IV11780: DEADLOCK IN HD_GS_VGSA_MERGE AND HD_SA_ONEREV APPLIES TO AIX 6100-07

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• Customer may encounter a deadlock in concurrent VG when
  HACMP processes LVM_IO_FAIL error.
  
  (0)> th -lk
                  SLOT NAME     STATE    TID PRI   RQ CPUID
   CL  WCHAN
  
  pvthread+01E600  486 gsclvmd  SLEEP 1E60037 03C    0
     0  F1000A02013A0C68 slist_table+000940
  pvthread+02E200  738 cl_vgsa_ SLEEP 2E20071 03C    2
     0  F1000A020039CE78
  
  (0)> f 486
  pvthread+01E600 STACK:
  [00535AE8]slock+000488 (0000000000009024,
  8000000000009032 [??])
  [00009558].simple_lock+000058 ()
  [045A10F8]hd_dev2vg+000058 (??, ??)
  [0455ED84]hd_gs_vgsa_merge+000044 (??, ??)
  [045520BC]hd_lvm_config+000F7C (??, ??, ??, ??)
  [0455C4B4]hd_cfg+0001D4 (??, ??, ??, ??)
  [00003850]ovlya_addr_sc_flih_main+000130 ()
  [kdb_get_virtual_memory] no real storage @ 30326A68
  [D0FD761C]D0FD761C ()
  [kdb_read_mem] no real storage @ FFFFFFFFFFF92D0
  
  (0)> f 738
  pvthread+02E200 STACK:
  [000D57F0]e_block_thread+000290 ()
  [000D6448]e_sleep_thread+0000E8 (??, ??, ??)
  [00014F50].kernel_add_gate_cstack+000030 ()
  [045A9558]hd_sa_onerev+0000F8 (??, ??, ??, ??)
  [045A8D1C]hd_sa_config+001B3C (??, ??, ??)
  [0456F38C]hd_ioctl+002F2C (??, ??, ??, ??, ??, ??)
  [00556D00]rdevioctl+0000C0 (??, ??, ??, ??, ??, ??)
  [007279C0]spec_ioctl+000080 (??, ??, ??, ??, ??, ??)
  [005873F0]vnop_ioctl+000050 (??, ??, ??, ??, ??, ??)
  [0059C3FC]vno_ioctl+00009C (??, ??, ??, ??, ??)
  [00664DF8]common_ioctl+0000F8 (??, ??, ??, ??)
  [00003850]ovlya_addr_sc_flih_main+000130 ()
  [kdb_get_virtual_memory] no real storage @ 2FF22848
  [D01310D4]D01310D4 ()
  [kdb_read_mem] no real storage @ FFFFFFFFFFF92D0
  

Local fix

• A reboot only clears the deadlock
  

Problem summary

• A concurrent VG in powerHA enviornment may become deadlocked
  if both nodes experience I/O failures at the same time for
  the VG, and with very specific timing.
  

Problem conclusion

• Do not wait for a particular lock in the VGSA_ONEREV codepath,
  so that it cannot deadlock with an active vgsa merge thread.
  

Temporary fix

Comments

APAR Information

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

  IV11021

• 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 U839508

     UP12/05/11 I 1000

PTF to Fileset Mapping

• U839508 bos.rte.lvm 6.1.7.15

• U846172 bos.rte.lvm 6.1.7.2