以太坊原始碼分析(34)eth-downloader原始碼分析

尹成發表於2018-05-14
原始碼
downloader主要負責區塊鏈最開始的同步工作，當前的同步有兩種模式，一種是傳統的fullmode,這種模式通過下載區塊頭，和區塊體來構建區塊鏈，同步的過程就和普通的區塊插入的過程一樣，包括區塊頭的驗證，交易的驗證，交易執行，賬戶狀態的改變等操作，這其實是一個比較消耗CPU和磁碟的一個過程。 另一種模式就是 快速同步的fast sync模式， 這種模式有專門的文件來描述。請參考fast sync的文件。簡單的說 fast sync的模式會下載區塊頭，區塊體和收據， 插入的過程不會執行交易，然後在一個區塊高度(最高的區塊高度 - 1024)的時候同步所有的賬戶狀態，後面的1024個區塊會採用fullmode的方式來構建。 這種模式會加區塊的插入時間，同時不會產生大量的歷史的賬戶資訊。會相對節約磁碟， 但是對於網路的消耗會更高。 因為需要下載收據和狀態。

## downloader 資料結構

    
    type Downloader struct {
        mode SyncMode // Synchronisation mode defining the strategy used (per sync cycle)
        mux *event.TypeMux // Event multiplexer to announce sync operation events
        // queue 物件用來排程 區塊頭，交易，和收據的下載，以及下載完之後的組裝
        queue *queue // Scheduler for selecting the hashes to download
        // 對端的集合
        peers *peerSet // Set of active peers from which download can proceed
        stateDB ethdb.Database
        // fast sync 中的 Pivot point區塊的頭
        fsPivotLock *types.Header // Pivot header on critical section entry (cannot change between retries)
        fsPivotFails uint32 // Number of subsequent fast sync failures in the critical section
        // 下載的往返時延
        rttEstimate uint64 // Round trip time to target for download requests
        rttConfidence uint64 // Confidence in the estimated RTT (unit: millionths to allow atomic ops) 估計RTT的信心(單位：允許原子操作的百萬分之一)
    
        // Statistics 統計資訊，
        syncStatsChainOrigin uint64 // Origin block number where syncing started at
        syncStatsChainHeight uint64 // Highest block number known when syncing started
        syncStatsState stateSyncStats
        syncStatsLock sync.RWMutex // Lock protecting the sync stats fields
    
        lightchain LightChain
        blockchain BlockChain
    
        // Callbacks
        dropPeer peerDropFn // Drops a peer for misbehaving
    
        // Status
        synchroniseMock func(id string, hash common.Hash) error // Replacement for synchronise during testing
        synchronising int32
        notified int32
    
        // Channels
        headerCh chan dataPack // [eth/62] Channel receiving inbound block headers header的輸入通道，從網路下載的header會被送到這個通道
        bodyCh chan dataPack // [eth/62] Channel receiving inbound block bodies bodies的輸入通道，從網路下載的bodies會被送到這個通道
        receiptCh chan dataPack // [eth/63] Channel receiving inbound receipts receipts的輸入通道，從網路下載的receipts會被送到這個通道
        bodyWakeCh chan bool // [eth/62] Channel to signal the block body fetcher of new tasks 用來傳輸body fetcher新任務的通道
        receiptWakeCh chan bool // [eth/63] Channel to signal the receipt fetcher of new tasks    用來傳輸receipt fetcher 新任務的通道
        headerProcCh chan []*types.Header // [eth/62] Channel to feed the header processor new tasks       通道為header處理者提供新的任務
    
        // for stateFetcher
        stateSyncStart chan *stateSync //用來啟動新的 state fetcher
        trackStateReq chan *stateReq    // TODO
        stateCh chan dataPack // [eth/63] Channel receiving inbound node state data   state的輸入通道，從網路下載的state會被送到這個通道               
        // Cancellation and termination
        cancelPeer string // Identifier of the peer currently being used as the master (cancel on drop)
        cancelCh chan struct{} // Channel to cancel mid-flight syncs
        cancelLock sync.RWMutex // Lock to protect the cancel channel and peer in delivers
    
        quitCh chan struct{} // Quit channel to signal termination
        quitLock sync.RWMutex // Lock to prevent double closes
    
        // Testing hooks
        syncInitHook func(uint64, uint64) // Method to call upon initiating a new sync run
        bodyFetchHook func([]*types.Header) // Method to call upon starting a block body fetch
        receiptFetchHook func([]*types.Header) // Method to call upon starting a receipt fetch
        chainInsertHook func([]*fetchResult) // Method to call upon inserting a chain of blocks (possibly in multiple invocations)
    }


構造方法
    
    
    // New creates a new downloader to fetch hashes and blocks from remote peers.
    func New(mode SyncMode, stateDb ethdb.Database, mux *event.TypeMux, chain BlockChain, lightchain LightChain, dropPeer peerDropFn) *Downloader {
        if lightchain == nil {
            lightchain = chain
        }
        dl := &Downloader{
            mode: mode,
            stateDB: stateDb,
            mux: mux,
            queue: newQueue(),
            peers: newPeerSet(),
            rttEstimate: uint64(rttMaxEstimate),
            rttConfidence: uint64(1000000),
            blockchain: chain,
            lightchain: lightchain,
            dropPeer: dropPeer,
            headerCh: make(chan dataPack, 1),
            bodyCh: make(chan dataPack, 1),
            receiptCh: make(chan dataPack, 1),
            bodyWakeCh: make(chan bool, 1),
            receiptWakeCh: make(chan bool, 1),
            headerProcCh: make(chan []*types.Header, 1),
            quitCh: make(chan struct{}),
            stateCh: make(chan dataPack),
            stateSyncStart: make(chan *stateSync),
            trackStateReq: make(chan *stateReq),
        }
        go dl.qosTuner() //簡單 主要用來計算rttEstimate和rttConfidence
        go dl.stateFetcher() //啟動stateFetcher的任務監聽，但是這個時候還沒有生成state fetcher的任務。
        return dl
    }

## 同步下載
Synchronise試圖和一個peer來同步，如果同步過程中遇到一些錯誤，那麼會刪除掉Peer。然後會被重試。

    // Synchronise tries to sync up our local block chain with a remote peer, both
    // adding various sanity checks as well as wrapping it with various log entries.
    func (d *Downloader) Synchronise(id string, head common.Hash, td *big.Int, mode SyncMode) error {
        err := d.synchronise(id, head, td, mode)
        switch err {
        case nil:
        case errBusy:
    
        case errTimeout, errBadPeer, errStallingPeer,
            errEmptyHeaderSet, errPeersUnavailable, errTooOld,
            errInvalidAncestor, errInvalidChain:
            log.Warn("Synchronisation failed, dropping peer", "peer", id, "err", err)
            d.dropPeer(id)
    
        default:
            log.Warn("Synchronisation failed, retrying", "err", err)
        }
        return err
    }


synchronise
    
    // synchronise will select the peer and use it for synchronising. If an empty string is given
    // it will use the best peer possible and synchronize if it's TD is higher than our own. If any of the
    // checks fail an error will be returned. This method is synchronous
    func (d *Downloader) synchronise(id string, hash common.Hash, td *big.Int, mode SyncMode) error {
        // Mock out the synchronisation if testing
        if d.synchroniseMock != nil {
            return d.synchroniseMock(id, hash)
        }
        // Make sure only one goroutine is ever allowed past this point at once
        // 這個方法同時只能執行一個， 檢查是否正在執行。
        if !atomic.CompareAndSwapInt32(&d.synchronising, 0, 1) {
            return errBusy
        }
        defer atomic.StoreInt32(&d.synchronising, 0)
    
        // Post a user notification of the sync (only once per session)
        if atomic.CompareAndSwapInt32(&d.notified, 0, 1) {
            log.Info("Block synchronisation started")
        }
        // Reset the queue, peer set and wake channels to clean any internal leftover state
        // 重置queue和peer的狀態。
        d.queue.Reset()
        d.peers.Reset()
        // 清空d.bodyWakeCh, d.receiptWakeCh
        for _, ch := range []chan bool{d.bodyWakeCh, d.receiptWakeCh} {
            select {
            case <-ch:
            default:
            }
        }
        // 清空d.headerCh, d.bodyCh, d.receiptCh
        for _, ch := range []chan dataPack{d.headerCh, d.bodyCh, d.receiptCh} {
            for empty := false; !empty; {
                select {
                case <-ch:
                default:
                    empty = true
                }
            }
        }
        // 清空headerProcCh
        for empty := false; !empty; {
            select {
            case <-d.headerProcCh:
            default:
                empty = true
            }
        }
        // Create cancel channel for aborting mid-flight and mark the master peer
        d.cancelLock.Lock()
        d.cancelCh = make(chan struct{})
        d.cancelPeer = id
        d.cancelLock.Unlock()
    
        defer d.Cancel() // No matter what, we can't leave the cancel channel open
    
        // Set the requested sync mode, unless it's forbidden
        d.mode = mode
        if d.mode == FastSync && atomic.LoadUint32(&d.fsPivotFails) >= fsCriticalTrials {
            d.mode = FullSync
        }
        // Retrieve the origin peer and initiate the downloading process
        p := d.peers.Peer(id)
        if p == nil {
            return errUnknownPeer
        }
        return d.syncWithPeer(p, hash, td)
    }

syncWithPeer

    // syncWithPeer starts a block synchronization based on the hash chain from the
    // specified peer and head hash.
    func (d *Downloader) syncWithPeer(p *peerConnection, hash common.Hash, td *big.Int) (err error) {
        ...
        // Look up the sync boundaries: the common ancestor and the target block
        // 使用hash指來獲取區塊頭，這個方法裡面會訪問網路
        latest, err := d.fetchHeight(p)
        if err != nil {
            return err
        }
        height := latest.Number.Uint64()
        // findAncestor試圖來獲取大家共同的祖先，以便找到一個開始同步的點。
        origin, err := d.findAncestor(p, height)
        if err != nil {
            return err
        }
        d.syncStatsLock.Lock()
        if d.syncStatsChainHeight <= origin || d.syncStatsChainOrigin > origin {
            d.syncStatsChainOrigin = origin
        }
        d.syncStatsChainHeight = height
        d.syncStatsLock.Unlock()
    
        // Initiate the sync using a concurrent header and content retrieval algorithm
        pivot := uint64(0)
        switch d.mode {
        case LightSync:
            pivot = height
        case FastSync:
            // Calculate the new fast/slow sync pivot point
            // 如果pivot這個點沒有被鎖定。
            if d.fsPivotLock == nil {
                pivotOffset, err := rand.Int(rand.Reader, big.NewInt(int64(fsPivotInterval)))
                if err != nil {
                    panic(fmt.Sprintf("Failed to access crypto random source: %v", err))
                }
                if height > uint64(fsMinFullBlocks)+pivotOffset.Uint64() {
                    pivot = height - uint64(fsMinFullBlocks) - pivotOffset.Uint64()
                }
            } else { // 如過這個點已經被鎖定了。那麼就使用這個點
                // Pivot point locked in, use this and do not pick a new one!
                pivot = d.fsPivotLock.Number.Uint64()
            }
            // If the point is below the origin, move origin back to ensure state download
            if pivot < origin {
                if pivot > 0 {
                    origin = pivot - 1
                } else {
                    origin = 0
                }
            }
            log.Debug("Fast syncing until pivot block", "pivot", pivot)
        }
        d.queue.Prepare(origin+1, d.mode, pivot, latest)
        if d.syncInitHook != nil {
            d.syncInitHook(origin, height)
        }
        // 啟動幾個fetcher 分別負責header,bodies,receipts,處理headers
        fetchers := []func() error{
            func() error { return d.fetchHeaders(p, origin+1) }, // Headers are always retrieved
            func() error { return d.fetchBodies(origin + 1) }, // Bodies are retrieved during normal and fast sync
            func() error { return d.fetchReceipts(origin + 1) }, // Receipts are retrieved during fast sync
            func() error { return d.processHeaders(origin+1, td) },
        }
        if d.mode == FastSync { //根據模式的不同，增加新的處理邏輯
            fetchers = append(fetchers, func() error { return d.processFastSyncContent(latest) })
        } else if d.mode == FullSync {
            fetchers = append(fetchers, d.processFullSyncContent)
        }
        err = d.spawnSync(fetchers)
        if err != nil && d.mode == FastSync && d.fsPivotLock != nil {
            // If sync failed in the critical section, bump the fail counter.
            atomic.AddUint32(&d.fsPivotFails, 1)
        }
        return err
    }

spawnSync給每個fetcher啟動一個goroutine, 然後阻塞的等待fetcher出錯。

    // spawnSync runs d.process and all given fetcher functions to completion in
    // separate goroutines, returning the first error that appears.
    func (d *Downloader) spawnSync(fetchers []func() error) error {
        var wg sync.WaitGroup
        errc := make(chan error, len(fetchers))
        wg.Add(len(fetchers))
        for _, fn := range fetchers {
            fn := fn
            go func() { defer wg.Done(); errc <- fn() }()
        }
        // Wait for the first error, then terminate the others.
        var err error
        for i := 0; i < len(fetchers); i++ {
            if i == len(fetchers)-1 {
                // Close the queue when all fetchers have exited.
                // This will cause the block processor to end when
                // it has processed the queue.
                d.queue.Close()
            }
            if err = <-errc; err != nil {
                break
            }
        }
        d.queue.Close()
        d.Cancel()
        wg.Wait()
        return err
    }

## headers的處理

fetchHeaders方法用來獲取header。 然後根據獲取的header去獲取body和receipt等資訊。

    // fetchHeaders keeps retrieving headers concurrently from the number
    // requested, until no more are returned, potentially throttling on the way. To
    // facilitate concurrency but still protect against malicious nodes sending bad
    // headers, we construct a header chain skeleton using the "origin" peer we are
    // syncing with, and fill in the missing headers using anyone else. Headers from
    // other peers are only accepted if they map cleanly to the skeleton. If no one
    // can fill in the skeleton - not even the origin peer - it's assumed invalid and
    // the origin is dropped.
    fetchHeaders不斷的重複這樣的操作，傳送header請求，等待所有的返回。直到完成所有的header請求。 為了提高併發性，同時仍然能夠防止惡意節點傳送錯誤的header，我們使用我們正在同步的“origin”peer構造一個標頭檔案鏈骨架，並使用其他人填充缺失的header。 其他peer的header只有在乾淨地對映到骨架上時才被接受。 如果沒有人能夠填充骨架 - 甚至origin peer也不能填充 - 它被認為是無效的，並且origin peer也被丟棄。
    func (d *Downloader) fetchHeaders(p *peerConnection, from uint64) error {
        p.log.Debug("Directing header downloads", "origin", from)
        defer p.log.Debug("Header download terminated")
    
        // Create a timeout timer, and the associated header fetcher
        skeleton := true // Skeleton assembly phase or finishing up
        request := time.Now() // time of the last skeleton fetch request
        timeout := time.NewTimer(0) // timer to dump a non-responsive active peer
        <-timeout.C // timeout channel should be initially empty
        defer timeout.Stop()
    
        var ttl time.Duration
        getHeaders := func(from uint64) {
            request = time.Now()
    
            ttl = d.requestTTL()
            timeout.Reset(ttl)
    
            if skeleton { //填充骨架
                p.log.Trace("Fetching skeleton headers", "count", MaxHeaderFetch, "from", from)
                go p.peer.RequestHeadersByNumber(from+uint64(MaxHeaderFetch)-1, MaxSkeletonSize, MaxHeaderFetch-1, false)
            } else { // 直接請求
                p.log.Trace("Fetching full headers", "count", MaxHeaderFetch, "from", from)
                go p.peer.RequestHeadersByNumber(from, MaxHeaderFetch, 0, false)
            }
        }
        // Start pulling the header chain skeleton until all is done
        getHeaders(from)
    
        for {
            select {
            case <-d.cancelCh:
                return errCancelHeaderFetch
    
            case packet := <-d.headerCh: //網路上返回的header會投遞到headerCh這個通道
                // Make sure the active peer is giving us the skeleton headers
                if packet.PeerId() != p.id {
                    log.Debug("Received skeleton from incorrect peer", "peer", packet.PeerId())
                    break
                }
                headerReqTimer.UpdateSince(request)
                timeout.Stop()
    
                // If the skeleton's finished, pull any remaining head headers directly from the origin
                if packet.Items() == 0 && skeleton {
                    skeleton = false
                    getHeaders(from)
                    continue
                }
                // If no more headers are inbound, notify the content fetchers and return
                // 如果沒有更多的返回了。 那麼告訴headerProcCh通道
                if packet.Items() == 0 {
                    p.log.Debug("No more headers available")
                    select {
                    case d.headerProcCh <- nil:
                        return nil
                    case <-d.cancelCh:
                        return errCancelHeaderFetch
                    }
                }
                headers := packet.(*headerPack).headers
    
                // If we received a skeleton batch, resolve internals concurrently
                if skeleton { // 如果是需要填充骨架，那麼在這個方法裡面填充好
                    filled, proced, err := d.fillHeaderSkeleton(from, headers)
                    if err != nil {
                        p.log.Debug("Skeleton chain invalid", "err", err)
                        return errInvalidChain
                    }
                    headers = filled[proced:]
                    // proced代表已經處理完了多少個了。 所以只需要proced:後面的headers了
                    from += uint64(proced)
                }
                // Insert all the new headers and fetch the next batch
                if len(headers) > 0 {
                    p.log.Trace("Scheduling new headers", "count", len(headers), "from", from)
                    //投遞到headerProcCh 然後繼續迴圈。
                    select {
                    case d.headerProcCh <- headers:
                    case <-d.cancelCh:
                        return errCancelHeaderFetch
                    }
                    from += uint64(len(headers))
                }
                getHeaders(from)
    
            case <-timeout.C:
                // Header retrieval timed out, consider the peer bad and drop
                p.log.Debug("Header request timed out", "elapsed", ttl)
                headerTimeoutMeter.Mark(1)
                d.dropPeer(p.id)
    
                // Finish the sync gracefully instead of dumping the gathered data though
                for _, ch := range []chan bool{d.bodyWakeCh, d.receiptWakeCh} {
                    select {
                    case ch <- false:
                    case <-d.cancelCh:
                    }
                }
                select {
                case d.headerProcCh <- nil:
                case <-d.cancelCh:
                }
                return errBadPeer
            }
        }
    }

processHeaders方法，這個方法從headerProcCh通道來獲取header。並把獲取到的header丟入到queue來進行排程，這樣body fetcher或者是receipt fetcher就可以領取到fetch任務。
    
    // processHeaders takes batches of retrieved headers from an input channel and
    // keeps processing and scheduling them into the header chain and downloader's
    // queue until the stream ends or a failure occurs.
    // processHeaders批量的獲取headers， 處理他們，並通過downloader的queue物件來排程他們。 直到錯誤發生或者處理結束。
    func (d *Downloader) processHeaders(origin uint64, td *big.Int) error {
        // Calculate the pivoting point for switching from fast to slow sync
        pivot := d.queue.FastSyncPivot()
    
        // Keep a count of uncertain headers to roll back
        // rollback 用來處理這種邏輯，如果某個點失敗了。那麼之前插入的2048個節點都要回滾。因為安全性達不到要求， 可以詳細參考fast sync的文件。
        rollback := []*types.Header{}
        defer func() { // 這個函式用來錯誤退出的時候進行回滾。 TODO
            if len(rollback) > 0 {
                // Flatten the headers and roll them back
                hashes := make([]common.Hash, len(rollback))
                for i, header := range rollback {
                    hashes[i] = header.Hash()
                }
                lastHeader, lastFastBlock, lastBlock := d.lightchain.CurrentHeader().Number, common.Big0, common.Big0
                if d.mode != LightSync {
                    lastFastBlock = d.blockchain.CurrentFastBlock().Number()
                    lastBlock = d.blockchain.CurrentBlock().Number()
                }
                d.lightchain.Rollback(hashes)
                curFastBlock, curBlock := common.Big0, common.Big0
                if d.mode != LightSync {
                    curFastBlock = d.blockchain.CurrentFastBlock().Number()
                    curBlock = d.blockchain.CurrentBlock().Number()
                }
                log.Warn("Rolled back headers", "count", len(hashes),
                    "header", fmt.Sprintf("%d->%d", lastHeader, d.lightchain.CurrentHeader().Number),
                    "fast", fmt.Sprintf("%d->%d", lastFastBlock, curFastBlock),
                    "block", fmt.Sprintf("%d->%d", lastBlock, curBlock))
    
                // If we're already past the pivot point, this could be an attack, thread carefully
                if rollback[len(rollback)-1].Number.Uint64() > pivot {
                    // If we didn't ever fail, lock in the pivot header (must! not! change!)
                    if atomic.LoadUint32(&d.fsPivotFails) == 0 {
                        for _, header := range rollback {
                            if header.Number.Uint64() == pivot {
                                log.Warn("Fast-sync pivot locked in", "number", pivot, "hash", header.Hash())
                                d.fsPivotLock = header
                            }
                        }
                    }
                }
            }
        }()
    
        // Wait for batches of headers to process
        gotHeaders := false
    
        for {
            select {
            case <-d.cancelCh:
                return errCancelHeaderProcessing
    
            case headers := <-d.headerProcCh:
                // Terminate header processing if we synced up
                if len(headers) == 0 { //處理完成
                    // Notify everyone that headers are fully processed
                    for _, ch := range []chan bool{d.bodyWakeCh, d.receiptWakeCh} {
                        select {
                        case ch <- false:
                        case <-d.cancelCh:
                        }
                    }
                    // If no headers were retrieved at all, the peer violated it's TD promise that it had a
                    // better chain compared to ours. The only exception is if it's promised blocks were
                    // already imported by other means (e.g. fecher):
                    //
                    // R <remote peer>, L <local node>: Both at block 10
                    // R: Mine block 11, and propagate it to L
                    // L: Queue block 11 for import
                    // L: Notice that R's head and TD increased compared to ours, start sync
                    // L: Import of block 11 finishes
                    // L: Sync begins, and finds common ancestor at 11
                    // L: Request new headers up from 11 (R's TD was higher, it must have something)
                    // R: Nothing to give
                    if d.mode != LightSync { // 對方的TD比我們大，但是沒有獲取到任何東西。 那麼認為對方是錯誤的對方。 會斷開和對方的聯絡
                        if !gotHeaders && td.Cmp(d.blockchain.GetTdByHash(d.blockchain.CurrentBlock().Hash())) > 0 {
                            return errStallingPeer
                        }
                    }
                    // If fast or light syncing, ensure promised headers are indeed delivered. This is
                    // needed to detect scenarios where an attacker feeds a bad pivot and then bails out
                    // of delivering the post-pivot blocks that would flag the invalid content.
                    //
                    // This check cannot be executed "as is" for full imports, since blocks may still be
                    // queued for processing when the header download completes. However, as long as the
                    // peer gave us something useful, we're already happy/progressed (above check).
                    if d.mode == FastSync || d.mode == LightSync {
                        if td.Cmp(d.lightchain.GetTdByHash(d.lightchain.CurrentHeader().Hash())) > 0 {
                            return errStallingPeer
                        }
                    }
                    // Disable any rollback and return
                    rollback = nil
                    return nil
                }
                // Otherwise split the chunk of headers into batches and process them
                gotHeaders = true
    
                for len(headers) > 0 {
                    // Terminate if something failed in between processing chunks
                    select {
                    case <-d.cancelCh:
                        return errCancelHeaderProcessing
                    default:
                    }
                    // Select the next chunk of headers to import
                    limit := maxHeadersProcess
                    if limit > len(headers) {
                        limit = len(headers)
                    }
                    chunk := headers[:limit]
    
                    // In case of header only syncing, validate the chunk immediately
                    if d.mode == FastSync || d.mode == LightSync { //如果是快速同步模式，或者是輕量級同步模式(只下載區塊頭)
                        // Collect the yet unknown headers to mark them as uncertain
                        unknown := make([]*types.Header, 0, len(headers))
                        for _, header := range chunk {
                            if !d.lightchain.HasHeader(header.Hash(), header.Number.Uint64()) {
                                unknown = append(unknown, header)
                            }
                        }
                        // If we're importing pure headers, verify based on their recentness
                        // 每隔多少個區塊驗證一次
                        frequency := fsHeaderCheckFrequency
                        if chunk[len(chunk)-1].Number.Uint64()+uint64(fsHeaderForceVerify) > pivot {
                            frequency = 1
                        }
                        // lightchain預設是等於chain的。 插入區塊頭。如果失敗那麼需要回滾。
                        if n, err := d.lightchain.InsertHeaderChain(chunk, frequency); err != nil {
                            // If some headers were inserted, add them too to the rollback list
                            if n > 0 {
                                rollback = append(rollback, chunk[:n]...)
                            }
                            log.Debug("Invalid header encountered", "number", chunk[n].Number, "hash", chunk[n].Hash(), "err", err)
                            return errInvalidChain
                        }
                        // All verifications passed, store newly found uncertain headers
                        rollback = append(rollback, unknown...)
                        if len(rollback) > fsHeaderSafetyNet {
                            rollback = append(rollback[:0], rollback[len(rollback)-fsHeaderSafetyNet:]...)
                        }
                    }
                    // If we're fast syncing and just pulled in the pivot, make sure it's the one locked in
                    if d.mode == FastSync && d.fsPivotLock != nil && chunk[0].Number.Uint64() <= pivot && chunk[len(chunk)-1].Number.Uint64() >= pivot { //如果PivotLock,檢查一下Hash是否相同。
                        if pivot := chunk[int(pivot-chunk[0].Number.Uint64())]; pivot.Hash() != d.fsPivotLock.Hash() {
                            log.Warn("Pivot doesn't match locked in one", "remoteNumber", pivot.Number, "remoteHash", pivot.Hash(), "localNumber", d.fsPivotLock.Number, "localHash", d.fsPivotLock.Hash())
                            return errInvalidChain
                        }
                    }
                    // Unless we're doing light chains, schedule the headers for associated content retrieval
                    // 如果我們處理完輕量級鏈。 排程header來進行相關資料的獲取。body，receipts
                    if d.mode == FullSync || d.mode == FastSync {
                        // If we've reached the allowed number of pending headers, stall a bit
                        // 如果當前queue的容量容納不下了。那麼等待。
                        for d.queue.PendingBlocks() >= maxQueuedHeaders || d.queue.PendingReceipts() >= maxQueuedHeaders {
                            select {
                            case <-d.cancelCh:
                                return errCancelHeaderProcessing
                            case <-time.After(time.Second):
                            }
                        }
                        // Otherwise insert the headers for content retrieval
                        // 呼叫Queue進行排程，下載body和receipts
                        inserts := d.queue.Schedule(chunk, origin)
                        if len(inserts) != len(chunk) {
                            log.Debug("Stale headers")
                            return errBadPeer
                        }
                    }
                    headers = headers[limit:]
                    origin += uint64(limit)
                }
                // Signal the content downloaders of the availablility of new tasks
                // 給通道d.bodyWakeCh, d.receiptWakeCh傳送訊息，喚醒處理執行緒。
                for _, ch := range []chan bool{d.bodyWakeCh, d.receiptWakeCh} {
                    select {
                    case ch <- true:
                    default:
                    }
                }
            }
        }
    }


## bodies處理
fetchBodies函式定義了一些閉包函式，然後呼叫了fetchParts函式
    
    // fetchBodies iteratively downloads the scheduled block bodies, taking any
    // available peers, reserving a chunk of blocks for each, waiting for delivery
    // and also periodically checking for timeouts.
    // fetchBodies 持續的下載區塊體，中間會使用到任何可以用的連結，為每一個連結保留一部分的區塊體，等待區塊被交付，並定期的檢查是否超時。
    func (d *Downloader) fetchBodies(from uint64) error {
        log.Debug("Downloading block bodies", "origin", from)
    
        var (
            deliver = func(packet dataPack) (int, error) { //下載完的區塊體的交付函式
                pack := packet.(*bodyPack)
                return d.queue.DeliverBodies(pack.peerId, pack.transactions, pack.uncles)
            }
            expire = func() map[string]int { return d.queue.ExpireBodies(d.requestTTL()) } //超時
            fetch = func(p *peerConnection, req *fetchRequest) error { return p.FetchBodies(req) } // fetch函式
            capacity = func(p *peerConnection) int { return p.BlockCapacity(d.requestRTT()) } // 對端的吞吐量
            setIdle = func(p *peerConnection, accepted int) { p.SetBodiesIdle(accepted) } // 設定peer為idle
        )
        err := d.fetchParts(errCancelBodyFetch, d.bodyCh, deliver, d.bodyWakeCh, expire,
            d.queue.PendingBlocks, d.queue.InFlightBlocks, d.queue.ShouldThrottleBlocks, d.queue.ReserveBodies,
            d.bodyFetchHook, fetch, d.queue.CancelBodies, capacity, d.peers.BodyIdlePeers, setIdle, "bodies")
    
        log.Debug("Block body download terminated", "err", err)
        return err
    }


fetchParts


    // fetchParts iteratively downloads scheduled block parts, taking any available
    // peers, reserving a chunk of fetch requests for each, waiting for delivery and
    // also periodically checking for timeouts.
    // fetchParts迭代地下載預定的塊部分，取得任何可用的對等體，為每個部分預留大量的提取請求，等待交付並且還定期檢查超時。
    // As the scheduling/timeout logic mostly is the same for all downloaded data
    // types, this method is used by each for data gathering and is instrumented with
    // various callbacks to handle the slight differences between processing them.
    // 由於排程/超時邏輯對於所有下載的資料型別大部分是相同的，所以這個方法被用於不同的區塊型別的下載，並且用各種回撥函式來處理它們之間的細微差別。
    // The instrumentation parameters:
    // - errCancel: error type to return if the fetch operation is cancelled (mostly makes logging nicer) 如果fetch操作被取消，會在這個通道上傳送資料
    // - deliveryCh: channel from which to retrieve downloaded data packets (merged from all concurrent peers) 資料被下載完成後投遞的目的地
    // - deliver: processing callback to deliver data packets into type specific download queues (usually within `queue`) 處理完成後資料被投遞到哪個佇列
    // - wakeCh: notification channel for waking the fetcher when new tasks are available (or sync completed) 用來通知fetcher 新的任務到來，或者是同步完成
    // - expire: task callback method to abort requests that took too long and return the faulty peers (traffic shaping) 因為超時來終止請求的回撥函式。
    // - pending: task callback for the number of requests still needing download (detect completion/non-completability) 還需要下載的任務的數量。
    // - inFlight: task callback for the number of in-progress requests (wait for all active downloads to finish) 正在處理過程中的請求數量
    // - throttle: task callback to check if the processing queue is full and activate throttling (bound memory use) 用來檢查處理佇列是否滿的回撥函式。
    // - reserve: task callback to reserve new download tasks to a particular peer (also signals partial completions) 用來為某個peer來預定任務的回撥函式
    // - fetchHook: tester callback to notify of new tasks being initiated (allows testing the scheduling logic)
    // - fetch: network callback to actually send a particular download request to a physical remote peer //傳送網路請求的回撥函式
    // - cancel: task callback to abort an in-flight download request and allow rescheduling it (in case of lost peer) 用來取消正在處理的任務的回撥函式
    // - capacity: network callback to retrieve the estimated type-specific bandwidth capacity of a peer (traffic shaping) 網路容量或者是頻寬。
    // - idle: network callback to retrieve the currently (type specific) idle peers that can be assigned tasks peer是否空閒的回撥函式
    // - setIdle: network callback to set a peer back to idle and update its estimated capacity (traffic shaping) 設定peer為空閒的回撥函式
    // - kind: textual label of the type being downloaded to display in log mesages 下載型別，用於日誌
    func (d *Downloader) fetchParts(errCancel error, deliveryCh chan dataPack, deliver func(dataPack) (int, error), wakeCh chan bool,
        expire func() map[string]int, pending func() int, inFlight func() bool, throttle func() bool, reserve func(*peerConnection, int) (*fetchRequest, bool, error),
        fetchHook func([]*types.Header), fetch func(*peerConnection, *fetchRequest) error, cancel func(*fetchRequest), capacity func(*peerConnection) int,
        idle func() ([]*peerConnection, int), setIdle func(*peerConnection, int), kind string) error {
    
        // Create a ticker to detect expired retrieval tasks
        ticker := time.NewTicker(100 * time.Millisecond)
        defer ticker.Stop()
    
        update := make(chan struct{}, 1)
    
        // Prepare the queue and fetch block parts until the block header fetcher's done
        finished := false
        for {
            select {
            case <-d.cancelCh:
                return errCancel
    
            case packet := <-deliveryCh:
                // If the peer was previously banned and failed to deliver it's pack
                // in a reasonable time frame, ignore it's message.
                // 如果peer在之前被禁止而且沒有在合適的時間deliver它的資料，那麼忽略這個資料
                if peer := d.peers.Peer(packet.PeerId()); peer != nil {
                    // Deliver the received chunk of data and check chain validity
                    accepted, err := deliver(packet)
                    if err == errInvalidChain {
                        return err
                    }
                    // Unless a peer delivered something completely else than requested (usually
                    // caused by a timed out request which came through in the end), set it to
                    // idle. If the delivery's stale, the peer should have already been idled.
                    if err != errStaleDelivery {
                        setIdle(peer, accepted)
                    }
                    // Issue a log to the user to see what's going on
                    switch {
                    case err == nil && packet.Items() == 0:
                        peer.log.Trace("Requested data not delivered", "type", kind)
                    case err == nil:
                        peer.log.Trace("Delivered new batch of data", "type", kind, "count", packet.Stats())
                    default:
                        peer.log.Trace("Failed to deliver retrieved data", "type", kind, "err", err)
                    }
                }
                // Blocks assembled, try to update the progress
                select {
                case update <- struct{}{}:
                default:
                }
    
            case cont := <-wakeCh:
                // The header fetcher sent a continuation flag, check if it's done
                // 當所有的任務完成的時候會寫入這個佇列。
                if !cont {
                    finished = true
                }
                // Headers arrive, try to update the progress
                select {
                case update <- struct{}{}:
                default:
                }
    
            case <-ticker.C:
                // Sanity check update the progress
                select {
                case update <- struct{}{}:
                default:
                }
    
            case <-update:
                // Short circuit if we lost all our peers
                if d.peers.Len() == 0 {
                    return errNoPeers
                }
                // Check for fetch request timeouts and demote the responsible peers
                for pid, fails := range expire() {
                    if peer := d.peers.Peer(pid); peer != nil {
                        // If a lot of retrieval elements expired, we might have overestimated the remote peer or perhaps
                        // ourselves. Only reset to minimal throughput but don't drop just yet. If even the minimal times
                        // out that sync wise we need to get rid of the peer.
                        //如果很多檢索元素過期，我們可能高估了遠端物件或者我們自己。 只能重置為最小的吞吐量，但不要丟棄。 如果即使最小的同步任然超時，我們需要刪除peer。
                        // The reason the minimum threshold is 2 is because the downloader tries to estimate the bandwidth
                        // and latency of a peer separately, which requires pushing the measures capacity a bit and seeing
                        // how response times reacts, to it always requests one more than the minimum (i.e. min 2).
                        // 最小閾值為2的原因是因為下載器試圖分別估計對等體的頻寬和等待時間，這需要稍微推動測量容量並且看到響應時間如何反應，總是要求比最小值（即，最小值2）。
                        if fails > 2 {
                            peer.log.Trace("Data delivery timed out", "type", kind)
                            setIdle(peer, 0)
                        } else {
                            peer.log.Debug("Stalling delivery, dropping", "type", kind)
                            d.dropPeer(pid)
                        }
                    }
                }
                // If there's nothing more to fetch, wait or terminate
                // 任務全部完成。 那麼退出
                if pending() == 0 { //如果沒有等待分配的任務， 那麼break。不用執行下面的程式碼了。
                    if !inFlight() && finished {
                        log.Debug("Data fetching completed", "type", kind)
                        return nil
                    }
                    break
                }
                // Send a download request to all idle peers, until throttled
                progressed, throttled, running := false, false, inFlight()
                idles, total := idle()
    
                for _, peer := range idles {
                    // Short circuit if throttling activated
                    if throttle() {
                        throttled = true
                        break
                    }
                    // Short circuit if there is no more available task.
                    if pending() == 0 {
                        break
                    }
                    // Reserve a chunk of fetches for a peer. A nil can mean either that
                    // no more headers are available, or that the peer is known not to
                    // have them.
                    // 為某個peer請求分配任務。
                    request, progress, err := reserve(peer, capacity(peer))
                    if err != nil {
                        return err
                    }
                    if progress {
                        progressed = true
                    }
                    if request == nil {
                        continue
                    }
                    if request.From > 0 {
                        peer.log.Trace("Requesting new batch of data", "type", kind, "from", request.From)
                    } else if len(request.Headers) > 0 {
                        peer.log.Trace("Requesting new batch of data", "type", kind, "count", len(request.Headers), "from", request.Headers[0].Number)
                    } else {
                        peer.log.Trace("Requesting new batch of data", "type", kind, "count", len(request.Hashes))
                    }
                    // Fetch the chunk and make sure any errors return the hashes to the queue
                    if fetchHook != nil {
                        fetchHook(request.Headers)
                    }
                    if err := fetch(peer, request); err != nil {
                        // Although we could try and make an attempt to fix this, this error really
                        // means that we've double allocated a fetch task to a peer. If that is the
                        // case, the internal state of the downloader and the queue is very wrong so
                        // better hard crash and note the error instead of silently accumulating into
                        // a much bigger issue.
                        panic(fmt.Sprintf("%v: %s fetch assignment failed", peer, kind))
                    }
                    running = true
                }
                // Make sure that we have peers available for fetching. If all peers have been tried
                // and all failed throw an error
                if !progressed && !throttled && !running && len(idles) == total && pending() > 0 {
                    return errPeersUnavailable
                }
            }
        }
    }


## receipt的處理
receipt的處理和body類似。
    
    // fetchReceipts iteratively downloads the scheduled block receipts, taking any
    // available peers, reserving a chunk of receipts for each, waiting for delivery
    // and also periodically checking for timeouts.
    func (d *Downloader) fetchReceipts(from uint64) error {
        log.Debug("Downloading transaction receipts", "origin", from)
    
        var (
            deliver = func(packet dataPack) (int, error) {
                pack := packet.(*receiptPack)
                return d.queue.DeliverReceipts(pack.peerId, pack.receipts)
            }
            expire = func() map[string]int { return d.queue.ExpireReceipts(d.requestTTL()) }
            fetch = func(p *peerConnection, req *fetchRequest) error { return p.FetchReceipts(req) }
            capacity = func(p *peerConnection) int { return p.ReceiptCapacity(d.requestRTT()) }
            setIdle = func(p *peerConnection, accepted int) { p.SetReceiptsIdle(accepted) }
        )
        err := d.fetchParts(errCancelReceiptFetch, d.receiptCh, deliver, d.receiptWakeCh, expire,
            d.queue.PendingReceipts, d.queue.InFlightReceipts, d.queue.ShouldThrottleReceipts, d.queue.ReserveReceipts,
            d.receiptFetchHook, fetch, d.queue.CancelReceipts, capacity, d.peers.ReceiptIdlePeers, setIdle, "receipts")
    
        log.Debug("Transaction receipt download terminated", "err", err)
        return err
    }


## processFastSyncContent 和 processFullSyncContent



    // processFastSyncContent takes fetch results from the queue and writes them to the
    // database. It also controls the synchronisation of state nodes of the pivot block.
    func (d *Downloader) processFastSyncContent(latest *types.Header) error {
        // Start syncing state of the reported head block.
        // This should get us most of the state of the pivot block.
        // 啟動狀態同步
        stateSync := d.syncState(latest.Root)
        defer stateSync.Cancel()
        go func() {
            if err := stateSync.Wait(); err != nil {
                d.queue.Close() // wake up WaitResults
            }
        }()
    
        pivot := d.queue.FastSyncPivot()
        for {
            results := d.queue.WaitResults() // 等待佇列輸出處理完成的區塊
            if len(results) == 0 {
                return stateSync.Cancel()
            }
            if d.chainInsertHook != nil {
                d.chainInsertHook(results)
            }
            P, beforeP, afterP := splitAroundPivot(pivot, results)
            // 插入fast sync的資料
            if err := d.commitFastSyncData(beforeP, stateSync); err != nil {
                return err
            }
            if P != nil {
                // 如果已經達到了 pivot point 那麼等待狀態同步完成，
                stateSync.Cancel()
                if err := d.commitPivotBlock(P); err != nil {
                    return err
                }
            }
            // 對於pivot point 之後的所有節點，都需要按照完全的處理。
            if err := d.importBlockResults(afterP); err != nil {
                return err
            }
        }
    }

processFullSyncContent,比較簡單。 從佇列裡面獲取區塊然後插入。
    
    // processFullSyncContent takes fetch results from the queue and imports them into the chain.
    func (d *Downloader) processFullSyncContent() error {
        for {
            results := d.queue.WaitResults()
            if len(results) == 0 {
                return nil
            }
            if d.chainInsertHook != nil {
                d.chainInsertHook(results)
            }
            if err := d.importBlockResults(results); err != nil {
                return err
            }
        }
    }




網址：http://www.qukuailianxueyuan.io/



欲領取造幣技術與全套虛擬機器資料

區塊鏈技術交流QQ群：756146052  備註：CSDN

尹成學院微信：備註：CSDN


相關文章