Post r230394, the Lookup RPC counts for both NFS clients increased

significantly. Upon investigation this was caused by name cache
misses for lookups of "..". For name cache entries for non-".."
directories, the cache entry serves double duty. It maps both the
named directory plus ".." for the parent of the directory. As such,
two ctime values (one for each of the directory and its parent) need
to be saved in the name cache entry.
This patch adds an entry for ctime of the parent directory to the
name cache. It also adds an additional uma zone for large entries
with this time value, in order to minimize memory wastage.
As well, it fixes a couple of cases where the mtime of the parent
directory was being saved instead of ctime for positive name cache
entries. With this patch, Lookup RPC counts return to values similar
to pre-r230394 kernels.

Reported by:	bde
Discussed with:	kib
Reviewed by:	jhb
MFC after:	2 weeks

1 files changed, 41 insertions, 5 deletions

diff --git a/sys/kern/vfs_cache.c b/sys/kern/vfs_cache.c
index 1f081ba..ea481f4 100644
--- a/sys/kern/vfs_cache.c
+++ b/sys/kern/vfs_cache.c
@@ -105,6 +105,10 @@ struct	namecache {
 /*
  * struct namecache_ts repeats struct namecache layout up to the
  * nc_nlen member.
+ * struct namecache_ts is used in place of struct namecache when time(s) need
+ * to be stored.  The nc_dotdottime field is used when a cache entry is mapping
+ * both a non-dotdot directory name plus dotdot for the directory's
+ * parent.
  */
 struct	namecache_ts {
 	LIST_ENTRY(namecache) nc_hash;	/* hash chain */
@@ -115,6 +119,7 @@ struct	namecache_ts {
 	u_char	nc_flag;		/* flag bits */
 	u_char	nc_nlen;		/* length of name */
 	struct	timespec nc_time;	/* timespec provided by fs */
+	struct	timespec nc_dotdottime;	/* dotdot timespec provided by fs */
 	int	nc_ticks;		/* ticks value when entry was added */
 	char	nc_name[0];		/* segment name + nul */
 };
@@ -125,6 +130,7 @@ struct	namecache_ts {
 #define NCF_WHITE	0x01
 #define NCF_ISDOTDOT	0x02
 #define	NCF_TS		0x04
+#define	NCF_DTS		0x08
 
 /*
  * Name caching works as follows:
@@ -190,6 +196,7 @@ RW_SYSINIT(vfscache, &cache_lock, "Name Cache");
 static uma_zone_t cache_zone_small;
 static uma_zone_t cache_zone_small_ts;
 static uma_zone_t cache_zone_large;
+static uma_zone_t cache_zone_large_ts;
 
 #define	CACHE_PATH_CUTOFF	35
 
@@ -197,8 +204,12 @@ static struct namecache *
 cache_alloc(int len, int ts)
 {
 
-	if (len > CACHE_PATH_CUTOFF)
-		return (uma_zalloc(cache_zone_large, M_WAITOK));
+	if (len > CACHE_PATH_CUTOFF) {
+		if (ts)
+			return (uma_zalloc(cache_zone_large_ts, M_WAITOK));
+		else
+			return (uma_zalloc(cache_zone_large, M_WAITOK));
+	}
 	if (ts)
 		return (uma_zalloc(cache_zone_small_ts, M_WAITOK));
 	else
@@ -218,7 +229,9 @@ cache_free(struct namecache *ncp)
 			uma_zfree(cache_zone_small_ts, ncp);
 		else
 			uma_zfree(cache_zone_small, ncp);
-	} else
+	} else if (ts)
+		uma_zfree(cache_zone_large_ts, ncp);
+	else
 		uma_zfree(cache_zone_large, ncp);
 }
 
@@ -521,6 +534,10 @@ retry_wlocked:
 			SDT_PROBE(vfs, namecache, lookup, hit, dvp, "..",
 			    *vpp, 0, 0);
 			cache_out_ts(ncp, tsp, ticksp);
+			if ((ncp->nc_flag & (NCF_ISDOTDOT | NCF_DTS)) ==
+			    NCF_DTS && tsp != NULL)
+				*tsp = ((struct namecache_ts *)ncp)->
+				    nc_dotdottime;
 			goto success;
 		}
 	}
@@ -686,11 +703,12 @@ unlock:
  * Add an entry to the cache.
  */
 void
-cache_enter_time(dvp, vp, cnp, tsp)
+cache_enter_time(dvp, vp, cnp, tsp, dtsp)
 	struct vnode *dvp;
 	struct vnode *vp;
 	struct componentname *cnp;
 	struct timespec *tsp;
+	struct timespec *dtsp;
 {
 	struct namecache *ncp, *n2;
 	struct namecache_ts *n3;
@@ -769,6 +787,10 @@ cache_enter_time(dvp, vp, cnp, tsp)
 		n3->nc_time = *tsp;
 		n3->nc_ticks = ticks;
 		n3->nc_flag |= NCF_TS;
+		if (dtsp != NULL) {
+			n3->nc_dotdottime = *dtsp;
+			n3->nc_flag |= NCF_DTS;
+		}
 	}
 	len = ncp->nc_nlen = cnp->cn_namelen;
 	hash = fnv_32_buf(cnp->cn_nameptr, len, FNV1_32_INIT);
@@ -794,6 +816,12 @@ cache_enter_time(dvp, vp, cnp, tsp)
 				    ((struct namecache_ts *)ncp)->nc_time;
 				n3->nc_ticks =
 				    ((struct namecache_ts *)ncp)->nc_ticks;
+				if (dtsp != NULL) {
+					n3->nc_dotdottime =
+					    ((struct namecache_ts *)ncp)->
+					    nc_dotdottime;
+					n3->nc_flag |= NCF_DTS;
+				}
 			}
 			CACHE_WUNLOCK();
 			cache_free(ncp);
@@ -823,6 +851,11 @@ cache_enter_time(dvp, vp, cnp, tsp)
 			ncp->nc_flag |= NCF_WHITE;
 	} else if (vp->v_type == VDIR) {
 		if (flag != NCF_ISDOTDOT) {
+			/*
+			 * For this case, the cache entry maps both the
+			 * directory name in it and the name ".." for the
+			 * directory's parent.
+			 */
 			if ((n2 = vp->v_cache_dd) != NULL &&
 			    (n2->nc_flag & NCF_ISDOTDOT) != 0)
 				cache_zap(n2);
@@ -886,6 +919,9 @@ nchinit(void *dummy __unused)
 	    sizeof(struct namecache_ts) + CACHE_PATH_CUTOFF + 1,
 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_ZINIT);
 	cache_zone_large = uma_zcreate("L VFS Cache",
+	    sizeof(struct namecache) + NAME_MAX + 1,
+	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_ZINIT);
+	cache_zone_large_ts = uma_zcreate("LTS VFS Cache",
 	    sizeof(struct namecache_ts) + NAME_MAX + 1,
 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_ZINIT);
 
@@ -1382,7 +1418,7 @@ void
 cache_enter(struct vnode *dvp, struct vnode *vp, struct componentname *cnp)
 {
 
-	cache_enter_time(dvp, vp, cnp, NULL);
+	cache_enter_time(dvp, vp, cnp, NULL, NULL);
 }
 
 /*