csjoey@knuth.MTSU.EDU (mr. joey carruthers) (09/19/90)
Maybe someone out there can help me with this problem on the RS/6000.
It occurs mostly at shutdown, or when I have to do a kill on a Smit that
has gotten hung. The problem has occurred on both a 320 and a 520
running AIX version 3, and is as follows:
The system will go dead, except for the console, which appears to have
a debugging screen. At the top of the screen, I get 4 lines which start
with GPR0, GPR8, GPR16, and GPR24. These GPR numbers are followed by
codes, and the GPR16 has 2 setc of code reading 'DEADBEEF'.
I then get a listing of the contents of some registers, and on all 3
systems this has happend on, there follows a dump of memory locations
00028590 thru 000285F0 (it has been this exact same location on all
three machines). There is a message at the bottom of the screen saying
"Illegal trap instruction interrupt in kernel", followd by a prompt. If
you do a quit at the prompt, you have to reboot the system.
Has anyone else seen this happen? What causes it, and is there a cure?
Thanks in advance.
Joey Carruthers
________________________________________________________________________
CAUTION: .signature under reconstruction
------------------------------------------------------------------------frank@gremlin.austin.ibm.com (Frank Feuerbacher) (09/20/90)
> The system will go dead, except for the console, which appears to have > a debugging screen. First, let me say that I am NOT the expert on this. But... The system has indeed trapped an illegal instruction. Whoever installed the machine selected to have the low-level debugger come up when such things occur. Are you running the released level of the software? If not, get it. If it is the released level, call the problem in to IBM. It will probably be useful to record what shows up on the debugger screen. There probably is a way to crete a dump that may be useful to IBM to track down the beast. Disclaimer: I don't speak for my employer, they don't speak for me.
matt@mathew.austin.ibm.com (Mathew Accapadi;3-3517) (09/20/90)
In article <csjoey.653717485@knuth> csjoey@knuth.MTSU.EDU (mr. joey carruthers) writes: >Maybe someone out there can help me with this problem on the RS/6000. >It occurs mostly at shutdown, or when I have to do a kill on a Smit that >has gotten hung. > The system will go dead, except for the console, which appears to have >a debugging screen. At the top of the screen, I get 4 lines which start >with GPR0, GPR8, GPR16, and GPR24. These GPR numbers are followed by >codes, and the GPR16 has 2 setc of code reading 'DEADBEEF'. >three machines). There is a message at the bottom of the screen saying >"Illegal trap instruction interrupt in kernel", followd by a prompt. > > Has anyone else seen this happen? What causes it, and is there a cure? The best thing to do is to have an IBM SE look at this and take a stack trace to determine what component is causing the problem. Also, they can record the address of the instruction so as to determine the offset into the code that 'asserted'. Another thing you can do is type in "quit dump" instead of just "quit". This will force a dump to your primary dump device which you can then hand over to your IBM rep. Regards, Matt ------------------------------------------------------------------------- Mathew Accapadi ...cs.utexas.edu!ibmaus!auschs!mathew.austin.ibm.com!matt 512-823-3517 Tie Line 793-3517 -------------------------------------------------------------------------