lan0 down after reboot

[Omar Alvi_1]

Hi,

I have a peculiar problem at a customer site.

When the system is rebooted, the lan0 interface shows its status as up. However it is not pingable. We cannot telnet to it or anything. However when we reinitialise the card, changing the status from ifconfig to down and then back up again, it regains proper functionality.

The card is a gigabit ethernet card, and has some additonal advanced options e.g max input buffer, something (if I recall) peak coal.

Also I note that in the output of ifconfig lan0, the flags are as follows

1843

first time I see this CKO ... and the other cards have only 843 set.

BACKGROUND

The system is a high end 8430 or something with GSP. We connect to the lan console without problem.

This system is a disaster recovery system, therefore it was cloned with the system whose data it is to backup daily; and this problem appeared after this cloning.

Thanks a lot for any help.

Regards,

-Alvi

[Sean OB_1]

Check the speed and duplex settings and verify that they match the port on the router/switch it is on. If they are not matched then you may not be able to hit the interface.

man lanadmin

[Todd McDaniel_1]

Check the speed set in your /etc/rc.config.d/btlanconf for my 11i box... It may be a different file for you depending on your OS version.

I had a situation where my box kept falling back to half duplex upon a reboot and I had to hardcode it there to fix it. There might have been a patch as well, but I can't remember.

HP_BTLAN_INTERFACE_NAME[0]=lan0
HP_BTLAN_STATION_ADDRESS[0]=
HP_BTLAN_SPEED[0]=100FD

Also, do a lanadmin -x 0 for lan0. IT will show you the current speed.

[Ron Kinner]

Sounds like a patch problem.

http://www1.itrc.hp.com/service/patch/search.do?BC=patch.breadcrumb.main|&pageContextName=hpux:::

Dial in your OS and search for gigabit or the card's model number.

Could also be the subnet violates an obsolete RFC which 11.0 checks for. Should cause an error message but you can check it with:
ndd -get /dev/ip ip_check_subnet_addr

If it says 1 try turning it off:

Edit /etc/rc.config.d/nddconf

TRANSPORT_NAME[0]=ip
NDD_NAME[0]=ip_check_subnet_addr
NDD_VALUE[0]=0

Use the next higher integer in the brackets if you already have something there.

Ron

[Jeff Schussele]

Hi Omar,

Couple of things:

1) Check the /etc/rc.log - specifically the net start section for error msgs.

2) Verify that you don't have a MAC address hardcoded in a conf file. It would be in /etc/rc.config.d & named either hpigelanconf or hpgelanconf. If the system you cloned from had it - this one will have it.

HTH,
Jeff

[Laurent Menase]

Hi Alvi,

CKO is checksum offload. some checksum are made onboard.

[Steven E. Protter]

Check the lights on the back of the card.

Check the lights on the switch or hub the box is connected to.

Change out the cables if practical.

Check with network administration and make sure the port is not set to autonegotiate. That will kill any 100 BT or 10 BT HP NIC.

check dmesg for failure message.

probe and exercize the card with mstm,cstm or my personal favorite xstm in X Windows.

Report back.

SEP

[Omar Alvi_1]

Thanks all, I'll check all these points.

I had checked the logfiles though, had no error messages in dmesg, rc.log, or syslog.log.

And setting for the lan card were full duplex, and auto negotiation off. The speed was set to 100 mbps because I think its connected using Cat5 UTP.

Thanks and Regards,

-Alvi

[Steven E. Protter]

It seems like a speed duplex issue.

Attaching a refernce file.

/etc/rc.config.d/hpbtlanconf

If the other folks have been helpful to you, please assign them some points.

I have enough though.

:-)

SEP

[Omar Alvi_1]

I will certainly assign points as all input is invaluable :)

I've checked whether MAC address might have been hard coded, or half duplex is initialised insted of full duplex on startup, however, from this point of view it seems fine.

HP_BTLAN_SPEED[2]=AUTO_ON
HP_BTLAN_SPEED[1]=100FD

In the above settings, lan2 is configured for autoneg and lan1 isn't, right? We don't need to explicitly set autoneg of, do we?

Now the nature of the problem at present

lan1 and lan0 work fine when only one is configured. (lan2 is not connected)

both lans also work fine when both are configured and their cables are connected.

The problem occurs when both are configured but only one cable is connected. And the problem occurs only on rebooting; reboot itself becomes time consuming.

The ip addresses are (similar to) 10.2.3.15 and 10.2.3.16 for the gigabit and 100MB lan cards respectively.

Any ideas?

-Alvi

[Jakes Louw]

We used to have a timing issue with FDDI, and we put a sleep into the RC sequence between the config of the FDDI cards and the Eth cards. This just helped the IP bind process complete, as it seemed to be hung up while binding an IP to the FDDI card.
Maybe you have something similar, with a clash between normal Eth and Gigabit?