From owner-freebsd-questions@FreeBSD.ORG Fri Oct 23 18:16:49 2009 Return-Path: Delivered-To: freebsd-questions@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 64C12106566B for ; Fri, 23 Oct 2009 18:16:49 +0000 (UTC) (envelope-from cpghost@cordula.ws) Received: from fw.farid-hajji.net (fw.farid-hajji.net [213.146.115.42]) by mx1.freebsd.org (Postfix) with ESMTP id 5CF558FC19 for ; Fri, 23 Oct 2009 18:16:48 +0000 (UTC) Received: from phenom.cordula.ws (phenom [192.168.254.60]) by fw.farid-hajji.net (Postfix) with ESMTP id 4AF4837466; Fri, 23 Oct 2009 20:16:44 +0200 (CEST) Date: Fri, 23 Oct 2009 20:16:43 +0200 From: cpghost To: Henrik Hudson Message-ID: <20091023181643.GB1565@phenom.cordula.ws> References: <20091021204448.GA7125@phenom.cordula.ws> <20091022232020.GA7661@alucard.int.rhavenn.net> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20091022232020.GA7661@alucard.int.rhavenn.net> User-Agent: Mutt/1.5.20 (2009-06-14) Cc: freebsd-questions@freebsd.org Subject: Re: rTorrent + FreeBSD + pf = freeze? X-BeenThere: freebsd-questions@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: User questions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 23 Oct 2009 18:16:49 -0000 On Thu, Oct 22, 2009 at 03:20:20PM -0800, Henrik Hudson wrote: > On Wed, 21 Oct 2009, cpghost wrote: > > I'm experiencing frequent crashes on my soekris net4801 home router > > for some months now, and I'm wondering if it could be some kind of > > pf-related bug similar to this on OpenBSD: > > > > http://www.mail-archive.com/misc@openbsd.org/msg58042.html > > > > More precisely, when I fire up rtorrent-devel on some *other* machine > > (not the router!), everything runs fine at first. It could also run > > very fine for many days. BUT should I start a torrent with a large > > number of seeders which could saturate my link for an extended period > > of time, the soekris router would suddenly freeze... but not > > immediately: more like a few hours (3 to 6) or so of relatively heavy > > traffic. Only a hard reboot of the router would help. Please note > > that rtorrent is NOT running on the router, only its traffic is being > > redirected through the router. > > > > So I'm suspecting some bug / resource leak in pf that would bring the > > kernel down somehow. What kind of resources should I monitor (and > > how)? Maybe that could bring some clues? > > > > Oh, before anybody asks: I have no crashdumps, the router freezes > > totally without panicking. And it doesn't recover automatically > > even after many hours. > > Possibly a heat issue? I've seen many a little dlink style or > similar router work fine until it has to churn through a lot of > packets and then it just can't handle it, starts getting warm > doing all the computation and then eventually freezes. I'm not > ruling out a memory leak or similar, but I'm currently doing the > same with a little atom ITX board and it handles all the torrents > for myself and the roomies without issue. I'm using rtorrent myself with > pf and 8.0-RC1-stable. I believe the pf code is backported to 7. > > Also, if it was just a memory leak it will still happen with > non-torrent traffic, just most likely slower. Have you tried > throttling back the amount of connections and speed that rtorrent > makes? I've suspected a heat issue too, but sysutils/env4801 logging every 1 minute didn't show anything suspicious prior to the crashes. The system crashes ONLY on bittorrent traffic. Saturating the link (in one or both directions) even for many days in a row with 5 to 10 concurrent TCP streams to fixed destinations didn't cause any crashes. Yes, I've played with bandwidth and nr. of connections in rtorrent, and, if at all, I have a feeling (but I can't proove it) that the number of concurrent connections doesn't harm, but that the higher the output bandwidth, the more likely the crash. The only thing I didn't test yet was to replace the original DC transformer with another one that is a tad better dimensioned. Those transformers that are sent with the net4801(s) tend to degrade over the years for some reason (drying capacitors?). If it's not a software issue, this could be the cause of the crashes. > henrik Thanks for the hints, -cpghost. -- Cordula's Web. http://www.cordula.ws/