From owner-freebsd-stable@FreeBSD.ORG Fri Dec 1 00:25:21 2006 Return-Path: X-Original-To: freebsd-stable@freebsd.org Delivered-To: freebsd-stable@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [69.147.83.52]) by hub.freebsd.org (Postfix) with ESMTP id 20A8E16A59B for ; Fri, 1 Dec 2006 00:25:06 +0000 (UTC) (envelope-from spork@bway.net) Received: from xena.bway.net (xena.bway.net [216.220.96.26]) by mx1.FreeBSD.org (Postfix) with ESMTP id EF0BF43CA5 for ; Fri, 1 Dec 2006 00:24:54 +0000 (GMT) (envelope-from spork@bway.net) Received: (qmail 30456 invoked by uid 0); 1 Dec 2006 00:25:04 -0000 Received: from unknown (HELO gee5.nat.fasttrackmonkey.com) (216.220.116.154) by smtp.bway.net with (DHE-RSA-AES256-SHA encrypted) SMTP; 1 Dec 2006 00:25:04 -0000 Date: Thu, 30 Nov 2006 19:25:04 -0500 (EST) From: Charles Sprickman X-X-Sender: spork@gee5.nat.fasttrackmonkey.com To: Mark Kirkwood In-Reply-To: <456F5952.9020305@paradise.net.nz> Message-ID: References: <456DA282.6000801@smo.de> <20061129183028.L23451@sporker.bway.net> <456E6C55.3070609@paradise.net.nz> <456F5952.9020305@paradise.net.nz> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed Cc: Philipp Ost , freebsd-stable@freebsd.org Subject: Re: vr speed issues X-BeenThere: freebsd-stable@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Production branch of FreeBSD source code List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 01 Dec 2006 00:25:21 -0000 On Fri, 1 Dec 2006, Mark Kirkwood wrote: > Mark Kirkwood wrote: >> Charles Sprickman wrote: >> >>> I also did a little more digging and noticed that once I start pinging >>> from one of these hosts using large packet sizes, I get about 50-60% >>> packet loss (ie: ping -s 1500 other.vr0.host). If I ping something with a >>> decent card, I get about 30-50% packet loss. There's no packet loss with >>> the default packet size. >>> >>> Anyone else with some vr cards feel like checking this out? >>> >> >> I've got a VIA Rhine III card that I can dig out and put in if the data >> would be of any use/interest etc - FWIW I seem to recall being able to get >> reasonably close to wire speed when I was using it. >> > > I plugged in the card today, and seem to get pretty reasonable performance > (8-10MB/s for scp - see attached). The two boxes are plugged into a Linksys > router via store made cat5 or cat6 cables. The second box has an Intel PRO > 100 (fxp) adapter. Interesting. I've had the vr hosts going through three different switches, patch cables, the in-house cabling, and the issue remains the same. I'd love to just replace a cable and be done with it, but that doesn't seem to be the issue. I can also eliminate it by just switching out cards... so I don't really suspect I've got a whole load of bad cables. > Removing the vr card and going back to fxp everywhere seems to provide better > performance (e.g. get 9MB/s in the last test), but the vr performance is > acceptable (maybe your router clashes with your card?). Mine borders on unusable. The packet loss really slows down and stalls TCP connections. I noticed you have a newer revision of the Via card: # pciconf -lv vr0@pci0:9:0: class=0x020000 card=0x14031186 chip=0x31061106 rev=0x86 hdr=0x00 vendor = 'VIA Technologies Inc' device = 'VT6105M/LOM Rhine III PCI Fast Ethernet Controller' class = network subclass = ethernet This is what I'm dealing with: hostb6@pci0:17:7: class=0x060000 card=0x287e1106 chip=0x287e1106 rev=0x00 hdr=0x00 vendor = 'VIA Technologies Inc' class = bridge subclass = HOST-PCI vr0@pci0:18:0: class=0x020000 card=0x80a71043 chip=0x30651106 rev=0x7c hdr=0x00 vendor = 'VIA Technologies Inc' device = 'VT6102 Rhine II PCI Fast Ethernet Controller' class = network subclass = ethernet Maybe I'll just round up as much info as I can at my next visit, grab some tcpdumps of the loss from both ends and do a send-pr and hope for the best. I can get along with replacing the cards, but they seem to be really common these days - basically any cheap system with a Via chipset and onboard ethernet will be using some variation on this controller. Charles > > Cheers > > Mark > >