OA40383: GIVESOCKET/TAKESOCKET CALLS IS NOT BEING CLEANED UP WHEN THE NOBODY TAKES IT

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• It appears that the underlying problem is a control structure
  associated with the REXX sockets use of GIVESOCKET/TAKESOCKET
  calls is not being cleaned up when the nobody takes it (and the
  giving task goes away).  I will be discussing this aspect with
  the developers, but I doubt there will be anything done about
  this in the current release.
  
  There are two suggestions I would make that would help avoid
  this scenario:
  
   - I can see that the giving task does a SELECT waiting on the
  EXCEPTION for the given socket.  I assume that when the SELECT
  times out without the exception occurring that this exec has an
  error path here (possibly producing a message).  If so, add
  performing a TAKESOCKET to retrieve what was given, then CLOSE
  both copies of that socket    (the original and the result of
  the TAKESOCKET).  This is the recommended practice when using
  the GIVESOCKET call, anyway.
  
   - Use different subtask IDs on every INITIALIZE call.  I
  already  mentioned this recommendation so that each active REXX
  environment has a unique identifier.  But if the main task could
  calculate a 'random' name (like including the time as a part of
  it), that would  significantly reduce the possibility of a
  stranded entry being a match.
  

Local fix

• n/a
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: All Tivoli Event Pump for z/OS users.        *
  ****************************************************************
  * PROBLEM DESCRIPTION: Tivoli Event Pump for z/OS hangs after  *
  *                      PURGE operator command and can be       *
  *                      stopped only by issuing CANCEL command. *
  *                      System log contains GTM5428E message    *
  *                      indicating TAKESOCKET issue for         *
  *                      internal Tivoli Event Pump for z/OS     *
  *                      HTTP server.                            *
  ****************************************************************
  * RECOMMENDATION: Apply the PTF.                               *
  ****************************************************************
  Tivoli Event Pump for z/OS generates the following error message
  in system log:
  
      GTM5428E HTTP SERVER : TAKESOCKET FOR SOCKET 2 FAILED WITH
      RC = 22, RS = EINVAL Invalid argument
  
  Any attempt to stop the Tivoli Event Pump for z/OS address space
  leads to hang of the application. CANCEL command can be used to
  stop hanged address space. The following abends can occure
  during the processing of the CANCEL command:
  
  IEA995I SYMPTOM DUMP OUTPUT  176
  SYSTEM COMPLETION CODE=0C4  REASON CODE=00000004
   TIME=18.04.12  SEQ=08017  CPU=0000  ASID=00FC
   PSW AT TIME OF ERROR  478D0000   C0147286  ILC 2  INTC 04
     ACTIVE LOAD MODULE          ADDRESS=40146F18  OFFSET=0000036E
     NAME=GTMAOP16
     DATA AT PSW  40147280 - 181F18E3  0E0E18F4  12FF4780
     AR/GR 0: 00000000/4200D000   1: 00000000/FFFF47BE
           2: 01FF0028/FFFFC65E   3: 00000000/09C03E58
           4: 00000000/00005840   5: 00000000/09C004B6
           6: 00000000/00000000   7: 00000000/41C37E58
           8: 00000000/401578F0   9: 00000000/40221CE0
           A: 00000000/401574B8   B: 00000000/C0146F18
           C: 00000000/40147F18   D: 00000000/41C37EAC
           E: 00000000/09C0BCF8   F: 00000C05/FFFF47BE
   END OF SYMPTOM DUMP
  BPXP018I THREAD 4062A00000000002, IN PROCESS 67633175, ENDED 184
  WITHOUT BEING UNDUBBED WITH COMPLETION CODE 040C4000
  , AND REASON CODE 00000004.
  BPXP018I THREAD 40656A000000000C, IN PROCESS 67633175, ENDED 210
  WITHOUT BEING UNDUBBED WITH COMPLETION CODE 040C4000
  , AND REASON CODE 00000004.
  IEA995I SYMPTOM DUMP OUTPUT  213
  SYSTEM COMPLETION CODE=0C4  REASON CODE=00000011
   TIME=18.04.14  SEQ=08068  CPU=0000  ASID=00FC
   PSW AT TIME OF ERROR  478D0000   89C31F56  ILC 4  INTC 11
     NO ACTIVE MODULE FOUND
     NAME=UNKNOWN
     DATA AT PSW  09C31F50 - 30074770  81665550  40004740
     AR/GR 0: 008E5DB4/42021284   1: 00000000/001015B0
           2: 00000000/8B000000   3: 00000000/4200B000
           4: 00000000/04F07170   5: 00000000/424DB421
           6: 00000000/41F23000   7: 00000000/00000001
           8: 00000000/89C31EA6   9: 00000000/424DB420
           A: 00000000/42023040   B: 00000000/00000200
           C: 00000000/09C2F8B0   D: 00000000/42021598
           E: 00000000/89C2FA56   F: 00000000/42023040
   END OF SYMPTOM DUMP
  IEA995I SYMPTOM DUMP OUTPUT  214
  SYSTEM COMPLETION CODE=0C4  REASON CODE=00000004
   TIME=18.04.14  SEQ=08068  CPU=0000  ASID=00FC
   PSW AT TIME OF ERROR  478D1000   89BFAC8E  ILC 4  INTC 04
     NO ACTIVE MODULE FOUND
     NAME=UNKNOWN
     DATA AT PSW  09BFAC88 - 06904490  C86A18E5  5EE0B198
     GR 0: 42022C90   1: 422E16E4
        2: 42023CB8   3: 00000001
        4: 00000000   5: 000000B4
        6: 00000004   7: 00000000
        8: 00000000   9: 0000003B
        A: 0000003C   B: 422E1560
        C: 09BFA460   D: 422E1560
        E: 42023CB4   F: 09BFA460
   END OF SYMPTOM DUMP
  BPXP018I THREAD 4062C2000000000A, IN PROCESS 67633175, ENDED 215
  WITHOUT BEING UNDUBBED WITH COMPLETION CODE 940C4000
  , AND REASON CODE 00000004.
  
  There is a problem with a control structure associated with the
  REXX sockets use of GIVESOCKET/TAKESOCKET calls which is not
  being cleaned up when the nobody takes it. HTTP server main task
  does not perform TAKESOCKET cleanup call when child tasks don't
  take given socket. HTTP server always uses the same subtask ID
  in socket set INITIALIZE call for a  different instances of
  child tasks. This can lead to incorrect selection of given
  sockets from a socket sets assotiated with Tivoli Event Pump
  for z/OS address space.
  

Problem conclusion

• TAKESOCKET call was added to parent task of HTTP server for
  cases when child tasks don't perform GIVESOCKET for a passed
  socket. This condition is considered as an error state and leads
  to stop HTTP server. Every socket set in a child task is
  initialized with unique subtask ID. This significantly reduce
  the possibility of issues reported in this APAR.
  

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:

  UA66930

Modules/Macros

• GTMIPCRH GTMIPSRV
  

Fix information

• Fixed component name

  EVENT PUMP FOR

• Fixed component ID

  5698B3400

Applicable component levels

• R422 PSY UA66930

     UP13/02/27 P F302

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.