invalid memory read access from kernel mode Tru64 5.1B pk3

[George R. White]

I have 2 node TruCluster runing 5.1B pk3 and I am getting the following output when it panics:

trap: invalid memory read access from kernel mode

faulting virtual address: 0x65206f74206e6f7d
pc of faulting instruction: 0xfffffc000078b8d8
ra contents at time of fault: 0xfffffc00007e5780
sp contents at time of fault: 0xfffffe048d1377e0

panic (cpu 8): kernel memory fault
syncing disks... dump device name: DK2-32, num: 0x13002f6, off: 0x0, len: 0x1000
00

LSM attempting to dump to device at major 19 minor 758

dump device name: DK1-31, num: 0x1300266, off: 0x0, len: 0x100000

LSM attempting to dump to device at major 19 minor 614

dump device name: DK4-34, num: 0x1300306, off: 0x0, len: 0x100000

LSM attempting to dump to device at major 19 minor 774

dump device name: DK3-33, num: 0x1300276, off: 0x0, len: 0x100000

LSM attempting to dump to device at major 19 minor 630

Is it safe to assume that the Advfs is corrupted on the boot disk?

[Michael Schulte zur Sur]

Hi,

I think, this calls for an update to pk5. This is more likely a bug.

greetings,

Michael

[Arch_Muthiah]

George,

I don't know you all recent patches installed with your system.

But there were patches related with "KMF" panic string, caused by ATM interface due to a simple lock time violation. You will have to find those patches or the Main stream patches are to be installed.

This patch solve the problem of ATM interface memory run out issues, that is "Kernel Memory Fault" KMF panic string.

Archunan

[Arch_Muthiah]

George,

This is the another patch ID: OSF410-405517, which ia also related with KMF panic string caused by Dig Unix ATM interface.


Archunan

[Arch_Muthiah]

George,

When I see your crashdc report, I see "fta0: Link Unavailable", please have a look into /var/adm/messages file.

Because if FDDI driver does not reset/initialize properily, then there is chance of getting KMF panic.

Archunan

[Arch_Muthiah]

George,

If you have kdbx debugging tool, you can find the text message stored in "message buffer" before panic occured on your CPU 8.

Because when the panic CPU sends the crash info to the other node in the cluster, it will save the reason in the message buffer and send only the panic string to the other nodes before it goes down.

Also you can drilldown the PC and stack using KDBX debugger, using these values and comparing the seguence events logged in the stack, you can get better idea about what is the process, and the related program caused the system panic.


Archunan

[Michael Schulte zur Sur]

George,

there are several panic patches like yours mentioned in the docs for pk5.

Michael

[George R. White]

Are you saying that this KMF is related to the ATM interface? Or do you think it is something else? I have not seen this problem at any of our other sites. Is there something I can do to work around this problem?

[Michael Schulte zur Sur]

Hi,

these kernel panics are occurring very sporadic. I would update to pk5.

Michael

[Arch_Muthiah]

George,

ATM interface is just one of the reason for KMF panic, but there are so many other reasons which causes system crash with this same KMF bugcheck.

As Michael suggestion, first update your system latest pk5, then lets see what will happen. Probably your issue may be covered by any one of the patch bundled in pk5.

you can find the patches details here...
http://www2.itrc.hp.com/service/patch/mainPage.do

Archunan

[George R. White]

Turned out to be a bad ATM card.