diff --git a/src/xpra/server/proxy_server.py b/src/xpra/server/proxy_server.py
index 1cc1ff4f55..15e3b89dab 100644
--- a/src/xpra/server/proxy_server.py
+++ b/src/xpra/server/proxy_server.py
@@ -37,7 +37,7 @@ class ProxyServer(ServerCore):
     """
 
     def __init__(self):
-        log("ProxyServer.__init__()")
+        debug("ProxyServer.__init__()")
         ServerCore.__init__(self)
         self._max_connections = MAX_CONCURRENT_CONNECTIONS
         self.main_loop = None
@@ -49,6 +49,7 @@ def __init__(self):
         self.timeout_add = gobject.timeout_add
         self.source_remove = gobject.source_remove
         self._socket_timeout = PROXY_SOCKET_TIMEOUT
+        self.control_commands = ["hello", "stop"]
         #ensure we cache the platform info before intercepting SIGCHLD
         #as this will cause a fork and SIGCHLD to be emitted: 
         from xpra.version_util import get_platform_info
@@ -56,7 +57,7 @@ def __init__(self):
         signal.signal(signal.SIGCHLD, self.sigchld)
 
     def init(self, opts):
-        log("ProxyServer.init(%s)", opts)
+        debug("ProxyServer.init(%s)", opts)
         if not opts.auth:
             raise Exception("The proxy server requires an authentication mode")
         ServerCore.init(self, opts)
@@ -71,15 +72,35 @@ def do_run(self):
         self.main_loop = gobject.MainLoop()
         self.main_loop.run()
 
+    def do_handle_command_request(self, proto, command, args):
+        if command in ("help", "hello"):
+            return ServerCore.do_handle_command_request(self, proto, command, args)
+        assert command=="stop"
+        if len(args)!=1:
+            return ServerCore.control_command_response(self, proto, command, 4, "invalid number of arguments, usage: 'xpra control stop DISPLAY'")
+        display = args[0]
+        debug("stop command: will try to find proxy process for display %s", display)
+        for process, v in list(self.processes.items()):
+            disp,mq = v
+            if disp==display:
+                pid = process.pid
+                log.info("stop command: found process %s with pid %s for display %s, sending it 'stop' request", process, pid, display)
+                mq.put("stop")
+                return self.control_command_response(proto, command, 0, "stopped proxy process with pid %s" % pid)
+        return self.control_command_response(proto, command, 14, "no proxy found for display %s" % display)
+
+
     def stop_all_proxies(self):
         processes = self.processes
         self.processes = {}
-        log("stop_all_proxies() will stop proxy processes: %s", processes)
+        debug("stop_all_proxies() will stop proxy processes: %s", processes)
         for process, v in processes.items():
+            if not process.is_alive():
+                continue
             disp,mq = v
-            log("stop_all_proxies() stopping process %s for display %s", process, disp)
+            debug("stop_all_proxies() stopping process %s for display %s", process, disp)
             mq.put("stop")
-        log("stop_all_proxies() done")
+        debug("stop_all_proxies() done")
 
     def cleanup(self):
         self.stop_all_proxies()
diff --git a/src/xpra/server/server_base.py b/src/xpra/server/server_base.py
index 174be49553..a3b5aa1b1e 100644
--- a/src/xpra/server/server_base.py
+++ b/src/xpra/server/server_base.py
@@ -76,6 +76,14 @@ def __init__(self):
         self.lossless_mode_encodings = []
         self.default_encoding = None
 
+        #control mode:
+        self.control_commands = ["hello", "help",
+                    "compression", "encoder", "refresh",
+                    "sound-output",
+                    "scaling",
+                    "suspend", "resume", "name",
+                    "client"]
+
         self.init_encodings()
         self.init_packet_handlers()
         self.init_aliases()
@@ -443,10 +451,6 @@ def hello_oked(self, proto, packet, c, auth_caps):
         if c.boolget("info_request", False):
             self.send_hello_info(proto)
             return
-        command_req = c.strlistget("command_request")
-        if len(command_req)>0:
-            self.handle_command_request(proto, command_req)
-            return
 
         #"normal" connection, so log welcome message:
         log.info("Handshake complete; enabling connection")
@@ -621,17 +625,8 @@ def send_hello(self, server_source, root_w, root_h, key_repeat, server_cipher):
         server_source.hello(capabilities)
 
 
-    def handle_command_request(self, proto, args):
-        try:
-            self.do_handle_command_request(proto, args)
-        except Exception, e:
-            log.error("error processing command %s", args, exc_info=True)
-            proto.send_now(("hello", {"command_response"  : (127, "error processing command: %s" % e)}))
-
-    def do_handle_command_request(self, proto, args):
-        assert len(args)>0
-        log("handle_command_request(%s, %s)", proto, args)
-        command = args[0]
+    def do_handle_command_request(self, proto, command, args):
+        log("handle_command_request(%s, %s, %s)", proto, command, args)
         def respond(error=0, response=""):
             log("command request response(%s)=%s", command, response)
             hello = {"command_response"  : (error, response)}
@@ -643,117 +638,123 @@ def arg_err(n, msg):
         def success():
             respond(0, "success")
 
-        commands = ("hello",
-                    "compression", "encoder",
-                    "sound-output",
-                    "scaling",
-                    "suspend", "resume", "name",
-                    "client")
-        if command=="help":
-            return respond(0, "control supports: %s" % (", ".join(commands)))
-
-        if command not in commands:
-            return respond(6, "invalid command")
-
-        if command=="hello":
-            return respond(0, "hello")
-        #from here on, we assume the command applies to the
-        #current client connection, of which there must only be one:
-        sss = list(self._server_sources.items())
-        if len(sss)==0:
-            return respond(2, "no client connected")
-        elif len(sss)>1:
-            return respond(3, "more than one client connected")
-        cproto, csource = sss[0]
-
-        def may_forward_client_command(client_command):
-            if client_command[0] not in csource.control_commands:
-                log.info("not forwarded to client (not supported)")
-                return  False
-            csource.send_client_command(*client_command)
-            return True
-
-        log("handle_command_request will apply to client: %s", csource)
-        if command=="compression":
-            if len(args)!=2:
-                return argn_err(2)
-            compression = args[1].lower()
+        sources = list(self._server_sources.values())
+        protos = list(self._server_sources.keys())
+        def forward_all_clients(client_command):
+            """ forwards the command to all clients """
+            for source in sources:
+                """ forwards to *the* client, if there is *one* """
+                if client_command[0] not in source.control_commands:
+                    log.info("client command '%s' not forwarded to client %s (not supported)", client_command, source)
+                    return  False
+                source.send_client_command(*client_command)
+
+        #handle commands that either don't require a client,
+        #or can work on more than one connected client:
+        if command in ("help", "hello"):
+            #generic case:
+            return ServerCore.do_handle_command_request(self, proto, command, args)
+        elif command=="name":
+            if len(args)!=1:
+                return argn_err(1)
+            self.session_name = args[0]
+            log.info("changed session name: %s", self.session_name)
+            forward_all_clients(["name"])
+            return respond(0, "session name set")
+        elif command=="compression":
+            if len(args)!=1:
+                return argn_err(1)
+            compression = args[0].lower()
             opts = ("lz4", "zlib")
             if compression=="lz4":
-                cproto.enable_lz4()
-                may_forward_client_command(["enable_lz4"])
+                for cproto in protos:
+                    cproto.enable_lz4()
+                forward_all_clients(["enable_lz4"])
                 return success()
             elif compression=="zlib":
-                cproto.enable_zlib()
-                may_forward_client_command(["enable_zlib"])
+                for cproto in protos:
+                    cproto.enable_zlib()
+                forward_all_clients(["enable_zlib"])
                 return success()
             return arg_err(1, "must be one of: %s" % (", ".join(opts)))
         elif command=="encoder":
-            if len(args)!=2:
-                return argn_err(2)
-            encoder = args[1].lower()
+            if len(args)!=1:
+                return argn_err(1)
+            encoder = args[0].lower()
             opts = ("bencode", "rencode")
             if encoder=="bencode":
-                cproto.enable_bencode()
-                may_forward_client_command(["enable_bencode"])
+                for cproto in protos:
+                    cproto.enable_bencode()
+                forward_all_clients(["enable_bencode"])
                 return success()
             elif encoder=="rencode":
-                cproto.enable_rencode()
-                may_forward_client_command(["enable_rencode"])
+                for cproto in protos:
+                    cproto.enable_rencode()
+                forward_all_clients(["enable_rencode"])
                 return success()
             return arg_err(1, "must be one of: %s" % (", ".join(opts)))
         elif command=="sound-output":
-            if len(args)<2:
+            if len(args)<1:
                 return argn_err("more than 1")
-            msg = csource.sound_control(*args[1:])
-            return respond(0, msg)
+            msg = []
+            for csource in sources:
+                msg.append("%s : %s" % (csource, csource.sound_control(*args[1:])))
+            return respond(0, ", ".join(msg))
         elif command=="suspend":
-            csource.suspend(True, self._id_to_window)
-            return respond(0, "suspended")
+            for csource in sources:
+                csource.suspend(True, self._id_to_window)
+            return respond(0, "suspended %s clients" % len(sources))
         elif command=="resume":
-            csource.resume(True, self._id_to_window)
-            return respond(0, "resumed")
+            for csource in sources:
+                csource.resume(True, self._id_to_window)
+            return respond(0, "resumed %s clients" % len(sources))
+        elif command=="refresh":
+            widwin = list(self._id_to_window.items())
+            for wid, window in widwin:
+                for csource in sources:
+                    csource.full_quality_refresh(wid, window, {})
+            return respond(0, "refreshed %s window for %s clients" % (len(widwin), len(sources)))
         elif command=="scaling":
-            if len(args)!=3:
-                return argn_err(3)
-            if args[1]=="*":
-                wids = csource.window_sources.keys()
+            if len(args)!=2:
+                return argn_err(2)
+            from xpra.server.window_video_source import parse_scaling_value
+            try:
+                scaling = parse_scaling_value(args[1])
+            except:
+                return respond(11, "invalid scaling value %s" % args[1])
+            if args[0]=="*":
+                wids = list(self._id_to_window.keys())
             else:
                 try:
-                    wid = int(args[1])
-                    csource.window_sources[wid]
+                    wid = int(args[0])
+                    window = self._id_to_window[wid]
                     wids = [wid]
                 except:
-                    return respond(10, "cannot find window id %s" % args[1])
-            try:
-                from xpra.server.window_video_source import parse_scaling_value
-                scaling = parse_scaling_value(args[2])
-            except:
-                return respond(11, "invalid scaling value %s" % args[2])
-            for wid in wids:
-                window = self._id_to_window.get(wid)
-                if not window:
-                    continue
-                ws = csource.window_sources.get(wid)
-                if ws:
-                    ws.set_scaling(scaling)
-                    csource.refresh(wid, window, {})
-            return respond(0, "scaling set to %s" % str(scaling))
-        elif command=="name":
-            if len(args)!=2:
-                return argn_err(1)
-            self.session_name = args[1]
-            log.info("changed session name: %s", self.session_name)
-            may_forward_client_command(["name"])
-            return respond(0, "session name set")
+                    return respond(10, "cannot find window id %s" % args[0])
+
+            for csource in sources:
+                for wid in wids:
+                    window = self._id_to_window.get(wid)
+                    if not window:
+                        continue
+                    ws = csource.window_sources.get(wid)
+                    if ws:
+                        ws.set_scaling(scaling)
+                        csource.refresh(wid, window, {})
+            return respond(0, "scaling set to %s on window %s for %s clients" % (str(scaling), args[0], len(sources)))
         elif command=="client":
-            if len(args)<2:
-                return argn_err("at least 2")
-            client_command = args[1:]
-            if client_command[0] not in csource.control_commands:
-                return respond(12, "client does not support control command '%s'" % client_command[0])
+            if len(args)==0:
+                return argn_err("at least 1")
+            client_command = args
+            count = 0
+            for source in sources:
+                if client_command[0] in source.control_commands:
+                    count += 1
+                    source.send_client_command(*client_command)
+                else:
+                    log.warn("client %s does not support client command %s", source, client_command[0])
             csource.send_client_command(*client_command)
-            return respond(0, "client control command '%s' forwarded" % (client_command[0]))
+            return respond(0, "client control command '%s' forwarded to %s clients" % (client_command[0], count))
         else:
             return respond(9, "internal state error: invalid command '%s'", command)
 
diff --git a/src/xpra/server/server_core.py b/src/xpra/server/server_core.py
index c5f36dae6a..c819f82bcd 100644
--- a/src/xpra/server/server_core.py
+++ b/src/xpra/server/server_core.py
@@ -119,6 +119,9 @@ def __init__(self):
         self.compression_level = 1
         self.exit_with_client = False
 
+        #control mode:
+        self.control_commands = ["hello"]
+
         self.init_packet_handlers()
         self.init_aliases()
 
@@ -420,6 +423,11 @@ def _process_hello(self, proto, packet):
                 log.info("processing info request from %s", proto._conn)
                 self.send_hello_info(proto)
                 return
+            command_req = c.strlistget("command_request")
+            if len(command_req)>0:
+                #call from UI thread:
+                self.idle_add(self.handle_command_request, proto, command_req)
+                return
             #continue processing hello packet:
             try:
                 self.hello_oked(proto, packet, c, auth_caps)
@@ -533,6 +541,32 @@ def hello_oked(self, proto, packet, c, auth_caps):
         pass
 
 
+    def control_command_response(self, proto, command, error=0, response=""):
+        log("control_command_response(%s)=%s", command, response)
+        hello = {"command_response"  : (error, response)}
+        proto.send_now(("hello", hello))
+
+    def handle_command_request(self, proto, args):
+        try:
+            assert len(args)>0
+            command = args[0]
+            if command not in self.control_commands:
+                log.warn("invalid command: %s (must be one of: %s)", command, self.control_commands)
+                return self.control_command_response(proto, command, 6, "invalid command")
+            self.do_handle_command_request(proto, command, args[1:])
+        except Exception, e:
+            log.error("error processing command %s", args, exc_info=True)
+            proto.send_now(("hello", {"command_response"  : (127, "error processing command: %s" % e)}))
+
+    def do_handle_command_request(self, proto, command, args):
+        log("handle_command_request(%s, %s, %s)", proto, command, args)
+        if command=="hello":
+            return self.control_command_response(proto, command, 0, "hello")
+        assert command=="help"
+        return self.control_command_response(proto, command, 0,
+                    "control supports: %s" % (", ".join(self.control_commands)))
+
+
     def accept_client(self, proto, c):
         #max packet size from client (the biggest we can get are clipboard packets)
         proto.max_packet_size = 1024*1024  #1MB
diff --git a/src/xpra/server/source.py b/src/xpra/server/source.py
index 8737c1ec18..0e79974ff5 100644
--- a/src/xpra/server/source.py
+++ b/src/xpra/server/source.py
@@ -1362,6 +1362,13 @@ def set_speed(self, speed):
         elog("set_speed(%s) prev_speed=%s, default_encoding_options=%s", speed, prev_speed, self.default_encoding_options)
         self.reconfigure(force_reload=(speed>99 and prev_speed<=99) or (speed<=99 and prev_speed>99))
 
+    def full_quality_refresh(self, wid, window, damage_options):
+        if not self.can_send_window(window):
+            return
+        ws = self.window_sources.get(wid)
+        if ws:
+            ws.full_quality_refresh(window, damage_options)
+
     def refresh(self, wid, window, opts):
         if not self.can_send_window(window):
             return