Index: [Article Count Order] [Thread]

Date: Wed, 16 Jul 2008 10:29:56 +0900
From: Toyo Abe <tabe@miraclelinux.com>
Subject: [users:00827] Re: Questions about Interoperability Test for phase2-core
To: Yukiyo Akisada <akisada@tahi.org>
Cc: users@tahi.org
Message-Id: <487D4F14.6030103@miraclelinux.com>
In-Reply-To: <20080714120336.9054abfa.akisada@tahi.org>
References: <4871AD9A.6010303@miraclelinux.com> <20080714120336.9054abfa.akisada@tahi.org>
X-Mail-Count: 00827

Akisada-san,

Thanks for your kind response.

Yukiyo Akisada wrote:
> Hi, Abe-san.
> 
> For 1st question)
> 
>     You can use them for your environment.
>     Your list satisfies IPv6 Ready Logo requirements,
>     but I have a suggestion about it.
>     I feel that using other OS (ex, Open BSD) is better than KAME
>     because KAME already had stopped the activity long time ago.
>     If you have the product other than free software OS,
>     using the product is the best.
>
Indeed. I'll consider it.

 
> For 2nd question)
> 
>     It is the correct behavior according to RFC 4861.
> 
>     7.2.2.  Sending Neighbor Solicitations
>     ------------------------------------------------------------------------
>     3440    While waiting for address resolution to complete, the sender MUST,
>     3441    for each neighbor, retain a small queue of packets waiting for
>     3442    address resolution to complete.  The queue MUST hold at least one
>     3443    packet, and MAY contain more.  However, the number of queued packets
>     3444    per neighbor SHOULD be limited to some small value.  When a queue
>     3445    overflows, the new arrival SHOULD replace the oldest entry.  Once
>     3446    address resolution completes, the node transmits any queued packets.
>     ------------------------------------------------------------------------
>
Thanks for the information.
BTW, I found the way to enlarge the queue length.
I'd like to note it for the record. Just run the following command.
  # sysctl -w net.inet6.icmp6.nd6_maxqueuelen=10
This works on FreeBSD6.1 or later.
On NetBSD4.0 or later, the command should be changed like this.
  # sysctl -w net.inet6.icmp6.maxqueuelen=10


Thank you,
-toyo
 
> Thanks,
> 
> 
> On Mon, 07 Jul 2008 14:46:02 +0900
> Toyo Abe <tabe@miraclelinux.com> wrote:
> 
>> Hello,
>>
>> I have two questions WRT interoperability test for host category of phase2-core.
>>
>> One is whether the test environment is acceptable or not.
>> My environment is as follows.
>> LOGO:
>>   Our product, which is Linux box.
>> REF-1:
>>   FreeBSD 6.0-RELEASE-p7
>> REF-2:
>>   FreeBSD 6.0-RELEASE-p7
>> TAR-Host1:
>>   FreeBSD 6.0-RELEASE-p7
>> TAR-Host2:
>>   USAGI(usagi-linux26-stable-20050714.tar.bz2) on Debian3.1r8
>> TAR-Router1:
>>   NetBSD 2.0.2
>> TAR-Router2:
>>   KAME(kame-20080601-netbsd20-snap.tgz) on NetBSD 2.0.2
>>
>>
>> Another one is related to IP6Interop.1.6(spec version is 2.8.4), which is
>> Path MTU Discovery and Fragmentation test.
>> This kind of question might be inappropriate here, but I'd like to know
>> if anyone had seen the same problem before or has a workaround.
>>
>> interop.1.6.x requires TAR/REF box to send fragmented ICMPv6 ECHO packets
>> to LOGO. However, all BSDs listed above conditionally fail to send out only
>> the 1st fragmented packet.
>> Usually they successfully send the fragmented pings. But they fail in case that
>> no NCE for ping destination exist. When there isn't NCE for destination, BSD box
>> initiates ND to figure out who is the destination. After it's completed, only 
>> the 1st fragmented packet appears to be dropped on BSD box.
>> I'm wondering why, and worrying that if this will be a serious problem for 
>> getting Logo application.
>>
>>
>> Thank you,
>> -toyo
>>
>>
> 
>