From: Jeff Layton Subject: Re: [PATCH] sunrpc: split the vs_hidden flag into TCP and UDP variants Date: Fri, 25 Jun 2010 09:22:35 -0400 Message-ID: <20100625092235.20a5f4bf@tlielax.poochiereds.net> References: <1277429621-9579-1-git-send-email-jlayton@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Cc: , To: Return-path: Received: from mx1.redhat.com ([209.132.183.28]:44796 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755793Ab0FYNWp (ORCPT ); Fri, 25 Jun 2010 09:22:45 -0400 In-Reply-To: Sender: linux-nfs-owner@vger.kernel.org List-ID: On Fri, 25 Jun 2010 08:58:12 -0400 wrote: > Does this mean that the Linux NFS server will no longer respond to NFSv4 > requests coming in over UDP? If it will continue to do so, wouldn't it > be more important to solve that correctly? > > Thanx... > > ps > I think that would be a good thing to do too, but that's a bit more invasive. The RPC layer doesn't really have infrastructure at the moment to say "don't service requests for this protocol version on this transport". I'd eventually like to see that happen, and I'm willing to work on it as well, but I think having the kernel not advertise v4 over UDP is a reasonable first step. The other piece of this (that I believe steved is working on) is to make mount.nfs not attempt to do v4 over UDP either. I'm not certain of how he's planning to implement it, but it seems reasonable that we can just make it skip to v3 when someone specifies "proto=udp" and doesn't specify a version explicitly. > > -----Original Message----- > From: linux-nfs-owner@vger.kernel.org > [mailto:linux-nfs-owner@vger.kernel.org] On Behalf Of Jeff Layton > Sent: Thursday, June 24, 2010 9:34 PM > To: bfields@fieldses.org > Cc: linux-nfs@vger.kernel.org > Subject: [PATCH] sunrpc: split the vs_hidden flag into TCP and UDP > variants > > RFC3530 states: > > Where an NFS version 4 implementation supports operation over the IP > network protocol, the supported transports between NFS and IP MUST be > among the IETF-approved congestion control transport protocols, which > include TCP and SCTP > > The NFS server currently registers the NFSv4 UDP port with the > portmapper, which it really shouldn't do. This patch splits the > vs_hidden flag into TCP and UDP variants, and sets vs_hidden_udp for the > NFSv4 service. > > Reported-by: Sachin Prabhu > Signed-off-by: Jeff Layton > --- > fs/nfs/callback_xdr.c | 3 ++- > fs/nfsd/nfs2acl.c | 1 - > fs/nfsd/nfs3acl.c | 1 - > fs/nfsd/nfs4proc.c | 1 + > include/linux/sunrpc/svc.h | 5 +++-- > net/sunrpc/svc.c | 15 +++++++++++---- > 6 files changed, 17 insertions(+), 9 deletions(-) > > diff --git a/fs/nfs/callback_xdr.c b/fs/nfs/callback_xdr.c > index 05af212..30baf97 100644 > --- a/fs/nfs/callback_xdr.c > +++ b/fs/nfs/callback_xdr.c > @@ -783,7 +783,8 @@ struct svc_version nfs4_callback_version1 = { > .vs_proc = nfs4_callback_procedures1, > .vs_xdrsize = NFS4_CALLBACK_XDRSIZE, > .vs_dispatch = NULL, > - .vs_hidden = 1, > + .vs_hidden_tcp = true, > + .vs_hidden_udp = true, > }; > > struct svc_version nfs4_callback_version4 = { > diff --git a/fs/nfsd/nfs2acl.c b/fs/nfsd/nfs2acl.c > index 6aa5590..ec602b4 100644 > --- a/fs/nfsd/nfs2acl.c > +++ b/fs/nfsd/nfs2acl.c > @@ -352,5 +352,4 @@ struct svc_version nfsd_acl_version2 = { > .vs_proc = nfsd_acl_procedures2, > .vs_dispatch = nfsd_dispatch, > .vs_xdrsize = NFS3_SVC_XDRSIZE, > - .vs_hidden = 0, > }; > diff --git a/fs/nfsd/nfs3acl.c b/fs/nfsd/nfs3acl.c > index a596e9d..1949a4c 100644 > --- a/fs/nfsd/nfs3acl.c > +++ b/fs/nfsd/nfs3acl.c > @@ -262,6 +262,5 @@ struct svc_version nfsd_acl_version3 = { > .vs_proc = nfsd_acl_procedures3, > .vs_dispatch = nfsd_dispatch, > .vs_xdrsize = NFS3_SVC_XDRSIZE, > - .vs_hidden = 0, > }; > > diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c > index 59ec449..4dfbf08 100644 > --- a/fs/nfsd/nfs4proc.c > +++ b/fs/nfsd/nfs4proc.c > @@ -1365,6 +1365,7 @@ struct svc_version nfsd_version4 = { > .vs_proc = nfsd_procedures4, > .vs_dispatch = nfsd_dispatch, > .vs_xdrsize = NFS4_SVC_XDRSIZE, > + .vs_hidden_udp = true, > }; > > /* > diff --git a/include/linux/sunrpc/svc.h b/include/linux/sunrpc/svc.h > index 5a3085b..27ff713 100644 > --- a/include/linux/sunrpc/svc.h > +++ b/include/linux/sunrpc/svc.h > @@ -370,8 +370,9 @@ struct svc_version { > struct svc_procedure * vs_proc; /* per-procedure info */ > u32 vs_xdrsize; /* xdrsize needed for > this version */ > > - unsigned int vs_hidden : 1; /* Don't register with > portmapper. > - * Only used for nfsacl > so far. */ > + /* these flags control whether program is registered with > rpcbind */ > + bool vs_hidden_tcp; > + bool vs_hidden_udp; > > /* Override dispatch function (e.g. when caching replies). > * A return value of 0 means drop the request. > diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c > index d9017d6..487792c 100644 > --- a/net/sunrpc/svc.c > +++ b/net/sunrpc/svc.c > @@ -868,6 +868,7 @@ int svc_register(const struct svc_serv *serv, const > int family, > struct svc_program *progp; > unsigned int i; > int error = 0; > + bool vs_hidden; > > BUG_ON(proto == 0 && port == 0); > > @@ -876,16 +877,20 @@ int svc_register(const struct svc_serv *serv, > const int family, > if (progp->pg_vers[i] == NULL) > continue; > > + vs_hidden = (proto == IPPROTO_UDP) ? > + progp->pg_vers[i]->vs_hidden_udp : > + progp->pg_vers[i]->vs_hidden_tcp; > + > dprintk("svc: svc_register(%sv%d, %s, %u, > %u)%s\n", > progp->pg_name, > i, > - proto == IPPROTO_UDP? "udp" : > "tcp", > + proto == IPPROTO_UDP ? "udp" : > "tcp", > port, > family, > - progp->pg_vers[i]->vs_hidden? > + vs_hidden ? > " (but not telling > portmap)" : ""); > > - if (progp->pg_vers[i]->vs_hidden) > + if (vs_hidden) > continue; > > error = __svc_register(progp->pg_name, > progp->pg_prog, > @@ -943,7 +948,9 @@ static void svc_unregister(const struct svc_serv > *serv) > for (i = 0; i < progp->pg_nvers; i++) { > if (progp->pg_vers[i] == NULL) > continue; > - if (progp->pg_vers[i]->vs_hidden) > + > + if (progp->pg_vers[i]->vs_hidden_tcp && > + progp->pg_vers[i]->vs_hidden_udp) > continue; > > __svc_unregister(progp->pg_prog, i, > progp->pg_name); -- Jeff Layton