PI89227: POTENTIAL DEADLOCK IN TCPIP

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• .
  The potential exists for two threads using TCPIP services
  (socket calls) to attempt to get control of two locks at the
  same time in different orders.  If there is a third thread
  needing exclusive control of one of those locks and its request
  occurs between the two points, processing within TCPIP will be
  suspended causing various responsiveness problems.
  
  The most likely external symptom will be an EZZ9674E TCPIP
  SYSPLEX PROCESSING WAS NOT RESPONSIVE message followed by an
  S4C5/77620405 (TcpSysplexUnresponsive) ABEND.  This will likely
  be accompanied by OMPROUTE Adjacency Failure messages and
  various applications not responding.  It is also likely that
  NETSTAT commands will hang.
  
  
  Verification:
  
  A TCPIPCS LOCK report from the resultant dump will show a DUCB
  holding the TCBMULTLIST lock and waiting for the XCF lock.  The
  holder of the XCF lock will also be waiting for the TCBMULTLIST
  lock.  The call list for the first DUCB will be:
  
     EZBPFCON
     EZBTCFCN
     EZBTCRD
     EZBTCPCN
     EZBITKOB
     EZBITDSS
     EZBITKSS
  

Local fix

•                                                                .
  If the application holding either of these locks is cancelled,
  processing will be allowed to continue.  Otherwise, a recycle of
  TCPIP will be required.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED:                                              *
  * All users of the IBM Communications Server for z/OS Version  *
  * 2 Release 3 IP: Dynamic VIPA (DVIPA)                         *
  ****************************************************************
  * PROBLEM DESCRIPTION:                                         *
  * AbendS4C5 0405 occurs because ezbxfpdx was unable to obtain  *
  * the XCF lock within the sysplex recovery interval.           *
  ****************************************************************
  * RECOMMENDATION:                                              *
  * Apply the PTF                                                *
  ****************************************************************
  Sysplex Autonomics detected that the XCF lock could not be
  obtained over the course of the sysplex recovery interval. This
  implies there is a thread holding the lock and not releasing it.
  While holding the XCF lock it was waiting for another lock. The
  other lock was held by a different thread who was waiting for
  the XCF lock. This deadlock is what prevents the XCF lock from
  being freed. The two locks are incorrectly being obtained in
  different order leading to the deadlock.
  

Problem conclusion

• Code was modified to ensure the two locks are obtained in the
  same order across all threads.
  

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:

  UI51680

Modules/Macros

• EZBTCPCN
  

Fix information

• Fixed component name

  TCP/IP MVS

• Fixed component ID

  5655HAL00

Applicable component levels

• R230 PSY UI51680

     UP18/01/03 P F801

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.