NetBSD-Bugs archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: kern/52554: IPv6 connections not routing to default gateway



On Mon, Oct 02, 2017 at 02:15:01PM +0000, Roy Bixler wrote:
>  I asked the network admin. and he tells me that that IPv6 routers are
>  advertising priorities and the router that works has the highest
>  priority.  He says that it's an option that was put into the Linux
>  kernel a few years ago.  The option, CONFIG_IPV6_ROUTER_PREF, quotes
>  RFC4191 as a reference.  From looking at the dhcp client logs, I can
>  see that it's getting advertisements from all of the routers and it
>  just picks the first one (i.e. sets its default route according to it)
>  and disregards the others.  I still wonder why this doesn't happen
>  with NetBSD 7.1.

I'm fairly convinced that the problem is with dhcpcd.  With NetBSD-8,
I tried an experiment, which set ip6mode to autohost and pass the -4
flag to dhcpcd.  I observed that the correct default router was
applied more often.  If it wasn't, then I could do an "ndp -R" command
one (or a few) time(s).  If it found the highest priority router
first, then "ndp -H" was enough to correct the default IPv6 router.
If some other IPv6 router was found first, then a router with higher
preference will not change the routing table.  However, similar
experiments I did with NetBSD 7.1 had the same result.

The difference is that, with NetBSD 7.1, dhcpcd will obey the router's
priority.  Even if a lower preference router advertises first and is
made the default IPv6 router, dhcpcd will correct that later if the
higher preference router advertises.  With NetBSD 8, dhcpcd strictly
takes the first IPv6 router that advertises and makes that the
default.  It will never update the default unless I manually do "ndp
-R" once or a few times, the highest priority router happens to
advertise first and then I do "ndp -H" to update the routing table.

-- 
Roy Bixler <rcbixler%nyx.net@localhost>
"The fundamental principle of science, the definition almost, is this: the
sole test of the validity of any idea is experiment."
-- Richard P. Feynman



Home | Main Index | Thread Index | Old Index