2 * Copyright (c) 2007-2012 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
19 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
21 #include <linux/skbuff.h>
24 #include <linux/openvswitch.h>
25 #include <linux/tcp.h>
26 #include <linux/udp.h>
27 #include <linux/in6.h>
28 #include <linux/if_arp.h>
29 #include <linux/if_vlan.h>
32 #include <net/checksum.h>
33 #include <net/dsfield.h>
38 static int do_execute_actions(struct datapath *dp, struct sk_buff *skb,
39 const struct nlattr *attr, int len, bool keep_skb);
41 static int make_writable(struct sk_buff *skb, int write_len)
43 if (!skb_cloned(skb) || skb_clone_writable(skb, write_len))
46 return pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
49 /* remove VLAN header from packet and update csum accordingly. */
50 static int __pop_vlan_tci(struct sk_buff *skb, __be16 *current_tci)
52 struct vlan_hdr *vhdr;
55 err = make_writable(skb, VLAN_ETH_HLEN);
59 if (skb->ip_summed == CHECKSUM_COMPLETE)
60 skb->csum = csum_sub(skb->csum, csum_partial(skb->data
61 + (2 * ETH_ALEN), VLAN_HLEN, 0));
63 vhdr = (struct vlan_hdr *)(skb->data + ETH_HLEN);
64 *current_tci = vhdr->h_vlan_TCI;
66 memmove(skb->data + VLAN_HLEN, skb->data, 2 * ETH_ALEN);
67 __skb_pull(skb, VLAN_HLEN);
69 vlan_set_encap_proto(skb, vhdr);
70 skb->mac_header += VLAN_HLEN;
71 skb_reset_mac_len(skb);
76 static int pop_vlan(struct sk_buff *skb)
81 if (likely(vlan_tx_tag_present(skb))) {
84 if (unlikely(skb->protocol != htons(ETH_P_8021Q) ||
85 skb->len < VLAN_ETH_HLEN))
88 err = __pop_vlan_tci(skb, &tci);
92 /* move next vlan tag to hw accel tag */
93 if (likely(skb->protocol != htons(ETH_P_8021Q) ||
94 skb->len < VLAN_ETH_HLEN))
97 err = __pop_vlan_tci(skb, &tci);
101 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), ntohs(tci));
105 static int push_vlan(struct sk_buff *skb, const struct ovs_action_push_vlan *vlan)
107 if (unlikely(vlan_tx_tag_present(skb))) {
110 /* push down current VLAN tag */
111 current_tag = vlan_tx_tag_get(skb);
113 if (!__vlan_put_tag(skb, skb->vlan_proto, current_tag))
116 if (skb->ip_summed == CHECKSUM_COMPLETE)
117 skb->csum = csum_add(skb->csum, csum_partial(skb->data
118 + (2 * ETH_ALEN), VLAN_HLEN, 0));
121 __vlan_hwaccel_put_tag(skb, vlan->vlan_tpid, ntohs(vlan->vlan_tci) & ~VLAN_TAG_PRESENT);
125 static int set_eth_addr(struct sk_buff *skb,
126 const struct ovs_key_ethernet *eth_key)
129 err = make_writable(skb, ETH_HLEN);
133 memcpy(eth_hdr(skb)->h_source, eth_key->eth_src, ETH_ALEN);
134 memcpy(eth_hdr(skb)->h_dest, eth_key->eth_dst, ETH_ALEN);
139 static void set_ip_addr(struct sk_buff *skb, struct iphdr *nh,
140 __be32 *addr, __be32 new_addr)
142 int transport_len = skb->len - skb_transport_offset(skb);
144 if (nh->protocol == IPPROTO_TCP) {
145 if (likely(transport_len >= sizeof(struct tcphdr)))
146 inet_proto_csum_replace4(&tcp_hdr(skb)->check, skb,
148 } else if (nh->protocol == IPPROTO_UDP) {
149 if (likely(transport_len >= sizeof(struct udphdr))) {
150 struct udphdr *uh = udp_hdr(skb);
152 if (uh->check || skb->ip_summed == CHECKSUM_PARTIAL) {
153 inet_proto_csum_replace4(&uh->check, skb,
156 uh->check = CSUM_MANGLED_0;
161 csum_replace4(&nh->check, *addr, new_addr);
166 static void update_ipv6_checksum(struct sk_buff *skb, u8 l4_proto,
167 __be32 addr[4], const __be32 new_addr[4])
169 int transport_len = skb->len - skb_transport_offset(skb);
171 if (l4_proto == IPPROTO_TCP) {
172 if (likely(transport_len >= sizeof(struct tcphdr)))
173 inet_proto_csum_replace16(&tcp_hdr(skb)->check, skb,
175 } else if (l4_proto == IPPROTO_UDP) {
176 if (likely(transport_len >= sizeof(struct udphdr))) {
177 struct udphdr *uh = udp_hdr(skb);
179 if (uh->check || skb->ip_summed == CHECKSUM_PARTIAL) {
180 inet_proto_csum_replace16(&uh->check, skb,
183 uh->check = CSUM_MANGLED_0;
189 static void set_ipv6_addr(struct sk_buff *skb, u8 l4_proto,
190 __be32 addr[4], const __be32 new_addr[4],
191 bool recalculate_csum)
193 if (recalculate_csum)
194 update_ipv6_checksum(skb, l4_proto, addr, new_addr);
197 memcpy(addr, new_addr, sizeof(__be32[4]));
200 static void set_ipv6_tc(struct ipv6hdr *nh, u8 tc)
202 nh->priority = tc >> 4;
203 nh->flow_lbl[0] = (nh->flow_lbl[0] & 0x0F) | ((tc & 0x0F) << 4);
206 static void set_ipv6_fl(struct ipv6hdr *nh, u32 fl)
208 nh->flow_lbl[0] = (nh->flow_lbl[0] & 0xF0) | (fl & 0x000F0000) >> 16;
209 nh->flow_lbl[1] = (fl & 0x0000FF00) >> 8;
210 nh->flow_lbl[2] = fl & 0x000000FF;
213 static void set_ip_ttl(struct sk_buff *skb, struct iphdr *nh, u8 new_ttl)
215 csum_replace2(&nh->check, htons(nh->ttl << 8), htons(new_ttl << 8));
219 static int set_ipv4(struct sk_buff *skb, const struct ovs_key_ipv4 *ipv4_key)
224 err = make_writable(skb, skb_network_offset(skb) +
225 sizeof(struct iphdr));
231 if (ipv4_key->ipv4_src != nh->saddr)
232 set_ip_addr(skb, nh, &nh->saddr, ipv4_key->ipv4_src);
234 if (ipv4_key->ipv4_dst != nh->daddr)
235 set_ip_addr(skb, nh, &nh->daddr, ipv4_key->ipv4_dst);
237 if (ipv4_key->ipv4_tos != nh->tos)
238 ipv4_change_dsfield(nh, 0, ipv4_key->ipv4_tos);
240 if (ipv4_key->ipv4_ttl != nh->ttl)
241 set_ip_ttl(skb, nh, ipv4_key->ipv4_ttl);
246 static int set_ipv6(struct sk_buff *skb, const struct ovs_key_ipv6 *ipv6_key)
253 err = make_writable(skb, skb_network_offset(skb) +
254 sizeof(struct ipv6hdr));
259 saddr = (__be32 *)&nh->saddr;
260 daddr = (__be32 *)&nh->daddr;
262 if (memcmp(ipv6_key->ipv6_src, saddr, sizeof(ipv6_key->ipv6_src)))
263 set_ipv6_addr(skb, ipv6_key->ipv6_proto, saddr,
264 ipv6_key->ipv6_src, true);
266 if (memcmp(ipv6_key->ipv6_dst, daddr, sizeof(ipv6_key->ipv6_dst))) {
267 unsigned int offset = 0;
268 int flags = IP6_FH_F_SKIP_RH;
269 bool recalc_csum = true;
271 if (ipv6_ext_hdr(nh->nexthdr))
272 recalc_csum = ipv6_find_hdr(skb, &offset,
273 NEXTHDR_ROUTING, NULL,
274 &flags) != NEXTHDR_ROUTING;
276 set_ipv6_addr(skb, ipv6_key->ipv6_proto, daddr,
277 ipv6_key->ipv6_dst, recalc_csum);
280 set_ipv6_tc(nh, ipv6_key->ipv6_tclass);
281 set_ipv6_fl(nh, ntohl(ipv6_key->ipv6_label));
282 nh->hop_limit = ipv6_key->ipv6_hlimit;
287 /* Must follow make_writable() since that can move the skb data. */
288 static void set_tp_port(struct sk_buff *skb, __be16 *port,
289 __be16 new_port, __sum16 *check)
291 inet_proto_csum_replace2(check, skb, *port, new_port, 0);
296 static void set_udp_port(struct sk_buff *skb, __be16 *port, __be16 new_port)
298 struct udphdr *uh = udp_hdr(skb);
300 if (uh->check && skb->ip_summed != CHECKSUM_PARTIAL) {
301 set_tp_port(skb, port, new_port, &uh->check);
304 uh->check = CSUM_MANGLED_0;
311 static int set_udp(struct sk_buff *skb, const struct ovs_key_udp *udp_port_key)
316 err = make_writable(skb, skb_transport_offset(skb) +
317 sizeof(struct udphdr));
322 if (udp_port_key->udp_src != uh->source)
323 set_udp_port(skb, &uh->source, udp_port_key->udp_src);
325 if (udp_port_key->udp_dst != uh->dest)
326 set_udp_port(skb, &uh->dest, udp_port_key->udp_dst);
331 static int set_tcp(struct sk_buff *skb, const struct ovs_key_tcp *tcp_port_key)
336 err = make_writable(skb, skb_transport_offset(skb) +
337 sizeof(struct tcphdr));
342 if (tcp_port_key->tcp_src != th->source)
343 set_tp_port(skb, &th->source, tcp_port_key->tcp_src, &th->check);
345 if (tcp_port_key->tcp_dst != th->dest)
346 set_tp_port(skb, &th->dest, tcp_port_key->tcp_dst, &th->check);
351 static int do_output(struct datapath *dp, struct sk_buff *skb, int out_port)
358 vport = ovs_vport_rcu(dp, out_port);
359 if (unlikely(!vport)) {
364 ovs_vport_send(vport, skb);
368 static int output_userspace(struct datapath *dp, struct sk_buff *skb,
369 const struct nlattr *attr)
371 struct dp_upcall_info upcall;
372 const struct nlattr *a;
375 upcall.cmd = OVS_PACKET_CMD_ACTION;
376 upcall.key = &OVS_CB(skb)->flow->key;
377 upcall.userdata = NULL;
380 for (a = nla_data(attr), rem = nla_len(attr); rem > 0;
381 a = nla_next(a, &rem)) {
382 switch (nla_type(a)) {
383 case OVS_USERSPACE_ATTR_USERDATA:
387 case OVS_USERSPACE_ATTR_PID:
388 upcall.portid = nla_get_u32(a);
393 return ovs_dp_upcall(dp, skb, &upcall);
396 static int sample(struct datapath *dp, struct sk_buff *skb,
397 const struct nlattr *attr)
399 const struct nlattr *acts_list = NULL;
400 const struct nlattr *a;
403 for (a = nla_data(attr), rem = nla_len(attr); rem > 0;
404 a = nla_next(a, &rem)) {
405 switch (nla_type(a)) {
406 case OVS_SAMPLE_ATTR_PROBABILITY:
407 if (net_random() >= nla_get_u32(a))
411 case OVS_SAMPLE_ATTR_ACTIONS:
417 return do_execute_actions(dp, skb, nla_data(acts_list),
418 nla_len(acts_list), true);
421 static int execute_set_action(struct sk_buff *skb,
422 const struct nlattr *nested_attr)
426 switch (nla_type(nested_attr)) {
427 case OVS_KEY_ATTR_PRIORITY:
428 skb->priority = nla_get_u32(nested_attr);
431 case OVS_KEY_ATTR_SKB_MARK:
432 skb->mark = nla_get_u32(nested_attr);
435 case OVS_KEY_ATTR_ETHERNET:
436 err = set_eth_addr(skb, nla_data(nested_attr));
439 case OVS_KEY_ATTR_IPV4:
440 err = set_ipv4(skb, nla_data(nested_attr));
443 case OVS_KEY_ATTR_IPV6:
444 err = set_ipv6(skb, nla_data(nested_attr));
447 case OVS_KEY_ATTR_TCP:
448 err = set_tcp(skb, nla_data(nested_attr));
451 case OVS_KEY_ATTR_UDP:
452 err = set_udp(skb, nla_data(nested_attr));
459 /* Execute a list of actions against 'skb'. */
460 static int do_execute_actions(struct datapath *dp, struct sk_buff *skb,
461 const struct nlattr *attr, int len, bool keep_skb)
463 /* Every output action needs a separate clone of 'skb', but the common
464 * case is just a single output action, so that doing a clone and
465 * then freeing the original skbuff is wasteful. So the following code
466 * is slightly obscure just to avoid that. */
468 const struct nlattr *a;
471 for (a = attr, rem = len; rem > 0;
472 a = nla_next(a, &rem)) {
475 if (prev_port != -1) {
476 do_output(dp, skb_clone(skb, GFP_ATOMIC), prev_port);
480 switch (nla_type(a)) {
481 case OVS_ACTION_ATTR_OUTPUT:
482 prev_port = nla_get_u32(a);
485 case OVS_ACTION_ATTR_USERSPACE:
486 output_userspace(dp, skb, a);
489 case OVS_ACTION_ATTR_PUSH_VLAN:
490 err = push_vlan(skb, nla_data(a));
491 if (unlikely(err)) /* skb already freed. */
495 case OVS_ACTION_ATTR_POP_VLAN:
499 case OVS_ACTION_ATTR_SET:
500 err = execute_set_action(skb, nla_data(a));
503 case OVS_ACTION_ATTR_SAMPLE:
504 err = sample(dp, skb, a);
514 if (prev_port != -1) {
516 skb = skb_clone(skb, GFP_ATOMIC);
518 do_output(dp, skb, prev_port);
519 } else if (!keep_skb)
525 /* Execute a list of actions against 'skb'. */
526 int ovs_execute_actions(struct datapath *dp, struct sk_buff *skb)
528 struct sw_flow_actions *acts = rcu_dereference(OVS_CB(skb)->flow->sf_acts);
530 return do_execute_actions(dp, skb, acts->actions,
531 acts->actions_len, false);