Re: Strange e1000

[Jeff Garzik <jgarzik@pobox.com>]

On Fri, Apr 04, 2003 at 03:45:25PM +0100, Justin Cormack wrote:
> On Fri, 2003-04-04 at 14:41, Paul Rolland wrote:
> > Hello,
> > 
> > > when I load the e1000 module, my NIC is recognized. Then, "pump -i 
> > > eth0" is called (DHCP-Client), the message "e1000: eth0 NIC 
> > > Link is Up 
> > > 1000 Mbps Full Duplex" appears and after some time I get the message 
> > > "operation failed".
> > > 
> > > When I sleep some time (currently 20 seconds) before doing 
> > > the "pump", 
> > > everything works as expected.
> > > 
> > > What the hell is happening here? Ok, I got it working with the 
> > > 20-sec-sleep but this is not the way it sould work...
> > > 
> > > My Board is a Gigabyte GA-7ZXR (1.0) and the Intel NIC is a PRO/1000 
> > > MT (should be the 82540OEM Chip). The NIC is attached to a NetGear 
> > > FSM726S Switch (24x100 + 2x1000). It is currenty the only box 
> > > attached 
> > 
> > Could it be possible that the 1000MBps FD on the e1000 side is
> > a local configuration, and that it needs some time to discuss with
> > the Netgear switch to negotiate correctly speed and duplex before 
> > working correctly ? (i.e. 20 sec = negotiation time)
> 
> It is probably something like this. For some reason the managed Netgear
> switches take a very long time to do anything. Log into the switch and
> watch the port status while this happens to confirm. I actually can't
> netboot off these switches because if this. Hopefully Netgear will come
> up with a fix.

In another thread, Scott Feldman (one of the e1000 team) asked if
spanning trees were enabled on the switch.  That could be a potential
cause.

	Jeff