[PATCH] knfsd: knfsd: cache ipmap per TCP socket

Speed up high call-rate workloads by caching the struct ip_map for the peer on
the connected struct svc_sock instead of looking it up in the ip_map cache
hashtable on every call.  This helps workloads using AUTH_SYS authentication
over TCP.

Testing was on a 4 CPU 4 NIC Altix using 4 IRIX clients, each with 16
synthetic client threads simulating an rsync (i.e.  recursive directory
listing) workload reading from an i386 RH9 install image (161480 regular files
in 10841 directories) on the server.  That tree is small enough to fill in the
server's RAM so no disk traffic was involved.  This setup gives a sustained
call rate in excess of 60000 calls/sec before being CPU-bound on the server.

Profiling showed strcmp(), called from ip_map_match(), was taking 4.8% of each
CPU, and ip_map_lookup() was taking 2.9%.  This patch drops both contribution
into the profile noise.

Note that the above result overstates this value of this patch for most
workloads.  The synthetic clients are all using separate IP addresses, so
there are 64 entries in the ip_map cache hash.  Because the kernel measured
contained the bug fixed in commit

commit 1f1e030b

and was running on 64bit little-endian machine, probably all of those 64
entries were on a single chain, thus increasing the cost of ip_map_lookup().

With a modern kernel you would need more clients to see the same amount of
performance improvement.  This patch has helped to scale knfsd to handle a
deployment with 2000 NFS clients.
Signed-off-by: NGreg Banks <gnb@melbourne.sgi.com>
Signed-off-by: NNeil Brown <neilb@suse.de>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

include/linux/sunrpc/cache.h

@@ -163,6 +163,17 @@ static inline void cache_put(struct cache_head *h, struct cache_detail *cd)
 	kref_put(&h->ref, cd->cache_put);
 }
 
+static inline int cache_valid(struct cache_head *h)
+{
+	/* If an item has been unhashed pending removal when
+	 * the refcount drops to 0, the expiry_time will be
+	 * set to 0.  We don't want to consider such items
+	 * valid in this context even though CACHE_VALID is
+	 * set.
+	 */
+	return (h->expiry_time != 0 && test_bit(CACHE_VALID, &h->flags));
+}
+
 extern int cache_check(struct cache_detail *detail,
 		       struct cache_head *h, struct cache_req *rqstp);
 extern void cache_flush(void);

include/linux/sunrpc/svcauth.h

@@ -126,6 +126,7 @@ extern struct auth_domain *auth_domain_find(char *name);
 extern struct auth_domain *auth_unix_lookup(struct in_addr addr);
 extern int auth_unix_forget_old(struct auth_domain *dom);
 extern void svcauth_unix_purge(void);
+extern void svcauth_unix_info_release(void *);
 
 static inline unsigned long hash_str(char *name, int bits)
 {

include/linux/sunrpc/svcsock.h

@@ -54,6 +54,9 @@ struct svc_sock {
 	int			sk_reclen;	/* length of record */
 	int			sk_tcplen;	/* current read length */
 	time_t			sk_lastrecv;	/* time of last received request */
+
+	/* cache of various info for TCP sockets */
+	void			*sk_info_authunix;
 };
 
 /*

net/sunrpc/svcauth_unix.c

@@ -9,6 +9,7 @@
 #include <linux/seq_file.h>
 #include <linux/hash.h>
 #include <linux/string.h>
+#include <net/sock.h>
 
 #define RPCDBG_FACILITY	RPCDBG_AUTH
 
@@ -375,6 +376,44 @@ void svcauth_unix_purge(void)
 	cache_purge(&ip_map_cache);
 }
 
+static inline struct ip_map *
+ip_map_cached_get(struct svc_rqst *rqstp)
+{
+	struct ip_map *ipm = rqstp->rq_sock->sk_info_authunix;
+	if (ipm != NULL) {
+		if (!cache_valid(&ipm->h)) {
+			/*
+			 * The entry has been invalidated since it was
+			 * remembered, e.g. by a second mount from the
+			 * same IP address.
+			 */
+			rqstp->rq_sock->sk_info_authunix = NULL;
+			cache_put(&ipm->h, &ip_map_cache);
+			return NULL;
+		}
+		cache_get(&ipm->h);
+	}
+	return ipm;
+}
+
+static inline void
+ip_map_cached_put(struct svc_rqst *rqstp, struct ip_map *ipm)
+{
+	struct svc_sock *svsk = rqstp->rq_sock;
+
+	if (svsk->sk_sock->type == SOCK_STREAM && svsk->sk_info_authunix == NULL)
+		svsk->sk_info_authunix = ipm;	/* newly cached, keep the reference */
+	else
+		cache_put(&ipm->h, &ip_map_cache);
+}
+
+void
+svcauth_unix_info_release(void *info)
+{
+	struct ip_map *ipm = info;
+	cache_put(&ipm->h, &ip_map_cache);
+}
+
 static int
 svcauth_unix_set_client(struct svc_rqst *rqstp)
 {
@@ -384,6 +423,8 @@ svcauth_unix_set_client(struct svc_rqst *rqstp)
 	if (rqstp->rq_proc == 0)
 		return SVC_OK;
 
+	ipm = ip_map_cached_get(rqstp);
+	if (ipm == NULL)
 		ipm = ip_map_lookup(rqstp->rq_server->sv_program->pg_class,
 				    rqstp->rq_addr.sin_addr);
 
@@ -400,7 +441,7 @@ svcauth_unix_set_client(struct svc_rqst *rqstp)
 		case 0:
 			rqstp->rq_client = &ipm->m_client->h;
 			kref_get(&rqstp->rq_client->ref);
-			cache_put(&ipm->h, &ip_map_cache);
+			ip_map_cached_put(rqstp, ipm);
 			break;
 	}
 	return SVC_OK;

net/sunrpc/svcsock.c

@@ -1610,6 +1610,8 @@ svc_delete_socket(struct svc_sock *svsk)
 			sockfd_put(svsk->sk_sock);
 		else
 			sock_release(svsk->sk_sock);
+		if (svsk->sk_info_authunix != NULL)
+			svcauth_unix_info_release(svsk->sk_info_authunix);
 		kfree(svsk);
 	} else {
 		spin_unlock_bh(&serv->sv_lock);