[PATCH 1/5] NFS: Use non-atomic bit ops when initialising struct nfs_client_initdata

[PATCH 1/5] NFS: Use non-atomic bit ops when initialising struct nfs_client_initdata

[Trond Myklebust]

We don't need atomic bit ops when initialising a local structure on the
stack.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/nfs3client.c | 2 +-
 fs/nfs/nfs4client.c | 6 +++---
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/fs/nfs/nfs3client.c b/fs/nfs/nfs3client.c
index 148ceb74d27c..178dc102442f 100644
--- a/fs/nfs/nfs3client.c
+++ b/fs/nfs/nfs3client.c
@@ -106,7 +106,7 @@ struct nfs_client *nfs3_set_ds_client(struct nfs_server *mds_srv,
 		cl_init.nconnect = mds_clp->cl_nconnect;
 
 	if (mds_srv->flags & NFS_MOUNT_NORESVPORT)
-		set_bit(NFS_CS_NORESVPORT, &cl_init.init_flags);
+		__set_bit(NFS_CS_NORESVPORT, &cl_init.init_flags);
 
 	/* Use the MDS nfs_client cl_ipaddr. */
 	nfs_init_timeout_values(&ds_timeout, ds_proto, ds_timeo, ds_retrans);
diff --git a/fs/nfs/nfs4client.c b/fs/nfs/nfs4client.c
index da6204025a2d..ebc960dd89ff 100644
--- a/fs/nfs/nfs4client.c
+++ b/fs/nfs/nfs4client.c
@@ -882,11 +882,11 @@ static int nfs4_set_client(struct nfs_server *server,
 	if (minorversion > 0 && proto == XPRT_TRANSPORT_TCP)
 		cl_init.nconnect = nconnect;
 	if (server->flags & NFS_MOUNT_NORESVPORT)
-		set_bit(NFS_CS_NORESVPORT, &cl_init.init_flags);
+		__set_bit(NFS_CS_NORESVPORT, &cl_init.init_flags);
 	if (server->options & NFS_OPTION_MIGRATION)
-		set_bit(NFS_CS_MIGRATION, &cl_init.init_flags);
+		__set_bit(NFS_CS_MIGRATION, &cl_init.init_flags);
 	if (test_bit(NFS_MIG_TSM_POSSIBLE, &server->mig_status))
-		set_bit(NFS_CS_TSM_POSSIBLE, &cl_init.init_flags);
+		__set_bit(NFS_CS_TSM_POSSIBLE, &cl_init.init_flags);
 	server->port = rpc_get_port(addr);
 
 	/* Allocate or find a client reference we can use */
-- 
2.21.0

[PATCH 2/5] NFS: Add a flag to tell nfs_client to set RPC_CLNT_CREATE_NOPING

[Trond Myklebust]

Add a flag to tell the nfs_client it should set RPC_CLNT_CREATE_NOPING when
creating the rpc client.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/client.c           | 2 ++
 include/linux/nfs_fs_sb.h | 1 +
 2 files changed, 3 insertions(+)

diff --git a/fs/nfs/client.c b/fs/nfs/client.c
index 30838304a0bf..fa7d92328c72 100644
--- a/fs/nfs/client.c
+++ b/fs/nfs/client.c
@@ -515,6 +515,8 @@ int nfs_create_rpc_client(struct nfs_client *clp,
 		args.flags |= RPC_CLNT_CREATE_NONPRIVPORT;
 	if (test_bit(NFS_CS_INFINITE_SLOTS, &clp->cl_flags))
 		args.flags |= RPC_CLNT_CREATE_INFINITE_SLOTS;
+	if (test_bit(NFS_CS_NOPING, &clp->cl_flags))
+		args.flags |= RPC_CLNT_CREATE_NOPING;
 
 	if (!IS_ERR(clp->cl_rpcclient))
 		return 0;
diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h
index 47266870a235..a50dd432475b 100644
--- a/include/linux/nfs_fs_sb.h
+++ b/include/linux/nfs_fs_sb.h
@@ -45,6 +45,7 @@ struct nfs_client {
 #define NFS_CS_INFINITE_SLOTS	3		/* - don't limit TCP slots */
 #define NFS_CS_NO_RETRANS_TIMEOUT	4	/* - Disable retransmit timeouts */
 #define NFS_CS_TSM_POSSIBLE	5		/* - Maybe state migration */
+#define NFS_CS_NOPING		6		/* - don't ping on connect */
 	struct sockaddr_storage	cl_addr;	/* server identifier */
 	size_t			cl_addrlen;
 	char *			cl_hostname;	/* hostname of server */
-- 
2.21.0

[PATCH 3/5] pNFS: nfs3_set_ds_client should set NFS_CS_NOPING

[Trond Myklebust]

Connecting to the DS is a non-interactive, asynchronous task, so there is
no reason to fire up an extra RPC null ping in order to ensure that the
server is up.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/nfs3client.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/fs/nfs/nfs3client.c b/fs/nfs/nfs3client.c
index 178dc102442f..793fa4273edb 100644
--- a/fs/nfs/nfs3client.c
+++ b/fs/nfs/nfs3client.c
@@ -108,6 +108,8 @@ struct nfs_client *nfs3_set_ds_client(struct nfs_server *mds_srv,
 	if (mds_srv->flags & NFS_MOUNT_NORESVPORT)
 		__set_bit(NFS_CS_NORESVPORT, &cl_init.init_flags);
 
+	__set_bit(NFS_CS_NOPING, &cl_init.init_flags);
+
 	/* Use the MDS nfs_client cl_ipaddr. */
 	nfs_init_timeout_values(&ds_timeout, ds_proto, ds_timeo, ds_retrans);
 	clp = nfs_get_client(&cl_init);
-- 
2.21.0

[PATCH 4/5] NFS/pnfs: Separate NFSv3 DS and MDS traffic

[Trond Myklebust]

If a NFSv3 server is being used as both a DS and as a regular NFSv3 server,
we may want to keep the IO traffic on a separate TCP connection, since
it will typically have very different timeout characteristics.

This patch therefore sets up a flag to separate the two modes of operation
for the nfs_client.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/client.c           | 6 ++++++
 fs/nfs/nfs3client.c       | 1 +
 include/linux/nfs_fs_sb.h | 1 +
 3 files changed, 8 insertions(+)

diff --git a/fs/nfs/client.c b/fs/nfs/client.c
index fa7d92328c72..bd6575ee3b8e 100644
--- a/fs/nfs/client.c
+++ b/fs/nfs/client.c
@@ -312,6 +312,12 @@ static struct nfs_client *nfs_match_client(const struct nfs_client_initdata *dat
 		/* Match nfsv4 minorversion */
 		if (clp->cl_minorversion != data->minorversion)
 			continue;
+
+		/* Match request for a dedicated DS */
+		if (test_bit(NFS_CS_DS, &data->init_flags) !=
+		    test_bit(NFS_CS_DS, &clp->cl_flags))
+			continue;
+
 		/* Match the full socket address */
 		if (!rpc_cmp_addr_port(sap, clap))
 			/* Match all xprt_switch full socket addresses */
diff --git a/fs/nfs/nfs3client.c b/fs/nfs/nfs3client.c
index 793fa4273edb..223904bc40a7 100644
--- a/fs/nfs/nfs3client.c
+++ b/fs/nfs/nfs3client.c
@@ -109,6 +109,7 @@ struct nfs_client *nfs3_set_ds_client(struct nfs_server *mds_srv,
 		__set_bit(NFS_CS_NORESVPORT, &cl_init.init_flags);
 
 	__set_bit(NFS_CS_NOPING, &cl_init.init_flags);
+	__set_bit(NFS_CS_DS, &cl_init.init_flags);
 
 	/* Use the MDS nfs_client cl_ipaddr. */
 	nfs_init_timeout_values(&ds_timeout, ds_proto, ds_timeo, ds_retrans);
diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h
index a50dd432475b..69e80cef5a81 100644
--- a/include/linux/nfs_fs_sb.h
+++ b/include/linux/nfs_fs_sb.h
@@ -46,6 +46,7 @@ struct nfs_client {
 #define NFS_CS_NO_RETRANS_TIMEOUT	4	/* - Disable retransmit timeouts */
 #define NFS_CS_TSM_POSSIBLE	5		/* - Maybe state migration */
 #define NFS_CS_NOPING		6		/* - don't ping on connect */
+#define NFS_CS_DS		7		/* - Server is a DS */
 	struct sockaddr_storage	cl_addr;	/* server identifier */
 	size_t			cl_addrlen;
 	char *			cl_hostname;	/* hostname of server */
-- 
2.21.0

[PATCH 5/5] NFSv4.1: Don't rebind to the same source port when reconnecting to the server

[Trond Myklebust]

NFSv2, v3 and NFSv4 servers often have duplicate replay caches that look
at the source port when deciding whether or not an RPC call is a replay
of a previous call. This requires clients to perform strange TCP gymnastics
in order to ensure that when they reconnect to the server, they bind
to the same source port.

NFSv4.1 and NFSv4.2 have sessions that provide proper replay semantics,
that do not look at the source port of the connection. This patch therefore
ensures they can ignore the rebind requirement.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/lockd/host.c             | 3 ++-
 fs/nfs/client.c             | 3 +++
 fs/nfs/nfs4client.c         | 5 ++++-
 include/linux/nfs_fs_sb.h   | 1 +
 include/linux/sunrpc/clnt.h | 1 +
 include/linux/sunrpc/xprt.h | 3 ++-
 net/sunrpc/clnt.c           | 7 ++++++-
 net/sunrpc/xprtsock.c       | 2 +-
 8 files changed, 20 insertions(+), 5 deletions(-)

diff --git a/fs/lockd/host.c b/fs/lockd/host.c
index 7d46fafdbbe5..0afb6d59bad0 100644
--- a/fs/lockd/host.c
+++ b/fs/lockd/host.c
@@ -464,7 +464,8 @@ nlm_bind_host(struct nlm_host *host)
 			.version	= host->h_version,
 			.authflavor	= RPC_AUTH_UNIX,
 			.flags		= (RPC_CLNT_CREATE_NOPING |
-					   RPC_CLNT_CREATE_AUTOBIND),
+					   RPC_CLNT_CREATE_AUTOBIND |
+					   RPC_CLNT_CREATE_REUSEPORT),
 			.cred		= host->h_cred,
 		};
 
diff --git a/fs/nfs/client.c b/fs/nfs/client.c
index bd6575ee3b8e..02110a30a49e 100644
--- a/fs/nfs/client.c
+++ b/fs/nfs/client.c
@@ -523,6 +523,8 @@ int nfs_create_rpc_client(struct nfs_client *clp,
 		args.flags |= RPC_CLNT_CREATE_INFINITE_SLOTS;
 	if (test_bit(NFS_CS_NOPING, &clp->cl_flags))
 		args.flags |= RPC_CLNT_CREATE_NOPING;
+	if (test_bit(NFS_CS_REUSEPORT, &clp->cl_flags))
+		args.flags |= RPC_CLNT_CREATE_REUSEPORT;
 
 	if (!IS_ERR(clp->cl_rpcclient))
 		return 0;
@@ -670,6 +672,7 @@ static int nfs_init_server(struct nfs_server *server,
 		.timeparms = &timeparms,
 		.cred = server->cred,
 		.nconnect = data->nfs_server.nconnect,
+		.init_flags = (1UL << NFS_CS_REUSEPORT),
 	};
 	struct nfs_client *clp;
 	int error;
diff --git a/fs/nfs/nfs4client.c b/fs/nfs/nfs4client.c
index ebc960dd89ff..abd5af77fe94 100644
--- a/fs/nfs/nfs4client.c
+++ b/fs/nfs/nfs4client.c
@@ -879,8 +879,11 @@ static int nfs4_set_client(struct nfs_server *server,
 	};
 	struct nfs_client *clp;
 
-	if (minorversion > 0 && proto == XPRT_TRANSPORT_TCP)
+	if (minorversion == 0)
+		__set_bit(NFS_CS_REUSEPORT, &cl_init.init_flags);
+	else if (proto == XPRT_TRANSPORT_TCP)
 		cl_init.nconnect = nconnect;
+
 	if (server->flags & NFS_MOUNT_NORESVPORT)
 		__set_bit(NFS_CS_NORESVPORT, &cl_init.init_flags);
 	if (server->options & NFS_OPTION_MIGRATION)
diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h
index 69e80cef5a81..df61ff8981e8 100644
--- a/include/linux/nfs_fs_sb.h
+++ b/include/linux/nfs_fs_sb.h
@@ -47,6 +47,7 @@ struct nfs_client {
 #define NFS_CS_TSM_POSSIBLE	5		/* - Maybe state migration */
 #define NFS_CS_NOPING		6		/* - don't ping on connect */
 #define NFS_CS_DS		7		/* - Server is a DS */
+#define NFS_CS_REUSEPORT	8		/* - reuse src port on reconnect */
 	struct sockaddr_storage	cl_addr;	/* server identifier */
 	size_t			cl_addrlen;
 	char *			cl_hostname;	/* hostname of server */
diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h
index abc63bd1be2b..ec52e78d432b 100644
--- a/include/linux/sunrpc/clnt.h
+++ b/include/linux/sunrpc/clnt.h
@@ -149,6 +149,7 @@ struct rpc_add_xprt_test {
 #define RPC_CLNT_CREATE_NO_IDLE_TIMEOUT	(1UL << 8)
 #define RPC_CLNT_CREATE_NO_RETRANS_TIMEOUT	(1UL << 9)
 #define RPC_CLNT_CREATE_SOFTERR		(1UL << 10)
+#define RPC_CLNT_CREATE_REUSEPORT	(1UL << 11)
 
 struct rpc_clnt *rpc_create(struct rpc_create_args *args);
 struct rpc_clnt	*rpc_bind_new_program(struct rpc_clnt *,
diff --git a/include/linux/sunrpc/xprt.h b/include/linux/sunrpc/xprt.h
index d783e15ba898..ccd35cf4fc41 100644
--- a/include/linux/sunrpc/xprt.h
+++ b/include/linux/sunrpc/xprt.h
@@ -207,7 +207,8 @@ struct rpc_xprt {
 	unsigned int		min_reqs;	/* min number of slots */
 	unsigned int		num_reqs;	/* total slots */
 	unsigned long		state;		/* transport state */
-	unsigned char		resvport   : 1; /* use a reserved port */
+	unsigned char		resvport   : 1,	/* use a reserved port */
+				reuseport  : 1; /* reuse port on reconnect */
 	atomic_t		swapper;	/* we're swapping over this
 						   transport */
 	unsigned int		bind_index;	/* bind function index */
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index f7f78566be46..5baf9b9be2e8 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -591,6 +591,9 @@ struct rpc_clnt *rpc_create(struct rpc_create_args *args)
 	xprt->resvport = 1;
 	if (args->flags & RPC_CLNT_CREATE_NONPRIVPORT)
 		xprt->resvport = 0;
+	xprt->reuseport = 0;
+	if (args->flags & RPC_CLNT_CREATE_REUSEPORT)
+		xprt->reuseport = 1;
 
 	clnt = rpc_create_xprt(args, xprt);
 	if (IS_ERR(clnt) || args->nconnect <= 1)
@@ -2906,7 +2909,7 @@ int rpc_clnt_add_xprt(struct rpc_clnt *clnt,
 	struct rpc_xprt *xprt;
 	unsigned long connect_timeout;
 	unsigned long reconnect_timeout;
-	unsigned char resvport;
+	unsigned char resvport, reuseport;
 	int ret = 0;
 
 	rcu_read_lock();
@@ -2918,6 +2921,7 @@ int rpc_clnt_add_xprt(struct rpc_clnt *clnt,
 		return -EAGAIN;
 	}
 	resvport = xprt->resvport;
+	reuseport = xprt->reuseport;
 	connect_timeout = xprt->connect_timeout;
 	reconnect_timeout = xprt->max_reconnect_timeout;
 	rcu_read_unlock();
@@ -2928,6 +2932,7 @@ int rpc_clnt_add_xprt(struct rpc_clnt *clnt,
 		goto out_put_switch;
 	}
 	xprt->resvport = resvport;
+	xprt->reuseport = reuseport;
 	if (xprt->ops->set_connect_timeout != NULL)
 		xprt->ops->set_connect_timeout(xprt,
 				connect_timeout,
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 70e52f567b2a..98e2d40b2d6a 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -1752,7 +1752,7 @@ static void xs_set_port(struct rpc_xprt *xprt, unsigned short port)
 
 static void xs_set_srcport(struct sock_xprt *transport, struct socket *sock)
 {
-	if (transport->srcport == 0)
+	if (transport->srcport == 0 && transport->xprt.reuseport)
 		transport->srcport = xs_sock_getport(sock);
 }
 
-- 
2.21.0