Date: Mon, 25 Dec 2006 00:22:35 +0300 From: "Andrew Pantyukhin" <infofarmer@FreeBSD.org> To: "Ruslan Ermilov" <ru@freebsd.org> Cc: David Xu <davidxu@freebsd.org>, current@freebsd.org, Kris Kennaway <kris@obsecurity.org> Subject: Re: vge(4) bad checksum Message-ID: <cb5206420612241322q25436f1dx8d6e6829b6d73d49@mail.gmail.com> In-Reply-To: <20061222080202.GB77429@rambler-co.ru> References: <cb5206420612171246q54ac783h1fd9d420b80ba84c@mail.gmail.com> <20061217205249.GA73132@xor.obsecurity.org> <20061222080202.GB77429@rambler-co.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
On 12/22/06, Ruslan Ermilov <ru@freebsd.org> wrote: > Hi, > > On Sun, Dec 17, 2006 at 03:52:49PM -0500, Kris Kennaway wrote: > > On Sun, Dec 17, 2006 at 11:46:24PM +0300, Andrew Pantyukhin wrote: > > > I'm not sure what it's all about, but with today's > > > current whatever goes out my vge interface (icmp/ > > > tcp/udp) has bad checksum: > > > > This is a FAQ; it's probably using hardware checksum offloading. > > > > Since the packet passed down to the NIC does not yet have the checksum > > computed, it looks to tcpdump like the checksum is incorrect. However > > if you look at the packet actually transmitted by the NIC > > (e.g. tcpdump on another host), you'll see that it has the correct > > checksum. > > > Kris, you probably missed a commit by csjp@ where it was fixed. > > : revision 1.220 > : date: 2006/11/18 23:17:22; author: csjp; state: Exp; lines: +40 -0 > : Currently, drivers that support hardware offload of VLAN tag > : processing are forced to toggle this functionality when the card > : is put in and out of promiscuous mode. The main reason for this > : is because the hardware strips the VLAN tag, making it impossible > : for the tag information to show up in network diagnostic tools like > : tcpdump(1). > : [...] > > Andrey, have you been able to narrow your problem down to > either this commit, my vge(4) commit (though you tested it > as well before it was committed), or to FAST_IPSEC? I tried to communicate with another fast_ipsec-enabled box (6-stable/rl0) and everything was okay, so I guess vge0 works well with fast_ipsec. I'll have to investi- gate re0 quirks. Thanks!
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?cb5206420612241322q25436f1dx8d6e6829b6d73d49>