diff --git a/client/embed/embed.go b/client/embed/embed.go index 0e8991be2..d0d88b177 100644 --- a/client/embed/embed.go +++ b/client/embed/embed.go @@ -279,9 +279,11 @@ func (c *Client) Start(startCtx context.Context) error { select { case <-startCtx.Done(): - // Cancel the client context before stopping: Engine.Start blocks on the - // signal stream while holding the engine mutex and only unblocks on - // cancellation. Stopping first would deadlock on that mutex. + // ConnectClient.Stop now cancels its own run context and waits for the + // run loop to tear the engine down, so this cancel() is no longer + // required to break the deadlock and could be removed. It is kept as a + // defensive belt-and-suspenders: cancelling the parent context first + // guarantees the run loop is unblocked even if Stop's contract regresses. cancel() if stopErr := client.Stop(); stopErr != nil { return fmt.Errorf("stop error after context done. Stop error: %w. Context done: %w", stopErr, startCtx.Err()) diff --git a/client/internal/connect.go b/client/internal/connect.go index e38bc2f58..faf20078a 100644 --- a/client/internal/connect.go +++ b/client/internal/connect.go @@ -11,6 +11,7 @@ import ( "runtime/debug" "strings" "sync" + "sync/atomic" "time" "github.com/cenkalti/backoff/v4" @@ -54,6 +55,10 @@ var androidRunOverride func(c *ConnectClient, runningChan chan struct{}, logPath type ConnectClient struct { ctx context.Context + runCancel context.CancelFunc + runExited chan struct{} + runOnce sync.Once + runStarted atomic.Bool config *profilemanager.Config statusRecorder *peer.Status @@ -70,8 +75,14 @@ func NewConnectClient( config *profilemanager.Config, statusRecorder *peer.Status, ) *ConnectClient { + // Derive the run context here so Stop owns the cancel that unblocks the run + // loop. runCancel is set once at construction, so Stop can call it without + // racing the run loop's startup. Callers therefore need not cancel before Stop. + runCtx, runCancel := context.WithCancel(ctx) return &ConnectClient{ - ctx: ctx, + ctx: runCtx, + runCancel: runCancel, + runExited: make(chan struct{}), config: config, statusRecorder: statusRecorder, engineMutex: sync.Mutex{}, @@ -132,6 +143,11 @@ func (c *ConnectClient) RunOniOS( } func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan struct{}, logPath string) error { + // Mark the loop as started and signal exit on return so Stop can wait for + // the loop to finish (and skip the wait if the loop never ran). + c.runStarted.Store(true) + defer c.runOnce.Do(func() { close(c.runExited) }) + defer func() { if r := recover(); r != nil { rec := c.statusRecorder @@ -287,7 +303,7 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan log.Debug(err) if s, ok := gstatus.FromError(err); ok && (s.Code() == codes.PermissionDenied) { state.Set(StatusNeedsLogin) - _ = c.Stop() + c.runCancel() return backoff.Permanent(wrapErr(err)) // unrecoverable error } return wrapErr(err) @@ -407,14 +423,10 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan c.engine = nil c.engineMutex.Unlock() - // todo: consider to remove this condition. Is not thread safe. - // We should always call Stop(), but we need to verify that it is idempotent - if engine.wgInterface != nil { - log.Infof("ensuring %s is removed, Netbird engine context cancelled", engine.wgInterface.Name()) + log.Infof("ensuring wg interface is removed, Netbird engine context cancelled") - if err := engine.Stop(); err != nil { - log.Errorf("Failed to stop engine: %v", err) - } + if err := engine.Stop(); err != nil { + log.Errorf("Failed to stop engine: %v", err) } c.statusRecorder.ClientTeardown() @@ -435,7 +447,7 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan log.Debugf("exiting client retry loop due to unrecoverable error: %s", err) if s, ok := gstatus.FromError(err); ok && (s.Code() == codes.PermissionDenied) { state.Set(StatusNeedsLogin) - _ = c.Stop() + c.runCancel() } return err } @@ -513,11 +525,9 @@ func (c *ConnectClient) Status() StatusType { } func (c *ConnectClient) Stop() error { - engine := c.Engine() - if engine != nil { - if err := engine.Stop(); err != nil { - return fmt.Errorf("stop engine: %w", err) - } + c.runCancel() + if c.runStarted.Load() { + <-c.runExited } return nil } diff --git a/client/server/server.go b/client/server/server.go index 32daf7718..b72a52c9d 100644 --- a/client/server/server.go +++ b/client/server/server.go @@ -988,6 +988,10 @@ func (s *Server) cleanupConnection() error { return nil } + // TODO: consider calling s.connectClient.Stop() instead of engine.Stop(). + // actCancel() lets the run loop stop the engine too, so both stop it + // concurrently; ConnectClient.Stop cancels and waits for the run loop, + // making the run loop the sole owner of engine shutdown. if engine != nil { if err := engine.Stop(); err != nil { return err