IV19341: SYSTEM CRASH DUE TO FREED SOCKET WHEN SOCKETPAIR() CALL USED APPLIES TO AIX 6100-08

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• When socketpair calls are used on the system we could see
  a crash in socket code path (close, send, etc.) due to
  one
  of the sockets in the freelist.
  Stack could be something like,
  
  [00009514].simple_lock+000014 ()
  [0049EB08]unp_disconnect@AF32_12+000228 (??, ??, ??, ??,
   ??)
  [0049C574]uipc_usrreq+0005D4 (??, ??, ??, ??, ??)
  [004C75A8]sodisconn+0001A8 (??)
  [004C6DE4]soclose2+000BE4 (??, ??)
  [004C8070]soclose+000010 (??)
  [004D2734]soo_close+000294 (??)
  [00630F7C]closef+00005C (??)
  [00547890]closefd+000150 (??, ??)
  [00547C44]closex+000304 (??, ??)
  [00547EBC]close+00011C (??)
  [00003850]ovlya_addr_sc_flih_main+000130 ()
  [kdb_get_virtual_memory] no real storage @ 111694A50
  [900000000B192F4]0900000000B192F4 ()
  [kdb_read_mem] no real storage @ FFFFFFFFFFF6380
  
  This is because the socket in unp_conn is already freed.
  

Local fix

Problem summary

• System crash in a system using socketpair() with a stack
  similar to:
  00009514 .simple_lock+000014 ()
  0049EB08 unp_disconnect@AF32_12+000228 (??, ??, ??, ??,
  ??)
  0049C574 uipc_usrreq+0005D4 (??, ??, ??, ??, ??)
  004C75A8 sodisconn+0001A8 (??)
  004C6DE4 soclose2+000BE4 (??, ??)
  004C8070 soclose+000010 (??)
  004D2734 soo_close+000294 (??)
  00630F7C closef+00005C (??)
  00547890 closefd+000150 (??, ??)
  00547C44 closex+000304 (??, ??)
  00547EBC close+00011C (??)
  00003850 ovlya_addr_sc_flih_main+000130 ()
  kdb_get_virtual_memory  no real storage @ 111694A50
  900000000B192F4 0900000000B192F4 ()
  kdb_read_mem  no real storage @ FFFFFFFFFFF6380
  The application involved is likely to be multithreaded.
  .
  Customers running Oracle RAC 11.2.0.3 or newer, with
  Oracle Cluster Health Monitor (CHM), are likely to see
  this problem.
  

Problem conclusion

• The code in socketpair() system call modifid so that in
  multithreaded environment, another thread is not able to close
  one of the sockets during critical pairing phase.
  

Temporary fix

Comments

APAR Information

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

  IV16603

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

Fix information

• Fixed component name

  AIX 610 STD EDI

• Fixed component ID

  5765G6200

Applicable component levels

• R610 PSY U845040

     UP12/10/23 I 1000

PTF to Fileset Mapping

• U845040 bos.mp64 6.1.8.0