Mailing List Archive

Support open source code!


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

tlug: NIC data glut



I've been using an old 75mhz 486DX machine as a masquerading firewall in
my office for the past few weeks (many thanks to list members for advice
in getting the network cards set up).  It works, but large transfers crash
the driver on one card.

The local interface is an old Western Digital card; the side exposed to
the LAN is an old Allied Telesys 1700.  It is the AT1700 driver that
crashes.

I have set the machine up with a 2.2.14 kernel compiled with the "slow
processor" option, that I assume is meant to help cope with overruns on
the NIC during data transfers from fast machines, but no joy.

In particular, this is a problem with fetchmail; a large message (over
100k, say) will crash the driver every time I attempt to fetch it.  When
this happens, I have to log onto the firewall machine through the
still-active local NIC, take down the LAN interface, de-install the
driver, re-install the driver and bring up the interface, then read the
message by other means on the server before I can continue fetching the
remainder of my mail through the masquerade machine. 

The one thing from the log files that may point at trouble (this may
solicit a groan or two, so I have my umbrella ready to catch any flying 
tomatoes) is a complaint about clearing the IRQ.  In /var/log/messages,
I get a trail like this when the driver goes down and I go through the
rigamarole mentioned above:

Mar 20 19:32:20 rm-325-1 kernel: eth1: transmit timed out with status 0000, network cable problem?
Mar 20 19:32:20 rm-325-1 kernel: eth1: timeout registers: 4000 8182 4106 e85a 0000 0000 4800 0001.
Mar 20 19:32:55 rm-325-1 kernel: eth1: transmit timed out with status 0000, network cable problem?
Mar 20 19:32:55 rm-325-1 kernel: eth1: timeout registers: 0000 8181 0606 ffff 8888 8181 acac 8383.
Mar 20 19:33:58 rm-325-1 kernel: Trying to free free IRQ15
Mar 20 19:34:06 rm-325-1 kernel: eth1: AT1700 found at 0x300, IRQ 15, address 0000f4d2ac93 10baseT interface.
Mar 20 19:34:06 rm-325-1 kernel: at1700.c:v1.15 4/7/98  Donald Becker (becker@example.com)
Mar 20 19:34:06 rm-325-1 kernel:   AT1700 at 0x300 is unusable due to a conflict onIRQ 15.

There is no complaint of a conflict when the driver is first installed. 
Despite the complaint upon reinstallation, the driver does work (until it
crashes during another large transfer).

While I was setting the machine up in the first place, I couldn't get the
AT1700 card to work at all until I disconnected the cable to the machine's
CD-ROM drive, which, it turned out, also used IRQ 15 (oops).  The
controller is still in the box, although the cable to the drive has been
disconnected.  Could this be related to the failure of the NIC under
"heavy" loads? 

For what it's worth, the installation is Debian "frozen" (potato).

Any advice will be received with great joy and general hooplah.

Cheers,
----
-x80
Frank G Bennett, Jr         @@
Faculty of Law, Nagoya Univ () email: bennett@example.com
Tel: +81[(0)52]789-2239     ()

--------------------------------------------------------------------
Next Nomikai Meeting: April 20 (Thu) Linux Conference 2000 Spring Ed.
Next Technical Meeting: May 13 (Sat) 13:30 Temple University Japan
* Topic: TBD
--------------------------------------------------------------------
more info: http://www.tlug.gr.jp        Sponsor: Global Online Japan


Home | Main Index | Thread Index

Home Page Mailing List Linux and Japan TLUG Members Links