Re: Linux 2.2.2pre4 SMP crash (wait_on_bh)

Simon Kirby (sim@netnation.com)
Sat, 20 Feb 1999 16:59:02 -0800 (PST)


On Sat, 20 Feb 1999, Jordan Mendelson wrote:

> Andrea Arcangeli wrote:
> >
> > On Sat, 20 Feb 1999, Terry Katz wrote:
> >
> > >I've experienced the same problem... 2.2.1, and 2.2.2pre2... There have been
> > >a number of people who have brought this up... No one has really been
> > >helpful in an answer as to how to fix. Linus suggested 2.2.2pre2 to me a
> > >short time ago, but just recently it happened with that so ...
> >
> > Are you using the shaper. The shaper had a SMP race that could cause
> > exactly this kind of SMP hang. I just fixed it and you can find my fix in
> > both linux-kernel and linux-net.
>
> No traffic shaper here.. my current running config is:

No shaper here, either. Because I noticed that the traces seem to be the
same and each has tcp_v4_hash(), and because I noticed that a machine we
have here with identical hardware that has less net traffic hasn't ever
had the problem, I ran some tests with multiple processes connecting and
closing thousands of TCP sockets simultaneously on my SMP celeron box at
home. It's been going for an hour or so now, and I haven't been able to
reproduce anything.

I noticed you are using a buslogic SCSI controller -- the machine that is
having problems here is using an AIC7xxx, so perhaps the low level driver
isn't the problem. What Ethernet card are you using? I noticed
"EEPRO100" in your conf, which is what we are using here.

Are you running a lot of net traffic through the machine? This machine
that has had the problem 4 times in the last month or so sends ands
receives about 500,000 packets per hour. It sends and receives mail,
handles POP3 clients, runs a big nameserver, and has a few people logged
in occasionally to read mail.

Hmm...*ponder* I was testing sockets locally. Perhaps it has to go over
the NIC...

Simon-

| Simon Kirby | Systems Administration |
| mailto:sim@netnation.com | NetNation Communications |
| http://www.netnation.com/ | Tech: (604) 684-6892 |

-
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/