Re: Flaw in 3c59x.c or in Kernel?

[parlin () STARBURSTSOFTWARE COM (Sonny Parlin)]

The only problem I have with ifconfiging down/up is that every time I've run
my test, the machine completely locks, no keyboard, no nothing... So
ifconfig is not an option at that point. Unfortunately, when it crashes, it
crashes BIG. Also, I should have mentioned in my first mail that when using
the 3com driver, this problem goes away. I haven't had a chance to look at
the differences yet, but I'm sure someone will now..   I don't believe the
3com driver ships with any distros (well, at least not redhat), so it's
still a pretty serious problem to consider.

-Sonny

-----Original Message-----
From: dwmalone () maths tcd ie [mailto:dwmalone () maths tcd ie]
Sent: Wednesday, January 05, 2000 5:43 AM
To: Sonny Parlin
Cc: BUGTRAQ () SECURITYFOCUS COM
Subject: Re: Flaw in 3c59x.c or in Kernel?

On Tue, Jan 04, 2000 at 09:21:36AM -0500, Sonny Parlin wrote:

> eth1: Too much work in interrupt, status e481. Temporarily disabling
> functions(7b7e).

We saw this with some Linux machines in college that were connected
to busy 100Mb/s ethernet. Bill Paul is right when he says ifconfiging
down and then up fixes the hang. To work around the problem we changed
max_interrupt_work from 20 to 200 and I don't think they've seen any
hangs since. (You can find this in the .c file for the driver).

These machines were also seeing hard lockups - keyboard stopped
responding. This seemed to stop too when we upped this variable.

        David.