[PATCH 4/5] NFSv4/pNFS: Don't call _nfs4_pnfs_v3_ds_connect multiple times

[trondmy@xxxxxxxxxx]

From: Trond Myklebust <trond.myklebust@xxxxxxxxxxxxxxx>

After we grab the lock in nfs4_pnfs_ds_connect(), there is no check for
whether or not ds->ds_clp has already been initialised, so we can end up
adding the same transports multiple times.

Fixes: fc821d59209d ("pnfs/NFSv4.1: Add multipath capabilities to pNFS flexfiles servers over NFSv3")
Signed-off-by: Trond Myklebust <trond.myklebust@xxxxxxxxxxxxxxx>
---
 fs/nfs/pnfs_nfs.c | 52 +++++++++++++++++++++++------------------------
 1 file changed, 26 insertions(+), 26 deletions(-)

diff --git a/fs/nfs/pnfs_nfs.c b/fs/nfs/pnfs_nfs.c
index 49d3389bd813..1c2c0d08614e 100644
--- a/fs/nfs/pnfs_nfs.c
+++ b/fs/nfs/pnfs_nfs.c
@@ -805,19 +805,16 @@ nfs4_pnfs_ds_add(struct list_head *dsaddrs, gfp_t gfp_flags)
 }
 EXPORT_SYMBOL_GPL(nfs4_pnfs_ds_add);
 
-static void nfs4_wait_ds_connect(struct nfs4_pnfs_ds *ds)
+static int nfs4_wait_ds_connect(struct nfs4_pnfs_ds *ds)
 {
 	might_sleep();
-	wait_on_bit(&ds->ds_state, NFS4DS_CONNECTING,
-			TASK_KILLABLE);
+	return wait_on_bit(&ds->ds_state, NFS4DS_CONNECTING, TASK_KILLABLE);
 }
 
 static void nfs4_clear_ds_conn_bit(struct nfs4_pnfs_ds *ds)
 {
 	smp_mb__before_atomic();
-	clear_bit(NFS4DS_CONNECTING, &ds->ds_state);
-	smp_mb__after_atomic();
-	wake_up_bit(&ds->ds_state, NFS4DS_CONNECTING);
+	clear_and_wake_up_bit(NFS4DS_CONNECTING, &ds->ds_state);
 }
 
 static struct nfs_client *(*get_v3_ds_connect)(
@@ -993,30 +990,33 @@ int nfs4_pnfs_ds_connect(struct nfs_server *mds_srv, struct nfs4_pnfs_ds *ds,
 {
 	int err;
 
-again:
-	err = 0;
-	if (test_and_set_bit(NFS4DS_CONNECTING, &ds->ds_state) == 0) {
-		if (version == 3) {
-			err = _nfs4_pnfs_v3_ds_connect(mds_srv, ds, timeo,
-						       retrans);
-		} else if (version == 4) {
-			err = _nfs4_pnfs_v4_ds_connect(mds_srv, ds, timeo,
-						       retrans, minor_version);
-		} else {
-			dprintk("%s: unsupported DS version %d\n", __func__,
-				version);
-			err = -EPROTONOSUPPORT;
-		}
+	do {
+		err = nfs4_wait_ds_connect(ds);
+		if (err || ds->ds_clp)
+			goto out;
+		if (nfs4_test_deviceid_unavailable(devid))
+			return -ENODEV;
+	} while (test_and_set_bit(NFS4DS_CONNECTING, &ds->ds_state) != 0);
 
-		nfs4_clear_ds_conn_bit(ds);
-	} else {
-		nfs4_wait_ds_connect(ds);
+	if (ds->ds_clp)
+		goto connect_done;
 
-		/* what was waited on didn't connect AND didn't mark unavail */
-		if (!ds->ds_clp && !nfs4_test_deviceid_unavailable(devid))
-			goto again;
+	switch (version) {
+	case 3:
+		err = _nfs4_pnfs_v3_ds_connect(mds_srv, ds, timeo, retrans);
+		break;
+	case 4:
+		err = _nfs4_pnfs_v4_ds_connect(mds_srv, ds, timeo, retrans,
+					       minor_version);
+		break;
+	default:
+		dprintk("%s: unsupported DS version %d\n", __func__, version);
+		err = -EPROTONOSUPPORT;
 	}
 
+connect_done:
+	nfs4_clear_ds_conn_bit(ds);
+out:
 	/*
 	 * At this point the ds->ds_clp should be ready, but it might have
 	 * hit an error.
-- 
2.31.1