Hi,

I know that there is a known problem with the Netgear FA311 and Redhat/Linux after reading various sources across the net I have been unable to find a cure for my problem.

A little bit of background information for you, I am a complete newbie to Linux, but I am trying to run a Celeron 533 machine that I have put together with 'Left over' parts as a Router for my DSL, Once this is done I will also be building a p3 833 into a file server.

System Info -

Celery 533, 128mb RAM, 2.1gig HD, 2xFA311's. RedHat 9, 2.4.20-8

When I boot, as soon as it fires up ethX (Where X = 0 or 1) I get this error

ethX: PCI error 0x800000

It just repeats over and over and I cant see anything. I can still login by just typing login and password then typing service network stop, eventually it stops.

If I do a ifconfig ethx, everything looks alright, but the overruns are like huge, like > 3mil but different on both cards.

Some interesting information, though I don't know if this has anything to do with it, the network config shows the drivers to be natsemi DP83815 yet the chip on the card is DP83816. Also the card has stamped on it FA311 Rev-D2 (checked this after reading that some came out with the wrong EPROM).

I have tried to install new drivers from the Netgear site, and from www scyld com [cant post a url yet ?!] (I don't know if I installed these properly as there wasent much instruction), and after reading on another forum I also tried some other drivers (netdrivers.tgz) from the same site that had a 'Make/Make Install' file attached (For newbs like me). The netgear drivers made no difference, while the scyld drivers changed the error message to 'Something Wicked Happened!' but thats all.

If i startx and enable the cards there (one as static ip one as PPPoE for DSL) they seem to work, I can ping internal machines and also connect to the net, but my processor is sitting at 100% and it takes about 5 mins to do anything. If I then <ctrl><alt><F1> I see the error filling the screen and the only way back is <ctrl><alt><F7> as I cannot type any commands.

I have been waiting on a response back from Netgear with help (As I have purchased these cards for 14 machines as well as a 16port switch from them) for five days now, and as I hope to have this all running for start of business on Monday coming I am getting desperate.

Does anyone know how I can fix this ?!?!

Thanks in advance!

Ed

Side note, ctrl+alt+f2, or f3 up til most likely f6, will get you a login prompt and let you type commands f1 is where startx is launched from, so you can't login there as it's still running.

Check dmesg for any odd/error messages, try the card in a different slot, etc (sounds like a bus-mastering problem to me, maybe you can turn that off). Most likely you've done all these things and more. perhaps lspci will show something amiss on the pci bus, modinfo <modulenameforthecard> may reveal possible parameters for the module you can set at load-time with an "options modulename var=val..." line in /etc/modules.conf.

Yeah if i do f2-6 I get the same error spam etc...

I have tried other slots, with both cards, I have also tried just running one card, same problem.

Out of interest, I stuck an old SMC network card in to see if the PCI slot was buggered, and it worked fine, worked a treat. No errors etc. If only I didnt buy so many of the fn things.... (It said on the Netgear site that they are supported in RedHat)....

What would the lspci command show me ? ( ) ! Is there something specific that I should be looking for ? Also, the command for modinfo would then be - modinfo natsemi dp83815 ?! or like it shows in the config ... natsemi|dp83815

Aslo, do you think that the info on the chip is pointing me to a possible problem with the driver ? (eg written on the card is dp83816, driver running is dp83815).

There should only be one module associated with the netcard, so the modinfo command would be "modinfo modulename" where the "modulename" is most likely what comes after "alias eth0.." in your /etc/modules.conf, if you've got more than one "alias eth0..." lines there with differing modules on them, then one or the other is wrong. You should see the same module name turn up when you do "lsmod", which shows a list of all currently loaded modules.

You've configured these cards with different ip addresses? ie: /etc/sysconfig/network-scripts/ifcfg-eth0 & ifcfg-eth1 are defined correctly?

out of curiosity what does the output of "route" show you?

lspci, would show everything on the pci bus as linux sees it, "lspci -v" shows better info and should be used instead. You'd be looking for weirdness in irq's and/or io and memory addresses that linux thinks the devices are using. Look for "unknown device" lines also, but don't get too excited if you see them where they wouldn't affect the network controller.

If you type "dmesg | less" and using the arrow keys scroll down looking for "no known interrupt for pin X of device X", those are not good.

Yeah IP's are set correctly.. The static IP on eth0 is set to 192.168.0.2 and eth1 is set up for PPPoE (Similar to DHCP, IP is provided by ISP).

I am currently at work, I will give you the output from 'route' and those other commands in a few hours when I get home.



And BTW, Thanks heaps for your help so far

Any other ideas ?

no other ideas as of yet, (of course I'll be passed out in a few hours, different timezones...) although I'll probably see the same docs you found I'll try a search on this device myself tomorrow.

Yeah a few searches I did were in google,

eth0: PCI error 0x80000 (About 5 pages of results)
natsemi dp83815 (About 10 pages)
FA311 RedHat 9 (About 10 pages)
dp83816 (about 10 pages - not helpful)


A few sites I visited

http://www.netgear.com/support/suppo...?prodNum=FA311

Download page for FA311 on Netgear website.

http://www.scyld.com/pipermail/netdr...ne/000031.html

Drivers I tried from scyld. This above link holds almost the same info I had in my first post.

eth0 = Static IP
eth1 = PPPoE for DSL

Bringing up interface eth0 [OK]
5 seconds later error starts.

Ok after blind login and network service stop I got this -

Route - Showed nothing other than the headings (No ip's nothing)

lspci -v -

00:09.0 Ethernet Controller : National Semiconductor Corperation DP83815 (MacPhyter) Ethernet Controller
Subsystem : Netgear : Unknown Device f311
Flags : bus master, medim devsel, latency 32, IRQ 10
I/O Ports at e000 [Size=256k]
Memory at de902000 (32-bit, non-prefetchable [Size=4k]
Expansion ROM at dc000000 [Disabled] [Size=64k]
Capabilities : [40] Power Management Version 2

00:0b.0 Ethernet Controller : National Semiconductor Corperation DP83815 (MacPhyter) Ethernet Controller
Subsystem : Netgear : Unknown Device f311
Flags : bus master, medim devsel, latency 32, IRQ 5
I/O Ports at e400 [Size=256k]
Memory at de901000 (32-bit, non-prefetchable [Size=4k]
Expansion ROM at dc000000 [Disabled] [Size=64k]
Capabilities : [40] Power Management Version 2

dmsg |less -

CI Error 0x800000
eth0: PCI error 0x800000 (thousands of times)

lsmod -

Size Used By
natsemi 19552 0

modules.conf -

Alias's ok. One each both natsemi.

Then xstart, used terminal to Network Restart.

Route -

192.168.0.0 * 255.255.255.0 u 0 0 0 eth0
169.254.0.0 * 255.255.0.0 u 0 0 0 eth0
127.0.0.0 * 255.0.0.0 u 0 0 0 lo
Default 192.168.0.254 0.0.0.0 UG 0 0 0 eth0

ifconfig eth0 - (only the tx and rx packets)

RX packets 0 Errors 0 Droopped 0 Overrunns 19735475
TX packets 26 Errors 0 Droopped 0 Overrunns 19735475

When I send network stop, CPU usage drops from 99.8% to 9%.

uhhh forgot this....

modinfo natsemi -

Filename : /lib/modules/2.4.20-8/kernel/drivers/net/netsemi.o
Description : "National Semiconductor DP8381x Series PCI Ethernet Driver"
Author : "Donald Becker <becker@scyld.com>"
Licence : "GPL"
Parm : max_interrupt_work init, description "DP8381x maximum events handled per interrupt"
Parm : mtu_int, description "DP8381x MTU (all boards)"
Parm : debug init, description "DP8381x copy breakpoint for copy-only-tiny-frames"
Parm : options int array (min = 1, max = 8), description "DP8381x bits 0-3 : media type, bit 17 : full duplex"
Parm : full_duplex init array (min = 1, max = 8, description "DP8381x full duplex setting(s) (1)"

what does this mean ??

Subsystem : Netgear : Unknown Device f311

Is it bad ?

Is there any other commands you need me to run ?

I have just been told by a work mate that these cards just simply wont work with linux . I just dont understand why... I thought it was simply drivers because they work just uber slow with the errors.

Thanks for your help Bevis, it looks like this is a no brainer and i'll have to go and get all new network cards....

Any suggestions for cards that work well with RH9 ?

It really pisses me off actually, I wonder what Netgear will say about this. Check this out - http://www.netgear.com/products/details/FA311.asp?view=

I bought this whole setup (Network Cards/Switch) because the rep said it was tried and tested, worked easy first go. Their tech support is hopeless, I still havent heard squat from them. It's great for the 24/7 support that comes with the product(s).

Wow, that is incredibly annoying, and you say that the card is listed as supported in linux on the netgear site? How can they say that?

Actually, and sadly, I didn't really help you at all, and getting stubborn hardware working in linux is my main pet project, to me, it's the one main obstacle linux needs to get over to gain wide acceptance. Windows (understandably) and Mac(can't figure out why) users invariably get a driver on the cd that comes in the box with the device, linux users have to come here, compile,install, and hopefully between all of us we can get them going.

You know what? send all those linux-rejects back, if the cards are that unfriendly to an OS that is so network-centric, then netgear doesn't deserve the revenue.

As for what to get, I'm partial to SMC cards, and/or no-name cards with a realtek chipset. The trick is to find out what chips are on the card *before* purchasing, which is difficult to do.

Yeah I tried a old SMC card that I had in my p100, it worked a treat.

Do SMC do full duplex (200) ? I would like to utilise this as one of the main reasons I go the switch was that I did full duplex

I'll see what I can dig up. I'll also see what the netgear people say when I send them back 16 network cards lol

It's not so much about the brand-name, or lack thereof, it's mostly about the chipset on the card. To me, realtek works and works best, for the money.

oops, didn't answer the "full-duplex" question...I bought a lot of realtek-based pci netcards (made in taiwan) for $5 each (US dollars) which go full-duplex no problem with stock linux realtek modules. As a result, I really like realtek chips.

As a last resort I just did a quick search on this forum for FA311, and there are heaps of threads with similar problems...

eg.

http://www.linuxquestions.org/questi...ighlight=fa311
http://www.linuxquestions.org/questi...1&pagenumber=1
http://www.linuxquestions.org/questi...ighlight=fa311

From reading thru the above plus a few more, I think that I need to use the natsemi driver from this page http://www.scyld.com/network/natsemi.html the natsemi.c file.

Few problems. First is I dont know if this driver is for RH9, second, I have no idea how to put this into the kernel and 'recompile' etc.

Can you help me out with this (last attempt before buying new network cards) driver ??