NetBSD Problem Report #9540
Received: (qmail 2845 invoked from network); 4 Mar 2000 23:09:16 -0000
Message-Id: <a04310100b4e7470202e0@[194.196.194.195]>
Date: Sun, 5 Mar 2000 00:08:59 +0100
From: Jakub Nesetril <nesetril@webworx.cz>
To: gnats-bugs@gnats.netbsd.org
Subject: Apache dropping the system into debugger under heavy load
>Number: 9540
>Category: port-i386
>Synopsis: Under heavy connection load, system drops into debugger
>Confidential: no
>Severity: critical
>Priority: high
>Responsible: port-i386-maintainer
>State: closed
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Sat Mar 04 15:12:00 +0000 2000
>Closed-Date: Mon Feb 18 18:06:45 +0000 2002
>Last-Modified: Mon Feb 18 18:06:45 +0000 2002
>Originator: Jakub Nesetril
>Release: 1.4.1
>Organization:
Webworx
>Environment:
AMD-K6-III, brand new install of NetBSD
System: NetBSD server.local.webworx.cz 1.4.1 NetBSD 1.4.1 (GENERIC)
#1: Tue Aug 10 00:03:09 MEST 1999
fvdl@struis:/usr/src/sys/arch/i386/compile/GENERIC i386
>Description:
In order to be sure about reproducing the problem, I used a newly
formated SCSI disc, installed NetBSD 1.4.1 from boot-floppy and CD,
set up network (editing only /etc/myname /etc/mygate
/etc/ifconfig.fxp0 /etc/resolv.conf). I unpacked a fresh pkgsrc tree
(29th of February 2000), switched to www/apache and compiled a brand
new Apache 1.3.11 downloaded from the FTP servers. Placed it into
rc.local, minimally edited the /usr/pkg/etc/httpd/httpd.conf, reboot.
Now, having a brand new fresh NetBSD with Apache, I ran the ab
(Apache benchmark) utility with following paramteres: ab -n 10000 -c
50 http://localhost/manual/index.html
Repeating this two or three times causes the httpd to refuse
connection, repeating it once more causes the system to drop into
debugger (the proccess marked as current is httpd). Errors in
/var/log/httpd/error_log are on 4 lines, all of them being:
[Error] (22)Invalid argument: getsockname
It should be noted that sometimes before dropping into debugger,
there is a syslogd message about system panic and some pmaloc
problems (I think) - certainly memory-related function. It should
also be noted that I have 64 megs, NetBSD uses about 4 megs normally,
when I run the benchmarks, it sometimes climbs to 20, but not more.
There is no disk activity (i.e. no swapping).
Also, there is very odd inetd behavior, which is harder to reproduce
exactly (upon ftp or telnet connection - waiting 5 minutes before
inetd launches ftpd or telnetd, socket is not connected errors and so
on).
HW configuration:
AMD-K6-III. PS/2 kbd and mouse. SCSI HD (Quantum) connected via
Adaptec AHA 2940 PCI card, network on Intel EtherExpress 10+/100B PCI
card. I'll send a /var/log/messages upon request.
>How-To-Repeat:
Plain new 1.4.1 i386 install, plain new Apache install, changing only
bare minimum, running ab -n 10000 -c 50 http://localhost/ a few
times.
>Fix:
>Release-Note:
>Audit-Trail:
State-Changed-From-To: open->closed
State-Changed-By: chs
State-Changed-When: Mon Feb 18 09:14:31 PST 2002
State-Changed-Why:
I can't reproduce this problem with either 1.5-branch or -current.
there have been a bunch of memory-management fixes since 1.4.x,
so it's reasonable to believe that this problem has been fixed.
>Unformatted:
X-send-pr-version: 3.95
(Contact us)
$NetBSD: query-full-pr,v 1.51 2026/08/10 02:28:17 riastradh Exp $
$NetBSD: gnats_config.sh,v 1.10 2026/05/13 22:00:09 riastradh Exp $
Copyright © 1994-2026
The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.