nfscl: Fix NFSv4.1/4.2 pnfs mounts using nconnect

When a mount with the "pnfs" and "nconnect" options specified
does an I/O operation, it erroneously uses a TCP connection
to the MDS when it is meant to be a DS operation and, as such,
needs to use a TCP connection to the DS.  This patch fixes this.

When the "pnfs" and "nconnect" options are specified for a
NFSv4.1/4.2 mount, there probably should be N connections
established to each DS for I/O RPCs.  This is a fair amount
of work and may be done in a future commit.

This problem was found during a recent IETF NFSv4 working
group testing event.

MFC after:	2 weeks

1 files changed, 10 insertions, 1 deletions

diff --git a/sys/fs/nfs/nfs_commonkrpc.c b/sys/fs/nfs/nfs_commonkrpc.c
index 358d77fe5b30..c1a5fab2a358 100644
--- a/sys/fs/nfs/nfs_commonkrpc.c
+++ b/sys/fs/nfs/nfs_commonkrpc.c
@@ -639,8 +639,17 @@ newnfs_request(struct nfsrv_descript *nd, struct nfsmount *nmp,
 	if (nrp->nr_client == NULL)
 		newnfs_connect(nmp, nrp, cred, td, 0, false, &nrp->nr_client);
 
+	/*
+	 * If the "nconnect" mount option was specified and this RPC is
+	 * one that can have a large RPC message and is being done through
+	 * the NFS/MDS server, use an additional connection. (When the RPC is
+	 * being done through the server/MDS, nrp == &nmp->nm_sockreq.)
+	 * The "nconnect" mount option normally has minimal effect when the
+	 * "pnfs" mount option is specified, since only Readdir RPCs are
+	 * normally done through the NFS/MDS server.
+	 */
 	nextconn_set = false;
-	if (nmp != NULL && nmp->nm_aconnect > 0 &&
+	if (nmp != NULL && nmp->nm_aconnect > 0 && nrp == &nmp->nm_sockreq &&
 	    (nd->nd_procnum == NFSPROC_READ ||
 	     nd->nd_procnum == NFSPROC_READDIR ||
 	     nd->nd_procnum == NFSPROC_READDIRPLUS ||