[vortex-bug] 3c59x 0.99H and 3c905B (Cyclone) problems

[Jonathan Feldman]

(Somewhat off-topic, but related to this problem) - Naren says below that you think it's bad to disable STP.  I'd have to disagree; if you don't _need_ spanning tree protocol, cisco discovery protocol, Mom's protocol, or whatever other protocol, it's a BAD IDEA to have it enabled.  Gratuitous protocols (and network traffic) are the root of all evil.

(Back on topic) Bob, you repeat that you are OK on your 10Mb hub; and other folks on this list have suggested that you turn off STP.  Have you done so?  Like others on this list, we've had problems with Cisco's STP implementation.

But you don't say whether your switch is "smart" (can you telnet to it and configure it?) or whether it's dumb/autosensing.  If it is smart, STP is likely; if it is not smart, STP is highly unlikely in your scenario, and the 10/100 sensing is more likely the culprit.

Try forcing 10 or 100Mbps in the 3com driver, in that case.

--Jonathan
Chief Technical Manager
Chatham County ICS Department
Author, "Teach Yourself Network Troubleshooting," "Network+ Exam Guide"
http://feldman.org

>>> Narendra Sankar <naren@xstreamlogic.com> 08/03 4:49 PM >>>
We had this problem here a long time ago, and this is what I discovered -

The 3com cards seem to have problems with the spanning tree protocol 
updates that most managed switches go through when they see a new link 
activated at 100Mbps. We had a cisco 2900 catalyst switch. There are two 
solutions. One is to disable STP on the switch if possible. I think this is 
bad to do. The solution we implemented was this. Go to the switch vendor 
and find out what their time to complete STP is. In case of the catalyst it 
is 30 secs. Then in your network startup scripts, in our case the syslog 
script, insert a sleep n ( where n is greater than the STP time). This 
forces linux to wait till the switch has finished initializing before 
continuing. This works perfectly for us using both the 3c59x driver and the 
3c90x driver from 3Com. Our NIC is a 3c905B. Interestingly we found this 
problem only using linux. Windows was quite happy to ignore the link not up.

Hope this helps

naren

At 04:39 PM 8/3/2000 -0400, Bob Ramstad wrote:
>howdy.  please send any replies directly to me as well as to the list
>as i am not a subscriber.
>
>i'm running the vanilla RedHat 6.2 distribution on a machine with a
>single 3c905B card and i'm having a fairly mysterious problem.
>
>the system in question was running just fine when it was connected to
>a 10 mbps hub.  we recently replaced this hub with an autosensing
>10/100 mbps switch and as a result, this particular system cannot see
>the LAN at all.  (this is our only Linux box.)
>
>i've tried rebooting, unloading and reloading the driver module and
>running linuxconf all to no avail.
>
>an interesting piece of data.  when i disconnect the cable from the
>10/100 mbps switch, and connect it to the 10 mbps hub, everything
>works great.  (this is true even without rebooting!)
>
>all of our Windows boxes are working great at 100 mpbs and are sharing
>files and printers no problem as well as accessing the Internet.
>
>i have taken a look at /var/log/messages and the only error there is
>
>eth0: Updating statistics failed, disabling stats as an interrupt
>source.
>
>this error only occurs when connecting to the 10/100 mbps switch, and
>does not occur when connecting to the 10 mbps hub.
>
>suggestions on things to try would be most helpful.  i'm happy to
>provide diagnostic data as well.  thanks in advance!
>
>-- Bob
>
>
>_______________________________________________
>vortex-bug mailing list
>vortex-bug@scyld.com 
>http://www.scyld.com/mailman/listinfo/vortex-bug 


_______________________________________________
vortex-bug mailing list
vortex-bug@scyld.com 
http://www.scyld.com/mailman/listinfo/vortex-bug