X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=datapath%2Fflow.c;h=c52081b2f7dfa57e67d04f9c449c36a1843a5f24;hb=a4fdb0f3bdbfff1924afefd19239260ed429c193;hp=abe678906a4fe1f6ecf4deb8657c5b89f3eac2dc;hpb=ac3e564e4392ef1fad53f329c14fa4972c87a952;p=sliver-openvswitch.git

diff --git a/datapath/flow.c b/datapath/flow.c
index abe678906..c52081b2f 100644
--- a/datapath/flow.c
+++ b/datapath/flow.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2007-2013 Nicira, Inc.
+ * Copyright (c) 2007-2014 Nicira, Inc.
  *
  * This program is free software; you can redistribute it and/or
  * modify it under the terms of version 2 of the GNU General Public
@@ -62,113 +62,112 @@ u64 ovs_flow_used_time(unsigned long flow_jiffies)
 
 #define TCP_FLAGS_BE16(tp) (*(__be16 *)&tcp_flag_word(tp) & htons(0x0FFF))
 
-void ovs_flow_stats_update(struct sw_flow *flow, struct sk_buff *skb)
+void ovs_flow_stats_update(struct sw_flow *flow, __be16 tcp_flags,
+			   struct sk_buff *skb)
 {
 	struct flow_stats *stats;
-	__be16 tcp_flags = 0;
-
-	if (!flow->stats.is_percpu)
-		stats = flow->stats.stat;
-	else
-		stats = this_cpu_ptr(flow->stats.cpu_stats);
-
-	if ((flow->key.eth.type == htons(ETH_P_IP) ||
-	     flow->key.eth.type == htons(ETH_P_IPV6)) &&
-	    flow->key.ip.frag != OVS_FRAG_TYPE_LATER &&
-	    flow->key.ip.proto == IPPROTO_TCP &&
-	    likely(skb->len >= skb_transport_offset(skb) + sizeof(struct tcphdr))) {
-		tcp_flags = TCP_FLAGS_BE16(tcp_hdr(skb));
+	int node = numa_node_id();
+
+	stats = rcu_dereference(flow->stats[node]);
+
+	/* Check if already have node-specific stats. */
+	if (likely(stats)) {
+		spin_lock(&stats->lock);
+		/* Mark if we write on the pre-allocated stats. */
+		if (node == 0 && unlikely(flow->stats_last_writer != node))
+			flow->stats_last_writer = node;
+	} else {
+		stats = rcu_dereference(flow->stats[0]); /* Pre-allocated. */
+		spin_lock(&stats->lock);
+
+		/* If the current NUMA-node is the only writer on the
+		 * pre-allocated stats keep using them.
+		 */
+		if (unlikely(flow->stats_last_writer != node)) {
+			/* A previous locker may have already allocated the
+			 * stats, so we need to check again.  If node-specific
+			 * stats were already allocated, we update the pre-
+			 * allocated stats as we have already locked them.
+			 */
+			if (likely(flow->stats_last_writer != NUMA_NO_NODE)
+			    && likely(!rcu_dereference(flow->stats[node]))) {
+				/* Try to allocate node-specific stats. */
+				struct flow_stats *new_stats;
+
+				new_stats =
+					kmem_cache_alloc_node(flow_stats_cache,
+							      GFP_THISNODE |
+							      __GFP_NOMEMALLOC,
+							      node);
+				if (likely(new_stats)) {
+					new_stats->used = jiffies;
+					new_stats->packet_count = 1;
+					new_stats->byte_count = skb->len;
+					new_stats->tcp_flags = tcp_flags;
+					spin_lock_init(&new_stats->lock);
+
+					rcu_assign_pointer(flow->stats[node],
+							   new_stats);
+					goto unlock;
+				}
+			}
+			flow->stats_last_writer = node;
+		}
 	}
 
-	spin_lock(&stats->lock);
 	stats->used = jiffies;
 	stats->packet_count++;
 	stats->byte_count += skb->len;
 	stats->tcp_flags |= tcp_flags;
+unlock:
 	spin_unlock(&stats->lock);
 }
 
-static void stats_read(struct flow_stats *stats, bool lock_bh,
-		       struct ovs_flow_stats *ovs_stats,
-		       unsigned long *used, __be16 *tcp_flags)
-{
-	if (lock_bh)
-		spin_lock_bh(&stats->lock);
-	else
-		spin_lock(&stats->lock);
-
-	if (time_after(stats->used, *used))
-		*used = stats->used;
-	*tcp_flags |= stats->tcp_flags;
-	ovs_stats->n_packets += stats->packet_count;
-	ovs_stats->n_bytes += stats->byte_count;
-
-	if (lock_bh)
-		spin_unlock_bh(&stats->lock);
-	else
-		spin_unlock(&stats->lock);
-}
-
-void ovs_flow_stats_get(struct sw_flow *flow, struct ovs_flow_stats *ovs_stats,
+/* Must be called with rcu_read_lock or ovs_mutex. */
+void ovs_flow_stats_get(const struct sw_flow *flow,
+			struct ovs_flow_stats *ovs_stats,
 			unsigned long *used, __be16 *tcp_flags)
 {
-	int cpu, cur_cpu;
+	int node;
 
 	*used = 0;
 	*tcp_flags = 0;
 	memset(ovs_stats, 0, sizeof(*ovs_stats));
 
-	if (!flow->stats.is_percpu) {
-		stats_read(flow->stats.stat, true, ovs_stats, used, tcp_flags);
-	} else {
-		cur_cpu = get_cpu();
-
-		for_each_possible_cpu(cpu) {
-			struct flow_stats *stats;
-			bool lock_bh;
+	for_each_node(node) {
+		struct flow_stats *stats = rcu_dereference_ovsl(flow->stats[node]);
 
-			stats = per_cpu_ptr(flow->stats.cpu_stats, cpu);
-			lock_bh = (cpu == cur_cpu);
-			stats_read(stats, lock_bh, ovs_stats, used, tcp_flags);
+		if (stats) {
+			/* Local CPU may write on non-local stats, so we must
+			 * block bottom-halves here.
+			 */
+			spin_lock_bh(&stats->lock);
+			if (!*used || time_after(stats->used, *used))
+				*used = stats->used;
+			*tcp_flags |= stats->tcp_flags;
+			ovs_stats->n_packets += stats->packet_count;
+			ovs_stats->n_bytes += stats->byte_count;
+			spin_unlock_bh(&stats->lock);
 		}
-		put_cpu();
 	}
 }
 
-static void stats_reset(struct flow_stats *stats, bool lock_bh)
-{
-	if (lock_bh)
-		spin_lock_bh(&stats->lock);
-	else
-		spin_lock(&stats->lock);
-
-	stats->used = 0;
-	stats->packet_count = 0;
-	stats->byte_count = 0;
-	stats->tcp_flags = 0;
-
-	if (lock_bh)
-		spin_unlock_bh(&stats->lock);
-	else
-		spin_unlock(&stats->lock);
-}
-
+/* Called with ovs_mutex. */
 void ovs_flow_stats_clear(struct sw_flow *flow)
 {
-	int cpu, cur_cpu;
-
-	if (!flow->stats.is_percpu) {
-		stats_reset(flow->stats.stat, true);
-	} else {
-		cur_cpu = get_cpu();
-
-		for_each_possible_cpu(cpu) {
-			bool lock_bh;
-
-			lock_bh = (cpu == cur_cpu);
-			stats_reset(per_cpu_ptr(flow->stats.cpu_stats, cpu), lock_bh);
+	int node;
+
+	for_each_node(node) {
+		struct flow_stats *stats = ovsl_dereference(flow->stats[node]);
+
+		if (stats) {
+			spin_lock_bh(&stats->lock);
+			stats->used = 0;
+			stats->packet_count = 0;
+			stats->byte_count = 0;
+			stats->tcp_flags = 0;
+			spin_unlock_bh(&stats->lock);
 		}
-		put_cpu();
 	}
 }
 
@@ -356,8 +355,8 @@ static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key,
 	/* The ICMPv6 type and code fields use the 16-bit transport port
 	 * fields, so we need to store them in 16-bit network byte order.
 	 */
-	key->ipv6.tp.src = htons(icmp->icmp6_type);
-	key->ipv6.tp.dst = htons(icmp->icmp6_code);
+	key->tp.src = htons(icmp->icmp6_type);
+	key->tp.dst = htons(icmp->icmp6_code);
 
 	if (icmp->icmp6_code == 0 &&
 	    (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION ||
@@ -396,14 +395,14 @@ static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key,
 			    && opt_len == 8) {
 				if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll)))
 					goto invalid;
-				memcpy(key->ipv6.nd.sll,
-				    &nd->opt[offset+sizeof(*nd_opt)], ETH_ALEN);
+				ether_addr_copy(key->ipv6.nd.sll,
+				    &nd->opt[offset+sizeof(*nd_opt)]);
 			} else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR
 				   && opt_len == 8) {
 				if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll)))
 					goto invalid;
-				memcpy(key->ipv6.nd.tll,
-				    &nd->opt[offset+sizeof(*nd_opt)], ETH_ALEN);
+				ether_addr_copy(key->ipv6.nd.tll,
+				    &nd->opt[offset+sizeof(*nd_opt)]);
 			}
 
 			icmp_len -= opt_len;
@@ -463,8 +462,8 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key)
 	 * header in the linear data area.
 	 */
 	eth = eth_hdr(skb);
-	memcpy(key->eth.src, eth->h_source, ETH_ALEN);
-	memcpy(key->eth.dst, eth->h_dest, ETH_ALEN);
+	ether_addr_copy(key->eth.src, eth->h_source);
+	ether_addr_copy(key->eth.dst, eth->h_dest);
 
 	__skb_pull(skb, 2 * ETH_ALEN);
 	/* We are going to push all headers that we pull, so no need to
@@ -518,21 +517,21 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key)
 		if (key->ip.proto == IPPROTO_TCP) {
 			if (tcphdr_ok(skb)) {
 				struct tcphdr *tcp = tcp_hdr(skb);
-				key->ipv4.tp.src = tcp->source;
-				key->ipv4.tp.dst = tcp->dest;
-				key->ipv4.tp.flags = TCP_FLAGS_BE16(tcp);
+				key->tp.src = tcp->source;
+				key->tp.dst = tcp->dest;
+				key->tp.flags = TCP_FLAGS_BE16(tcp);
 			}
 		} else if (key->ip.proto == IPPROTO_UDP) {
 			if (udphdr_ok(skb)) {
 				struct udphdr *udp = udp_hdr(skb);
-				key->ipv4.tp.src = udp->source;
-				key->ipv4.tp.dst = udp->dest;
+				key->tp.src = udp->source;
+				key->tp.dst = udp->dest;
 			}
 		} else if (key->ip.proto == IPPROTO_SCTP) {
 			if (sctphdr_ok(skb)) {
 				struct sctphdr *sctp = sctp_hdr(skb);
-				key->ipv4.tp.src = sctp->source;
-				key->ipv4.tp.dst = sctp->dest;
+				key->tp.src = sctp->source;
+				key->tp.dst = sctp->dest;
 			}
 		} else if (key->ip.proto == IPPROTO_ICMP) {
 			if (icmphdr_ok(skb)) {
@@ -540,8 +539,8 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key)
 				/* The ICMP type and code fields use the 16-bit
 				 * transport port fields, so we need to store
 				 * them in 16-bit network byte order. */
-				key->ipv4.tp.src = htons(icmp->type);
-				key->ipv4.tp.dst = htons(icmp->code);
+				key->tp.src = htons(icmp->type);
+				key->tp.dst = htons(icmp->code);
 			}
 		}
 
@@ -561,8 +560,8 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key)
 				key->ip.proto = ntohs(arp->ar_op);
 			memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src));
 			memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst));
-			memcpy(key->ipv4.arp.sha, arp->ar_sha, ETH_ALEN);
-			memcpy(key->ipv4.arp.tha, arp->ar_tha, ETH_ALEN);
+			ether_addr_copy(key->ipv4.arp.sha, arp->ar_sha);
+			ether_addr_copy(key->ipv4.arp.tha, arp->ar_tha);
 		}
 	} else if (key->eth.type == htons(ETH_P_IPV6)) {
 		int nh_len;             /* IPv6 Header + Extensions */
@@ -587,21 +586,21 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key)
 		if (key->ip.proto == NEXTHDR_TCP) {
 			if (tcphdr_ok(skb)) {
 				struct tcphdr *tcp = tcp_hdr(skb);
-				key->ipv6.tp.src = tcp->source;
-				key->ipv6.tp.dst = tcp->dest;
-				key->ipv6.tp.flags = TCP_FLAGS_BE16(tcp);
+				key->tp.src = tcp->source;
+				key->tp.dst = tcp->dest;
+				key->tp.flags = TCP_FLAGS_BE16(tcp);
 			}
 		} else if (key->ip.proto == NEXTHDR_UDP) {
 			if (udphdr_ok(skb)) {
 				struct udphdr *udp = udp_hdr(skb);
-				key->ipv6.tp.src = udp->source;
-				key->ipv6.tp.dst = udp->dest;
+				key->tp.src = udp->source;
+				key->tp.dst = udp->dest;
 			}
 		} else if (key->ip.proto == NEXTHDR_SCTP) {
 			if (sctphdr_ok(skb)) {
 				struct sctphdr *sctp = sctp_hdr(skb);
-				key->ipv6.tp.src = sctp->source;
-				key->ipv6.tp.dst = sctp->dest;
+				key->tp.src = sctp->source;
+				key->tp.dst = sctp->dest;
 			}
 		} else if (key->ip.proto == NEXTHDR_ICMP) {
 			if (icmp6hdr_ok(skb)) {