Strange network problem, overrun and frame errors

I was wondering if anyone might have an idea on the reason of some
error messages and traffic slowdown I sometimes experience
on my Linux system (Mandrake 8.2).
I get a couple of frame/overrun errors per day

eth0 Link encap:Ethernet HWaddr 00:10C:60:C7C
inet addr:192.168.1.100 Bcast:192.168.1.255 Mask:255.255.255.0
UP BROADCAST NOTRAILERS RUNNING MULTICAST MTU:1500 Metric:1
RX packets:128485 errors:23 dropped:0 overruns:23 frame:23
TX packets:99175 errors:54 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:100
RX bytes:126063867 (120.2 Mb) TX bytes:15081824 (14.3 Mb)
Interrupt:11 Base address:0xec00

and in the log file I sometimes get messages of type:

Oct 12 22:00:21 kaktus kernel: NETDEV WATCHDOG: eth0: transmit timed
out
Oct 12 22:00:21 kaktus kernel: eth0: Transmit timed out, status fc664
010, CSR12 00000000, resetting...

and a significant slowdown. I never seem to completely lose the connection
though. A significant slowdown seems to happen when there is a lot of
network demanding applications running simultaneously (ftp,apache,
sshd,telnet,browsing). Sometimes the NIC even get stalled and the only
way to recover is through reboot.

The network driver I have is tulip. The NIC is connected to a Linksys
BEFSR41 Router/switch.
I have searched for information and some suggested that the errors
might be caused by duplex mismatch between the NIC and the router
but it looks like they both are set to auto negotiation mode and the
connection seems to be up in full duplex 100 Mps (as indicated
by the control LED's on the router.
I tried mii-tool -F 100baseTx-FD without success.

I noticed that the NIC is assigned Interrupt 11, which is
also used by USB drivers. (The interrupt assignments are
all from the default settings. My system had Mandrake preinstalled).
Can the interrupt sharing cause a conflict or deadlock under some
conditions ? I was also thinking about forcing the connection mode
into the old 10 Mps/HD mode but haven't tried that yet.

Any help or references would be appreciated.

/
Karl D.

On 15 Oct 2003 00:31:40 -0700, Karl D <(E-Mail Removed)> wrote:
> I was wondering if anyone might have an idea on the reason of some
> error messages and traffic slowdown I sometimes experience
> on my Linux system (Mandrake 8.2).
> I get a couple of frame/overrun errors per day
>
> eth0 Link encap:Ethernet HWaddr 00:10C:60:C7C
> inet addr:192.168.1.100 Bcast:192.168.1.255 Mask:255.255.255.0
> UP BROADCAST NOTRAILERS RUNNING MULTICAST MTU:1500 Metric:1
> RX packets:128485 errors:23 dropped:0 overruns:23 frame:23
> TX packets:99175 errors:54 dropped:0 overruns:0 carrier:0
> collisions:0 txqueuelen:100
> RX bytes:126063867 (120.2 Mb) TX bytes:15081824 (14.3 Mb)
> Interrupt:11 Base address:0xec00
>
> and in the log file I sometimes get messages of type:
>
> Oct 12 22:00:21 kaktus kernel: NETDEV WATCHDOG: eth0: transmit timed
> out
> Oct 12 22:00:21 kaktus kernel: eth0: Transmit timed out, status fc664
> 010, CSR12 00000000, resetting...
>
> and a significant slowdown. I never seem to completely lose the connection
> though. A significant slowdown seems to happen when there is a lot of
> network demanding applications running simultaneously (ftp,apache,
> sshd,telnet,browsing). Sometimes the NIC even get stalled and the only
> way to recover is through reboot.
>
> The network driver I have is tulip. The NIC is connected to a Linksys
> BEFSR41 Router/switch.

A floating MTU does sound strange. If you just had packet loss, I would
think it would be more irregular (just some lost pings).

What nic (or related output from 'lspci -v')? I had trouble using Netgear
FA310TX with newer tulip modules (so I stuck with old_tulip module in
latest 2.2 kernel until PC died). Newer tulip modules with that nic would
either not work until it received traffic (latest 2.2 tulip) or would stop
working overnight (early 2.4 tulip).

I also had trouble with some distros trying to use tulip module for old
10baseT integrated real DEC chip (in which case de4x5 module worked
better). But that does not sound like your problem if it is 10/100 nic.

I have had incompatiblities with switch ports on Linksys EFSP42
switch/printserver. The ethernet port of my Linksys WAP11 v2.2 will not
work on it, and a 3com nic gets 20-100% packet loss when connected to it.
I also had an SMC router/switch that was incompatible with PC card nics
unless they were switched to 10baseT (full duplex worked). You might try
following just to see if it makes a difference:

mii-tool -A 10baseT-FD,10baseT-HD eth0

--
David Efflandt - All spam ignored http://www.de-srv.com/
http://www.autox.chicago.il.us/ http://www.berniesfloral.net/
http://cgi-help.virtualave.net/ http://hammer.prohosting.com/~cgi-wiz/

David, thanks for your input.

Some more information:

> What nic (or related output from 'lspci -v')? I had trouble using Netgear
> FA310TX with newer tulip modules (so I stuck with old_tulip module in
> latest 2.2 kernel until PC died).

lspci -v
:
00:0f.0 Ethernet controller: Linksys Network Everywhere Fast Ethernet
10/100 model NC100 (rev 11)
Subsystem: Accton Technology Corporation: Unknown device 1216
Flags: bus master, medium devsel, latency 32, IRQ 11
I/O ports at ec00 [size=256]
Memory at d8000000 (32-bit, non-prefetchable) [size=1K]
Expansion ROM at <unassigned> [disabled] [size=128K]
Capabilities: <available only to root>

The PC came with this built in ethernet controller. I don't even know
what
chip set is used in this NC100 controller.
Driver version (This was the one distributed in MDK 8.2, Kernel is
2.4.18):
dmesg | grep Tulip
Linux Tulip driver version 0.9.15-pre9 (Nov 6, 2001)

A side note:
After some experimentation with mii-tool setting yesterday
and harddrake (Mdk's network administration tool)
the connection now seems to have been permanently broken.
I cannot even start dhcpcd to connect with the router.
Running ifup returns: Link is down after a long long time.
I'm checking all the config files and network-scripts
in /etc from back-up CD but I haven't found any differences.
I'v tried connecting the ethernet cable to other ports of the
router without success. (The router is working since I have both
a Win98 machine as well as a Sun Sparc10 connected and working
in dhcp mode).
Is it possible that various reconfiguration through the
tulip driver can cause permanent damages/changes on the NIC,
which is not resolved through power off/reboot? I guess
my question is if the NIC settings are stored in
some non-destructive memory (EEprom/Flash).

Thanks

/
Karl D.

(E-Mail Removed) (David Efflandt) wrote in message news:<(E-Mail Removed)>...
> On 15 Oct 2003 00:31:40 -0700, Karl D <(E-Mail Removed)> wrote:
> > I was wondering if anyone might have an idea on the reason of some
> > error messages and traffic slowdown I sometimes experience
> > on my Linux system (Mandrake 8.2).
> > I get a couple of frame/overrun errors per day
> >
> > eth0 Link encap:Ethernet HWaddr 00:10C:60:C7C
> > inet addr:192.168.1.100 Bcast:192.168.1.255 Mask:255.255.255.0
> > UP BROADCAST NOTRAILERS RUNNING MULTICAST MTU:1500 Metric:1
> > RX packets:128485 errors:23 dropped:0 overruns:23 frame:23
> > TX packets:99175 errors:54 dropped:0 overruns:0 carrier:0
> > collisions:0 txqueuelen:100
> > RX bytes:126063867 (120.2 Mb) TX bytes:15081824 (14.3 Mb)
> > Interrupt:11 Base address:0xec00
> >
> > and in the log file I sometimes get messages of type:
> >
> > Oct 12 22:00:21 kaktus kernel: NETDEV WATCHDOG: eth0: transmit timed
> > out
> > Oct 12 22:00:21 kaktus kernel: eth0: Transmit timed out, status fc664
> > 010, CSR12 00000000, resetting...
> >
> > and a significant slowdown. I never seem to completely lose the connection
> > though. A significant slowdown seems to happen when there is a lot of
> > network demanding applications running simultaneously (ftp,apache,
> > sshd,telnet,browsing). Sometimes the NIC even get stalled and the only
> > way to recover is through reboot.
> >
> > The network driver I have is tulip. The NIC is connected to a Linksys
> > BEFSR41 Router/switch.
>
> A floating MTU does sound strange. If you just had packet loss, I would
> think it would be more irregular (just some lost pings).
>
> What nic (or related output from 'lspci -v')? I had trouble using Netgear
> FA310TX with newer tulip modules (so I stuck with old_tulip module in
> latest 2.2 kernel until PC died). Newer tulip modules with that nic would
> either not work until it received traffic (latest 2.2 tulip) or would stop
> working overnight (early 2.4 tulip).
>
> I also had trouble with some distros trying to use tulip module for old
> 10baseT integrated real DEC chip (in which case de4x5 module worked
> better). But that does not sound like your problem if it is 10/100 nic.
>
> I have had incompatiblities with switch ports on Linksys EFSP42
> switch/printserver. The ethernet port of my Linksys WAP11 v2.2 will not
> work on it, and a 3com nic gets 20-100% packet loss when connected to it.
> I also had an SMC router/switch that was incompatible with PC card nics
> unless they were switched to 10baseT (full duplex worked). You might try
> following just to see if it makes a difference:
>
> mii-tool -A 10baseT-FD,10baseT-HD eth0