PM66484: DFHKERRQ AMODE 64 PROBLEM CAUSES 0C4S IN VARIOUS MODULES

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• DFHPG0001 AN ABEND (CODE 0C4/AKEA) HAS OCCURRED AT
  OFFSET X'FFFF' IN MODULE DFHPGLE
  .
  There is an AMODE 64 problem in DFHKERRQ .
  .
  DFHPGLE is in control and gets back a purged response on a
  DFHLMLM Lock call and so the DFHLMLMM macro turns that into a
  DFHKERN FUNCTION(RECOVERY_REQUEST) which will invoke DFHKERRQ
  in 64-bit mode using these instructions:
  .
           AGHI  15,1                Set Amode 64.
           BASSM 14,15               Call Program.
  .
  Those instruction leave the low order 1 bit on in R15.
  .
  DFHKERRQ doesn't deal with that and uses R15 to set the R3 base
  register, still with that low-order 1 bit on. Then when DFHKERRQ
  uses R3 to load another address to branch to, R3 is off by 1 and
  so a bad address is loaded causing the 0C4.
  .
  Additional Symptoms / Search Keywords:
  KIXREVNDB S0C4 abend0C4 abendS0C4
  Problem happens similarly when CICS is at MXT and the CICS
  region has run out of task numbers. DFHXMAT issues a
  DFHKERN FUNCTION(RECOVERY_REQUEST) call which turns the low-
  order 1 bit on in Reg15, then branches to DFHKERRQ. The 0C4
  happens due to the invalid branch from DFHKERRQ but the
  resulting message will be:
  .
  DFHXM0001 An abend (code 0C4/AKEA) has occurred at offset
   X'FFFF' in module DFHXMAT.
  .
  The BEAR typically identifies a branch from offset x'18E':
  .
  Breaking Event Address: 00000000_xxxxxxxx - offset 0000018E in
  module DFHKERRQ
  .
  Problem also happens when tasks in LMQUEUE suspends are
  purged. In this case, the needed lock was the NQLOCK
  requested by DFHNQED. After the purge, DFHNQED issues the
  DFHKERN function(RECOVERY_REQUEST) which causes the same
  0C4 in DFHKERRQ.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED: All                                          *
  ****************************************************************
  * PROBLEM DESCRIPTION: DFHPG0001 An abend 0c4 occurred at      *
  *                      offset X'FFFF' in module DFHPGLE.       *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  DFHPGLE is in control and gets back a purged response on a
  DFHLMLMM Lock call and DFHLMLMW turns that into a
  DFHKERN FUNCTION(RECOVERY_REQUEST) which will invoke DFHKERRQ
  in AMODE 64, setting the low order bit on in the branch
  address in R15.
    DFHKERRQ uses R15 to set R3 as its base register. This still
  has the low-order 1 bit on. When DFHKERRQ later uses R3 to
  load another address to branch to, R3 is off by 1 and so a
  bad address is loaded, causing the S0C4.
  KEYWORDS: 0c4 s0c4 abend0c4 abends0c4 MSGDFHPG0001 PG0001
  

Problem conclusion

• DFHKERRQ has been changed to set its base register correctly.
  

Temporary fix

• FIX AVAILABLE BY PTF ONLY
  

Comments

APAR Information

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

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

  UK79865

Modules/Macros

•    DFHKERRQ
  

Fix information

• Fixed component name

  CICS TS Z/OS V4

• Fixed component ID

  5655S9700

Applicable component levels

• R700 PSY UK79865

     UP12/07/11 P F207

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.