2 * Copyright (c) 2007-2011 Nicira, Inc.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
21 #include <linux/uaccess.h>
22 #include <linux/netdevice.h>
23 #include <linux/etherdevice.h>
24 #include <linux/if_ether.h>
25 #include <linux/if_vlan.h>
26 #include <net/llc_pdu.h>
27 #include <linux/kernel.h>
28 #include <linux/jhash.h>
29 #include <linux/jiffies.h>
30 #include <linux/llc.h>
31 #include <linux/module.h>
33 #include <linux/rcupdate.h>
34 #include <linux/if_arp.h>
36 #include <linux/ipv6.h>
37 #include <linux/tcp.h>
38 #include <linux/udp.h>
39 #include <linux/icmp.h>
40 #include <linux/icmpv6.h>
41 #include <linux/rculist.h>
44 #include <net/ndisc.h>
48 static struct kmem_cache *flow_cache;
50 static int check_header(struct sk_buff *skb, int len)
52 if (unlikely(skb->len < len))
54 if (unlikely(!pskb_may_pull(skb, len)))
59 static bool arphdr_ok(struct sk_buff *skb)
61 return pskb_may_pull(skb, skb_network_offset(skb) +
62 sizeof(struct arp_eth_header));
65 static int check_iphdr(struct sk_buff *skb)
67 unsigned int nh_ofs = skb_network_offset(skb);
71 err = check_header(skb, nh_ofs + sizeof(struct iphdr));
75 ip_len = ip_hdrlen(skb);
76 if (unlikely(ip_len < sizeof(struct iphdr) ||
77 skb->len < nh_ofs + ip_len))
80 skb_set_transport_header(skb, nh_ofs + ip_len);
84 static bool tcphdr_ok(struct sk_buff *skb)
86 int th_ofs = skb_transport_offset(skb);
89 if (unlikely(!pskb_may_pull(skb, th_ofs + sizeof(struct tcphdr))))
92 tcp_len = tcp_hdrlen(skb);
93 if (unlikely(tcp_len < sizeof(struct tcphdr) ||
94 skb->len < th_ofs + tcp_len))
100 static bool udphdr_ok(struct sk_buff *skb)
102 return pskb_may_pull(skb, skb_transport_offset(skb) +
103 sizeof(struct udphdr));
106 static bool icmphdr_ok(struct sk_buff *skb)
108 return pskb_may_pull(skb, skb_transport_offset(skb) +
109 sizeof(struct icmphdr));
112 u64 ovs_flow_used_time(unsigned long flow_jiffies)
114 struct timespec cur_ts;
117 ktime_get_ts(&cur_ts);
118 idle_ms = jiffies_to_msecs(jiffies - flow_jiffies);
119 cur_ms = (u64)cur_ts.tv_sec * MSEC_PER_SEC +
120 cur_ts.tv_nsec / NSEC_PER_MSEC;
122 return cur_ms - idle_ms;
125 #define SW_FLOW_KEY_OFFSET(field) \
126 (offsetof(struct sw_flow_key, field) + \
127 FIELD_SIZEOF(struct sw_flow_key, field))
129 static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key,
132 unsigned int nh_ofs = skb_network_offset(skb);
140 *key_lenp = SW_FLOW_KEY_OFFSET(ipv6.label);
142 err = check_header(skb, nh_ofs + sizeof(*nh));
147 nexthdr = nh->nexthdr;
148 payload_ofs = (u8 *)(nh + 1) - skb->data;
150 key->ip.proto = NEXTHDR_NONE;
151 key->ip.tos = ipv6_get_dsfield(nh);
152 key->ip.ttl = nh->hop_limit;
153 key->ipv6.label = *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL);
154 key->ipv6.addr.src = nh->saddr;
155 key->ipv6.addr.dst = nh->daddr;
157 payload_ofs = ipv6_skip_exthdr(skb, payload_ofs, &nexthdr, &frag_off);
158 if (unlikely(payload_ofs < 0))
162 if (frag_off & htons(~0x7))
163 key->ip.frag = OVS_FRAG_TYPE_LATER;
165 key->ip.frag = OVS_FRAG_TYPE_FIRST;
168 nh_len = payload_ofs - nh_ofs;
169 skb_set_transport_header(skb, nh_ofs + nh_len);
170 key->ip.proto = nexthdr;
174 static bool icmp6hdr_ok(struct sk_buff *skb)
176 return pskb_may_pull(skb, skb_transport_offset(skb) +
177 sizeof(struct icmp6hdr));
180 #define TCP_FLAGS_OFFSET 13
181 #define TCP_FLAG_MASK 0x3f
183 void ovs_flow_used(struct sw_flow *flow, struct sk_buff *skb)
187 if ((flow->key.eth.type == htons(ETH_P_IP) ||
188 flow->key.eth.type == htons(ETH_P_IPV6)) &&
189 flow->key.ip.proto == IPPROTO_TCP &&
190 likely(skb->len >= skb_transport_offset(skb) + sizeof(struct tcphdr))) {
191 u8 *tcp = (u8 *)tcp_hdr(skb);
192 tcp_flags = *(tcp + TCP_FLAGS_OFFSET) & TCP_FLAG_MASK;
195 spin_lock(&flow->lock);
196 flow->used = jiffies;
197 flow->packet_count++;
198 flow->byte_count += skb->len;
199 flow->tcp_flags |= tcp_flags;
200 spin_unlock(&flow->lock);
203 struct sw_flow_actions *ovs_flow_actions_alloc(int size)
205 struct sw_flow_actions *sfa;
207 if (size > MAX_ACTIONS_BUFSIZE)
208 return ERR_PTR(-EINVAL);
210 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
212 return ERR_PTR(-ENOMEM);
214 sfa->actions_len = 0;
218 struct sw_flow *ovs_flow_alloc(void)
220 struct sw_flow *flow;
222 flow = kmem_cache_alloc(flow_cache, GFP_KERNEL);
224 return ERR_PTR(-ENOMEM);
226 spin_lock_init(&flow->lock);
227 flow->sf_acts = NULL;
232 static struct hlist_head *find_bucket(struct flow_table *table, u32 hash)
234 hash = jhash_1word(hash, table->hash_seed);
235 return flex_array_get(table->buckets,
236 (hash & (table->n_buckets - 1)));
239 static struct flex_array *alloc_buckets(unsigned int n_buckets)
241 struct flex_array *buckets;
244 buckets = flex_array_alloc(sizeof(struct hlist_head *),
245 n_buckets, GFP_KERNEL);
249 err = flex_array_prealloc(buckets, 0, n_buckets, GFP_KERNEL);
251 flex_array_free(buckets);
255 for (i = 0; i < n_buckets; i++)
256 INIT_HLIST_HEAD((struct hlist_head *)
257 flex_array_get(buckets, i));
262 static void free_buckets(struct flex_array *buckets)
264 flex_array_free(buckets);
267 struct flow_table *ovs_flow_tbl_alloc(int new_size)
269 struct flow_table *table = kmalloc(sizeof(*table), GFP_KERNEL);
274 table->buckets = alloc_buckets(new_size);
276 if (!table->buckets) {
280 table->n_buckets = new_size;
283 table->keep_flows = false;
284 get_random_bytes(&table->hash_seed, sizeof(u32));
289 void ovs_flow_tbl_destroy(struct flow_table *table)
296 if (table->keep_flows)
299 for (i = 0; i < table->n_buckets; i++) {
300 struct sw_flow *flow;
301 struct hlist_head *head = flex_array_get(table->buckets, i);
302 struct hlist_node *node, *n;
303 int ver = table->node_ver;
305 hlist_for_each_entry_safe(flow, node, n, head, hash_node[ver]) {
306 hlist_del_rcu(&flow->hash_node[ver]);
312 free_buckets(table->buckets);
316 static void flow_tbl_destroy_rcu_cb(struct rcu_head *rcu)
318 struct flow_table *table = container_of(rcu, struct flow_table, rcu);
320 ovs_flow_tbl_destroy(table);
323 void ovs_flow_tbl_deferred_destroy(struct flow_table *table)
328 call_rcu(&table->rcu, flow_tbl_destroy_rcu_cb);
331 struct sw_flow *ovs_flow_tbl_next(struct flow_table *table, u32 *bucket, u32 *last)
333 struct sw_flow *flow;
334 struct hlist_head *head;
335 struct hlist_node *n;
339 ver = table->node_ver;
340 while (*bucket < table->n_buckets) {
342 head = flex_array_get(table->buckets, *bucket);
343 hlist_for_each_entry_rcu(flow, n, head, hash_node[ver]) {
358 static void __flow_tbl_insert(struct flow_table *table, struct sw_flow *flow)
360 struct hlist_head *head;
361 head = find_bucket(table, flow->hash);
362 hlist_add_head_rcu(&flow->hash_node[table->node_ver], head);
366 static void flow_table_copy_flows(struct flow_table *old, struct flow_table *new)
371 old_ver = old->node_ver;
372 new->node_ver = !old_ver;
374 /* Insert in new table. */
375 for (i = 0; i < old->n_buckets; i++) {
376 struct sw_flow *flow;
377 struct hlist_head *head;
378 struct hlist_node *n;
380 head = flex_array_get(old->buckets, i);
382 hlist_for_each_entry(flow, n, head, hash_node[old_ver])
383 __flow_tbl_insert(new, flow);
385 old->keep_flows = true;
388 static struct flow_table *__flow_tbl_rehash(struct flow_table *table, int n_buckets)
390 struct flow_table *new_table;
392 new_table = ovs_flow_tbl_alloc(n_buckets);
394 return ERR_PTR(-ENOMEM);
396 flow_table_copy_flows(table, new_table);
401 struct flow_table *ovs_flow_tbl_rehash(struct flow_table *table)
403 return __flow_tbl_rehash(table, table->n_buckets);
406 struct flow_table *ovs_flow_tbl_expand(struct flow_table *table)
408 return __flow_tbl_rehash(table, table->n_buckets * 2);
411 void ovs_flow_free(struct sw_flow *flow)
416 kfree((struct sf_flow_acts __force *)flow->sf_acts);
417 kmem_cache_free(flow_cache, flow);
420 /* RCU callback used by ovs_flow_deferred_free. */
421 static void rcu_free_flow_callback(struct rcu_head *rcu)
423 struct sw_flow *flow = container_of(rcu, struct sw_flow, rcu);
428 /* Schedules 'flow' to be freed after the next RCU grace period.
429 * The caller must hold rcu_read_lock for this to be sensible. */
430 void ovs_flow_deferred_free(struct sw_flow *flow)
432 call_rcu(&flow->rcu, rcu_free_flow_callback);
435 /* RCU callback used by ovs_flow_deferred_free_acts. */
436 static void rcu_free_acts_callback(struct rcu_head *rcu)
438 struct sw_flow_actions *sf_acts = container_of(rcu,
439 struct sw_flow_actions, rcu);
443 /* Schedules 'sf_acts' to be freed after the next RCU grace period.
444 * The caller must hold rcu_read_lock for this to be sensible. */
445 void ovs_flow_deferred_free_acts(struct sw_flow_actions *sf_acts)
447 call_rcu(&sf_acts->rcu, rcu_free_acts_callback);
450 static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key)
453 __be16 eth_type; /* ETH_P_8021Q */
456 struct qtag_prefix *qp;
458 if (unlikely(skb->len < sizeof(struct qtag_prefix) + sizeof(__be16)))
461 if (unlikely(!pskb_may_pull(skb, sizeof(struct qtag_prefix) +
465 qp = (struct qtag_prefix *) skb->data;
466 key->eth.tci = qp->tci | htons(VLAN_TAG_PRESENT);
467 __skb_pull(skb, sizeof(struct qtag_prefix));
472 static __be16 parse_ethertype(struct sk_buff *skb)
474 struct llc_snap_hdr {
475 u8 dsap; /* Always 0xAA */
476 u8 ssap; /* Always 0xAA */
481 struct llc_snap_hdr *llc;
484 proto = *(__be16 *) skb->data;
485 __skb_pull(skb, sizeof(__be16));
487 if (ntohs(proto) >= 1536)
490 if (skb->len < sizeof(struct llc_snap_hdr))
491 return htons(ETH_P_802_2);
493 if (unlikely(!pskb_may_pull(skb, sizeof(struct llc_snap_hdr))))
496 llc = (struct llc_snap_hdr *) skb->data;
497 if (llc->dsap != LLC_SAP_SNAP ||
498 llc->ssap != LLC_SAP_SNAP ||
499 (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0)
500 return htons(ETH_P_802_2);
502 __skb_pull(skb, sizeof(struct llc_snap_hdr));
503 return llc->ethertype;
506 static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key,
507 int *key_lenp, int nh_len)
509 struct icmp6hdr *icmp = icmp6_hdr(skb);
513 /* The ICMPv6 type and code fields use the 16-bit transport port
514 * fields, so we need to store them in 16-bit network byte order.
516 key->ipv6.tp.src = htons(icmp->icmp6_type);
517 key->ipv6.tp.dst = htons(icmp->icmp6_code);
518 key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
520 if (icmp->icmp6_code == 0 &&
521 (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION ||
522 icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) {
523 int icmp_len = skb->len - skb_transport_offset(skb);
527 key_len = SW_FLOW_KEY_OFFSET(ipv6.nd);
529 /* In order to process neighbor discovery options, we need the
532 if (unlikely(icmp_len < sizeof(*nd)))
534 if (unlikely(skb_linearize(skb))) {
539 nd = (struct nd_msg *)skb_transport_header(skb);
540 key->ipv6.nd.target = nd->target;
541 key_len = SW_FLOW_KEY_OFFSET(ipv6.nd);
543 icmp_len -= sizeof(*nd);
545 while (icmp_len >= 8) {
546 struct nd_opt_hdr *nd_opt =
547 (struct nd_opt_hdr *)(nd->opt + offset);
548 int opt_len = nd_opt->nd_opt_len * 8;
550 if (unlikely(!opt_len || opt_len > icmp_len))
553 /* Store the link layer address if the appropriate
554 * option is provided. It is considered an error if
555 * the same link layer option is specified twice.
557 if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR
559 if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll)))
561 memcpy(key->ipv6.nd.sll,
562 &nd->opt[offset+sizeof(*nd_opt)], ETH_ALEN);
563 } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR
565 if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll)))
567 memcpy(key->ipv6.nd.tll,
568 &nd->opt[offset+sizeof(*nd_opt)], ETH_ALEN);
579 memset(&key->ipv6.nd.target, 0, sizeof(key->ipv6.nd.target));
580 memset(key->ipv6.nd.sll, 0, sizeof(key->ipv6.nd.sll));
581 memset(key->ipv6.nd.tll, 0, sizeof(key->ipv6.nd.tll));
589 * ovs_flow_extract - extracts a flow key from an Ethernet frame.
590 * @skb: sk_buff that contains the frame, with skb->data pointing to the
592 * @in_port: port number on which @skb was received.
593 * @key: output flow key
594 * @key_lenp: length of output flow key
596 * The caller must ensure that skb->len >= ETH_HLEN.
598 * Returns 0 if successful, otherwise a negative errno value.
600 * Initializes @skb header pointers as follows:
602 * - skb->mac_header: the Ethernet header.
604 * - skb->network_header: just past the Ethernet header, or just past the
605 * VLAN header, to the first byte of the Ethernet payload.
607 * - skb->transport_header: If key->dl_type is ETH_P_IP or ETH_P_IPV6
608 * on output, then just past the IP header, if one is present and
609 * of a correct length, otherwise the same as skb->network_header.
610 * For other key->dl_type values it is left untouched.
612 int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key,
616 int key_len = SW_FLOW_KEY_OFFSET(eth);
619 memset(key, 0, sizeof(*key));
621 key->phy.priority = skb->priority;
622 if (OVS_CB(skb)->tun_key)
623 memcpy(&key->tun_key, OVS_CB(skb)->tun_key, sizeof(key->tun_key));
624 key->phy.in_port = in_port;
625 key->phy.skb_mark = skb_get_mark(skb);
627 skb_reset_mac_header(skb);
629 /* Link layer. We are guaranteed to have at least the 14 byte Ethernet
630 * header in the linear data area.
633 memcpy(key->eth.src, eth->h_source, ETH_ALEN);
634 memcpy(key->eth.dst, eth->h_dest, ETH_ALEN);
636 __skb_pull(skb, 2 * ETH_ALEN);
638 if (vlan_tx_tag_present(skb))
639 key->eth.tci = htons(vlan_get_tci(skb));
640 else if (eth->h_proto == htons(ETH_P_8021Q))
641 if (unlikely(parse_vlan(skb, key)))
644 key->eth.type = parse_ethertype(skb);
645 if (unlikely(key->eth.type == htons(0)))
648 skb_reset_network_header(skb);
649 __skb_push(skb, skb->data - skb_mac_header(skb));
652 if (key->eth.type == htons(ETH_P_IP)) {
656 key_len = SW_FLOW_KEY_OFFSET(ipv4.addr);
658 error = check_iphdr(skb);
659 if (unlikely(error)) {
660 if (error == -EINVAL) {
661 skb->transport_header = skb->network_header;
668 key->ipv4.addr.src = nh->saddr;
669 key->ipv4.addr.dst = nh->daddr;
671 key->ip.proto = nh->protocol;
672 key->ip.tos = nh->tos;
673 key->ip.ttl = nh->ttl;
675 offset = nh->frag_off & htons(IP_OFFSET);
677 key->ip.frag = OVS_FRAG_TYPE_LATER;
680 if (nh->frag_off & htons(IP_MF) ||
681 skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
682 key->ip.frag = OVS_FRAG_TYPE_FIRST;
684 /* Transport layer. */
685 if (key->ip.proto == IPPROTO_TCP) {
686 key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
687 if (tcphdr_ok(skb)) {
688 struct tcphdr *tcp = tcp_hdr(skb);
689 key->ipv4.tp.src = tcp->source;
690 key->ipv4.tp.dst = tcp->dest;
692 } else if (key->ip.proto == IPPROTO_UDP) {
693 key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
694 if (udphdr_ok(skb)) {
695 struct udphdr *udp = udp_hdr(skb);
696 key->ipv4.tp.src = udp->source;
697 key->ipv4.tp.dst = udp->dest;
699 } else if (key->ip.proto == IPPROTO_ICMP) {
700 key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
701 if (icmphdr_ok(skb)) {
702 struct icmphdr *icmp = icmp_hdr(skb);
703 /* The ICMP type and code fields use the 16-bit
704 * transport port fields, so we need to store
705 * them in 16-bit network byte order. */
706 key->ipv4.tp.src = htons(icmp->type);
707 key->ipv4.tp.dst = htons(icmp->code);
711 } else if ((key->eth.type == htons(ETH_P_ARP) ||
712 key->eth.type == htons(ETH_P_RARP)) && arphdr_ok(skb)) {
713 struct arp_eth_header *arp;
715 arp = (struct arp_eth_header *)skb_network_header(skb);
717 if (arp->ar_hrd == htons(ARPHRD_ETHER)
718 && arp->ar_pro == htons(ETH_P_IP)
719 && arp->ar_hln == ETH_ALEN
720 && arp->ar_pln == 4) {
722 /* We only match on the lower 8 bits of the opcode. */
723 if (ntohs(arp->ar_op) <= 0xff)
724 key->ip.proto = ntohs(arp->ar_op);
725 memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src));
726 memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst));
727 memcpy(key->ipv4.arp.sha, arp->ar_sha, ETH_ALEN);
728 memcpy(key->ipv4.arp.tha, arp->ar_tha, ETH_ALEN);
729 key_len = SW_FLOW_KEY_OFFSET(ipv4.arp);
731 } else if (key->eth.type == htons(ETH_P_IPV6)) {
732 int nh_len; /* IPv6 Header + Extensions */
734 nh_len = parse_ipv6hdr(skb, key, &key_len);
735 if (unlikely(nh_len < 0)) {
736 if (nh_len == -EINVAL)
737 skb->transport_header = skb->network_header;
743 if (key->ip.frag == OVS_FRAG_TYPE_LATER)
745 if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
746 key->ip.frag = OVS_FRAG_TYPE_FIRST;
748 /* Transport layer. */
749 if (key->ip.proto == NEXTHDR_TCP) {
750 key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
751 if (tcphdr_ok(skb)) {
752 struct tcphdr *tcp = tcp_hdr(skb);
753 key->ipv6.tp.src = tcp->source;
754 key->ipv6.tp.dst = tcp->dest;
756 } else if (key->ip.proto == NEXTHDR_UDP) {
757 key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
758 if (udphdr_ok(skb)) {
759 struct udphdr *udp = udp_hdr(skb);
760 key->ipv6.tp.src = udp->source;
761 key->ipv6.tp.dst = udp->dest;
763 } else if (key->ip.proto == NEXTHDR_ICMP) {
764 key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
765 if (icmp6hdr_ok(skb)) {
766 error = parse_icmpv6(skb, key, &key_len, nh_len);
778 static u32 ovs_flow_hash(const struct sw_flow_key *key, int key_start, int key_len)
780 return jhash2((u32 *)((u8 *)key + key_start),
781 DIV_ROUND_UP(key_len - key_start, sizeof(u32)), 0);
784 static int flow_key_start(struct sw_flow_key *key)
786 if (key->tun_key.ipv4_dst)
789 return offsetof(struct sw_flow_key, phy);
792 struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *table,
793 struct sw_flow_key *key, int key_len)
795 struct sw_flow *flow;
796 struct hlist_node *n;
797 struct hlist_head *head;
802 key_start = flow_key_start(key);
803 hash = ovs_flow_hash(key, key_start, key_len);
805 _key = (u8 *) key + key_start;
806 head = find_bucket(table, hash);
807 hlist_for_each_entry_rcu(flow, n, head, hash_node[table->node_ver]) {
809 if (flow->hash == hash &&
810 !memcmp((u8 *)&flow->key + key_start, _key, key_len - key_start)) {
817 void ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow,
818 struct sw_flow_key *key, int key_len)
820 flow->hash = ovs_flow_hash(key, flow_key_start(key), key_len);
821 memcpy(&flow->key, key, sizeof(flow->key));
822 __flow_tbl_insert(table, flow);
825 void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow)
827 hlist_del_rcu(&flow->hash_node[table->node_ver]);
829 BUG_ON(table->count < 0);
832 /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
833 const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
834 [OVS_KEY_ATTR_ENCAP] = -1,
835 [OVS_KEY_ATTR_PRIORITY] = sizeof(u32),
836 [OVS_KEY_ATTR_IN_PORT] = sizeof(u32),
837 [OVS_KEY_ATTR_SKB_MARK] = sizeof(u32),
838 [OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet),
839 [OVS_KEY_ATTR_VLAN] = sizeof(__be16),
840 [OVS_KEY_ATTR_ETHERTYPE] = sizeof(__be16),
841 [OVS_KEY_ATTR_IPV4] = sizeof(struct ovs_key_ipv4),
842 [OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6),
843 [OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp),
844 [OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp),
845 [OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp),
846 [OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6),
847 [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp),
848 [OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd),
849 [OVS_KEY_ATTR_TUNNEL] = -1,
852 [OVS_KEY_ATTR_TUN_ID] = sizeof(__be64),
855 static int ipv4_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_len,
856 const struct nlattr *a[], u64 *attrs)
858 const struct ovs_key_icmp *icmp_key;
859 const struct ovs_key_tcp *tcp_key;
860 const struct ovs_key_udp *udp_key;
862 switch (swkey->ip.proto) {
864 if (!(*attrs & (1 << OVS_KEY_ATTR_TCP)))
866 *attrs &= ~(1 << OVS_KEY_ATTR_TCP);
868 *key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
869 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
870 swkey->ipv4.tp.src = tcp_key->tcp_src;
871 swkey->ipv4.tp.dst = tcp_key->tcp_dst;
875 if (!(*attrs & (1 << OVS_KEY_ATTR_UDP)))
877 *attrs &= ~(1 << OVS_KEY_ATTR_UDP);
879 *key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
880 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
881 swkey->ipv4.tp.src = udp_key->udp_src;
882 swkey->ipv4.tp.dst = udp_key->udp_dst;
886 if (!(*attrs & (1 << OVS_KEY_ATTR_ICMP)))
888 *attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
890 *key_len = SW_FLOW_KEY_OFFSET(ipv4.tp);
891 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
892 swkey->ipv4.tp.src = htons(icmp_key->icmp_type);
893 swkey->ipv4.tp.dst = htons(icmp_key->icmp_code);
900 static int ipv6_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_len,
901 const struct nlattr *a[], u64 *attrs)
903 const struct ovs_key_icmpv6 *icmpv6_key;
904 const struct ovs_key_tcp *tcp_key;
905 const struct ovs_key_udp *udp_key;
907 switch (swkey->ip.proto) {
909 if (!(*attrs & (1 << OVS_KEY_ATTR_TCP)))
911 *attrs &= ~(1 << OVS_KEY_ATTR_TCP);
913 *key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
914 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
915 swkey->ipv6.tp.src = tcp_key->tcp_src;
916 swkey->ipv6.tp.dst = tcp_key->tcp_dst;
920 if (!(*attrs & (1 << OVS_KEY_ATTR_UDP)))
922 *attrs &= ~(1 << OVS_KEY_ATTR_UDP);
924 *key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
925 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
926 swkey->ipv6.tp.src = udp_key->udp_src;
927 swkey->ipv6.tp.dst = udp_key->udp_dst;
931 if (!(*attrs & (1 << OVS_KEY_ATTR_ICMPV6)))
933 *attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
935 *key_len = SW_FLOW_KEY_OFFSET(ipv6.tp);
936 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
937 swkey->ipv6.tp.src = htons(icmpv6_key->icmpv6_type);
938 swkey->ipv6.tp.dst = htons(icmpv6_key->icmpv6_code);
940 if (swkey->ipv6.tp.src == htons(NDISC_NEIGHBOUR_SOLICITATION) ||
941 swkey->ipv6.tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
942 const struct ovs_key_nd *nd_key;
944 if (!(*attrs & (1 << OVS_KEY_ATTR_ND)))
946 *attrs &= ~(1 << OVS_KEY_ATTR_ND);
948 *key_len = SW_FLOW_KEY_OFFSET(ipv6.nd);
949 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
950 memcpy(&swkey->ipv6.nd.target, nd_key->nd_target,
951 sizeof(swkey->ipv6.nd.target));
952 memcpy(swkey->ipv6.nd.sll, nd_key->nd_sll, ETH_ALEN);
953 memcpy(swkey->ipv6.nd.tll, nd_key->nd_tll, ETH_ALEN);
961 static int parse_flow_nlattrs(const struct nlattr *attr,
962 const struct nlattr *a[], u64 *attrsp)
964 const struct nlattr *nla;
969 nla_for_each_nested(nla, attr, rem) {
970 u16 type = nla_type(nla);
973 if (type > OVS_KEY_ATTR_MAX || attrs & (1ULL << type))
976 expected_len = ovs_key_lens[type];
977 if (nla_len(nla) != expected_len && expected_len != -1)
980 attrs |= 1ULL << type;
990 int ipv4_tun_from_nlattr(const struct nlattr *attr,
991 struct ovs_key_ipv4_tunnel *tun_key)
997 memset(tun_key, 0, sizeof(*tun_key));
999 nla_for_each_nested(a, attr, rem) {
1000 int type = nla_type(a);
1001 static const u32 ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
1002 [OVS_TUNNEL_KEY_ATTR_ID] = sizeof(u64),
1003 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = sizeof(u32),
1004 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = sizeof(u32),
1005 [OVS_TUNNEL_KEY_ATTR_TOS] = 1,
1006 [OVS_TUNNEL_KEY_ATTR_TTL] = 1,
1007 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = 0,
1008 [OVS_TUNNEL_KEY_ATTR_CSUM] = 0,
1011 if (type > OVS_TUNNEL_KEY_ATTR_MAX ||
1012 ovs_tunnel_key_lens[type] != nla_len(a))
1016 case OVS_TUNNEL_KEY_ATTR_ID:
1017 tun_key->tun_id = nla_get_be64(a);
1018 tun_key->tun_flags |= OVS_TNL_F_KEY;
1020 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
1021 tun_key->ipv4_src = nla_get_be32(a);
1023 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
1024 tun_key->ipv4_dst = nla_get_be32(a);
1026 case OVS_TUNNEL_KEY_ATTR_TOS:
1027 tun_key->ipv4_tos = nla_get_u8(a);
1029 case OVS_TUNNEL_KEY_ATTR_TTL:
1030 tun_key->ipv4_ttl = nla_get_u8(a);
1033 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
1034 tun_key->tun_flags |= OVS_TNL_F_DONT_FRAGMENT;
1036 case OVS_TUNNEL_KEY_ATTR_CSUM:
1037 tun_key->tun_flags |= OVS_TNL_F_CSUM;
1047 if (!tun_key->ipv4_dst)
1056 int ipv4_tun_to_nlattr(struct sk_buff *skb,
1057 const struct ovs_key_ipv4_tunnel *tun_key)
1061 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
1065 if (tun_key->tun_flags & OVS_TNL_F_KEY &&
1066 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, tun_key->tun_id))
1068 if (tun_key->ipv4_src &&
1069 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC, tun_key->ipv4_src))
1071 if (nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST, tun_key->ipv4_dst))
1073 if (tun_key->ipv4_tos &&
1074 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, tun_key->ipv4_tos))
1076 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, tun_key->ipv4_ttl))
1078 if ((tun_key->tun_flags & OVS_TNL_F_DONT_FRAGMENT) &&
1079 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
1081 if ((tun_key->tun_flags & OVS_TNL_F_CSUM) &&
1082 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
1085 nla_nest_end(skb, nla);
1090 * ovs_flow_from_nlattrs - parses Netlink attributes into a flow key.
1091 * @swkey: receives the extracted flow key.
1092 * @key_lenp: number of bytes used in @swkey.
1093 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1096 int ovs_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp,
1097 const struct nlattr *attr)
1099 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1100 const struct ovs_key_ethernet *eth_key;
1105 memset(swkey, 0, sizeof(struct sw_flow_key));
1106 key_len = SW_FLOW_KEY_OFFSET(eth);
1108 err = parse_flow_nlattrs(attr, a, &attrs);
1112 /* Metadata attributes. */
1113 if (attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
1114 swkey->phy.priority = nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]);
1115 attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
1117 if (attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
1118 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
1119 if (in_port >= DP_MAX_PORTS)
1121 swkey->phy.in_port = in_port;
1122 attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
1124 swkey->phy.in_port = DP_MAX_PORTS;
1126 if (attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
1127 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
1128 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,20) && !defined(CONFIG_NETFILTER)
1132 swkey->phy.skb_mark = mark;
1133 attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
1136 if (attrs & (1ULL << OVS_KEY_ATTR_TUN_ID) &&
1137 attrs & (1ULL << OVS_KEY_ATTR_TUNNEL)) {
1140 err = ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], &swkey->tun_key);
1144 if (!(swkey->tun_key.tun_flags & OVS_TNL_F_KEY))
1147 tun_id = nla_get_be64(a[OVS_KEY_ATTR_TUN_ID]);
1148 if (tun_id != swkey->tun_key.tun_id)
1151 attrs &= ~(1ULL << OVS_KEY_ATTR_TUN_ID);
1152 attrs &= ~(1ULL << OVS_KEY_ATTR_TUNNEL);
1153 } else if (attrs & (1ULL << OVS_KEY_ATTR_TUNNEL)) {
1155 err = ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], &swkey->tun_key);
1159 attrs &= ~(1ULL << OVS_KEY_ATTR_TUNNEL);
1162 /* Data attributes. */
1163 if (!(attrs & (1 << OVS_KEY_ATTR_ETHERNET)))
1165 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
1167 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
1168 memcpy(swkey->eth.src, eth_key->eth_src, ETH_ALEN);
1169 memcpy(swkey->eth.dst, eth_key->eth_dst, ETH_ALEN);
1171 if (attrs & (1u << OVS_KEY_ATTR_ETHERTYPE) &&
1172 nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q)) {
1173 const struct nlattr *encap;
1176 if (attrs != ((1 << OVS_KEY_ATTR_VLAN) |
1177 (1 << OVS_KEY_ATTR_ETHERTYPE) |
1178 (1 << OVS_KEY_ATTR_ENCAP)))
1181 encap = a[OVS_KEY_ATTR_ENCAP];
1182 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1183 if (tci & htons(VLAN_TAG_PRESENT)) {
1184 swkey->eth.tci = tci;
1186 err = parse_flow_nlattrs(encap, a, &attrs);
1190 /* Corner case for truncated 802.1Q header. */
1194 swkey->eth.type = htons(ETH_P_8021Q);
1195 *key_lenp = key_len;
1202 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1203 swkey->eth.type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1204 if (ntohs(swkey->eth.type) < 1536)
1206 attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1208 swkey->eth.type = htons(ETH_P_802_2);
1211 if (swkey->eth.type == htons(ETH_P_IP)) {
1212 const struct ovs_key_ipv4 *ipv4_key;
1214 if (!(attrs & (1 << OVS_KEY_ATTR_IPV4)))
1216 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
1218 key_len = SW_FLOW_KEY_OFFSET(ipv4.addr);
1219 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
1220 if (ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX)
1222 swkey->ip.proto = ipv4_key->ipv4_proto;
1223 swkey->ip.tos = ipv4_key->ipv4_tos;
1224 swkey->ip.ttl = ipv4_key->ipv4_ttl;
1225 swkey->ip.frag = ipv4_key->ipv4_frag;
1226 swkey->ipv4.addr.src = ipv4_key->ipv4_src;
1227 swkey->ipv4.addr.dst = ipv4_key->ipv4_dst;
1229 if (swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1230 err = ipv4_flow_from_nlattrs(swkey, &key_len, a, &attrs);
1234 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1235 const struct ovs_key_ipv6 *ipv6_key;
1237 if (!(attrs & (1 << OVS_KEY_ATTR_IPV6)))
1239 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1241 key_len = SW_FLOW_KEY_OFFSET(ipv6.label);
1242 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
1243 if (ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX)
1245 swkey->ipv6.label = ipv6_key->ipv6_label;
1246 swkey->ip.proto = ipv6_key->ipv6_proto;
1247 swkey->ip.tos = ipv6_key->ipv6_tclass;
1248 swkey->ip.ttl = ipv6_key->ipv6_hlimit;
1249 swkey->ip.frag = ipv6_key->ipv6_frag;
1250 memcpy(&swkey->ipv6.addr.src, ipv6_key->ipv6_src,
1251 sizeof(swkey->ipv6.addr.src));
1252 memcpy(&swkey->ipv6.addr.dst, ipv6_key->ipv6_dst,
1253 sizeof(swkey->ipv6.addr.dst));
1255 if (swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1256 err = ipv6_flow_from_nlattrs(swkey, &key_len, a, &attrs);
1260 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
1261 swkey->eth.type == htons(ETH_P_RARP)) {
1262 const struct ovs_key_arp *arp_key;
1264 if (!(attrs & (1 << OVS_KEY_ATTR_ARP)))
1266 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1268 key_len = SW_FLOW_KEY_OFFSET(ipv4.arp);
1269 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1270 swkey->ipv4.addr.src = arp_key->arp_sip;
1271 swkey->ipv4.addr.dst = arp_key->arp_tip;
1272 if (arp_key->arp_op & htons(0xff00))
1274 swkey->ip.proto = ntohs(arp_key->arp_op);
1275 memcpy(swkey->ipv4.arp.sha, arp_key->arp_sha, ETH_ALEN);
1276 memcpy(swkey->ipv4.arp.tha, arp_key->arp_tha, ETH_ALEN);
1281 *key_lenp = key_len;
1287 * ovs_flow_metadata_from_nlattrs - parses Netlink attributes into a flow key.
1288 * @in_port: receives the extracted input port.
1289 * @tun_id: receives the extracted tunnel ID.
1290 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1293 * This parses a series of Netlink attributes that form a flow key, which must
1294 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1295 * get the metadata, that is, the parts of the flow key that cannot be
1296 * extracted from the packet itself.
1299 int ovs_flow_metadata_from_nlattrs(struct sw_flow *flow, int key_len, const struct nlattr *attr)
1301 struct ovs_key_ipv4_tunnel *tun_key = &flow->key.tun_key;
1302 const struct nlattr *nla;
1306 flow->key.phy.in_port = DP_MAX_PORTS;
1307 flow->key.phy.priority = 0;
1308 flow->key.phy.skb_mark = 0;
1309 memset(tun_key, 0, sizeof(flow->key.tun_key));
1311 nla_for_each_nested(nla, attr, rem) {
1312 int type = nla_type(nla);
1314 if (type <= OVS_KEY_ATTR_MAX && ovs_key_lens[type] > 0) {
1317 if (nla_len(nla) != ovs_key_lens[type])
1321 case OVS_KEY_ATTR_PRIORITY:
1322 flow->key.phy.priority = nla_get_u32(nla);
1325 case OVS_KEY_ATTR_TUN_ID:
1326 tun_id = nla_get_be64(nla);
1328 if (tun_key->ipv4_dst) {
1329 if (!(tun_key->tun_flags & OVS_TNL_F_KEY))
1331 if (tun_key->tun_id != tun_id)
1335 tun_key->tun_id = tun_id;
1336 tun_key->tun_flags |= OVS_TNL_F_KEY;
1340 case OVS_KEY_ATTR_TUNNEL:
1341 if (tun_key->tun_flags & OVS_TNL_F_KEY) {
1342 tun_id = tun_key->tun_id;
1343 err = ipv4_tun_from_nlattr(nla, tun_key);
1347 if (!(tun_key->tun_flags & OVS_TNL_F_KEY))
1350 if (tun_key->tun_id != tun_id)
1353 err = ipv4_tun_from_nlattr(nla, tun_key);
1359 case OVS_KEY_ATTR_IN_PORT:
1360 if (nla_get_u32(nla) >= DP_MAX_PORTS)
1362 flow->key.phy.in_port = nla_get_u32(nla);
1365 case OVS_KEY_ATTR_SKB_MARK:
1366 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,20) && !defined(CONFIG_NETFILTER)
1367 if (nla_get_u32(nla) != 0)
1370 flow->key.phy.skb_mark = nla_get_u32(nla);
1378 flow->hash = ovs_flow_hash(&flow->key,
1379 flow_key_start(&flow->key), key_len);
1384 int ovs_flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb)
1386 struct ovs_key_ethernet *eth_key;
1387 struct nlattr *nla, *encap;
1389 if (swkey->phy.priority &&
1390 nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, swkey->phy.priority))
1391 goto nla_put_failure;
1393 if (swkey->tun_key.ipv4_dst &&
1394 ipv4_tun_to_nlattr(skb, &swkey->tun_key))
1395 goto nla_put_failure;
1397 if ((swkey->tun_key.tun_flags & OVS_TNL_F_KEY) &&
1398 nla_put_be64(skb, OVS_KEY_ATTR_TUN_ID, swkey->tun_key.tun_id))
1399 goto nla_put_failure;
1401 if (swkey->phy.in_port != DP_MAX_PORTS &&
1402 nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, swkey->phy.in_port))
1403 goto nla_put_failure;
1405 if (swkey->phy.skb_mark &&
1406 nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, swkey->phy.skb_mark))
1407 goto nla_put_failure;
1409 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1411 goto nla_put_failure;
1412 eth_key = nla_data(nla);
1413 memcpy(eth_key->eth_src, swkey->eth.src, ETH_ALEN);
1414 memcpy(eth_key->eth_dst, swkey->eth.dst, ETH_ALEN);
1416 if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
1417 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, htons(ETH_P_8021Q)) ||
1418 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, swkey->eth.tci))
1419 goto nla_put_failure;
1420 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1421 if (!swkey->eth.tci)
1427 if (swkey->eth.type == htons(ETH_P_802_2))
1430 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, swkey->eth.type))
1431 goto nla_put_failure;
1433 if (swkey->eth.type == htons(ETH_P_IP)) {
1434 struct ovs_key_ipv4 *ipv4_key;
1436 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1438 goto nla_put_failure;
1439 ipv4_key = nla_data(nla);
1440 ipv4_key->ipv4_src = swkey->ipv4.addr.src;
1441 ipv4_key->ipv4_dst = swkey->ipv4.addr.dst;
1442 ipv4_key->ipv4_proto = swkey->ip.proto;
1443 ipv4_key->ipv4_tos = swkey->ip.tos;
1444 ipv4_key->ipv4_ttl = swkey->ip.ttl;
1445 ipv4_key->ipv4_frag = swkey->ip.frag;
1446 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1447 struct ovs_key_ipv6 *ipv6_key;
1449 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1451 goto nla_put_failure;
1452 ipv6_key = nla_data(nla);
1453 memcpy(ipv6_key->ipv6_src, &swkey->ipv6.addr.src,
1454 sizeof(ipv6_key->ipv6_src));
1455 memcpy(ipv6_key->ipv6_dst, &swkey->ipv6.addr.dst,
1456 sizeof(ipv6_key->ipv6_dst));
1457 ipv6_key->ipv6_label = swkey->ipv6.label;
1458 ipv6_key->ipv6_proto = swkey->ip.proto;
1459 ipv6_key->ipv6_tclass = swkey->ip.tos;
1460 ipv6_key->ipv6_hlimit = swkey->ip.ttl;
1461 ipv6_key->ipv6_frag = swkey->ip.frag;
1462 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
1463 swkey->eth.type == htons(ETH_P_RARP)) {
1464 struct ovs_key_arp *arp_key;
1466 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1468 goto nla_put_failure;
1469 arp_key = nla_data(nla);
1470 memset(arp_key, 0, sizeof(struct ovs_key_arp));
1471 arp_key->arp_sip = swkey->ipv4.addr.src;
1472 arp_key->arp_tip = swkey->ipv4.addr.dst;
1473 arp_key->arp_op = htons(swkey->ip.proto);
1474 memcpy(arp_key->arp_sha, swkey->ipv4.arp.sha, ETH_ALEN);
1475 memcpy(arp_key->arp_tha, swkey->ipv4.arp.tha, ETH_ALEN);
1478 if ((swkey->eth.type == htons(ETH_P_IP) ||
1479 swkey->eth.type == htons(ETH_P_IPV6)) &&
1480 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1482 if (swkey->ip.proto == IPPROTO_TCP) {
1483 struct ovs_key_tcp *tcp_key;
1485 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1487 goto nla_put_failure;
1488 tcp_key = nla_data(nla);
1489 if (swkey->eth.type == htons(ETH_P_IP)) {
1490 tcp_key->tcp_src = swkey->ipv4.tp.src;
1491 tcp_key->tcp_dst = swkey->ipv4.tp.dst;
1492 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1493 tcp_key->tcp_src = swkey->ipv6.tp.src;
1494 tcp_key->tcp_dst = swkey->ipv6.tp.dst;
1496 } else if (swkey->ip.proto == IPPROTO_UDP) {
1497 struct ovs_key_udp *udp_key;
1499 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1501 goto nla_put_failure;
1502 udp_key = nla_data(nla);
1503 if (swkey->eth.type == htons(ETH_P_IP)) {
1504 udp_key->udp_src = swkey->ipv4.tp.src;
1505 udp_key->udp_dst = swkey->ipv4.tp.dst;
1506 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1507 udp_key->udp_src = swkey->ipv6.tp.src;
1508 udp_key->udp_dst = swkey->ipv6.tp.dst;
1510 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1511 swkey->ip.proto == IPPROTO_ICMP) {
1512 struct ovs_key_icmp *icmp_key;
1514 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1516 goto nla_put_failure;
1517 icmp_key = nla_data(nla);
1518 icmp_key->icmp_type = ntohs(swkey->ipv4.tp.src);
1519 icmp_key->icmp_code = ntohs(swkey->ipv4.tp.dst);
1520 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1521 swkey->ip.proto == IPPROTO_ICMPV6) {
1522 struct ovs_key_icmpv6 *icmpv6_key;
1524 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1525 sizeof(*icmpv6_key));
1527 goto nla_put_failure;
1528 icmpv6_key = nla_data(nla);
1529 icmpv6_key->icmpv6_type = ntohs(swkey->ipv6.tp.src);
1530 icmpv6_key->icmpv6_code = ntohs(swkey->ipv6.tp.dst);
1532 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1533 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1534 struct ovs_key_nd *nd_key;
1536 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1538 goto nla_put_failure;
1539 nd_key = nla_data(nla);
1540 memcpy(nd_key->nd_target, &swkey->ipv6.nd.target,
1541 sizeof(nd_key->nd_target));
1542 memcpy(nd_key->nd_sll, swkey->ipv6.nd.sll, ETH_ALEN);
1543 memcpy(nd_key->nd_tll, swkey->ipv6.nd.tll, ETH_ALEN);
1550 nla_nest_end(skb, encap);
1558 /* Initializes the flow module.
1559 * Returns zero if successful or a negative error code. */
1560 int ovs_flow_init(void)
1562 flow_cache = kmem_cache_create("sw_flow", sizeof(struct sw_flow), 0,
1564 if (flow_cache == NULL)
1570 /* Uninitializes the flow module. */
1571 void ovs_flow_exit(void)
1573 kmem_cache_destroy(flow_cache);