Fix race conditions and improve lifecycle safety (#150) (#152)

## Summary
Comprehensive fixes for race conditions, concurrency bugs, and code quality improvements in the lifecycle module and related test files.

## Changes

### Race Condition Fixes
- Fixed data race in `SetChannel()` and `Channel()` by adding mutex protection
- Fixed data race in `StartedAt()` by adding mutex protection
- Fixed test race conditions in `transport` and `interaction` packages where variables were shared between goroutines

### Concurrency Safety Improvements
- Refactored `Close()` to release mutex before calling external code, preventing potential deadlocks
- Made `SetChannel()` reject calls after `Close()` to prevent setting channel on torn-down lifecycle
- Made `SetStatus()` ignore calls after `Close()` to ensure CLOSED is a terminal state
- Added nil check to `SetChannel()` to prevent masking upstream bugs

### Code Quality Improvements
- Refactored `Close()` to extract helper methods: `closeChannel()`, `closeConnection()`, `cleanupRegistry()`, `cleanupForwarder()`
- Fixed `isClosedError()` to remove redundant string comparison
- Added `PortRegistry` interface for better interface segregation (removed dependency on full `port.Port`)
- Made `cleanupRegistry()` conditional on non-empty slug to avoid unnecessary registry calls
- Fixed `startedAt` to be set when session starts running instead of at object creation

Reviewed-on: #150
Co-authored-by: Bagas <bagas@fossy.my.id>
Co-committed-by: Bagas <bagas@fossy.my.id>

Reviewed-on: #152

internal/grpc/client/client_test.go

@@ -13,7 +13,6 @@ import (
 	"tunnel_pls/internal/session/slug"
 	"tunnel_pls/internal/types"
 
-	"tunnel_pls/internal/port"
 	"tunnel_pls/internal/registry"
 
 	proto "git.fossy.my.id/bagas/tunnel-please-grpc/gen"
@@ -885,16 +884,16 @@ func (m *mockLifecycle) Connection() ssh.Conn {
 	return args.Get(0).(ssh.Conn)
 }
 func (m *mockLifecycle) User() string                         { return m.Called().String(0) }
-func (m *mockLifecycle) SetChannel(channel ssh.Channel)       { m.Called(channel) }
+func (m *mockLifecycle) SetChannel(channel ssh.Channel) error { return m.Called(channel).Error(0) }
 func (m *mockLifecycle) SetStatus(status types.SessionStatus) { m.Called(status) }
 func (m *mockLifecycle) IsActive() bool                       { return m.Called().Bool(0) }
 func (m *mockLifecycle) StartedAt() time.Time                 { return m.Called().Get(0).(time.Time) }
-func (m *mockLifecycle) PortRegistry() port.Port {
+func (m *mockLifecycle) PortRegistry() lifecycle.PortRegistry {
 	args := m.Called()
 	if args.Get(0) == nil {
 		return nil
 	}
-	return args.Get(0).(port.Port)
+	return args.Get(0).(lifecycle.PortRegistry)
 }
 
 type mockEventServiceClient struct {

internal/registry/registry_test.go

@@ -4,7 +4,6 @@ import (
 	"sync"
 	"testing"
 	"time"
-	"tunnel_pls/internal/port"
 	"tunnel_pls/internal/session/forwarder"
 	"tunnel_pls/internal/session/interaction"
 	"tunnel_pls/internal/session/lifecycle"
@@ -78,15 +77,15 @@ func (ml *mockLifecycle) Connection() ssh.Conn {
 	return args.Get(0).(ssh.Conn)
 }
 
-func (ml *mockLifecycle) PortRegistry() port.Port {
+func (ml *mockLifecycle) PortRegistry() lifecycle.PortRegistry {
 	args := ml.Called()
 	if args.Get(0) == nil {
 		return nil
 	}
-	return args.Get(0).(port.Port)
+	return args.Get(0).(lifecycle.PortRegistry)
 }
 
-func (ml *mockLifecycle) SetChannel(channel ssh.Channel)       { ml.Called(channel) }
+func (ml *mockLifecycle) SetChannel(channel ssh.Channel) error { return ml.Called(channel).Error(0) }
 func (ml *mockLifecycle) SetStatus(status types.SessionStatus) { ml.Called(status) }
 func (ml *mockLifecycle) IsActive() bool                       { return ml.Called().Bool(0) }
 func (ml *mockLifecycle) StartedAt() time.Time                 { return ml.Called().Get(0).(time.Time) }

internal/session/interaction/interaction_test.go

@@ -1922,10 +1922,6 @@ func TestInteraction_Start_ProtocolSelection(t *testing.T) {
 			time.Sleep(50 * time.Millisecond)
 
 			i := mockInteraction.(*interaction)
-			if i.program != nil {
-				assert.NotNil(t, i.program, "program should be initialized")
-			}
-
 			i.Stop()
 
 			mockConfig.AssertExpectations(t)

internal/session/lifecycle/lifecycle.go

@@ -2,6 +2,7 @@ package lifecycle
 
 import (
 	"errors"
+	"fmt"
 	"io"
 	"net"
 	"sync"
@@ -9,8 +10,6 @@ import (
 	"tunnel_pls/internal/session/slug"
 	"tunnel_pls/internal/types"
 
-	portUtil "tunnel_pls/internal/port"
-
 	"golang.org/x/crypto/ssh"
 )
 
@@ -24,6 +23,12 @@ type SessionRegistry interface {
 	Remove(key types.SessionKey)
 }
 
+type PortRegistry interface {
+	Unassigned() (uint16, bool)
+	Claim(port uint16) bool
+	SetStatus(port uint16, assigned bool) error
+}
+
 type lifecycle struct {
 	mu              sync.Mutex
 	status          types.SessionStatus
@@ -34,18 +39,18 @@ type lifecycle struct {
 	slug            slug.Slug
 	startedAt       time.Time
 	sessionRegistry SessionRegistry
-	portRegistry    portUtil.Port
+	portRegistry    PortRegistry
 	user            string
 }
 
-func New(conn ssh.Conn, forwarder Forwarder, slugManager slug.Slug, port portUtil.Port, sessionRegistry SessionRegistry, user string) Lifecycle {
+func New(conn ssh.Conn, forwarder Forwarder, slugManager slug.Slug, port PortRegistry, sessionRegistry SessionRegistry, user string) Lifecycle {
 	return &lifecycle{
 		status:          types.SessionStatusINITIALIZING,
 		conn:            conn,
 		channel:         nil,
 		forwarder:       forwarder,
 		slug:            slugManager,
-		startedAt:       time.Now(),
+		startedAt:       time.Time{},
 		sessionRegistry: sessionRegistry,
 		portRegistry:    port,
 		user:            user,
@@ -55,16 +60,16 @@ func New(conn ssh.Conn, forwarder Forwarder, slugManager slug.Slug, port portUti
 type Lifecycle interface {
 	Connection() ssh.Conn
 	Channel() ssh.Channel
-	PortRegistry() portUtil.Port
+	PortRegistry() PortRegistry
 	User() string
-	SetChannel(channel ssh.Channel)
+	SetChannel(channel ssh.Channel) error
 	SetStatus(status types.SessionStatus)
 	IsActive() bool
 	StartedAt() time.Time
 	Close() error
 }
 
-func (l *lifecycle) PortRegistry() portUtil.Port {
+func (l *lifecycle) PortRegistry() PortRegistry {
 	return l.portRegistry
 }
 
@@ -72,11 +77,25 @@ func (l *lifecycle) User() string {
 	return l.user
 }
 
-func (l *lifecycle) SetChannel(channel ssh.Channel) {
+func (l *lifecycle) SetChannel(channel ssh.Channel) error {
+	l.mu.Lock()
+	defer l.mu.Unlock()
+	if l.status == types.SessionStatusCLOSED {
+		return fmt.Errorf("lifecycle is closed")
+	}
+	if channel == nil {
+		return fmt.Errorf("channel cannot be nil")
+	}
+	if l.channel != nil {
+		return fmt.Errorf("channel already set")
+	}
 	l.channel = channel
+	return nil
 }
 
 func (l *lifecycle) Channel() ssh.Channel {
+	l.mu.Lock()
+	defer l.mu.Unlock()
 	return l.channel
 }
 
@@ -87,7 +106,13 @@ func (l *lifecycle) Connection() ssh.Conn {
 func (l *lifecycle) SetStatus(status types.SessionStatus) {
 	l.mu.Lock()
 	defer l.mu.Unlock()
+	if l.status == types.SessionStatusCLOSED {
+		return
+	}
 	l.status = status
+	if status == types.SessionStatusRUNNING && l.startedAt.IsZero() {
+		l.startedAt = time.Now()
+	}
 }
 
 func (l *lifecycle) IsActive() bool {
@@ -98,50 +123,74 @@ func (l *lifecycle) IsActive() bool {
 
 func (l *lifecycle) Close() error {
 	l.mu.Lock()
-	defer l.mu.Unlock()
 	if l.status == types.SessionStatusCLOSED {
-		return l.closeErr
+		closeErr := l.closeErr
+		l.mu.Unlock()
+		return closeErr
 	}
 	l.status = types.SessionStatusCLOSED
 
+	channel := l.channel
+	conn := l.conn
+	l.mu.Unlock()
+
 	var errs []error
-	tunnelType := l.forwarder.TunnelType()
-
-	if l.channel != nil {
-		if err := l.channel.Close(); err != nil && !isClosedError(err) {
+	if channel != nil {
+		if err := channel.Close(); err != nil && !isClosedError(err) {
+			errs = append(errs, err)
+		}
+	}
+	if conn != nil {
+		if err := conn.Close(); err != nil && !isClosedError(err) {
 			errs = append(errs, err)
 		}
 	}
 
-	if l.conn != nil {
-		if err := l.conn.Close(); err != nil && !isClosedError(err) {
-			errs = append(errs, err)
-		}
+	l.cleanupRegistry()
+	if err := l.cleanupForwarder(); err != nil {
+		errs = append(errs, err)
 	}
 
-	clientSlug := l.slug.String()
+	closeErr := errors.Join(errs...)
+
+	l.mu.Lock()
+	l.closeErr = closeErr
+	l.mu.Unlock()
+
+	return closeErr
+}
+
+func (l *lifecycle) cleanupRegistry() {
+	slugStr := l.slug.String()
+	if slugStr == "" {
+		return
+	}
 	key := types.SessionKey{
-		Id:   clientSlug,
-		Type: tunnelType,
+		Id:   slugStr,
+		Type: l.forwarder.TunnelType(),
 	}
 	l.sessionRegistry.Remove(key)
+}
 
-	if tunnelType == types.TunnelTypeTCP {
-		errs = append(errs, l.PortRegistry().SetStatus(l.forwarder.ForwardedPort(), false))
-		errs = append(errs, l.forwarder.Close())
+func (l *lifecycle) cleanupForwarder() error {
+	if l.forwarder.TunnelType() != types.TunnelTypeTCP {
+		return nil
 	}
-
-	l.closeErr = errors.Join(errs...)
-	return l.closeErr
+	var errs []error
+	errs = append(errs, l.portRegistry.SetStatus(l.forwarder.ForwardedPort(), false))
+	errs = append(errs, l.forwarder.Close())
+	return errors.Join(errs...)
 }
 
 func isClosedError(err error) bool {
 	if err == nil {
 		return false
 	}
-	return errors.Is(err, io.EOF) || errors.Is(err, net.ErrClosed) || err.Error() == "EOF"
+	return errors.Is(err, io.EOF) || errors.Is(err, net.ErrClosed)
 }
 
 func (l *lifecycle) StartedAt() time.Time {
+	l.mu.Lock()
+	defer l.mu.Unlock()
 	return l.startedAt
 }

internal/session/lifecycle/lifecycle_test.go

@@ -5,6 +5,7 @@ import (
 	"errors"
 	"io"
 	"net"
+	"sync"
 	"testing"
 	"tunnel_pls/internal/types"
 
@@ -177,8 +178,14 @@ func TestLifecycle_SetChannel(t *testing.T) {
 
 	mockSSHChannel := &MockSSHChannel{}
 
-	mockLifecycle.SetChannel(mockSSHChannel)
+	err := mockLifecycle.SetChannel(mockSSHChannel)
+	assert.NoError(t, err)
+	assert.Equal(t, mockSSHChannel, mockLifecycle.Channel())
 
+	anotherChannel := &MockSSHChannel{}
+	err = mockLifecycle.SetChannel(anotherChannel)
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "channel already set")
 	assert.Equal(t, mockSSHChannel, mockLifecycle.Channel())
 }
 
@@ -276,14 +283,15 @@ func TestLifecycle_Close(t *testing.T) {
 			mockLifecycle := New(mockSSHConn, mockForwarder, mockSlug, mockPort, mockSessionRegistry, "mas-fuad")
 
 			mockLifecycle.SetStatus(types.SessionStatusRUNNING)
-			mockLifecycle.SetChannel(mockSSHChannel)
+			err := mockLifecycle.SetChannel(mockSSHChannel)
+			assert.NoError(t, err)
 
 			if tt.alreadyClosed {
-				err := mockLifecycle.Close()
+				err = mockLifecycle.Close()
 				assert.NoError(t, err)
 			}
 
-			err := mockLifecycle.Close()
+			err = mockLifecycle.Close()
 
 			if tt.expectErr {
 				assert.Error(t, err)
@@ -301,3 +309,116 @@ func TestLifecycle_Close(t *testing.T) {
 		})
 	}
 }
+
+func TestLifecycle_ConcurrentClose(t *testing.T) {
+	mockSSHConn := &MockSSHConn{}
+	mockSSHConn.On("Close").Return(nil)
+
+	mockForwarder := &MockForwarder{}
+	mockForwarder.On("TunnelType").Return(types.TunnelTypeHTTP)
+
+	mockSlug := &MockSlug{}
+	mockSlug.On("String").Return("test-slug")
+
+	mockPort := &MockPort{}
+
+	mockSessionRegistry := &MockSessionRegistry{}
+	mockSessionRegistry.On("Remove", mock.Anything).Return()
+
+	mockSSHChannel := &MockSSHChannel{}
+	mockSSHChannel.On("Close").Return(nil)
+
+	mockLifecycle := New(mockSSHConn, mockForwarder, mockSlug, mockPort, mockSessionRegistry, "mas-fuad")
+	mockLifecycle.SetStatus(types.SessionStatusRUNNING)
+	err := mockLifecycle.SetChannel(mockSSHChannel)
+	assert.NoError(t, err)
+
+	const numGoroutines = 10
+	var wg sync.WaitGroup
+	errChan := make(chan error, numGoroutines)
+
+	for i := 0; i < numGoroutines; i++ {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			err := mockLifecycle.Close()
+			errChan <- err
+		}()
+	}
+
+	wg.Wait()
+	close(errChan)
+
+	for err := range errChan {
+		assert.NoError(t, err)
+	}
+
+	assert.False(t, mockLifecycle.IsActive())
+}
+
+func TestLifecycle_SetChannel_AfterClose(t *testing.T) {
+	mockSSHConn := new(MockSSHConn)
+	mockSSHConn.On("Close").Return(nil)
+	mockForwarder := &MockForwarder{}
+	mockForwarder.On("TunnelType").Return(types.TunnelTypeHTTP)
+	mockSlug := &MockSlug{}
+	mockSlug.On("String").Return("test-slug")
+	mockPort := &MockPort{}
+	mockSessionRegistry := &MockSessionRegistry{}
+	mockSessionRegistry.On("Remove", mock.Anything).Return()
+	mockSSHChannel := &MockSSHChannel{}
+	mockSSHChannel.On("Close").Return(nil)
+
+	mockLifecycle := New(mockSSHConn, mockForwarder, mockSlug, mockPort, mockSessionRegistry, "mas-fuad")
+	mockLifecycle.SetStatus(types.SessionStatusRUNNING)
+	err := mockLifecycle.SetChannel(mockSSHChannel)
+	assert.NoError(t, err)
+
+	err = mockLifecycle.Close()
+	assert.NoError(t, err)
+
+	anotherChannel := &MockSSHChannel{}
+	err = mockLifecycle.SetChannel(anotherChannel)
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "lifecycle is closed")
+}
+
+func TestLifecycle_SetChannel_Nil(t *testing.T) {
+	mockSSHConn := new(MockSSHConn)
+	mockForwarder := &MockForwarder{}
+	mockSlug := &MockSlug{}
+	mockPort := &MockPort{}
+	mockSessionRegistry := &MockSessionRegistry{}
+
+	mockLifecycle := New(mockSSHConn, mockForwarder, mockSlug, mockPort, mockSessionRegistry, "mas-fuad")
+
+	err := mockLifecycle.SetChannel(nil)
+	assert.Error(t, err)
+	assert.Contains(t, err.Error(), "channel cannot be nil")
+}
+
+func TestLifecycle_SetStatus_AfterClose(t *testing.T) {
+	mockSSHConn := new(MockSSHConn)
+	mockSSHConn.On("Close").Return(nil)
+	mockForwarder := &MockForwarder{}
+	mockForwarder.On("TunnelType").Return(types.TunnelTypeHTTP)
+	mockSlug := &MockSlug{}
+	mockSlug.On("String").Return("test-slug")
+	mockPort := &MockPort{}
+	mockSessionRegistry := &MockSessionRegistry{}
+	mockSessionRegistry.On("Remove", mock.Anything).Return()
+	mockSSHChannel := &MockSSHChannel{}
+	mockSSHChannel.On("Close").Return(nil)
+
+	mockLifecycle := New(mockSSHConn, mockForwarder, mockSlug, mockPort, mockSessionRegistry, "mas-fuad")
+	mockLifecycle.SetStatus(types.SessionStatusRUNNING)
+	err := mockLifecycle.SetChannel(mockSSHChannel)
+	assert.NoError(t, err)
+
+	err = mockLifecycle.Close()
+	assert.NoError(t, err)
+	assert.False(t, mockLifecycle.IsActive())
+
+	mockLifecycle.SetStatus(types.SessionStatusRUNNING)
+	assert.False(t, mockLifecycle.IsActive(), "SetStatus should be ignored after Close")
+}

internal/session/session.go

@@ -164,7 +164,9 @@ func (s *session) setupInteractiveMode(channel ssh.NewChannel) error {
 		}
 	}()
 
-	s.lifecycle.SetChannel(ch)
+	if err = s.lifecycle.SetChannel(ch); err != nil {
+		return err
+	}
 	s.interaction.SetChannel(ch)
 	s.interaction.SetMode(types.InteractiveModeINTERACTIVE)
 

internal/transport/http_test.go

@@ -55,8 +55,7 @@ func TestHTTPServer_Serve(t *testing.T) {
 
 	go func() {
 		time.Sleep(100 * time.Millisecond)
-		err = listener.Close()
-		assert.NoError(t, err)
+		_ = listener.Close()
 	}()
 
 	err = srv.Serve(listener)

internal/transport/https_test.go

@@ -63,8 +63,7 @@ func TestHTTPSServer_Serve(t *testing.T) {
 
 	go func() {
 		time.Sleep(100 * time.Millisecond)
-		err = listener.Close()
-		assert.NoError(t, err)
+		_ = listener.Close()
 	}()
 
 	err = srv.Serve(listener)

internal/transport/tcp_test.go

@@ -45,8 +45,7 @@ func TestTCPServer_Serve(t *testing.T) {
 
 	go func() {
 		time.Sleep(100 * time.Millisecond)
-		err = listener.Close()
-		assert.NoError(t, err)
+		_ = listener.Close()
 	}()
 
 	err = srv.Serve(listener)