From owner-freebsd-current Fri Oct 26 10:12:35 2001 Delivered-To: freebsd-current@freebsd.org Received: from hal-5.inet.it (hal-5.inet.it [213.92.5.24]) by hub.freebsd.org (Postfix) with ESMTP id DF7A937B407 for ; Fri, 26 Oct 2001 10:12:27 -0700 (PDT) Received: (from root@localhost) by hal-5.inet.it (8.11.1/8.11.1) id f9QHCRP175292 for ; Fri, 26 Oct 2001 19:12:27 +0200 Received: from unknown(213.92.1.190) by hal-5.inet.it via I-SMTP id s-213.92.1.190-jYbIPg; Fri Oct 26 19:12:27 2001 Received: (qmail 17942 invoked by uid 1000); 26 Oct 2001 17:12:24 -0000 Date: Fri, 26 Oct 2001 19:12:24 +0200 From: Andrea Campi To: David Malone Cc: freebsd-current@FreeBSD.org Subject: Re: -CURRENT freeze under high load Message-ID: <20011026191224.B7800@webcom.it> References: <20011024153151.A75454@webcom.it> <20011026181612.A7800@webcom.it> <20011026175237.A33082@walton.maths.tcd.ie> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: <20011026175237.A33082@walton.maths.tcd.ie>; from dwmalone@maths.tcd.ie on Fri, Oct 26, 2001 at 05:52:37PM +0100 X-Echelon: BND CIA NSA Mossad KGB MI6 IRA detonator nuclear assault strike Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG On Fri, Oct 26, 2001 at 05:52:37PM +0100, David Malone wrote: > On Fri, Oct 26, 2001 at 06:16:12PM +0200, Andrea Campi wrote: > > All my problems are now gone. This sort of makes sense to me, as the culprit, > > qmail, is quite socket intensive. > > > > Anybody has any idea how to properly fix? > > This patch changed quite a few things, so it's not obvious exactly > what is causing the problem. I know. I'd like to look deeper into the issue, but from a quick glance at the code, I don't think I could figure out a way to separate those "things" and try each one. Do you happen to have separate patches for them, that I could try? > Do you know if qmail does any discriptor passing? The code makes > discriptor passing a bit more mbuf intensive, so it's possible that > you're running your machine out of mbufs. I know qmail tends to > run machines as hard as it can, so it may have run the machine into > the ground. I'm not 100% sure of how to check, but a grep SOL_SOCKET * in the sources didn't return anything. Also, from what I can understand without really reading all of that #@#@ DJB code, qmail mainly uses pipe and 2 or 3 fifos. AFAIK your commit wasn't intended to change that, but is it possible that a bug did sneak in? Anyway, both ways I can trigger the bug (find . -type f | xargs mutt, and actually running fetchmail -a) do generate a LOT of work, so it's actually possible that your diagnosis (mbuf exhaustion) is correct; trouble is, this shouln't hurt the machine to the point I can't even enter DDB. > > Also, are you running on alpha or i386? i386, IBM Thinkpad 570E (not that it being a laptop makes any difference, of course ;-)) Bye, Andrea -- Intel: where Quality is job number 0.9998782345! To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message