I setup a NFS server between my two boxes in a LAN.
I put some Documents and my MP3 collection on NFS share so both computers can use it, but when i run XMMS on the client machine, the music starts up fine but after about 3 minutes i get:
Aug 8 15:33:02 SlackBox kernel: NETDEV WATCHDOG: eth0: transmit timed out
on the Server machine, and the music stops playing on the Client.
So my question is how can i fix this, so i can listen to MP3's over NFS?
Thank you in advance!
-NSKL

A search in Google for NETDEV WATCHDOG show some posts
that suggest that the problem can be related to a conflict with irq conflicts.

So a check in /proc/interrupts or lspci -v can help.

Hmm, i also noticed my Network dies sometimes, and i have to bring down eth0 interface on both computers to restart it (ifconfig eth0 down). This is maybe the real cause of NFS stopping to deliver files to Client box.
lspci -v lists all my PCI devices, but not their IRQs.
less /proc/interrupts gives this:
CPU0
0: 317355 XT-PIC timer
1: 6565 XT-PIC keyboard
2: 0 XT-PIC cascade
3: 120301 XT-PIC serial
5: 3198 XT-PIC eth0
8: 1 XT-PIC rtc
9: 40814 XT-PIC usb-uhci, usb-uhci, usb-uhci
11: 263915 XT-PIC EMU10K1, nvidia
14: 35780 XT-PIC ide0
15: 256150 XT-PIC ide1
NMI: 0
ERR: 0

So how can i find out if there are any IRQ conflicts
(that are not in /proc/interrupts)? Or what else could be causing this problem?
Any suggestions?
Thank you
-NSKL

What model of ethernet card do you have?

To check if there are a irq conflict with the lan card i suggest you
to remove the card, reboot and do cat /proc/interrupts to check if the irq 5 is used by other device.

also in:

http://www.ibiblio.org/pub/Linux/doc...Ethernet-HOWTO

there are something like your problem:

Problem: The ne2k-pci.c driver reports error messages like timeout waiting for Tx RDC with a PCI NE2000 card and doesn't work right.

Reason: Your card and/or the card to PCI bus link can't handle the long word I/O optimization used in this driver.

Solution: Firstly, check the settings available in the BIOS/CMOS setup to see if any related to PCI bus timing are too aggressive for
reliable operation. Otherwise using the ISA/PCI ne.c driver (or removing the #define USE_LONGIO from ne2k-pci.c) should let you use the card.

It's a Cnet Fast Ethernet
Theres a floppy disk with drivers that comes with it (for Linux) so i'll try installing those first (since now it's using Tulip kernel module), and then i'll try removing the card as you suggested.
Thanks for help
-NSKL

All right, i removed the cards to check for conflicts, and a thing i don't understand happens..
When i remove the card, usb-uhci moves to interrup 5, where eth0 was, but when the card is plugged into the PCI port, usb-uhci moves back to interrup 9 and eth0 is back to interrupt 5 where it was before. Could this be the problem that sometimes halts my network (eth0 card)?

On the other computer if i remove the eth card nothing takes it's place in /proc/interrupts. So i guess that one is OK.

Now what should i do, if this is the problem at all?

I also tried loading the particular modules for this card, but i couldnt. The module is on a floppy disk, it's cnet~1.gz. I gunzipped it and put it in /lib/modules/kernelversion/kernel/drivers/net but modprobe reports module not found when i try to load it. I also think it is not necessary to load this module since the card works even with the Tulip module...

Any help appreciated. thanks again
-NSKL

This site has drivers with source code, so you can compile it to tailor to your box.
Please check if they have more recent drives:

http://www.cnet.com.tw/download/index-f100.html

About your problem with loading the modules, this is from the ethernet howto:

"
The first thing you need in your conf.modules file is something to
tell modprobe what driver to use for the eth0 (and eth1 and...)
network interface. You use the alias command for this. For example,
if you have an ISA SMC EtherEZ card which uses the smc-ultra.o driver
module, you need to alias this driver to eth0 by adding the line:

alias eth0 smc-ultra

The other thing you may need is an options line indicating what
options are to be used with a particular module (or module alias).
Continuing with the above example, if you only used the single alias
line with no options line, the kernel would warn you (see dmesg) that
autoprobing for ISA cards is not a good idea. To get rid of this
warning, you would add another line telling the module what I/O base
the card is configured to, in this case say the hexidecimal address
0x280 for example.

options smc-ultra io=0x280 irq=5

"

It can help if you add these lines and use an irq not used in your box, as irq 4, 6 or 7.

Seems to me that there arent irq conflicts, because you dont have reported problems with another device and
the device usb-uhci can use other irq so it dont conflicts with the lan card,
but i suggest to change the irq of the lan card (using the options line in conf.modules) to discard this posibility.


This other site have a great amount of information about Linux Ethernet device drivers:

http://www.scyld.com/network/

At last, you can use another lan card to see if the problem is related to the lan card or not.

If nothing of this help, please post the information show by dmesg, ifconfig and the hardware-software
configuration of both boxs, hoping that this gave us more clues about your problem.

OK, i was looking at dmesg more closely.
When network stops working i get that "NetDev: Watchdog: Transmit times out" on the server, and "NFS server not responding, still trying" on the client
Then i bring down and back up eth0 card on the server and after a couple of secinds the client recognizes NFS server is back up and continues working most of the time, but sometimes it happens that dmesg reports "NFS: Can't get a request slot" on the client machine, and in that case i have to reboot the client machine. In any case, im quite sure the problem is on the server machine. I will try loading the module with a different IRQ, etc... because it was setup that way in Slack 8.0 and these things didnt happen. (Now im in slack 8.1)

Thanks again for suggestions and help, i will try these things and see if it fixes my problem.
Thanks again
-NSKL

I remenber i have randomly this error message before in my home lan:

NFS: Can't get a request slot

and the transfer rate was very inestable.

After endless experimentation without results a fortuitous success
(my sister damage the lan cable) forced me to build another cable.
This time all errors dissapear, so i suppose the problem was the first cable (im not very good with manual work) ;-)

I suggest you to replace the cable to see if in your case the problem dissapear.

Also, if you want to try change the irq of the lan card,
another option to do it is using the ifconfig command.
See the manpages for the syntax.

I'm sorry that my recomendation to take the lan card out the box have created more problems (printer and RTC) to you . . . . . ^_^

zelgadis: Ok, as soon as some store opens (in August in Italy, everything is closed!) I'll get another cable, and see if that helps. It's possible, because if i leave the LAN alone (no traffic) for a hour or so, and then try to ping, the first time i get is horrible! about 900 ms and the usual is about 1, 2 ms so the cable might be damaged.
Ok, thanks for help Again and dont worry about othr problems, its always fun fixing them.. im just wondering will i ever get to the point to actually use my Linux for normal everyday use not only configuring it.
Thanks again,
-NSKL