VM65971: UNRESPONSIVE MASTER PROCESSOR LOOPING BUT NO ABEND

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• Unresponsive processor detection runs on the Master so it can't
  detect a looping Master processor. The system hangs as a result.
  

Local fix

• If a system hang is observed, use SYSTEM RESTART to obtain an
  ABENDSVC002 dump and restart the system rather than SNAPDUMP.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: All Users of z/VM                            *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  ****************************************************************
  * RECOMMENDATION: APPLY PTF                                    *
  ****************************************************************
  The master processor becomes unresponsive which causes the
  entire system to become unresponsive and no longer do
  productive work. The most common cause of an unresponsive
  Master processor is that it is looping continuously so the
  Master processor will show very high utilization. The system
  will appear hung and unlikely to be able to process commands.
  
  Further evidence of this problem is the lack of MSHCPMPG9152E
  messages to report unresponsive non-Master processors. The
  existing unresponsive processor detection function runs on the
  master processor so it is unable to detect that the master
  processor has become unresponsive.
  
  If the OPERATOR continues to receive message MSHCPMPG9152E,
  then the Master processor is responsive.  The text of the
  MSHCPMPG9152E message indicates which non-Master processor
  is unresponsive.
  
  The problem that this APAR addresses can only occur if there
  is another prior error that causes the Master processor to
  become unresponsive. A variety of conditions could cause this
  but typically the processor is in a spin loop in CP code so
  that it never returns to the dispatcher to do other work and
  never opens interrupt windows.
  

Problem conclusion

• A mechanism for non-Master processors to detect an unresponsive
  Master processor has been added to areas of the system likely
  to execute when the Master is looping. If an unresponsive
  Master processor is detected, an ABENDMCW002 dump is generated.
  To provide a way to disable the detection of unresponsive
  processors on second level test systems, a new operand on the
  FEATURES statement in the system configuration file is added:
  UNRESPONSIVE_PROCESSOR_DETECTION.
  
  All processors repeatedly mark themselves as responsive in
  PFXDETUP which is tested by both the existing and new
  unresponsive processor checks.  The checking mechanism sets
  PFXDETUP indicating it has been tested for responsiveness.
  Processors that are operating normally reset PFXDETUP to a
  state indicating they are responsive in the Dispatcher.  A
  looping processor will not execute the frequent code path that
  resets PFXDETUP so they will appear unresponsive when the
  unresponsive processor checking code checks PFXDETUP again.
  When PFXDETUP is not reset by the processor for longer than
  allowed, it is considered unresponsive and an ABENDMCW002 dump
  is generated.
  
  HCPDSP checks whether the Master is responsive on entry to
  wait state.  HCPDSP, HCPSXL and HCPSYN check responsiveness of
  the Master in code that checks for processor malfunctions
  while looping attempting to acquire a spinlock.
  
  This APAR provides FFDC (first-failure data capture) and
  availability improvements.  The ABENDMCW002 dump for the
  unresponsive Master processor condition provides diagnostic
  data used to determine the reason the Master processor was
  unresponsive.  The system is re-IPLed as soon as the dump
  generation completes which greatly reduces the length of time
  the system is unavailable compared to the hang condition that
  required manual intervention for recovery actions.
  
  The detection is only applicable in a multiprocessor partition
  when there is at least one alternate processor in addition to
  the master processor.
  
  The documentation for system configuration FEATURES statement
  in the following books will be changed to include information
  about the new UNRESPONSIVE_PROCESSOR_DETECTION operand.
  - SC24-6178-14 z/VM V640 "CP Planning and Administration".
  - SC24-6271-03 z/VM V710 "CP Planning and Administration".
  
     FEATURES DISABLE UNRESPONSIVE_PROCESSOR_DETECTION
  
       specifies that detection of unresponsive processors will
       not occur on second level systems.  A common cause of an
       unresponsive processor is that it is looping continuously
       and is no longer doing productive work.  If the master
       processor is unresponsive, the system will appear hung and
       unlikely to be able to process commands.  Disabling
       unresponsive processor detection prevents the CP from
       recognizing an unresponsive master or non-master processor
       and from initiating normal error recovery, which could be
       to restart the unresponsive processor or could be to abend
       the system.
  
       This option is intended for use in diagnostic situations
       when running as a second level system where CP might be
       stopped for long periods of time when using the CP TRACE
       command or similar facilities to debug code within the CP
       nucleus or a CPXLOADed nucleus extension.  Virtual
       processors that aren't being traced could run enough to
       recognize the traced processor is being unresponsive.
       Specifying DISABLE UNRESPONSIVE_PROCESSOR_DETECTION only
       affects CP when it is running second level; when running
       first, this specification is ignored.
  
     FEATURES ENABLE UNRESPONSIVE_PROCESSOR_DETECTION
  
       specifies that detection of unresponsive processors
       occurs.  A common cause of an unresponsive processor is
       that it is looping continuously and is no longer doing
       productive work.  If the master processor is unresponsive,
       the system will appear hung and unlikely to be able to
       process commands.  Detection of unresponsive processors
       is the default and allows CP to initiate normal error
       recovery, which could be to restart the unresponsive
       processor or could be to abend the system.
  
  *** Warning *** When enabling or disabling unresponsive
  processor detection in the z/VM system configuration file, it
  is recommended to add a separate FEATURES statement on a new
  line rather than include UNRESPONSIVE_PROCESSOR_DETECTION on an
  existing FEATURES line.  This will avoid issues with existing
  FEATURES statements when IPLing CPLOAD MODULES which do not
  include this Apar.
  
  The documentation for Abend MCW002 in the following books
  will be changed.
    GC24-6177-12 z/VM V640 "CP Messages and Codes" book.
    GC24-6270-03 z/VM V710 "CP Messages and Codes" book.
  
   MCW002
  
   Explanation
   Either the master processor or an alternate processor was
   discovered to be unresponsive.  In the case when the master
   processor is unresponsive, no recovery can be attempted so
   the system was terminated.  In the case of an unresponsive
   alternate processor, a reset was done but the processor was
   doing system work when it was reset, so recovery was not
   possible.  This abend can occur if running first level, or
   if running second level and unresponsive processor detection
   is enabled.
  
   User response
   See z/VM: Diagnosis Guide for information on gathering the
   documentation you need to assist IBM in diagnosing the
   problem; then contact your IBM Support Center personnel.
  

Temporary fix

• *********
  * HIPER *
  *********
  FOR RELEASE VM/ESA CP/ESA R640 :
  PREREQ: VM65776 VM65988 VM66248
  CO-REQ: NONE
  IF-REQ: NONE
  FOR RELEASE VM/ESACP/ESAR710 :
  PREREQ: VM66265 VM66283
  CO-REQ: NONE
  IF-REQ: NONE
  

Comments

APAR Information

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

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

  UM35459 UM35460

Modules/Macros

• HCPCKMST HCPDSP   HCPMPC   HCPOM2   HCPSRMBK HCPSXL   HCPSYN
  HCPSYS   HCPSYSCM HCPZSC
  

Fix information

• Fixed component name

  VM CP

• Fixed component ID

  568411202

Applicable component levels

• R640 PSY UM35459

     UP19/09/30 P 2001 ¢

• R710 PSY UM35460

     UP19/09/30 P 2101 ¢

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.