PI99580: CICS REGION GOES INTO TIGHT LOOP IN DFHDSKE IN ROUTINE FREE_OPEN_TCBS DUE TO AN S8 TCB POINTING TO ITSELF

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• You are running CICS Transaction server 5.4, using
  IPIC connections.  When running a large volume
  workload, the region becomes unresponsive.
  A dump taken at the time of this hang showed the
  QR TCB was in a loop in module DFHDSKE.  The
  stack for the running task shows it is in the process
  of terminating, when DFHDSKE attempts to free
  TCB's associated with the task. The modules from the
  stack are:
           NAME
           DFHKETA
           DFHDSKE
           RESUME_AWAIT
  .
  It is attempting to free the SSL pool TCBs or S8 TCBs, which
  are located at offset x'170' into the tasks DTA control
  block. The S8 TCB is represented by a DFHDSDS_TCB
  control block, and offset x'98' is field
  OWNED_FWD -Which is a pointer to tcbs of same open mode.
  In the dump, this field points to itself.
  There is a problem with the CISR transaction where it
  references an S8 TCB, that is being used by a different
  task, and incorrectly frees it putting it on the free chain,
  and leads to this loop.
  
  Additional Symptom(s) Search Keyword(s): KIXREVRJL
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED: All CICS users.                              *
  ****************************************************************
  * PROBLEM DESCRIPTION: CICS region becomes unresponsive due    *
  *                      to a loop in DFHDSKE caused by an S8    *
  *                      TCB on the open TCB free chain pointing *
  *                      to itself.                              *
  ****************************************************************
  CICS is processing a heavy workload over an IPIC connection that
  is using SSL.  A mirror task is sending back a large response to
  the partner system.  At the same time a new request arrives on
  the same socket and CISR is resumed to handle it.
  
  The mirror task and CISR need to switch between the QR, SO and
  S8 TCBs to perform the socket I/O.  Due to an error in the TCB
  management within the CICS sockets domain, it is possible for
  the CISR task to end up running on the same S8 TCB as the mirror
  task.  When both tasks finish using the S8 TCB they issue a
  CHANGE_MODE with DISASSOCIATE_TCB(YES).  Each of these cause the
  S8 TCB to be added to the open TCB free chain.  The S8 TCB is
  now on the free chain twice and most likely points to itself.
  
  A loop will occur the next time the open TCB free chain gets
  processed and CICS will appear to be hung.  Formatting the DS
  section of a dump that was taken at the time of the hang will
  cause IPCS to enter a loop.
  

Problem conclusion

• The CICS web and sockets domain code used to process IPIC
  requests has been changed to correctly manage the TCB state when
  multiple tasks use the same socket.
  

Temporary fix

Comments

• ×**** PE18/11/20 FIX IN ERROR. SEE APAR PH05536  FOR DESCRIPTION
  

APAR Information

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

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

  UI58045 PH10004

Modules/Macros

• DFHISCO  DFHISEM  DFHISRR  DFHISSR  DFHMNXM  DFHSOCK  DFHWBCL
  DFHWBSO  DFHWBSV  DFHWBSVT DFHWBXM
  

Fix information

• Fixed component name

  CICS TS Z/OS V5

• Fixed component ID

  5655Y0400

Applicable component levels

• R100 PSY UI58045

     UP18/08/25 P F808

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.