GNU General Public License
+// as published by the Free Software Foundation, either version 3 of
+// the License, or (at your option) any later version.
+//
+// thin-provisioning-tools is distributed in the hope that it will be
+// useful, but WITHOUT ANY WARRANTY; without even the implied warranty
+// of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License along
+// with thin-provisioning-tools.  If not, see
+// <http://www.gnu.org/licenses/>.
+
+#include "thin-provisioning/thin_journal.h"
+
+#include <algorithm>
+
+using namespace thin_provisioning;
+using namespace persistent_data;
+using namespace std;
+
+//----------------------------------------------------------------
+
+byte_stream::byte_stream(block_manager<JOURNAL_BLOCK_SIZE>::ptr bm)
+		: bm_(bm),
+		  current_block_(0),
+		  cursor_(0)
+{
+}
+
+void
+byte_stream::read_bytes(uint8_t *b, uint8_t *e)
+{
+	while (b != e)
+		b += read_some_(b, e);
+
+	assert(b == e);
+}
+
+void
+byte_stream::next_block_()
+{
+	current_block_++;
+	cursor_ = 0;
+}
+
+size_t
+byte_stream::read_some_(uint8_t *b, uint8_t *e)
+{
+	assert(cursor_ <= JOURNAL_BLOCK_SIZE);
+
+	if (cursor_ == JOURNAL_BLOCK_SIZE)
+		next_block_();
+
+	size_t len = min<uint64_t>(e - b, JOURNAL_BLOCK_SIZE - cursor_);
+	auto rr = bm_->read_lock(current_block_);
+
+	uint8_t const *data_begin = reinterpret_cast<uint8_t const *>(rr.data()) + cursor_;
+	memcpy(b, data_begin, len);
+	cursor_ += len;
+
+	return len;
+}
+
+//---------------------------------
+
+journal_msg::journal_msg(bool success)
+	: success_(success)
+{
+}
+
+open_journal_msg::open_journal_msg(uint64_t nr_metadata_blocks)
+	: journal_msg(true),
+	  nr_metadata_blocks_(nr_metadata_blocks)
+{
+}
+
+void
+open_journal_msg::visit(journal_visitor &v) const
+{
+	v.visit(*this);
+}
+
+close_journal_msg::close_journal_msg()
+	: journal_msg(true)
+{
+}
+
+void
+close_journal_msg::visit(journal_visitor &v) const
+{
+	v.visit(*this);
+}
+
+block_msg::block_msg(bool success, uint64_t index) 
+	: journal_msg(success), index_(index)
+{
+}
+
+read_lock_msg::read_lock_msg(bool success, uint64_t index) 
+	: block_msg(success, index)
+{
+}
+
+void
+read_lock_msg::visit(journal_visitor &v) const
+{
+	v.visit(*this);
+}
+
+write_lock_msg::write_lock_msg(bool success, uint64_t index) 
+	: block_msg(success, index)
+{
+}
+
+void
+write_lock_msg::visit(journal_visitor &v) const
+{
+	v.visit(*this);
+}
+
+zero_lock_msg::zero_lock_msg(bool success, uint64_t index) 
+	: block_msg(success, index)
+{
+}
+
+void
+zero_lock_msg::visit(journal_visitor &v) const
+{
+	v.visit(*this);
+}
+
+try_read_lock_msg::try_read_lock_msg(bool success, uint64_t index) 
+	: block_msg(success, index)
+{
+}
+
+void
+try_read_lock_msg::visit(journal_visitor &v) const
+{
+	v.visit(*this);
+}
+
+unlock_msg::unlock_msg(bool success, uint64_t index, delta_list const &deltas)
+	: block_msg(success, index),
+	  deltas_(deltas)
+{
+}
+
+void
+unlock_msg::visit(journal_visitor &v) const
+{
+	v.visit(*this);
+}
+
+verify_msg::verify_msg(bool success, uint64_t index) 
+	: block_msg(success, index)
+{
+}
+
+void
+verify_msg::visit(journal_visitor &v) const
+{
+	v.visit(*this);
+}
+
+prepare_msg::prepare_msg(bool success, uint64_t index)
+	: block_msg(success, index)
+{
+}
+
+void
+prepare_msg::visit(journal_visitor &v) const
+{
+	v.visit(*this);
+}
+
+flush_msg::flush_msg(bool success)
+	: journal_msg(success)
+{
+}
+
+void
+flush_msg::visit(journal_visitor &v) const
+{
+	v.visit(*this);
+}
+
+flush_and_unlock_msg::flush_and_unlock_msg(bool success, uint64_t index, delta_list const &deltas)
+	: block_msg(success, index),
+	  deltas_(deltas)
+{
+}
+
+void
+flush_and_unlock_msg::visit(journal_visitor &v) const
+{
+	v.visit(*this);
+}
+
+prefetch_msg::prefetch_msg(bool success, uint64_t index)
+	: block_msg(success, index)
+{
+}
+
+void
+prefetch_msg::visit(journal_visitor &v) const
+{
+	v.visit(*this);
+}
+
+set_read_only_msg::set_read_only_msg()
+	: journal_msg(true)
+{
+}
+
+void
+set_read_only_msg::visit(journal_visitor &v) const
+{
+	v.visit(*this);
+}
+
+set_read_write_msg::set_read_write_msg()
+	: journal_msg(true)
+{
+}
+
+void
+set_read_write_msg::visit(journal_visitor &v) const
+{
+	v.visit(*this);
+}
+
+//------------------------------------------
+
+journal::journal(block_manager<JOURNAL_BLOCK_SIZE>::ptr bm)
+	: in_(bm)
+{
+}
+
+void
+journal::read_journal(struct journal_visitor &v)
+{
+	while (read_one_(v))
+		;
+}
+
+bool
+journal::read_one_(struct journal_visitor &v)
+{
+	uint8_t header = read_<uint8_t>();
+	uint8_t t = header >> 1;
+	uint8_t success = header & 0x1;
+	uint64_t index, nr_blocks;
+
+	switch (static_cast<msg_type>(t)) {
+	case MT_OPEN_JOURNAL:
+		nr_blocks = read_<uint64_t>();
+		v.visit(open_journal_msg(nr_blocks));
+                break;
+
+	case MT_CLOSE_JOURNAL:
+                v.visit(close_journal_msg());
+		return false;
+
+	case MT_READ_LOCK:
+		index = read_<uint64_t>();
+		v.visit(read_lock_msg(success, index));
+		break;
+
+	case MT_WRITE_LOCK:
+		index = read_<uint64_t>();
+		v.visit(write_lock_msg(success, index));
+		break;
+
+	case MT_ZERO_LOCK:
+		index = read_<uint64_t>();
+		v.visit(zero_lock_msg(success, index));
+		break;
+
+	case MT_TRY_READ_LOCK:
+		index = read_<uint64_t>();
+		v.visit(try_read_lock_msg(success, index));
+		break;
+
+	case MT_UNLOCK: {
+		index = read_<uint64_t>();
+		auto deltas = read_deltas_();
+		v.visit(unlock_msg(success, index, deltas));
+	}
+		break;
+
+	case MT_VERIFY:
+		index = read_<uint64_t>();
+                v.visit(verify_msg(success, index));
+		break;
+
+	case MT_PREPARE:
+                index = read_<uint64_t>();
+                v.visit(prepare_msg(success, index));
+		break;
+
+	case MT_FLUSH:
+                v.visit(flush_msg(success));
+		break;
+
+	case MT_FLUSH_AND_UNLOCK: {
+		cerr << "reading flush_and_unlock msg\n";
+		index = read_<uint64_t>();
+		auto deltas = read_deltas_();
+                v.visit(flush_and_unlock_msg(success, index, deltas));
+		}
+		break;
+
+	case MT_PREFETCH:
+		index = read_<uint64_t>();
+                v.visit(prefetch_msg(success, index));
+		break;
+
+	case MT_SET_READ_ONLY:
+                v.visit(set_read_only_msg());
+		break;
+
+	case MT_SET_READ_WRITE:
+                v.visit(set_read_write_msg());
+		break;
+	}
+
+	return true;
+}
+
+bool
+journal::read_delta_(delta_list &ds)
+{
+	uint16_t chunk = read_<uint16_t>();
+
+	if (chunk == 0xffff)
+		return false;
+
+	assert(chunk < JOURNAL_NR_CHUNKS);
+
+	auto bytes = vector<uint8_t>(JOURNAL_CHUNK_SIZE, 0);
+	in_.read_bytes(bytes.data(), bytes.data() + JOURNAL_CHUNK_SIZE);
+	ds.push_back(delta(chunk * JOURNAL_CHUNK_SIZE, bytes));
+
+	return true;
+}
+
+thin_provisioning::delta_list
+journal::read_deltas_()
+{
+	delta_list ds;
+
+	while (read_delta_(ds))
+		;
+
+	return ds;
+}
+
+//----------------------------------------------------------------
+

thin-provisioning/thin_journal.h

@@ -0,0 +1,218 @@
+// Copyright (C) 2018 Red Hat, Inc. All rights reserved.
+//
+// This file is part of the thin-provisioning-tools source.
+//
+// thin-provisioning-tools is free software: you can redistribute it
+// and/or modify it under the terms of the GNU General Public License
+// as published by the Free Software Foundation, either version 3 of
+// the License, or (at your option) any later version.
+//
+// thin-provisioning-tools is distributed in the hope that it will be
+// useful, but WITHOUT ANY WARRANTY; without even the implied warranty
+// of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License along
+// with thin-provisioning-tools.  If not, see
+// <http://www.gnu.org/licenses/>.
+
+#ifndef THIN_PROVISIONING_THIN_JOURNAL_H
+#define THIN_PROVISIONING_THIN_JOURNAL_H
+
+#include "persistent-data/block.h"
+
+#include <vector>
+
+//----------------------------------------------------------------
+
+namespace thin_provisioning {
+	uint32_t const JOURNAL_BLOCK_SIZE = 256 * 1024;
+	uint32_t const JOURNAL_CHUNK_SIZE = 32;
+	uint32_t const JOURNAL_NR_CHUNKS = (4096 / JOURNAL_CHUNK_SIZE);
+
+	class byte_stream {
+	public:
+		byte_stream(persistent_data::block_manager<JOURNAL_BLOCK_SIZE>::ptr bm);
+
+		void read_bytes(uint8_t *b, uint8_t *e);
+
+	private:
+		void next_block_();
+		size_t read_some_(uint8_t *b, uint8_t *e);
+
+		persistent_data::block_manager<JOURNAL_BLOCK_SIZE>::ptr bm_;
+
+		uint64_t current_block_;
+		uint64_t cursor_;
+	};
+
+	//---------------------------------
+
+	class journal_visitor;
+
+	struct journal_msg {
+		journal_msg(bool success);
+
+		virtual ~journal_msg() {}
+		virtual void visit(journal_visitor &v) const = 0;
+
+		bool success_;
+	};
+
+	struct open_journal_msg : public journal_msg {
+		open_journal_msg(uint64_t nr_metadata_blocks);
+		virtual void visit(journal_visitor &v) const;
+		uint64_t nr_metadata_blocks_;
+	};
+
+	struct close_journal_msg : public journal_msg {
+		close_journal_msg();
+		virtual void visit(journal_visitor &v) const;
+	};
+
+	struct block_msg : public journal_msg {
+		block_msg(bool success, uint64_t index);
+		uint64_t index_;
+	};
+
+	struct read_lock_msg : public block_msg {
+		read_lock_msg(bool success, uint64_t index);
+		virtual void visit(journal_visitor &v) const;
+	};
+
+	struct write_lock_msg : public block_msg {
+		write_lock_msg(bool success, uint64_t index);
+		virtual void visit(journal_visitor &v) const;
+	};
+
+	struct zero_lock_msg : public block_msg {
+		zero_lock_msg(bool success, uint64_t index);
+		virtual void visit(journal_visitor &v) const;
+	};
+
+	struct try_read_lock_msg : public block_msg {
+		try_read_lock_msg(bool success, uint64_t index);
+		virtual void visit(journal_visitor &v) const;
+	};
+
+	struct delta {
+		delta(uint32_t offset, std::vector<uint8_t> &bytes)
+			: offset_(offset),
+			  bytes_(bytes) {
+		}
+
+		uint32_t offset_;
+		std::vector<uint8_t> bytes_;
+	};
+
+	using delta_list = std::vector<delta>;
+
+	struct unlock_msg : public block_msg {
+		unlock_msg(bool success, uint64_t index, delta_list const &deltas);
+		virtual void visit(journal_visitor &v) const;
+
+		delta_list deltas_;
+	};
+
+	struct verify_msg : public block_msg {
+		verify_msg(bool success, uint64_t index);
+		virtual void visit(journal_visitor &v) const;
+	};
+
+	struct prepare_msg : public block_msg {
+		prepare_msg(bool success, uint64_t index);
+		virtual void visit(journal_visitor &v) const;
+	};
+
+	struct flush_msg : public journal_msg {
+		flush_msg(bool success);
+		virtual void visit(journal_visitor &v) const;
+	};
+
+	struct flush_and_unlock_msg : public block_msg {
+		flush_and_unlock_msg(bool success, uint64_t index, delta_list const &deltas);
+		virtual void visit(journal_visitor &v) const;
+
+		delta_list deltas_;
+	};
+
+	struct prefetch_msg : public block_msg {
+		prefetch_msg(bool success, uint64_t index);
+		virtual void visit(journal_visitor &v) const;
+	};
+
+	struct set_read_only_msg : public journal_msg {
+		set_read_only_msg();
+		virtual void visit(journal_visitor &v) const;
+	};
+
+	struct set_read_write_msg : public journal_msg {
+		set_read_write_msg();
+		virtual void visit(journal_visitor &v) const;
+	};
+
+	struct journal_visitor {
+	public:
+		virtual ~journal_visitor() {};
+
+		void visit(journal_msg const &msg) {
+			msg.visit(*this);
+		}
+
+		virtual void visit(open_journal_msg const &msg) = 0;
+		virtual void visit(close_journal_msg const &msg) = 0;
+		virtual void visit(read_lock_msg const &msg) = 0;
+		virtual void visit(write_lock_msg const &msg) = 0;
+		virtual void visit(zero_lock_msg const &msg) = 0;
+		virtual void visit(try_read_lock_msg const &msg) = 0;
+		virtual void visit(unlock_msg const &msg) = 0;
+		virtual void visit(verify_msg const &msg) = 0;
+		virtual void visit(prepare_msg const &msg) = 0;
+		virtual void visit(flush_msg const &msg) = 0;
+		virtual void visit(flush_and_unlock_msg const &msg) = 0;
+		virtual void visit(prefetch_msg const &msg) = 0;
+		virtual void visit(set_read_only_msg const &msg) = 0;
+		virtual void visit(set_read_write_msg const &msg) = 0;
+	};
+
+	enum msg_type {
+		MT_OPEN_JOURNAL,
+		MT_CLOSE_JOURNAL,
+
+		MT_READ_LOCK,
+		MT_WRITE_LOCK,
+		MT_ZERO_LOCK,
+		MT_TRY_READ_LOCK,
+		MT_UNLOCK,
+		MT_VERIFY,
+		MT_PREPARE,
+		MT_FLUSH,
+		MT_FLUSH_AND_UNLOCK,
+		MT_PREFETCH,
+		MT_SET_READ_ONLY,
+		MT_SET_READ_WRITE,
+	};
+
+	class journal {
+	public:
+		journal(persistent_data::block_manager<JOURNAL_BLOCK_SIZE>::ptr bm);
+		void read_journal(struct journal_visitor &v);
+
+	private:
+		bool read_delta_(delta_list &ds);
+		delta_list read_deltas_();
+		bool read_one_(struct journal_visitor &v);
+
+		template <typename T> T read_() {
+			T r;
+			in_.read_bytes(reinterpret_cast<uint8_t *>(&r), reinterpret_cast<uint8_t *>(&r + 1));
+			return r;
+		}
+
+		byte_stream in_;
+	};
+}
+
+//----------------------------------------------------------------
+
+#endif

thin-provisioning/thin_journal_check.cc

@@ -0,0 +1,461 @@
+// Copyright (C) 2018 Red Hat, Inc. All rights reserved.
+//
+// This file is part of the thin-provisioning-tools source.
+//
+// thin-provisioning-tools is free software: you can redistribute it
+// and/or modify it under the terms of the GNU General Public License
+// as published by the Free Software Foundation, either version 3 of
+// the License, or (at your option) any later version.
+//
+// thin-provisioning-tools is distributed in the hope that it will be
+// useful, but WITHOUT ANY WARRANTY; without even the implied warranty
+// of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License along
+// with thin-provisioning-tools.  If not, see
+// <http://www.gnu.org/licenses/>.
+
+#include <iostream>
+#include <getopt.h>
+#include <libgen.h>
+#include <fcntl.h>
+
+#include <boost/lexical_cast.hpp>
+#include <boost/optional.hpp>
+
+#include <map>
+#include <set>
+
+#include "version.h"
+
+#include "base/application.h"
+#include "base/error_state.h"
+#include "base/file_utils.h"
+#include "base/nested_output.h"
+#include "persistent-data/data-structures/btree_counter.h"
+#include "persistent-data/space-maps/core.h"
+#include "persistent-data/space-maps/disk.h"
+#include "persistent-data/file_utils.h"
+#include "thin-provisioning/metadata.h"
+#include "thin-provisioning/device_tree.h"
+#include "thin-provisioning/mapping_tree.h"
+#include "thin-provisioning/metadata_counter.h"
+#include "thin-provisioning/superblock.h"
+#include "thin-provisioning/commands.h"
+#include "thin-provisioning/thin_journal.h"
+
+using namespace base;
+using namespace boost;
+using namespace file_utils;
+using namespace std;
+using namespace thin_provisioning;
+
+//----------------------------------------------------------------
+
+namespace {
+	class journal_display : public journal_visitor {
+	public:
+		journal_display(journal_visitor &inner)
+			: inner_(inner) {
+		}
+
+		virtual void visit(open_journal_msg const &msg) {
+			cout << "open_journal\n";
+			inner_.visit(msg);
+		}
+
+		virtual void visit(close_journal_msg const &msg) {
+			cout << "close_journal\n";
+			inner_.visit(msg);
+		}
+
+		virtual void visit(read_lock_msg const &msg) {
+			if (interesting(msg.index_))
+				cout << "read_lock " << msg.index_ << "\n";
+			inner_.visit(msg);
+		}
+
+		virtual void visit(write_lock_msg const &msg) {
+			if (interesting(msg.index_))
+				cout << "write_lock " << msg.index_ << "\n";
+			inner_.visit(msg);
+		}
+
+		virtual void visit(zero_lock_msg const &msg) {
+			if (interesting(msg.index_))
+				cout << "zero_lock " << msg.index_ << "\n";
+			inner_.visit(msg);
+		}
+
+		virtual void visit(try_read_lock_msg const &msg) {
+			if (interesting(msg.index_))
+				cout << "try_read_lock " << msg.index_ << "\n";
+			inner_.visit(msg);
+		}
+
+		virtual void visit(unlock_msg const &msg) {
+			if (interesting(msg.index_))
+				cout << "unlock " << msg.index_ << "\n";
+			inner_.visit(msg);
+		}
+
+		virtual void visit(verify_msg const &msg) {
+			if (interesting(msg.index_))
+				cout << "verify " << msg.index_ << "\n";
+			inner_.visit(msg);
+		}
+
+		virtual void visit(prepare_msg const &msg) {
+			if (interesting(msg.index_))
+				cout << "prepare " << msg.index_ << "\n";
+			inner_.visit(msg);
+		}
+
+		virtual void visit(flush_msg const &msg) {
+			cout << "flush\n";
+			inner_.visit(msg);
+		}
+
+		virtual void visit(flush_and_unlock_msg const &msg) {
+			if (interesting(msg.index_))
+				cout << "flush_and_unlock " << msg.index_ << "\n";
+			inner_.visit(msg);
+		}
+
+		virtual void visit(prefetch_msg const &msg) {
+			if (interesting(msg.index_))
+				cout << "prefetch " << msg.index_ << "\n";
+			inner_.visit(msg);
+		}
+
+		virtual void visit(set_read_only_msg const &msg) {
+			cout << "set_read_only\n";
+			inner_.visit(msg);
+		}
+
+		virtual void visit(set_read_write_msg const &msg) {
+			cout << "set_read_write\n";
+			inner_.visit(msg);
+		}
+
+		bool interesting(block_address b) const {
+			return true;
+		}
+
+		journal_visitor &inner_;
+	};
+
+	unsigned const MAX_HELD_LOCKS = 16;
+
+	// We use a temporary file to hold all the deltas.  Assume the metadata initially starts zeroed.
+	// Need to introduce notion of 'time' that increments everytime a write lock is taken.
+	// Need to track updates to the superblock to define transactions.
+	class checker : public journal_visitor {
+	public:
+		virtual void visit(open_journal_msg const &msg) {
+			bm_.reset(new block_manager<>("metadata.tmp", msg.nr_metadata_blocks_,
+                                                      MAX_HELD_LOCKS, block_manager<>::CREATE));
+		}
+
+		virtual void visit(close_journal_msg const &msg) {
+			// noop
+		}
+
+		virtual void visit(read_lock_msg const &msg) {
+			if (msg.success_)
+				read_lock_(msg.index_);
+		}
+
+		virtual void visit(write_lock_msg const &msg) {
+			if (msg.success_)
+				write_lock_(msg.index_);
+		}
+
+		virtual void visit(zero_lock_msg const &msg) {
+			if (msg.success_) {
+				write_lock_(msg.index_);
+				zero_(msg.index_);
+			}
+		}
+
+		virtual void visit(try_read_lock_msg const &msg) {
+			if (msg.success_)
+				read_lock_(msg.index_);
+		}
+
+		virtual void visit(unlock_msg const &msg) {
+			bool write_locked = is_write_locked_(msg.index_);
+
+			unlock_(msg.index_, msg.deltas_);
+
+			if (write_locked && msg.index_ == superblock_detail::SUPERBLOCK_LOCATION)
+				commit_();
+		}
+
+		virtual void visit(verify_msg const &msg) {
+			// noop
+		}
+
+		virtual void visit(prepare_msg const &msg) {
+			// noop
+		}
+
+		virtual void visit(flush_msg const &msg) {
+			cout << "WARN: spurious flush()\n";
+		}
+
+		virtual void visit(flush_and_unlock_msg const &msg) {
+			if (msg.index_ != superblock_detail::SUPERBLOCK_LOCATION) {
+				cout << "ERROR: flush_and_unlock received for block " << msg.index_
+				     << ", which isn't the superblock\n";
+			}
+
+			unlock_(msg.index_, msg.deltas_);
+			commit_();
+		}
+
+		virtual void visit(prefetch_msg const &msg) {
+			// ignore
+		}
+
+		virtual void visit(set_read_only_msg const &msg) {
+			// ignore
+		}
+
+		virtual void visit(set_read_write_msg const &msg) {
+			// ignore
+		}
+
+	private:
+		void read_lock_(block_address b) {
+			auto it = locks_.find(b);
+			if (it == locks_.end())
+				locks_.insert(make_pair(b, -1));
+
+			else if (it->second > 0) {
+				cout << "WARN: read lock taken concurrently with write lock for block "
+				     << b << "\n";
+
+			} else
+				--it->second;
+		}
+
+		void write_lock_(block_address b) {
+			if (is_superblock_(b)) {
+				if (locks_.size())
+					cout << "WARN: superblock taken when locks still held\n";
+
+			} else if (active_.count(b)) {
+				cout << "ERROR: write lock taken for block "
+				     << b
+				     << ", but it is still in the active transaction\n";
+				throw runtime_error("bad write_lock");
+			}
+
+			auto it = locks_.find(b);
+			if (it == locks_.end())
+				locks_.insert(make_pair(b, 1));
+
+			else if (it->second < 0) {
+				cout << "WARN: write lock requested for read locked block "
+				     << b << "\n";
+			} else
+				it->second++;
+		}
+
+		bool is_write_locked_(block_address b) const {
+			auto it = locks_.find(b);
+			return it != locks_.end() && it->second > 0;
+		}
+
+		void unlock_(block_address b, delta_list const &deltas) {
+			auto it = locks_.find(b);
+			if (it == locks_.end() || !it->second) {
+				cout << "ERROR: unlock requested on block " << b << ", which isn't locked\n";
+				throw runtime_error("bad unlock");
+			}
+
+			if (it->second < 0) {
+				it->second++;
+
+				if (deltas.size()) {
+					cout << "ERROR: unlocking a read lock for " << b << ", yet there are " << deltas.size() << " deltas\n";
+					throw runtime_error("bad unlock");
+				}
+			} else {
+				auto wr = bm_->write_lock(b);
+
+				for (auto &&d : deltas) {
+					uint8_t *data = static_cast<uint8_t *>(wr.data());
+
+					if (d.offset_ + d.bytes_.size() > 4096) {
+						cout << "ERROR: delta for block " << b << " is out of range ("
+                                                     << d.offset_ << ", " << d.offset_ + d.bytes_.size() << "]\n";
+                                                throw runtime_error("bad unlock");
+                                        }
+
+					memcpy(data + d.offset_, d.bytes_.data(), d.bytes_.size());
+				}
+
+				it->second--;
+			}
+
+			if (!it->second)
+				locks_.erase(it);
+		}
+
+		void zero_(block_address b) {
+			auto wr = bm_->write_lock_zero(b);
+		}
+
+		void commit_() {
+			using namespace thin_provisioning::superblock_detail;
+
+			// At this point the only lock held should be the superblock,
+			// and that should be a write lock.
+			if (locks_.size() != 0) {
+				cout << "ERROR: committing when the following locks are still held:\n";
+				for (auto &&p : locks_)
+					if (p.first != SUPERBLOCK_LOCATION)
+						cerr << p.first << "\n";
+				throw runtime_error("bad commit");
+			}
+
+			build_active_set_();
+		}
+
+		void build_active_set_() {
+			using namespace thin_provisioning::superblock_detail;
+
+			cerr << "build active set\n";
+			superblock sb = read_superblock(bm_);
+			block_counter bc;
+
+			auto tm = open_tm(bm_, SUPERBLOCK_LOCATION);
+			auto sm = open_metadata_sm(*tm, &sb.metadata_space_map_root_);
+			tm->set_sm(sm);
+
+			// FIXME: check we don't have a space leak from a cycle between the sm and tm
+
+			count_metadata(tm, sb, bc);
+
+			active_.clear();
+			active_.insert(SUPERBLOCK_LOCATION);
+
+			for (auto &&p : bc.get_counts()) {
+				if (!p.second) {
+					cout << "weird zero count for block " << p.first << "\n";
+				}
+
+				active_.insert(p.first);
+			}
+		}
+
+		bool is_superblock_(block_address b) const {
+			return b == superblock_detail::SUPERBLOCK_LOCATION;
+		}
+
+		typedef set<block_address> block_set;
+
+		// write locks positive, unlocked 0, read locks negative
+		typedef map<block_address, int> block_map;
+
+		block_set active_;
+		block_map locks_;
+
+		block_manager<>::ptr bm_;
+		transaction_manager::ptr tm_;
+	};
+
+	struct flags {
+		flags()
+			: quiet(false) {
+		}
+
+		bool quiet;
+	};
+
+	void check(string const &path) {
+		block_address journal_size = get_file_length(path) / JOURNAL_BLOCK_SIZE;
+		block_manager<JOURNAL_BLOCK_SIZE>::ptr bm(
+			new block_manager<JOURNAL_BLOCK_SIZE>(path, journal_size, 4,
+                                          block_manager<JOURNAL_BLOCK_SIZE>::READ_ONLY));
+		journal j(bm);
+		checker c;
+		journal_display dc(c);
+
+		j.read_journal(c);
+	}
+}
+
+//----------------------------------------------------------------
+
+thin_journal_cmd::thin_journal_cmd()
+	: command("thin_journal_check")
+{
+}
+
+void
+thin_journal_cmd::usage(std::ostream &out) const
+{
+	out << "Usage: " << get_name() << " [options] {device|file}" << endl
+	    << "Options:\n"
+	    << "  {-q|--quiet}\n"
+	    << "  {-h|--help}\n"
+	    << "  {-V|--version}\n";
+}
+
+int
+thin_journal_cmd::run(int argc, char **argv)
+{
+	int c;
+	flags fs;
+
+	char const shortopts[] = "qhV";
+	option const longopts[] = {
+		{ "quiet", no_argument, NULL, 'q'},
+		{ "help", no_argument, NULL, 'h'},
+		{ "version", no_argument, NULL, 'V'},
+		{ NULL, no_argument, NULL, 0 }
+	};
+
+	while ((c = getopt_long(argc, argv, shortopts, longopts, NULL)) != -1) {
+		switch(c) {
+		case 'h':
+			usage(cout);
+			return 0;
+
+		case 'q':
+			fs.quiet = true;
+			break;
+
+		case 'V':
+			cout << THIN_PROVISIONING_TOOLS_VERSION << endl;
+			return 0;
+
+		default:
+			usage(cerr);
+			return 1;
+		}
+	}
+
+	if (argc - optind != 1) {
+		if (!fs.quiet)
+			usage(cerr);
+
+		exit(1);
+	}
+
+	try {
+		check(argv[optind]);
+
+	} catch (std::exception &e) {
+		cerr << e.what() << "\n";
+		return 1;
+	}
+
+	return 0;
+}
+
+//----------------------------------------------------------------