[Ns-bugs] [Bug 426] TCP: close does not send RST

[code@nsnam.ece.gatech.edu]

http://www.nsnam.org/bugzilla/show_bug.cgi?id=426





--- Comment #6 from Tom Henderson <tomh at tomh.org>  2009-03-30 10:06:43 EDT ---
(From update of attachment 407)
>diff -r 80601ddf444a examples/tcp-large-transfer.cc
>--- a/examples/tcp-large-transfer.cc	Mon Mar 23 00:06:28 2009 -0400
>+++ b/examples/tcp-large-transfer.cc	Mon Mar 30 01:49:51 2009 -0400
>@@ -212,5 +212,5 @@
>         return;
>       }
>   }
>-  localSocket->Close ();
>+  localSocket->ShutdownSend ();
> }

I would not necessarily change the above; Close() is still valid here.  The
program otherwise doesn't call Close() or ShutdownRecv() so it seems
insufficient to just make the above change.

>diff -r 80601ddf444a src/internet-stack/tcp-socket-impl.cc
>--- a/src/internet-stack/tcp-socket-impl.cc	Mon Mar 23 00:06:28 2009 -0400
>+++ b/src/internet-stack/tcp-socket-impl.cc	Mon Mar 30 01:49:51 2009 -0400
>@@ -73,6 +73,7 @@
>     m_closeRequestNotified (false),
>     m_closeOnEmpty (false),
>     m_pendingClose (false),
>+    m_closeCalled (false),
>     m_nextTxSequence (0),
>     m_highTxMark (0),
>     m_highestRxAck (0),
>@@ -112,6 +113,7 @@
>     m_closeRequestNotified (sock.m_closeRequestNotified),
>     m_closeOnEmpty (sock.m_closeOnEmpty),
>     m_pendingClose (sock.m_pendingClose),
>+    m_closeCalled (sock.m_closeCalled),
>     m_nextTxSequence (sock.m_nextTxSequence),
>     m_highTxMark (sock.m_highTxMark),
>     m_highestRxAck (sock.m_highestRxAck),
>@@ -294,6 +296,16 @@
> TcpSocketImpl::ShutdownSend (void)
> {
>   NS_LOG_FUNCTION_NOARGS ();
>+  if (m_pendingData && m_pendingData->Size() != 0)
>+  { // App close with pending data must wait until all data transmitted
>+    m_closeOnEmpty = true;
>+    NS_LOG_LOGIC("Socket " << this << 
>+                 " deferring close, state " << m_state);
>+    return 0;
>+  }
>+  
>+  Actions_t action  = ProcessEvent (APP_CLOSE);
>+  ProcessAction (action);
>   m_shutdownSend = true;
>   return 0;
> }
>@@ -308,18 +320,15 @@
> int
> TcpSocketImpl::Close (void)
> {
>-  NS_LOG_FUNCTION_NOARGS ();

do not delete the log statement

>-  if (m_pendingData && m_pendingData->Size() != 0)
>-    { // App close with pending data must wait until all data transmitted
>-      m_closeOnEmpty = true;
>-      NS_LOG_LOGIC("Socket " << this << 
>-                   " deferring close, state " << m_state);
>-      return 0;
>-    }
>-
>-  Actions_t action  = ProcessEvent (APP_CLOSE);
>-  ProcessAction (action);
>-  ShutdownSend ();
>+  if (GetRxAvailable () > 0)
>+  {
>+    ProcessAction (RST_TX);
>+  }
>+  else 
>+  {
>+    ShutdownSend();
>+  }
>+  m_closeCalled = true;

doesn't ShutdownRecv() need to be called also?  (and fix ShutdownRecv() as you
previously mentioned)

I am not sure the code will honor shutdownRecv; it seems that RecvFrom ignores
this.  After close(), you should not be able to read from the socket.

>   return 0;
> }
> 
>@@ -1125,6 +1134,11 @@
>                 " seq " << tcpHeader.GetSequenceNumber() <<
>                 " ack " << tcpHeader.GetAckNumber() <<
>                 " p.size is " << p->GetSize());
>+  if(m_closeCalled) //send RST accordingly
>+  {
>+    ProcessAction(RST_TX);
>+    return;
>+  }

Do you send RST for each NewRx?  I don't know if these are rate limited in any
way.

>   States_t origState = m_state;
>   if (RxBufferFreeSpace() < p->GetSize()) 
>     { //if not enough room, fragment
>diff -r 80601ddf444a src/internet-stack/tcp-socket-impl.h
>--- a/src/internet-stack/tcp-socket-impl.h	Mon Mar 23 00:06:28 2009 -0400
>+++ b/src/internet-stack/tcp-socket-impl.h	Mon Mar 30 01:49:51 2009 -0400
>@@ -191,6 +191,7 @@
>   bool m_closeRequestNotified;
>   bool m_closeOnEmpty;
>   bool m_pendingClose;
>+  bool m_closeCalled;
> 
>   
>   //sequence info, sender side


-- 
Configure bugmail: http://www.nsnam.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.