We are setting up our modular servers with a new OS version (ubuntu Server 11.04 w/ 2.6.38 kernel) and are experiencing wierd problems with the e1000e driver.
The hardware is composed of 2 80003ES2 controllers and 2 82571EB controllers (on mezzanine boards).
The ethernet driver loads ok after a modprobe e1000e, but once we configure it e.g. with dhcllient eth3 (one of the 82571EB ports), it starts to loop in up-down state and resetting. The 80003ES2 ports behave well.
The failure happens with at least driver vers. 1.3.17 and 1.3.10a. The error handling routine seems to consistent with line 4989 of netdev.c, if this helps, but we could not yet figure out which pre-conditions causes it.
Someone a clue?
Sorry to hear of your issue but we have some questions. Has this ever worked on the this exact HW? Does ist work with another OS like RH, SuSE or Centos? Since it's a new system it may have never worked and you're just seeing it. What type of blade server is this? Please supply as many details of the system as you can. What type of switch is the port that is seeing the issue plugged into? Do you see the issue if you plug the NIC into another NIC (back-to-back)?
Please let us know and we might be able to help.
Actually we solved the problem today by comparing the code of v 1.2.20 (which worked) and 1.3.17 (not working). The interface is ok and stable, but we are not sure, if this change impacts other configurations. We'll provide the diff asap on the list.
The HW is a Intel Modular Server and the blade itself is a MFS5000SI. The dual ethernet controller is on the mezzanine board and the switch is the integrated one.