----- Original Message ----- From: "Edward Dekkers" <edward@xxxxxxxxxxxxxxxxxxxx>
Sent: Tuesday, February 17, 2004 5:24 PM
Subject: Re: RH Fedora network interfaces
Standard issue Supermicro server with dual 10/100 network ports built in.
RH9
and Fedora Core 1 see both network interfaces on cold-boot and everything
works
great. If I do a soft reboot (without physically powering the server down
for a
few seconds), the network cards are not seen as active and kudzu complains
about
removed hardware. RH 7.3 used to work just fine for a year on this machine.
[...]
Is there anything I can do in the configuration somewhere to tell kudzu that
the
network cards are just fine, not to detect them again, and not to mess with
the
ocnfiguration? Cold booting works fine every time, so I don't get why a warm-boot makes it go crazy on both RH9 and Fedora.
Once a server is up and running properly, and you don't plan to make any configuration changes, why not just disable kudzu altogether?
just run ntsysv from a console, and disable it.
Ed,
Thanks for the suggestion, but that only disables kudzu during bootup - but does not resolve the issue that on warm-boot RH9 and Fedora do not recognize the network interfaces on this server
Well, mate I was only answering your question above, "Is there anything I can do in the configuration somewhere to tell kudzu that the network cards are just fine, not to detect them again, and not to mess with the
ocnfiguration?"
Not starting Kudzu will do the above, but see below for my thoughts.
(standard Intel Pro 10/100 82559). With kudzu
disabled it simply says eth0 and eth1 are not there. If you just power it down and power it up again (cold boot) it detects both network interfaces properly and everything is working 100%.
So again, the problem is not with kudzu - it's with RH9 and Fedora. On cold-boot they both see the network interfaces and work PERFECTLY, but on warm-boot it's as if they do not get initialized or something and come up with eth0/eth1 not present. Both ports are built into the motherboard, not PCI cards, and both have worked perfectly fine on RH7.3 for over a year - cold and warm boot. That's why the problem is with RH9 and Fedora and how they initialize the network ports on bootup.
I'd just like to know what the difference is on warm-boot that RH9 and Fedora complain about and why they can't detect these ports properly. :-(
Any other ideas? Any way to issue a 'forced reset' on these ports somehow on shutdown or bootup, something like a cold-boot would normally do?
Chris
Chris, I don't think it is so much an initialization at start-up that's failing, I'm thinking that on a warm boot as eth0 and 1 shut down, somehow the card is left in a jammed state. A warm boot does not remove power from the PCI bus (yes, even if they're on-board they're on the PCI bus). A cold boot does, hence in effect re-setting the card back to blank slate so to speak.
I'm thinking the module for the card is not shutting the card back to blank slate mode on a warm boot.
If you bring them down manually, then up again, what happens?
Or more basic level, what about removing the module and re-installing it?
Does Intel provide a Linux driver for that card?
Those are the things I would look at.
Regards, Ed.
-- redhat-list mailing list unsubscribe mailto:redhat-list-request@xxxxxxxxxx?subject=unsubscribe https://www.redhat.com/mailman/listinfo/redhat-list
