[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-id: <172160430713.18529.16870160544594676027@noble.neil.brown.name>
Date: Mon, 22 Jul 2024 09:25:07 +1000
From: "NeilBrown" <neilb@...e.de>
To: "Chuck Lever" <chuck.lever@...cle.com>
Cc: "Jeff Layton" <jlayton@...nel.org>, "Olga Kornievskaia" <kolga@...app.com>,
"Dai Ngo" <Dai.Ngo@...cle.com>, "Tom Talpey" <tom@...pey.com>,
"Lorenzo Bianconi" <lorenzo@...nel.org>, linux-nfs@...r.kernel.org,
linux-kernel@...r.kernel.org, "Steve Dickson" <steved@...hat.com>
Subject:
Re: [PATCH] nfsd: don't set SVC_SOCK_ANONYMOUS when creating nfsd sockets
On Sat, 20 Jul 2024, Chuck Lever wrote:
> On Fri, Jul 19, 2024 at 02:55:53PM -0400, Jeff Layton wrote:
> > When creating nfsd sockets via the netlink interface, we do want to
> > register with the portmapper. Don't set SVC_SOCK_ANONYMOUS.
>
> NFSD's RDMA transports don't register with rpcbind, for example.
For RDMA, what does inet->sk_family and in->sk_protocol contain in
svc_setup_socket()?
Could that code detect that it doesn't make sense to register the socket
with rpcbind?
NeilBrown
Powered by blists - more mailing lists