Hi, Srinidhi.
Since I already replied you about this issue at <contact@tahi.org>,
I hope that you already solve this problem
by using "Possible Problems" section in the test scenario.
And I have a suggestion to perform this test.
When you execute ping6 command,
you should specify packet size 1452 bytes instead of 1500 bytest.
The test scenarion says "original packet size is 1500 byte".
To make this packet size,
you must subtract IPv6 Header (40 bytes) and ICMPv6 Header (8 bytes).
Thanks,
On Tue, 26 May 2009 12:25:22 +0530
Srinidhi Gopal <srinidhigopal@gmail.com> wrote:
> Hi,
>
> Thanks for the responses. Have attached the tcpdumps on REF_HOST1 & TH1.
>
> * Test Setup:*
>
> TH1 (NetBSD 4.0.1)
> | bge0
> -------------------------- (Network1) MTU 1500
> | eth0
> TR1 (CPE Router - RUT under test)
> | eth1
> --------------------------- (Network2) MTU 1280
> | eth1
> Ref-Host1 (Fedora)
>
> *Problem Description:
> **
> ** When REF-HOST1 sends a packet of size 1500, NETBSD (TH1) fragments it by
> default considering 1280 bytes as Max MTU (1232 | 276) though ifconfig bge0
> shows the MTU to be 1500:
> *
> NETBSD# ifconfig bge0*
> bge0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> *mtu 1500 *
> capabilities=3f80<TSO4,IP4CSUM_Rx,IP4CSUM_Tx,TCP4CSUM_Rx,TCP4CSUM_Tx,UDP4CSUM_Rx,UDP4CSUM_Tx>
> enabled=0
> address: 00:12:3f:47:79:af
> media: Ethernet autoselect (100baseTX
> full-duplex,flowcontrol,rxpause,txpause)
> status: active
> inet6 fe80::212:3fff:fe47:79af%bge0 prefixlen 64 scopeid 0x1
> inet6 3ffe:501:ffff:100:212:3fff:fe47:79af prefixlen 64*
> **
> Irrespective of the MTU setting on Network2 (1280 or 1500), NETBSD always
> fragments packets of size 1500 as ( 1232 | 276 ) instead of (1448 | 60).*
>
> I don't face any issue with FreeBSD. The same testcase passes if i use
> FreeBSD instead of NetBSD.
>
> Kindly help in resolving this issue.
>
> On Tue, May 26, 2009 at 10:14 AM, Arul Murugan <umarul@gmail.com> wrote:
>
> > Hi Srinidhi,
> > I think, when Ref1 host pings it knows its MTU, ping with size of 1500
> > will be fragmented from REF1 host itself, so there will not be any "too big
> > message" at network2. But as the target host MTU is 1500, it will try to
> > respond reply 1500 MTU size, that time Tar router will send a Too big
> > message to Network1.
> >
> > Are you not seeing "too big messaage at Network1"?
> > What is the Tar router you are using?
> > If you share your network1 and network2 dump, i may dig bit more on this.
> >
> > Regards,
> > Arul
> >
> > On Tue, May 26, 2009 at 7:03 AM, wang_jiabo <jiabwang@redhat.com> wrote:
> >
> >> Srinidhi Gopal wrote:
> >>
> >>> Hi,
> >>>
> >>> Has anyone used NetBSD machine for IPv6 Phase 2 Interoperability Testing?
> >>>
> >>> We are using NetBSD 4.0.1 as the other Target node for testing our
> >>> product.
> >>>
> >>> The following 2 test cases under PMTU category are failing.
> >>>
> >>> *Failures:*
> >>>
> >>> IP6Interop 1.6a
> >>> IP6Interop 1.6b
> >>>
> >>> *Test Setup:*
> >>>
> >>> TH1 / TR2 |
> >>> -------------------------- (Network1) MTU 1500
> >>> |
> >>> TR1
> >>> |
> >>> --------------------------- (Network2) MTU 1280
> >>> |
> >>> Ref-Host1
> >>> *Observations:*
> >>>
> >>> * The max PMTU size in NetBSD is 1280.
> >>>
> >>> * When we transmit echo requests from Ref-Host1 to TH1/TR2 with data
> >>> size 1500, No "Packet Too Big" Messages are sent by our Router (TR1) since
> >>> the packets (size greater than 1280) are by default fragmented irrespective
> >>> of the MTU size configured on the interface.
> >>>
> >>> * Because of this behaviour of NetBSD stack, We are unable to complete
> >>> the interop testing of our product.
> >>>
> >>> Kindly help us in resolving this issue.
> >>>
> >>> --
> >>> Regards,
> >>> Srinidhi
> >>>
> >> I used NetBSD is no problem, I think REF-Host1 can any OS such as windows
> >> 2008, redhat linux.
> >> only TAR-Host2 should only use NetBSD and FreeBSD.
> >>
> >>
> >
>
>
> --
> Regards,
> Srinidhi Gopal
>
--
Yukiyo Akisada <akisada@tahi.org>