兄弟连Go语言+区块链技术培训以太坊源码分析49p2ptablego源码分析.docx

1、兄弟连Go语言+区块链技术培训以太坊源码分析49p2ptablego源码分析兄弟连Go语言+区块链技术培训以太坊源码分析(49）p2p-table.go源码分析table.go主要实现了p2p的Kademlia协议。# Kademlia协议简介(建议阅读references里面的pdf文档)Kademlia协议（以下简称Kad） 是美国纽约大学的PetarP. Maymounkov和David Mazieres.在2002年发布的一项研究结果Kademlia: A peerto -peer information system based onthe XOR metric。简单的说， Kad

2、是一种分布式哈希表（ DHT） 技术， 不过和其他 DHT 实现技术比较，如Chord、 CAN、 Pastry 等， Kad 通过独特的以异或算法（ XOR）为距离度量基础，建立了一种全新的 DHT 拓扑结构，相比于其他算法，大大提高了路由查询速度。# table的结构和字段const (alpha = 3 / Kademlia concurrency factorbucketSize = 16 / Kademlia bucket sizehashBits = len(common.Hash) * 8nBuckets = hashBits + 1 / Number of bucketsmax

3、BondingPingPongs = 16maxFindnodeFailures = 5autoRefreshInterval = 1 * time.HourseedCount = 30seedMaxAge = 5 * 24 * time.Hour)type Table struct mutex sync.Mutex / protects buckets, their content, and nurserybuckets nBuckets*bucket / index of known nodes by distancenursery *Node / bootstrap nodesdb *n

4、odeDB / database of known nodesrefreshReq chan chan structcloseReq chan structclosed chan structbondmu sync.Mutexbonding mapNodeID*bondprocbondslots chan struct / limits total number of active bonding processesnodeAddedHook func(*Node) / for testingnet transportself *Node / metadata of the local nod

5、e# 初始化func newTable(t transport, ourID NodeID, ourAddr *net.UDPAddr, nodeDBPath string) (*Table, error) / If no node database was given, use an in-memory one/这个在之前的database.go里面有介绍。 打开leveldb。如果path为空。那么打开一个基于内存的dbdb, err := newNodeDB(nodeDBPath, Version, ourID)if err != nil return nil, errtab := &T

6、ablenet: t,db: db,self: NewNode(ourID, ourAddr.IP, uint16(ourAddr.Port), uint16(ourAddr.Port),bonding: make(mapNodeID*bondproc),bondslots: make(chan struct, maxBondingPingPongs),refreshReq: make(chan chan struct),closeReq: make(chan struct),closed: make(chan struct),for i := 0; i cap(tab.bondslots);

7、 i+ tab.bondslots - structfor i := range tab.buckets tab.bucketsi = new(bucket)go tab.refreshLoop()return tab, nil上面的初始化启动了一个goroutine refreshLoop()，这个函数主要完成以下的工作。1.每一个小时进行一次刷新工作(autoRefreshInterval)2.如果接收到refreshReq请求。那么进行刷新工作。3.如果接收到关闭消息。那么进行关闭。所以函数主要的工作就是启动刷新工作。doRefresh/ refreshLoop schedules do

8、Refresh runs and coordinates shutdown.func (tab *Table) refreshLoop() var (timer = time.NewTicker(autoRefreshInterval)waiting chan struct / accumulates waiting callers while doRefresh runsdone chan struct / where doRefresh reports completion)loop:for select case -timer.C:if done = nil done = make(ch

9、an struct)go tab.doRefresh(done)case req := -tab.refreshReq:waiting = append(waiting, req)if done = nil done = make(chan struct)go tab.doRefresh(done)case -done:for _, ch := range waiting close(ch)waiting = nildone = nilcase -tab.closeReq:break loopif != nil .close()if done != nil -donefor _, ch :=

10、range waiting close(ch)tab.db.close()close(tab.closed)doRefresh函数/ doRefresh performs a lookup for a random target to keep buckets/ full. seed nodes are inserted if the table is empty (initial/ bootstrap or discarded faulty peers)./ doRefresh 随机查找一个目标，以便保持buckets是满的。如果table是空的，那么种子节点会插入。 （比如最开始的启动或者

11、是删除错误的节点之后）func (tab *Table) doRefresh(done chan struct) defer close(done)/ The Kademlia paper specifies that the bucket refresh should/ perform a lookup in the least recently used bucket. We cannot/ adhere to this because the findnode target is a 512bit value/ (not hash-sized) and it is not easily possible to generate a/ sha3 preimage that falls into a chosen bucket./ We perform a lookup with a random target instead./这里暂时没看懂var target NodeIDrand.Read(target:)result := tab.lookup(targ