IPv6 on Ben Nanonote

David Kuehling dvdkhlng at gmx.de
Sun Jun 17 00:01:23 EDT 2012

>>>>> "Lluís" == Lluís Batlle i Rossell <viric at viric.name> writes:

> On Sat, Jun 16, 2012 at 02:17:06AM +0200, David Kuehling wrote:
>> >>>>> "Xiangfu" == Xiangfu Liu <xiangfu at sharism.cc> writes: ssh
>> root at fe80::8ccf:67ff:fe82:98e6%usb0
>> Next problem: the IPv6 link-local address changes after every boot!
>> The IPv6 address is generated from the MAC-address; the MAC adress
>> changes, too.  Why???

> Link local addresses are generated at random, just as ipv4 autoip
> addresses, only to have some id that does not collide with any in the
> network. And their prefix does not indicate anything to the kernel
> routing table, hence you need to specify the out interface.

While you're maybe technically right, that IPv6 link-local addresses
*might* be generated at random, they're non-random for all Linux systems
I recently used.  The suffix /is/ generated from the MAC—address, which
can easily be seen when looking at ifconfig output.  

Also for nanonote the IPv6 link-local address is generated from the
MAC-address, and it is the MAC-address that is random, probably because
nobody cared to assign and store unique addresses at the factory.  After
all USB-based ethernet is somewhat a virtual interface.

> You need some ipv6 dhcp server, or set some easy addresses in both
> sides. Note that ipv6 interfaces are meant to have *more* than one ip,
> so it's only about *adding* a fixed address, a *dhcp* address,
> whatever.

IPv6 DHCP? Noooooooo.  Let's not repeat these mistakes.  The right way
would be for the host (and/or nanonote) to implement an IPv6 subnet
(probably using a unique local address), and run radvd to broadcast
presence of the network.  Then the Linux kernel on the other side would
automatically configure a corresponding address once it picks up the
broadcast.  We don't need (nor want) DHCP on IPv6.

I'm against running any daemons on the nanonote to save resources.  So
with only the host having radvd, we'd still not know which address the
nanonote would get, as long as it's using a random MAC.  Using a static
IPv6 address on the nanonote somewhat defeats the idea of using IPv6 in
the first place.

Looking here [1] I think the nanonote may be able to send router
solicitations directly from the Kernel, without any daemon required.
Though making the NanoNote a "router" is also somewhat ugly?


GnuPG public key: http://dvdkhlng.users.sourceforge.net/dk.gpg
Fingerprint: B17A DC95 D293 657B 4205  D016 7DEF 5323 C174 7D40

[1] http://www.tldp.org/HOWTO/Linux+IPv6-HOWTO/proc-sys-net-ipv6..html
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: not available
URL: <http://lists.en.qi-hardware.com/pipermail/discussion/attachments/20120617/556d4aed/attachment.pgp>

More information about the discussion mailing list