IV65589: VIOS CRASHED IN EMUL_SCHEDULE APPLIES TO AIX 6100-08

APAR status

• Closed as program error.

Error description

• VIOS Crashed at following stack..
  
  
  CPU 1 CSA F00000002FF47600 at time of crash, error code
  for LEDs:30000000
  pvthread+003500 STACK:
  [F1000000C0313784]vscsi_host:emul_schedule+000044
  (700DFEED00007262,
  F1000A00E0E04A08 [??])
  [F1000000C030DF20]vscsi_host:emul_no_dev_scheduler+000020
   (??, ??, ??)
  [F1000000C0310C1C]vscsi_host:emul_schedule_queue+0001BC
  (??, ??, ??)
  [F1000000C03226FC]vscsi_host:target_edasd_iodone+000F1C
  (??)
  [00014D70].hkey_legacy_gate+00004C ()
  [00262D20]iodone+000340 (??)
  [F1000000C031A5E8]vscsi_host:hdasd_issue_open+000168 (??)
  [F1000000C032BD9C]vscsi_host:target_kproc+00015C (??, ??,
  ??)
  [00014D70].hkey_legacy_gate+00004C ()
  
  
  (1)> print dev_info F1000A002B64BE00
  struct dev_info {
      struct q_manage {
          struct q_manage *next   = 0x0000000000000000;
          struct q_manage *prev   = 0x0000000000000000;
      } global_dlist;
      uint64_t eye        = 0x56626C6B44657600;
      uint64_t *dev_trace = 0x0000000000000000;
      void *dd_private    = 0x0000000000000000;
      struct task_set_t *task_set = 0xF1000A002B614600;
      unknown num_act_cmds        = 0x00000000;
      unknown dev_trc_index       = 0x00000000;
      unknown flags       = 0x0000;
      uchar device_type   = 0x08;
      uchar pad[0]        = 0x00;
      uchar pad[1]        = 0x00;
      uchar pad[2]        = 0x00;
      uchar pad[3]        = 0x00;
      uchar pad[4]        = 0x00;
      union Simple_lock {
          simple_lock_data _slock = 0x0000000000000000;
          struct lock_data_instrumented *_slockp  =
  0x0000000000000000;
      } dev_lock;
      struct q_manage *lun_list   = 0x0000000000000000;
      void *emulation_priv        = 0x0000000000000000;
      void (*scheduler)() = 0xF1000000C03AD158;
      struct cmd_elem *(*abort_io)()      =
  0xF1000000C03AD170;
      long (*wait_done)() = 0xF1000000C03AD188;
      void (*error_devstrat)()    = 0x0000000000000000;
      void (*delete_lun)()        = 0xF1000000C03AD1A0;
      void (*initiator_login)()   = 0xF1000000C03AD1B8;
      void (*unit_attention)()    = 0xF1000000C03AD1B8;
      long (*need_thread)()       = 0xF1000000C03AD1E8;
      void (*set_vtd)()   = 0xF1000000C03AD1D0;
      void (*get_vtd)()   = 0xF1000000C03AD1D0;
      long (*dev_dep_fun)()       = 0xF1000000C03ACAF8;
      void (*vbsd_init)() = 0x0000000000000000;
      long (*finish_init_lun)()   = 0x0000000000000000;
      unknown use_count   = 0x00000000;
      unknown blk_size    = 0x00000000;
      unknown num_blks    = 0x0000000000000000;
      dev_t blck_devno    = 0x0000000000000000;
      struct lun_reserve {
          unknown reserved        = 0x00000000;
          dev_t adpt_devno        = 0x0000000000000000;
      } reserve;
      char *copy_buf      = 0x0000000000000000;
      ulong copy_buf_len  = 0x0000000000000000;
      struct cdt *cdt     = 0x0000000000000000;
      void *ras_cb        = 0x700DFEED00007262;
  <---------------
  
  
  
       | 000000                           PDEF
  emul_schedule
      0|                                  PROC
  lun,cmd,gr3,gr4
      0| 008180 std      FBE1FFF8   1     ST8
  #stack(gr1,-8)=gr31
  ...
    697| 0081A0 ld       EB830028   1     L8
  gr28=(*)_vadapter_lu._vadapter_lu.lun(gr3,40)
  ...
    703| 0081B8 ld       E87C0100   1     L8
  gr3=(*)dev_info.dev_info.ras_cb(gr28,256)
  ...
    703| 0081C4 lha      A8030018   1     L2A
  gr0=(*)Crasr_block.rasr_block.rrb_trace_privlevel(gr3,24)
  <-----------
  ...
  
  gr3 is loaded with dev_info.ras_cb, which is,
  0x700DFEED00007262.  In
  
  case of other dev_info.ras_cb, it is set to valid values.
  We need to understand why dev_info for this is not
  properly
  initialized.
  

Local fix

Problem summary

• When a crash is caused by this problem, emul_schedule_queue
  will always be somewhere in the stack, though the actual
  function at the top of the stack may vary.  If a listing is
  used to see precisely where in the code the crash occurred, it
  will be while trying to reference the dkstat structure.
  

Problem conclusion

• Fix problem which corrupts command queues.
  

Temporary fix

Comments

• 6100-08 - use AIX APAR IV65589
  6100-09 - use AIX APAR IV65788
  

APAR Information

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

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

  U870211

Fix information

• Fixed component name

  VIRTUAL I/O SER

• Fixed component ID

  5765G3400

Applicable component levels