Slashdot Mirror
Networking Problems w/ an Intel Ethernet Express Pro
An Anonymous Coward submitted the following question:
"Some one gave me a Packard Bell Legend 425 (486SX) and
I wanted to put slackware 3.5 on it, which I did, then I
bought an Intel Ethernet Express Pro 10+ ISA network card.
If I ping the linux box from Windows it shows 100% packetloss and vice versa for the
linux box. Could any one tell me what is wrong?
I have tried using: route add -net 131.204.207.0 eth0.
Any suggestions? And yes, I have read the net how-to."
There are more details. Click in to the article if you can
help!
Here are more details on the submittor's network
configuration:
"It was a Plug and play card and my 486 did not have PnP bios so I had to configure it in my K6-2 350 and then I put the card in my 486 and booted from a DOS floppy and ran the Intel tests on it and the tests came back that the card was working fine and it had a network connection. I booted to linux and loaded up the eepro.o module (with the irq=10 and io=0x300,0x30F) and it reported it found the card. Then I uncommented the modprobe line in /etc/rc.d/rc.modules and added the IRQ and IO settings to the line and dmesg shows the card as working. I have the ip on the machine set to 131.204.207.202 (this is just a lan) and a windows machine set at 131.204.207.201."
Just as an aside, I usually do not run submissions that don't have a valid e-mail address associated with it. Most Ask Slashdot submissions I try to handle myself, and if I can't get back to the submittor to ask for more information about the question, then it gets shoved into the round file. PLEASE: If you want to make Ask Slashdot submissions and want to have a decent chance of getting SOME form of answer, please include a working e-mail (or a URL where I can get one) so I can respond if the need arises. If not, then submissions will have to be really interesting for me to submit them.
"It was a Plug and play card and my 486 did not have PnP bios so I had to configure it in my K6-2 350 and then I put the card in my 486 and booted from a DOS floppy and ran the Intel tests on it and the tests came back that the card was working fine and it had a network connection. I booted to linux and loaded up the eepro.o module (with the irq=10 and io=0x300,0x30F) and it reported it found the card. Then I uncommented the modprobe line in /etc/rc.d/rc.modules and added the IRQ and IO settings to the line and dmesg shows the card as working. I have the ip on the machine set to 131.204.207.202 (this is just a lan) and a windows machine set at 131.204.207.201."
Just as an aside, I usually do not run submissions that don't have a valid e-mail address associated with it. Most Ask Slashdot submissions I try to handle myself, and if I can't get back to the submittor to ask for more information about the question, then it gets shoved into the round file. PLEASE: If you want to make Ask Slashdot submissions and want to have a decent chance of getting SOME form of answer, please include a working e-mail (or a URL where I can get one) so I can respond if the need arises. If not, then submissions will have to be really interesting for me to submit them.
This should be one of the last things to check, but it could be a bad card... Have you tried any others? Do they work?
As an aside, you should be using 192.168.*.* for your ip addresses. This could also possibly be causing your troubles, if one of your machines was connected to the 'net at the time. 192.168.* is specifically reserved for uses such as this, so you'll never run into any problems with multiple interfaces.
Ping the IP that you assign your card. If it doesn't work, you are in trouble. If it works, then it is a problem at another layer, such as cabling, setup, etc.
If you have any network sniffer software for linux, ifconfig eth0 promisc, and then fire up the sniffer and see if you can see any traffic.
Good Luck
If you read the source for the driver for this card, it tells that the driver really does not work well. I have two of these cards that gave me nothing but trouble on several machines. I tossed them for two NE2000 cards at $15 each. No more problems.
Please, do yourself a $30 favor.
Where I work, we bought a batch of them... about half turned out to be duds, with the Intel diagnostics showing mysterious transmit failures. The bad cards also made machines very unstable. We ended up replacing the bad cards with 3com and Linksys cards, which cleared up all our troubles.
FWIW, the PRO/100B is *not* the same as the 10 mbps cards.
you're problem is you need to download a new driver
check out this page:
http://titan.cs.uni-bonn.de/~canavan/ee pro/
What is a good 10/100 NIC then? (good quality and well supported under Linux). Is there something like ne2000 standard for 100 MB/s network cards?
P.S. Don't say 3Com. I currently have 10base2 network with generic ne2000 card and no problems.
You can't run 100 over 10-2.
I've never seen a 100/10/Coax combo that actually worked.
The best 100BT card under Linux is IMHO the 3com.
Second best is a DEC tuip chip.
The 2.1 intel drivers arn't that bad, but the dec is fast and cheap soo...
Ne2000s are garbage, the whole design is stupid and should have been forgotten long ago. They are slow and cause significant CPU wastage.
The Pro10+ was created in the "growning" PnP days, and actually can run in a jumperless mode. Go to Intel's site, and look for config utilities for it. It can be softset via dos tools (just a boot floppy in your 486), and then sent along its merry way. I've got one running in my masqdialer box for some time now, and it's given me zero problems.
Still to lazy to account for myself,
-=Brian (brian@maybe.net)
The driver source for the eepro was cleaned up in the new 2.1 kernels.
A friend of mine was having problems with the same card, so we helped the creator of the driver get it working in 2.1. Someone else cleaned it after that.
2.2 should have a good, working driver for the eepro.
The Intel EtherExpress 10+ and 100/Pro cards are some of the best I've used. If the drivers are that fucked up, run FreeBSD which has good support for these cards.
Posted by VRBrain:
This shows another reason why Intel NICs are a pain in the you-know-where. This summer during a major rehaul of our computer system, we bought about 120 EtherExpress 10/100 cards. Looked real nice on paper. However, they quickly turned into a night-mare. None of the cards worked with our SMC hubs- Kingston hubs were fine. Took about 10 services calls and three or four days to figure out the problem. Ow.
Whats in syslog? Often vital clues get burried in there.
Try an
# ifconfig eth0 IPADDR up
and tell me what happens.
Try pinging the ip address of the linux machine from the linux machine. (if this fails fix your linux problems, I can help here)
Try pinging your windows machine ip address on the windows machine. Problems here indicate windows problems, which is very common, try rebooting.
I've seen problems like this before, and I have no idea what would be wrong, but after playing around I get it working eventially. I do recall when I installed slackware (2.0.13) it didn't like my etherexpress card, but switching to a 3com card worked just fine. Maybe switch the cards between your two machines, if the bus fits.
Please send discarded NIC cards to me!
It is, of course, only for the sake of the environment that I suggest you avoid burdening the landfills and stock my spare parts collection instead.
I see even classic Slashdot is now pretty much unusable on dial up anymore.
I have use Etherexpress pro before, the main issue for the card are crappy drivers. The good news is,Intel Etherexpress fast ethernet is much cheaper than 3com ethernet card. The bad news is, crappy driver will ruins your day.
Redhat 5.2 have better support for this adapter.
I had a similar problem before, and my symptoms were identical. It was on a box that also ran windows, and when windows auto-configured my PnP sound card, it gave it a different value for the IRQ than I was using in Linux. The value I used in Linux was the same as the IRQ for my legacy ethernet card. I changed a jumper on the nic, and viola, it worked. Obviously, your problem is not identical, but I would bet that it is a IRQ or resource problem.
I don't have one of these cards (thank God) so I may be grossly misinformed here.
/proc/kmesg and watch for error messages.
You say it's a plug-and-pray card. It doesn't matter that your machine does not have a plug-and-play bios. That's okay. What you need to remember is that (unless the card has some FLASH ROM for the settings, which I doubt) the card DOES NOT RETAIN SETTINGS between reboots, may they be warm or cold. Essentially, it has no idea what IRQ, IO, DMA?, etc it's supposed to be using.
First of all, install isapnptools. Read the docs provided on how to set up your card.
Next, recompile your kernel with the card modular. Install and reboot. Ignore evil messages about the netcard.
Then, try running isapnp, then insmod'ing the netcard. The best way to see if it worked is to kill syslogd and klogd, then run a tail -f
-= PsychLo =- x86?? xor sp,sp inc sp push sp
I am running a legacy EtherExpress Pro 10+ card on a PCI machine, and for some reason the IRQ and I/O port settings I choose using the DOS utilities don't carry over when I boot into Linux. I can set the IRQ to, say, 10 using the DOS utility, but if I try to tell eepro.o to use those settings it won't work. However, when I just let eepro.o probe for the card, it sets it up on IRQ 7 for some reason and works fine. Have you tried that yet? Taking all the hard-coded irq and i/o port stuff out? Other than that, I would take all the advice listed above (ping the loopback, ping its own address, etc.) And one other thing -- did you check the cabling? In my experience a surprising number of network problems end up being that.
.config file by hand to make it compile in eepro.c. Or am I missing something here?
On a side note, why is the EtherExpress Pro driver disabled in the kernel 2.2.0pre4 make xconfig? The code's all there, but for whatever reason I had to go in and edit the
I had the same problem with mine a few years ago. I had to edit eepro.c, and in the first couple lines, it has a list of 'valid' addresses or io locations (or whatever it is) that didn't include 210 (where mine wanted to go). i just added 210 to the list, rebuilt, and it worked fine
this was 3+ years ago, so i would have assumed someone else had run into this problem and fixed it for real...
I think you have a resource conflict. Plug and Play is problematic for this card under Linux - disable PnP and set the IRQ manually using the Softset2 utility. Flash programming should also be disabled. I have two of these cards that I've used successfully under RH5.0 with the 0.11 eepro.c driver. There is a newer version (0.11d) that has been modified for multiple cards, and for Kernel 2.1.xx, although I haven't tried it yet. The old driver does "burp" occasionally when transfering large files - it displays various error messages on the console, but I have never lost any data. Somebody replying to your post was talking about the Pro/10+ PCI version, which is different.
I just had a similar problem. As someone else mentioned, you need to get a new version of eepro.c from http://titan.cs.uni-bonn.de/~canavan/eepro/ /usr/src/linux/drivers/net
and copy it into your
directory. Rebuild and install the modules.
The card is perfectly fine with the right driver.
Assuming you have already turned off PnP, and set the card to a sane configuration like io=0x300 and irq=5 with softset2.
/proc/interrupts. If it's zero, then the driver just doesn't work with your card, except probably (but i can't remember exaclty) at irq 5. Try mine (to be found at titan.cs.bonn.edu/~canavan).
Have a look at
Rainer