I am seeing the following error after my machine has been up
for a while. My eth0 is connected to a switched, local
subnet. There is not a lot of traffic on the interface, maybe
a few 100 Mbytes or so. Taking the interface down and then up
again fixes the problem (until it happens again :)
Here is the relevant section from my kernel log
Mar 1 10:48:44 tela kernel: NETDEV WATCHDOG: eth0: transmit timed out
Mar 1 10:48:44 tela kernel: eth0: transmit timed out, tx_status 00 status e000.
Mar 1 10:48:44 tela kernel: diagnostics: net 0ec0 media 4810 dma 00000021.
Mar 1 10:48:44 tela kernel: Flags; bus-master 1, full 1; dirty 87959(7) current 87975(7).
Mar 1 10:48:44 tela kernel: Transmit list 01252270 vs. c1252270.
Mar 1 10:48:44 tela kernel: 0: @c1252200 length 800000f7 status 000000f7
Mar 1 10:48:44 tela kernel: 1: @c1252210 length 8000010c status 0000010c
Mar 1 10:48:44 tela kernel: 2: @c1252220 length 800000f7 status 000000f7
Mar 1 10:48:44 tela kernel: 3: @c1252230 length 8000010c status 0000010c
Mar 1 10:48:44 tela kernel: 4: @c1252240 length 800000f7 status 000000f7
Mar 1 10:48:44 tela kernel: 5: @c1252250 length 8000002a status 8000002a
Mar 1 10:48:44 tela kernel: 6: @c1252260 length 8000002a status 8000002a
Mar 1 10:48:44 tela kernel: 7: @c1252270 length 8000010c status 0000010c
Mar 1 10:48:44 tela kernel: 8: @c1252280 length 800000f7 status 000000f7
Mar 1 10:48:44 tela kernel: 9: @c1252290 length 8000010c status 0000010c
Mar 1 10:48:44 tela kernel: 10: @c12522a0 length 800000f7 status 000000f7
Mar 1 10:48:44 tela kernel: 11: @c12522b0 length 8000010c status 0000010c
Mar 1 10:48:44 tela kernel: 12: @c12522c0 length 800000f7 status 000000f7
Mar 1 10:48:44 tela kernel: 13: @c12522d0 length 8000010c status 0000010c
Mar 1 10:48:44 tela kernel: 14: @c12522e0 length 800000f7 status 000000f7
Mar 1 10:48:44 tela kernel: 15: @c12522f0 length 8000010c status 0000010c
Mar 1 10:48:44 tela kernel: eth0: Resetting the Tx ring pointer.
Then a similar dump repeats until the interface is recycled.
It appears that the interface was not functioning for some
hours before the message was generated, and it was my attempt
to ping a host on the local subnet that caused the NETDEV
WATCHDOG error to be generated (e.g. the card locked up, but
the kernel didn't notice until I tried to send something on
the wire).
The card is:
eth0: 3Com PCI 3c900 Boomerang 10Mbps Combo at 0x1400,
00:60:08:bd:ab:0e, IRQ 9
I am running kernel 2.4.2, and have seen this error in 2.4.1
as well; not sure about 2.4.0. I do not ever recall
encountering this error with the 2.2.x kernels, though my
network topology has changed, but not my hardware. I know of
at least one other person who gets this same error with a
eth0: 3Com PCI 3c905B Cyclone 100baseTx card. The system is a
P2-300, 128 Mb RAM, running various versions of Linux very
happily for 3 years.
FWIW, IRQ 9 is shared with the bttv module, though the network
lockup doesn't seem to be related to my use of that module. I
was using xawtv last night while the interface was stil active
and functioning. The lockup happened this morning.
Sorry for the long-winded post. Is this a known bug?
Anything I can do to help track it down and squash it if so?
--
cae at bklyn dot org | Caleb Epstein | bklyn . org | Brooklyn Dust Bunny Mfg.
Caleb Epstein wrote:
>
> I am seeing the following error after my machine has been up
> for a while. My eth0 is connected to a switched, local
> subnet. There is not a lot of traffic on the interface, maybe
> a few 100 Mbytes or so. Taking the interface down and then up
> again fixes the problem (until it happens again :)
>
> Here is the relevant section from my kernel log
>
> Mar 1 10:48:44 tela kernel: NETDEV WATCHDOG: eth0: transmit timed out
My guess would be that the driver has decided there's no
link beat on the 10baseT interface and has flopped over
to using 10base2. A fix for this exists in 2.4.2-ac5+,
in the zerocopy patch and in
http://www.uow.edu.au/~andrewm/linux/3c59x.c-2.4.2-pre4.gz
but not in 2.4.2.
You'll need to use
options 3c59x options=0
in /etc/modules.conf to pin the driver down to using a
particular physical interface - disable autoselection.
So could you please upgrade the driver? If problems
remain, please send me a report, as described in the
final section of Documentation/networking/vortex.txt.
Thanks.
-