Data page fault crash

[Cristian_4]

Hi,

I got a crash in one of my servers, "Data page fault" and im not really sure what to do to find out which process is the guilty

mordor root:/var/adm/crash/crash.1 510# echo "msgbuf+8/s" |adb -m vmunix . | more
hostname mordor
model 9000/785/J5600
mem size 80000000(x)
panic Data page fault
dumptime 1100755412 Thu Nov 18 05:23:32 UTC 2004
savetime 1100756284 Thu Nov 18 05:38:04 UTC 2004
release @(#)B2352B/9245XB HP-UX (B.11.00) #1: Wed Nov 5 22:38:19 PST 1997
msgbuf+8: e Jul 6 19:18:39 UTC 2004
_release_version: @(#)B2352B/9245XB HP-UX (B.11.00) #1: Wed Nov 5 22:38:19 PS
T 1997
__kern_ci_revision: $Header: kern_sequence.c,v 1.6.106.512 97/11/05 18:01:46 mso
sa Exp $

panic: Data page fault

PC-Offset Stack Trace (read across, top of stack is 1st):
0x0038248c 0x00395ee4 0x003955ec
0x001576b0 0x000d3ec4 0x000c4a3c
0x00130570 0x00130688 0x001a37ac
0x00154eb8 0x00156ac8 0x0015768c
0x0014ffc8 0x001526c4
End Of Stack

NOT sync'ing disks (on the ICS)
Interrupt Type 15 (Data page fault):
Instruction Address (pcsq.pcoq) = 0x0.0x3ffe5c
Instruction (iir) = 0x4bfd0000 (load/store)
Target Address (isr.ior) = 0x0.0x0000000000000001
Base Register (gr31) = 0x0000000000000001
Savestate Ptr (ssp) = 0x0.0x00000000008390d0
Savestate Return Pointer (ss_rp) = 0x3a0e64

linkstamp: Tue Jul 6 19:18:39 UTC 2004


i already used q4 and found nothing.

any clues ?

[hpuxrox]

I usually open a call for all crash dumps. They have more information on such errors that are not readily available to us on the internet.

[Cristian_4]

i would love to do that, but i dont have support contract with hp :(

[Steven E. Protter]

q4 dump analysis is something I always do with the response center.

This situation is a good reason to have a contract. HP will analyze it for you for time and materials. Call the response center and they'll give you an estimate. At least I believe they will.

I don't think it really takes them a long time to do this.

Having a software contract also gives you the right to new OS releases and application cd's. Its worth the money.

SEP

[Cristian_4]

yeah, i know, it's just a budget thing

[hpuxrox]

Well in that case, I would recommend a full patch anyalise for your system. Find out what patches you need and grep for "panic: Data page fault" in the decriptions.

[Cristian_4]

i had a new crash today, but it's hardware because i have a isr.ior panic in the /etc/shutdownlog file

i couldn't save the crash dump because the /var fs it was almost full

i do have the /var/tombstone file

how can a i know what hardware is failing ?

[Bill Hassell]

A program or application almost NEVER crashes the operating system. It is very likely due to lack of patches. A data page fault is simply a kernel addressing error. The cause cannot be determined without access to crash dump tools, lots of training and a large database of crash signatures. You don't need a support contract to get a crash analysis. HP will analyze your crash for a fixed fee and recommend needed patches.

Or you can apply the latest SupportPlus packages (HWE and QPK) and see if the problem disappears.

First, save off the first set of crash files onto tape. Then empty the /var/adm/crash directory. Now the next crash can be captured (if necessary). Now the panic message is very likely a software failure, not hardware. The reason is that the fault is not listed as an HPMC (High Priority Machine Check). Most (not all) hardware failures are caught with internal processor monitors and generate the HPMC trap when for instance, a data cache parity failure occurs. However, you can check the ts99 file in /var/tombstones. ts99 is the most recent hardware register dump.

Note to budget planners: a software and hardware support contract is an insurance policy. If you never use it, then it appears to be a waste of money. But just like insurance, when you don't have support, you can pay time and materials for needed services. For a workstation, it may be cheaper to just buy a couple of used ones and keep them for parts (assuming you're handy with tools...)