<html>
<head>
</head>
<body>
I've tried with and without aliases - I am going to check to see if the
system does a low level bios scan during startup (I think it does) maybe
I can disable it - (I'm running 2.4.8) --<br>
<br>
Thanks,<br>
Mark<br>
<br>
<br>
M. Mueller (bhu5nji) wrote:<br>
<blockquote type="cite" cite="mid:E16UFA8-0005mc-00@barry.mail.mindspring.net">
<pre wrap="">On Friday 25 January 2002 03:15 pm, you wrote:<br></pre>
<blockquote type="cite">
<pre wrap="">Mark,<br>I have a dual homed machine with heterogenious NICs as well. The<br>solution I found is to use a modular kernel with the card's drivers as<br>modules. Then add two lines to your modules.conf:<br>alias eth0 pcnet32<br>alias eth1 eepro100<br></pre>
</blockquote>
<pre wrap=""><!----><br>If you have a system with one Ethernet card (I/F-A) which is a pcnet32, then <br>the system will associate eth0 with I/F-A, and the alias above will associate <br>the pcnet32.o driver in /lib/modules/blah-blah-blah/net with eth0, so when <br>the kernel references eth0, pcnet32.o will be loaded (if not already) and the <br>kernel will execute pcnet32.o code.<br><br>If you add another Ethernet card to your system (I/F-B) which is a eepro100, <br>it is possible that I/F-A will become eth1 and I/F-B will be the new eth0. <br>Then the aliases above will lead to failure. The fix would be to reverse the <br>statements as:<br><br>alias eth1 pcnet32<br>alias eth0 eepro100<br><br>Unless the association between I/F-A and eth0 can be controlled (and I don't <br>think it can be on all machines), then a trial-and-error method with the <br>aliases as described above is a necessary process.<br><br>My 2.2 systems can get by without having the alias for the net dri
vers - even <br>when I change card types (tulip vs rtl8139). There is something else at work <br>connecting the symbol eth<x> to it appropriate driver. I suspect that it is <br>a PCI scanning function. In my 2.2 systems, I am using pci-scan.o <br>(<a class="moz-txt-link-freetext" href="http://www.scyld.com/network/updates.html#pci-scan">http://www.scyld.com/network/updates.html#pci-scan</a>). In my 2.4.8 system, I <br>found /lib/modules/2.4.8-26mdk/modules.pcimap (which contain both pcnet32 and <br>eepro100 references).<br><br>Have you tried running without aliases in your /etc/conf.modules?<br>What is your kernel version? If you have a 2.2 then maybe updating to new <br>drivers that use pci-scan.o will help eliminate the need for aliases. If you <br>have a 2.4, then it seems that the scanning function is already built in to <br>the system.<br><br>mjm<br><br></pre>
<blockquote type="cite">
<pre wrap="">Now when you reference eth0, the kernel will use the pcnet32 driver and<br>find a pcnet32 card to attach it to. This works for me.<br><br>Ken<br><br>On Fri, 2002-01-25 at 08:00, Mark wrote:<br></pre>
<blockquote type="cite">
<pre wrap="">I am having a bizzare problem - I have a netfinity-5100 with dual nic's<br>- It has an on board (eth0) pcnet32 --- the second is in pci slot 1<br>(eepro100) (eth1) ... When the system boots up, eth0 expects, eepro100<br>and eth1 exepect pcnet32 - so I bring it up in Single user mode, change<br>the drivers and do an init 5 (all is well) -- if i try to reboot, then I<br>start all over. Has anyone any idea why the devices/software switch<br>positions? Could it be the bios?<br><br>tia,<br>mark<br><br><br>_______________________________________________<br>TriLUG mailing list<br><a class="moz-txt-link-freetext" href="http://www.trilug.org/mailman/listinfo/trilug">http://www.trilug.org/mailman/listinfo/trilug</a><br></pre>
</blockquote>
</blockquote>
<pre wrap=""><!---->_______________________________________________<br>TriLUG mailing list<br><a class="moz-txt-link-freetext" href="http://www.trilug.org/mailman/listinfo/trilug">http://www.trilug.org/mailman/listinfo/trilug</a><br><br></pre>
</blockquote>
<br>
</body>
</html>