Use two sockets for video and control

The socket used the device-to-computer direction to stream the video and
the computer-to-device direction to send control events.

Some features, like copy-paste from device to computer, require to send
non-video data from the device to the computer.

To make them possible, use two sockets:
 - one for streaming the video from the device to the client;
 - one for control/events in both directions.

app/src/controller.c

@@ -7,7 +7,7 @@
 #include "log.h"
 
 bool
-controller_init(struct controller *controller, socket_t video_socket) {
+controller_init(struct controller *controller, socket_t control_socket) {
     cbuf_init(&controller->queue);
 
     if (!(controller->mutex = SDL_CreateMutex())) {
@@ -19,7 +19,7 @@ controller_init(struct controller *controller, socket_t video_socket) {
         return false;
     }
 
-    controller->video_socket = video_socket;
+    controller->control_socket = control_socket;
     controller->stopped = false;
 
     return true;
@@ -57,7 +57,7 @@ process_event(struct controller *controller,
     if (!length) {
         return false;
     }
-    int w = net_send_all(controller->video_socket, serialized_event, length);
+    int w = net_send_all(controller->control_socket, serialized_event, length);
     return w == length;
 }
 

app/src/controller.h

@@ -12,7 +12,7 @@
 struct control_event_queue CBUF(struct control_event, 64);
 
 struct controller {
-    socket_t video_socket;
+    socket_t control_socket;
     SDL_Thread *thread;
     SDL_mutex *mutex;
     SDL_cond *event_cond;
@@ -21,7 +21,7 @@ struct controller {
 };
 
 bool
-controller_init(struct controller *controller, socket_t video_socket);
+controller_init(struct controller *controller, socket_t control_socket);
 
 void
 controller_destroy(struct controller *controller);

app/src/scrcpy.c

@@ -300,15 +300,13 @@ scrcpy(const struct scrcpy_options *options) {
         goto end;
     }
 
-    socket_t device_socket = server.device_socket;
-
     char device_name[DEVICE_NAME_FIELD_LENGTH];
     struct size frame_size;
 
     // screenrecord does not send frames when the screen content does not
     // change therefore, we transmit the screen size before the video stream,
     // to be able to init the window immediately
-    if (!device_read_info(device_socket, device_name, &frame_size)) {
+    if (!device_read_info(server.video_socket, device_name, &frame_size)) {
         goto end;
     }
 
@@ -344,7 +342,7 @@ scrcpy(const struct scrcpy_options *options) {
 
     av_log_set_callback(av_log_callback);
 
-    stream_init(&stream, device_socket, dec, rec);
+    stream_init(&stream, server.video_socket, dec, rec);
 
     // now we consumed the header values, the socket receives the video stream
     // start the stream
@@ -355,7 +353,7 @@ scrcpy(const struct scrcpy_options *options) {
 
     if (display) {
         if (control) {
-            if (!controller_init(&controller, device_socket)) {
+            if (!controller_init(&controller, server.control_socket)) {
                 goto end;
             }
 

app/src/server.c

@@ -222,8 +222,14 @@ server_start(struct server *server, const char *serial,
 bool
 server_connect_to(struct server *server) {
     if (!server->tunnel_forward) {
-        server->device_socket = net_accept(server->server_socket);
-        if (server->device_socket == INVALID_SOCKET) {
+        server->video_socket = net_accept(server->server_socket);
+        if (server->video_socket == INVALID_SOCKET) {
+            return false;
+        }
+
+        server->control_socket = net_accept(server->server_socket);
+        if (server->control_socket == INVALID_SOCKET) {
+            // the video_socket will be clean up on destroy
             return false;
         }
 
@@ -232,9 +238,16 @@ server_connect_to(struct server *server) {
     } else {
         uint32_t attempts = 100;
         uint32_t delay = 100; // ms
-        server->device_socket = connect_to_server(server->local_port, attempts,
-                                                  delay);
-        if (server->device_socket == INVALID_SOCKET) {
+        server->video_socket =
+            connect_to_server(server->local_port, attempts, delay);
+        if (server->video_socket == INVALID_SOCKET) {
+            return false;
+        }
+
+        // we know that the device is listening, we don't need several attempts
+        server->control_socket =
+            net_connect(IPV4_LOCALHOST, server->local_port);
+        if (server->control_socket == INVALID_SOCKET) {
             return false;
         }
     }
@@ -251,8 +264,11 @@ server_stop(struct server *server) {
     if (server->server_socket != INVALID_SOCKET) {
         close_socket(&server->server_socket);
     }
-    if (server->device_socket != INVALID_SOCKET) {
-        close_socket(&server->device_socket);
+    if (server->video_socket != INVALID_SOCKET) {
+        close_socket(&server->video_socket);
+    }
+    if (server->control_socket != INVALID_SOCKET) {
+        close_socket(&server->control_socket);
     }
 
     SDL_assert(server->process != PROCESS_NONE);

app/src/server.h

@@ -11,7 +11,8 @@ struct server {
     char *serial;
     process_t process;
     socket_t server_socket; // only used if !tunnel_forward
-    socket_t device_socket;
+    socket_t video_socket;
+    socket_t control_socket;
     uint16_t local_port;
     bool tunnel_enabled;
     bool tunnel_forward; // use "adb forward" instead of "adb reverse"
@@ -22,7 +23,8 @@ struct server {
     .serial = NULL,                   \
     .process = PROCESS_NONE,          \
     .server_socket = INVALID_SOCKET,  \
-    .device_socket = INVALID_SOCKET,  \
+    .video_socket = INVALID_SOCKET,   \
+    .control_socket = INVALID_SOCKET, \
     .local_port = 0,                  \
     .tunnel_enabled = false,          \
     .tunnel_forward = false,          \