IV53430: DEADLOCK IN JFS2 INTERNAL SNAPSHOT CODE CAN CAUSE HANG

APAR status

• Closed as program error.

Error description

• A dead lock can occur when working with internal jfs2
  shapshots
  As a result commands on the associated jfs2 filesystem
  can hang.
  Both kernel stacks involved with dead lock will contain
  the function
  siWriterReadSMap with stacks similar to:
  
  (0)> f 190
  pvthread+00BE00 STACK:
  [000E47F8]e_block_thread+000298 ()
  [000E5368]e_sleep_thread+0000E8 (??, ??, ??)
  [002A76D8]bmAssign+000778 (??, ??, ??, ??, ??, ??)
  [002A6960]bmRead+0000A0 (??, ??, ??, ??, ??, ??)
  [00303BD4]xtSearch+0005D4 (??, ??, ??, ??, ??)
  [002C4F88]siWriterReadSMap+000348 (F10001023D4CA880,
  0000000001ACC0A0,0FFFFFFFF4FA7B48, 0000000000000000)
  [002C4720]siCOWLookupSMap+000060 (??, ??, ??, ??, ??)
  [002CCC70]siCOW+000270 (??, ??)
  [002BBB9C]j2PagerService+00051C (??)
  [002B7954]j2PagerThread+0001F4 (??)
  [00387774]threadentry+000094 (??, ??, ??, ??)
  [kdb_read_mem] no real storage @ FFFFFFFFFFF8D30
  
  and
  0> f 81
  [00574454]complex_lock_sleep_ppc+0001D4
  (0000000000574454,
  8000000000001032,
     0000000088024024, 0FFFFFFFF42B7670 [??])
  [005763C0]lock_write_ppc+0001A0 (??)
  [002C5048]siWriterReadSMap+000408 (F10001023D4CA880,
  00000000032C8BA0,
     0FFFFFFFF42B7B48, 0000000000000000)
  [002C4720]siCOWLookupSMap+000060 (??, ??, ??, ??, ??)
  [002CCC70]siCOW+000270 (??, ??)
  [002BBB9C]j2PagerService+00051C (??)
  [002B7954]j2PagerThread+0001F4 (??)
  [00387774]threadentry+000094 (??, ??, ??, ??)
  [kdb_read_mem] no real storage @ FFFFFFFFFFF8D30
  
  Code changes in IV29780 with abstract:
  Deadlock hang in snapshot code doing chdir and remove
  operations
  
  causes the race condition which can lead to this deadlock
  

Local fix

• Work around is to use external snapshot instead of
  internal
  snapshots
  

Problem summary

• Database begins blocking on writes. Slowly brings IO to a stop.
  

Problem conclusion

• Change locking serialization
  

Temporary fix

Comments

APAR Information

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

  IV46121

• 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

     UP