JR56344: IN INFINIBAND BONDING ENVIRONMENT, INTERMITTENLY LLM APPLICATION CRASH IN IBEL_THREAD

APAR status

• Closed as program error.

Error description

• In Infiniband bonding environment, if IB interfaces is disabled,
  intermittenly LLM application crash in ibel_thread.
  Call stack shows the following :-
  
  (gdb) bt
  #0  0x00007f2ff15e536b in ibel_thread (arg=0x0)
      at
  /home/ausbldu/mds_rm11c/currbld/src/mds.rm.shared/src/SockMgm.c:
  8146
  #1  0x00000037abc079d1 in start_thread () from
  /lib64/libpthread.so.0
  #2  0x00000037ab8e89dd in clone () from /lib64/libc.so.6
  

Local fix

• -NA
  

Problem summary

• In Infiniband bonding environment, if IB interfaces is disabled,
   intermittenly LLM application crash in ibel_thread. The problem
   was that while rebuilding the interface for the IB bond the
  IBEL thread could access resources that have already been freed.
   This is a timing issue, and so may not happen on every attempt.
  

Problem conclusion

• Changes has been implemented to ensure it does not access the
  resources that has already been freed.
  

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:

Fix information

• Fixed component name

  WS LOW LATENCY

• Fixed component ID

  5724T2100

Applicable component levels

• R260 PSY

     UP