Inserting an extra nic blocks ethernet

I have a network where the server (slackware 10.1) has a dial up connection, and a nic for the local network. I want to add an extra nic to connect to the adsl modem i will have within a few days. When i insert the second nic (different make) and boot i lose connection to my network, even without configuring anything on the new nic. Taking it out and booting makes everything work fine again.

ifconfig -a does not report the new nic, of course. I can do ifconfig eth1 and append an ip-address to it, and after that i can ping it, so it seems to me to be no hardware problem.
I can also ping eth0, but not a machine on the local network. I have checked that eth0 is still the same card, it is.

without the new nic inserted pinging the clients on the local network goes ok, and with the new nic i cannot ping my local network anymore (dest. host unreachable).

After inserting the extra nic and without configuring it the output of route -n stays exactly the same as it was before the insert of the card:
root@hpserver:/home/rob# route -n
Kernel IP routing table
Destination Gateway Genmask Flags Metric Ref Use Iface
192.168.1.0 0.0.0.0 255.255.255.0 U 0 0 0 eth0
195.121.0.0 0.0.0.0 255.255.255.0 U 0 0 0 ippp0
127.0.0.0 0.0.0.0 255.0.0.0 U 0 0 0 lo
0.0.0.0 195.121.0.174 0.0.0.0 UG 0 0 0 ippp0

I do not understand why just inserting a card stops the throughput through eth0.

any help will be highly appriciated.

rob

Are these NIC's the same make/model?
If not, try removing the module running the second card, and see if you regain throughput...

thanks for your suggestion. The cards are of a different type.
Here's what i have done so far:
Unloading the modules loaded for the new inserted card failed ("device or resource busy") even though the icard is not given an ip address or device name like eth1.
I tried rmmod crc32, rmmod mii rmmod 8139too and rmmod -r 8139too, in different orders but the unloading keeps failing.

So i thought this is an item too far beyond my skills, and i bought me another card of a different brand. This means that i now have three cards (3com.realtek and intel). It does not matter what card i use for eth0, as soon as i put in the second lan card that is to be eth1 it stops the possibility to ping machines in the local network. The eth0 card has ip address 192.168.1.100. As noted before ping 192.168.1.100 works, but ping 192.168.1.51 does not (dest. host unreachable).

unloading the isdn modules after bringing down ippp0 does not change this situation.

I keep checking whether the cards are changed place, but so far they are not.

And now i don't know where the problem could be ???????????

accidently solved the problem by removing the new nic, bootup, homenet works, shutdown and insert a second nic and connect it with een extra cable to my switch. Voila, it works.

I then tried to reproduce the problem which was not a very bright idea. Inserted my two favorite cards and could not get it to work anymore, with whatever combination of cards i tried.

To get to internet i removed the second card, let homenet work again, and then reinserted a second card, and then it worked again.

So i have reached the conclusion that if i shutdown with a non working configuration and change something it will not work. Shutdown with a working configuration with one card and then add a second (and connect it with a cable before boot) did the trick for me.

I still do not understand what happens, possibly leaves a non working configuration traces that make things fail after a reboot. But so far so good, i can live happily without that knowledge.

rob

You should NEVER insert a PCI card into a machine while it is running, unless you have a machine that is capable of hot-swapping, which I very much doubt.

This is more than likely why your install was not working. You should always shut the computer down, install the card, and then start the machine.

You are right, of course. I did not insert cards into a running machine. That is so normal to me that i did not mention shutting off and rebooting explicit.

You know, it sounds to me like some weird PCI IRQ mapping problem. Although it sounds like you are up and running, if it's not working I'd try different PCI slots and/or resetting the BIOS PCI assignment (reset configuration data, or similiar option in the BIOS.)

Does dmesg report any problem with available ipv6 router when both NICs are installed ?

It has been a while, but now i am back. As noted before my setup has worked for a while, but due to other reasons i did a new install of slack 10.1, and ran into the same problem again. I have tried a lot of things, but to my surprise it turned out to be a problem with irq's.

In the problematic set-up cat /proc/interrupts reported no shared irq's and no problems. Because i could not figure out anything else to cause the problem I tried to figure out if Darin is right, and he is.

Went playing around in my bios, in pci setting had to change the routing algorithm to fixed, and then i could assign a fixed irq to couples of pci slots, like slots 3 and 7 and 4 and 8. So i gave the slot with the non-working card a fixed irq, and now my nic's work fine both.

To my surprise cat /proc/interrupts now gives following output (as far as is relevant):

9: 3897 XT-PIC megaraid,eth0,usb-uhci
10: 188 XT-PIC ncr53c8xx, eth1

As far as i understand things this is not a preferable result, and the former output with just one device per line is commonly seen as saver when it comes to possible conflicts. But somehow it works fine, so I will leave it this way.

Thank you all for your help, rob

p.s the machine is an old hp netserver LH3.