Test Station console loses contact with V-Class Server

[Jeff Fall]

Hello all Wizards,
I have been taking care of a V-class. I am having a lot of trouble installing Emulex LP9000 Fibre Channel cards. Their driver makes my V-Class on HPUX-11.00 produce a "data page fault" upon device probing.

As their driver works in L-Class HPUX 11.00 supposedly, I figured I would start upgrading our V-class.

I find it is running TSSW Exemplar Diagnostics Version 5.1.

I found this is a few revs behind so, I start to load in the next rev via a patch.
The next rev of TSSW is 5.1.2.
I found the patch to handle that.

The patch is PF_CV220512.
I followed the instructions and let the patch expand into a directory on the 712 test station for the V-Class,
(not on the V-Class itself).
Then, I installed the actual software of the patch
This command was
sh CV220512.inst

After I ran the command, it moved all the firware files out of my patches directory
into the /spp directory.

I saw my console X window then
close and never to be seen again.

if I try to bring up the console now it complains like this:

./sppconsole
console: connect: 2015@localhost: Connection refused

I see a jf-node_info program in spp/bin
# ./jf-node_info
IP Address UDP Port Version String
--------------- ------------- -----------------------------------------------
15.99.111.100 0x089a (2202) jtag 5.0 1999/03/24 20:17:14

But, I'm really lost.
I just want and need my console back. The V-class can't boot without it.

I would like to understand how the upgrade would disturb any linkages between the teststation and the V-Class.

I don't find any documentation anywhere on how to setup the testlan or the JTAG and get things working from broken.
cd

I found a ts.install script and it looks pretty powerful.
It fixed networking files all over the 10.20 directory tree on my teststation.
And, it added a bunch of entries to my /etc/hosts file for the v-class. But, still getting this..
# /spp/scripts/sppconsole
console: localhost: access from your host refused

Here is my etc/hosts file:
# The following lines were added by ts.install:
127.0.0.1 localhost
15.99.111.99 tsdart-d
15.99.111.255 dartbcst
15.99.111.100 mu-0000
15.99.111.101 mu-0001
15.99.111.102 mu-0002
15.99.111.103 mu-0003
15.99.111.104 mu-0004
15.99.111.105 mu-0005
15.99.111.106 mu-0006
15.99.111.107 mu-0007
15.99.111.108 mu-0008
15.99.111.109 mu-0009
15.99.111.110 mu-000a
15.99.111.111 mu-000b
15.99.111.112 mu-000c
15.99.111.113 mu-000d
15.99.111.114 mu-000e
15.99.111.115 mu-000f
15.99.111.150 obp-0000
15.99.111.151 obp-0001
15.99.111.152 obp-0002
15.99.111.155 obp-0005
15.99.111.156 obp-0006
15.99.111.157 obp-0007
15.99.111.158 obp-0008
15.99.111.159 obp-0009
15.99.111.160 obp-000a
15.99.111.161 obp-000b
15.99.111.162 obp-000c
15.99.111.163 obp-000d
15.99.111.164 obp-000e
15.99.111.165 obp-000f
# End of ts.install changes.

I can't ping any of those ip addresses and I believe they are HP internal addresses anyway.

Any help would be much appreciated.
Kind Regards,
Jeff

[Patrick Wessel]

Jeff,
usually does the V-Class boots without a console. If you have aotuboot disabled or need to interact in the Boot Cponsole Hndler, you can connect a " normal" serial console to the console port of the V-Class
But I'm sure that we get your teststation back to work. Did you post the whole /etc/hosts or are there any further entries in the file?
Try the following ping commands, what do they return?
ping mu-0000
ping 15.99.111.100
ping localhost
is the conserver running? ps -ef | grep conserv
can you please post the contense of the following file: /spp/data/conserver.cf
Do you use DNS or NIS on the teststaion?
Plese post the output of: nslookup localhost

Please come back with this information and we see what we can do for you.

[Bill McNAMARA_1]

those addresses are from the hp reserved block, but are used for the V class and superdome dedicated lans also.

They are not physically connected to the HP network.

Later,
Bill

[Jeff Fall]

Welcome to Engineering life.
You celebrate a victory and are slapped with the next failure.

I tried the upgrade firmware script..

Here is what the teststation did:

# /spp/scripts/load_CV220512
**********************************************************************
*** Starting the firmware download and initialization process.
**********************************************************************
load_CV220512: Version 5.1.2 Date: 1999/08/30 16:00:00

**********************************************************************
*** The system will now be reset to OBP. Please wait.
**********************************************************************
Executing "do_reset all 1 obp" command.

**********************************************************************
*** Saving NVRAM contents and beginning firmware download via OBP.
**********************************************************************
Issuing the following OBP commands:
nvsave
source /core@f0,f0000000/lan@0,d30000;15.99.111.99:/spp/scripts/CV220512-diag

And, here is what the V-Class did (via WORKING yea ) console window

[0:B] ok
[0:B] ok .obp
OBP Release 4.1.2, compiled 99/04/12 19:02:47
[0:B] ok nvsave
Each '.' represents one NVRAM option value read
................................................................................
................................................................................
...................................................................
All current configuration options have now been read.
5451 bytes being written to flash RAM NVSAVE file.
nvsave done.
[0:B] ok source /core@f0,f0000000/lan@0,d30000;15.99.111.99:/spp/scripts/CV22051
2-diags
DNS server : 15.99.111.99 (f636f63)
NFS server : 15.99.111.99 (f636f63)
Mounting : /spp/
nfs: rpc call receive failed, retrying

nfs: rpc call receive failed, retrying

nfs: rpc call receive failed, retrying

nfs: rpc call receive failed, retrying

nfs: rpc call receive failed, retrying

nfs: rpc call receive failed, retrying
nfs: rpc call failed.
nfs: mount call failed

nfs-lookup failed with status : stale
nfs: nfs-resolve failed before `CV220512-diags'
Cannot access file: /spp/scripts/CV220512-diags
from device: /core@f0,f0000000/lan@0,d30000
[0:B] ok

Hopefully, I did not fry the V-Class. But, knowing my luck..

Jeff

[Patrick Wessel]

Jeff,
Looks like you get really exceeded about your console functionality ;-)
Lets try to get some more light in the dark:
Yes, there are always two conserver processes running, this is nothing unusual. ping mu-0000 is the easiest way to verify if the network connection between teststation and v-class is Ok (private LAN). The other network connector on the core utility board obp-0000 response to a ping only if the OS is up and running.
You didn't need my help to point you to the interesting line in conserver.cf, that's good, because it saves us a day. I wouldn???t' start to mess with the conserver.cf but would create a /etc/nsswitch.conf file that defines first look into /etc/hosts and than use DNS because the teststation software isn't designed to work with DNS. If you have currently no /etc/nsswitch.conf or an empty file include just the following line:
hosts: files[NOTFOUND=continue] dns

You will find a few pages about console trouble shooting for the V-Class. I wrote it once down for our customer engineers and a coworker translated it to English. I hope it helps to answer some of your questions and gives you food for thoughts???

And your next problem with the firmware download??? well, to be honest??? I would suggest to get your teststation straight before we try to find the problem. As a rule of thumb: The teststation is a tool, not a toy.
But what we can say about the firmware upgrade problem is??? the OBP of the v-class tries to NFS mount the /spp directory of the teststation. That didn't succeed.

And a last point: why do you try to install an old firmware version and not the current one?

[Patrick Wessel · Accepted solution]

oops, I missed the attachmend in the first answer. So here we go

[Jeff Fall]

Hi Patrick.
I really appreciate your .pdf you attached. I did change the switch file for DNS settings and it did allow me to run the firmware upgrade script and the V-Class from the firmware level was able to download firmware from the teststation over NFS. The instructions warn about it taking 6 minutes and it took seconds.

Here are my current levels:

PDC_ENTRY version 4.1.4.0
POST Revision: 5.1.0.0
OBP Release 4.1.4, compiled 99/08/13 17:09:58 (3)
SPP_PDC Release 2.0.0 (04/12/99 18:55:48)
Command:

My test station is now at:

# ./diag_version
Exemplar Diagnostics, Version 5.1.2

All of the firmware patches I see listed from this point on all require

Exemplar Diagnostics, Version 5.2 or better.

They warn very strongly of not upgradiing from any level lower then 5.2.

But, I don't see a path from 5.1.2. up to 5.2.
Any ideas?

My real hope is that to get the PDC in the V-Class up to rev level 3.0.0 or 3.0.2 will correct PCI problems to allow me to boot HPUX 11.00 with the March 2002 quality pack
and to allow me to use an Emulex 9002 card.

To date, when I try this card, it causes HPUX 11.00 to fail during the probe.

The company - Emulex says it works on L-Class, so I figured maybe it is just a firmware aspect of the V-Class.

Here is a fresh snapshot of vmunix crashing due to this driver.

I did configure an early dump a few weeks ago and actually have the binary of the dump files here to send to someone to look at. I did fire up Q4 and take a look myself and all it really says is that a line in the probe function in lpfc driver is causing the data page fault.. This is why I have been so ambitious to
bring the firmware up to the most recent levels.

The card is a 64 bit PCI card and it is in the 32 bit slot of the V-Class.

The N-Class got a PCI-X face lift I was reading. I wonder if at least there is a 64 bit PCI slot I/O cage available for V-Class?? Anyone know?

We are trying the LP9000.
Here is a link to the Emulex
vendor site..
Anyone used this in a V-Class ever?

http://emulex.com/products/modelguide.html


Probing epic2
Probe of epic2 complete
2 epic
2/0/0 c720
2/0/0.7 tgt
2/0/0.7.0 sctl
2/1/0 c720
2/1/0.7 tgt
2/1/0.7.0 sctl
2/2/0 lpfc
lpfc0: Adapter initialization failed, Mailbox command 88 0 0 0

Trap Type 15 (Data page fault):
Instruction Address (pcsq.pcoq) = 0x0.0x5dbed8
Instruction (iir) = 0x0e931200 (load/store)
Target Address (isr.ior) = 0x0.0x0000000000002114
Base Register (gr20) = 0x0000000000002114
Savestate Ptr (ssp) = 0xebc4800.0x400003ffffff0fa0
Savestate Return Pointer (ss_rp) = 0x5dbd2c

System Panic:

OBP reboot Thu Mar 7 21:21:18 PST 2002
_release_version: @(#)B2352B/9245XB HP-UX (B.11.00) #1: Wed Nov 5 22:38:19

[Patrick Wessel]

Jeff,
Great to hear that your V-class does again its duty.
If I remember right is 5.3 the current firmware for the V22x0 (you will find the release note attached). The whole bundle is 250Mbyte so it's to much to attach it or send it as an e-mail. Usually it's a CD with the part no. B6044-10038.
The V2500 and V2600 have 64bit PCI slots. There are no plans to upgrade the V22x0 to 64 bit PCI.