topic Re: network bonding on Red Hat Enterprise 3 in Operating System - Linux

network bonding on Red Hat Enterprise 3

Roy Yeow_2 — Tue, 17 Feb 2004 21:09:15 GMT

Hi there,

I am seeing this error on my server (dmesg) when I configure network bonding:
eth0: HP NC7781 Gigabit Server Adapter found at mem f7ef0000, IRQ 11, node addr 000e7f215f97
eth0: Broadcom BCM5703 Integrated Copper transceiver found
eth0: Scatter-gather ON, 64-bit DMA ON, Tx Checksum ON, Rx Checksum ON, 802.1Q VLAN ON, NAPI ON
divert: allocating divert_blk for eth1
eth1: HP NC7781 Gigabit Server Adapter found at mem f7ee0000, IRQ 15, node addr 000e7f215f96
eth1: Broadcom BCM5703 Integrated Copper transceiver found
eth1: Scatter-gather ON, 64-bit DMA ON, Tx Checksum ON, Rx Checksum ON, 802.1Q VLAN ON, NAPI ON
bond_enslave(): failed to get speed/duplex from eth0, speed forced to 100Mbps, duplex forced to Full.
bond0: enslaving eth0 as an active interface with an up link.
bond_enslave(): failed to get speed/duplex from eth1, speed forced to 100Mbps, duplex forced to Full.
bond0: enslaving eth1 as a backup interface with an up link.
bcm5700: eth0 NIC Link is UP, 1000 Mbps full duplex, receive & transmit flow control ON
bcm5700: eth1 NIC Link is UP, 1000 Mbps full duplex, receive & transmit flow control ON

Is there a reason why bond_enslave is failing to get the speed/duplex setting? How can I resolve this? If I restart the network after it comes up, I won't see the error again, it looks like it is some sequence issue.

Here is my modules.conf:
alias eth0 bcm5700
alias eth1 bcm5700
alias scsi_hostadapter cciss
alias usb-controller usb-ohci
alias bond0 bonding
options bcm5700 auto_speed=1 line_speed=1000 full_duplex=1
options bond0 miimon=100 mode=active-backup

Thanks!

regards,
Roy

Re: network bonding on Red Hat Enterprise 3

Steven E. Protter — Tue, 17 Feb 2004 21:36:48 GMT

Most Gigabit adapters I am familiar with require auto negotiate. Looks like you have that, but if you are hard coding the settings, try not doing that.

This message looks like the switch is incapable of going any faster than 100 BaseT. I'd check that first.

It could be hard coded settings on the switch. The switch port needs to be auto negotiate. Check with the switch admin on that.

SEP

Re: network bonding on Red Hat Enterprise 3

Mark Travis — Tue, 17 Feb 2004 22:11:58 GMT

There should be a file called LINUX_SRC/Documentation/networking/bonding.txt in your kernel. Read it. What's probably happening is that the broadcom driver doesn't support mii or ethtool link status reporting.

The bonding driver doesn't work the same with all network interface drivers. It appears as though the existing broadcom drivers don't provide some of the functionality that the bond module needs -- namely, the ability to query the state of the interface and the ability to change the state of the interface. That's a step beyond just moving traffic through the interface.

It is very important that the bonding driver be able to shut the interface on and off in order for failover to work, and it is strange that it seems to be defaulting to 100FDX. It would be nice to be able to nail that by hand, but that might not be possible.

I'm sorry I haven't used bonding with broadcom nics. But when I ran into similar issues with Intel E1000 nics, I RTFKS and luckily found newer drivers.

Re: network bonding on Red Hat Enterprise 3

Roy Yeow_2 — Tue, 17 Feb 2004 22:43:00 GMT

I am using a pair of 6509 with 48 ports Gigabits Copper port module.
I have set everything to autonegotiate, but it is still giving the same error.
I tested with tg3 driver before this, and it is giving the same error.

Re: network bonding on Red Hat Enterprise 3

Bill Wood_3 — Wed, 18 Feb 2004 09:24:28 GMT

Please try adding the following line to /etc/modules.conf:

"probeall bond0 eth0 eth1 bonding"

Re: network bonding on Red Hat Enterprise 3

Olivier Drouin — Wed, 18 Feb 2004 10:58:27 GMT

bcm5700 is usually good for bonding. Should be your settings...

Re: network bonding on Red Hat Enterprise 3

Emory Cowan_1 — Wed, 18 Feb 2004 13:08:22 GMT

try bugzilla.redhat.com

Bugzilla Bug 112449: (TG3) driver doesn't work properly with bcm5700 nic