Re: eth0 carrier wrongly seen on

[Chuck Anderson <cra WPI EDU>]

On Fri, Jan 16, 2009 at 03:24:23PM -0500, Dan Williams wrote:
> On Fri, 2009-01-16 at 21:20 +0100, Sébastien Mazy wrote:
> > Hello,
> > 
> > I updated today from Linux 2.6.27.8 to 2.6.28. Since this update, NM
> > 0.7.0 has seen my eth0 (driven by e1000e kernel module) carrier as on,
> > even where there was no cable plugged.
> > 
> > What do you think? Is it a e1000e bug or does NM need changes to work
> > properly with Linux 2.6.28?
> > 
> > 'cat /sys/class/net/eth0/carrier' --> '0' after the 'connection' on eth0
> > is achieved and no cable was plugged in.
> 
> Carrier detection is driven by netlink events, so the issue could be in
> a number of places.  To do top-level debugging, when you plug a cable in
> and unplug it again, does anything happen in the log?  Failing that,
> we'll need to make a netlink carrier test tool to show what's happening
> on netlink and see if the kernel/driver is lying to us or not delivering
> events.

FYI, I've been seeing this issue too since 2.6.28.  With no cable 
plugged in on bootup, NM thinks the device is active.  Until I 
unchecked "automatically connect to this network" on my multiple 
static configurations, it would choose one of the static configs and 
set the default route to it, even though no cable was plugged in.

Also using e1000e on a Thinkpad T61.