Add a timeout for the negative name cache entries in the NFS client.

This avoids a bogus negative name cache entry from persisting forever
when another client creates an entry with the same name within the
same NFS server time of day clock tick. The mount option negnametimeo
can be used to override the default timeout interval on a
per-mount-point basis. Setting negnametimeo to 0 disables negative
name caching for the mount point.
I also fixed one obvious typo where args.timeo should be
args.maxgrouplist.

Submitted by:	jhb (earlier version)
Reviewed by:	jhb
MFC after:	2 weeks

4 files changed, 33 insertions, 8 deletions

diff --git a/sys/nfsclient/nfs_vfsops.c b/sys/nfsclient/nfs_vfsops.c
index 2346d41..72ca2ce 100644
--- a/sys/nfsclient/nfs_vfsops.c
+++ b/sys/nfsclient/nfs_vfsops.c
@@ -114,7 +114,7 @@ static void	nfs_decode_args(struct mount *mp, struct nfsmount *nmp,
 		    struct nfs_args *argp, const char *hostname);
 static int	mountnfs(struct nfs_args *, struct mount *,
 		    struct sockaddr *, char *, struct vnode **,
-		    struct ucred *cred);
+		    struct ucred *cred, int);
 static vfs_mount_t nfs_mount;
 static vfs_cmount_t nfs_cmount;
 static vfs_unmount_t nfs_unmount;
@@ -551,7 +551,7 @@ nfs_mountdiskless(char *path,
 
 	nam = sodupsockaddr((struct sockaddr *)sin, M_WAITOK);
 	if ((error = mountnfs(args, mp, nam, path, vpp,
-	    td->td_ucred)) != 0) {
+	    td->td_ucred, NFS_DEFAULT_NEGNAMETIMEO)) != 0) {
 		printf("nfs_mountroot: mount %s on /: %d\n", path, error);
 		return (error);
 	}
@@ -778,7 +778,7 @@ static const char *nfs_opts[] = { "from", "nfs_args",
     "readdirsize", "soft", "hard", "mntudp", "tcp", "udp", "wsize", "rsize",
     "retrans", "acregmin", "acregmax", "acdirmin", "acdirmax", 
     "deadthresh", "hostname", "timeout", "addr", "fh", "nfsv3", "sec",
-    "maxgroups", "principal",
+    "maxgroups", "principal", "negnametimeo",
     NULL };
 
 /*
@@ -827,6 +827,7 @@ nfs_mount(struct mount *mp)
 	size_t len;
 	u_char nfh[NFSX_V3FHMAX];
 	char *opt;
+	int negnametimeo = NFS_DEFAULT_NEGNAMETIMEO;
 
 	has_nfs_args_opt = 0;
 	has_addr_opt = 0;
@@ -1029,7 +1030,7 @@ nfs_mount(struct mount *mp)
 	}
 	if (vfs_getopt(mp->mnt_optnew, "maxgroups", (void **)&opt, NULL) == 0) {
 		ret = sscanf(opt, "%d", &args.maxgrouplist);
-		if (ret != 1 || args.timeo <= 0) {
+		if (ret != 1 || args.maxgrouplist <= 0) {
 			vfs_mount_error(mp, "illegal maxgroups: %s",
 			    opt);
 			error = EINVAL;
@@ -1037,6 +1038,16 @@ nfs_mount(struct mount *mp)
 		}
 		args.flags |= NFSMNT_MAXGRPS;
 	}
+	if (vfs_getopt(mp->mnt_optnew, "negnametimeo", (void **)&opt, NULL)
+	    == 0) {
+		ret = sscanf(opt, "%d", &negnametimeo);
+		if (ret != 1 || negnametimeo < 0) {
+			vfs_mount_error(mp, "illegal negnametimeo: %s",
+			    opt);
+			error = EINVAL;
+			goto out;
+		}
+	}
 	if (vfs_getopt(mp->mnt_optnew, "addr", (void **)&args.addr,
 		&args.addrlen) == 0) {
 		has_addr_opt = 1;
@@ -1125,7 +1136,7 @@ nfs_mount(struct mount *mp)
 		}
 	}
 	error = mountnfs(&args, mp, nam, args.hostname, &vp,
-	    curthread->td_ucred);
+	    curthread->td_ucred, negnametimeo);
 out:
 	if (!error) {
 		MNT_ILOCK(mp);
@@ -1167,7 +1178,7 @@ nfs_cmount(struct mntarg *ma, void *data, int flags)
  */
 static int
 mountnfs(struct nfs_args *argp, struct mount *mp, struct sockaddr *nam,
-    char *hst, struct vnode **vpp, struct ucred *cred)
+    char *hst, struct vnode **vpp, struct ucred *cred, int negnametimeo)
 {
 	struct nfsmount *nmp;
 	struct nfsnode *np;
@@ -1217,6 +1228,7 @@ mountnfs(struct nfs_args *argp, struct mount *mp, struct sockaddr *nam,
 	nmp->nm_numgrps = NFS_MAXGRPS;
 	nmp->nm_readahead = NFS_DEFRAHEAD;
 	nmp->nm_deadthresh = NFS_MAXDEADTHRESH;
+	nmp->nm_negnametimeo = negnametimeo;
 	nmp->nm_tprintf_delay = nfs_tprintf_delay;
 	if (nmp->nm_tprintf_delay < 0)
 		nmp->nm_tprintf_delay = 0;
diff --git a/sys/nfsclient/nfs_vnops.c b/sys/nfsclient/nfs_vnops.c
index 7d6bb30..c74a6be 100644
--- a/sys/nfsclient/nfs_vnops.c
+++ b/sys/nfsclient/nfs_vnops.c
@@ -982,8 +982,13 @@ nfs_lookup(struct vop_lookup_args *ap)
 		 * modification time of the parent directory matches
 		 * our cached copy.  Otherwise, we discard all of the
 		 * negative cache entries for this directory.
+		 * negative cache entries for this directory. We also
+		 * only trust -ve cache entries for less than
+		 * nm_negative_namecache_timeout seconds.
 		 */
-		if (VOP_GETATTR(dvp, &vattr, cnp->cn_cred) == 0 &&
+		if ((u_int)(ticks - np->n_dmtime_ticks) <
+		    (nmp->nm_negnametimeo * hz) &&
+		    VOP_GETATTR(dvp, &vattr, cnp->cn_cred) == 0 &&
 		    vattr.va_mtime.tv_sec == np->n_dmtime) {
 			nfsstats.lookupcache_hits++;
 			return (ENOENT);
@@ -1157,8 +1162,10 @@ nfsmout:
 			 */
 			mtx_lock(&np->n_mtx);
 			if (np->n_dmtime <= dmtime) {
-				if (np->n_dmtime == 0)
+				if (np->n_dmtime == 0) {
 					np->n_dmtime = dmtime;
+					np->n_dmtime_ticks = ticks;
+				}
 				mtx_unlock(&np->n_mtx);
 				cache_enter(dvp, NULL, cnp);
 			} else
diff --git a/sys/nfsclient/nfsmount.h b/sys/nfsclient/nfsmount.h
index 47d7ef3..d47957c 100644
--- a/sys/nfsclient/nfsmount.h
+++ b/sys/nfsclient/nfsmount.h
@@ -85,6 +85,7 @@ struct	nfsmount {
 	struct rpc_timers nm_timers[NFS_MAX_TIMER]; /* RTT Timers for rpcs */
 	char	nm_principal[MNAMELEN];	/* GSS-API principal of server */
 	gss_OID	nm_mech_oid;		/* OID of selected GSS-API mechanism */
+	int	nm_negnametimeo;	/* timeout for -ve entries (sec) */
 
 	/* NFSv4 */
 	uint64_t nm_clientid;
@@ -107,6 +108,10 @@ struct	nfsmount {
 #define NFS_TPRINTF_DELAY               30
 #endif
 
+#ifndef NFS_DEFAULT_NEGNAMETIMEO
+#define NFS_DEFAULT_NEGNAMETIMEO	60
+#endif
+
 #define	NFS_PCATCH	(PCATCH | PBDRY)
 
 #endif
diff --git a/sys/nfsclient/nfsnode.h b/sys/nfsclient/nfsnode.h
index 29198d8..402fc79 100644
--- a/sys/nfsclient/nfsnode.h
+++ b/sys/nfsclient/nfsnode.h
@@ -114,6 +114,7 @@ struct nfsnode {
 	struct timespec		n_mtime;	/* Prev modify time. */
 	time_t			n_ctime;	/* Prev create time. */
 	time_t			n_dmtime;	/* Prev dir modify time. */
+	int			n_dmtime_ticks;	/* Tick of -ve cache entry */
 	time_t			n_expiry;	/* Lease expiry time */
 	nfsfh_t			*n_fhp;		/* NFS File Handle */
 	struct vnode		*n_vnode;	/* associated vnode */