[OpenAFS-devel] AIX system hang

[Hans-Werner Paulsen]

Dear OpenAFS user on AIX,
in the last year we had about 10 system hangs on
IBM machines running AIX 4.3.3 and OpenAFS 1.2.3-5
Now we forced the machines to write a system dump file,
and there is a deadlock with "gil" and "afsd".
Is there anyone who can help?

crash> thread 13 62
SLT ST    TID      PID    CPUID  POLICY PRI CPU    EVENT  PROCNAME
 13 s     d1b      810  unbound      RR  25   0                gil
        t_flags:  sig_avail kthread
 62 s    3e7d     3060  unbound   other  25   0               afsd
        t_flags: 

crash> lock
...
Threads waiting on locks:
  TSLOT  WAITING ON  HELD BY  SYMBOL OF ADDRESS
            ADDRESS    TSLOT  (if found)
     10  0x0554fb44       13  <.[afs.ext.iauth:DATA]+148a4> 
     11  0x33f63764       13  <__ublock+4028364>            
     12  0x33f63764       13  <__ublock+4028364>            
     13  0x05547bf0       62  <.[afs.ext.iauth:DATA]+c950>  
     15  0x05547bf0       62  <.[afs.ext.iauth:DATA]+c950>  
     42  0x05547bf0       62  <.[afs.ext.iauth:DATA]+c950>  
     52  0x05547bf0       62  <.[afs.ext.iauth:DATA]+c950>  
     59  0x05547bf0       62  <.[afs.ext.iauth:DATA]+c950>  
     61  0x05547bf0       62  <.[afs.ext.iauth:DATA]+c950>  
     62  0x33f63764       13  <__ublock+4028364>            
     65  0x05547bf0       62  <.[afs.ext.iauth:DATA]+c950>  
...

crash> trace -mk 62
Skipping first MST

MST STACK TRACE:
0x2ff3b400 (excpt=00000000:42000000:00044011:344f8000:00000106) (intpri=0)
        IAR:      .slock_ppc+25c (00150590):    cmpi   cr0,0x0,r19,0x0
        LR:       .slock_ppc+25c (00150590)
        2ff3aa60: .simple_lock+64 (00009564)
        2ff3aaa0: .[afs.ext.iauth:rx_WritevAlloc]+2c (054e70b4)
        2ff3aaf0: .[afs.ext.iauth:afs_MemCacheStoreProc]+d0 (05515e40)
        2ff3ab50: .[afs.ext.iauth:afs_StoreAllSegments]+938 (0550b2a8)
        2ff3ae50: .[afs.ext.iauth:afs_StoreOnLastReference]+12c (0551ded4)
        2ff3aeb0: .[afs.ext.iauth:BStore]+c8 (0551a99c)
        2ff3af20: .[afs.ext.iauth:afs_BackgroundDaemon]+240 (05519d4c)
        2ff3af80: .[afs.ext.iauth:afs_syscall_call]+278 (054be89c)
        2ff3b360: .[afs.ext.iauth:syscall]+ac (054bdd7c)
        2ff3b3c0: .sys_call_ret+0 (00003a90)


-- 
Hans-Werner Paulsen		hans@MPA-Garching.MPG.DE
MPI für Astrophysik		Tel 089-30000-2602
Karl-Schwarzschild-Str. 1	Fax 089-30000-2235	
D-85741 Garching