Received-SPF: pass (google.com: domain of linux-nfs+bounces-2960-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
From: Jeff Layton <jlayton@kernel.org>
Subject: [PATCH nfs-utils v3 0/3] nfsdctl: new nfs-utils tool for managing
 the kernel NFS server
Date: Tue, 23 Apr 2024 12:31:52 -0400
Message-Id: <20240423-nfsdctl-v3-0-9e68181c846d@kernel.org>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit
To: Steve Dickson <steved@redhat.com>, 
 Lorenzo Bianconi <lorenzo@kernel.org>, Chuck Lever <chuck.lever@oracle.com>, 
 Neil Brown <neilb@suse.de>
Cc: linux-nfs@vger.kernel.org, Jeff Layton <jlayton@kernel.org>

Lorenzo posted an updated version of his netlink interface patches [1],
so here is an updated version of the companion userland patches. To be
clear, to test these, you need a kernel with his netlink patches.

The series adds a new tool to nfs-utils called nfsdctl, intended
as an eventual replacement for rpc.nfsd. It's a subcommand based
interface like nmcli or virsh, so we can easily expand the interface
later to deal with new sorts of configuration.

This version of the tool should be at feature parity with rpc.nfsd, at
least as far as autostarting the server. This posting also includes a
manpage and an update to the nfs-server systemctl service, to start
using the new interface when possible.

I've also included a patch that adds the manpage source. It's much nicer
to edit that and regenerate it if we have to update it later. We can
drop that patch if you just want to keep the result though.

The one thing that's not quite right here is the way the nfsd_netlink.h
file is handled. This set includes a copy of the proposed header, but it
would be better to build against the UAPI header in the kernel-headers
package instead. Older kernels have a subset of the new interface
though, so we can't build this against that file universally.

Is there a good way to test for the presence of an enum value in
autoconf? I didn't see any macros for it, but maybe there is some
generic test for C symbols we can use?

[1]: https://lore.kernel.org/linux-nfs/cover.1713878413.git.lorenzo@kernel.org/T/#m5fd847189894f58e93706c40340e18858f242a27

Signed-off-by: Jeff Layton <jlayton@kernel.org>
---
Changes in v3:
- split nfsdctl.h so we can include the UAPI header as-is
- squash the patches together that added Lorenzo's version and convert
  it to the new interface
- adapt to latest version of netlink interface changes
  + have THREADS_SET/GET report an array of thread counts (one per pool)
  + pass scope in as a string to THREADS_SET instead of using unshare() trick

Changes in v2:
- Adapt to latest kernel netlink interface changes (in particular, send
  the leastime and gracetime when they are set in the config).
- More help text for different subcommands
- New nfsdctl(8) manpage
- Patch to make systemd preferentially use nfsdctl instead of rpc.nfsd
- Link to v1: https://lore.kernel.org/r/20240412-nfsdctl-v1-0-efd6dcebcc04@kernel.org

---
Jeff Layton (2):
      nfsdctl: asciidoc source for the manpage
      systemd: use nfsdctl to start and stop the nfs server

Lorenzo Bianconi (1):
      nfsdctl: add the nfsdctl utility to nfs-utils

 configure.ac                 |   13 +
 systemd/nfs-server.service   |    4 +-
 utils/Makefile.am            |    4 +
 utils/nfsdctl/Makefile.am    |   13 +
 utils/nfsdctl/nfsd_netlink.h |   86 +++
 utils/nfsdctl/nfsdctl.8      |  274 ++++++++
 utils/nfsdctl/nfsdctl.adoc   |  140 +++++
 utils/nfsdctl/nfsdctl.c      | 1426 ++++++++++++++++++++++++++++++++++++++++++
 utils/nfsdctl/nfsdctl.h      |   93 +++
 9 files changed, 2051 insertions(+), 2 deletions(-)
---
base-commit: c6aa75d25b79121c4cf83ae09a04f8728c4e6593
change-id: 20240412-nfsdctl-fa8bd8430cfd

Best regards,
-- 
Jeff Layton <jlayton@kernel.org>