Specifying IPv6 off-link for a subnet

[Scott]

On Mon, Jul 19, 2021 at 12:40:35PM -0700, Mel Pilgrim wrote:
> On 2021-07-18 17:07, Scott wrote:
> > On Wed, Jul 07, 2021 at 07:43:48PM -0700, Mel Pilgrim wrote:
> >> On 2021-07-05 8:45, Scott wrote:
> >>> I have manually configured IPv6 on the interfaces:
> >>> ifconfig_vmx0_ipv6="inet6 xx::yy prefer_source accept_rtadv -autoconf no_radr"
> [...]
> >>> Here's the received RA:
> [...]
> >>> 	  prefix info option (3), length 32 (4): xx::/64, Flags [none], valid
> >>> 	  time 3600s, pref. time 60s
> >>>
> >>> Note that there is no L flag set, so the prefix is off-link.
> >>
> >> If set, the L flag indicates the prefix can be used for on-link
> >> determination. But if the L bit is unset, the PIO isn't making any
> >> assertion about the prefix being on-link or off-link. More importantly,
> >> a host must not use a PIO with L=0 to conclude a prefix is off-link.
> >>
> >> IOW, it is correct for a host to do nothing when it receives an RA PIO
> >> with no flags for an already-configured prefix.
> > 
> > so for the case of manually addressed devices on a private vLAN/IPPL there
> > must be a way to specify that the configured prefix is not on-Link.
> 
> Yes, by not configuring that prefix.  Numbering vmx0 into xx::/64 made 
> xx::/64 on-link by definition.
> 
> This seems like too intuitive of an answer, though.  Can you step back 
> and explain what you need to do at a higher level?
> 

My need is pretty simple: to have FreeBSD servers with an IPv6 stack 
operating correctly on a private vLAN (IPPL).  Specifically an isolated 
private vLAN.

In IPv6 (as I understand it), this should be done by specifying that the 
prefix/subnet of the configured address is not on-link, and therefore should 
*always* use the router as the next hop.  (Similar to an NBMA network).

I have this working for IPv4 - which uses different mechanisms such as local 
proxy ARP on the router (such mechanisms do no apply to v6).

Thanks