PM90215: DEVELOPMENT APAR

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• Development APAR
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * All user of IBM z/OS Communications Server for z/OS V2R1     *
  * E2827/K                                                      *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * 15483: AB/S0C4 EZBPTSRC +0710 TNPRC923 Running TN3270        *
  * connections over Links PM87339                               *
  * 15542: R130515O AB/S00C4 0011 EZBTCSND(HIP6210               *
  * 13.071)+0054C8                                               *
  * 15554: R130515O AB/S00C4 0004 EZBSCINB(BOOTSNT               *
  * 13.120)+0061A4                                               *
  * 15455: 5013 internal error during smc connection setup       *
  * 15500: ZZ020245 zOS 2.1 CommServer dumps EZBTIINI interface  *
  * workloads after APARs installed                              *
  * 15175: CLC decline message going to appl (JIST) when get     *
  * decline 5106                                                 *
  * 14608: Client and Distributor on the same stack, not using   *
  * new links when it should be eligible                         *
  * 15502: ZZ020253 TCPIP Dump (ITKRA -- Lock Release Error)     *
  * after VTAM recycle                                           *
  * 15599: AB/S00C4 EZBSCUTL(BOOTSNT 13.137)+001980              *
  * 15613: IESP problem: Netstat DEvlinks/-d for VIPA6 shows the *
  * associated interface                                         *
  * 15456: AB/S04C5 2706 EZBSRLLC(BOOTSNT 13.122)+000C94         *
  * 15563: ZZ020282 Netstat command to TCPIP stack hung z/OS     *
  * 15610: ZZ020352 AB/S04C5 2506 EZBITSTO +306A REGS/0B5DE      *
  * FFFFF                                                        *
  * 15608: ZZ020334 MP65 GA2 ROCE - AB/S04C5 402 EZBTIINI        *
  * EZBSCINB EZBITDRC                                            *
  * 15210: EZBSRUTL(BOOTSNT 13.118)+000BF8 S4C5/74300            *
  * 15644: AB/S00C4 0004 EZBSCINB(BOOTSNT 13.141)+00084A         *
  * 15335: ABEND=S0DC2,REASON=55006C20                           *
  * 15673: ITKRA -- Lock Release Error after start/stop          *
  * workloads                                                    *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * Apply Fix                                                    *
  ****************************************************************
  (15175) When a DM_Act_Req failure is detected on the Client side
  of a link then it is attempting to allow the pending TCP
  connections to fall back to IP. If both the Client and Server
  endpoints of this connection both fail on the DM_Act_Req then
  this can allow a CMC Decline to flow after a connection has
  fallen back to IP. This causes the CMC Decline to be treated as
  real application data.
  (15542) A TCP connection over a link may still use the TCP
  connection for KEEPALIVE probes. For data
  still to be sent the tcb_Send_Qlen field will reflect the amount
  to send but the tcb_Send_Q@ will
  be null. The data is queued in a control block and not the TCB.
  This causes a problem
  when replying to a KEEPALIVE probe. The stack attempts to send
  what it thinks is queued data from
  the tcb_send_Q@ but the address is zero.
  (15554) The final rendezvous message (CONFIRM) has not been
  acknowledged by
  the peer when queued application data is sent over the link.
  tcb_recv_qlen is
  greater than zero after sending all queued data. An incorrect
  address for the queued
  staging buffer is then used.
  (15455) The stack surfaces reason code 5013 as an "internal
  error".  This reason code actually means that the first attempt
  to send data over the adapter connection encountered a failure
  (which often means a switch configuration error).
  (15599) Referencing the TCB was causing an addressing exception
  at offset x'1980' in module EZBSCUTL.
  (15456) The stack incorrectly serializes the issing of Test Link
  LLC messages on a link. This could cause a 4C5 abend in EZBSRLLC
  trying to set the Test Link timer.
  (15613) Netstat issues the 'Associated ......' message for the
  interface under wrong conditions.
  Netstat relied on the mpaChpidType only be set to the proper
  value for a interface type is IPAQENET or IPAQENET6. But from
  the Netstat trace can see it is not true. It could be a value
  from previous entry. The mpaChpidType is valid only when the
  interface type is IPAQENET or IPAQENET6, Netstat needs to check
  the interface type is IPAQENET or IPAQENET6 first, and then
  check the mpaChpidType.
  (15664) Segments may be presented to the receiving stack out of
  order due to multi-threaded processing. This works fine
  until the receiving stack receives a close for the connection
  and frees the control block. Any subsequent out-of-order
  segments can cause an abend.
  (15210) When there are a large number of Confirm Rkey LLC
  signals being exchanged along with other dependent LLC signals
  there are timing scenarios where we can trigger a 4C5 Abend
  because the storage has not been registered to all active
  links. Many of these Abends were reflected in EZBSRUTL at offset
  x'B904' but could be reflected at other offsets within EZBSRUTL.
  (14608) Sysplex distributed connections fail to utilize links
  when the client and distributor are on the same TCP stack.
  The client does not use a enabled OSD to send
  the SYN request to the distributing stack and therefore
  does not send the TCP option with the SYN request.
  In this case, the connection uses TCP/IP and not link.
  (15500) The stack does not properly handle ECSA shortages during
  negotiation which can lead to storage overlays and abends.
  (15502) f a VIPARANGE DVIPA is created when VTAM is down,
  return from error leg processing does not UNLOCK the XCF lock.
  
  Results in Lock release error abend.
  
  (15563) EZBSCINB is looping calling EZBSROUT infinitely and
  EZBSROUT is returning a negative RC.
  (15608) This is caused by a small timing window in link setup.
  When a new  link is being established
  it is possible for one side to get notified of link
  establishment and then send data followed by a close to
  the peer. This can occur, in rare cases, before the peer has
  received notification of link establishment.
  This leads to the unexpected abend.
  (15610) The status field for a connection is not being reset
  when the connection ends after the local application issues a
  shutdown(). This causes the stack to attempt to reference
  information that has been freed and causes an abend.
  (15483) When Packet trace is active and a Delete Rkey LLC signal
  is transmitted (buffer contraction) then no valid PIF address is
  passed. This cause the packet trace routine EZBPTSRC to Abend
  with an 0C4 when trying to reference fields in the PIF.
  (15335) The stack incorrectly serializes the deactivation of an
  interface with the deregistration of  staging buffers. This can
  lead to an error when deactivating an interface or deregistering
  staging buffers.
  (15673) The stack does not release the TCB lock when old
  information arrives on a connection that has already closed.
  This leads to the EZBITC13 FFST probe hitting.
  

Problem conclusion

• (15175) Change EZBSRUTL to terminate the TCP connections that
  are awaiting activation on the client side when a failure is
  detected on the DM_Act_Req. Modifications were also made to
  EZBSCUTL (EZBSCSRT subroutine) to correctly handle multiple TCP
  connections that need to be terminated.   Small change to
  ISTSRRRV to handle case of no VLANs being supported (adapter
  simulation code).
  (15542) Before attempting to send data over the TCP connection a
  check will be made to ensure the tcb_Send_Q@
  is not zero. If so, it will be treated as if tcp_send_Qlen is
  zero.
  (15554) A check has been added to ensure a valid staging buffer
  address is used before attempting
  to send data.
  (15455) Enhance the Netstat, NMI, SMF, and packet trace
  externals to surface reason code 5013 as an "connectivity
  failure" and document this in the books.
  (15599) Change EZBSCUTL to correct basing of the TCB . Routine
  EZBSCSRT (within EZBSCUTL) has the TCB based on the pointer TCB@
  but this pointer was not being set.
  (15456) The Test Link code is changed to get the lock exclusive
  to properly serialize the Test Link messages.
  (15613) Add code to Netstat to correct the conditions for
  displaying the associated interface.
  (15664) Change EZBSCINB to handle out-of-order segments by
  ensuring the CB has not been freed before
  processing the interrupt.
  (15210) Change EZBSRUTL, EZBSRLLC, and EZBSRUT2  to synchronize
  the exchange of the Confirm Rkey (CRK) LLC command with the
  registration of the storage to IOS' and currently utilized
  interfaces.
  (14608) The distributing stack will now attempt to use an link
  for this particular connection before falling back
  to TCP/IP.
  (15500) Changed to check for errors during negotiation and
  properly handle them.
  (15502) EZBXFDYN
  Unlock of XCF lock added before returning from error leg.
  EZBXFUDV
  Changed to not call EZBXFDYN if Bind call is to define a DVIPA
  and VTAM is down, instead it will fail immediately.
  
  (15563) Change EZBSCINB to terminate the send loop whenever
  EZBSROUT returns a negative RC.
  (15608) Allow for this timing window by setting the state of the
  control block to *_ApplCloseWait1
  if a close is received before the link enablement completes.
  This will allow EZBTCFRD to check for
  data and process the not active state.
  EZBSCUTL is also modified to check the state of the CB and not
  set it to active when the
  link enablement finally completes for this case.
  (15610) The shutdown(0 code was not resetting the tcb_smcstatus
  field to zero on connection termination. This caused ezbtcutl to
  try to purge the send queue for a n CCB which had already been
  freed. ezbtcfsh is changed to reset the status field.
  (15483) Change EZBSRLLC to correctly pass the PIF address for
  the DRK (Delete Rkey) LLC signal being transmitted to the remote
  peer.
  (15335) If a CM_DEACT is in progress when we attempt to contract
  the number of staging buffers, we will skip contraction.
  (15673) Release the TCB lock.
  

Temporary fix

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

• APAR is sysrouted TO one or more of the following:

  UK95117 UK95118

Modules/Macros

• EZBCTFME EZASMF   EZBSRUTL EZBTCFWR EZBTCSYN EZBTCFCN EZBIFIUM
  EZBT6SND EZBTLFWR EZBXFDYN EZBIFIUT EZBNMRHA EZBTCPCN EZASMF77
  EZBSCINB EZBNMRHC EZBPTDAT EZBSRUT2 EZBXFUDV EZACDNE0 EZBITSCU
  EZBTCSND EZBSCUTL EZBT6FWR EZACDNE6 EZBTSFWR EZBITSTO EZBSRLLC
  EZBT6LWR EZBIFSTC EZBTCFSH
  

Fix information

• Fixed component name

  TCP/IP V3 MVS

• Fixed component ID

  5655HAL00

Applicable component levels

• R210 PSY UK95117

     UP13/07/03 P F307

Fix is available

• Select the PTF appropriate for your component level. You will be required to sign in. Distribution on physical media is not available in all countries.