X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=datapath%2Factions.c;h=945c461cdc01c6dd965ddc9a9a20ecb24ac569e7;hb=6d23c6f42206809f985d3fc7d95fca0589522836;hp=b6b713534565bc6961c514a324df56dfebb0990d;hpb=007948177581f3b3dad188221593d0e4bdca6ba0;p=openvswitch

diff --git a/datapath/actions.c b/datapath/actions.c
index b6b71353..945c461c 100644
--- a/datapath/actions.c
+++ b/datapath/actions.c
@@ -8,6 +8,8 @@
 
 /* Functions for executing flow actions. */
 
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
 #include <linux/skbuff.h>
 #include <linux/in.h>
 #include <linux/ip.h>
@@ -23,104 +25,107 @@
 #include "actions.h"
 #include "checksum.h"
 #include "datapath.h"
-#include "loop_counter.h"
 #include "openvswitch/datapath-protocol.h"
 #include "vlan.h"
 #include "vport.h"
 
-static int do_execute_actions(struct datapath *, struct sk_buff *,
-			      struct sw_flow_actions *acts);
+static int do_execute_actions(struct datapath *dp, struct sk_buff *skb,
+			const struct nlattr *attr, int len, bool keep_skb);
 
-static struct sk_buff *make_writable(struct sk_buff *skb, unsigned min_headroom)
+static int make_writable(struct sk_buff *skb, int write_len)
 {
-	if (skb_cloned(skb)) {
-		struct sk_buff *nskb;
-		unsigned headroom = max(min_headroom, skb_headroom(skb));
+	if (!skb_cloned(skb) || skb_clone_writable(skb, write_len))
+		return 0;
 
-		nskb = skb_copy_expand(skb, headroom, skb_tailroom(skb), GFP_ATOMIC);
-		if (nskb) {
-			set_skb_csum_bits(skb, nskb);
-			kfree_skb(skb);
-			return nskb;
-		}
-	} else {
-		unsigned int hdr_len = (skb_transport_offset(skb)
-					+ sizeof(struct tcphdr));
-		if (pskb_may_pull(skb, min(hdr_len, skb->len)))
-			return skb;
-	}
-	kfree_skb(skb);
-	return NULL;
+	return pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
 }
 
-static struct sk_buff *strip_vlan(struct sk_buff *skb)
+/* remove VLAN header from packet and update csum accrodingly. */
+static int __pop_vlan_tci(struct sk_buff *skb, __be16 *current_tci)
 {
 	struct ethhdr *eh;
+	struct vlan_ethhdr *veth;
+	int err;
 
-	if (vlan_tx_tag_present(skb)) {
-		vlan_set_tci(skb, 0);
-		return skb;
-	}
-
-	if (unlikely(vlan_eth_hdr(skb)->h_vlan_proto != htons(ETH_P_8021Q) ||
-	    skb->len < VLAN_ETH_HLEN))
-		return skb;
-
-	skb = make_writable(skb, 0);
-	if (unlikely(!skb))
-		return NULL;
+	err = make_writable(skb, VLAN_ETH_HLEN);
+	if (unlikely(err))
+		return err;
 
 	if (get_ip_summed(skb) == OVS_CSUM_COMPLETE)
 		skb->csum = csum_sub(skb->csum, csum_partial(skb->data
 					+ ETH_HLEN, VLAN_HLEN, 0));
 
+	veth = (struct vlan_ethhdr *) skb->data;
+	*current_tci = veth->h_vlan_TCI;
+
 	memmove(skb->data + VLAN_HLEN, skb->data, 2 * ETH_ALEN);
 
-	eh = (struct ethhdr *)skb_pull(skb, VLAN_HLEN);
+	eh = (struct ethhdr *)__skb_pull(skb, VLAN_HLEN);
 
 	skb->protocol = eh->h_proto;
 	skb->mac_header += VLAN_HLEN;
 
-	return skb;
+	return 0;
 }
 
-static struct sk_buff *modify_vlan_tci(struct sk_buff *skb, __be16 tci)
+static int pop_vlan(struct sk_buff *skb)
 {
-	struct vlan_ethhdr *vh;
-	__be16 old_tci;
+	__be16 tci;
+	int err;
+
+	if (likely(vlan_tx_tag_present(skb))) {
+		vlan_set_tci(skb, 0);
+	} else {
+		if (unlikely(skb->protocol != htons(ETH_P_8021Q) ||
+		    skb->len < VLAN_ETH_HLEN))
+			return 0;
 
-	if (vlan_tx_tag_present(skb) || skb->protocol != htons(ETH_P_8021Q))
-		return __vlan_hwaccel_put_tag(skb, ntohs(tci));
+		err = __pop_vlan_tci(skb, &tci);
+		if (err)
+			return err;
+	}
+	/* move next vlan tag to hw accel tag */
+	if (likely(skb->protocol != htons(ETH_P_8021Q) ||
+	    skb->len < VLAN_ETH_HLEN))
+		return 0;
 
-	skb = make_writable(skb, 0);
-	if (unlikely(!skb))
-		return NULL;
+	err = __pop_vlan_tci(skb, &tci);
+	if (unlikely(err))
+		return err;
 
-	if (unlikely(skb->len < VLAN_ETH_HLEN))
-		return skb;
+	__vlan_hwaccel_put_tag(skb, ntohs(tci));
+	return 0;
+}
 
-	vh = vlan_eth_hdr(skb);
+static int push_vlan(struct sk_buff *skb, __be16 new_tci)
+{
+	if (unlikely(vlan_tx_tag_present(skb))) {
+		u16 current_tag;
 
-	old_tci = vh->h_vlan_TCI;
-	vh->h_vlan_TCI = tci;
+		/* push down current VLAN tag */
+		current_tag = vlan_tx_tag_get(skb);
 
-	if (get_ip_summed(skb) == OVS_CSUM_COMPLETE) {
-		__be16 diff[] = { ~old_tci, vh->h_vlan_TCI };
-		skb->csum = ~csum_partial((char *)diff, sizeof(diff), ~skb->csum);
-	}
+		if (!__vlan_put_tag(skb, current_tag))
+			return -ENOMEM;
 
-	return skb;
+		if (get_ip_summed(skb) == OVS_CSUM_COMPLETE)
+			skb->csum = csum_add(skb->csum, csum_partial(skb->data
+					+ ETH_HLEN, VLAN_HLEN, 0));
+
+	}
+	__vlan_hwaccel_put_tag(skb, ntohs(new_tci));
+	return 0;
 }
 
 static bool is_ip(struct sk_buff *skb)
 {
-	return (OVS_CB(skb)->flow->key.dl_type == htons(ETH_P_IP) &&
+	return (OVS_CB(skb)->flow->key.eth.type == htons(ETH_P_IP) &&
 		skb->transport_header > skb->network_header);
 }
 
 static __sum16 *get_l4_checksum(struct sk_buff *skb)
 {
-	u8 nw_proto = OVS_CB(skb)->flow->key.nw_proto;
+	u8 nw_proto = OVS_CB(skb)->flow->key.ip.proto;
 	int transport_len = skb->len - skb_transport_offset(skb);
 	if (nw_proto == IPPROTO_TCP) {
 		if (likely(transport_len >= sizeof(struct tcphdr)))
@@ -132,22 +137,24 @@ static __sum16 *get_l4_checksum(struct sk_buff *skb)
 	return NULL;
 }
 
-static struct sk_buff *set_nw_addr(struct sk_buff *skb, const struct nlattr *a)
+static int set_nw_addr(struct sk_buff *skb, const struct nlattr *a)
 {
 	__be32 new_nwaddr = nla_get_be32(a);
 	struct iphdr *nh;
 	__sum16 *check;
 	__be32 *nwaddr;
+	int err;
 
 	if (unlikely(!is_ip(skb)))
-		return skb;
+		return 0;
 
-	skb = make_writable(skb, 0);
-	if (unlikely(!skb))
-		return NULL;
+	err = make_writable(skb, skb_network_offset(skb) +
+				 sizeof(struct iphdr));
+	if (unlikely(err))
+		return err;
 
 	nh = ip_hdr(skb);
-	nwaddr = nla_type(a) == ODP_ACTION_ATTR_SET_NW_SRC ? &nh->saddr : &nh->daddr;
+	nwaddr = nla_type(a) == OVS_ACTION_ATTR_SET_NW_SRC ? &nh->saddr : &nh->daddr;
 
 	check = get_l4_checksum(skb);
 	if (likely(check))
@@ -158,47 +165,52 @@ static struct sk_buff *set_nw_addr(struct sk_buff *skb, const struct nlattr *a)
 
 	*nwaddr = new_nwaddr;
 
-	return skb;
+	return 0;
 }
 
-static struct sk_buff *set_nw_tos(struct sk_buff *skb, u8 nw_tos)
+static int set_nw_tos(struct sk_buff *skb, u8 nw_tos)
 {
+	struct iphdr *nh = ip_hdr(skb);
+	u8 old, new;
+	int err;
+
 	if (unlikely(!is_ip(skb)))
-		return skb;
-
-	skb = make_writable(skb, 0);
-	if (skb) {
-		struct iphdr *nh = ip_hdr(skb);
-		u8 *f = &nh->tos;
-		u8 old = *f;
-		u8 new;
-
-		/* Set the DSCP bits and preserve the ECN bits. */
-		new = nw_tos | (nh->tos & INET_ECN_MASK);
-		csum_replace4(&nh->check, (__force __be32)old,
-					  (__force __be32)new);
-		*f = new;
-	}
-	return skb;
+		return 0;
+
+	err = make_writable(skb, skb_network_offset(skb) +
+				 sizeof(struct iphdr));
+	if (unlikely(err))
+		return err;
+
+	/* Set the DSCP bits and preserve the ECN bits. */
+	old = nh->tos;
+	new = nw_tos | (nh->tos & INET_ECN_MASK);
+	csum_replace4(&nh->check, (__force __be32)old,
+				  (__force __be32)new);
+	nh->tos = new;
+
+	return 0;
 }
 
-static struct sk_buff *set_tp_port(struct sk_buff *skb, const struct nlattr *a)
+static int set_tp_port(struct sk_buff *skb, const struct nlattr *a)
 {
 	struct udphdr *th;
 	__sum16 *check;
 	__be16 *port;
+	int err;
 
 	if (unlikely(!is_ip(skb)))
-		return skb;
+		return 0;
 
-	skb = make_writable(skb, 0);
-	if (unlikely(!skb))
-		return NULL;
+	err = make_writable(skb, skb_transport_offset(skb) +
+				 sizeof(struct tcphdr));
+	if (unlikely(err))
+		return err;
 
 	/* Must follow make_writable() since that can move the skb data. */
 	check = get_l4_checksum(skb);
 	if (unlikely(!check))
-		return skb;
+		return 0;
 
 	/*
 	 * Update port and checksum.
@@ -208,80 +220,69 @@ static struct sk_buff *set_tp_port(struct sk_buff *skb, const struct nlattr *a)
 	 * supports those protocols.
 	 */
 	th = udp_hdr(skb);
-	port = nla_type(a) == ODP_ACTION_ATTR_SET_TP_SRC ? &th->source : &th->dest;
+	port = nla_type(a) == OVS_ACTION_ATTR_SET_TP_SRC ? &th->source : &th->dest;
 	inet_proto_csum_replace2(check, skb, *port, nla_get_be16(a), 0);
 	*port = nla_get_be16(a);
 	skb_clear_rxhash(skb);
 
-	return skb;
+	return 0;
 }
 
-/**
- * is_spoofed_arp - check for invalid ARP packet
- *
- * @skb: skbuff containing an Ethernet packet, with network header pointing
- * just past the Ethernet and optional 802.1Q header.
- *
- * Returns true if @skb is an invalid Ethernet+IPv4 ARP packet: one with screwy
- * or truncated header fields or one whose inner and outer Ethernet address
- * differ.
- */
-static bool is_spoofed_arp(struct sk_buff *skb)
+static int do_output(struct datapath *dp, struct sk_buff *skb, int out_port)
 {
-	struct arp_eth_header *arp;
+	struct vport *vport;
 
-	if (OVS_CB(skb)->flow->key.dl_type != htons(ETH_P_ARP))
-		return false;
+	if (unlikely(!skb))
+		return -ENOMEM;
 
-	if (skb_network_offset(skb) + sizeof(struct arp_eth_header) > skb->len)
-		return true;
+	vport = rcu_dereference(dp->ports[out_port]);
+	if (unlikely(!vport)) {
+		kfree_skb(skb);
+		return -ENODEV;
+	}
 
-	arp = (struct arp_eth_header *)skb_network_header(skb);
-	return (arp->ar_hrd != htons(ARPHRD_ETHER) ||
-		arp->ar_pro != htons(ETH_P_IP) ||
-		arp->ar_hln != ETH_ALEN ||
-		arp->ar_pln != 4 ||
-		compare_ether_addr(arp->ar_sha, eth_hdr(skb)->h_source));
+	vport_send(vport, skb);
+	return 0;
 }
 
-static void do_output(struct datapath *dp, struct sk_buff *skb, int out_port)
+static int output_userspace(struct datapath *dp, struct sk_buff *skb, u64 arg)
 {
-	struct vport *p;
-
-	if (!skb)
-		goto error;
-
-	p = rcu_dereference(dp->ports[out_port]);
-	if (!p)
-		goto error;
-
-	vport_send(p, skb);
-	return;
+	struct dp_upcall_info upcall;
 
-error:
-	kfree_skb(skb);
+	upcall.cmd = OVS_PACKET_CMD_ACTION;
+	upcall.key = &OVS_CB(skb)->flow->key;
+	upcall.userdata = arg;
+	return dp_upcall(dp, skb, &upcall);
 }
 
-static int output_control(struct datapath *dp, struct sk_buff *skb, u64 arg)
+static int sample(struct datapath *dp, struct sk_buff *skb,
+		  const struct nlattr *attr)
 {
-	struct dp_upcall_info upcall;
+	const struct nlattr *acts_list = NULL;
+	const struct nlattr *a;
+	int rem;
 
-	skb = skb_clone(skb, GFP_ATOMIC);
-	if (!skb)
-		return -ENOMEM;
+	for (a = nla_data(attr), rem = nla_len(attr); rem > 0;
+		 a = nla_next(a, &rem)) {
+		switch (nla_type(a)) {
+		case OVS_SAMPLE_ATTR_PROBABILITY:
+			if (net_random() >= nla_get_u32(a))
+				return 0;
+			break;
 
-	upcall.cmd = ODP_PACKET_CMD_ACTION;
-	upcall.key = &OVS_CB(skb)->flow->key;
-	upcall.userdata = arg;
-	upcall.sample_pool = 0;
-	upcall.actions = NULL;
-	upcall.actions_len = 0;
-	return dp_upcall(dp, skb, &upcall);
+		case OVS_SAMPLE_ATTR_ACTIONS:
+			acts_list = a;
+			break;
+		}
+	}
+
+	return do_execute_actions(dp, skb, nla_data(acts_list),
+						 nla_len(acts_list), true);
 }
 
 /* Execute a list of actions against 'skb'. */
 static int do_execute_actions(struct datapath *dp, struct sk_buff *skb,
-			      struct sw_flow_actions *acts)
+			const struct nlattr *attr, int len, bool keep_skb)
 {
 	/* Every output action needs a separate clone of 'skb', but the common
 	 * case is just a single output action, so that doing a clone and
@@ -290,117 +291,114 @@ static int do_execute_actions(struct datapath *dp, struct sk_buff *skb,
 	int prev_port = -1;
 	u32 priority = skb->priority;
 	const struct nlattr *a;
-	int rem, err;
+	int rem;
 
-	for (a = acts->actions, rem = acts->actions_len; rem > 0;
+	for (a = attr, rem = len; rem > 0;
 	     a = nla_next(a, &rem)) {
+		int err = 0;
+
 		if (prev_port != -1) {
 			do_output(dp, skb_clone(skb, GFP_ATOMIC), prev_port);
 			prev_port = -1;
 		}
 
 		switch (nla_type(a)) {
-		case ODP_ACTION_ATTR_OUTPUT:
+		case OVS_ACTION_ATTR_OUTPUT:
 			prev_port = nla_get_u32(a);
 			break;
 
-		case ODP_ACTION_ATTR_CONTROLLER:
-			err = output_control(dp, skb, nla_get_u64(a));
-			if (err) {
-				kfree_skb(skb);
-				return err;
-			}
+		case OVS_ACTION_ATTR_USERSPACE:
+			output_userspace(dp, skb, nla_get_u64(a));
 			break;
 
-		case ODP_ACTION_ATTR_SET_TUNNEL:
+		case OVS_ACTION_ATTR_SET_TUNNEL:
 			OVS_CB(skb)->tun_id = nla_get_be64(a);
 			break;
 
-		case ODP_ACTION_ATTR_SET_DL_TCI:
-			skb = modify_vlan_tci(skb, nla_get_be16(a));
+		case OVS_ACTION_ATTR_PUSH_VLAN:
+			err = push_vlan(skb, nla_get_be16(a));
+			if (unlikely(err)) /* skb already freed */
+				return err;
 			break;
 
-		case ODP_ACTION_ATTR_STRIP_VLAN:
-			skb = strip_vlan(skb);
+		case OVS_ACTION_ATTR_POP_VLAN:
+			err = pop_vlan(skb);
 			break;
 
-		case ODP_ACTION_ATTR_SET_DL_SRC:
-			skb = make_writable(skb, 0);
-			if (!skb)
-				return -ENOMEM;
-			memcpy(eth_hdr(skb)->h_source, nla_data(a), ETH_ALEN);
+		case OVS_ACTION_ATTR_SET_DL_SRC:
+			err = make_writable(skb, ETH_HLEN);
+			if (likely(!err))
+				memcpy(eth_hdr(skb)->h_source, nla_data(a), ETH_ALEN);
 			break;
 
-		case ODP_ACTION_ATTR_SET_DL_DST:
-			skb = make_writable(skb, 0);
-			if (!skb)
-				return -ENOMEM;
-			memcpy(eth_hdr(skb)->h_dest, nla_data(a), ETH_ALEN);
+		case OVS_ACTION_ATTR_SET_DL_DST:
+			err = make_writable(skb, ETH_HLEN);
+			if (likely(!err))
+				memcpy(eth_hdr(skb)->h_dest, nla_data(a), ETH_ALEN);
 			break;
 
-		case ODP_ACTION_ATTR_SET_NW_SRC:
-		case ODP_ACTION_ATTR_SET_NW_DST:
-			skb = set_nw_addr(skb, a);
+		case OVS_ACTION_ATTR_SET_NW_SRC:
+		case OVS_ACTION_ATTR_SET_NW_DST:
+			err = set_nw_addr(skb, a);
 			break;
 
-		case ODP_ACTION_ATTR_SET_NW_TOS:
-			skb = set_nw_tos(skb, nla_get_u8(a));
+		case OVS_ACTION_ATTR_SET_NW_TOS:
+			err = set_nw_tos(skb, nla_get_u8(a));
 			break;
 
-		case ODP_ACTION_ATTR_SET_TP_SRC:
-		case ODP_ACTION_ATTR_SET_TP_DST:
-			skb = set_tp_port(skb, a);
+		case OVS_ACTION_ATTR_SET_TP_SRC:
+		case OVS_ACTION_ATTR_SET_TP_DST:
+			err = set_tp_port(skb, a);
 			break;
 
-		case ODP_ACTION_ATTR_SET_PRIORITY:
+		case OVS_ACTION_ATTR_SET_PRIORITY:
 			skb->priority = nla_get_u32(a);
 			break;
 
-		case ODP_ACTION_ATTR_POP_PRIORITY:
+		case OVS_ACTION_ATTR_POP_PRIORITY:
 			skb->priority = priority;
 			break;
 
-		case ODP_ACTION_ATTR_DROP_SPOOFED_ARP:
-			if (unlikely(is_spoofed_arp(skb)))
-				goto exit;
+		case OVS_ACTION_ATTR_SAMPLE:
+			err = sample(dp, skb, a);
 			break;
+
+		}
+		if (unlikely(err)) {
+			kfree_skb(skb);
+			return err;
 		}
-		if (!skb)
-			return -ENOMEM;
 	}
-exit:
-	if (prev_port != -1)
+
+	if (prev_port != -1) {
+		if (keep_skb)
+			skb = skb_clone(skb, GFP_ATOMIC);
+
 		do_output(dp, skb, prev_port);
-	else
-		kfree_skb(skb);
+	} else if (!keep_skb)
+		consume_skb(skb);
+
 	return 0;
 }
 
-static void sflow_sample(struct datapath *dp, struct sk_buff *skb,
-			 struct sw_flow_actions *acts)
-{
-	struct sk_buff *nskb;
-	struct vport *p = OVS_CB(skb)->vport;
-	struct dp_upcall_info upcall;
-
-	if (unlikely(!p))
-		return;
+/* We limit the number of times that we pass into execute_actions()
+ * to avoid blowing out the stack in the event that we have a loop. */
+#define MAX_LOOPS 5
 
-	atomic_inc(&p->sflow_pool);
-	if (net_random() >= dp->sflow_probability)
-		return;
+struct loop_counter {
+	u8 count;		/* Count. */
+	bool looping;		/* Loop detected? */
+};
 
-	nskb = skb_clone(skb, GFP_ATOMIC);
-	if (unlikely(!nskb))
-		return;
+static DEFINE_PER_CPU(struct loop_counter, loop_counters);
 
-	upcall.cmd = ODP_PACKET_CMD_SAMPLE;
-	upcall.key = &OVS_CB(skb)->flow->key;
-	upcall.userdata = 0;
-	upcall.sample_pool = atomic_read(&p->sflow_pool);
-	upcall.actions = acts->actions;
-	upcall.actions_len = acts->actions_len;
-	dp_upcall(dp, nskb, &upcall);
+static int loop_suppress(struct datapath *dp, struct sw_flow_actions *actions)
+{
+	if (net_ratelimit())
+		pr_warn("%s: flow looped %d times, dropping\n",
+				dp_name(dp), MAX_LOOPS);
+	actions->actions_len = 0;
+	return -ELOOP;
 }
 
 /* Execute a list of actions against 'skb'. */
@@ -411,7 +409,7 @@ int execute_actions(struct datapath *dp, struct sk_buff *skb)
 	int error;
 
 	/* Check whether we've looped too much. */
-	loop = loop_get_counter();
+	loop = &__get_cpu_var(loop_counters);
 	if (unlikely(++loop->count > MAX_LOOPS))
 		loop->looping = true;
 	if (unlikely(loop->looping)) {
@@ -420,11 +418,9 @@ int execute_actions(struct datapath *dp, struct sk_buff *skb)
 		goto out_loop;
 	}
 
-	/* Really execute actions. */
-	if (dp->sflow_probability)
-		sflow_sample(dp, skb, acts);
 	OVS_CB(skb)->tun_id = 0;
-	error = do_execute_actions(dp, skb, acts);
+	error = do_execute_actions(dp, skb, acts->actions,
+					 acts->actions_len, false);
 
 	/* Check whether sub-actions looped too much. */
 	if (unlikely(loop->looping))
@@ -434,7 +430,6 @@ out_loop:
 	/* Decrement loop counter. */
 	if (!--loop->count)
 		loop->looping = false;
-	loop_put_counter();
 
 	return error;
 }