Kernel hang on Ignite

[Bill Brandt]

Hi folks. Has anyone seen this before? Got a new rx6600, imaged it with FOE 11.23-0712, installed Integrity VM 3.50, patched OS to current level with anything that looks IVM-related, configured the unused array disks for a large raid-0 virtual disk, configured a single VM (2-cpus, 20%, etc.), configured the backing store to be a virtual FileDisk on the raid-0 disk with hpvmdevmgmt, started an ignite on the VM and halfway through the install, get this:

* Installing fileset "Asian-UTILITY.MFCONV-JPN-COM,r=B.11.31" (660 of
1378).
* Installing fileset "Asian-UTILITY.MFCONV-JPN-RUN,r=B.11.31" (661 of
1378).

Stored message buffer up to system crash:

MFS is not defined.
krs_read_mfs: Error 5 opening MFS.
Found adjacent data tr. Growing size. 0x21d9000 -> 0x61d9000.
Loaded ACPI revision 2.0 tables.
krs_read_mfs: Error 5 opening MFS.


MCA recovery subsystem disabled, not supported on this platform.
montecito_proc_features: PROC_GET_FEATURES returned 0xfffffffffffffff8

bad_kern_reference: 0xec6031.0x6, fault = 0x1c

===========After reboot==========

panic: Fault when executing in kernel mode
Stack Trace:
IP Function Name
0xe000000002767460
0xe0000000013cdd00
0xe000000000ce0010
0xe000000002268780
0xe000000000cb25a1
0xe000000000d45520
0xe000000000d13e30
0xe000000000d14550
0xe000000000ce0010
0xe000000002268780
End of Stack Trace

linkstamp: Wed Feb 06 17:16:26 MST 2008
_release_version: @(#) $Revision: vmunix: B.11.31_LR FLAVOR=perf
Calling function e000000001c9c6c0 for Shutdown State 1 type 0x2


THE SPINLOCK CODE HAS NOT FAILED!
You are seeing this message because some kernel code has
attempted to yield the processor holding spinlocks.
Please look for the code that did not drop the spinlock!

It appears to be a bug but maybe not? I previously set up the array with distinct virtual disks and had no problem installing the VM's.

[Tim Nelson]

Latest version of Ignite ?

Even the latest release quotes some IVM issues.

http://www.docs.hp.com/en/IUX/docs/release_note.pdf

[Charlie123]

Hi all,
I recently used ignite to backup & restore, I got the same error msg as this:
MFS is not defined.

krs_read_mfs: Error 5 opening MFS.
Found adjacent data tr. Growing size. 0x2131000 -> 0x6131000.
Loaded ACPI revision 2.0 tables.
krs_read_mfs: Error 5 opening MFS.

Can you explain a solution, thanks a lot.
Charlie

[Bill Brandt]

I tried upgrading Ignite to C.7.6.100. That made no difference. It appears that the SAS Controller is at the heart of the problem, or at least my problem.

I have several rx6600's. Those with the P400 array card work fine with IVM-those that just use the SAS Controller card fail miserably at IVM but work fine as standalone machines. I ended up simply using the machine above as a single host.

[Don Morris_1]

Hmm... how much memory did you give the IVM guests you were igniting with the SAS cards?

[John Maus]

I got essentially the same error message and I am not using Ignite. I have an rx8640 divided into 4 nPars. nPar0 is loaded with HP-UX 11iv2 Dec 2007 and is an HPVM Host using HPVM version 3.5 (AVIO), 32GB and 8 processors. I successfully created one 11iv2 FOE Guest installing from HP media using the real DVD drive on this nPar. The backing store was an internal 146GB disk. While trying to create a second Guest with 11iv3 DCOE Mar 2008 (backing store a 32GB raw logical volume on the Host boot disk), after putting in the second installation DVD, I got:
* Installing fileset "hpuxwsXML.BATIK,r=A.2.03" (1472 of 1476).
* Installing fileset "hpuxwsXML.COCOON,r=A.2.03" (1473 of 1476).
* Installing fileset "hpuxwsXML.FOP,r=A.2.03" (1474 of 1476).
* Installing fileset "hpuxwsXML.XALANJ,r=A.2.03" (1475 of 1476).
* Installing fileset "hpuxwsXML.XERCESJ,r=A.2.03" (1476 of 1476).

Stored message buffer up to system crash:

MFS is not defined.
krs_read_mfs: Error 5 opening MFS.
Found adjacent data tr. Growing size. 0x21d9000 -> 0x61d9000.
Loaded ACPI revision 2.0 tables.
krs_read_mfs: Error 5 opening MFS.


MCA recovery subsystem disabled, not supported on this platform.
montecito_proc_features: PROC_GET_FEATURES returned 0xfffffffffffffff8

The system then re-booted and died. I've attached a complete screenshot.

[Bill Brandt]

In response to Don's question: I started off with 2-2-4-4-8 (the host has 32GB). I fell back to all 4GB later with no change in the problem.

Side note: I found a reference somewhere to corrupted 11.31 install disks on the 1207 release. I just got the 0308 set but haven't tried them yet.

[Don Morris_1]

Ok, probably not related to what I was wondering about then. (There was an issue we fixed in the 0803 timeframe where v3 Guests with less than 4Gb of RAM but several I/O cards which wanted 32-bit DMA safe physical addresses were actually failing allocations during boot [counter-intuitive, but hey.. if it was obvious it likely wouldn't have been a bug in the first place] -- I was wondering if Ignite was getting to the point of trying to use a driver on a controller that hadn't properly initialized since you mentioned that it seemed to vary with underlying controller).

The screen shot posted with the similar issue shows 12Gb (and most of yours were 4Gb or more without a report of different behavior), so it isn't likely the same.

Too bad there's only the addresses on the stack... the screen shot does interestingly show a SCSI bus reset just before the NULL pointer dereference in the kernel.. but that doesn't really narrow anything down without a dump (which you won't get during install without jumping through a lot of hoops Support would need to walk you through).

Be interested to know if the updated media addresses things -- otherwise, please use any support channels you have to try to get this into analysis.