Conference smurf::dec_mls_plus

From:	US2RMC::"NLIANG@us.oracle.com" "NLIANG.US.ORACLE.COM"  2-JUN-1997 16:56:02.03
To:	smurf::bat
CC:	
Subj:	stack of hung problem

 
Barbara, 
 
Here is the stack trace of the hung process: 
 
[2] record output foo.tmp (0 lines) 
(dbx) 
 
(dbx) (dbx) >  0 thread_block() 
["../../../../src/kernel/kern/sched_prim.c":2063, 0xfffffc00002ac7ac 
] 
   1 msg_dequeue(0xfffffc0001dcea80, 0xfffffc0007f9aa48, 0xffffffff80309020, 
0xffffffff883738e0, 0x0 
) ["../../../../src/kernel/kern/ipc_basics.c":869, 0xfffffc0000299cf0] 
   2 msg_receive_trap(0x3ffc0182c50, 0x3ffc0187700, 0x3ffc0082620, 
0x1400e5710, 0x400) ["../../../.. 
/src/kernel/kern/ipc_basics.c":1235, 0xfffffc000029a3b4] 
   3 _Xsyscall(0x8, 0x3ff8053ea44, 0x3ffc017b660, 0x1400e5710, 0x400) 
["../../../../src/kernel/arch/ 
alpha/locore.s":1333, 0xfffffc000051adf8] 
(dbx) (dbx) >  0 thread_block() 
["../../../../src/kernel/kern/sched_prim.c":2063, 0xfffffc00002ac7ac 
] 
    [ another dbm: ]
    
   1 msg_dequeue(0xfffffc0003232a80, 0xfffffc000423a048, 0xffffffff80309020, 
0xffffffff882538e0, 0x0 
) ["../../../../src/kernel/kern/ipc_basics.c":869, 0xfffffc0000299cf0] 
   2 msg_receive_trap(0x3ffc0182c50, 0x3ffc0187700, 0x3ffc0082620, 
0x1400e5710, 0x400) ["../../../.. 
/src/kernel/kern/ipc_basics.c":1235, 0xfffffc000029a3b4] 
   3 _Xsyscall(0x8, 0x3ff8053ea44, 0x3ffc017b660, 0x1400e5710, 0x400) 
["../../../../src/kernel/arch/ 
alpha/locore.s":1333, 0xfffffc000051adf8] 
(dbx) (dbx) >  0 thread_block() 
["../../../../src/kernel/kern/sched_prim.c":2063, 0xfffffc00002ac7ac 
] 
    
    [ yet another dbm hung on the same system: ]
    
   1 msg_dequeue(0xfffffc0005c71500, 0xfffffc000656e4a8, 0xffffffff80309020, 
0xffffffff8816f8e0, 0x0 
) ["../../../../src/kernel/kern/ipc_basics.c":869, 0xfffffc0000299cf0] 
   2 msg_receive_trap(0x3ffc0182c50, 0x3ffc0187700, 0x3ffc0082620, 
0x1400e5710, 0x400) ["../../../.. 
/src/kernel/kern/ipc_basics.c":1235, 0xfffffc000029a3b4] 
   3 _Xsyscall(0x8, 0x3ff8053ea44, 0x3ffc017b660, 0x1400e5710, 0x400) 
["../../../../src/kernel/arch/ 
alpha/locore.s":1333, 0xfffffc000051adf8] 
(dbx) (dbx) >  0 thread_block() 
["../../../../src/kernel/kern/sched_prim.c":2063, 0xfffffc00002ac7ac 
] 
    
    [ yet another dbm hung on the same system: ]
    
   1 msg_dequeue(0xfffffc0005c71500, 0xfffffc000656e4a8, 0xffffffff80309020, 
0xffffffff8816f8e0, 0x0 
) ["../../../../src/kernel/kern/ipc_basics.c":869, 0xfffffc0000299cf0] 
   2 msg_receive_trap(0x3ffc0182c50, 0x3ffc0187700, 0x3ffc0082620, 
0x1400e5710, 0x400) ["../../../.. 
/src/kernel/kern/ipc_basics.c":1235, 0xfffffc000029a3b4] 
   3 _Xsyscall(0x8, 0x3ff8053ea44, 0x3ffc017b660, 0x1400e5710, 0x400) 
["../../../../src/kernel/arch/ 
alpha/locore.s":1333, 0xfffffc000051adf8] 
(dbx) 

    I spoke to Kris this morning to get a pointer to the right place to ask
    about how to find this.  She said
    
    1.	Dave Long is the right guy to ask about thread issues.
    
    2.	Try building them a debug kernel:
    
    Ask them what software options they have in their kernel (ask them for
    their conf file), then build them a genvmunix with those options and
    with CFLAGS=g3 and no optimizations, and send it to them.  
    
    Then, next time these dbm's hang, they should force a crash and send
    the crash dump files here.

    Norman sent me the kernel options list; I've archived it in ~ftp/pub/oracle
    

    In further discussing this hang, Norman said that he had had to remove
    the calls to the lock manager their code was using... he said because 
    "they couldn't find the 'Digital Lock Manager' code in V4... e.g.,
    /usr/include/sys/dlm.h and the dlm_detach, etc., routines... where are
    they?"

From:	US2RMC::"NLIANG@us.oracle.com" "NLIANG.US.ORACLE.COM"  3-JUN-1997 22:23:47.95
To:	smurf::bat
CC:	ASCHEN@us.oracle.com, NLIANG@us.oracle.com
Subj:	Re: RE: multiple dbm hangs on thread_block


--=_ORCL_38673501_0_11919706032002120
Content-Transfer-Encoding:7bit
Content-Type:text/plain; charset="us-ascii"

The hung situation still persists. And I cannot relate it to any Oracle's 
potential problem. The only thing I could think of now is latch problem. Since 
latch is implemented in assembly language and the "as" compiler seems output a 
significant small .o for me, where in 3.1, I got a much bigger object file. 
 
Anyway, I've attached the files you need. 
 
(I need more semaphores in order for several database running together.) 
 
Norman Liang 
Oracle Corporation

From:	US2RMC::"NLIANG@us.oracle.com" "NLIANG.US.ORACLE.COM"  3-JUN-1997 22:28:01.30
To:	smurf::bat
CC:	
Subj:	Re: RE: multiple dbm hangs on thread_block


--=_ORCL_38673663_0_11919706032006300
Content-Transfer-Encoding:7bit
Content-Type:text/plain; charset="us-ascii"

One more thing you need to know. We're using a multi-processors machine and 
It's more likely to hang when I'm using parallel query option from Oracle. 
 
Norman Liang 
Oracle Coporation

    I just found out that dlm stuff is in the kernel, for TruClusters --
    still don't know where the pool is.