From owner-freebsd-hackers Tue Mar 26 15:41:30 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id PAA06732 for hackers-outgoing; Tue, 26 Mar 1996 15:41:30 -0800 (PST) Received: from jbrann.dialup.access.net (jbrann.dialup.access.net [166.84.193.118]) by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id PAA06724 Tue, 26 Mar 1996 15:41:25 -0800 (PST) Received: (from jbrann@localhost) by jbrann.dialup.access.net (8.6.12/8.6.12) id SAA02049; Tue, 26 Mar 1996 18:40:02 -0500 Message-Id: <199603262340.SAA02049@jbrann.dialup.access.net> Subject: iijppp problems - related to interrupts? To: questions@freebsd.org Date: Tue, 26 Mar 1996 18:40:01 -0500 (EST) Cc: hackers@freebsd.org From: John Brann Reply-To: John Brann Organisation: Not while I'm at home X-Mailer: ELM [version 2.4ME+ PL13 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-hackers@freebsd.org X-Loop: FreeBSD.org Precedence: bulk Hi, I've been using iijppp, and its dial-on-demand facility, for over six months, pretty much since it first became available in a snapshot release. I have encountered occasional sig 10s - which is annoying since it forces me to restart the ppp program manually. I've tried investigating this problem, without much success. I rebuilt the program with -g, and tried gdb on the core file, with no consistent results. However, one trend is definite. The lifetime of a ppp process before I get a sig 10 is related to the uptime of my machine. Abends in the first 24 hours are very rare, but by the time the machine has been up for a week, ppp won't usually last 24 hours. One other thing is consistent - the crash occurs when the ppp program is very busy, for instance when it first connects to my ISP and goes through the handshaking procedure (this is the most common point for a failure) or transferring a large file or web page from a fast site. Today I was looking at my system performance using Xperfmon++, to se if that could give me a clue - and it may have done, hence the title of my e-mail (finally, he gets to the point :-). When not connected through ppp, my machine runs at about 250 interrupts / sec. This seems to be a fairly sensible level for a machine running X with half a dozen windows open. If I ping a remote site, ppp dials and connects, and at the moment of connection, Xperfmon++ reports over 4500 interrupts / sec. This seems a bit high. Contacting a graphics-intensive web site can cause spikes at over 11000 /sec. All this is happening on a single dial-up line which normally connects at 14.4. port speed is 115200, kernel reports of overflows are very rare (maybe once a month, usually when I'm doing a big make and downloading at the same time). Hardware is Pentium 60 with 32Mb memory and Adaptec 2940 driving the peripherals. The async card is (very) generic. Is anyone else suffering similar sig 10 abends with ppp? Is the interrupt level extraordinary? If so, what can I do - is it a hardware problem? I'm copying this to 'hackers' since I think it's a bit advanced for 'questions'. TIA John -- Beavis and Butt-Head; Vladimir and Estragon for the '90s. finger jbrann@panix.com for pgp public key