From: "Christoph Böhmwalder" <christoph.boehmwalder@linbit.com>
To: Jakub Kicinski <kuba@kernel.org>
Cc: Philipp Reisner <philipp.reisner@linbit.com>,
"David S. Miller" <davem@davemloft.net>,
Eric Dumazet <edumazet@google.com>,
Paolo Abeni <pabeni@redhat.com>, Simon Horman <horms@kernel.org>,
Alok Tiwari <alok.a.tiwari@oracle.com>,
Kees Cook <kees@kernel.org>,
netdev@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [RFC PATCH] genetlink: add multi-version family support for protocol negotiation
Date: Wed, 18 Mar 2026 15:54:04 +0100 [thread overview]
Message-ID: <b17134a2-c95e-4019-800e-ab6ec6982ea3@linbit.com> (raw)
In-Reply-To: <20260317151745.4562eefc@kernel.org>
Am 17.03.26 um 23:17 schrieb Jakub Kicinski:
> On Tue, 17 Mar 2026 16:47:44 +0100 Christoph Böhmwalder wrote:
>> Am 16.03.26 um 19:37 schrieb Jakub Kicinski:
>>> What's the delta between the families? Do you have any examples of
>>> the the version ends up being used?
>>
>> The main difference between DRBD 8 and 9 in general is that 8 only
>> replicates to a single peer, while 9 has support for multiple peers.
>> This also introduces major changes in the genl family. For example, the
>> "connect" command works completely differently between the v1 and v2
>> families.
>> In v1, it takes all the net_conf parameters and immediately sets up the
>> connection.
>> In v2, you need a new_peer / new_path command first; the net_conf is
>> passed to new_peer in this case. The connect command then only
>> "activates" the connection.
>> So the semantics are completely different.
>>
>> Other commands (like "state_info") were removed entirely and split up
>> into different commands.
>>
>> The specific exact diff is available as part of our prototype DRBD 9
>> branch (included in linux-next):
>> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/diff/include/linux/drbd_genl.h?id=11b8887a3efae2868037f2bd8dbbc68a8591f66c
>
> This uAPI header wrapped in DRDB magic tells me nothing.
>
> The macro wrapping that DRBD does never gained broader adoption and
> should be considered deprecated. Please take a look at YNL and the YAML
> specs. (I mean this as a nack).
We are actually working on exactly that as part of our upstreaming
efforts (implementing YNL-based generation of the genl headers for
DRBD). So the macro magic will be removed soon.
>> As for the usage: the plan is for drbd to register its family with
>> min_version=1, max_version=2. Then we would dispatch the v2 commands
>> normally, and the v1 commands through a thin compat layer.
>> In the in-tree DRBD, the version is currently unused. In the (still
>> out-of-tree) DRBD 9 though, we do check that the version is what we expect:
>> https://github.com/LINBIT/drbd/blob/master/drbd/drbd_nl.c#L426
>>
>> This would be replaced by the mentioned compat dispatch logic once we
>> have in-kernel infrastructure for supporting multiple versions.
>
> I still feel a little bit in the dark as to what you're doing but the
> normal path for Netlink is to add new attributes and commands in a
> backward-compat manner. In case of DRBD given how "special" the
> existing code is we could probably start a new family using modern
> constructs.
The new family indeed sounds like a way out. So we would register two
families: the old "drbd" for the compat case, and a new "drbd2" with the
new command set. Would that be more in line with the "standard approach"?
Thanks,
Christoph
--
Christoph Böhmwalder
LINBIT | Keeping the Digital World Running
DRBD HA — Disaster Recovery — Software defined Storage
next prev parent reply other threads:[~2026-03-18 14:54 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-03-16 15:15 [RFC PATCH] genetlink: add multi-version family support for protocol negotiation Christoph Böhmwalder
2026-03-16 18:37 ` Jakub Kicinski
2026-03-17 15:47 ` Christoph Böhmwalder
2026-03-17 22:17 ` Jakub Kicinski
2026-03-18 14:54 ` Christoph Böhmwalder [this message]
2026-03-18 22:27 ` Jakub Kicinski
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=b17134a2-c95e-4019-800e-ab6ec6982ea3@linbit.com \
--to=christoph.boehmwalder@linbit.com \
--cc=alok.a.tiwari@oracle.com \
--cc=davem@davemloft.net \
--cc=edumazet@google.com \
--cc=horms@kernel.org \
--cc=kees@kernel.org \
--cc=kuba@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=philipp.reisner@linbit.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox