Changing a LAN card Driver

[Andrew Pollard]

Hi,

We have 2 identical RX2620 servers running 11.23.
The only difference is that the first server (that is working correctly) has the following LAN card:
HP AB465-60001 PCI/P
CI-X 1000Base-T 2-port 2Gb FC/2-port 1000B-T Combo Adapter using the "igelan" driver.
The second server (that's not working) has the following LAN card:
HP A7012-60001 PCI/PCI-X
1000Base-T Dual-port Adapter using the "iether" driver.

Can anyone tell me how I can change the driver for the second server to see if the "igelan" driver makes a difference?

Thanks

Andrew

[Steven E. Protter]

Shalom Andrew,

There are limits but you should be able to renumber your hardware instance to meet these needs.

http://forums.itrc.hp.com/service/forums/questionanswer.do?threadId=695058

http://h21007.www2.hp.com/dspp/files/unprotected/ddk/drivermigrationguide/net-mig.pdf

Summary, this is dangerous and can make your system unbootable.

You will need to remove and reinitialize your /etc/ioinit file.

Be careful not to change things you do not need to change, it can get hairy.

SEP

[Hemmetter]

Hi Andrew

// first consult man pages.

// find the HW-PATH of your NIC" with
# ioscan -fnkClan


// "remove special files"
# rmsf -H "HW-Path to nic"

// "bind new driver to Hardware"
# ioscan -M igelan -H "HW-Path to nic"

// "recreate special files."
# insf -H "HW-Path to nic"



rgds
HGH

[Andrew Pollard]

Hi,

I decided to go with the not so "dangerous" suggestion first and got this message:
# rmsf -H 0/4/1/1
rmsf: Specified hardware path has no devices

Any ideas?

Andrew

[Enrico P. · Accepted solution]

Hi,
your driver is correct:

http://www.docs.hp.com/en/5971-4261/ch01s01.html#d0e166

Your problem is something else (cable, NIC, switch port, IP/subnet configuration ...)

Enrico

[Andrew Pollard]

Hi Enrico,

You are correct, it is something else. We are just not sure what.

We have 6 LAN cards, 2 onboard and 5 of the 6 will not work.

We can ping from another system and a sniffer shows traffic going to the server but it is not responding, even though the network can see the MAC address of the sever that isn't responding.

It's a blackhole :-)

Thanks for all the suggestions.

Andrew

[Torsten.]

Try linkloop to ensure the connectivity.

[rick jones]

Just for the record - NIC drivers in HP-UX are not interchangable. If you were used to being able to say switch from bcm5700 to tg3 in Linux that does not happen in HP-UX. And besides, the hardware in question is more like "e1000 vs tg3" in Linux-speak - one is based on an Intel chip (iether driver) and the other a broadcom (igelan).

You mention "blackhole" - any chance you have multiple ports configured into the same IP subnet? By default when you do that, HP-UX assumes that each NIC/port has equal connectivity to the rest of the network. Packets will be accepted on all ports in the subnet, but will be transmited out only one of them.

If you want multiple NICs in the same subnet, the "best" thing to use is Auto Port Aggregation. Barring that you need to consider static routes pointing out specific ports, or setting ip_strong_es_model to a value of either 1 or 2.

So, checking netstat -rn might be goodness.

I take it the other two GbE's already on the motherboard are being used for something?

[Andrew Pollard]

Hi Rick,

We currently only have 2 ports configured. The first is lan0 and is a GbE on the motherboard. The other configured port is lan1 and is the second GbE on the motherboard and it is not working.

As I had mentioned before, a sniffer can see traffic going out but there is no return traffic. We have verified that routing is configured correctly.

It's a mystery.

Andrew

[rick jones]

How about routing on the remote, or the "router"? Routing is afterall a symmetric problem :)

[Torsten.]

First you tell us about installed combo cards and now about the build in cards.

Are you sure you configured the right cards connected to the network?

You mentioned 5 of the 6 LANs will not work.
Did you configure different subnets to each?

[Andrew Pollard]

Hi,

In a nutshell, I'm trying to use a second subnet. The one working port is a built in card and is configured for our "production" subnet (.20)

I then wanted to configure a second port (a combo card) to use our "backup" subnet (.120), and when that card did not work, I tried the other 3 combo cards. None of them worked and that's when I posted this question.

I then tried configuring the second built in card to use the "backup" subnet and it still does not work.

I hope that explains it a little better.

Andrew

[rick jones]

OK, we need to see the output of netstat -rn.

[Torsten.]

Are you sure the configured lanx corresponds to the port with the cable plugged in?

Sorry for this question, but I remember a thread some days ago with exactly this reason.

[Andrew Pollard]

Hi Guys,

I have moved the cable around a lot and configured all 5 remaining ports, but I'm sure the configured port is the one that the cable is plugged in to.

Attached is the output of "netstat -rn". I had to replace the first part of the IP with "*" for security reasons.


Thanks

Andrew

[Torsten.]

Please send a "lanscan".

[Andrew Pollard]

Attached is the lanscan.

Thanks

Andrew

[Jeff Schussele]

Hi Andrew,

Few more things to consider...

1) What subnet mask are you using on that static route? Check with netstat -rvn You may have a VLAN or supernet in play here. Check with your networking group to be sure.

2) If using ping outbound make sure you're using ping -i xxx.xxx.120.201 to "force" the ping out the proper I/F

3) If using traceroute, again use -i I/F or -s addr to force it out the proper I/F

Rgds,
Jeff

[Patrick Wallek]

2 more things that may be helpful.

1) Output of 'netstat -in'
2) Output of 'ioscan -kfnC lan'

[Torsten.]

Just to clarify this part.

0/1/2/0 lan0 build-in A
0/1/2/1 lan1 build-in B
0/3/1/x lan2/3 second card from top
0/4/1/x lan4/5 top card

make sure you configure only one interface to one subnet at a time.

You can also check the connectivity with "linkloop" (as already mentioned).

[Andrew Pollard]

Hi,

Attached is a file with all the commands recently mentioned.

Torsten,
You are correct about the configuration of the cards, and yes I am pretty sure there is only one interface to one subnet configured at a time.

Thanks

Andrew

[Patrick Wallek]

What is the subnet mask for these cards? Are you using a class B (255.255.0.0) or a class C (255.255.255.0) mask?

If it's 255.255.0.0 then that could be your problem if the 1st 2 octets of each network are similar.

If it's 255.255.255.0, then you shouldn't have any problems.

[Andrew Pollard]

Hi,

Our subnet mask is 255.255.255.0

[Torsten.]

linkloop is using the MAC address, not the IP address. see man linkloop

[Andrew Pollard]

Hi,

Here is the output of the linkloop.

# linkloop -i 2 -v 0x0015605E72C9
Link connectivity to LAN station: 0x0015605E72C9
error: expected primitive 0x30, got DL_ERROR_ACK
dl_error_primitive = 0x2d
dl_errno = 0x04
dl_unix_errno = 57
error - did not receive data part of message
#
#
# linkloop -i 1 -v 0x0016353E9D28
Link connectivity to LAN station: 0x0016353E9D28
error: get_msg2 getmsg failed, errno = 4
-- FAILED
frames sent : 1
frames received correctly : 0
reads that timed out : 1

The first linkloop was run on the card configured for ***.***.120.201 and the second was run on the ***.***.20.201 card.

Thanks

Andrew