From owner-freebsd-hackers Sat Jul 28 13:28:10 2001 Delivered-To: freebsd-hackers@freebsd.org Received: from scaup.mail.pas.earthlink.net (scaup.mail.pas.earthlink.net [207.217.121.49]) by hub.freebsd.org (Postfix) with ESMTP id D64B637B401; Sat, 28 Jul 2001 13:28:05 -0700 (PDT) (envelope-from tlambert2@mindspring.com) Received: from mindspring.com (dialup-209.247.137.129.Dial1.SanJose1.Level3.net [209.247.137.129]) by scaup.mail.pas.earthlink.net (EL-8_9_3_3/8.9.3) with ESMTP id NAA02282; Sat, 28 Jul 2001 13:27:50 -0700 (PDT) Message-ID: <3B63206D.4377EDD9@mindspring.com> Date: Sat, 28 Jul 2001 13:28:29 -0700 From: Terry Lambert Reply-To: tlambert2@mindspring.com X-Mailer: Mozilla 4.7 [en]C-CCK-MCD {Sony} (Win98; U) X-Accept-Language: en MIME-Version: 1.0 To: Chris Dillon Cc: Julian Elischer , "Eugene L. Vorokov" , Soren Kristensen , freebsd-hackers@FreeBSD.ORG, freebsd-net@FreeBSD.ORG Subject: Re: Why two cards on the same segment... References: Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG Chris Dillon wrote: > > We saw the error with multiple 10.x addresses, with subnet masks > > which should have logically seperated the subnets, but failed to > > do the job correctly, when using two cards on the same segment, > > with different subnet masks which should have rendered them > > non-intersecting. I can probably get the configuration data for > > you, if you are truly interested (this is on a 4.3 derived > > system). > > Not that being 10.x addresses would matter any, but it would be > interesting to look at. It wouldn't be hard for me to put another NIC > in this box and play around with that scenario. What exactly was > going wrong in the above setup you're talking about? The ARP response came back on the wrong interface because it was sent on the wrong interface, and the kernel bitched about it coming back from the wrong place. If it didn't want the response coming back on that interface, it shouldn't have sent the request from that interface. You can duplicate this pretty easily by: 1) Set up a 10.x address with a netmask of 255.255.0.0 on one card. 2) Make this card your default route 3) Set up a 10.7.y (x != 7) address with a netmask of 255.255.255.0 on a second card attached to the same wire (we used a Netgear hub, but a Netgear switch and some other equipment shows the same behaviour) 4) Config the second interface to force a proxy arp for its own address 5) Watch the ARP be sent by the first interface instead of the second 6) Watch the kernel complain about the response coming back to the MAC which it was sent from - by the kernel If you need exact numbers and netmasks and hardware, I can give you that information on Monday or Tuesday (it wasn't my personal box that had this problem). -- Terry To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message