e1000 driver fails on 2.6.11.4 kernel distribution build

Support for the other architectures (X86_64, s390(x) and PowerPC)
Post Reply
mhrich
Posts: 66
Joined: 2007/01/24 19:54:51

e1000 driver fails on 2.6.11.4 kernel distribution build

Post by mhrich » 2007/02/06 19:08:53

[Second try - I messed somthing up the first time....]

I am trying to bring up 2.6.11.4 on a box that runs 2.6.9 just fine. There's an e1000 ethernet controller builtinto the box, but the 2.6.11.4 kernel has problems with this. If I build the driver into the kernel, it fails, usually in silence. If I build it as a loadable module, I get the error: "Different MAC address than expected" and the driver fails (no ethernet connection). This only happens on the x86_64 build - 32-bits builds (i386) work just fine on the same box, as does the 2.6.9 64 bit build.

Has anyone else seen this / have a solution?

We do not have the option to move up past 2.6.11.4 because we need the Infiniband driver and my information is that this is not supported beyond 2.6.11.4. If this is not so, please tell me how high up the infiniband driver is supported.

Many thanks.

mhr

Lenard
Posts: 2283
Joined: 2005/11/29 02:35:25
Location: Indiana

e1000 driver fails on 2.6.11.4 kernel distribution build

Post by Lenard » 2007/02/07 00:35:00

Edit the /etc/sysconfig/network-scripts file for the device and empty the MAC address info from the 'HWADDR=' line, any text editor as root will do. When done with the edit and saving the file as root restart the network

service network restart

FYI: it is a good idea to build some hardware kernel modules (like this one in your case).

mhrich
Posts: 66
Joined: 2007/01/24 19:54:51

Re: e1000 driver fails on 2.6.11.4 kernel distribution build

Post by mhrich » 2007/02/07 00:42:27

Don't know if this will make a difference, but I got the message wrong - here is the exact mesage from the system log:

Feb 5 10:16:15 sparenode1 sysctl: net.ipv4.conf.default.rp_filter = 1
Feb 5 10:16:15 sparenode1 sysctl: net.ipv4.conf.default.accept_source_route = 0
Feb 5 10:16:15 sparenode1 sysctl: kernel.sysrq = 0
Feb 5 10:16:15 sparenode1 sysctl: kernel.core_uses_pid = 1
Feb 5 10:16:15 sparenode1 network: Setting network parameters: succeeded
Feb 5 10:16:15 sparenode1 network: Bringing up loopback interface: succeeded
Feb 5 10:16:15 sparenode1 ifup: Device eth0 has different MAC address than expected, ignoring.
Feb 5 10:16:15 sparenode1 network: Bringing up interface eth0: failed
Feb 5 10:16:15 sparenode1 ifup: Device eth1 has different MAC address than expected, ignoring.
Feb 5 10:16:15 sparenode1 network: Bringing up interface eth1: failed

It turns out that these are the exact same messages that the kernel with the built-in driver generates.

Previous builds of questionable handiwork generated these messages:

Jan 30 02:27:21 sparenode1 kernel: Intel(R) PRO/1000 Network Driver - version 7.3.15-NAPI
Jan 30 02:27:21 sparenode1 kernel: Copyright (c) 1999-2006 Intel Corporation.
Jan 30 02:27:21 sparenode1 kernel: kobject_register failed for e1000 (-17) Jan 30 02:27:21 sparenode1 kernel:
Jan 30 02:27:21 sparenode1 kernel: Call Trace:{kobject_register+70} {bus_add_driver+107} Jan 30 02:27:21 sparenode1 kernel: {pci_register_driver+131} {:e1000:e1000_init_module+62}
Jan 30 02:27:21 sparenode1 kernel: {sys_init_module+344} {system_call+126}
Jan 30 02:27:21 sparenode1 kernel: Jan 30 02:27:50 sparenode1 kernel: Intel(R) PRO/1000 Network Driver - version 7.3.15-NAPI
Jan 30 02:27:50 sparenode1 kernel: Copyright (c) 1999-2006 Intel Corporation.
Jan 30 02:27:50 sparenode1 kernel: kobject_register failed for e1000 (-17)
Jan 30 02:27:50 sparenode1 kernel:
Jan 30 02:27:50 sparenode1 kernel: Call Trace:{kobject_register+70} {bus_add_driver+107}
Jan 30 02:27:50 sparenode1 kernel: {pci_register_driver+131} {:e1000:e1000_init_module+62}
Jan 30 02:27:50 sparenode1 kernel: {sys_init_module+344} {system_call+126}
Jan 30 02:27:50 sparenode1 kernel:

And so on. I suspect that this kernel (not of my making) was not built properly....

Suggestions?

mhr

Lenard
Posts: 2283
Joined: 2005/11/29 02:35:25
Location: Indiana

Re: e1000 driver fails on 2.6.11.4 kernel distribution build

Post by Lenard » 2007/02/07 11:35:23

Build your own, from what I saw this 2.6.11.4 kernel was not built correctly. If you did not build it where did it come from??

FYI: the latest released for CentOS has all the updates from upstream that are security and enhancement related (some from the 2.6.19 upstream series). The speed differences between the original 2.6.9 RH kernel release and the latest are slightly faster then the 2.6.11/2 vanilla series.

mhrich
Posts: 66
Joined: 2007/01/24 19:54:51

Re: e1000 driver fails on 2.6.11.4 kernel distribution build

Post by mhrich » 2007/02/07 23:45:32

I built the kernel myself. What looked like it as incorrectly built? (Details, please? :-)

I will see if there is an alternative (42.0.8 or newer kernel.org kernel).

Thanks.

mhr

Post Reply