IV30703: SYSTEM CRASH IN SEMSLEEP() WHEN CREATING/DELETING SEMAPHORES APPLIES TO AIX 6100-06

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• System crash with following stack:
  
  (0)> f
  pvthread+0A4200 STACK:
  [000BF964]remove_e_list+000024 (F100010105974C00,
  F1000101025BB8D8 [??])
  [000D7BCC]e_block_thread+00050C ()
  [00609920]rtipc_tsleep_block+0001A0 (??, ??, ??)
  [0060C44C]rtsem_tsleep_thread+00024C (??, ??, ??, ??, ??,
  ??)
  [0060D09C]semsleep+0000DC (??, ??, ??, ??, ??, ??)
  [0060D670]atomic+000450 (??, ??, ??, ??, ??, ??, ??)
  [0060E350]rsemop+0002D0 (33E0005233E00052,
  070000038BFE88E0,
     0000000000000001, 0000000000000000, 0000000000000000,
  0000000000000000)
  [00610014]__semop+000074 (33E0005233E00052,
  070000038BFE88E0,
     0000000000000001)
  [00003850]ovlya_addr_sc_flih_main+000130 ()
  [kdb_get_virtual_memory] no real storage @
  70000038BFE87A0
  [900000000255B1C]0900000000255B1C ()
  [kdb_read_mem] no real storage @ FFFFFFFFFFF9500
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED:
  * Systems running the 6100-06 Technology Level with
  * bos.mp64 below the 6.1.6.22 level.
  ****************************************************************
  * PROBLEM DESCRIPTION:
  * System can crash when certain workloads create and delete
  * semaphores at a very high rate.
  * One known cause of this when db2stop or stopsap is run on a
  * system where DB2 and SAP are running on the same host.
  *
  * Semaphore structures can be incorrectly overwritten while
  * threads using those structures are sleeping, and when they
  * wake up, they can make an invalid reference and crash the
  * system with a stack trace similar to this:
  *
  *      000BF964 remove_e_list+000024
  *      000D7BCC e_block_thread+00050C
  *      00609920 rtipc_tsleep_block+0001A0
  *      0060C44C rtsem_tsleep_thread+00024C
  *      0060D09C semsleep+0000DC
  *      0060D670 atomic+000450
  *      0060E350 rsemop+0002D0
  *      00610014 __semop+000074
  ****************************************************************
  * RECOMMENDATION:
  * Install APAR IV30703.
  * Prior to fix availability, an interim fix is available from
  * ftp://public.dhe.ibm.com/aix/ifixes/iv30703/
  ****************************************************************
  

Problem conclusion

• Extend the possible range of semaphore sequence numbers to
  avoid potential of overwriting the wrong structure.
  

Temporary fix

• *********
  * HIPER *
  *********
  

Comments

• 6100-06 - use AIX APAR IV30703
  

APAR Information

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

  IV27680

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

  IV37083 IV37084 U858664

Fix information

• Fixed component name

  AIX 610 STD EDI

• Fixed component ID

  5765G6200

Applicable component levels

• R610 PSY U858664

     UP13/03/28 I 1000