network problems

[K.C. Chan]

lost connection on remote boxs, when restarting network, things are working ok. Using ifconfig, I noticed that there's a great number of error on the NIC: eth0 Link encap:Ethernet HWaddr 00:01:03:E0:6E:2C
inet addr:24.151.28.208 Bcast:24.151.28.255 Mask:255.255.255.0
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:9694698 errors:0 dropped:0 overruns:367521 frame:0
TX packets:493389 errors:18916 dropped:0 overruns:0 carrier:1
collisions:17088 txqueuelen:100
Interrupt:11 Base address:0xe400

lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
UP LOOPBACK RUNNING MTU:16436 Metric:1
RX packets:9581 errors:0 dropped:0 overruns:0 frame:0
TX packets:9581 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0

Is there a command to determine if the card is operation at 100 MB or 10 MB. and also if it's operation at HD or FD. This is what I get from the dmesg: eth0: Resetting the Tx ring pointer.
NETDEV WATCHDOG: eth0: transmit timed out
eth0: transmit timed out, tx_status 00 status e000.
diagnostics: net 0cd8 media 8c80 dma 000000a0.
Flags; bus-master 1, dirty 493073(1) current 493089(1)
Transmit list 1f672240 vs. df672240.
0: @df672200 length 8000002a status 8000002a
1: @df672240 length 80000047 status 00000047
2: @df672280 length 80000045 status 00000045
3: @df6722c0 length 80000047 status 00000047
4: @df672300 length 80000045 status 00000045
5: @df672340 length 8000004a status 0000004a
6: @df672380 length 8000004a status 0000004a
7: @df6723c0 length 8000002a status 0000002a
8: @df672400 length 8000002a status 0000002a
9: @df672440 length 8000002a status 0000002a
10: @df672480 length 8000002a status 0000002a
11: @df6724c0 length 8000002a status 0000002a
12: @df672500 length 8000002a status 0000002a
13: @df672540 length 8000002a status 0000002a
14: @df672580 length 8000002a status 0000002a
15: @df6725c0 length 8000002a status 8000002a
Does any one know what this mean. Thanks.

[I_M]

Hi

What is NIC driver you are using?

If you are using Intel Etherpro100 as NIC,
and e100 as driver, you may find the information on the dmesg. When the driver start to initialize the NIC, driver show you the information.
But if you are using eepro100 as driver for the same NIC, you can not find the information.
Because it is depend on the driver....

So my suggestion is, check your driver to find where you can find connection speed...say dmesg or /proc? or you may want to try to find the option to "force" the speed and half or full.
You will be able to find such an information inside of driver's source code.....

Happy new year

[Michael Chan_3]

iida, I am using 3com NIC, the driver being loaded is 3c59x. So I don't think it's the driver. I suspect that the port on the hub is hardcoded to 100MB while the DSL Modem only supports 10MB. That's why I want to verify if it's running as 10 or 100.

[Darrell Allen]

Hi Michael,

This doesn't really answer your questions but...

A speed mismatch (100Mb vs 10Mb) between a NIC and a hub (or switch) will not let the devices communicate. A duplex mismatch (Full vs Half) will not keep them from communicating as long as the speed matches. However, collisions will be frequent and collisions will increase rapidly when the volume of traffic increases. Throughput on the link will be greatly affected.

If you can communicate at all then the speed matches but the duplex setting may not.

Darrell

[K.C. Chan]

All, thanks for your help. But does any-one knows what this error: Dec 31 14:11:54 01mgr1 kernel: NETDEV WATCHDOG: "eth0: transmit timed out
Dec 31 14:11:54 01mgr1 kernel: eth0: transmit timed out, tx_status 00 status e000" means.
I am trying to determined why we lost network connections and when restarting the network, everything seems to be ok. No error listed in the log file since then. thanks.

[Paulo A G Fessel]

Michael, check the program vortex-diag.c from this page:

http://www.scyld.com/diag/index.html#pci-diags

There will be instructions of how to compile it to check your link status with great detail. Anyway, I do think you're having duplex-mismatching problem. If you're loading the driver as a module, add the following option to /etc/modules.conf (assuming that you have only a 3c59x card):

options 3c59x full_duplex=x

where x=1 forces full duplex operation, x=0 will not force full duplex operation. If you have compiled the driver statically, then use the following option when booting the kernel

ether=0,0,0x200,eth0

For more details, see the driver's page at http://www.scyld.com/network/vortex.html. If all of this fails, there are links for a 3Com implementation of the drivers for this chipset.

Good luck,

Paulo