2 * Copyright (c) 2007-2013 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/sctp.h>
26 #include <linux/tcp.h>
27 #include <linux/udp.h>
28 #include <linux/in6.h>
29 #include <linux/if_arp.h>
30 #include <linux/if_vlan.h>
33 #include <net/checksum.h>
34 #include <net/dsfield.h>
35 #include <net/sctp/checksum.h>
40 static int do_execute_actions(struct datapath
*dp
, struct sk_buff
*skb
,
41 struct sw_flow_key
*key
,
42 const struct nlattr
*attr
, int len
);
44 static int make_writable(struct sk_buff
*skb
, int write_len
)
46 if (!pskb_may_pull(skb
, write_len
))
49 if (!skb_cloned(skb
) || skb_clone_writable(skb
, write_len
))
52 return pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
);
55 /* remove VLAN header from packet and update csum accordingly. */
56 static int __pop_vlan_tci(struct sk_buff
*skb
, __be16
*current_tci
)
58 struct vlan_hdr
*vhdr
;
61 err
= make_writable(skb
, VLAN_ETH_HLEN
);
65 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
66 skb
->csum
= csum_sub(skb
->csum
, csum_partial(skb
->data
67 + (2 * ETH_ALEN
), VLAN_HLEN
, 0));
69 vhdr
= (struct vlan_hdr
*)(skb
->data
+ ETH_HLEN
);
70 *current_tci
= vhdr
->h_vlan_TCI
;
72 memmove(skb
->data
+ VLAN_HLEN
, skb
->data
, 2 * ETH_ALEN
);
73 __skb_pull(skb
, VLAN_HLEN
);
75 vlan_set_encap_proto(skb
, vhdr
);
76 skb
->mac_header
+= VLAN_HLEN
;
77 if (skb_network_offset(skb
) < ETH_HLEN
)
78 skb_set_network_header(skb
, ETH_HLEN
);
79 skb_reset_mac_len(skb
);
84 static int pop_vlan(struct sk_buff
*skb
)
89 if (likely(vlan_tx_tag_present(skb
))) {
92 if (unlikely(skb
->protocol
!= htons(ETH_P_8021Q
) ||
93 skb
->len
< VLAN_ETH_HLEN
))
96 err
= __pop_vlan_tci(skb
, &tci
);
100 /* move next vlan tag to hw accel tag */
101 if (likely(skb
->protocol
!= htons(ETH_P_8021Q
) ||
102 skb
->len
< VLAN_ETH_HLEN
))
105 err
= __pop_vlan_tci(skb
, &tci
);
109 __vlan_hwaccel_put_tag(skb
, htons(ETH_P_8021Q
), ntohs(tci
));
113 static int push_vlan(struct sk_buff
*skb
, const struct ovs_action_push_vlan
*vlan
)
115 if (unlikely(vlan_tx_tag_present(skb
))) {
118 /* push down current VLAN tag */
119 current_tag
= vlan_tx_tag_get(skb
);
121 if (!__vlan_put_tag(skb
, skb
->vlan_proto
, current_tag
))
124 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
125 skb
->csum
= csum_add(skb
->csum
, csum_partial(skb
->data
126 + (2 * ETH_ALEN
), VLAN_HLEN
, 0));
129 __vlan_hwaccel_put_tag(skb
, vlan
->vlan_tpid
, ntohs(vlan
->vlan_tci
) & ~VLAN_TAG_PRESENT
);
133 static int set_eth_addr(struct sk_buff
*skb
,
134 const struct ovs_key_ethernet
*eth_key
)
137 err
= make_writable(skb
, ETH_HLEN
);
141 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_ALEN
* 2);
143 ether_addr_copy(eth_hdr(skb
)->h_source
, eth_key
->eth_src
);
144 ether_addr_copy(eth_hdr(skb
)->h_dest
, eth_key
->eth_dst
);
146 ovs_skb_postpush_rcsum(skb
, eth_hdr(skb
), ETH_ALEN
* 2);
151 static void set_ip_addr(struct sk_buff
*skb
, struct iphdr
*nh
,
152 __be32
*addr
, __be32 new_addr
)
154 int transport_len
= skb
->len
- skb_transport_offset(skb
);
156 if (nh
->protocol
== IPPROTO_TCP
) {
157 if (likely(transport_len
>= sizeof(struct tcphdr
)))
158 inet_proto_csum_replace4(&tcp_hdr(skb
)->check
, skb
,
160 } else if (nh
->protocol
== IPPROTO_UDP
) {
161 if (likely(transport_len
>= sizeof(struct udphdr
))) {
162 struct udphdr
*uh
= udp_hdr(skb
);
164 if (uh
->check
|| skb
->ip_summed
== CHECKSUM_PARTIAL
) {
165 inet_proto_csum_replace4(&uh
->check
, skb
,
168 uh
->check
= CSUM_MANGLED_0
;
173 csum_replace4(&nh
->check
, *addr
, new_addr
);
178 static void update_ipv6_checksum(struct sk_buff
*skb
, u8 l4_proto
,
179 __be32 addr
[4], const __be32 new_addr
[4])
181 int transport_len
= skb
->len
- skb_transport_offset(skb
);
183 if (l4_proto
== IPPROTO_TCP
) {
184 if (likely(transport_len
>= sizeof(struct tcphdr
)))
185 inet_proto_csum_replace16(&tcp_hdr(skb
)->check
, skb
,
187 } else if (l4_proto
== IPPROTO_UDP
) {
188 if (likely(transport_len
>= sizeof(struct udphdr
))) {
189 struct udphdr
*uh
= udp_hdr(skb
);
191 if (uh
->check
|| skb
->ip_summed
== CHECKSUM_PARTIAL
) {
192 inet_proto_csum_replace16(&uh
->check
, skb
,
195 uh
->check
= CSUM_MANGLED_0
;
201 static void set_ipv6_addr(struct sk_buff
*skb
, u8 l4_proto
,
202 __be32 addr
[4], const __be32 new_addr
[4],
203 bool recalculate_csum
)
205 if (recalculate_csum
)
206 update_ipv6_checksum(skb
, l4_proto
, addr
, new_addr
);
209 memcpy(addr
, new_addr
, sizeof(__be32
[4]));
212 static void set_ipv6_tc(struct ipv6hdr
*nh
, u8 tc
)
214 nh
->priority
= tc
>> 4;
215 nh
->flow_lbl
[0] = (nh
->flow_lbl
[0] & 0x0F) | ((tc
& 0x0F) << 4);
218 static void set_ipv6_fl(struct ipv6hdr
*nh
, u32 fl
)
220 nh
->flow_lbl
[0] = (nh
->flow_lbl
[0] & 0xF0) | (fl
& 0x000F0000) >> 16;
221 nh
->flow_lbl
[1] = (fl
& 0x0000FF00) >> 8;
222 nh
->flow_lbl
[2] = fl
& 0x000000FF;
225 static void set_ip_ttl(struct sk_buff
*skb
, struct iphdr
*nh
, u8 new_ttl
)
227 csum_replace2(&nh
->check
, htons(nh
->ttl
<< 8), htons(new_ttl
<< 8));
231 static int set_ipv4(struct sk_buff
*skb
, const struct ovs_key_ipv4
*ipv4_key
)
236 err
= make_writable(skb
, skb_network_offset(skb
) +
237 sizeof(struct iphdr
));
243 if (ipv4_key
->ipv4_src
!= nh
->saddr
)
244 set_ip_addr(skb
, nh
, &nh
->saddr
, ipv4_key
->ipv4_src
);
246 if (ipv4_key
->ipv4_dst
!= nh
->daddr
)
247 set_ip_addr(skb
, nh
, &nh
->daddr
, ipv4_key
->ipv4_dst
);
249 if (ipv4_key
->ipv4_tos
!= nh
->tos
)
250 ipv4_change_dsfield(nh
, 0, ipv4_key
->ipv4_tos
);
252 if (ipv4_key
->ipv4_ttl
!= nh
->ttl
)
253 set_ip_ttl(skb
, nh
, ipv4_key
->ipv4_ttl
);
258 static int set_ipv6(struct sk_buff
*skb
, const struct ovs_key_ipv6
*ipv6_key
)
265 err
= make_writable(skb
, skb_network_offset(skb
) +
266 sizeof(struct ipv6hdr
));
271 saddr
= (__be32
*)&nh
->saddr
;
272 daddr
= (__be32
*)&nh
->daddr
;
274 if (memcmp(ipv6_key
->ipv6_src
, saddr
, sizeof(ipv6_key
->ipv6_src
)))
275 set_ipv6_addr(skb
, ipv6_key
->ipv6_proto
, saddr
,
276 ipv6_key
->ipv6_src
, true);
278 if (memcmp(ipv6_key
->ipv6_dst
, daddr
, sizeof(ipv6_key
->ipv6_dst
))) {
279 unsigned int offset
= 0;
280 int flags
= IP6_FH_F_SKIP_RH
;
281 bool recalc_csum
= true;
283 if (ipv6_ext_hdr(nh
->nexthdr
))
284 recalc_csum
= ipv6_find_hdr(skb
, &offset
,
285 NEXTHDR_ROUTING
, NULL
,
286 &flags
) != NEXTHDR_ROUTING
;
288 set_ipv6_addr(skb
, ipv6_key
->ipv6_proto
, daddr
,
289 ipv6_key
->ipv6_dst
, recalc_csum
);
292 set_ipv6_tc(nh
, ipv6_key
->ipv6_tclass
);
293 set_ipv6_fl(nh
, ntohl(ipv6_key
->ipv6_label
));
294 nh
->hop_limit
= ipv6_key
->ipv6_hlimit
;
299 /* Must follow make_writable() since that can move the skb data. */
300 static void set_tp_port(struct sk_buff
*skb
, __be16
*port
,
301 __be16 new_port
, __sum16
*check
)
303 inet_proto_csum_replace2(check
, skb
, *port
, new_port
, 0);
308 static void set_udp_port(struct sk_buff
*skb
, __be16
*port
, __be16 new_port
)
310 struct udphdr
*uh
= udp_hdr(skb
);
312 if (uh
->check
&& skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
313 set_tp_port(skb
, port
, new_port
, &uh
->check
);
316 uh
->check
= CSUM_MANGLED_0
;
323 static int set_udp(struct sk_buff
*skb
, const struct ovs_key_udp
*udp_port_key
)
328 err
= make_writable(skb
, skb_transport_offset(skb
) +
329 sizeof(struct udphdr
));
334 if (udp_port_key
->udp_src
!= uh
->source
)
335 set_udp_port(skb
, &uh
->source
, udp_port_key
->udp_src
);
337 if (udp_port_key
->udp_dst
!= uh
->dest
)
338 set_udp_port(skb
, &uh
->dest
, udp_port_key
->udp_dst
);
343 static int set_tcp(struct sk_buff
*skb
, const struct ovs_key_tcp
*tcp_port_key
)
348 err
= make_writable(skb
, skb_transport_offset(skb
) +
349 sizeof(struct tcphdr
));
354 if (tcp_port_key
->tcp_src
!= th
->source
)
355 set_tp_port(skb
, &th
->source
, tcp_port_key
->tcp_src
, &th
->check
);
357 if (tcp_port_key
->tcp_dst
!= th
->dest
)
358 set_tp_port(skb
, &th
->dest
, tcp_port_key
->tcp_dst
, &th
->check
);
363 static int set_sctp(struct sk_buff
*skb
,
364 const struct ovs_key_sctp
*sctp_port_key
)
368 unsigned int sctphoff
= skb_transport_offset(skb
);
370 err
= make_writable(skb
, sctphoff
+ sizeof(struct sctphdr
));
375 if (sctp_port_key
->sctp_src
!= sh
->source
||
376 sctp_port_key
->sctp_dst
!= sh
->dest
) {
377 __le32 old_correct_csum
, new_csum
, old_csum
;
379 old_csum
= sh
->checksum
;
380 old_correct_csum
= sctp_compute_cksum(skb
, sctphoff
);
382 sh
->source
= sctp_port_key
->sctp_src
;
383 sh
->dest
= sctp_port_key
->sctp_dst
;
385 new_csum
= sctp_compute_cksum(skb
, sctphoff
);
387 /* Carry any checksum errors through. */
388 sh
->checksum
= old_csum
^ old_correct_csum
^ new_csum
;
396 static int do_output(struct datapath
*dp
, struct sk_buff
*skb
, int out_port
)
403 vport
= ovs_vport_rcu(dp
, out_port
);
404 if (unlikely(!vport
)) {
409 ovs_vport_send(vport
, skb
);
413 static int output_userspace(struct datapath
*dp
, struct sk_buff
*skb
,
414 struct sw_flow_key
*key
, const struct nlattr
*attr
)
416 struct dp_upcall_info upcall
;
417 const struct nlattr
*a
;
420 upcall
.cmd
= OVS_PACKET_CMD_ACTION
;
422 upcall
.userdata
= NULL
;
425 for (a
= nla_data(attr
), rem
= nla_len(attr
); rem
> 0;
426 a
= nla_next(a
, &rem
)) {
427 switch (nla_type(a
)) {
428 case OVS_USERSPACE_ATTR_USERDATA
:
432 case OVS_USERSPACE_ATTR_PID
:
433 upcall
.portid
= nla_get_u32(a
);
438 return ovs_dp_upcall(dp
, skb
, &upcall
);
441 static bool last_action(const struct nlattr
*a
, int rem
)
443 return a
->nla_len
== rem
;
446 static int sample(struct datapath
*dp
, struct sk_buff
*skb
,
447 struct sw_flow_key
*key
, const struct nlattr
*attr
)
449 const struct nlattr
*acts_list
= NULL
;
450 const struct nlattr
*a
;
451 struct sk_buff
*sample_skb
;
454 for (a
= nla_data(attr
), rem
= nla_len(attr
); rem
> 0;
455 a
= nla_next(a
, &rem
)) {
456 switch (nla_type(a
)) {
457 case OVS_SAMPLE_ATTR_PROBABILITY
:
458 if (prandom_u32() >= nla_get_u32(a
))
462 case OVS_SAMPLE_ATTR_ACTIONS
:
468 rem
= nla_len(acts_list
);
469 a
= nla_data(acts_list
);
471 /* Actions list is either empty or only contains a single user-space
472 * action, the latter being a special case as it is the only known
473 * usage of the sample action.
474 * In these special cases don't clone the skb as there are no
475 * side-effects in the nested actions.
476 * Otherwise, clone in case the nested actions have side effects.
478 if (likely(rem
== 0 || (nla_type(a
) == OVS_ACTION_ATTR_USERSPACE
&&
479 last_action(a
, rem
)))) {
483 sample_skb
= skb_clone(skb
, GFP_ATOMIC
);
484 if (!sample_skb
) /* Skip sample action when out of memory. */
488 /* Note that do_execute_actions() never consumes skb.
489 * In the case where skb has been cloned above it is the clone that
490 * is consumed. Otherwise the skb_get(skb) call prevents
491 * consumption by do_execute_actions(). Thus, it is safe to simply
492 * return the error code and let the caller (also
493 * do_execute_actions()) free skb on error.
495 return do_execute_actions(dp
, sample_skb
, key
, a
, rem
);
498 static int execute_set_action(struct sk_buff
*skb
,
499 const struct nlattr
*nested_attr
)
503 switch (nla_type(nested_attr
)) {
504 case OVS_KEY_ATTR_PRIORITY
:
505 skb
->priority
= nla_get_u32(nested_attr
);
508 case OVS_KEY_ATTR_SKB_MARK
:
509 skb
->mark
= nla_get_u32(nested_attr
);
512 case OVS_KEY_ATTR_IPV4_TUNNEL
:
513 OVS_CB(skb
)->egress_tun_key
= nla_data(nested_attr
);
516 case OVS_KEY_ATTR_ETHERNET
:
517 err
= set_eth_addr(skb
, nla_data(nested_attr
));
520 case OVS_KEY_ATTR_IPV4
:
521 err
= set_ipv4(skb
, nla_data(nested_attr
));
524 case OVS_KEY_ATTR_IPV6
:
525 err
= set_ipv6(skb
, nla_data(nested_attr
));
528 case OVS_KEY_ATTR_TCP
:
529 err
= set_tcp(skb
, nla_data(nested_attr
));
532 case OVS_KEY_ATTR_UDP
:
533 err
= set_udp(skb
, nla_data(nested_attr
));
536 case OVS_KEY_ATTR_SCTP
:
537 err
= set_sctp(skb
, nla_data(nested_attr
));
544 /* Execute a list of actions against 'skb'. */
545 static int do_execute_actions(struct datapath
*dp
, struct sk_buff
*skb
,
546 struct sw_flow_key
*key
,
547 const struct nlattr
*attr
, int len
)
549 /* Every output action needs a separate clone of 'skb', but the common
550 * case is just a single output action, so that doing a clone and
551 * then freeing the original skbuff is wasteful. So the following code
552 * is slightly obscure just to avoid that. */
554 const struct nlattr
*a
;
557 for (a
= attr
, rem
= len
; rem
> 0;
558 a
= nla_next(a
, &rem
)) {
561 if (prev_port
!= -1) {
562 do_output(dp
, skb_clone(skb
, GFP_ATOMIC
), prev_port
);
566 switch (nla_type(a
)) {
567 case OVS_ACTION_ATTR_OUTPUT
:
568 prev_port
= nla_get_u32(a
);
571 case OVS_ACTION_ATTR_USERSPACE
:
572 output_userspace(dp
, skb
, key
, a
);
575 case OVS_ACTION_ATTR_PUSH_VLAN
:
576 err
= push_vlan(skb
, nla_data(a
));
577 if (unlikely(err
)) /* skb already freed. */
581 case OVS_ACTION_ATTR_POP_VLAN
:
585 case OVS_ACTION_ATTR_SET
:
586 err
= execute_set_action(skb
, nla_data(a
));
589 case OVS_ACTION_ATTR_SAMPLE
:
590 err
= sample(dp
, skb
, key
, a
);
591 if (unlikely(err
)) /* skb already freed. */
603 do_output(dp
, skb
, prev_port
);
610 /* Execute a list of actions against 'skb'. */
611 int ovs_execute_actions(struct datapath
*dp
, struct sk_buff
*skb
,
612 struct sw_flow_key
*key
)
614 struct sw_flow_actions
*acts
= rcu_dereference(OVS_CB(skb
)->flow
->sf_acts
);
616 return do_execute_actions(dp
, skb
, key
,
617 acts
->actions
, acts
->actions_len
);