cmd/utils, rpc/comms: stop XEth when IPC connection ends

There are a bunch of changes required to make this work:

- in miner: allow unregistering agents, fix RemoteAgent.Stop
- in eth/filters: make FilterSystem.Stop not crash
- in rpc/comms: move listen loop to platform-independent code

Fixes #1930. I ran the shell loop there for a few minutes and didn't see
any changes in the memory profile.

cmd/utils/flags.go

@@ -627,17 +627,14 @@ func StartIPC(eth *eth.Ethereum, ctx *cli.Context) error {
 		Endpoint: IpcSocketPath(ctx),
 	}
 
-	initializer := func(conn net.Conn) (shared.EthereumApi, error) {
+	initializer := func(conn net.Conn) (comms.Stopper, shared.EthereumApi, error) {
 		fe := useragent.NewRemoteFrontend(conn, eth.AccountManager())
 		xeth := xeth.New(eth, fe)
-		codec := codec.JSON
+		apis, err := api.ParseApiString(ctx.GlobalString(IPCApiFlag.Name), codec.JSON, xeth, eth)
-
-		apis, err := api.ParseApiString(ctx.GlobalString(IPCApiFlag.Name), codec, xeth, eth)
 		if err != nil {
-			return nil, err
+			return nil, nil, err
 		}
-
+		return xeth, api.Merge(apis...), nil
-		return api.Merge(apis...), nil
 	}
 
 	return comms.StartIpc(config, codec.JSON, initializer)

eth/filters/filter_system.go

@@ -31,30 +31,32 @@ import (
 // block, transaction and log events. The Filtering system can be used to listen
 // for specific LOG events fired by the EVM (Ethereum Virtual Machine).
 type FilterSystem struct {
-	eventMux *event.TypeMux
-
 	filterMu sync.RWMutex
 	filterId int
 	filters  map[int]*Filter
 	created  map[int]time.Time
-
+	sub      event.Subscription
-	quit chan struct{}
 }
 
 // NewFilterSystem returns a newly allocated filter manager
 func NewFilterSystem(mux *event.TypeMux) *FilterSystem {
 	fs := &FilterSystem{
-		eventMux: mux,
 		filters: make(map[int]*Filter),
 		created: make(map[int]time.Time),
 	}
+	fs.sub = mux.Subscribe(
+		//core.PendingBlockEvent{},
+		core.ChainEvent{},
+		core.TxPreEvent{},
+		vm.Logs(nil),
+	)
 	go fs.filterLoop()
 	return fs
 }
 
 // Stop quits the filter loop required for polling events
 func (fs *FilterSystem) Stop() {
-	close(fs.quit)
+	fs.sub.Unsubscribe()
 }
 
 // Add adds a filter to the filter manager
@@ -89,26 +91,7 @@ func (fs *FilterSystem) Get(id int) *Filter {
 // filterLoop waits for specific events from ethereum and fires their handlers
 // when the filter matches the requirements.
 func (fs *FilterSystem) filterLoop() {
-	// Subscribe to events
+	for event := range fs.sub.Chan() {
-	eventCh := fs.eventMux.Subscribe(
-		//core.PendingBlockEvent{},
-		core.ChainEvent{},
-		core.TxPreEvent{},
-		vm.Logs(nil),
-	).Chan()
-
-out:
-	for {
-		select {
-		case <-fs.quit:
-			break out
-		case event, ok := <-eventCh:
-			if !ok {
-				// Event subscription closed, set the channel to nil to stop spinning
-				eventCh = nil
-				continue
-			}
-			// A real event arrived, notify the registered filters
 		switch ev := event.Data.(type) {
 		case core.ChainEvent:
 			fs.filterMu.RLock()
@@ -142,4 +125,3 @@ out:
 		}
 	}
 }
-}

miner/miner.go

@@ -133,10 +133,13 @@ func (self *Miner) Register(agent Agent) {
 	if self.Mining() {
 		agent.Start()
 	}
-
 	self.worker.register(agent)
 }
 
+func (self *Miner) Unregister(agent Agent) {
+	self.worker.unregister(agent)
+}
+
 func (self *Miner) Mining() bool {
 	return atomic.LoadInt32(&self.mining) > 0
 }
@@ -146,7 +149,7 @@ func (self *Miner) HashRate() (tot int64) {
 	// do we care this might race? is it worth we're rewriting some
 	// aspects of the worker/locking up agents so we can get an accurate
 	// hashrate?
-	for _, agent := range self.worker.agents {
+	for agent := range self.worker.agents {
 		tot += agent.GetHashRate()
 	}
 	return

miner/remote_agent.go

@@ -48,9 +48,10 @@ type RemoteAgent struct {
 }
 
 func NewRemoteAgent() *RemoteAgent {
-	agent := &RemoteAgent{work: make(map[common.Hash]*Work), hashrate: make(map[common.Hash]hashrate)}
+	return &RemoteAgent{
-
+		work:     make(map[common.Hash]*Work),
-	return agent
+		hashrate: make(map[common.Hash]hashrate),
+	}
 }
 
 func (a *RemoteAgent) SubmitHashrate(id common.Hash, rate uint64) {
@@ -75,9 +76,13 @@ func (a *RemoteAgent) Start() {
 }
 
 func (a *RemoteAgent) Stop() {
+	if a.quit != nil {
 		close(a.quit)
+	}
+	if a.workCh != nil {
 		close(a.workCh)
 	}
+}
 
 // GetHashRate returns the accumulated hashrate of all identifier combined
 func (a *RemoteAgent) GetHashRate() (tot int64) {

miner/worker.go

@@ -92,7 +92,7 @@ type Result struct {
 type worker struct {
 	mu sync.Mutex
 
-	agents []Agent
+	agents map[Agent]struct{}
 	recv   chan *Result
 	mux    *event.TypeMux
 	quit   chan struct{}
@@ -136,6 +136,7 @@ func newWorker(coinbase common.Address, eth core.Backend) *worker {
 		coinbase:       coinbase,
 		txQueue:        make(map[common.Hash]*types.Transaction),
 		quit:           make(chan struct{}),
+		agents:         make(map[Agent]struct{}),
 		fullValidation: false,
 	}
 	go worker.update()
@@ -180,7 +181,7 @@ func (self *worker) start() {
 	atomic.StoreInt32(&self.mining, 1)
 
 	// spin up agents
-	for _, agent := range self.agents {
+	for agent := range self.agents {
 		agent.Start()
 	}
 }
@@ -190,16 +191,14 @@ func (self *worker) stop() {
 	defer self.mu.Unlock()
 
 	if atomic.LoadInt32(&self.mining) == 1 {
-		var keep []Agent
+		// Stop all agents.
-		// stop all agents
+		for agent := range self.agents {
-		for _, agent := range self.agents {
 			agent.Stop()
-			// keep all that's not a cpu agent
+			// Remove CPU agents.
-			if _, ok := agent.(*CpuAgent); !ok {
+			if _, ok := agent.(*CpuAgent); ok {
-				keep = append(keep, agent)
+				delete(self.agents, agent)
 			}
 		}
-		self.agents = keep
 	}
 
 	atomic.StoreInt32(&self.mining, 0)
@@ -209,10 +208,17 @@ func (self *worker) stop() {
 func (self *worker) register(agent Agent) {
 	self.mu.Lock()
 	defer self.mu.Unlock()
-	self.agents = append(self.agents, agent)
+	self.agents[agent] = struct{}{}
 	agent.SetReturnCh(self.recv)
 }
 
+func (self *worker) unregister(agent Agent) {
+	self.mu.Lock()
+	defer self.mu.Unlock()
+	delete(self.agents, agent)
+	agent.Stop()
+}
+
 func (self *worker) update() {
 	eventSub := self.mux.Subscribe(core.ChainHeadEvent{}, core.ChainSideEvent{}, core.TxPreEvent{})
 	defer eventSub.Unsubscribe()
@@ -341,11 +347,9 @@ func (self *worker) push(work *Work) {
 			glog.Infoln("You turn back and abort mining")
 			return
 		}
-
 		// push new work to agents
-		for _, agent := range self.agents {
+		for agent := range self.agents {
 			atomic.AddInt32(&self.atWork, 1)
-
 			if agent.Work() != nil {
 				agent.Work() <- work
 			}

rpc/comms/ipc.go

@@ -20,13 +20,22 @@ import (
 	"fmt"
 	"math/rand"
 	"net"
+	"os"
 
 	"encoding/json"
 
+	"github.com/ethereum/go-ethereum/logger"
+	"github.com/ethereum/go-ethereum/logger/glog"
 	"github.com/ethereum/go-ethereum/rpc/codec"
 	"github.com/ethereum/go-ethereum/rpc/shared"
 )
 
+type Stopper interface {
+	Stop()
+}
+
+type InitFunc func(conn net.Conn) (Stopper, shared.EthereumApi, error)
+
 type IpcConfig struct {
 	Endpoint string
 }
@@ -90,8 +99,38 @@ func NewIpcClient(cfg IpcConfig, codec codec.Codec) (*ipcClient, error) {
 }
 
 // Start IPC server
-func StartIpc(cfg IpcConfig, codec codec.Codec, initializer func(conn net.Conn) (shared.EthereumApi, error)) error {
+func StartIpc(cfg IpcConfig, codec codec.Codec, initializer InitFunc) error {
-	return startIpc(cfg, codec, initializer)
+	l, err := ipcListen(cfg)
+	if err != nil {
+		return err
+	}
+	go ipcLoop(cfg, codec, initializer, l)
+	return nil
+}
+
+func ipcLoop(cfg IpcConfig, codec codec.Codec, initializer InitFunc, l net.Listener) {
+	glog.V(logger.Info).Infof("IPC service started (%s)\n", cfg.Endpoint)
+	defer os.Remove(cfg.Endpoint)
+	defer l.Close()
+	for {
+		conn, err := l.Accept()
+		if err != nil {
+			glog.V(logger.Debug).Infof("accept: %v", err)
+			return
+		}
+		id := newIpcConnId()
+		go func() {
+			defer conn.Close()
+			glog.V(logger.Debug).Infof("new connection with id %06d started", id)
+			stopper, api, err := initializer(conn)
+			if err != nil {
+				glog.V(logger.Error).Infof("Unable to initialize IPC connection: %v", err)
+				return
+			}
+			defer stopper.Stop()
+			handle(id, conn, api, codec)
+		}()
+	}
 }
 
 func newIpcConnId() int {

rpc/comms/ipc_unix.go

@@ -23,8 +23,6 @@ import (
 	"os"
 	"path/filepath"
 
-	"github.com/ethereum/go-ethereum/logger"
-	"github.com/ethereum/go-ethereum/logger/glog"
 	"github.com/ethereum/go-ethereum/rpc/codec"
 	"github.com/ethereum/go-ethereum/rpc/shared"
 	"github.com/ethereum/go-ethereum/rpc/useragent"
@@ -69,44 +67,16 @@ func (self *ipcClient) reconnect() error {
 	return err
 }
 
-func startIpc(cfg IpcConfig, codec codec.Codec, initializer func(conn net.Conn) (shared.EthereumApi, error)) error {
+func ipcListen(cfg IpcConfig) (net.Listener, error) {
 	// Ensure the IPC path exists and remove any previous leftover
 	if err := os.MkdirAll(filepath.Dir(cfg.Endpoint), 0751); err != nil {
-		return err
+		return nil, err
 	}
 	os.Remove(cfg.Endpoint)
-
+	l, err := net.Listen("unix", cfg.Endpoint)
-	l, err := net.ListenUnix("unix", &net.UnixAddr{Name: cfg.Endpoint, Net: "unix"})
 	if err != nil {
-		return err
+		return nil, err
 	}
 	os.Chmod(cfg.Endpoint, 0600)
-
+	return l, nil
-	go func() {
-		for {
-			conn, err := l.AcceptUnix()
-			if err != nil {
-				glog.V(logger.Error).Infof("Error accepting ipc connection - %v\n", err)
-				continue
-			}
-
-			id := newIpcConnId()
-			glog.V(logger.Debug).Infof("New IPC connection with id %06d started\n", id)
-
-			api, err := initializer(conn)
-			if err != nil {
-				glog.V(logger.Error).Infof("Unable to initialize IPC connection - %v\n", err)
-				conn.Close()
-				continue
-			}
-
-			go handle(id, conn, api, codec)
-		}
-
-		os.Remove(cfg.Endpoint)
-	}()
-
-	glog.V(logger.Info).Infof("IPC service started (%s)\n", cfg.Endpoint)
-
-	return nil
 }

rpc/comms/ipc_windows.go

@@ -28,8 +28,6 @@ import (
 	"time"
 	"unsafe"
 
-	"github.com/ethereum/go-ethereum/logger"
-	"github.com/ethereum/go-ethereum/logger/glog"
 	"github.com/ethereum/go-ethereum/rpc/codec"
 	"github.com/ethereum/go-ethereum/rpc/shared"
 	"github.com/ethereum/go-ethereum/rpc/useragent"
@@ -688,40 +686,12 @@ func (self *ipcClient) reconnect() error {
 	return err
 }
 
-func startIpc(cfg IpcConfig, codec codec.Codec, initializer func(conn net.Conn) (shared.EthereumApi, error)) error {
+func ipcListen(cfg IpcConfig) (net.Listener, error) {
 	os.Remove(cfg.Endpoint) // in case it still exists from a previous run
-
 	l, err := Listen(cfg.Endpoint)
 	if err != nil {
-		return err
+		return nil, err
 	}
 	os.Chmod(cfg.Endpoint, 0600)
-
+	return l, nil
-	go func() {
-		for {
-			conn, err := l.Accept()
-			if err != nil {
-				glog.V(logger.Error).Infof("Error accepting ipc connection - %v\n", err)
-				continue
-			}
-
-			id := newIpcConnId()
-			glog.V(logger.Debug).Infof("New IPC connection with id %06d started\n", id)
-
-			api, err := initializer(conn)
-			if err != nil {
-				glog.V(logger.Error).Infof("Unable to initialize IPC connection - %v\n", err)
-				conn.Close()
-				continue
-			}
-
-			go handle(id, conn, api, codec)
-		}
-
-		os.Remove(cfg.Endpoint)
-	}()
-
-	glog.V(logger.Info).Infof("IPC service started (%s)\n", cfg.Endpoint)
-
-	return nil
 }

xeth/xeth.go

@@ -113,19 +113,15 @@ func New(ethereum *eth.Ethereum, frontend Frontend) *XEth {
 	if frontend == nil {
 		xeth.frontend = dummyFrontend{}
 	}
-	state, err := xeth.backend.BlockChain().State()
+	state, _ := xeth.backend.BlockChain().State()
-	if err != nil {
-		return nil
-	}
 	xeth.state = NewState(xeth, state)
-
 	go xeth.start()
-
 	return xeth
 }
 
 func (self *XEth) start() {
 	timer := time.NewTicker(2 * time.Second)
+	defer timer.Stop()
 done:
 	for {
 		select {
@@ -171,8 +167,12 @@ done:
 	}
 }
 
-func (self *XEth) stop() {
+// Stop releases any resources associated with self.
+// It may not be called more than once.
+func (self *XEth) Stop() {
 	close(self.quit)
+	self.filterManager.Stop()
+	self.backend.Miner().Unregister(self.agent)
 }
 
 func cAddress(a []string) []common.Address {