[PATCH Version 10 00/12] pNFS file layout session trunking

[PATCH Version 10 00/12] pNFS file layout session trunking

[andros]

From: Andy Adamson <andros@netapp.com>

version-10:
- rebased to Anna's linux-next branch
- responded to Anna's version-9 cleanup review comments which effects
  these patches:
	NFS detect session trunking
	NFS test session trunking with exchange id
	NFS pnfs data server multipath session trunking	

version-9 RESEND: 

- added put_rpccred to to _nfs4_pnfs_v4_ds_connect in
  "NFS pnfs data server multipath session trunking"

version-9:
- renamed rpc_clnt_xprt_switch_find_addr
- spelling nits
- moved rpc_cnt_xprt_switch_add_xprt into nfs4_detect_session_trunking


Andy Adamson (12):
  NFS setup async exchange_id
  NFS refactor nfs4_match_clientids
  NFS refactor nfs4_check_serverowner_major_id
  NFS detect session trunking
  SUNRPC remove rpc_task_release_client from rpc_task_set_client
  SUNRPC rpc_clnt_xprt_switch_put
  SUNRPC rpc_clnt_xprt_switch_add_xprt
  SUNRPC search xprt switch for sockaddr
  SUNRPC: rpc_clnt_add_xprt setup function for NFS layer
  NFS add xprt switch addrs test to match client
  NFS test session trunking with exchange id
  NFS pnfs data server multipath session trunking

 fs/nfs/client.c                      |   5 +-
 fs/nfs/internal.h                    |   3 +
 fs/nfs/nfs4_fs.h                     |   8 +
 fs/nfs/nfs4client.c                  | 115 ++++++++++++--
 fs/nfs/nfs4proc.c                    | 287 ++++++++++++++++++++++++-----------
 fs/nfs/pnfs_nfs.c                    |  58 +++++--
 include/linux/sunrpc/clnt.h          |  17 +++
 include/linux/sunrpc/xprtmultipath.h |   2 +
 net/sunrpc/clnt.c                    |  93 +++++++++++-
 net/sunrpc/xprtmultipath.c           |  24 ++-
 10 files changed, 498 insertions(+), 114 deletions(-)

-- 
1.8.3.1

[PATCH Version 10 01/12] NFS setup async exchange_id

[andros]

From: Andy Adamson <andros@netapp.com>

Testing an rpc_xprt for session trunking should not delay application
progress over already established transports.
Setup exchange_id to be able to be an async call to test an rpc_xprt
for session trunking use.

Signed-off-by: Andy Adamson <andros@netapp.com>
---
 fs/nfs/nfs4proc.c | 215 ++++++++++++++++++++++++++++++++++--------------------
 1 file changed, 134 insertions(+), 81 deletions(-)

diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 7f360e6..98bd30e 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -7048,6 +7048,80 @@ static int nfs4_sp4_select_mode(struct nfs_client *clp,
 	return 0;
 }
 
+struct nfs41_exchange_id_data {
+	struct nfs41_exchange_id_res res;
+	struct nfs41_exchange_id_args args;
+	int rpc_status;
+};
+
+static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
+{
+	struct nfs41_exchange_id_data *cdata =
+					(struct nfs41_exchange_id_data *)data;
+	struct nfs_client *clp = cdata->args.client;
+	int status = task->tk_status;
+
+	trace_nfs4_exchange_id(clp, status);
+
+	if (status == 0)
+		status = nfs4_check_cl_exchange_flags(cdata->res.flags);
+	if (status  == 0)
+		status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
+
+	if (status == 0) {
+		clp->cl_clientid = cdata->res.clientid;
+		clp->cl_exchange_flags = cdata->res.flags;
+		/* Client ID is not confirmed */
+		if (!(cdata->res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
+			clear_bit(NFS4_SESSION_ESTABLISHED,
+			&clp->cl_session->session_state);
+			clp->cl_seqid = cdata->res.seqid;
+		}
+
+		kfree(clp->cl_serverowner);
+		clp->cl_serverowner = cdata->res.server_owner;
+		cdata->res.server_owner = NULL;
+
+		/* use the most recent implementation id */
+		kfree(clp->cl_implid);
+		clp->cl_implid = cdata->res.impl_id;
+		cdata->res.impl_id = NULL;
+
+		if (clp->cl_serverscope != NULL &&
+		    !nfs41_same_server_scope(clp->cl_serverscope,
+					cdata->res.server_scope)) {
+			dprintk("%s: server_scope mismatch detected\n",
+				__func__);
+			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
+			kfree(clp->cl_serverscope);
+			clp->cl_serverscope = NULL;
+		}
+
+		if (clp->cl_serverscope == NULL) {
+			clp->cl_serverscope = cdata->res.server_scope;
+			cdata->res.server_scope = NULL;
+		}
+	}
+	cdata->rpc_status = status;
+}
+
+static void nfs4_exchange_id_release(void *data)
+{
+	struct nfs41_exchange_id_data *cdata =
+					(struct nfs41_exchange_id_data *)data;
+
+	nfs_put_client(cdata->args.client);
+	kfree(cdata->res.impl_id);
+	kfree(cdata->res.server_scope);
+	kfree(cdata->res.server_owner);
+	kfree(cdata);
+}
+
+static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
+	.rpc_call_done = nfs4_exchange_id_done,
+	.rpc_release = nfs4_exchange_id_release,
+};
+
 /*
  * _nfs4_proc_exchange_id()
  *
@@ -7057,66 +7131,60 @@ static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
 	u32 sp4_how)
 {
 	nfs4_verifier verifier;
-	struct nfs41_exchange_id_args args = {
-		.verifier = &verifier,
-		.client = clp,
-#ifdef CONFIG_NFS_V4_1_MIGRATION
-		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
-			 EXCHGID4_FLAG_BIND_PRINC_STATEID |
-			 EXCHGID4_FLAG_SUPP_MOVED_MIGR,
-#else
-		.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
-			 EXCHGID4_FLAG_BIND_PRINC_STATEID,
-#endif
-	};
-	struct nfs41_exchange_id_res res = {
-		0
-	};
-	int status;
 	struct rpc_message msg = {
 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
-		.rpc_argp = &args,
-		.rpc_resp = &res,
 		.rpc_cred = cred,
 	};
+	struct rpc_task_setup task_setup_data = {
+		.rpc_client = clp->cl_rpcclient,
+		.callback_ops = &nfs4_exchange_id_call_ops,
+		.rpc_message = &msg,
+		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
+	};
+	struct nfs41_exchange_id_data *calldata;
+	struct rpc_task *task;
+	int status = -EIO;
+
+	if (!atomic_inc_not_zero(&clp->cl_count))
+		goto out;
+
+	status = -ENOMEM;
+	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
+	if (!calldata)
+		goto out;
 
 	nfs4_init_boot_verifier(clp, &verifier);
 
 	status = nfs4_init_uniform_client_string(clp);
 	if (status)
-		goto out;
+		goto out_calldata;
 
 	dprintk("NFS call  exchange_id auth=%s, '%s'\n",
 		clp->cl_rpcclient->cl_auth->au_ops->au_name,
 		clp->cl_owner_id);
 
-	res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
-					GFP_NOFS);
-	if (unlikely(res.server_owner == NULL)) {
-		status = -ENOMEM;
-		goto out;
-	}
+	calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
+						GFP_NOFS);
+	status = -ENOMEM;
+	if (unlikely(calldata->res.server_owner == NULL))
+		goto out_calldata;
 
-	res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
+	calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
 					GFP_NOFS);
-	if (unlikely(res.server_scope == NULL)) {
-		status = -ENOMEM;
+	if (unlikely(calldata->res.server_scope == NULL))
 		goto out_server_owner;
-	}
 
-	res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
-	if (unlikely(res.impl_id == NULL)) {
-		status = -ENOMEM;
+	calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
+	if (unlikely(calldata->res.impl_id == NULL))
 		goto out_server_scope;
-	}
 
 	switch (sp4_how) {
 	case SP4_NONE:
-		args.state_protect.how = SP4_NONE;
+		calldata->args.state_protect.how = SP4_NONE;
 		break;
 
 	case SP4_MACH_CRED:
-		args.state_protect = nfs4_sp4_mach_cred_request;
+		calldata->args.state_protect = nfs4_sp4_mach_cred_request;
 		break;
 
 	default:
@@ -7126,55 +7194,30 @@ static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
 		goto out_impl_id;
 	}
 
-	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
-	trace_nfs4_exchange_id(clp, status);
-	if (status == 0)
-		status = nfs4_check_cl_exchange_flags(res.flags);
-
-	if (status == 0)
-		status = nfs4_sp4_select_mode(clp, &res.state_protect);
-
-	if (status == 0) {
-		clp->cl_clientid = res.clientid;
-		clp->cl_exchange_flags = res.flags;
-		/* Client ID is not confirmed */
-		if (!(res.flags & EXCHGID4_FLAG_CONFIRMED_R)) {
-			clear_bit(NFS4_SESSION_ESTABLISHED,
-					&clp->cl_session->session_state);
-			clp->cl_seqid = res.seqid;
-		}
-
-		kfree(clp->cl_serverowner);
-		clp->cl_serverowner = res.server_owner;
-		res.server_owner = NULL;
-
-		/* use the most recent implementation id */
-		kfree(clp->cl_implid);
-		clp->cl_implid = res.impl_id;
-		res.impl_id = NULL;
-
-		if (clp->cl_serverscope != NULL &&
-		    !nfs41_same_server_scope(clp->cl_serverscope,
-					     res.server_scope)) {
-			dprintk("%s: server_scope mismatch detected\n",
-				__func__);
-			set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
-			kfree(clp->cl_serverscope);
-			clp->cl_serverscope = NULL;
-		}
+	calldata->args.verifier = &verifier;
+	calldata->args.client = clp;
+#ifdef CONFIG_NFS_V4_1_MIGRATION
+	calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
+	EXCHGID4_FLAG_BIND_PRINC_STATEID |
+	EXCHGID4_FLAG_SUPP_MOVED_MIGR,
+#else
+	calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
+	EXCHGID4_FLAG_BIND_PRINC_STATEID,
+#endif
+	msg.rpc_argp = &calldata->args;
+	msg.rpc_resp = &calldata->res;
+	task_setup_data.callback_data = calldata;
 
-		if (clp->cl_serverscope == NULL) {
-			clp->cl_serverscope = res.server_scope;
-			res.server_scope = NULL;
-		}
+	task = rpc_run_task(&task_setup_data);
+	if (IS_ERR(task)) {
+	status = PTR_ERR(task);
+		goto out_impl_id;
 	}
 
-out_impl_id:
-	kfree(res.impl_id);
-out_server_scope:
-	kfree(res.server_scope);
-out_server_owner:
-	kfree(res.server_owner);
+	status = rpc_wait_for_completion_task(task);
+	if (!status)
+		status = calldata->rpc_status;
+	rpc_put_task(task);
 out:
 	if (clp->cl_implid != NULL)
 		dprintk("NFS reply exchange_id: Server Implementation ID: "
@@ -7184,6 +7227,16 @@ out:
 			clp->cl_implid->date.nseconds);
 	dprintk("NFS reply exchange_id: %d\n", status);
 	return status;
+
+out_impl_id:
+	kfree(calldata->res.impl_id);
+out_server_scope:
+	kfree(calldata->res.server_scope);
+out_server_owner:
+	kfree(calldata->res.server_owner);
+out_calldata:
+	kfree(calldata);
+	goto out;
 }
 
 /*
-- 
1.8.3.1

[PATCH Version 10 02/12] NFS refactor nfs4_match_clientids

[andros]

From: Andy Adamson <andros@netapp.com>

For session trunking, to compare nfs41_exchange_id_res with
exiting nfs_client.

Signed-off-by: Andy Adamson <andros@netapp.com>
---
 fs/nfs/nfs4client.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/fs/nfs/nfs4client.c b/fs/nfs/nfs4client.c
index 8d7d08d..d77a1ad 100644
--- a/fs/nfs/nfs4client.c
+++ b/fs/nfs/nfs4client.c
@@ -562,15 +562,15 @@ out:
 /*
  * Returns true if the client IDs match
  */
-static bool nfs4_match_clientids(struct nfs_client *a, struct nfs_client *b)
+static bool nfs4_match_clientids(u64 a, u64 b)
 {
-	if (a->cl_clientid != b->cl_clientid) {
+	if (a != b) {
 		dprintk("NFS: --> %s client ID %llx does not match %llx\n",
-			__func__, a->cl_clientid, b->cl_clientid);
+			__func__, a, b);
 		return false;
 	}
 	dprintk("NFS: --> %s client ID %llx matches %llx\n",
-		__func__, a->cl_clientid, b->cl_clientid);
+		__func__, a, b);
 	return true;
 }
 
@@ -650,7 +650,7 @@ int nfs41_walk_client_list(struct nfs_client *new,
 		if (pos->cl_cons_state != NFS_CS_READY)
 			continue;
 
-		if (!nfs4_match_clientids(pos, new))
+		if (!nfs4_match_clientids(pos->cl_clientid, new->cl_clientid))
 			continue;
 
 		/*
-- 
1.8.3.1

[PATCH Version 10 03/12] NFS refactor nfs4_check_serverowner_major_id

[andros]

From: Andy Adamson <andros@netapp.com>

For session trunking, to compare nfs41_exchange_id_res with
existing nfs_client

Signed-off-by: Andy Adamson <andros@netapp.com>
---
 fs/nfs/nfs4client.c | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/fs/nfs/nfs4client.c b/fs/nfs/nfs4client.c
index d77a1ad..f356d50 100644
--- a/fs/nfs/nfs4client.c
+++ b/fs/nfs/nfs4client.c
@@ -578,17 +578,15 @@ static bool nfs4_match_clientids(u64 a, u64 b)
  * Returns true if the server major ids match
  */
 static bool
-nfs4_check_clientid_trunking(struct nfs_client *a, struct nfs_client *b)
+nfs4_check_serverowner_major_id(struct nfs41_server_owner *o1,
+				struct nfs41_server_owner *o2)
 {
-	struct nfs41_server_owner *o1 = a->cl_serverowner;
-	struct nfs41_server_owner *o2 = b->cl_serverowner;
-
 	if (o1->major_id_sz != o2->major_id_sz)
 		goto out_major_mismatch;
 	if (memcmp(o1->major_id, o2->major_id, o1->major_id_sz) != 0)
 		goto out_major_mismatch;
 
-	dprintk("NFS: --> %s server owners match\n", __func__);
+	dprintk("NFS: --> %s server owner major IDs match\n", __func__);
 	return true;
 
 out_major_mismatch:
@@ -658,7 +656,8 @@ int nfs41_walk_client_list(struct nfs_client *new,
 		 * client id trunking. In either case, we want to fall back
 		 * to using the existing nfs_client.
 		 */
-		if (!nfs4_check_clientid_trunking(pos, new))
+		if (!nfs4_check_serverowner_major_id(pos->cl_serverowner,
+						     new->cl_serverowner))
 			continue;
 
 		/* Unlike NFSv4.0, we know that NFSv4.1 always uses the
-- 
1.8.3.1

[PATCH Version 10 04/12] NFS detect session trunking

[andros]

From: Andy Adamson <andros@netapp.com>

Signed-off-by: Andy Adamson <andros@netapp.com>
---
 fs/nfs/nfs4_fs.h    |  2 ++
 fs/nfs/nfs4client.c | 90 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 92 insertions(+)

diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h
index f230aa6..cf744aa 100644
--- a/fs/nfs/nfs4_fs.h
+++ b/fs/nfs/nfs4_fs.h
@@ -278,6 +278,8 @@ extern int nfs4_proc_get_lease_time(struct nfs_client *clp,
 		struct nfs_fsinfo *fsinfo);
 extern int nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data,
 				  bool sync);
+extern int nfs4_detect_session_trunking(struct nfs_client *clp,
+		struct nfs41_exchange_id_res *res, struct rpc_xprt *xprt);
 
 static inline bool
 is_ds_only_client(struct nfs_client *clp)
diff --git a/fs/nfs/nfs4client.c b/fs/nfs/nfs4client.c
index f356d50..3371395 100644
--- a/fs/nfs/nfs4client.c
+++ b/fs/nfs/nfs4client.c
@@ -595,6 +595,96 @@ out_major_mismatch:
 	return false;
 }
 
+/*
+ * Returns true if server minor ids match
+ */
+static bool
+nfs4_check_serverowner_minor_id(struct nfs41_server_owner *o1,
+				struct nfs41_server_owner *o2)
+{
+	/* Check eir_server_owner so_minor_id */
+	if (o1->minor_id != o2->minor_id)
+		goto out_minor_mismatch;
+
+	dprintk("NFS: --> %s server owner minor IDs match\n", __func__);
+	return true;
+
+out_minor_mismatch:
+	dprintk("NFS: --> %s server owner minor IDs do not match\n", __func__);
+	return false;
+}
+
+/*
+ * Returns true if the server scopes match
+ */
+static bool
+nfs4_check_server_scope(struct nfs41_server_scope *s1,
+			struct nfs41_server_scope *s2)
+{
+	if (s1->server_scope_sz != s2->server_scope_sz)
+		goto out_scope_mismatch;
+	if (memcmp(s1->server_scope, s2->server_scope,
+		   s1->server_scope_sz) != 0)
+		goto out_scope_mismatch;
+
+	dprintk("NFS: --> %s server scopes match\n", __func__);
+	return true;
+
+out_scope_mismatch:
+	dprintk("NFS: --> %s server scopes do not match\n",
+		__func__);
+	return false;
+}
+
+/**
+ * nfs4_detect_session_trunking - Checks for session trunking called
+ * after a successful EXCHANGE_ID testing a multi-addr connection to be
+ * potentially added as a session trunk
+ *
+ * @clp:    original mount nfs_client
+ * @res:    result structure from an exchange_id using the original mount
+ *          nfs_client with a new multi_addr transport
+ *
+ * Returns zero on success, otherwise -EINVAL
+ *
+ * Note: since the exchange_id for the new multi_addr transport uses the
+ * same nfs_client from the original mount, the cl_owner_id is reused,
+ * so eir_clientowner is the same.
+ */
+int nfs4_detect_session_trunking(struct nfs_client *clp,
+				 struct nfs41_exchange_id_res *res,
+				 struct rpc_xprt *xprt)
+{
+	/* Check eir_clientid */
+	if (!nfs4_match_clientids(clp->cl_clientid, res->clientid))
+		goto out_err;
+
+	/* Check eir_server_owner so_major_id */
+	if (!nfs4_check_serverowner_major_id(clp->cl_serverowner,
+					     res->server_owner))
+		goto out_err;
+
+	/* Check eir_server_owner so_minor_id */
+	if (!nfs4_check_serverowner_minor_id(clp->cl_serverowner,
+					     res->server_owner))
+		goto out_err;
+
+	/* Check eir_server_scope */
+	if (!nfs4_check_server_scope(clp->cl_serverscope, res->server_scope))
+		goto out_err;
+
+	pr_info("NFS:  %s: Session trunking succeeded for %s\n",
+		clp->cl_hostname,
+		xprt->address_strings[RPC_DISPLAY_ADDR]);
+
+	return 0;
+out_err:
+	pr_info("NFS:  %s: Session trunking failed for %s\n", clp->cl_hostname,
+		xprt->address_strings[RPC_DISPLAY_ADDR]);
+
+	return -EINVAL;
+}
+
 /**
  * nfs41_walk_client_list - Find nfs_client that matches a client/server owner
  *
-- 
1.8.3.1

[PATCH Version 10 05/12] SUNRPC remove rpc_task_release_client from rpc_task_set_client

[andros]

From: Andy Adamson <andros@netapp.com>

rpc_task_set_client is only called from rpc_run_task after
rpc_new_task and rpc_task_release_client is not needed as the
task is new.

When called from rpc_new_task, rpc_task_set_client also removed the
assigned rpc_xprt which is not desired.

Signed-off-by: Andy Adamson <andros@netapp.com>
---
 net/sunrpc/clnt.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index 6747265..4abc631 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -987,7 +987,6 @@ void rpc_task_set_client(struct rpc_task *task, struct rpc_clnt *clnt)
 {
 
 	if (clnt != NULL) {
-		rpc_task_release_client(task);
 		if (task->tk_xprt == NULL)
 			task->tk_xprt = xprt_iter_get_next(&clnt->cl_xpi);
 		task->tk_client = clnt;
-- 
1.8.3.1

[PATCH Version 10 06/12] SUNRPC rpc_clnt_xprt_switch_put

[andros]

From: Andy Adamson <andros@netapp.com>

Give the NFS layer access to the xprt_switch_put function

Signed-off-by: Andy Adamson <andros@netapp.com>
---
 include/linux/sunrpc/clnt.h | 2 ++
 net/sunrpc/clnt.c           | 6 ++++++
 2 files changed, 8 insertions(+)

diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h
index 5c02b06..c12f86b 100644
--- a/include/linux/sunrpc/clnt.h
+++ b/include/linux/sunrpc/clnt.h
@@ -199,5 +199,7 @@ void		rpc_cap_max_reconnect_timeout(struct rpc_clnt *clnt,
 			unsigned long timeo);
 
 const char *rpc_proc_name(const struct rpc_task *task);
+
+void rpc_clnt_xprt_switch_put(struct rpc_clnt *);
 #endif /* __KERNEL__ */
 #endif /* _LINUX_SUNRPC_CLNT_H */
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index 4abc631..138f777 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -2695,6 +2695,12 @@ rpc_cap_max_reconnect_timeout(struct rpc_clnt *clnt, unsigned long timeo)
 }
 EXPORT_SYMBOL_GPL(rpc_cap_max_reconnect_timeout);
 
+void rpc_clnt_xprt_switch_put(struct rpc_clnt *clnt)
+{
+	xprt_switch_put(rcu_dereference(clnt->cl_xpi.xpi_xpswitch));
+}
+EXPORT_SYMBOL_GPL(rpc_clnt_xprt_switch_put);
+
 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
 static void rpc_show_header(void)
 {
-- 
1.8.3.1

[PATCH Version 10 07/12] SUNRPC rpc_clnt_xprt_switch_add_xprt

[andros]

From: Andy Adamson <andros@netapp.com>

Give the NFS layer access to the rpc_xprt_switch_add_xprt function

Signed-off-by: Andy Adamson <andros@netapp.com>
---
 include/linux/sunrpc/clnt.h | 1 +
 net/sunrpc/clnt.c           | 7 +++++++
 2 files changed, 8 insertions(+)

diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h
index c12f86b..b069d6e 100644
--- a/include/linux/sunrpc/clnt.h
+++ b/include/linux/sunrpc/clnt.h
@@ -201,5 +201,6 @@ void		rpc_cap_max_reconnect_timeout(struct rpc_clnt *clnt,
 const char *rpc_proc_name(const struct rpc_task *task);
 
 void rpc_clnt_xprt_switch_put(struct rpc_clnt *);
+void rpc_clnt_xprt_switch_add_xprt(struct rpc_clnt *, struct rpc_xprt *);
 #endif /* __KERNEL__ */
 #endif /* _LINUX_SUNRPC_CLNT_H */
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index 138f777..6db0d00 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -2701,6 +2701,13 @@ void rpc_clnt_xprt_switch_put(struct rpc_clnt *clnt)
 }
 EXPORT_SYMBOL_GPL(rpc_clnt_xprt_switch_put);
 
+void rpc_clnt_xprt_switch_add_xprt(struct rpc_clnt *clnt, struct rpc_xprt *xprt)
+{
+	rpc_xprt_switch_add_xprt(rcu_dereference(clnt->cl_xpi.xpi_xpswitch),
+				 xprt);
+}
+EXPORT_SYMBOL_GPL(rpc_clnt_xprt_switch_add_xprt);
+
 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
 static void rpc_show_header(void)
 {
-- 
1.8.3.1

[PATCH Version 10 08/12] SUNRPC search xprt switch for sockaddr

[andros]

From: Andy Adamson <andros@netapp.com>

Signed-off-by: Andy Adamson <andros@netapp.com>
---
 include/linux/sunrpc/clnt.h          |  2 ++
 include/linux/sunrpc/xprtmultipath.h |  2 ++
 net/sunrpc/clnt.c                    | 15 +++++++++++++++
 net/sunrpc/xprtmultipath.c           | 24 +++++++++++++++++++++++-
 4 files changed, 42 insertions(+), 1 deletion(-)

diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h
index b069d6e..35cc539 100644
--- a/include/linux/sunrpc/clnt.h
+++ b/include/linux/sunrpc/clnt.h
@@ -202,5 +202,7 @@ const char *rpc_proc_name(const struct rpc_task *task);
 
 void rpc_clnt_xprt_switch_put(struct rpc_clnt *);
 void rpc_clnt_xprt_switch_add_xprt(struct rpc_clnt *, struct rpc_xprt *);
+bool rpc_clnt_xprt_switch_has_addr(struct rpc_clnt *clnt,
+			const struct sockaddr *sap);
 #endif /* __KERNEL__ */
 #endif /* _LINUX_SUNRPC_CLNT_H */
diff --git a/include/linux/sunrpc/xprtmultipath.h b/include/linux/sunrpc/xprtmultipath.h
index 5a9acff..507418c 100644
--- a/include/linux/sunrpc/xprtmultipath.h
+++ b/include/linux/sunrpc/xprtmultipath.h
@@ -66,4 +66,6 @@ extern struct rpc_xprt *xprt_iter_xprt(struct rpc_xprt_iter *xpi);
 extern struct rpc_xprt *xprt_iter_get_xprt(struct rpc_xprt_iter *xpi);
 extern struct rpc_xprt *xprt_iter_get_next(struct rpc_xprt_iter *xpi);
 
+extern bool rpc_xprt_switch_has_addr(struct rpc_xprt_switch *xps,
+		const struct sockaddr *sap);
 #endif
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index 6db0d00..282a5e7 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -2708,6 +2708,21 @@ void rpc_clnt_xprt_switch_add_xprt(struct rpc_clnt *clnt, struct rpc_xprt *xprt)
 }
 EXPORT_SYMBOL_GPL(rpc_clnt_xprt_switch_add_xprt);
 
+bool rpc_clnt_xprt_switch_has_addr(struct rpc_clnt *clnt,
+				   const struct sockaddr *sap)
+{
+	struct rpc_xprt_switch *xps;
+	bool ret;
+
+	xps = rcu_dereference(clnt->cl_xpi.xpi_xpswitch);
+
+	rcu_read_lock();
+	ret = rpc_xprt_switch_has_addr(xps, sap);
+	rcu_read_unlock();
+	return ret;
+}
+EXPORT_SYMBOL_GPL(rpc_clnt_xprt_switch_has_addr);
+
 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
 static void rpc_show_header(void)
 {
diff --git a/net/sunrpc/xprtmultipath.c b/net/sunrpc/xprtmultipath.c
index 66c9d63..ae92a9e 100644
--- a/net/sunrpc/xprtmultipath.c
+++ b/net/sunrpc/xprtmultipath.c
@@ -15,6 +15,7 @@
 #include <asm/cmpxchg.h>
 #include <linux/spinlock.h>
 #include <linux/sunrpc/xprt.h>
+#include <linux/sunrpc/addr.h>
 #include <linux/sunrpc/xprtmultipath.h>
 
 typedef struct rpc_xprt *(*xprt_switch_find_xprt_t)(struct list_head *head,
@@ -49,7 +50,8 @@ void rpc_xprt_switch_add_xprt(struct rpc_xprt_switch *xps,
 	if (xprt == NULL)
 		return;
 	spin_lock(&xps->xps_lock);
-	if (xps->xps_net == xprt->xprt_net || xps->xps_net == NULL)
+	if ((xps->xps_net == xprt->xprt_net || xps->xps_net == NULL) &&
+	    !rpc_xprt_switch_has_addr(xps, (struct sockaddr *)&xprt->addr))
 		xprt_switch_add_xprt_locked(xps, xprt);
 	spin_unlock(&xps->xps_lock);
 }
@@ -232,6 +234,26 @@ struct rpc_xprt *xprt_iter_current_entry(struct rpc_xprt_iter *xpi)
 	return xprt_switch_find_current_entry(head, xpi->xpi_cursor);
 }
 
+bool rpc_xprt_switch_has_addr(struct rpc_xprt_switch *xps,
+			      const struct sockaddr *sap)
+{
+	struct list_head *head;
+	struct rpc_xprt *pos;
+
+	if (xps == NULL || sap == NULL)
+		return false;
+
+	head = &xps->xps_xprt_list;
+	list_for_each_entry_rcu(pos, head, xprt_switch) {
+		if (rpc_cmp_addr_port(sap, (struct sockaddr *)&pos->addr)) {
+			pr_info("RPC:   addr %s already in xprt switch\n",
+				pos->address_strings[RPC_DISPLAY_ADDR]);
+			return true;
+		}
+	}
+	return false;
+}
+
 static
 struct rpc_xprt *xprt_switch_find_next_entry(struct list_head *head,
 		const struct rpc_xprt *cur)
-- 
1.8.3.1

[PATCH Version 10 09/12] SUNRPC: rpc_clnt_add_xprt setup function for NFS layer

[andros]

From: Andy Adamson <andros@netapp.com>

Use a setup function to call into the NFS layer to test an rpc_xprt
for session trunking so as to not leak the rpc_xprt_switch into
the nfs layer.

Search for the address in the rpc_xprt_switch first so as not to
put an unnecessary EXCHANGE_ID on the wire.

Signed-off-by: Andy Adamson <andros@netapp.com>
---
 include/linux/sunrpc/clnt.h | 12 +++++++++
 net/sunrpc/clnt.c           | 64 +++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 76 insertions(+)

diff --git a/include/linux/sunrpc/clnt.h b/include/linux/sunrpc/clnt.h
index 35cc539..85cc819 100644
--- a/include/linux/sunrpc/clnt.h
+++ b/include/linux/sunrpc/clnt.h
@@ -125,6 +125,13 @@ struct rpc_create_args {
 	struct svc_xprt		*bc_xprt;	/* NFSv4.1 backchannel */
 };
 
+struct rpc_add_xprt_test {
+	int (*add_xprt_test)(struct rpc_clnt *,
+		struct rpc_xprt *,
+		void *calldata);
+	void *data;
+};
+
 /* Values for "flags" field */
 #define RPC_CLNT_CREATE_HARDRTRY	(1UL << 0)
 #define RPC_CLNT_CREATE_AUTOBIND	(1UL << 2)
@@ -198,6 +205,11 @@ int		rpc_clnt_add_xprt(struct rpc_clnt *, struct xprt_create *,
 void		rpc_cap_max_reconnect_timeout(struct rpc_clnt *clnt,
 			unsigned long timeo);
 
+int		rpc_clnt_setup_test_and_add_xprt(struct rpc_clnt *,
+			struct rpc_xprt_switch *,
+			struct rpc_xprt *,
+			void *);
+
 const char *rpc_proc_name(const struct rpc_task *task);
 
 void rpc_clnt_xprt_switch_put(struct rpc_clnt *);
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index 282a5e7..913a283 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -2614,6 +2614,70 @@ int rpc_clnt_test_and_add_xprt(struct rpc_clnt *clnt,
 EXPORT_SYMBOL_GPL(rpc_clnt_test_and_add_xprt);
 
 /**
+ * rpc_clnt_setup_test_and_add_xprt()
+ *
+ * This is an rpc_clnt_add_xprt setup() function which returns 1 so:
+ *   1) caller of the test function must dereference the rpc_xprt_switch
+ *   and the rpc_xprt.
+ *   2) test function must call rpc_xprt_switch_add_xprt, usually in
+ *   the rpc_call_done routine.
+ *
+ * Upon success (return of 1), the test function adds the new
+ * transport to the rpc_clnt xprt switch
+ *
+ * @clnt: struct rpc_clnt to get the new transport
+ * @xps:  the rpc_xprt_switch to hold the new transport
+ * @xprt: the rpc_xprt to test
+ * @data: a struct rpc_add_xprt_test pointer that holds the test function
+ *        and test function call data
+ */
+int rpc_clnt_setup_test_and_add_xprt(struct rpc_clnt *clnt,
+				     struct rpc_xprt_switch *xps,
+				     struct rpc_xprt *xprt,
+				     void *data)
+{
+	struct rpc_cred *cred;
+	struct rpc_task *task;
+	struct rpc_add_xprt_test *xtest = (struct rpc_add_xprt_test *)data;
+	int status = -EADDRINUSE;
+
+	xprt = xprt_get(xprt);
+	xprt_switch_get(xps);
+
+	if (rpc_xprt_switch_has_addr(xps, (struct sockaddr *)&xprt->addr))
+		goto out_err;
+
+	/* Test the connection */
+	cred = authnull_ops.lookup_cred(NULL, NULL, 0);
+	task = rpc_call_null_helper(clnt, xprt, cred,
+				    RPC_TASK_SOFT | RPC_TASK_SOFTCONN,
+				    NULL, NULL);
+	put_rpccred(cred);
+	if (IS_ERR(task)) {
+		status = PTR_ERR(task);
+		goto out_err;
+	}
+	status = task->tk_status;
+	rpc_put_task(task);
+
+	if (status < 0)
+		goto out_err;
+
+	/* rpc_xprt_switch and rpc_xprt are deferrenced by add_xprt_test() */
+	xtest->add_xprt_test(clnt, xprt, xtest->data);
+
+	/* so that rpc_clnt_add_xprt does not call rpc_xprt_switch_add_xprt */
+	return 1;
+out_err:
+	xprt_put(xprt);
+	xprt_switch_put(xps);
+	pr_info("RPC:   rpc_clnt_test_xprt failed: %d addr %s not added\n",
+		status, xprt->address_strings[RPC_DISPLAY_ADDR]);
+	return status;
+}
+EXPORT_SYMBOL_GPL(rpc_clnt_setup_test_and_add_xprt);
+
+/**
  * rpc_clnt_add_xprt - Add a new transport to a rpc_clnt
  * @clnt: pointer to struct rpc_clnt
  * @xprtargs: pointer to struct xprt_create
-- 
1.8.3.1

[PATCH Version 10 10/12] NFS add xprt switch addrs test to match client

[andros]

From: Andy Adamson <andros@netapp.com>

Signed-off-by: Andy Adamson <andros@netapp.com>
---
 fs/nfs/client.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/fs/nfs/client.c b/fs/nfs/client.c
index ef9e70b..b4a8eb0c 100644
--- a/fs/nfs/client.c
+++ b/fs/nfs/client.c
@@ -313,7 +313,10 @@ static struct nfs_client *nfs_match_client(const struct nfs_client_initdata *dat
 			continue;
 		/* Match the full socket address */
 		if (!rpc_cmp_addr_port(sap, clap))
-			continue;
+			/* Match all xprt_switch full socket addresses */
+			if (!rpc_clnt_xprt_switch_has_addr(clp->cl_rpcclient,
+							   sap))
+				continue;
 
 		atomic_inc(&clp->cl_count);
 		return clp;
-- 
1.8.3.1

[PATCH Version 10 11/12] NFS test session trunking with exchange id

[andros]

From: Andy Adamson <andros@netapp.com>

Use an async exchange id call to test for session trunking

To conform with RFC 5661 section 18.35.4, the Non-Update on
Existing Clientid case, save the exchange id verifier in
cl_confirm and use it for the session trunking exhange id test.

Signed-off-by: Andy Adamson <andros@netapp.com>
---
 fs/nfs/nfs4client.c | 10 +++++++---
 fs/nfs/nfs4proc.c   | 45 +++++++++++++++++++++++++++++++++++++--------
 2 files changed, 44 insertions(+), 11 deletions(-)

diff --git a/fs/nfs/nfs4client.c b/fs/nfs/nfs4client.c
index 3371395..5a45d07 100644
--- a/fs/nfs/nfs4client.c
+++ b/fs/nfs/nfs4client.c
@@ -637,9 +637,10 @@ out_scope_mismatch:
 }
 
 /**
- * nfs4_detect_session_trunking - Checks for session trunking called
- * after a successful EXCHANGE_ID testing a multi-addr connection to be
- * potentially added as a session trunk
+ * nfs4_detect_session_trunking - Checks for session trunking.
+ *
+ * Called after a successful EXCHANGE_ID on a multi-addr connection.
+ * Upon success, add the transport.
  *
  * @clp:    original mount nfs_client
  * @res:    result structure from an exchange_id using the original mount
@@ -673,6 +674,9 @@ int nfs4_detect_session_trunking(struct nfs_client *clp,
 	if (!nfs4_check_server_scope(clp->cl_serverscope, res->server_scope))
 		goto out_err;
 
+	/* Session trunking passed, add the xprt */
+	rpc_clnt_xprt_switch_add_xprt(clp->cl_rpcclient, xprt);
+
 	pr_info("NFS:  %s: Session trunking succeeded for %s\n",
 		clp->cl_hostname,
 		xprt->address_strings[RPC_DISPLAY_ADDR]);
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 98bd30e..0cea0e4 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -7051,6 +7051,7 @@ static int nfs4_sp4_select_mode(struct nfs_client *clp,
 struct nfs41_exchange_id_data {
 	struct nfs41_exchange_id_res res;
 	struct nfs41_exchange_id_args args;
+	struct rpc_xprt *xprt;
 	int rpc_status;
 };
 
@@ -7065,6 +7066,13 @@ static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
 
 	if (status == 0)
 		status = nfs4_check_cl_exchange_flags(cdata->res.flags);
+
+	if (cdata->xprt && status == 0) {
+		status = nfs4_detect_session_trunking(clp, &cdata->res,
+						      cdata->xprt);
+		goto out;
+	}
+
 	if (status  == 0)
 		status = nfs4_sp4_select_mode(clp, &cdata->res.state_protect);
 
@@ -7101,8 +7109,13 @@ static void nfs4_exchange_id_done(struct rpc_task *task, void *data)
 			clp->cl_serverscope = cdata->res.server_scope;
 			cdata->res.server_scope = NULL;
 		}
+		/* Save the EXCHANGE_ID verifier session trunk tests */
+		memcpy(clp->cl_confirm.data, cdata->args.verifier->data,
+		       sizeof(clp->cl_confirm.data));
 	}
+out:
 	cdata->rpc_status = status;
+	return;
 }
 
 static void nfs4_exchange_id_release(void *data)
@@ -7111,6 +7124,10 @@ static void nfs4_exchange_id_release(void *data)
 					(struct nfs41_exchange_id_data *)data;
 
 	nfs_put_client(cdata->args.client);
+	if (cdata->xprt) {
+		xprt_put(cdata->xprt);
+		rpc_clnt_xprt_switch_put(cdata->args.client->cl_rpcclient);
+	}
 	kfree(cdata->res.impl_id);
 	kfree(cdata->res.server_scope);
 	kfree(cdata->res.server_owner);
@@ -7128,7 +7145,7 @@ static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
  * Wrapper for EXCHANGE_ID operation.
  */
 static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
-	u32 sp4_how)
+			u32 sp4_how, struct rpc_xprt *xprt)
 {
 	nfs4_verifier verifier;
 	struct rpc_message msg = {
@@ -7153,7 +7170,8 @@ static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
 	if (!calldata)
 		goto out;
 
-	nfs4_init_boot_verifier(clp, &verifier);
+	if (!xprt)
+		nfs4_init_boot_verifier(clp, &verifier);
 
 	status = nfs4_init_uniform_client_string(clp);
 	if (status)
@@ -7193,8 +7211,15 @@ static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
 		status = -EINVAL;
 		goto out_impl_id;
 	}
-
-	calldata->args.verifier = &verifier;
+	if (xprt) {
+		calldata->xprt = xprt;
+		task_setup_data.rpc_xprt = xprt;
+		task_setup_data.flags =
+				RPC_TASK_SOFT|RPC_TASK_SOFTCONN|RPC_TASK_ASYNC;
+		calldata->args.verifier = &clp->cl_confirm;
+	} else {
+		calldata->args.verifier = &verifier;
+	}
 	calldata->args.client = clp;
 #ifdef CONFIG_NFS_V4_1_MIGRATION
 	calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
@@ -7214,9 +7239,13 @@ static int _nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred,
 		goto out_impl_id;
 	}
 
-	status = rpc_wait_for_completion_task(task);
-	if (!status)
+	if (!xprt) {
+		status = rpc_wait_for_completion_task(task);
+		if (!status)
+			status = calldata->rpc_status;
+	} else	/* session trunking test */
 		status = calldata->rpc_status;
+
 	rpc_put_task(task);
 out:
 	if (clp->cl_implid != NULL)
@@ -7259,13 +7288,13 @@ int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
 	/* try SP4_MACH_CRED if krb5i/p	*/
 	if (authflavor == RPC_AUTH_GSS_KRB5I ||
 	    authflavor == RPC_AUTH_GSS_KRB5P) {
-		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
+		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED, NULL);
 		if (!status)
 			return 0;
 	}
 
 	/* try SP4_NONE */
-	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
+	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
 }
 
 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
-- 
1.8.3.1

[PATCH Version 10 12/12] NFS pnfs data server multipath session trunking

[andros]

From: Andy Adamson <andros@netapp.com>

Try all multipath addresses for a data server. The first address that
successfully connects and creates a session is the DS mount address.
All subsequent addresses are tested for session trunking and
added as aliases.

Signed-off-by: Andy Adamson <andros@netapp.com>
---
 fs/nfs/internal.h |  3 +++
 fs/nfs/nfs4_fs.h  |  6 ++++++
 fs/nfs/nfs4proc.c | 33 +++++++++++++++++++++++++++++++
 fs/nfs/pnfs_nfs.c | 58 +++++++++++++++++++++++++++++++++++++++++--------------
 4 files changed, 86 insertions(+), 14 deletions(-)

diff --git a/fs/nfs/internal.h b/fs/nfs/internal.h
index 7ce5e02..9ced091 100644
--- a/fs/nfs/internal.h
+++ b/fs/nfs/internal.h
@@ -569,6 +569,9 @@ extern int nfs40_walk_client_list(struct nfs_client *clp,
 extern int nfs41_walk_client_list(struct nfs_client *clp,
 				struct nfs_client **result,
 				struct rpc_cred *cred);
+extern int nfs4_test_session_trunk(struct rpc_clnt *,
+				struct rpc_xprt *,
+				void *);
 
 static inline struct inode *nfs_igrab_and_active(struct inode *inode)
 {
diff --git a/fs/nfs/nfs4_fs.h b/fs/nfs/nfs4_fs.h
index cf744aa..8041134 100644
--- a/fs/nfs/nfs4_fs.h
+++ b/fs/nfs/nfs4_fs.h
@@ -59,6 +59,7 @@ struct nfs4_minor_version_ops {
 			struct nfs4_lock_state *);
 	struct nfs_seqid *
 		(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
+	int	(*session_trunk)(struct rpc_clnt *, struct rpc_xprt *, void *);
 	const struct rpc_call_ops *call_sync_ops;
 	const struct nfs4_state_recovery_ops *reboot_recovery_ops;
 	const struct nfs4_state_recovery_ops *nograce_recovery_ops;
@@ -203,6 +204,11 @@ struct nfs4_state_recovery_ops {
 		struct rpc_cred *);
 };
 
+struct nfs4_add_xprt_data {
+	struct nfs_client	*clp;
+	struct rpc_cred		*cred;
+};
+
 struct nfs4_state_maintenance_ops {
 	int (*sched_state_renewal)(struct nfs_client *, struct rpc_cred *, unsigned);
 	struct rpc_cred * (*get_state_renewal_cred_locked)(struct nfs_client *);
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 0cea0e4..ad35b5e 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -7297,6 +7297,37 @@ int nfs4_proc_exchange_id(struct nfs_client *clp, struct rpc_cred *cred)
 	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE, NULL);
 }
 
+/**
+ * nfs4_test_session_trunk
+ *
+ * This is an add_xprt_test() test function called from
+ * rpc_clnt_setup_test_and_add_xprt.
+ *
+ * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
+ * and is dereferrenced in nfs4_exchange_id_release
+ *
+ * Upon success, add the new transport to the rpc_clnt
+ *
+ * @clnt: struct rpc_clnt to get new transport
+ * @xprt: the rpc_xprt to test
+ * @data: call data for _nfs4_proc_exchange_id.
+ */
+int nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
+			    void *data)
+{
+	struct nfs4_add_xprt_data *adata = (struct nfs4_add_xprt_data *)data;
+	u32 sp4_how;
+
+	dprintk("--> %s try %s\n", __func__,
+		xprt->address_strings[RPC_DISPLAY_ADDR]);
+
+	sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
+
+	/* Test connection for session trunking. Async exchange_id call */
+	return  _nfs4_proc_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
+}
+EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
+
 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
 		struct rpc_cred *cred)
 {
@@ -8874,6 +8905,7 @@ static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
 	.find_root_sec = nfs41_find_root_sec,
 	.free_lock_state = nfs41_free_lock_state,
 	.alloc_seqid = nfs_alloc_no_seqid,
+	.session_trunk = nfs4_test_session_trunk,
 	.call_sync_ops = &nfs41_call_sync_ops,
 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
@@ -8903,6 +8935,7 @@ static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
 	.free_lock_state = nfs41_free_lock_state,
 	.call_sync_ops = &nfs41_call_sync_ops,
 	.alloc_seqid = nfs_alloc_no_seqid,
+	.session_trunk = nfs4_test_session_trunk,
 	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
 	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
 	.state_renewal_ops = &nfs41_state_renewal_ops,
diff --git a/fs/nfs/pnfs_nfs.c b/fs/nfs/pnfs_nfs.c
index f3468b5..53b4705 100644
--- a/fs/nfs/pnfs_nfs.c
+++ b/fs/nfs/pnfs_nfs.c
@@ -690,13 +690,50 @@ static int _nfs4_pnfs_v4_ds_connect(struct nfs_server *mds_srv,
 		dprintk("%s: DS %s: trying address %s\n",
 			__func__, ds->ds_remotestr, da->da_remotestr);
 
-		clp = nfs4_set_ds_client(mds_srv,
-					(struct sockaddr *)&da->da_addr,
-					da->da_addrlen, IPPROTO_TCP,
-					timeo, retrans, minor_version,
-					au_flavor);
-		if (!IS_ERR(clp))
-			break;
+		if (!IS_ERR(clp) && clp->cl_mvops->session_trunk) {
+			struct xprt_create xprt_args = {
+				.ident = XPRT_TRANSPORT_TCP,
+				.net = clp->cl_net,
+				.dstaddr = (struct sockaddr *)&da->da_addr,
+				.addrlen = da->da_addrlen,
+				.servername = clp->cl_hostname,
+			};
+			struct nfs4_add_xprt_data xprtdata = {
+				.clp = clp,
+				.cred = nfs4_get_clid_cred(clp),
+			};
+			struct rpc_add_xprt_test rpcdata = {
+				.add_xprt_test = clp->cl_mvops->session_trunk,
+				.data = &xprtdata,
+			};
+
+			/**
+			* Test this address for session trunking and
+			* add as an alias
+			*/
+			rpc_clnt_add_xprt(clp->cl_rpcclient, &xprt_args,
+					  rpc_clnt_setup_test_and_add_xprt,
+					  &rpcdata);
+			if (xprtdata.cred)
+				put_rpccred(xprtdata.cred);
+		} else {
+			clp = nfs4_set_ds_client(mds_srv,
+						(struct sockaddr *)&da->da_addr,
+						da->da_addrlen, IPPROTO_TCP,
+						timeo, retrans, minor_version,
+						au_flavor);
+			if (IS_ERR(clp))
+				continue;
+
+			status = nfs4_init_ds_session(clp,
+					mds_srv->nfs_client->cl_lease_time);
+			if (status) {
+				nfs_put_client(clp);
+				clp = ERR_PTR(-EIO);
+				continue;
+			}
+
+		}
 	}
 
 	if (IS_ERR(clp)) {
@@ -704,18 +741,11 @@ static int _nfs4_pnfs_v4_ds_connect(struct nfs_server *mds_srv,
 		goto out;
 	}
 
-	status = nfs4_init_ds_session(clp, mds_srv->nfs_client->cl_lease_time);
-	if (status)
-		goto out_put;
-
 	smp_wmb();
 	ds->ds_clp = clp;
 	dprintk("%s [new] addr: %s\n", __func__, ds->ds_remotestr);
 out:
 	return status;
-out_put:
-	nfs_put_client(clp);
-	goto out;
 }
 
 /*
-- 
1.8.3.1