Re: High latency for IPv6 on netbsd-8

[Gua Chung Lim <gua.chunglim%gmail.com@localhost>]

Thanks for your kind response, though I don't know much about what you are explaining.

* Robert Elz (kre%munnari.OZ.AU@localhost) wrote:
> Just guessing, but this sounds to me like the "prohibiit sending from 
> tentative addresses" BS that was added a while ago.
> 
> I'd suspect that it goes something like this ...
> 
> host boots, configures itself link-local addresses, and starts DaD
> on them - addesses are tentative.
> 
> rtsol (v7) or dhcpcd (v8) starts, and sends a RS to request a
> RA from the router.   In v7 that is just sent.  In v8 it is not, as
> DaD isn't finished yet.
I previously had rtsol in /etc/ifconfig.wm0 i.e.

% cat /etc/ifconfig.wm0
up
!rtsol $int

But during boot, it complains as rtsol is obsolete, so I remove the second line. So only "up" is still there in the file. Anyway, I saw "Waiting for DAD to complete..." right before router solicitation. So I guess that DAD complete before RS.
https://pastebin.com/udwFzp3N

> RS's don't get ack'd, the RA is not that, but even if the sender
> decides to send another after not seeing a RA fairly soon, it might
> still be within the DaD period.
> 
> The v7 host gets the RA, configures global v6 addresses, and all
> works, the v8 host does not (quickly).
> 
> Routers send RAs periodically - every 10 minutes is a common config,
> so eventually a RA appears, the v8 host configures its v6 addrs (finally)
> and all is OK (the actual delay will vary of course, depending upon where
> in the router's time sequence of regular RA sending the host boots).
Sometimes I have to wait for 30 minutes.

> To see if this guess is likely enough to be raised to be a hypothesis, do
> two things...
> 
> First, soon after the boot, ping the router's link local address.   Since LL
> (v6) addresses tend to be obnoxious things, it would be best to have a
> script pre-prepared, and tested, for this in advance.
Immediately after boot, I can ping6 fe80::1. (I believe it's LL.)

> If that works (give it at least a second or two after boot - by the time you
> can log in should be OK, but don't automate it in rc.local or anything), and
> a ping6 of a global v6 address does not (nb: have the address calculatted
> in advance, so it cannot be a DNS issue, and use -n on the pings - always...)
> then this is a clue.
Right after boot, where ping6 canonical name fails, ping6 global ipv6 address also fails. So it should not be a DNS issue.

> Then add
> 
> 	net.inet6.ip6.dad_count=0
> 
> to /etc/sysctl.conf and reboot, and try again.   If the global v6 ping works
> this time, (and the LL ping both times) then this is likely to be the problem.
Adding this line to /etc/sysctl.conf and reboot doesn't fix the problem. ping6 LL works but ping6 the public ipv6 address or canonical name doesn't.

> On the other hand if a global v6 ping to an address works, and one to a
> hostname does not, then that would be evidence of DNS issues, and we
> would need to look in a different direction.
Both ping6 global address and canonical name fail.

> ps: I hope you are on the list, as gmail almost always rejects mail from me.
I am on the list. :-)

Thank you,

-- 
Gua Chung Lim
 
"UNIX is basically a simple operating system,
but you have to be a genius to understand the simplicity."
-- Dennis M. Ritchie