Hi, try reducing the tun1 mtu further, to see whether it makes the issue go away - I'm not sure right now if "40" is the right size for ipipv6 (it's the extra IPv4 headers that you need to count, not the IPv6 header)
I'm not exactly sure how to test interface mtu on linux right now -
"ifconfig tun1" should show, though...
gert
On Tue, Nov 05, 2013 at 08:43:16AM -0500, Pietro Paolini wrote:
> Hi,
> thanks for your response, the MTU of my input interface is 1500 and the MTU
> of the tunnel interface is 1460 but I still have this problem.
>
>
> If I try to put another printk:
>
>
>
>
> static int ip_finish_output(struct sk_buff *skb)
> {
> ....
> if ((skb->ipsec_offload == 0) &&
> skb->len > ip_skb_dst_mtu(skb) && !skb_is_gso(skb)) {
> printk(KERN_ERR "%s: %d > %d Fragment packet input device
> (%s)\n",
>
> __FUNCTION__,
>
> skb->len,
>
> dst_mtu(skb_dst(skb)),
>
> skb->dev->name);
> return ip_fragment(skb, ip_finish_output2);
> }
> ..
> }
>
>
>
>
> I can see that when I send a packet of 1433 - UDP payload, total size 1461 -
> ( I am using hping3 and I verify with wireshark ), the result is
>
>
> ip_fragment: 1461 > 1460 Fragment packet input device (dslite-vlan1)
>
>
> then the packet is fragmented by the IPv4 Stack even if the MTU 1500 vs 1460
> (40 is the size of ipv6 headers)
>
>
> Any ideas ?
>
>
>
>
>
> Pietro Paolini
> pulsarpie...@aol.com
>
>
>
>
> -----Original Message-----
> From: Gert Doering <g...@greenie.muc.de>
> To: OpenWrt Development List <openwrt-devel@lists.openwrt.org>
> Sent: Tue, Nov 5, 2013 12:02 pm
> Subject: Re: [OpenWrt-Devel] IPIP6 Fragmentation
>
>
> Hi,
>
> On Mon, Nov 04, 2013 at 07:52:57AM -0500, Pietro Paolini wrote:
> > I am working for an embedded Linux firmware running on a CPE and I am
> > using an IP4 over IPv6 tunnel configured as following:
> >
> > ip -f inet6 tunnel add tun1 mode ipip6 remote XX:XX:XX local XX:XX:XX
> > dev eth0
>
> By setting an appropriate MTU on the tun1 interface ("outer mtu minus
> ipv6 header size") you should see the fragmentation happen only on
> the IPv6 side of things.
>
> What you are doing is telling the stack "yes, tun1 can take my 1500 byte
> packets just fine!", and then the resulting packet will be too large for
> the IPv4 path it takes (which is not necessarily so, you could have an
> IPv4 MTU of 9000 bytes... who knows), so it needs to fragment IPv4,
> but at that point in time, IPv6 is "just payload" so it can't do anything
> about the IPv6 packet anymore (aka "send back ICMP PTB").
>
> [..]
> > Is there any options to disable the fragmentation on IPV4 ?
>
> In that case, you'd need to drop the IPv4 packet. Gain? Zero :-)
>
> gert
> --
> USENET is *not* the non-clickable part of WWW!
> //www.muc.de/~gert/
> Gert Doering - Munich, Germany g...@greenie.muc.de
> fax: +49-89-35655025 g...@net.informatik.tu-muenchen.de
>
>
> _______________________________________________
> openwrt-devel mailing list
> openwrt-devel@lists.openwrt.org
> https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
>
>
--
USENET is *not* the non-clickable part of WWW!
//www.muc.de/~gert/
Gert Doering - Munich, Germany g...@greenie.muc.de
fax: +49-89-35655025 g...@net.informatik.tu-muenchen.de
pgpkohMZAcoLv.pgp
Description: PGP signature
_______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
