All of lore.kernel.org
 help / color / mirror / Atom feed
From: Stefano Brivio <sbrivio@redhat.com>
To: Tj <tj.iam.tj@proton.me>
Cc: "Íñigo Huguet" <ihuguet@redhat.com>,
	"Thorsten Leemhuis" <regressions@leemhuis.info>,
	"Fernando Fernandez Mancera" <fmancera@suse.de>,
	"Jakub Kicinski" <kuba@kernel.org>,
	netdev@vger.kernel.org, "Yumei Huang" <yuhuang@redhat.com>,
	"Ido Schimmel" <idosch@idosch.org>,
	"Justin Iurman" <justin.iurman@gmail.com>,
	"David Ahern" <dsahern@kernel.org>,
	"David Gibson" <david@gibson.dropbear.id.au>,
	"Linux kernel regressions list" <regressions@lists.linux.dev>,
	"Beniamino Galvani" <bgalvani@redhat.com>
Subject: Re: Problem with IPv6 privacy addresses in 7.0
Date: Mon, 01 Jun 2026 10:08:27 +0200 (CEST)	[thread overview]
Message-ID: <20260601100826.4222a657@elisabeth> (raw)
In-Reply-To: <3f0d5431-60b0-42dc-8192-2f8a3c6da50e@proton.me>

On Fri, 29 May 2026 20:18:55 +0000
Tj <tj.iam.tj@proton.me> wrote:

> On 29/05/2026 20:04, Stefano Brivio wrote:
> > On Fri, 29 May 2026 18:28:58 +0000
> > Tj <tj.iam.tj@proton.me> wrote:
> >  
> >> I believe I hit this on a router using Debian 13 with v7.0.* kernel this week that
> >> uses systemd-networkd to configure IPV6 RA and prefix delegation after
> >> moving from v6.19.*.
> >>
> >> Symptom was the router could no longer reach public IPv6 addresses
> >> itself but forwarding was unaffected.
> >>
> >> The ISP (Starlink) provides a /64 prefix via RA and a /56 via DHCPv6. networkd
> >> allocates a static suffix address from both to the WAN-side interface.
> >>
> >> I discovered after much experimentation that instead of the usual /56
> >> address being the source it was choosing the /64 and failing.  
> > Do you really mean an address configured as /56, or a /64 address that
> > systemd-networkd derives from a /56 delegated prefix?
>
> Apologies - I should have been clearer. I meant addresses generated 
> *from the* respective prefixes - the addresses themselves are /64, yes!

Thanks for clarifying. Meanwhile, Jakub applied Fernando's revert patch
(I'm actually relieved, thanks Jakub ;)) but I had a look anyway into
how systemd-networkd handles these two cases.

They both seem to be hitting in the same way configure_address()
(src/network/networkd-address.c) with no attempt to define a given
order. After all, the only way to force a specific order would be to
remove and add addresses back, which systemd-networkd doesn't do.

So, whether the setup worked before the patch (and after the revert)
depends on the order of events / configurations, which in some sense
makes it even more important that we keep a stable order by default.

Contrary to NetworkManager, here, there's nothing we could really "fix".

-- 
Stefano


  reply	other threads:[~2026-06-01  8:08 UTC|newest]

Thread overview: 36+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-05-29 18:28 Problem with IPv6 privacy addresses in 7.0 Tj
2026-05-29 20:04 ` Stefano Brivio
2026-05-29 20:18   ` Tj
2026-06-01  8:08     ` Stefano Brivio [this message]
  -- strict thread matches above, loose matches on Subject: below --
2026-05-21 13:53 Chris Adams
2026-05-27  0:57 ` Jakub Kicinski
2026-05-27  1:06   ` Chris Adams
2026-05-27  1:31     ` Jakub Kicinski
2026-05-27 21:13       ` Chris Adams
2026-05-27 21:16         ` Fernando Fernandez Mancera
2026-05-27 21:51         ` Jakub Kicinski
2026-05-27 21:51       ` Chris Adams
2026-05-27 21:59         ` Fernando Fernandez Mancera
2026-05-27 23:07           ` Jakub Kicinski
2026-05-28  5:38           ` Stefano Brivio
2026-05-28 10:46             ` Fernando Fernandez Mancera
2026-05-28 11:12               ` Stefano Brivio
2026-05-28 11:29                 ` Fernando Fernandez Mancera
2026-05-28 12:29                   ` Thorsten Leemhuis
2026-05-28 13:32                     ` Stefano Brivio
2026-05-28 14:02                       ` Thorsten Leemhuis
2026-05-28 14:15                       ` Íñigo Huguet
2026-05-28 14:53                         ` Stefano Brivio
2026-05-28 15:24                           ` Íñigo Huguet
2026-05-28 16:01                             ` Beniamino Galvani
2026-05-28 17:21                               ` Stefano Brivio
2026-05-28 18:42                                 ` Fernando Fernandez Mancera
2026-05-28 18:50                                   ` Fernando Fernandez Mancera
2026-05-28 19:22                                     ` Stefano Brivio
2026-05-29  4:47                                       ` David Gibson
2026-05-29  8:40                                       ` Beniamino Galvani
2026-05-29 17:40                                         ` Stefano Brivio
2026-05-28 14:34                       ` Andrew Lunn
2026-05-28 15:17                         ` Stefano Brivio
2026-05-29  4:48                           ` David Gibson
2026-05-29  9:42                             ` Stefano Brivio

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260601100826.4222a657@elisabeth \
    --to=sbrivio@redhat.com \
    --cc=bgalvani@redhat.com \
    --cc=david@gibson.dropbear.id.au \
    --cc=dsahern@kernel.org \
    --cc=fmancera@suse.de \
    --cc=idosch@idosch.org \
    --cc=ihuguet@redhat.com \
    --cc=justin.iurman@gmail.com \
    --cc=kuba@kernel.org \
    --cc=netdev@vger.kernel.org \
    --cc=regressions@leemhuis.info \
    --cc=regressions@lists.linux.dev \
    --cc=tj.iam.tj@proton.me \
    --cc=yuhuang@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.