MFC: r317345

Make the NFSv4 client to use a write open for reading if allowed by the server.

An NFSv4 server has the option of allowing a Read to be done using a Write
Open. If this is not allowed, the server will return NFSERR_OPENMODE.
This patch attempts the read with a write open and then disables this
if the server replies NFSERR_OPENMODE.
This change will avoid some uses of the special stateids. This will be
useful for pNFS/DS Reads, since they cannot use special stateids.
It will also be useful for any NFSv4 server that does not support reading
via the special stateids. It has been tested against both types of NFSv4 server.

3 files changed, 42 insertions, 10 deletions

diff --git a/sys/fs/nfs/nfsport.h b/sys/fs/nfs/nfsport.h
index 8c43f3f..8bd0f9e 100644
--- a/sys/fs/nfs/nfsport.h
+++ b/sys/fs/nfs/nfsport.h
@@ -899,6 +899,7 @@ int newnfs_realign(struct mbuf **, int);
  */
 #define	NFSSTA_HASWRITEVERF	0x00040000  /* Has write verifier */
 #define	NFSSTA_GOTFSINFO	0x00100000  /* Got the fsinfo */
+#define	NFSSTA_OPENMODE		0x00200000  /* Must use correct open mode */
 #define	NFSSTA_NOLAYOUTCOMMIT	0x04000000  /* Don't do LayoutCommit */
 #define	NFSSTA_SESSPERSIST	0x08000000  /* Has a persistent session */
 #define	NFSSTA_TIMEO		0x10000000  /* Experiencing a timeout */
@@ -929,6 +930,7 @@ int newnfs_realign(struct mbuf **, int);
 #define	NFSHASNOLAYOUTCOMMIT(n)	((n)->nm_state & NFSSTA_NOLAYOUTCOMMIT)
 #define	NFSHASSESSPERSIST(n)	((n)->nm_state & NFSSTA_SESSPERSIST)
 #define	NFSHASPNFS(n)		((n)->nm_state & NFSSTA_PNFS)
+#define	NFSHASOPENMODE(n)	((n)->nm_state & NFSSTA_OPENMODE)
 #define	NFSHASONEOPENOWN(n)	(((n)->nm_flag & NFSMNT_ONEOPENOWN) != 0 &&	\
 				    (n)->nm_minorvers > 0)
 
diff --git a/sys/fs/nfsclient/nfs_clrpcops.c b/sys/fs/nfsclient/nfs_clrpcops.c
index 32c6188..38d2f73 100644
--- a/sys/fs/nfsclient/nfs_clrpcops.c
+++ b/sys/fs/nfsclient/nfs_clrpcops.c
@@ -1134,6 +1134,11 @@ nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
 		else
 			error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
 			    stuff);
+		if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
+			NFSLOCKMNT(nmp);
+			nmp->nm_state |= NFSSTA_OPENMODE;
+			NFSUNLOCKMNT(nmp);
+		}
 		if (error == NFSERR_STALESTATEID)
 			nfscl_initiate_recovery(nmp->nm_clp);
 		if (lckp != NULL)
@@ -1154,7 +1159,9 @@ nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
 	    error == NFSERR_BADSESSION ||
 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
-	     expireret == 0 && clidrev != 0 && retrycnt < 4));
+	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
+	    (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
+	     retrycnt < 4));
 	if (error && retrycnt >= 4)
 		error = EIO;
 	return (error);
@@ -1391,6 +1398,11 @@ nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
 			    &lckp);
 		error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
 		    attrflagp, stuff);
+		if (error == NFSERR_OPENMODE) {
+			NFSLOCKMNT(nmp);
+			nmp->nm_state |= NFSSTA_OPENMODE;
+			NFSUNLOCKMNT(nmp);
+		}
 		if (error == NFSERR_STALESTATEID)
 			nfscl_initiate_recovery(nmp->nm_clp);
 		if (lckp != NULL)
@@ -1409,7 +1421,8 @@ nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
 	    error == NFSERR_BADSESSION ||
 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
-	     expireret == 0 && clidrev != 0 && retrycnt < 4));
+	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
+	    (error == NFSERR_OPENMODE && retrycnt < 4));
 	if (error && retrycnt >= 4)
 		error = EIO;
 	if (NFSHASNFSV4(nmp))
@@ -5594,6 +5607,11 @@ nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
 					if (lastbyte > layp->nfsly_lastbyte)
 						layp->nfsly_lastbyte = lastbyte;
 					NFSUNLOCKCLSTATE();
+				} else if (error == NFSERR_OPENMODE &&
+				    rwaccess == NFSV4OPEN_ACCESSREAD) {
+					NFSLOCKMNT(nmp);
+					nmp->nm_state |= NFSSTA_OPENMODE;
+					NFSUNLOCKMNT(nmp);
 				}
 			} else
 				error = EIO;
diff --git a/sys/fs/nfsclient/nfs_clstate.c b/sys/fs/nfsclient/nfs_clstate.c
index f12757d..0842e52 100644
--- a/sys/fs/nfsclient/nfs_clstate.c
+++ b/sys/fs/nfsclient/nfs_clstate.c
@@ -500,10 +500,11 @@ nfscl_getstateid(vnode_t vp, u_int8_t *nfhp, int fhlen, u_int32_t mode,
 {
 	struct nfsclclient *clp;
 	struct nfsclowner *owp;
-	struct nfsclopen *op = NULL;
+	struct nfsclopen *op = NULL, *top;
 	struct nfscllockowner *lp;
 	struct nfscldeleg *dp;
 	struct nfsnode *np;
+	struct nfsmount *nmp;
 	u_int8_t own[NFSV4CL_LOCKNAMELEN];
 	int error, done;
 
@@ -521,8 +522,9 @@ nfscl_getstateid(vnode_t vp, u_int8_t *nfhp, int fhlen, u_int32_t mode,
 	if (vnode_vtype(vp) != VREG)
 		return (EISDIR);
 	np = VTONFS(vp);
+	nmp = VFSTONFS(vnode_mount(vp));
 	NFSLOCKCLSTATE();
-	clp = nfscl_findcl(VFSTONFS(vnode_mount(vp)));
+	clp = nfscl_findcl(nmp);
 	if (clp == NULL) {
 		NFSUNLOCKCLSTATE();
 		return (EACCES);
@@ -592,23 +594,33 @@ nfscl_getstateid(vnode_t vp, u_int8_t *nfhp, int fhlen, u_int32_t mode,
 	}
 	if (op == NULL) {
 		/* If not found, just look for any OpenOwner that will work. */
+		top = NULL;
 		done = 0;
 		owp = LIST_FIRST(&clp->nfsc_owner);
 		while (!done && owp != NULL) {
 			LIST_FOREACH(op, &owp->nfsow_open, nfso_list) {
 				if (op->nfso_fhlen == fhlen &&
-				    !NFSBCMP(op->nfso_fh, nfhp, fhlen) &&
-				    (mode & op->nfso_mode) == mode) {
-					done = 1;
-					break;
+				    !NFSBCMP(op->nfso_fh, nfhp, fhlen)) {
+					if (top == NULL && (op->nfso_mode &
+					    NFSV4OPEN_ACCESSWRITE) != 0 &&
+					    (mode & NFSV4OPEN_ACCESSREAD) != 0)
+						top = op;
+					if ((mode & op->nfso_mode) == mode) {
+						done = 1;
+						break;
+					}
 				}
 			}
 			if (!done)
 				owp = LIST_NEXT(owp, nfsow_list);
 		}
 		if (!done) {
-			NFSUNLOCKCLSTATE();
-			return (ENOENT);
+			NFSCL_DEBUG(2, "openmode top=%p\n", top);
+			if (top == NULL || NFSHASOPENMODE(nmp)) {
+				NFSUNLOCKCLSTATE();
+				return (ENOENT);
+			} else
+				op = top;
 		}
 		/*
 		 * For read aheads or write behinds, use the open cred.