Here's an interesting perspective on the Two eepro100 card post from earlier
that might be enlightening. I am not sure that this was what was happening
with my setup-- I still think it was hardware... as I believe I tried
swapping cables between NICs just to test this type of scenario.
I do recall getting results back from the "wrong" card in one test, but was
unable to recreate that particular situation.
Definitely a story to consider. I was pretty sure I had tried everything,
including checking the MAC addresses shown by ifconfig to make sure I had
the right card noted as the right interface. Works now, with 2 different
cards. Not going to mess with it.
Rod Hauser Curriculum Manager - Linux
[EMAIL PROTECTED] Wave Technologies
314-692-1916 St Louis MO, 63141
----- Original Message -----
From: kevin morris <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Thursday, September 09, 1999 7:14 PM
Subject: Re: Two eepro100 cards...won't both work (long story)
> Rod,
>
> I found your post to the linux-net forum very interesting since I had
> the exact same problem with intel Ethernet Express pro 100s. However,
> my problem was with three eepro100s and I solved it after many hours of
> dead end trial and error.
>
> My linux server arrived with Redhat 6.0 and two eepro100s already
> installed by Penguin Computing. I set up a firewall, and several proxy
> servers for various services and everything was humming along nicely.
> Then I realized I needed to add another subnet to the internet gateway,
> but decided "Hey instead of buying a router why not just add another NIC
> to the linux server." Five hours later I still didn't have it.
>
> The next morning I came in and to make a long story short, I discovered
> this--when I installed the third NIC, Linux switched my interfaces so
> the new card became eth0, the card that was eth0 became eth2 and eth1
> stayed the same! My subnets we're now attached to the wrong interface
> cards!
>
> I realized this by looking at the HW addr: entries output by ifconfig.
> Then I unhooked all the ethernet cables, and begain pinging the IP of
> the subnet I wanted to connect to, letting it run while I tried each
> card with that subnets' cable until I found the card that hit. I then
> repeated it with the second one subnet, by process of elimination the
> third one was the one left over.
>
> Here is a pictorial description...
>
> The linux server before adding third card:
>
> TOP OF CHASSIS
> ---------
> |
> PCI5 | eth0
> === ---------- 170.1.112.227 => 170.1.112.*
> |
> PCI4 | eth1
> === ----------- 192.168.3.1 => 192.168.3.*
> |
> PCI3 |
> ===
> . |
> . |
> ---------
> * *
>
> Everythings cool until after adding third card:
>
>
> TOP OF CHASSIS
> ---------
> |
> PCI5 | eth2
> === ---------- 10.0.0.20 => 170.1.112.* (oops!)
> |
> PCI4 | eth1
> === ----------- 192.168.3.1 => 192.168.3.*
> |
> PCI3 | eth0
> === ----------- 170.1.112.227 => 10.0.0.* (oops!)
> |
> ---------
> * *
>
> I hope you can post this to the list since I'm not a subscriber, and so
> noone has to live through what I just went though.
>
> Regards,
> Kevin
>
> ------------------
>
> Each card will work fine in either PCI slot, but when _both_ cards are
> in
> the machine, neither will work. I'm pretty sure it has something to do
> with
> the kernel loadable modules, but I ran out of patience yesterday. Just
> as a
> spoiler, the machine is up and running with 2 NICs right now-- one
> eepro100
> and one epic100. But I want to _know_why_...
> Here's my hardware and what I tried:
>
> Hardware: HP Vectra486/66 w/ 20MB RAM etc. 2 PCI slots and one on-board
> 10BT
> LAN port (currently enabled in the BIOS-- anyone know offhand which
> driver
> would address this port? Not important)
> Software: RedHat 6.0, pretty much a standard install. As a note, the RH
> install detected the eepro100 fine, but I only had one in the machine at
> the
> time of the install. I installed the second NIC, and decided to
> experiment
> with linuxconf to add it. Seemed to work, ifconfig showed both
> interfaces
> up, but no response from either NIC as far as getting anything from the
> network. So here's what I did to troubleshoot.
>
> 1. Found an error (probably from using linuxconf...) in
> /etc/conf.modules
> and removed duplicate lines -- essentially the file repeated itself.
> Here's
> fixed conf.modules
> alias eth0 eepro100
> alias eth1 eepro100
> alias parport_lowlevel parport_pc
> pre-install pcmcia_core /etc/rc.d/init.d/pcmcia start
>
> 2. Went and read Ethernet how-to again, as well as Don's shorter version
>
> linked there-from.
> 3. Specified io base and irq options in lilo, with ether=
> 4. Reversed lilo specification because a) it didn't work and ii) it's
> the
> wrong place - they're modules, not in the kernel
> 5. Specified iobase as options in /etc/conf.modules. as such:
> alias eth0 eepro100
> alias eth1 eepro100
> options eepro100 io=0xfce0,0xfcc0
> 6. That didn't work so I double-checked the iobase that ifconfig gave
> me,
> using modprobe. Correct.
> 7. Tried making the options distinct for each card as such:
> alias eth0 eepro100
> alias eth1 eepro100
> options eth0 -o eepro100-0 io=0xfce0
> options eth1 -o eepro100-1 io=0xfcc0
> 8. Tried adding irq's to above, still no success. Double-checked irq
> availability with a cat /proc/interrupts. Correct
> 9. Looking back at the howto (p.2 for me of Ethernet howto) for doing
> this,
> I saw that this was for ne2000 ISA cards, so I took out options line in
> conf.modules on the following assumption: the loadable module must not
> accept these options, perhaps it's written to always probe, because PCI
> devices are always successful? (unlike ISA)
> 10. Backed up then removed /etc/sysconfig/network-scripts/ifcfg-eth1 and
>
> rebooted to let the machine bring up only eth0, but with both cards
> present
> in the machine. No joy.
> 11. Pulled the second eepro100 card out and put in an epic100, edited
> conf.modules. Booted successfully immediately. Sigh.
>
> Obviously, there were some reboots along the line there, and several
> times I
> downed all the interfaces (even lo) and brought them up one at a time,
> or
> one or the other, with ifconfig. One item of note is that, booting with
> one
> or both eepro100 cards installed, ifconfig showed (one or both) active,
> but
> route would hang, while route -n would complete. Here's a pair of
> samples:
> Kernel IP routing table
> Destination Gateway Genmask Flags Metric Ref Use
> Iface
> 10.0.1.254 * 255.255.255.255 UH 0 0 0
> eth1
> 192.168.1.80 * 255.255.255.255 UH 0 0 0
> eth0
> 192.168.1.0 * 255.255.255.0 U 0 0 0
> eth0
> 10.0.0.0 * 255.255.0.0 U 0 0 0
> eth1
> 127.0.0.0 * 255.0.0.0 U 0 0 0
> lo
> (hangs)
>
> at boot without eth1:
> [root@bozo/root]# route
> Kernel IP routing table
> Destination Gateway Genmask Flags Metric Ref Use
> Iface
> 192.168.1.80 * 255.255.255.255 UH 0 0 0
> eth0
> 192.168.1.0 * 255.255.255.0 U 0 0 0
> eth0
> 127.0.0.0 * 255.0.0.0 U 0 0 0
> lo
> (hangs)
>
> _but_ after bringing down one or the other NIC, route would complete,
> showing an accurate routing table without hanging. Here's a one NIC
> sample:
> [root@bozo/root]# ifconfig eth0 down
> [root@bozo /root]# ifconfig eth0 10.0.1.254 netmask 255.255.0.0 up
> [root@bozo/root]# route
> Kernel IP routing table
> Destination Gateway Genmask Flags Metric Ref Use
> Iface
> 10.0.0.0 * 255.255.0.0 U 0 0 0
> eth0
> 127.0.0.0 * 255.0.0.0 U 0 0 0
> lo
>
> This (getting route to not hang) would work after bringing down one or
> both
> interfaces and bringing them back up, even if I had both cards
> installed,
> but in any case where I had both eepro100's installed, I could get
> ifconfig
> and route to show working fine, but no network connectivity. The only
> exception to that was when I did a tcpdump with both NICs in, I did get
> some
> traffic, but there was no way I could find to get ping to respond (even
> to
> hosts on that segment, no gateway needed) if I had both eepro100's in.
>
> All's well now, but I want to know... what did I miss? I initially
> _chose_
> an identical second NIC because the first one was working fine... I
> know
> that the routing tables look a little different after using ifconfig to
> manually reconfigure, but that's not particularly pertinent, just
> because I
> didn't use the -host specification, and the current routing table works
> fine: the two routing tables are identical, except for the fact that the
> one
> that didn't work with two eepro100's hung right before the default gw
> line.
> I think this is just that it was trying to do a DNS lookup and getting
> no
> communication from the NIC. Which leads me to believe that the whole
> "route
> hangs" was just a dead end, another way of saying "Your NIC isn't
> working"
> Here's the working routing table:
> Kernel IP routing table
> Destination Gateway Genmask Flags Metric Ref Use
> Iface
> 10.0.1.254 * 255.255.255.255 UH 0 0 0
> eth1
> 192.168.1.80 * 255.255.255.255 UH 0 0 0
> eth0
> 192.168.1.0 * 255.255.255.0 U 0 0 0
> eth0
> 10.0.0.0 * 255.255.0.0 U 0 0 0
> eth1
> 127.0.0.0 * 255.0.0.0 U 0 0 0
> lo
> default 192.168.1.65 0.0.0.0 UG 0 0 0
> eth0
>
>
> The only thing I can think of after sleeping on it, is that I could've
> copied /lib/modules/##/net/eepro100.o to eepro100-2.o and specified it
> as if
> it were a "different" loadable module from conf.modules...
>
> Shouldn't I be able to run two identical PCI NIC's with the same
> loadable
> module? Doesn't it load twice in the initial scenario? I considered
> messing
> with the BIOS, but didn't as these are PCI cards, not ISA, and the two
> _different_ cards work fine without tweaking the BIOS.
>
> Other ideas? Thanks.
>
>
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
