[Xitami] Xitami aborts on UnixWare
Bowden, Russell
xitami@lists.xitami.org
Thu, 11 Apr 2002 11:32:07 -0500
Xitami aborts frequently (sometimes after a few minutes, sometimes after a few hours) on UnixWare 7.1.1. I posted about this several months ago and Ewen suggested trying dbx on the core dumps to look at the stack, etc. I tried this ('debug' now instead of dbx) and didn't see anything that looked useful to me. Now I am running 2.5b5 on one port and 2.4d9 on another port (same machine), and each gets around 10000 transactions per hour. The problem has gotten much worse as the load increased. And often when one Xitami dies the other one dies, so I think it may be some kind of threshold in UnixWare. But sometimes it happens at 3 a.m. when there is practically no activity at all. All these transactions go through an LRWP processing module, with 25 copies of the module attached to each Xitami. Below is debug info from a 2.5b5 core dump:
Warning: No debugging information in Xitami
Core image of Xitami (process p1) created
CORE FILE [_kill]
Signal: sigabrt
0xbffa1e8c (_kill+12:) jb +0x67e <bffa2510> [ _cerror ]
debug> ps
Program ID Pid Thread State Function Location Comman
d
*xitami p1 -1 -1 Core _kill 0xbffa1e8c ./xita
mi -b 5800
debug> t
Stack Trace for p1, Program xitami
*[0] _kill(0x5994, 0x6) [0xbffa1e8c]
[1] abort(presumed: 0x8047b08, 0x8074768, 0x80ba618) [0xbffd69ca]
[2] sys_assert(0x80ba618, 0x11d) [0x80a45dc]
[3] terminate_the_thread(0x8170710) [0x8074763]
[4] execute_thread(0x8170710) [0x805b008]
[5] smt_exec_step(presumed: 0x8047b6c, 0x804ae86, 0x8047b8c) [0x805aa1a]
[6] smt_exec_full(presumed: 0x8047b8c, 0x80bd394, 0x8047b9c) [0x805a9f0]
[7] main(0x3, 0x8047b8c, 0x8047b9c) [0x804ae81]
[8] _start() [0x804a5ac]
debug> regs
Register Contents for p1, program xitami
%eax 0x00000000 %ecx 0x0007d030 %edx 0xbff90aa0
%ebx 0xbfffdc4c %esp 0x08047ab8 %ebp 0x08047af8
%esi 0x081293a0 %edi 0x08170710 %eip 0xbffa1e8c
%eflags 0x00000206 %cs 0x00000017 %ss 0x0000001f
%ds 0x0000001f %es 0x0000001f %fs 0x00000000
%gs 0x00000000 %fpsw 0xffff0000 %fpcw 0xffff137f
%fpip 0xbffa47fa %fpdp 0x080478e8
%st0 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st1 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st2 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st3 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st4 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st5 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st6 [ EMPTY ] 0x4003 b000 0000 0000 0000 == 22
%st7 [ EMPTY ] 0x4001 a000 0000 0000 0000 == 5
debug> q
This is one from 2.4d9:
************************ core.20475 *****************************
Warning: No debugging information in xitami
Core image of xitami (process p1) created
CORE FILE [_kill]
Signal: sigabrt
0xbffa1e8c (_kill+12:) jb +0x67e <bffa2510> [ _cerror ]
debug> ps
Program ID Pid Thread State Function Location Comman
d
*xitami p1 -1 -1 Core _kill 0xbffa1e8c ./xita
mi -b 5900
debug> t
Stack Trace for p1, Program xitami
*[0] _kill(0x4ffb, 0x6) [0xbffa1e8c]
[1] abort(0x80d16ac, 0x80d16ac, 0x808f610, 0x808e2c8, 0x11d, 0x80d169c) [
0xbffd69ca]
[2] sys_assert(0x808e2c8, 0x11d) [0x8089dad]
[3] terminate_the_thread(0x8116c38) [0x806abda]
[4] execute_thread(0x8116c38) [0x8057812]
[5] smt_exec_step(presumed: 0x804a852, 0x80d8d30, 0x80d2bf8) [0x80571a5]
[6] smt_exec_full(presumed: 0x80d8d30, 0x80d2bf8, 0x80d2c48) [0x8057180]
[7] main(0x3, 0x8047b8c, 0x8047b9c) [0x804a84d]
[8] _start() [0x804a53c]
debug> regs
Register Contents for p1, program xitami
%eax 0x00000000 %ecx 0x0007d030 %edx 0xbff90aa0
%ebx 0xbfffdc4c %esp 0x08047a8c %ebp 0x080c21ec
%esi 0x080c8a34 %edi 0x00000000 %eip 0xbffa1e8c
%eflags 0x00000202 %cs 0x00000017 %ss 0x0000001f
%ds 0x0000001f %es 0x0000001f %fs 0x00000000
%gs 0x00000000 %fpsw 0xffff0000 %fpcw 0xffff137f
%fpip 0xbffa47fa %fpdp 0x080478c8
%st0 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st1 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st2 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st3 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st4 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st5 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st6 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st7 [ EMPTY ] 0x4001 8000 0000 0000 0000 == 4
debug> q
This is what I usually get (2.4d9):
************************ core.22523 *****************************
Warning: No debugging information in xitami
Core image of xitami (process p1) created
CORE FILE [_kill]
Signal: sigabrt
0xbffa1e8c (_kill+12:) jb +0x67e <bffa2510> [ _cerror ]
debug> ps
Program ID Pid Thread State Function Location Comman
d
*xitami p1 -1 -1 Core _kill 0xbffa1e8c ./xita
mi -b 5900
debug> t
Stack Trace for p1, Program xitami
*[0] _kill(0x57fb, 0x6) [0xbffa1e8c]
[1] abort(0x80d17ac, 0x80d17ac, 0x808f3e0, 0x808f3f0, 0x80d16ac, 0x808f3c0, 0x8
0d16ac, 0x808f394, 0x808b3a8, 0xd0, 0x80d169c) [0xbffd69ca]
[2] mem_assert_(0x808b3a8, 0xd0, 0x80d21a0, 0x80d8d30, 0x80d2bf8, 0x80d2c48) [
0x8084555]
[3] main(0x3, 0x8047b8c, 0x8047b9c) [0x804a86d]
[4] _start() [0x804a53c]
debug> regs
Register Contents for p1, program xitami
%eax 0x00000000 %ecx 0x0007d030 %edx 0xbff90aa0
%ebx 0xbfffdc4c %esp 0x08047ab0 %ebp 0x00000001
%esi 0x00000003 %edi 0x00000000 %eip 0xbffa1e8c
%eflags 0x00000212 %cs 0x00000017 %ss 0x0000001f
%ds 0x0000001f %es 0x0000001f %fs 0x00000000
%gs 0x00000000 %fpsw 0xffff0000 %fpcw 0xffff137f
%fpip 0xbffa47fa %fpdp 0x080478c8
%st0 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st1 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st2 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st3 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st4 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st5 [ EMPTY ] 0x0000 0000 0000 0000 0000 == 0
%st6 [ EMPTY ] 0x4003 b000 0000 0000 0000 == 22
%st7 [ EMPTY ] 0x4000 c000 0000 0000 0000 == 3
debug> q
********************************
Any help or suggestions would be greatly appreciated.
Russell