PQ06771: TIME-WAIT, CLOSING, CLOSE-WAIT, TCP/IP TIMER FUNCTION STOPS

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• The TCP/IP timer function stops working when EZAT2NTF issues a
  STIMER to wait for SUSPEND to complete.  The loss of the timer
  function causes connections in time-wait, closing, close-wait to
  remain in these states indefinitely.
  VERIFICATION STEPS:
  
  The customer must have at least 1 HPNS enabled application
  running to encounter this problem.
  A Netstat will show connections in a x-wait condition.
  From a dump of TCP/IP:
  1) issue a find for TCTIMER and find x'58F0F008' prior to
     this label.  Equate this location to TCTIMER.
  2) locate TCTIMER + 840??
  3) this is the array of timer ACBs
  4) the first ACB is at offset +8 in the array
  5) the stck value at +24 is when the ACB should be removed from
  the array
  6) this ACB time should be lower than the system time
  KEYWORDS:
  TIME CLOSE WAIT TIMEWAIT CLOSEWAIT
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED: All users of TCP/IP for MVS V3R2 with at     *
  *                 least one HPNS application.                  *
  ****************************************************************
  * PROBLEM DESCRIPTION: TCP/IP stops processing TimerQ elements *
  *                      resulting in FIN-WAIT, TIME-WAIT,       *
  *                      Closing, and CLOSE-WAIT states that     *
  *                      are never completed.                    *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  As part of the TCP/IP Rel. 3.2 performance improvements, all
  timer elements are processed in heap storage with interrupts
  generated only when the top (oldest) entry in the heap changes.
  On large multi-processor systems, the TCP/IP timer code did
  not completely synchronize multiple tasks (STIMER Cancels,
  STIMER exits, and IRBs) executing asynchronously causing timer
  processing to discard certain interrupts prematurely.  Since
  the timer heap elements represent work to be processed in the
  future, once the top entry was not processed and became older
  than the current time, no other work would ever change the top
  heap entry causing the timer processing to stop functioning.
  

Problem conclusion

• The TCP/IP timer queue processing has been amended to handle
  multiple asynchronous tasks on multi-processing platforms.
  
  * Cross Reference between External and Internal Names
  EZAAE00C (CMCSRTNS)  EZABB03D (TCSCHED )  EZABB03I (TCTIMER )
  EZACA002 (MVPTIINT)  EZACA01T (MVPIOINT)  EZACA02D (MVPTIMER)
  EZACA02K (MVPUTIL )
  

Temporary fix

• *********
  * HIPER *
  *********
  

Comments

APAR Information

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

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

  UQ11864 PQ12945

Modules/Macros

• EZAAE00C EZABB03D EZABB03I EZACA002 EZACA01T
  EZACA02D EZACA02K
  

Fix information

• Fixed component name

  TCP/IP V3 MVS

• Fixed component ID

  5655HAL00

Applicable component levels

• R320 PSY UQ11864

     UP98/01/08 P F801 Ž

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.