[vortex] Re: Query: Any solution to 3c905B problem?

Iain O'Cain ec@intranet.org
Thu Jan 10 08:50:01 2002


On Sun, 6 Jan 2002, Bogdan Costescu wrote:

> > > I've been having similar trouble, first with a Tulip card, and now with a
> > > 3c905C.  Now I'm upgrading to 2.4.17 and hoping the problem goes away, but
> > > every 20-30 seconds, I get a ping of 20-30 seconds.  :-(
>
> Please confirm: with both a Tulip and a 3Com card you observed the
> problem ? What is at the other end of the cable ? A Cisco switch by
> any chance ? When the problem appears, could you please run "arp -n"
> and see if you get any "incomplete" entries ?

Okay, I've just caught this in progress.  It seems the same with the Tulip
card I had in there, and with the 3Com 905C card I replaced that with.
I've also upgraded from a 2.4.10 kernel to 2.4.17 with no apparent result,
and even recompiled 2.4.17 again with 586 optimization instead of K6-2,
since I vaguely recall someone saying the latter could introduce problems
(but if that was ever true, I suspect it's long since fixed).  In the more
recent 2.4.17 kernel, I also turned off any other option that seemed to me
like it might reasonably introduce this problem.

Yes, "arp -n" reports "incomplete" entries:

Address                  HWtype  HWaddress           Flags Mask            Iface
10.0.0.10                        (incomplete)                              eth0
10.0.0.1                 ether   00:60:67:32:69:64   C                     eth0
10.0.0.2                         (incomplete)                              eth0

Where there's actually no host at "10.0.0.1" since months ago (so why do I
have a MAC address for it?), and I get a line like the other two for any
host I try to ping, whether there's really a host at the address or not.

What I get from mii-daig is:

Using the default interface 'eth0'.
Basic registers of MII PHY #24:  3000 782d 0040 6176 05e1 45e1 0003 0000.
 The autonegotiated capability is 01e0.
The autonegotiated media type is 100baseTx-FD.
 Basic mode control register 0x3000: Auto-negotiation enabled.
 You have link beat, and everything is working OK.
 Your link partner advertised 45e1: Flow-control 100baseTx-FD 100baseTx
10baseT-FD 10baseT, w/ 802.3X flow control.
   End of basic transceiver informaion.

The switch it's talking to is the cheapo D-Link 16-port unit with the two
Broadcomm chipset cards in it.  None of the other systems on that switch
are having any trouble with Tulip or RTL8139 (yuck) cards and the same
(first) 2.4.17 kernel.  None of the others are identically configured,
either, though.

Any pointers would be most welcome!  I'm not a list subscriber, but I'll
watch the archives for the next while.  A Cc would still help.  :-)

Thanks -

- Iain

-- 
Iain O'Cain <ec@intranet.org>       - IO10 -      http://ec.intranet.org/

"MS is a disease of the central nervous system which affects Canadians
more frequently than most other countries in the world."
  		-- http://www.mssociety.ca/en/