| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | // Copyright 2019 The go-ethereum Authors | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | // This file is part of the go-ethereum library. | 
					
						
							|  |  |  | // | 
					
						
							|  |  |  | // The go-ethereum library is free software: you can redistribute it and/or modify | 
					
						
							|  |  |  | // it under the terms of the GNU Lesser General Public License as published by | 
					
						
							|  |  |  | // the Free Software Foundation, either version 3 of the License, or | 
					
						
							|  |  |  | // (at your option) any later version. | 
					
						
							|  |  |  | // | 
					
						
							|  |  |  | // The go-ethereum library 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 Lesser General Public License for more details. | 
					
						
							|  |  |  | // | 
					
						
							|  |  |  | // You should have received a copy of the GNU Lesser General Public License | 
					
						
							|  |  |  | // along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>. | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | package rawdb | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | import ( | 
					
						
							| 
									
										
										
										
											2021-09-07 12:31:17 +02:00
										 |  |  | 	"bytes" | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	"encoding/binary" | 
					
						
							|  |  |  | 	"errors" | 
					
						
							|  |  |  | 	"fmt" | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | 	"io" | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	"os" | 
					
						
							|  |  |  | 	"path/filepath" | 
					
						
							|  |  |  | 	"sync" | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	"sync/atomic" | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 
 | 
					
						
							|  |  |  | 	"github.com/ethereum/go-ethereum/common" | 
					
						
							|  |  |  | 	"github.com/ethereum/go-ethereum/log" | 
					
						
							|  |  |  | 	"github.com/ethereum/go-ethereum/metrics" | 
					
						
							|  |  |  | 	"github.com/golang/snappy" | 
					
						
							|  |  |  | ) | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | var ( | 
					
						
							|  |  |  | 	// errClosed is returned if an operation attempts to read from or write to the | 
					
						
							|  |  |  | 	// freezer table after it has already been closed. | 
					
						
							|  |  |  | 	errClosed = errors.New("closed") | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | 	// errOutOfBounds is returned if the item requested is not contained within the | 
					
						
							|  |  |  | 	// freezer table. | 
					
						
							|  |  |  | 	errOutOfBounds = errors.New("out of bounds") | 
					
						
							| 
									
										
										
											
												all: integrate the freezer with fast sync
* all: freezer style syncing
core, eth, les, light: clean up freezer relative APIs
core, eth, les, trie, ethdb, light: clean a bit
core, eth, les, light: add unit tests
core, light: rewrite setHead function
core, eth: fix downloader unit tests
core: add receipt chain insertion test
core: use constant instead of hardcoding table name
core: fix rollback
core: fix setHead
core/rawdb: remove canonical block first and then iterate side chain
core/rawdb, ethdb: add hasAncient interface
eth/downloader: calculate ancient limit via cht first
core, eth, ethdb: lots of fixes
* eth/downloader: print ancient disable log only for fast sync
											
										 
											2019-04-25 22:59:48 +08:00
										 |  |  | 
 | 
					
						
							|  |  |  | 	// errNotSupported is returned if the database doesn't support the required operation. | 
					
						
							|  |  |  | 	errNotSupported = errors.New("this operation is not supported") | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | ) | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | // indexEntry contains the number/id of the file that the data resides in, aswell as the | 
					
						
							|  |  |  | // offset within the file to the end of the data | 
					
						
							|  |  |  | // In serialized form, the filenum is stored as uint16. | 
					
						
							|  |  |  | type indexEntry struct { | 
					
						
							|  |  |  | 	filenum uint32 // stored as uint16 ( 2 bytes) | 
					
						
							|  |  |  | 	offset  uint32 // stored as uint32 ( 4 bytes) | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | const indexEntrySize = 6 | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2021-09-07 12:31:17 +02:00
										 |  |  | // unmarshalBinary deserializes binary b into the rawIndex entry. | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | func (i *indexEntry) unmarshalBinary(b []byte) error { | 
					
						
							|  |  |  | 	i.filenum = uint32(binary.BigEndian.Uint16(b[:2])) | 
					
						
							|  |  |  | 	i.offset = binary.BigEndian.Uint32(b[2:6]) | 
					
						
							|  |  |  | 	return nil | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2021-09-07 12:31:17 +02:00
										 |  |  | // append adds the encoded entry to the end of b. | 
					
						
							|  |  |  | func (i *indexEntry) append(b []byte) []byte { | 
					
						
							|  |  |  | 	offset := len(b) | 
					
						
							|  |  |  | 	out := append(b, make([]byte, indexEntrySize)...) | 
					
						
							|  |  |  | 	binary.BigEndian.PutUint16(out[offset:], uint16(i.filenum)) | 
					
						
							|  |  |  | 	binary.BigEndian.PutUint32(out[offset+2:], i.offset) | 
					
						
							|  |  |  | 	return out | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | // bounds returns the start- and end- offsets, and the file number of where to | 
					
						
							|  |  |  | // read there data item marked by the two index entries. The two entries are | 
					
						
							|  |  |  | // assumed to be sequential. | 
					
						
							|  |  |  | func (start *indexEntry) bounds(end *indexEntry) (startOffset, endOffset, fileId uint32) { | 
					
						
							|  |  |  | 	if start.filenum != end.filenum { | 
					
						
							|  |  |  | 		// If a piece of data 'crosses' a data-file, | 
					
						
							|  |  |  | 		// it's actually in one piece on the second data-file. | 
					
						
							|  |  |  | 		// We return a zero-indexEntry for the second file as start | 
					
						
							|  |  |  | 		return 0, end.offset, end.filenum | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	return start.offset, end.offset, end.filenum | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | // freezerTable represents a single chained data table within the freezer (e.g. blocks). | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | // It consists of a data file (snappy encoded arbitrary data blobs) and an indexEntry | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | // file (uncompressed 64 bit indices into the data file). | 
					
						
							|  |  |  | type freezerTable struct { | 
					
						
							| 
									
										
										
										
											2019-05-17 01:45:05 +03:00
										 |  |  | 	// WARNING: The `items` field is accessed atomically. On 32 bit platforms, only | 
					
						
							|  |  |  | 	// 64-bit aligned fields can be atomic. The struct is guaranteed to be so aligned, | 
					
						
							|  |  |  | 	// so take advantage of that (https://golang.org/pkg/sync/atomic/#pkg-note-BUG). | 
					
						
							|  |  |  | 	items uint64 // Number of items stored in the table (including items removed from tail) | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2022-01-18 10:29:38 +01:00
										 |  |  | 	noCompression bool // if true, disables snappy compression. Note: does not work retroactively | 
					
						
							|  |  |  | 	readonly      bool | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	maxFileSize   uint32 // Max file size for data-files | 
					
						
							|  |  |  | 	name          string | 
					
						
							|  |  |  | 	path          string | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	head   *os.File            // File descriptor for the data head of the table | 
					
						
							|  |  |  | 	files  map[uint32]*os.File // open files | 
					
						
							|  |  |  | 	headId uint32              // number of the currently active head file | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 	tailId uint32              // number of the earliest file | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	index  *os.File            // File descriptor for the indexEntry file of the table | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 	// In the case that old items are deleted (from the tail), we use itemOffset | 
					
						
							|  |  |  | 	// to count how many historic items have gone missing. | 
					
						
							|  |  |  | 	itemOffset uint32 // Offset (number of discarded items) | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2021-09-07 12:31:17 +02:00
										 |  |  | 	headBytes  int64         // Number of bytes written to the head file | 
					
						
							| 
									
										
										
										
											2019-09-10 14:39:07 +03:00
										 |  |  | 	readMeter  metrics.Meter // Meter for measuring the effective amount of data read | 
					
						
							|  |  |  | 	writeMeter metrics.Meter // Meter for measuring the effective amount of data written | 
					
						
							|  |  |  | 	sizeGauge  metrics.Gauge // Gauge for tracking the combined size of all freezer tables | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 
 | 
					
						
							|  |  |  | 	logger log.Logger   // Logger with database path and table name ambedded | 
					
						
							|  |  |  | 	lock   sync.RWMutex // Mutex protecting the data file descriptors | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2021-01-26 12:17:11 +01:00
										 |  |  | // NewFreezerTable opens the given path as a freezer table. | 
					
						
							| 
									
										
										
										
											2022-01-18 10:29:38 +01:00
										 |  |  | func NewFreezerTable(path, name string, disableSnappy, readonly bool) (*freezerTable, error) { | 
					
						
							|  |  |  | 	return newTable(path, name, metrics.NilMeter{}, metrics.NilMeter{}, metrics.NilGauge{}, freezerTableSize, disableSnappy, readonly) | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | // openFreezerFileForAppend opens a freezer table file and seeks to the end | 
					
						
							|  |  |  | func openFreezerFileForAppend(filename string) (*os.File, error) { | 
					
						
							|  |  |  | 	// Open the file without the O_APPEND flag | 
					
						
							|  |  |  | 	// because it has differing behaviour during Truncate operations | 
					
						
							|  |  |  | 	// on different OS's | 
					
						
							|  |  |  | 	file, err := os.OpenFile(filename, os.O_RDWR|os.O_CREATE, 0644) | 
					
						
							|  |  |  | 	if err != nil { | 
					
						
							|  |  |  | 		return nil, err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	// Seek to end for append | 
					
						
							|  |  |  | 	if _, err = file.Seek(0, io.SeekEnd); err != nil { | 
					
						
							|  |  |  | 		return nil, err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	return file, nil | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | // openFreezerFileForReadOnly opens a freezer table file for read only access | 
					
						
							|  |  |  | func openFreezerFileForReadOnly(filename string) (*os.File, error) { | 
					
						
							|  |  |  | 	return os.OpenFile(filename, os.O_RDONLY, 0644) | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | // openFreezerFileTruncated opens a freezer table making sure it is truncated | 
					
						
							|  |  |  | func openFreezerFileTruncated(filename string) (*os.File, error) { | 
					
						
							|  |  |  | 	return os.OpenFile(filename, os.O_RDWR|os.O_CREATE|os.O_TRUNC, 0644) | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | // truncateFreezerFile resizes a freezer table file and seeks to the end | 
					
						
							|  |  |  | func truncateFreezerFile(file *os.File, size int64) error { | 
					
						
							|  |  |  | 	if err := file.Truncate(size); err != nil { | 
					
						
							|  |  |  | 		return err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	// Seek to end for append | 
					
						
							|  |  |  | 	if _, err := file.Seek(0, io.SeekEnd); err != nil { | 
					
						
							|  |  |  | 		return err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	return nil | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2021-09-07 12:31:17 +02:00
										 |  |  | // newTable opens a freezer table, creating the data and index files if they are | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | // non existent. Both files are truncated to the shortest common length to ensure | 
					
						
							|  |  |  | // they don't go out of sync. | 
					
						
							| 
									
										
										
										
											2022-01-18 10:29:38 +01:00
										 |  |  | func newTable(path string, name string, readMeter metrics.Meter, writeMeter metrics.Meter, sizeGauge metrics.Gauge, maxFilesize uint32, noCompression, readonly bool) (*freezerTable, error) { | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	// Ensure the containing directory exists and open the indexEntry file | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	if err := os.MkdirAll(path, 0755); err != nil { | 
					
						
							|  |  |  | 		return nil, err | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	var idxName string | 
					
						
							|  |  |  | 	if noCompression { | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | 		// Raw idx | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		idxName = fmt.Sprintf("%s.ridx", name) | 
					
						
							|  |  |  | 	} else { | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | 		// Compressed idx | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		idxName = fmt.Sprintf("%s.cidx", name) | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2022-01-18 10:29:38 +01:00
										 |  |  | 	var ( | 
					
						
							|  |  |  | 		err     error | 
					
						
							|  |  |  | 		offsets *os.File | 
					
						
							|  |  |  | 	) | 
					
						
							|  |  |  | 	if readonly { | 
					
						
							|  |  |  | 		// Will fail if table doesn't exist | 
					
						
							|  |  |  | 		offsets, err = openFreezerFileForReadOnly(filepath.Join(path, idxName)) | 
					
						
							|  |  |  | 	} else { | 
					
						
							|  |  |  | 		offsets, err = openFreezerFileForAppend(filepath.Join(path, idxName)) | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	if err != nil { | 
					
						
							|  |  |  | 		return nil, err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	// Create the table and repair any past inconsistency | 
					
						
							|  |  |  | 	tab := &freezerTable{ | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		index:         offsets, | 
					
						
							|  |  |  | 		files:         make(map[uint32]*os.File), | 
					
						
							|  |  |  | 		readMeter:     readMeter, | 
					
						
							|  |  |  | 		writeMeter:    writeMeter, | 
					
						
							| 
									
										
										
										
											2019-09-10 14:39:07 +03:00
										 |  |  | 		sizeGauge:     sizeGauge, | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		name:          name, | 
					
						
							|  |  |  | 		path:          path, | 
					
						
							|  |  |  | 		logger:        log.New("database", path, "table", name), | 
					
						
							|  |  |  | 		noCompression: noCompression, | 
					
						
							| 
									
										
										
										
											2022-01-18 10:29:38 +01:00
										 |  |  | 		readonly:      readonly, | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		maxFileSize:   maxFilesize, | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	} | 
					
						
							|  |  |  | 	if err := tab.repair(); err != nil { | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		tab.Close() | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 		return nil, err | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-06-10 14:21:02 +03:00
										 |  |  | 	// Initialize the starting size counter | 
					
						
							|  |  |  | 	size, err := tab.sizeNolock() | 
					
						
							|  |  |  | 	if err != nil { | 
					
						
							|  |  |  | 		tab.Close() | 
					
						
							|  |  |  | 		return nil, err | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-09-10 14:39:07 +03:00
										 |  |  | 	tab.sizeGauge.Inc(int64(size)) | 
					
						
							| 
									
										
										
										
											2019-06-10 14:21:02 +03:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	return tab, nil | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | // repair cross checks the head and the index file and truncates them to | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | // be in sync with each other after a potential crash / data loss. | 
					
						
							|  |  |  | func (t *freezerTable) repair() error { | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	// Create a temporary offset buffer to init files with and read indexEntry into | 
					
						
							|  |  |  | 	buffer := make([]byte, indexEntrySize) | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	// If we've just created the files, initialize the index with the 0 indexEntry | 
					
						
							|  |  |  | 	stat, err := t.index.Stat() | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	if err != nil { | 
					
						
							|  |  |  | 		return err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	if stat.Size() == 0 { | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		if _, err := t.index.Write(buffer); err != nil { | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 			return err | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	// Ensure the index is a multiple of indexEntrySize bytes | 
					
						
							|  |  |  | 	if overflow := stat.Size() % indexEntrySize; overflow != 0 { | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | 		truncateFreezerFile(t.index, stat.Size()-overflow) // New file can't trigger this path | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	} | 
					
						
							|  |  |  | 	// Retrieve the file sizes and prepare for truncation | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	if stat, err = t.index.Stat(); err != nil { | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 		return err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	offsetsSize := stat.Size() | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	// Open the head file | 
					
						
							|  |  |  | 	var ( | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 		firstIndex  indexEntry | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		lastIndex   indexEntry | 
					
						
							|  |  |  | 		contentSize int64 | 
					
						
							|  |  |  | 		contentExp  int64 | 
					
						
							|  |  |  | 	) | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 	// Read index zero, determine what file is the earliest | 
					
						
							|  |  |  | 	// and what item offset to use | 
					
						
							|  |  |  | 	t.index.ReadAt(buffer, 0) | 
					
						
							|  |  |  | 	firstIndex.unmarshalBinary(buffer) | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2020-06-17 15:41:07 +08:00
										 |  |  | 	t.tailId = firstIndex.filenum | 
					
						
							|  |  |  | 	t.itemOffset = firstIndex.offset | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	t.index.ReadAt(buffer, offsetsSize-indexEntrySize) | 
					
						
							|  |  |  | 	lastIndex.unmarshalBinary(buffer) | 
					
						
							| 
									
										
										
										
											2022-01-18 10:29:38 +01:00
										 |  |  | 	if t.readonly { | 
					
						
							|  |  |  | 		t.head, err = t.openFile(lastIndex.filenum, openFreezerFileForReadOnly) | 
					
						
							|  |  |  | 	} else { | 
					
						
							|  |  |  | 		t.head, err = t.openFile(lastIndex.filenum, openFreezerFileForAppend) | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	if err != nil { | 
					
						
							|  |  |  | 		return err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	if stat, err = t.head.Stat(); err != nil { | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 		return err | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	contentSize = stat.Size() | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 
 | 
					
						
							|  |  |  | 	// Keep truncating both files until they come in sync | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	contentExp = int64(lastIndex.offset) | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 
 | 
					
						
							|  |  |  | 	for contentExp != contentSize { | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		// Truncate the head file to the last offset pointer | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 		if contentExp < contentSize { | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 			t.logger.Warn("Truncating dangling head", "indexed", common.StorageSize(contentExp), "stored", common.StorageSize(contentSize)) | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | 			if err := truncateFreezerFile(t.head, contentExp); err != nil { | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 				return err | 
					
						
							|  |  |  | 			} | 
					
						
							|  |  |  | 			contentSize = contentExp | 
					
						
							|  |  |  | 		} | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		// Truncate the index to point within the head file | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 		if contentExp > contentSize { | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 			t.logger.Warn("Truncating dangling indexes", "indexed", common.StorageSize(contentExp), "stored", common.StorageSize(contentSize)) | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | 			if err := truncateFreezerFile(t.index, offsetsSize-indexEntrySize); err != nil { | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 				return err | 
					
						
							|  |  |  | 			} | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 			offsetsSize -= indexEntrySize | 
					
						
							|  |  |  | 			t.index.ReadAt(buffer, offsetsSize-indexEntrySize) | 
					
						
							|  |  |  | 			var newLastIndex indexEntry | 
					
						
							|  |  |  | 			newLastIndex.unmarshalBinary(buffer) | 
					
						
							|  |  |  | 			// We might have slipped back into an earlier head-file here | 
					
						
							|  |  |  | 			if newLastIndex.filenum != lastIndex.filenum { | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | 				// Release earlier opened file | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 				t.releaseFile(lastIndex.filenum) | 
					
						
							| 
									
										
										
										
											2019-07-22 09:51:13 +02:00
										 |  |  | 				if t.head, err = t.openFile(newLastIndex.filenum, openFreezerFileForAppend); err != nil { | 
					
						
							|  |  |  | 					return err | 
					
						
							|  |  |  | 				} | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 				if stat, err = t.head.Stat(); err != nil { | 
					
						
							|  |  |  | 					// TODO, anything more we can do here? | 
					
						
							|  |  |  | 					// A data file has gone missing... | 
					
						
							|  |  |  | 					return err | 
					
						
							|  |  |  | 				} | 
					
						
							|  |  |  | 				contentSize = stat.Size() | 
					
						
							|  |  |  | 			} | 
					
						
							|  |  |  | 			lastIndex = newLastIndex | 
					
						
							|  |  |  | 			contentExp = int64(lastIndex.offset) | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 		} | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2022-01-18 10:29:38 +01:00
										 |  |  | 	// Sync() fails for read-only files on windows. | 
					
						
							|  |  |  | 	if !t.readonly { | 
					
						
							|  |  |  | 		// Ensure all reparation changes have been written to disk | 
					
						
							|  |  |  | 		if err := t.index.Sync(); err != nil { | 
					
						
							|  |  |  | 			return err | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 		if err := t.head.Sync(); err != nil { | 
					
						
							|  |  |  | 			return err | 
					
						
							|  |  |  | 		} | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	} | 
					
						
							|  |  |  | 	// Update the item and byte counters and return | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 	t.items = uint64(t.itemOffset) + uint64(offsetsSize/indexEntrySize-1) // last indexEntry points to the end of the data file | 
					
						
							| 
									
										
										
										
											2021-09-07 12:31:17 +02:00
										 |  |  | 	t.headBytes = contentSize | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	t.headId = lastIndex.filenum | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	// Close opened files and preopen all files | 
					
						
							|  |  |  | 	if err := t.preopen(); err != nil { | 
					
						
							|  |  |  | 		return err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	t.logger.Debug("Chain freezer table opened", "items", t.items, "size", common.StorageSize(t.headBytes)) | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	return nil | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | // preopen opens all files that the freezer will need. This method should be called from an init-context, | 
					
						
							|  |  |  | // since it assumes that it doesn't have to bother with locking | 
					
						
							|  |  |  | // The rationale for doing preopen is to not have to do it from within Retrieve, thus not needing to ever | 
					
						
							|  |  |  | // obtain a write-lock within Retrieve. | 
					
						
							|  |  |  | func (t *freezerTable) preopen() (err error) { | 
					
						
							|  |  |  | 	// The repair might have already opened (some) files | 
					
						
							|  |  |  | 	t.releaseFilesAfter(0, false) | 
					
						
							|  |  |  | 	// Open all except head in RDONLY | 
					
						
							| 
									
										
										
										
											2019-05-16 14:30:11 +03:00
										 |  |  | 	for i := t.tailId; i < t.headId; i++ { | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | 		if _, err = t.openFile(i, openFreezerFileForReadOnly); err != nil { | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 			return err | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2022-01-18 10:29:38 +01:00
										 |  |  | 	if t.readonly { | 
					
						
							|  |  |  | 		t.head, err = t.openFile(t.headId, openFreezerFileForReadOnly) | 
					
						
							|  |  |  | 	} else { | 
					
						
							|  |  |  | 		// Open head in read/write | 
					
						
							|  |  |  | 		t.head, err = t.openFile(t.headId, openFreezerFileForAppend) | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	return err | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | // truncate discards any recent data above the provided threshold number. | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | func (t *freezerTable) truncate(items uint64) error { | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	t.lock.Lock() | 
					
						
							|  |  |  | 	defer t.lock.Unlock() | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 
 | 
					
						
							|  |  |  | 	// If our item count is correct, don't do anything | 
					
						
							| 
									
										
										
										
											2020-08-25 08:45:41 +03:00
										 |  |  | 	existing := atomic.LoadUint64(&t.items) | 
					
						
							|  |  |  | 	if existing <= items { | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 		return nil | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-06-10 14:21:02 +03:00
										 |  |  | 	// We need to truncate, save the old size for metrics tracking | 
					
						
							|  |  |  | 	oldSize, err := t.sizeNolock() | 
					
						
							|  |  |  | 	if err != nil { | 
					
						
							|  |  |  | 		return err | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	// Something's out of sync, truncate the table's offset index | 
					
						
							| 
									
										
										
										
											2020-08-25 08:45:41 +03:00
										 |  |  | 	log := t.logger.Debug | 
					
						
							|  |  |  | 	if existing > items+1 { | 
					
						
							|  |  |  | 		log = t.logger.Warn // Only loud warn if we delete multiple items | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	log("Truncating freezer table", "items", existing, "limit", items) | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | 	if err := truncateFreezerFile(t.index, int64(items+1)*indexEntrySize); err != nil { | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 		return err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	// Calculate the new expected size of the data file and truncate it | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	buffer := make([]byte, indexEntrySize) | 
					
						
							|  |  |  | 	if _, err := t.index.ReadAt(buffer, int64(items*indexEntrySize)); err != nil { | 
					
						
							|  |  |  | 		return err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	var expected indexEntry | 
					
						
							|  |  |  | 	expected.unmarshalBinary(buffer) | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	// We might need to truncate back to older files | 
					
						
							|  |  |  | 	if expected.filenum != t.headId { | 
					
						
							|  |  |  | 		// If already open for reading, force-reopen for writing | 
					
						
							|  |  |  | 		t.releaseFile(expected.filenum) | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | 		newHead, err := t.openFile(expected.filenum, openFreezerFileForAppend) | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		if err != nil { | 
					
						
							|  |  |  | 			return err | 
					
						
							|  |  |  | 		} | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | 		// Release any files _after the current head -- both the previous head | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		// and any files which may have been opened for reading | 
					
						
							|  |  |  | 		t.releaseFilesAfter(expected.filenum, true) | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | 		// Set back the historic head | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		t.head = newHead | 
					
						
							| 
									
										
										
										
											2021-09-07 12:31:17 +02:00
										 |  |  | 		t.headId = expected.filenum | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | 	if err := truncateFreezerFile(t.head, int64(expected.offset)); err != nil { | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 		return err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	// All data files truncated, set internal counters and return | 
					
						
							| 
									
										
										
										
											2021-09-07 12:31:17 +02:00
										 |  |  | 	t.headBytes = int64(expected.offset) | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	atomic.StoreUint64(&t.items, items) | 
					
						
							| 
									
										
										
										
											2019-06-10 14:21:02 +03:00
										 |  |  | 
 | 
					
						
							|  |  |  | 	// Retrieve the new size and update the total size counter | 
					
						
							|  |  |  | 	newSize, err := t.sizeNolock() | 
					
						
							|  |  |  | 	if err != nil { | 
					
						
							|  |  |  | 		return err | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-09-10 14:39:07 +03:00
										 |  |  | 	t.sizeGauge.Dec(int64(oldSize - newSize)) | 
					
						
							| 
									
										
										
										
											2019-06-10 14:21:02 +03:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	return nil | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | // Close closes all opened files. | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | func (t *freezerTable) Close() error { | 
					
						
							|  |  |  | 	t.lock.Lock() | 
					
						
							|  |  |  | 	defer t.lock.Unlock() | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | 	var errs []error | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	if err := t.index.Close(); err != nil { | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 		errs = append(errs, err) | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	t.index = nil | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	for _, f := range t.files { | 
					
						
							|  |  |  | 		if err := f.Close(); err != nil { | 
					
						
							|  |  |  | 			errs = append(errs, err) | 
					
						
							|  |  |  | 		} | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	t.head = nil | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 
 | 
					
						
							|  |  |  | 	if errs != nil { | 
					
						
							|  |  |  | 		return fmt.Errorf("%v", errs) | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	return nil | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | // openFile assumes that the write-lock is held by the caller | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | func (t *freezerTable) openFile(num uint32, opener func(string) (*os.File, error)) (f *os.File, err error) { | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	var exist bool | 
					
						
							|  |  |  | 	if f, exist = t.files[num]; !exist { | 
					
						
							|  |  |  | 		var name string | 
					
						
							|  |  |  | 		if t.noCompression { | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 			name = fmt.Sprintf("%s.%04d.rdat", t.name, num) | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		} else { | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 			name = fmt.Sprintf("%s.%04d.cdat", t.name, num) | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		} | 
					
						
							| 
									
										
										
										
											2019-06-10 11:45:12 +02:00
										 |  |  | 		f, err = opener(filepath.Join(t.path, name)) | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 		if err != nil { | 
					
						
							|  |  |  | 			return nil, err | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 		t.files[num] = f | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	return f, err | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | // releaseFile closes a file, and removes it from the open file cache. | 
					
						
							|  |  |  | // Assumes that the caller holds the write lock | 
					
						
							|  |  |  | func (t *freezerTable) releaseFile(num uint32) { | 
					
						
							|  |  |  | 	if f, exist := t.files[num]; exist { | 
					
						
							|  |  |  | 		delete(t.files, num) | 
					
						
							|  |  |  | 		f.Close() | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | // releaseFilesAfter closes all open files with a higher number, and optionally also deletes the files | 
					
						
							|  |  |  | func (t *freezerTable) releaseFilesAfter(num uint32, remove bool) { | 
					
						
							|  |  |  | 	for fnum, f := range t.files { | 
					
						
							|  |  |  | 		if fnum > num { | 
					
						
							|  |  |  | 			delete(t.files, fnum) | 
					
						
							|  |  |  | 			f.Close() | 
					
						
							|  |  |  | 			if remove { | 
					
						
							|  |  |  | 				os.Remove(f.Name()) | 
					
						
							|  |  |  | 			} | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | // getIndices returns the index entries for the given from-item, covering 'count' items. | 
					
						
							|  |  |  | // N.B: The actual number of returned indices for N items will always be N+1 (unless an | 
					
						
							|  |  |  | // error is returned). | 
					
						
							|  |  |  | // OBS: This method assumes that the caller has already verified (and/or trimmed) the range | 
					
						
							|  |  |  | // so that the items are within bounds. If this method is used to read out of bounds, | 
					
						
							|  |  |  | // it will return error. | 
					
						
							|  |  |  | func (t *freezerTable) getIndices(from, count uint64) ([]*indexEntry, error) { | 
					
						
							|  |  |  | 	// Apply the table-offset | 
					
						
							|  |  |  | 	from = from - uint64(t.itemOffset) | 
					
						
							|  |  |  | 	// For reading N items, we need N+1 indices. | 
					
						
							|  |  |  | 	buffer := make([]byte, (count+1)*indexEntrySize) | 
					
						
							|  |  |  | 	if _, err := t.index.ReadAt(buffer, int64(from*indexEntrySize)); err != nil { | 
					
						
							|  |  |  | 		return nil, err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	var ( | 
					
						
							|  |  |  | 		indices []*indexEntry | 
					
						
							|  |  |  | 		offset  int | 
					
						
							|  |  |  | 	) | 
					
						
							|  |  |  | 	for i := from; i <= from+count; i++ { | 
					
						
							|  |  |  | 		index := new(indexEntry) | 
					
						
							|  |  |  | 		index.unmarshalBinary(buffer[offset:]) | 
					
						
							|  |  |  | 		offset += indexEntrySize | 
					
						
							|  |  |  | 		indices = append(indices, index) | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	if from == 0 { | 
					
						
							| 
									
										
										
										
											2020-06-17 15:41:07 +08:00
										 |  |  | 		// Special case if we're reading the first item in the freezer. We assume that | 
					
						
							|  |  |  | 		// the first item always start from zero(regarding the deletion, we | 
					
						
							|  |  |  | 		// only support deletion by files, so that the assumption is held). | 
					
						
							|  |  |  | 		// This means we can use the first item metadata to carry information about | 
					
						
							|  |  |  | 		// the 'global' offset, for the deletion-case | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | 		indices[0].offset = 0 | 
					
						
							|  |  |  | 		indices[0].filenum = indices[1].filenum | 
					
						
							| 
									
										
										
										
											2020-06-17 15:41:07 +08:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | 	return indices, nil | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | // Retrieve looks up the data offset of an item with the given number and retrieves | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | // the raw binary blob from the data file. | 
					
						
							|  |  |  | func (t *freezerTable) Retrieve(item uint64) ([]byte, error) { | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | 	items, err := t.RetrieveItems(item, 1, 0) | 
					
						
							| 
									
										
										
											
												core/rawdb: fix datarace in freezer (#22728)
The Append / truncate operations were racy. When a datafile reaches 2Gb, a new file is needed. For this operation, we require a writelock, which is not needed in the 99.99% of all cases where the data does fit in the current head-file.
This transition from readlock to writelock was incorrect, and as the readlock was released, a truncate operation could slip in between, and truncate the data. This would have been fine, however, the Append operation continued writing as if no truncation had occurred, e.g writing item 5 where item 0 should reside.
This PR changes the behaviour, so that if when we run into the situation that a new file is needed, it aborts, and retries, this time with a writelock.
The outcome of the situation described above, running on this PR, would instead be that the Append operation exits with a failure.
											
										 
											2021-04-26 18:19:07 +02:00
										 |  |  | 	if err != nil { | 
					
						
							|  |  |  | 		return nil, err | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | 	return items[0], nil | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | // RetrieveItems returns multiple items in sequence, starting from the index 'start'. | 
					
						
							|  |  |  | // It will return at most 'max' items, but will abort earlier to respect the | 
					
						
							|  |  |  | // 'maxBytes' argument. However, if the 'maxBytes' is smaller than the size of one | 
					
						
							|  |  |  | // item, it _will_ return one element and possibly overflow the maxBytes. | 
					
						
							|  |  |  | func (t *freezerTable) RetrieveItems(start, count, maxBytes uint64) ([][]byte, error) { | 
					
						
							|  |  |  | 	// First we read the 'raw' data, which might be compressed. | 
					
						
							|  |  |  | 	diskData, sizes, err := t.retrieveItems(start, count, maxBytes) | 
					
						
							|  |  |  | 	if err != nil { | 
					
						
							|  |  |  | 		return nil, err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	var ( | 
					
						
							|  |  |  | 		output     = make([][]byte, 0, count) | 
					
						
							|  |  |  | 		offset     int // offset for reading | 
					
						
							|  |  |  | 		outputSize int // size of uncompressed data | 
					
						
							|  |  |  | 	) | 
					
						
							|  |  |  | 	// Now slice up the data and decompress. | 
					
						
							|  |  |  | 	for i, diskSize := range sizes { | 
					
						
							|  |  |  | 		item := diskData[offset : offset+diskSize] | 
					
						
							|  |  |  | 		offset += diskSize | 
					
						
							|  |  |  | 		decompressedSize := diskSize | 
					
						
							|  |  |  | 		if !t.noCompression { | 
					
						
							|  |  |  | 			decompressedSize, _ = snappy.DecodedLen(item) | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 		if i > 0 && uint64(outputSize+decompressedSize) > maxBytes { | 
					
						
							|  |  |  | 			break | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 		if !t.noCompression { | 
					
						
							|  |  |  | 			data, err := snappy.Decode(nil, item) | 
					
						
							|  |  |  | 			if err != nil { | 
					
						
							|  |  |  | 				return nil, err | 
					
						
							|  |  |  | 			} | 
					
						
							|  |  |  | 			output = append(output, data) | 
					
						
							|  |  |  | 		} else { | 
					
						
							|  |  |  | 			output = append(output, item) | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 		outputSize += decompressedSize | 
					
						
							| 
									
										
										
											
												core/rawdb: fix datarace in freezer (#22728)
The Append / truncate operations were racy. When a datafile reaches 2Gb, a new file is needed. For this operation, we require a writelock, which is not needed in the 99.99% of all cases where the data does fit in the current head-file.
This transition from readlock to writelock was incorrect, and as the readlock was released, a truncate operation could slip in between, and truncate the data. This would have been fine, however, the Append operation continued writing as if no truncation had occurred, e.g writing item 5 where item 0 should reside.
This PR changes the behaviour, so that if when we run into the situation that a new file is needed, it aborts, and retries, this time with a writelock.
The outcome of the situation described above, running on this PR, would instead be that the Append operation exits with a failure.
											
										 
											2021-04-26 18:19:07 +02:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | 	return output, nil | 
					
						
							| 
									
										
										
											
												core/rawdb: fix datarace in freezer (#22728)
The Append / truncate operations were racy. When a datafile reaches 2Gb, a new file is needed. For this operation, we require a writelock, which is not needed in the 99.99% of all cases where the data does fit in the current head-file.
This transition from readlock to writelock was incorrect, and as the readlock was released, a truncate operation could slip in between, and truncate the data. This would have been fine, however, the Append operation continued writing as if no truncation had occurred, e.g writing item 5 where item 0 should reside.
This PR changes the behaviour, so that if when we run into the situation that a new file is needed, it aborts, and retries, this time with a writelock.
The outcome of the situation described above, running on this PR, would instead be that the Append operation exits with a failure.
											
										 
											2021-04-26 18:19:07 +02:00
										 |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | // retrieveItems reads up to 'count' items from the table. It reads at least | 
					
						
							|  |  |  | // one item, but otherwise avoids reading more than maxBytes bytes. | 
					
						
							|  |  |  | // It returns the (potentially compressed) data, and the sizes. | 
					
						
							|  |  |  | func (t *freezerTable) retrieveItems(start, count, maxBytes uint64) ([]byte, []int, error) { | 
					
						
							| 
									
										
										
										
											2020-04-14 17:13:47 +02:00
										 |  |  | 	t.lock.RLock() | 
					
						
							| 
									
										
										
											
												core/rawdb: fix datarace in freezer (#22728)
The Append / truncate operations were racy. When a datafile reaches 2Gb, a new file is needed. For this operation, we require a writelock, which is not needed in the 99.99% of all cases where the data does fit in the current head-file.
This transition from readlock to writelock was incorrect, and as the readlock was released, a truncate operation could slip in between, and truncate the data. This would have been fine, however, the Append operation continued writing as if no truncation had occurred, e.g writing item 5 where item 0 should reside.
This PR changes the behaviour, so that if when we run into the situation that a new file is needed, it aborts, and retries, this time with a writelock.
The outcome of the situation described above, running on this PR, would instead be that the Append operation exits with a failure.
											
										 
											2021-04-26 18:19:07 +02:00
										 |  |  | 	defer t.lock.RUnlock() | 
					
						
							| 
									
										
										
										
											2021-09-07 12:31:17 +02:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	// Ensure the table and the item is accessible | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	if t.index == nil || t.head == nil { | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | 		return nil, nil, errClosed | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | 	itemCount := atomic.LoadUint64(&t.items) // max number | 
					
						
							|  |  |  | 	// Ensure the start is written, not deleted from the tail, and that the | 
					
						
							|  |  |  | 	// caller actually wants something | 
					
						
							|  |  |  | 	if itemCount <= start || uint64(t.itemOffset) > start || count == 0 { | 
					
						
							|  |  |  | 		return nil, nil, errOutOfBounds | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | 	if start+count > itemCount { | 
					
						
							|  |  |  | 		count = itemCount - start | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | 	var ( | 
					
						
							|  |  |  | 		output     = make([]byte, maxBytes) // Buffer to read data into | 
					
						
							|  |  |  | 		outputSize int                      // Used size of that buffer | 
					
						
							|  |  |  | 	) | 
					
						
							|  |  |  | 	// readData is a helper method to read a single data item from disk. | 
					
						
							|  |  |  | 	readData := func(fileId, start uint32, length int) error { | 
					
						
							|  |  |  | 		// In case a small limit is used, and the elements are large, may need to | 
					
						
							|  |  |  | 		// realloc the read-buffer when reading the first (and only) item. | 
					
						
							|  |  |  | 		if len(output) < length { | 
					
						
							|  |  |  | 			output = make([]byte, length) | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 		dataFile, exist := t.files[fileId] | 
					
						
							|  |  |  | 		if !exist { | 
					
						
							|  |  |  | 			return fmt.Errorf("missing data file %d", fileId) | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 		if _, err := dataFile.ReadAt(output[outputSize:outputSize+length], int64(start)); err != nil { | 
					
						
							|  |  |  | 			return err | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 		outputSize += length | 
					
						
							|  |  |  | 		return nil | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | 	// Read all the indexes in one go | 
					
						
							|  |  |  | 	indices, err := t.getIndices(start, count) | 
					
						
							|  |  |  | 	if err != nil { | 
					
						
							|  |  |  | 		return nil, nil, err | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | 	var ( | 
					
						
							|  |  |  | 		sizes      []int               // The sizes for each element | 
					
						
							|  |  |  | 		totalSize  = 0                 // The total size of all data read so far | 
					
						
							|  |  |  | 		readStart  = indices[0].offset // Where, in the file, to start reading | 
					
						
							|  |  |  | 		unreadSize = 0                 // The size of the as-yet-unread data | 
					
						
							|  |  |  | 	) | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | 	for i, firstIndex := range indices[:len(indices)-1] { | 
					
						
							|  |  |  | 		secondIndex := indices[i+1] | 
					
						
							|  |  |  | 		// Determine the size of the item. | 
					
						
							|  |  |  | 		offset1, offset2, _ := firstIndex.bounds(secondIndex) | 
					
						
							|  |  |  | 		size := int(offset2 - offset1) | 
					
						
							|  |  |  | 		// Crossing a file boundary? | 
					
						
							|  |  |  | 		if secondIndex.filenum != firstIndex.filenum { | 
					
						
							|  |  |  | 			// If we have unread data in the first file, we need to do that read now. | 
					
						
							|  |  |  | 			if unreadSize > 0 { | 
					
						
							|  |  |  | 				if err := readData(firstIndex.filenum, readStart, unreadSize); err != nil { | 
					
						
							|  |  |  | 					return nil, nil, err | 
					
						
							|  |  |  | 				} | 
					
						
							|  |  |  | 				unreadSize = 0 | 
					
						
							|  |  |  | 			} | 
					
						
							|  |  |  | 			readStart = 0 | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 		if i > 0 && uint64(totalSize+size) > maxBytes { | 
					
						
							|  |  |  | 			// About to break out due to byte limit being exceeded. We don't | 
					
						
							|  |  |  | 			// read this last item, but we need to do the deferred reads now. | 
					
						
							|  |  |  | 			if unreadSize > 0 { | 
					
						
							|  |  |  | 				if err := readData(secondIndex.filenum, readStart, unreadSize); err != nil { | 
					
						
							|  |  |  | 					return nil, nil, err | 
					
						
							|  |  |  | 				} | 
					
						
							|  |  |  | 			} | 
					
						
							|  |  |  | 			break | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 		// Defer the read for later | 
					
						
							|  |  |  | 		unreadSize += size | 
					
						
							|  |  |  | 		totalSize += size | 
					
						
							|  |  |  | 		sizes = append(sizes, size) | 
					
						
							|  |  |  | 		if i == len(indices)-2 || uint64(totalSize) > maxBytes { | 
					
						
							|  |  |  | 			// Last item, need to do the read now | 
					
						
							|  |  |  | 			if err := readData(secondIndex.filenum, readStart, unreadSize); err != nil { | 
					
						
							|  |  |  | 				return nil, nil, err | 
					
						
							|  |  |  | 			} | 
					
						
							|  |  |  | 			break | 
					
						
							|  |  |  | 		} | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 	} | 
					
						
							| 
									
										
										
										
											2021-08-13 10:51:01 +02:00
										 |  |  | 	return output[:outputSize], sizes, nil | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
											
												all: integrate the freezer with fast sync
* all: freezer style syncing
core, eth, les, light: clean up freezer relative APIs
core, eth, les, trie, ethdb, light: clean a bit
core, eth, les, light: add unit tests
core, light: rewrite setHead function
core, eth: fix downloader unit tests
core: add receipt chain insertion test
core: use constant instead of hardcoding table name
core: fix rollback
core: fix setHead
core/rawdb: remove canonical block first and then iterate side chain
core/rawdb, ethdb: add hasAncient interface
eth/downloader: calculate ancient limit via cht first
core, eth, ethdb: lots of fixes
* eth/downloader: print ancient disable log only for fast sync
											
										 
											2019-04-25 22:59:48 +08:00
										 |  |  | // has returns an indicator whether the specified number data | 
					
						
							|  |  |  | // exists in the freezer table. | 
					
						
							|  |  |  | func (t *freezerTable) has(number uint64) bool { | 
					
						
							|  |  |  | 	return atomic.LoadUint64(&t.items) > number | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
											
												core, cmd, vendor: fixes and database inspection tool (#15)
* core, eth: some fixes for freezer
* vendor, core/rawdb, cmd/geth: add db inspector
* core, cmd/utils: check ancient store path forceily
* cmd/geth, common, core/rawdb: a few fixes
* cmd/geth: support windows file rename and fix rename error
* core: support ancient plugin
* core, cmd: streaming file copy
* cmd, consensus, core, tests: keep genesis in leveldb
* core: write txlookup during ancient init
* core: bump database version
											
										 
											2019-05-14 22:07:44 +08:00
										 |  |  | // size returns the total data size in the freezer table. | 
					
						
							|  |  |  | func (t *freezerTable) size() (uint64, error) { | 
					
						
							|  |  |  | 	t.lock.RLock() | 
					
						
							|  |  |  | 	defer t.lock.RUnlock() | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-06-10 14:21:02 +03:00
										 |  |  | 	return t.sizeNolock() | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | // sizeNolock returns the total data size in the freezer table without obtaining | 
					
						
							|  |  |  | // the mutex first. | 
					
						
							|  |  |  | func (t *freezerTable) sizeNolock() (uint64, error) { | 
					
						
							| 
									
										
										
											
												core, cmd, vendor: fixes and database inspection tool (#15)
* core, eth: some fixes for freezer
* vendor, core/rawdb, cmd/geth: add db inspector
* core, cmd/utils: check ancient store path forceily
* cmd/geth, common, core/rawdb: a few fixes
* cmd/geth: support windows file rename and fix rename error
* core: support ancient plugin
* core, cmd: streaming file copy
* cmd, consensus, core, tests: keep genesis in leveldb
* core: write txlookup during ancient init
* core: bump database version
											
										 
											2019-05-14 22:07:44 +08:00
										 |  |  | 	stat, err := t.index.Stat() | 
					
						
							|  |  |  | 	if err != nil { | 
					
						
							|  |  |  | 		return 0, err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 	total := uint64(t.maxFileSize)*uint64(t.headId-t.tailId) + uint64(t.headBytes) + uint64(stat.Size()) | 
					
						
							|  |  |  | 	return total, nil | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2021-09-07 12:31:17 +02:00
										 |  |  | // advanceHead should be called when the current head file would outgrow the file limits, | 
					
						
							|  |  |  | // and a new file must be opened. The caller of this method must hold the write-lock | 
					
						
							|  |  |  | // before calling this method. | 
					
						
							|  |  |  | func (t *freezerTable) advanceHead() error { | 
					
						
							|  |  |  | 	t.lock.Lock() | 
					
						
							|  |  |  | 	defer t.lock.Unlock() | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | 	// We open the next file in truncated mode -- if this file already | 
					
						
							|  |  |  | 	// exists, we need to start over from scratch on it. | 
					
						
							|  |  |  | 	nextID := t.headId + 1 | 
					
						
							|  |  |  | 	newHead, err := t.openFile(nextID, openFreezerFileTruncated) | 
					
						
							|  |  |  | 	if err != nil { | 
					
						
							|  |  |  | 		return err | 
					
						
							|  |  |  | 	} | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | 	// Close old file, and reopen in RDONLY mode. | 
					
						
							|  |  |  | 	t.releaseFile(t.headId) | 
					
						
							|  |  |  | 	t.openFile(t.headId, openFreezerFileForReadOnly) | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | 	// Swap out the current head. | 
					
						
							|  |  |  | 	t.head = newHead | 
					
						
							|  |  |  | 	t.headBytes = 0 | 
					
						
							|  |  |  | 	t.headId = nextID | 
					
						
							|  |  |  | 	return nil | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | // Sync pushes any pending data from memory out to disk. This is an expensive | 
					
						
							|  |  |  | // operation, so use it with care. | 
					
						
							|  |  |  | func (t *freezerTable) Sync() error { | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	if err := t.index.Sync(); err != nil { | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | 		return err | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2019-03-26 12:28:23 +01:00
										 |  |  | 	return t.head.Sync() | 
					
						
							| 
									
										
										
										
											2019-03-08 15:56:20 +02:00
										 |  |  | } | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2021-04-13 15:45:30 +02:00
										 |  |  | // DumpIndex is a debug print utility function, mainly for testing. It can also | 
					
						
							|  |  |  | // be used to analyse a live freezer table index. | 
					
						
							|  |  |  | func (t *freezerTable) DumpIndex(start, stop int64) { | 
					
						
							| 
									
										
										
										
											2021-09-07 12:31:17 +02:00
										 |  |  | 	t.dumpIndex(os.Stdout, start, stop) | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | func (t *freezerTable) dumpIndexString(start, stop int64) string { | 
					
						
							|  |  |  | 	var out bytes.Buffer | 
					
						
							|  |  |  | 	out.WriteString("\n") | 
					
						
							|  |  |  | 	t.dumpIndex(&out, start, stop) | 
					
						
							|  |  |  | 	return out.String() | 
					
						
							|  |  |  | } | 
					
						
							|  |  |  | 
 | 
					
						
							|  |  |  | func (t *freezerTable) dumpIndex(w io.Writer, start, stop int64) { | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 	buf := make([]byte, indexEntrySize) | 
					
						
							|  |  |  | 
 | 
					
						
							| 
									
										
										
										
											2021-09-07 12:31:17 +02:00
										 |  |  | 	fmt.Fprintf(w, "| number | fileno | offset |\n") | 
					
						
							|  |  |  | 	fmt.Fprintf(w, "|--------|--------|--------|\n") | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 
 | 
					
						
							| 
									
										
										
										
											2021-04-13 15:45:30 +02:00
										 |  |  | 	for i := uint64(start); ; i++ { | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 		if _, err := t.index.ReadAt(buf, int64(i*indexEntrySize)); err != nil { | 
					
						
							|  |  |  | 			break | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 		var entry indexEntry | 
					
						
							|  |  |  | 		entry.unmarshalBinary(buf) | 
					
						
							| 
									
										
										
										
											2021-09-07 12:31:17 +02:00
										 |  |  | 		fmt.Fprintf(w, "|  %03d   |  %03d   |  %03d   | \n", i, entry.filenum, entry.offset) | 
					
						
							| 
									
										
										
										
											2021-04-13 15:45:30 +02:00
										 |  |  | 		if stop > 0 && i >= uint64(stop) { | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | 			break | 
					
						
							|  |  |  | 		} | 
					
						
							|  |  |  | 	} | 
					
						
							| 
									
										
										
										
											2021-09-07 12:31:17 +02:00
										 |  |  | 	fmt.Fprintf(w, "|--------------------------|\n") | 
					
						
							| 
									
										
										
										
											2019-04-14 21:25:32 +02:00
										 |  |  | } |