rfc5549 support in illumos

rfc5549 support in illumos

[John E. Barfield]

[-- Attachment #1: Type: text/plain, Size: 314 bytes --]

Greetings!

Is there support or a roadmap to support ipv4 reachability over ipv6 in illumos?

reference:

https://datatracker.ietf.org/doc/html/rfc5549

Many large network vendors are supporting this now and Linux/BSD supports it natively.


John Barfield
Serial Innovator
9b - Wolfpeak Technologies

[-- Attachment #2: Type: text/html, Size: 1154 bytes --]

Re: [developer] rfc5549 support in illumos

[Dan McDonald]

On May 11, 2024, at 10:45 PM, John E. Barfield via illumos-developer <developer@lists.illumos.org> wrote:
> 
> Greetings!
> 
> Is there support or a roadmap to support ipv4 reachability over ipv6 in illumos?
> 
> reference:
> 
> https://datatracker.ietf.org/doc/html/rfc5549

This looks like an extension to MP-BGP.  illumos proper doesn't have BGP in it; that's more of a distro thing.

I *do* know there's work afoot to make other improvements in illumos for MP-BGP (see recent pushes in TCP, e.g. https://github.com/illumos/illumos-gate/commit/1edba515a3484e0f74b638b203d462b3112ac84d ), but the RFC you cite is something you'd see in a BGP daemon (e.g. openbgp).

Dan

Re: [developer] rfc5549 support in illumos

[Bill Sommerfeld]

On 5/12/24 07:37, Dan McDonald wrote:
> On May 11, 2024, at 10:45 PM, John E. Barfield via illumos-developer <developer@lists.illumos.org> wrote:
>>
>> Greetings!
>>
>> Is there support or a roadmap to support ipv4 reachability over ipv6 in illumos?
>>
>> reference:
>>
>> https://datatracker.ietf.org/doc/html/rfc5549
> 
> This looks like an extension to MP-BGP.  illumos proper doesn't have BGP in it; that's more of a distro thing.
> 
> I *do* know there's work afoot to make other improvements in illumos for MP-BGP (see recent pushes in TCP, e.g. https://github.com/illumos/illumos-gate/commit/1edba515a3484e0f74b638b203d462b3112ac84d ), but the RFC you cite is something you'd see in a BGP daemon (e.g. openbgp).

So to actually use what you learn via the RFC5549 BGP extensions, you 
need some sort of v4-over-v6 multipoint tunnel interface that can 
dynamically tunnel V4 over V6 based on nexthop information learned from 
BGP, and I don't think we have all the ingredients for that.

Not sure off the top of my head about what the right way to build this 
into our stack is -- perhaps some way to add a v6 nexthop to a v4 route 
which invokes the tunneling mechanism.   Perhaps a plugin for 
overlay(7), perhaps something else.

Re: [developer] rfc5549 support in illumos

[Dan McDonald]

On May 12, 2024, at 12:05 PM, Bill Sommerfeld via illumos-developer <developer@lists.illumos.org> wrote:
> 
> So to actually use what you learn via the RFC5549 BGP extensions, you need some sort of v4-over-v6 multipoint tunnel interface that can dynamically tunnel V4 over V6 based on nexthop information learned from BGP, and I don't think we have all the ingredients for that.
> 
> Not sure off the top of my head about what the right way to build this into our stack is -- perhaps some way to add a v6 nexthop to a v4 route which invokes the tunneling mechanism.   Perhaps a plugin for overlay(7), perhaps something else.

We have 6to4 for iptun(4D), if we're straight route based tunneling, seems iptun would be the best place for this since BGP is discovering things.

Unless BGP is being used as a resolver, which makes an overlay plugin to employ BGP might make sense.

Dan

Sent from my iPhone (typos, autocorrect, and all)

Re: [developer] rfc5549 support in illumos

[John E. Barfield]

[-- Attachment #1: Type: text/plain, Size: 1763 bytes --]

I saw an announcement recently on the smartos list suggesting an old tcp bug had been resolved which affected bgp in illumos…do you think that openbgp would/should work properly now in a native zone? If this is better suited for the smartos/pkgsrc list lmk and ill ask there.


John Barfield
Serial Innovator
9b - Wolfpeak Technologies
________________________________
From: Dan McDonald <danmcd@mnx.io>
Sent: Sunday, May 12, 2024 12:00:49 PM
To: illumos-developer <developer@lists.illumos.org>
Subject: Re: [developer] rfc5549 support in illumos

On May 12, 2024, at 12:05 PM, Bill Sommerfeld via illumos-developer <developer@lists.illumos.org> wrote:
>
> So to actually use what you learn via the RFC5549 BGP extensions, you need some sort of v4-over-v6 multipoint tunnel interface that can dynamically tunnel V4 over V6 based on nexthop information learned from BGP, and I don't think we have all the ingredients for that.
>
> Not sure off the top of my head about what the right way to build this into our stack is -- perhaps some way to add a v6 nexthop to a v4 route which invokes the tunneling mechanism.   Perhaps a plugin for overlay(7), perhaps something else.

We have 6to4 for iptun(4D), if we're straight route based tunneling, seems iptun would be the best place for this since BGP is discovering things.

Unless BGP is being used as a resolver, which makes an overlay plugin to employ BGP might make sense.

Dan

Sent from my iPhone (typos, autocorrect, and all)


------------------------------------------
illumos: illumos-developer
Permalink: https://illumos.topicbox.com/groups/developer/T77062f8668d4313f-Mea7d71725e9ac5d6f7225300
Delivery options: https://illumos.topicbox.com/groups/developer/subscription

[-- Attachment #2: Type: text/html, Size: 2963 bytes --]

Re: [developer] rfc5549 support in illumos

[Dan McDonald]

On May 12, 2024, at 11:46 PM, John E. Barfield via illumos-developer <developer@lists.illumos.org> wrote:
> 
> I saw an announcement recently on the smartos list suggesting an old tcp bug had been resolved which affected bgp in illumos…do you think that openbgp would/should work properly now in a native zone? If this is better suited for the smartos/pkgsrc list lmk and ill ask there.

Two parts appear to be necessary:

1.) One needs a bgp.  For SmartOS this would come from pkgsrc at first glance.

2.) One needs the forwarding IPv4-over-IPv6 Bill S. mentioned.  THAT is certainly relevant to this list.

Dan

Re: [developer] rfc5549 support in illumos

[Joshua M. Clulow]

On Sun, 12 May 2024 at 20:46, John E. Barfield via illumos-developer
<developer@lists.illumos.org> wrote:
> I saw an announcement recently on the smartos list suggesting an old tcp bug had been resolved which affected bgp in illumos…do you think that openbgp would/should work properly now in a native zone? If this is better suited for the smartos/pkgsrc list lmk and ill ask there.

FWIW, I have been using an illumos port of the portable OpenBGPD in
native zones for quite some time.  The port is in the "illumos" branch
of:

    https://github.com/eait-itig/openbgpd-portable

I take the patches and extra files from in there, and apply them to
OpenBGPD 7.9 and build it.


Cheers.

-- 
Joshua M. Clulow
http://blog.sysmgr.org

Re: [developer] rfc5549 support in illumos

[Joshua M. Clulow]

On Sun, 12 May 2024 at 10:01, Dan McDonald <danmcd@mnx.io> wrote:
> On May 12, 2024, at 12:05 PM, Bill Sommerfeld via illumos-developer <developer@lists.illumos.org> wrote:
> > So to actually use what you learn via the RFC5549 BGP extensions, you need some sort of v4-over-v6 multipoint tunnel interface that can dynamically tunnel V4 over V6 based on nexthop information learned from BGP, and I don't think we have all the ingredients for that.
> >
> > Not sure off the top of my head about what the right way to build this into our stack is -- perhaps some way to add a v6 nexthop to a v4 route which invokes the tunneling mechanism.   Perhaps a plugin for overlay(7), perhaps something else.
>
> We have 6to4 for iptun(4D), if we're straight route based tunneling, seems iptun would be the best place for this since BGP is discovering things.
>
> Unless BGP is being used as a resolver, which makes an overlay plugin to employ BGP might make sense.

With the caveat that I do not yet completely understand most of the
RFC, is tunnelling really required, or is it just the requirement that
a system be able to support an IPv6 router address as a next hop for
IPv4 routing decisions?  It feels like that would be much simpler to
implement, and isn't really even BGP-specific; i.e., you could imagine
point-to-point links between routers that just use IPv6 link local
addresses on either end, and yet are still willing to _forward_ IPv4
traffic they get from other interfaces.  The system would use NDP to
locate the router MAC address and then just do what it would have done
with an IPv4 next hop anyway: punt the frame out there with the right
target MAC.

-- 
Joshua M. Clulow
http://blog.sysmgr.org