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 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_ALEN
* 2);
135 memcpy(eth_hdr(skb
)->h_source
, eth_key
->eth_src
, ETH_ALEN
);
136 memcpy(eth_hdr(skb
)->h_dest
, eth_key
->eth_dst
, ETH_ALEN
);
138 ovs_skb_postpush_rcsum(skb
, eth_hdr(skb
), ETH_ALEN
* 2);
143 static void set_ip_addr(struct sk_buff
*skb
, struct iphdr
*nh
,
144 __be32
*addr
, __be32 new_addr
)
146 int transport_len
= skb
->len
- skb_transport_offset(skb
);
148 if (nh
->protocol
== IPPROTO_TCP
) {
149 if (likely(transport_len
>= sizeof(struct tcphdr
)))
150 inet_proto_csum_replace4(&tcp_hdr(skb
)->check
, skb
,
152 } else if (nh
->protocol
== IPPROTO_UDP
) {
153 if (likely(transport_len
>= sizeof(struct udphdr
))) {
154 struct udphdr
*uh
= udp_hdr(skb
);
156 if (uh
->check
|| skb
->ip_summed
== CHECKSUM_PARTIAL
) {
157 inet_proto_csum_replace4(&uh
->check
, skb
,
160 uh
->check
= CSUM_MANGLED_0
;
165 csum_replace4(&nh
->check
, *addr
, new_addr
);
170 static void update_ipv6_checksum(struct sk_buff
*skb
, u8 l4_proto
,
171 __be32 addr
[4], const __be32 new_addr
[4])
173 int transport_len
= skb
->len
- skb_transport_offset(skb
);
175 if (l4_proto
== IPPROTO_TCP
) {
176 if (likely(transport_len
>= sizeof(struct tcphdr
)))
177 inet_proto_csum_replace16(&tcp_hdr(skb
)->check
, skb
,
179 } else if (l4_proto
== IPPROTO_UDP
) {
180 if (likely(transport_len
>= sizeof(struct udphdr
))) {
181 struct udphdr
*uh
= udp_hdr(skb
);
183 if (uh
->check
|| skb
->ip_summed
== CHECKSUM_PARTIAL
) {
184 inet_proto_csum_replace16(&uh
->check
, skb
,
187 uh
->check
= CSUM_MANGLED_0
;
193 static void set_ipv6_addr(struct sk_buff
*skb
, u8 l4_proto
,
194 __be32 addr
[4], const __be32 new_addr
[4],
195 bool recalculate_csum
)
197 if (recalculate_csum
)
198 update_ipv6_checksum(skb
, l4_proto
, addr
, new_addr
);
201 memcpy(addr
, new_addr
, sizeof(__be32
[4]));
204 static void set_ipv6_tc(struct ipv6hdr
*nh
, u8 tc
)
206 nh
->priority
= tc
>> 4;
207 nh
->flow_lbl
[0] = (nh
->flow_lbl
[0] & 0x0F) | ((tc
& 0x0F) << 4);
210 static void set_ipv6_fl(struct ipv6hdr
*nh
, u32 fl
)
212 nh
->flow_lbl
[0] = (nh
->flow_lbl
[0] & 0xF0) | (fl
& 0x000F0000) >> 16;
213 nh
->flow_lbl
[1] = (fl
& 0x0000FF00) >> 8;
214 nh
->flow_lbl
[2] = fl
& 0x000000FF;
217 static void set_ip_ttl(struct sk_buff
*skb
, struct iphdr
*nh
, u8 new_ttl
)
219 csum_replace2(&nh
->check
, htons(nh
->ttl
<< 8), htons(new_ttl
<< 8));
223 static int set_ipv4(struct sk_buff
*skb
, const struct ovs_key_ipv4
*ipv4_key
)
228 err
= make_writable(skb
, skb_network_offset(skb
) +
229 sizeof(struct iphdr
));
235 if (ipv4_key
->ipv4_src
!= nh
->saddr
)
236 set_ip_addr(skb
, nh
, &nh
->saddr
, ipv4_key
->ipv4_src
);
238 if (ipv4_key
->ipv4_dst
!= nh
->daddr
)
239 set_ip_addr(skb
, nh
, &nh
->daddr
, ipv4_key
->ipv4_dst
);
241 if (ipv4_key
->ipv4_tos
!= nh
->tos
)
242 ipv4_change_dsfield(nh
, 0, ipv4_key
->ipv4_tos
);
244 if (ipv4_key
->ipv4_ttl
!= nh
->ttl
)
245 set_ip_ttl(skb
, nh
, ipv4_key
->ipv4_ttl
);
250 static int set_ipv6(struct sk_buff
*skb
, const struct ovs_key_ipv6
*ipv6_key
)
257 err
= make_writable(skb
, skb_network_offset(skb
) +
258 sizeof(struct ipv6hdr
));
263 saddr
= (__be32
*)&nh
->saddr
;
264 daddr
= (__be32
*)&nh
->daddr
;
266 if (memcmp(ipv6_key
->ipv6_src
, saddr
, sizeof(ipv6_key
->ipv6_src
)))
267 set_ipv6_addr(skb
, ipv6_key
->ipv6_proto
, saddr
,
268 ipv6_key
->ipv6_src
, true);
270 if (memcmp(ipv6_key
->ipv6_dst
, daddr
, sizeof(ipv6_key
->ipv6_dst
))) {
271 unsigned int offset
= 0;
272 int flags
= IP6_FH_F_SKIP_RH
;
273 bool recalc_csum
= true;
275 if (ipv6_ext_hdr(nh
->nexthdr
))
276 recalc_csum
= ipv6_find_hdr(skb
, &offset
,
277 NEXTHDR_ROUTING
, NULL
,
278 &flags
) != NEXTHDR_ROUTING
;
280 set_ipv6_addr(skb
, ipv6_key
->ipv6_proto
, daddr
,
281 ipv6_key
->ipv6_dst
, recalc_csum
);
284 set_ipv6_tc(nh
, ipv6_key
->ipv6_tclass
);
285 set_ipv6_fl(nh
, ntohl(ipv6_key
->ipv6_label
));
286 nh
->hop_limit
= ipv6_key
->ipv6_hlimit
;
291 /* Must follow make_writable() since that can move the skb data. */
292 static void set_tp_port(struct sk_buff
*skb
, __be16
*port
,
293 __be16 new_port
, __sum16
*check
)
295 inet_proto_csum_replace2(check
, skb
, *port
, new_port
, 0);
300 static void set_udp_port(struct sk_buff
*skb
, __be16
*port
, __be16 new_port
)
302 struct udphdr
*uh
= udp_hdr(skb
);
304 if (uh
->check
&& skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
305 set_tp_port(skb
, port
, new_port
, &uh
->check
);
308 uh
->check
= CSUM_MANGLED_0
;
315 static int set_udp(struct sk_buff
*skb
, const struct ovs_key_udp
*udp_port_key
)
320 err
= make_writable(skb
, skb_transport_offset(skb
) +
321 sizeof(struct udphdr
));
326 if (udp_port_key
->udp_src
!= uh
->source
)
327 set_udp_port(skb
, &uh
->source
, udp_port_key
->udp_src
);
329 if (udp_port_key
->udp_dst
!= uh
->dest
)
330 set_udp_port(skb
, &uh
->dest
, udp_port_key
->udp_dst
);
335 static int set_tcp(struct sk_buff
*skb
, const struct ovs_key_tcp
*tcp_port_key
)
340 err
= make_writable(skb
, skb_transport_offset(skb
) +
341 sizeof(struct tcphdr
));
346 if (tcp_port_key
->tcp_src
!= th
->source
)
347 set_tp_port(skb
, &th
->source
, tcp_port_key
->tcp_src
, &th
->check
);
349 if (tcp_port_key
->tcp_dst
!= th
->dest
)
350 set_tp_port(skb
, &th
->dest
, tcp_port_key
->tcp_dst
, &th
->check
);
355 static int do_output(struct datapath
*dp
, struct sk_buff
*skb
, int out_port
)
362 vport
= ovs_vport_rcu(dp
, out_port
);
363 if (unlikely(!vport
)) {
368 ovs_vport_send(vport
, skb
);
372 static int output_userspace(struct datapath
*dp
, struct sk_buff
*skb
,
373 const struct nlattr
*attr
)
375 struct dp_upcall_info upcall
;
376 const struct nlattr
*a
;
379 upcall
.cmd
= OVS_PACKET_CMD_ACTION
;
380 upcall
.key
= &OVS_CB(skb
)->flow
->key
;
381 upcall
.userdata
= NULL
;
384 for (a
= nla_data(attr
), rem
= nla_len(attr
); rem
> 0;
385 a
= nla_next(a
, &rem
)) {
386 switch (nla_type(a
)) {
387 case OVS_USERSPACE_ATTR_USERDATA
:
391 case OVS_USERSPACE_ATTR_PID
:
392 upcall
.portid
= nla_get_u32(a
);
397 return ovs_dp_upcall(dp
, skb
, &upcall
);
400 static int sample(struct datapath
*dp
, struct sk_buff
*skb
,
401 const struct nlattr
*attr
)
403 const struct nlattr
*acts_list
= NULL
;
404 const struct nlattr
*a
;
407 for (a
= nla_data(attr
), rem
= nla_len(attr
); rem
> 0;
408 a
= nla_next(a
, &rem
)) {
409 switch (nla_type(a
)) {
410 case OVS_SAMPLE_ATTR_PROBABILITY
:
411 if (net_random() >= nla_get_u32(a
))
415 case OVS_SAMPLE_ATTR_ACTIONS
:
421 return do_execute_actions(dp
, skb
, nla_data(acts_list
),
422 nla_len(acts_list
), true);
425 static int execute_set_action(struct sk_buff
*skb
,
426 const struct nlattr
*nested_attr
)
430 switch (nla_type(nested_attr
)) {
431 case OVS_KEY_ATTR_PRIORITY
:
432 skb
->priority
= nla_get_u32(nested_attr
);
435 case OVS_KEY_ATTR_SKB_MARK
:
436 skb
->mark
= nla_get_u32(nested_attr
);
439 case OVS_KEY_ATTR_IPV4_TUNNEL
:
440 OVS_CB(skb
)->tun_key
= nla_data(nested_attr
);
443 case OVS_KEY_ATTR_ETHERNET
:
444 err
= set_eth_addr(skb
, nla_data(nested_attr
));
447 case OVS_KEY_ATTR_IPV4
:
448 err
= set_ipv4(skb
, nla_data(nested_attr
));
451 case OVS_KEY_ATTR_IPV6
:
452 err
= set_ipv6(skb
, nla_data(nested_attr
));
455 case OVS_KEY_ATTR_TCP
:
456 err
= set_tcp(skb
, nla_data(nested_attr
));
459 case OVS_KEY_ATTR_UDP
:
460 err
= set_udp(skb
, nla_data(nested_attr
));
467 /* Execute a list of actions against 'skb'. */
468 static int do_execute_actions(struct datapath
*dp
, struct sk_buff
*skb
,
469 const struct nlattr
*attr
, int len
, bool keep_skb
)
471 /* Every output action needs a separate clone of 'skb', but the common
472 * case is just a single output action, so that doing a clone and
473 * then freeing the original skbuff is wasteful. So the following code
474 * is slightly obscure just to avoid that. */
476 const struct nlattr
*a
;
479 for (a
= attr
, rem
= len
; rem
> 0;
480 a
= nla_next(a
, &rem
)) {
483 if (prev_port
!= -1) {
484 do_output(dp
, skb_clone(skb
, GFP_ATOMIC
), prev_port
);
488 switch (nla_type(a
)) {
489 case OVS_ACTION_ATTR_OUTPUT
:
490 prev_port
= nla_get_u32(a
);
493 case OVS_ACTION_ATTR_USERSPACE
:
494 output_userspace(dp
, skb
, a
);
497 case OVS_ACTION_ATTR_PUSH_VLAN
:
498 err
= push_vlan(skb
, nla_data(a
));
499 if (unlikely(err
)) /* skb already freed. */
503 case OVS_ACTION_ATTR_POP_VLAN
:
507 case OVS_ACTION_ATTR_SET
:
508 err
= execute_set_action(skb
, nla_data(a
));
511 case OVS_ACTION_ATTR_SAMPLE
:
512 err
= sample(dp
, skb
, a
);
522 if (prev_port
!= -1) {
524 skb
= skb_clone(skb
, GFP_ATOMIC
);
526 do_output(dp
, skb
, prev_port
);
527 } else if (!keep_skb
)
533 /* Execute a list of actions against 'skb'. */
534 int ovs_execute_actions(struct datapath
*dp
, struct sk_buff
*skb
)
536 struct sw_flow_actions
*acts
= rcu_dereference(OVS_CB(skb
)->flow
->sf_acts
);
538 OVS_CB(skb
)->tun_key
= NULL
;
539 return do_execute_actions(dp
, skb
, acts
->actions
,
540 acts
->actions_len
, false);