[Cerowrt-devel] ipv6 confusion with source specific gateways.

[Cerowrt-devel] ipv6 confusion with source specific gateways.

[Dave Taht]

Some days I just lose it with prefix math. Or maybe we have a bug.

So here is a default gateway. So far as I know it is getting a /56
actually, not a /60, but it is requesting a /60 and thus distributing
a source specific route to that /60.

default from :: via fe80::201:5cff:ee62:b646 dev ge00  proto static  metric 1024
# should I even have a default route at all?
default from 2001:558:6045:e9:44a9:4a25:ece9:55c3 via
fe80::201:5cff:ee62:b646 dev ge00  proto static  metric 1024
# my uplink, looks sane

default from 2601:c:ce00:9d0::/60 via fe80::201:5cff:ee62:b646 dev
ge00  proto static  metric 1024
# and see that /60...

# and then there is the default ipv6 address assignements on cero.

2601:c:ce00:9d1::/64 dev gw10  proto kernel  metric 256  expires 326362sec

Now, where my brain crashes: shouldn't 2601:c:ce00:9d1::/64 actually
be something like

2601:c:ce00:9d01::/64 not, 2601:c:ce00:9d1::/64?

(see the 01?)?

OK, so, like, I have another babel source specifc gateway a few hops
in from there, I add

ip -6 addr add 2601:c:ce00:9d01::1/64 dev wlan0-1

that doesnt work. I make it be

ip -6 addr add 2601:c:ce00:9d6::1/64 dev wlan0-1

and that does.


-- 
Dave Täht

http://www.bufferbloat.net/projects/bloat/wiki/Upcoming_Talks

Re: [Cerowrt-devel] ipv6 confusion with source specific gateways.

[Toke Høiland-Jørgensen]

>default from :: via fe80::201:5cff:ee62:b646 dev ge00  proto static 
>metric 1024
># should I even have a default route at all?

The source on this is actually ::/128 so it's not a default route in the sense that it will forward arbitrary traffic through it. It is needed for IPv6 packets that originate on the host but where the application doesn't specify an address to bind to. To see that, try removing it and then ping6 something you don't have a specific route to...


>Now, where my brain crashes: shouldn't 2601:c:ce00:9d1::/64 actually
>be something like
>
>2601:c:ce00:9d01::/64 not, 2601:c:ce00:9d1::/64?
>
>(see the 01?)?

2601:c:ce00:9d01 is 2601:000c:ce00:9d01
2601:c:ce00:9d1 is 2601:000c:ce00:09d1

So no? :)

-Toke

Re: [Cerowrt-devel] ipv6 confusion with source specific gateways.

[Robert Bradley]

On 26/11/14 07:40, Toke Høiland-Jørgensen wrote:
> 2601:c:ce00:9d01 is 2601:000c:ce00:9d01
> 2601:c:ce00:9d1 is 2601:000c:ce00:09d1
>
> So no? :)

The issue is that this is supposedly a /60, and so only addresses in the
range 2601:c:ce00:9d0x::/60 are valid.  The question is actually about
how CeroWRT subnet hints cope with nybble alignment of subnets as
opposed to byte/octet alignments.

-- 
Robert Bradley

Re: [Cerowrt-devel] [Babel-users] ipv6 confusion with source specific gateways.

[Juliusz Chroboczek]

>> default from :: via fe80::201:5cff:ee62:b646 dev ge00 proto static metric
>> 1024 # should I even have a default route at all?
>
> The source on this is actually ::/128

You mean ::/0.

> It is needed for IPv6 packets that originate on the host but where the
> application doesn't specify an address to bind to.

Yes, it is also needed to ensure interoperability between Babel-S and
stock Babel.  (I really wish OpenWRT installed a non-specific route in
that case, by the way.)

-- Juliusz

Re: [Cerowrt-devel] [Babel-users] ipv6 confusion with source specific gateways.

[Dave Taht]

On Mon, Dec 1, 2014 at 6:15 PM, Juliusz Chroboczek
<jch@pps.univ-paris-diderot.fr> wrote:
>>> default from :: via fe80::201:5cff:ee62:b646 dev ge00 proto static metric
>>> 1024 # should I even have a default route at all?
>>
>> The source on this is actually ::/128
>
> You mean ::/0.
>
>> It is needed for IPv6 packets that originate on the host but where the
>> application doesn't specify an address to bind to.
>
> Yes, it is also needed to ensure interoperability between Babel-S and
> stock Babel.  (I really wish OpenWRT installed a non-specific route in
> that case, by the way.)

?

An issue I have had with owrt is that it used route to enter routes, which
does not support various tables (like dhcp, or babel-pinger).

I think, but am not sure, that netifd in chaos calmer actually calls
the routing commands directly now, so that
behavior could be more correct. I don't know what udhcp uses...

The unreachable syntax was also recently added to netifd, so exporting
covering routes has just got easier.

see: https://github.com/sbyx/hnetd/issues/23
for details.






>
> -- Juliusz



-- 
Dave Täht

thttp://www.bufferbloat.net/projects/bloat/wiki/Upcoming_Talks

Re: [Cerowrt-devel] [Babel-users] ipv6 confusion with source specific gateways.

[Toke Høiland-Jørgensen]

Juliusz Chroboczek <jch@pps.univ-paris-diderot.fr> writes:

>>> default from :: via fe80::201:5cff:ee62:b646 dev ge00 proto static metric
>>> 1024 # should I even have a default route at all?
>>
>> The source on this is actually ::/128
>
> You mean ::/0.

No, I do not, actually. That would make it just a catch-all default
route (and it would show up as such). In this case it is a
source-specific route that only matches packets originating on the host
that don't specify a source IP.

-Toke

Re: [Cerowrt-devel] ipv6 confusion with source specific gateways.

[Chuck Anderson]

On Tue, Nov 25, 2014 at 11:11:17PM -0800, Dave Taht wrote:
> Some days I just lose it with prefix math. Or maybe we have a bug.

Prefix math, there is no bug.

> So here is a default gateway. So far as I know it is getting a /56
> actually, not a /60, but it is requesting a /60 and thus distributing
> a source specific route to that /60.

Looks like a /60.

> default from :: via fe80::201:5cff:ee62:b646 dev ge00  proto static  metric 1024
> # should I even have a default route at all?
> default from 2001:558:6045:e9:44a9:4a25:ece9:55c3 via
> fe80::201:5cff:ee62:b646 dev ge00  proto static  metric 1024
> # my uplink, looks sane
> 
> default from 2601:c:ce00:9d0::/60 via fe80::201:5cff:ee62:b646 dev
> ge00  proto static  metric 1024
> # and see that /60...
> 
> # and then there is the default ipv6 address assignements on cero.
> 
> 2601:c:ce00:9d1::/64 dev gw10  proto kernel  metric 256  expires 326362sec
> 
> Now, where my brain crashes: shouldn't 2601:c:ce00:9d1::/64 actually
> be something like
> 
> 2601:c:ce00:9d01::/64 not, 2601:c:ce00:9d1::/64?

No, it is correct.  The /60 is:

2601:c:ce00:9d0::/60

which is really this when expanded out (zero-pad each : section):

2601:000c:ce00:09d0::/60


Now the 60 bit prefix is:

2601:000c:ce00:09dX:XXXX:XXXX:XXXX:XXXX

where X is the part which can be assigned to your LANs.  The first X
is a 4-bit nibble that was assigned "1" by Cero creating a /64 subnet
prefix, leaving the remaining 64 bits for your Interface IDs...

Re: [Cerowrt-devel] ipv6 confusion with source specific gateways.

[Dave Taht]

On Wed, Nov 26, 2014 at 7:21 PM, Chuck Anderson <cra@wpi.edu> wrote:
> On Tue, Nov 25, 2014 at 11:11:17PM -0800, Dave Taht wrote:
>> Some days I just lose it with prefix math. Or maybe we have a bug.
>
> Prefix math, there is no bug.
>
>> So here is a default gateway. So far as I know it is getting a /56
>> actually, not a /60, but it is requesting a /60 and thus distributing
>> a source specific route to that /60.
>
> Looks like a /60.
>
>> default from :: via fe80::201:5cff:ee62:b646 dev ge00  proto static  metric 1024
>> # should I even have a default route at all?
>> default from 2001:558:6045:e9:44a9:4a25:ece9:55c3 via
>> fe80::201:5cff:ee62:b646 dev ge00  proto static  metric 1024
>> # my uplink, looks sane
>>
>> default from 2601:c:ce00:9d0::/60 via fe80::201:5cff:ee62:b646 dev
>> ge00  proto static  metric 1024
>> # and see that /60...
>>
>> # and then there is the default ipv6 address assignements on cero.
>>
>> 2601:c:ce00:9d1::/64 dev gw10  proto kernel  metric 256  expires 326362sec
>>
>> Now, where my brain crashes: shouldn't 2601:c:ce00:9d1::/64 actually
>> be something like
>>
>> 2601:c:ce00:9d01::/64 not, 2601:c:ce00:9d1::/64?
>
> No, it is correct.  The /60 is:
>
> 2601:c:ce00:9d0::/60
>
> which is really this when expanded out (zero-pad each : section):
>
> 2601:000c:ce00:09d0::/60

You are absolutely correct. Thank you!

>
>
> Now the 60 bit prefix is:
>
> 2601:000c:ce00:09dX:XXXX:XXXX:XXXX:XXXX
>
> where X is the part which can be assigned to your LANs.  The first X
> is a 4-bit nibble that was assigned "1" by Cero creating a /64 subnet
> prefix, leaving the remaining 64 bits for your Interface IDs...

It turned out that I had a non-source-specific babeld between me and
it, also (bloody meshy routing!),
which was messing things up. Worse, that is on a roof where I cant
update it easily.

Also I AM getting a /56 from a comcast business class account
elsewhere on the net...

but grokking hnetd is killing me.

> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel



-- 
Dave Täht

thttp://www.bufferbloat.net/projects/bloat/wiki/Upcoming_Talks