Make sure that a TCP socket is closed only once

[htibosch]

Description

If during the SYN phase, a sockets goes to the eCLOSED state, it must be closed/released. But at that moment, it may happen that the application does a successful call to FreeRTOS_accept(), and gets the ownership of the socket.
The IP-task in a meanwhile is planning to close the socket as well. See the function vSocketCloseNextTime()

Here is the proposed change:

     if( ( eTCPState == eCLOSED ) ||
         ( eTCPState == eCLOSE_WAIT ) )
     {
         /* Socket goes to status eCLOSED because of a RST.
          * When nobody owns the socket yet, delete it. */
+        vTaskSuspendAll();
         {
             if( ( pxSocket->u.xTCP.bits.bPassQueued != pdFALSE_UNSIGNED ) ||
                 ( pxSocket->u.xTCP.bits.bPassAccept != pdFALSE_UNSIGNED ) )
             {
+                if( pxSocket->u.xTCP.bits.bReuseSocket == pdFALSE_UNSIGNED )
+                {
+                    /* Make sure the the ownership will not be transferred. */
+                    pxSocket->u.xTCP.bits.bPassQueued = pdFALSE_UNSIGNED;
+                    pxSocket->u.xTCP.bits.bPassAccept = pdFALSE_UNSIGNED;
+                }
+                xTaskResumeAll();
     
                 FreeRTOS_debug_printf( ( "vTCPStateChange: Closing socket\n" ) );
     
                 if( pxSocket->u.xTCP.bits.bReuseSocket == pdFALSE_UNSIGNED )
                 {
                     /* Plan to close the socket. */
                     vSocketCloseNextTime( pxSocket );
                 }
             }
+            else
+            {
+                xTaskResumeAll();
+            }
         }
     }

The application can "take" the socket until vTaskSuspendAll() is called. After that the IP-task will "take the ownership" by clearing both bPass bits.

This PR was tested in a set-up where the DUT received a TCP reset packet during the SYN phase.

Thanks Paul Helter, who first discovered and described this problem in PR #571. Mentioned PR was not merged because I missed a proper diagnosis of the cause. Only last week we saw exactly what happened, and could repair it.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[amazonKamath]

Closing this PR since PR #707 took care of fixing unit tests on top of this PR and got merged.