From 8a8fae0dbbdead7d3acd87334fd8cd7289dc1022 Mon Sep 17 00:00:00 2001
From: moret <moret@0c269be4-1314-0410-8aa9-9f06e86f4224>
Date: Fri, 25 Jul 2008 14:05:30 +0000
Subject: [PATCH] Switch to better network error management in realtime process

git-svn-id: http://subversion.jackaudio.org/jack/jack2/trunk/jackmp@2741 0c269be4-1314-0410-8aa9-9f06e86f4224
---
 common/JackNetDriver.cpp  |  8 +++++---
 common/JackNetManager.cpp | 21 ++++++++++++++-------
 common/JackNetManager.h   |  2 +-
 3 files changed, 20 insertions(+), 11 deletions(-)

diff --git a/common/JackNetDriver.cpp b/common/JackNetDriver.cpp
index ecfff43d..62ae13d1 100644
--- a/common/JackNetDriver.cpp
+++ b/common/JackNetDriver.cpp
@@ -482,7 +482,7 @@ namespace Jack
         if ( rx_bytes == SOCKET_ERROR )
         {
             net_error_t error = fSocket.GetError();
-            //just tell there is no data and return 0 instead of SOCKET_ERROR
+            //no data isn't really an error in realtime processing, so just return 0
             if ( error == NET_NO_DATA )
             {
                 jack_error ( "No data, is the master still running ?" );
@@ -492,7 +492,7 @@ namespace Jack
             else if ( error == NET_CONN_ERROR )
                 throw JackDriverException ( "Connection lost." );
             else
-                jack_error ( "Error in receive : %s", StrError ( NET_ERROR_CODE ) );
+                jack_error ( "Fatal error in receive : %s", StrError ( NET_ERROR_CODE ) );
         }
         return rx_bytes;
     }
@@ -508,7 +508,7 @@ namespace Jack
             if ( error == NET_CONN_ERROR )
                 throw JackDriverException ( "Connection lost." );
             else
-                jack_error ( "Error in send : %s", StrError ( NET_ERROR_CODE ) );
+                jack_error ( "Fatal error in send : %s", StrError ( NET_ERROR_CODE ) );
         }
         return tx_bytes;
     }
@@ -621,6 +621,8 @@ namespace Jack
         memset ( fTxData, 0, fPayloadSize );
         SetSyncPacket();
         tx_bytes = Send ( fParams.fMtu, 0 );
+        if ( tx_bytes == SOCKET_ERROR )
+            return 0;
 
 #ifdef JACK_MONITOR
         fMeasure[fMeasureId++] = ( ( float ) ( GetMicroSeconds() - JackDriver::fBeginDateUst ) / ( float ) fEngineControl->fPeriodUsecs ) * 100.f;
diff --git a/common/JackNetManager.cpp b/common/JackNetManager.cpp
index 920e5ac8..07cfbe74 100644
--- a/common/JackNetManager.cpp
+++ b/common/JackNetManager.cpp
@@ -345,6 +345,8 @@ namespace Jack
             }
             else
                 jack_error ( "Error in send : %s", StrError ( NET_ERROR_CODE ) );
+            //we can't stop the process here, the slave wouldn't see what happens, just return 0
+            return 0;
         }
         return tx_bytes;
     }
@@ -357,24 +359,27 @@ namespace Jack
             net_error_t error = fSocket.GetError();
             if ( error == NET_NO_DATA )
             {
-                //too much receive failure, exit
+                //too much receive failure, react
                 if ( ++fNetJumpCnt == 100 )
                 {
                     jack_error ( "No data from %s...", fParams.fName );
                     fNetJumpCnt = 0;
                 }
-                //we don't want the process to stop here, so just return 0
-                return 0;
             }
             else if ( error == NET_CONN_ERROR )
             {
                 //fatal connection issue, exit
                 jack_error ( "'%s' : %s, network connection with '%s' broken, exiting.",
-                             fParams.fName, StrError ( NET_ERROR_CODE ), fParams.fSlaveNetName );
+                             fParams.fName, StrError ( NET_ERROR_CODE ), fParams.fSlaveNetName );   
+                //ask to the manager to properly remove the master
                 Exit();
             }
             else
                 jack_error ( "Error in receive : %s", StrError ( NET_ERROR_CODE ) );
+            //if we stop the process now, the socket will be closed and deleted to prematurely
+            //so the slave won't see the connection is down - just return 0, thus the master's process don't stop now
+            //by this way, the slave will have enough time to get the ICMP "connection refused" msg
+            return 0;
         }
         return rx_bytes;
     }
@@ -690,8 +695,8 @@ namespace Jack
                         jack_info ( "Waiting for a slave..." );
                         break;
                     case KILL_MASTER:
-                        KillMaster ( &params );
-                        jack_info ( "Waiting for a slave..." );
+                        if ( KillMaster ( &params ) )
+                            jack_info ( "Waiting for a slave..." );
                         break;
                     default:
                         break;
@@ -751,7 +756,7 @@ namespace Jack
         return it;
     }
 
-    void JackNetMasterManager::KillMaster ( session_params_t* params )
+    int JackNetMasterManager::KillMaster ( session_params_t* params )
     {
         jack_log ( "JackNetMasterManager::KillMaster, ID %u.", params->fID );
         master_list_it_t master = FindMaster ( params->fID );
@@ -759,7 +764,9 @@ namespace Jack
         {
             fMasterList.erase ( master );
             delete *master;
+            return 1;
         }
+        return 0;
     }
 }//namespace
 
diff --git a/common/JackNetManager.h b/common/JackNetManager.h
index 077e8c06..117cb3d7 100644
--- a/common/JackNetManager.h
+++ b/common/JackNetManager.h
@@ -127,7 +127,7 @@ namespace Jack
             void Run();
             JackNetMaster* MasterInit ( session_params_t& params );
             master_list_it_t FindMaster ( uint32_t client_id );
-            void KillMaster ( session_params_t* params );
+            int KillMaster ( session_params_t* params );
             void SetSlaveName ( session_params_t& params );
 
             int SyncCallback ( jack_transport_state_t state, jack_position_t* pos );