Mr. Mendelson,
Does your 3c905B share an interrupt with another device?
If so:
You can, in Mr. Donald Beckers drivers, just increase the
"Max_Work_At_Interrupt = 20;" statement to a 1000.
This is a pragmatic fix but the driver is not ready for the
shared irq case and Alan Cox has been nice enough to
inform me that it would be a hard job to rewrite the driver
to actually be able to share an irq with another peripheral.
[I would like to see Mr. Becker comment on this once
for I am curious if not one of his beowulf machines with
a 3com NIC has ever 'suffered' from a shared interrupt.]
Yours,
Luuk van der Duim
----- Original Message -----
From: "Jordan Mendelson" <jordy@napster.com>
To: <linux-kernel@vger.rutgers.edu>
Sent: Monday, March 06, 2000 7:11 PM
Subject: 3c59x & Cat 4003 Ethernet Lockups
> I'm not sure this is a Linux problem, however I thought I'd shoot out
> some
> email anyway.
>
> We have a set of Catalyst 4003 switches which we connect to various
> dual-processor Linux machines w/ 3com 590B ethernet adapters. We've been
> seeing random lockups of the switch ports where both the switch and the
> Linux boxes appear to see eachother, but refuse to forward packets
> between them.
>
> Doing an ifconfig eth0 down up and readding the default route appears to
> fix the problem, however I've found it odd that rebooting the switch
> does not indicating that the Linux boxes network driver must be stuck in
> a vary strange state.
>
> Kernel is 2.2.14 SMP w/ various internal fixes to handle the number of
> persistant connections we have. We've tried both Donald Becker's driver
> as well as 3com's (which is what we are currently using).
>
> One interesting thing is that with the 3com driver, the 'carrier'
> counter on ifconfig keeps incrementing constantly. All the cables have
> been replaced numerous times so we are quite sure it's not that. The
> ports have been forced into 100 Mbps full-duplex. Also it migh be
> interesting to note that only the ports connecting to our messaging
> servers, which handle > 10K simultaneous connections are locking.
> Identical machines running other services aren't.
>
> I'd appreciate any help you could give,
>
> Jordan Mendelson
> Napster, Inc.
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.rutgers.edu
> Please read the FAQ at http://www.tux.org/lkml/
>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue Mar 07 2000 - 21:00:20 EST