lockd: Create client-side nlm_host cache

NFS clients don't need the garbage collection processing that is
performed on nlm_host structures.  The client picks up an nlm_host at
mount time and holds a reference to it until the file system is
unmounted.

Servers, on the other hand, don't have a precise way to tell when an
nlm_host is no longer being used, so zero refcount nlm_host entries
are left to expire in the cache after a time.

Basically there's nothing holding a reference to an nlm_host between
individual server-side NLM requests, but we can't afford the expense
of recreating them for every new NLM request from a client.  The
nlm_host cache adds some lifetime hysteresis to entries in the cache
so the next time a particular nlm_host is needed, it's likely to be
discovered by a lookup rather than created from whole cloth.

With the new implementation, client nlm_host cache items are no longer
garbage collected, and are destroyed directly by a new release
function specialized for client entries, nlmclnt_release_host().  They
are cached in their own data structure, and have their own lookup
logic, simplified and specialized for client nlm_host entries.

However, the client nlm_host cache still shares reboot recovery logic
with the server nlm_host cache.  The NSM "peer rebooted" downcall for
clients and servers still come through the same RPC call.  This is a
legacy formal API that would be difficult to alter, and besides, the
user space NSM implementation can't tell the difference between peers
that are clients or servers.

For this reason, the client cache continues to share the
nlm_host_mutex (and reboot recovery logic) with the server cache.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

3 files changed, 77 insertions, 14 deletions

diff --git a/fs/lockd/clntlock.c b/fs/lockd/clntlock.c
index 25509eb..8d4ea83 100644
--- a/fs/lockd/clntlock.c
+++ b/fs/lockd/clntlock.c
@@ -79,7 +79,7 @@ EXPORT_SYMBOL_GPL(nlmclnt_init);
  */
 void nlmclnt_done(struct nlm_host *host)
 {
-	nlm_release_host(host);
+	nlmclnt_release_host(host);
 	lockd_down();
 }
 EXPORT_SYMBOL_GPL(nlmclnt_done);
@@ -273,7 +273,7 @@ restart:
 	spin_unlock(&nlm_blocked_lock);
 
 	/* Release host handle after use */
-	nlm_release_host(host);
+	nlmclnt_release_host(host);
 	lockd_down();
 	return 0;
 }
diff --git a/fs/lockd/clntproc.c b/fs/lockd/clntproc.c
index fbc6617..adb45ec 100644
--- a/fs/lockd/clntproc.c
+++ b/fs/lockd/clntproc.c
@@ -58,7 +58,7 @@ static void nlm_put_lockowner(struct nlm_lockowner *lockowner)
 		return;
 	list_del(&lockowner->list);
 	spin_unlock(&lockowner->host->h_lock);
-	nlm_release_host(lockowner->host);
+	nlmclnt_release_host(lockowner->host);
 	kfree(lockowner);
 }
 
@@ -207,7 +207,7 @@ struct nlm_rqst *nlm_alloc_call(struct nlm_host *host)
 		printk("nlm_alloc_call: failed, waiting for memory\n");
 		schedule_timeout_interruptible(5*HZ);
 	}
-	nlm_release_host(host);
+	nlmclnt_release_host(host);
 	return NULL;
 }
 
@@ -215,7 +215,7 @@ void nlmclnt_release_call(struct nlm_rqst *call)
 {
 	if (!atomic_dec_and_test(&call->a_count))
 		return;
-	nlm_release_host(call->a_host);
+	nlmclnt_release_host(call->a_host);
 	nlmclnt_release_lockargs(call);
 	kfree(call);
 }
diff --git a/fs/lockd/host.c b/fs/lockd/host.c
index e58e142..c6942fb 100644
--- a/fs/lockd/host.c
+++ b/fs/lockd/host.c
@@ -26,6 +26,7 @@
 #define NLM_HOST_COLLECT	(120 * HZ)
 
 static struct hlist_head	nlm_hosts[NLM_HOST_NRHASH];
+static struct hlist_head	nlm_client_hosts[NLM_HOST_NRHASH];
 
 #define for_each_host(host, pos, chain, table) \
 	for ((chain) = (table); \
@@ -288,12 +289,76 @@ struct nlm_host *nlmclnt_lookup_host(const struct sockaddr *sap,
 		.hostname_len	= strlen(hostname),
 		.noresvport	= noresvport,
 	};
+	struct hlist_head *chain;
+	struct hlist_node *pos;
+	struct nlm_host	*host;
+	struct nsm_handle *nsm = NULL;
 
 	dprintk("lockd: %s(host='%s', vers=%u, proto=%s)\n", __func__,
 			(hostname ? hostname : "<none>"), version,
 			(protocol == IPPROTO_UDP ? "udp" : "tcp"));
 
-	return nlm_lookup_host(&ni);
+	mutex_lock(&nlm_host_mutex);
+
+	chain = &nlm_client_hosts[nlm_hash_address(sap)];
+	hlist_for_each_entry(host, pos, chain, h_hash) {
+		if (!rpc_cmp_addr(nlm_addr(host), sap))
+			continue;
+
+		/* Same address. Share an NSM handle if we already have one */
+		if (nsm == NULL)
+			nsm = host->h_nsmhandle;
+
+		if (host->h_proto != protocol)
+			continue;
+		if (host->h_version != version)
+			continue;
+
+		nlm_get_host(host);
+		dprintk("lockd: %s found host %s (%s)\n", __func__,
+			host->h_name, host->h_addrbuf);
+		goto out;
+	}
+
+	host = nlm_alloc_host(&ni, nsm);
+	if (unlikely(host == NULL))
+		goto out;
+
+	hlist_add_head(&host->h_hash, chain);
+	nrhosts++;
+
+	dprintk("lockd: %s created host %s (%s)\n", __func__,
+		host->h_name, host->h_addrbuf);
+
+out:
+	mutex_unlock(&nlm_host_mutex);
+	return host;
+}
+
+/**
+ * nlmclnt_release_host - release client nlm_host
+ * @host: nlm_host to release
+ *
+ */
+void nlmclnt_release_host(struct nlm_host *host)
+{
+	if (host == NULL)
+		return;
+
+	dprintk("lockd: release client host %s\n", host->h_name);
+
+	BUG_ON(atomic_read(&host->h_count) < 0);
+	BUG_ON(host->h_server);
+
+	if (atomic_dec_and_test(&host->h_count)) {
+		BUG_ON(!list_empty(&host->h_lockowners));
+		BUG_ON(!list_empty(&host->h_granted));
+		BUG_ON(!list_empty(&host->h_reclaim));
+
+		mutex_lock(&nlm_host_mutex);
+		nlm_destroy_host_locked(host);
+		mutex_unlock(&nlm_host_mutex);
+	}
 }
 
 /**
@@ -515,16 +580,14 @@ void nlm_host_rebooted(const struct nlm_reboot *info)
 	 * To avoid processing a host several times, we match the nsmstate.
 	 */
 	while ((host = next_host_state(nlm_hosts, nsm, info)) != NULL) {
-		if (host->h_server) {
-			/* We're server for this guy, just ditch
-			 * all the locks he held. */
-			nlmsvc_free_host_resources(host);
-		} else {
-			/* He's the server, initiate lock recovery. */
-			nlmclnt_recovery(host);
-		}
+		nlmsvc_free_host_resources(host);
 		nlm_release_host(host);
 	}
+	while ((host = next_host_state(nlm_client_hosts, nsm, info)) != NULL) {
+		nlmclnt_recovery(host);
+		nlmclnt_release_host(host);
+	}
+
 	nsm_release(nsm);
 }