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>
42 static int do_execute_actions(struct datapath
*dp
, struct sk_buff
*skb
,
43 const struct nlattr
*attr
, int len
, bool keep_skb
);
45 static int make_writable(struct sk_buff
*skb
, int write_len
)
47 if (!skb_cloned(skb
) || skb_clone_writable(skb
, write_len
))
50 return pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
);
53 /* remove VLAN header from packet and update csum accordingly. */
54 static int __pop_vlan_tci(struct sk_buff
*skb
, __be16
*current_tci
)
56 struct vlan_hdr
*vhdr
;
59 err
= make_writable(skb
, VLAN_ETH_HLEN
);
63 if (get_ip_summed(skb
) == OVS_CSUM_COMPLETE
)
64 skb
->csum
= csum_sub(skb
->csum
, csum_partial(skb
->data
65 + (2 * ETH_ALEN
), VLAN_HLEN
, 0));
67 vhdr
= (struct vlan_hdr
*)(skb
->data
+ ETH_HLEN
);
68 *current_tci
= vhdr
->h_vlan_TCI
;
70 memmove(skb
->data
+ VLAN_HLEN
, skb
->data
, 2 * ETH_ALEN
);
71 __skb_pull(skb
, VLAN_HLEN
);
73 vlan_set_encap_proto(skb
, vhdr
);
74 skb
->mac_header
+= VLAN_HLEN
;
75 skb_reset_mac_len(skb
);
80 static int pop_vlan(struct sk_buff
*skb
)
85 if (likely(vlan_tx_tag_present(skb
))) {
88 if (unlikely(skb
->protocol
!= htons(ETH_P_8021Q
) ||
89 skb
->len
< VLAN_ETH_HLEN
))
92 err
= __pop_vlan_tci(skb
, &tci
);
96 /* move next vlan tag to hw accel tag */
97 if (likely(skb
->protocol
!= htons(ETH_P_8021Q
) ||
98 skb
->len
< VLAN_ETH_HLEN
))
101 err
= __pop_vlan_tci(skb
, &tci
);
105 __vlan_hwaccel_put_tag(skb
, htons(ETH_P_8021Q
), ntohs(tci
));
109 static int push_vlan(struct sk_buff
*skb
, const struct ovs_action_push_vlan
*vlan
)
111 if (unlikely(vlan_tx_tag_present(skb
))) {
114 /* push down current VLAN tag */
115 current_tag
= vlan_tx_tag_get(skb
);
117 if (!__vlan_put_tag(skb
, skb
->vlan_proto
, current_tag
))
120 if (get_ip_summed(skb
) == OVS_CSUM_COMPLETE
)
121 skb
->csum
= csum_add(skb
->csum
, csum_partial(skb
->data
122 + (2 * ETH_ALEN
), VLAN_HLEN
, 0));
125 __vlan_hwaccel_put_tag(skb
, vlan
->vlan_tpid
, ntohs(vlan
->vlan_tci
) & ~VLAN_TAG_PRESENT
);
129 static int set_eth_addr(struct sk_buff
*skb
,
130 const struct ovs_key_ethernet
*eth_key
)
133 err
= make_writable(skb
, ETH_HLEN
);
137 if (get_ip_summed(skb
) == OVS_CSUM_COMPLETE
)
138 skb
->csum
= csum_sub(skb
->csum
, csum_partial(eth_hdr(skb
),
141 memcpy(eth_hdr(skb
)->h_source
, eth_key
->eth_src
, ETH_ALEN
);
142 memcpy(eth_hdr(skb
)->h_dest
, eth_key
->eth_dst
, ETH_ALEN
);
144 if (get_ip_summed(skb
) == OVS_CSUM_COMPLETE
)
145 skb
->csum
= csum_add(skb
->csum
, csum_partial(eth_hdr(skb
),
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
);
165 get_ip_summed(skb
) == OVS_CSUM_PARTIAL
) {
166 inet_proto_csum_replace4(&uh
->check
, skb
,
169 uh
->check
= CSUM_MANGLED_0
;
174 csum_replace4(&nh
->check
, *addr
, new_addr
);
175 skb_clear_rxhash(skb
);
179 static void update_ipv6_checksum(struct sk_buff
*skb
, u8 l4_proto
,
180 __be32 addr
[4], const __be32 new_addr
[4])
182 int transport_len
= skb
->len
- skb_transport_offset(skb
);
184 if (l4_proto
== IPPROTO_TCP
) {
185 if (likely(transport_len
>= sizeof(struct tcphdr
)))
186 inet_proto_csum_replace16(&tcp_hdr(skb
)->check
, skb
,
188 } else if (l4_proto
== IPPROTO_UDP
) {
189 if (likely(transport_len
>= sizeof(struct udphdr
))) {
190 struct udphdr
*uh
= udp_hdr(skb
);
193 get_ip_summed(skb
) == OVS_CSUM_PARTIAL
) {
194 inet_proto_csum_replace16(&uh
->check
, skb
,
197 uh
->check
= CSUM_MANGLED_0
;
203 static void set_ipv6_addr(struct sk_buff
*skb
, u8 l4_proto
,
204 __be32 addr
[4], const __be32 new_addr
[4],
205 bool recalculate_csum
)
207 if (recalculate_csum
)
208 update_ipv6_checksum(skb
, l4_proto
, addr
, new_addr
);
210 skb_clear_rxhash(skb
);
211 memcpy(addr
, new_addr
, sizeof(__be32
[4]));
214 static void set_ipv6_tc(struct ipv6hdr
*nh
, u8 tc
)
216 nh
->priority
= tc
>> 4;
217 nh
->flow_lbl
[0] = (nh
->flow_lbl
[0] & 0x0F) | ((tc
& 0x0F) << 4);
220 static void set_ipv6_fl(struct ipv6hdr
*nh
, u32 fl
)
222 nh
->flow_lbl
[0] = (nh
->flow_lbl
[0] & 0xF0) | (fl
& 0x000F0000) >> 16;
223 nh
->flow_lbl
[1] = (fl
& 0x0000FF00) >> 8;
224 nh
->flow_lbl
[2] = fl
& 0x000000FF;
227 static void set_ip_ttl(struct sk_buff
*skb
, struct iphdr
*nh
, u8 new_ttl
)
229 csum_replace2(&nh
->check
, htons(nh
->ttl
<< 8), htons(new_ttl
<< 8));
233 static int set_ipv4(struct sk_buff
*skb
, const struct ovs_key_ipv4
*ipv4_key
)
238 err
= make_writable(skb
, skb_network_offset(skb
) +
239 sizeof(struct iphdr
));
245 if (ipv4_key
->ipv4_src
!= nh
->saddr
)
246 set_ip_addr(skb
, nh
, &nh
->saddr
, ipv4_key
->ipv4_src
);
248 if (ipv4_key
->ipv4_dst
!= nh
->daddr
)
249 set_ip_addr(skb
, nh
, &nh
->daddr
, ipv4_key
->ipv4_dst
);
251 if (ipv4_key
->ipv4_tos
!= nh
->tos
)
252 ipv4_change_dsfield(nh
, 0, ipv4_key
->ipv4_tos
);
254 if (ipv4_key
->ipv4_ttl
!= nh
->ttl
)
255 set_ip_ttl(skb
, nh
, ipv4_key
->ipv4_ttl
);
260 static int set_ipv6(struct sk_buff
*skb
, const struct ovs_key_ipv6
*ipv6_key
)
267 err
= make_writable(skb
, skb_network_offset(skb
) +
268 sizeof(struct ipv6hdr
));
273 saddr
= (__be32
*)&nh
->saddr
;
274 daddr
= (__be32
*)&nh
->daddr
;
276 if (memcmp(ipv6_key
->ipv6_src
, saddr
, sizeof(ipv6_key
->ipv6_src
)))
277 set_ipv6_addr(skb
, ipv6_key
->ipv6_proto
, saddr
,
278 ipv6_key
->ipv6_src
, true);
280 if (memcmp(ipv6_key
->ipv6_dst
, daddr
, sizeof(ipv6_key
->ipv6_dst
))) {
281 unsigned int offset
= 0;
282 int flags
= OVS_IP6T_FH_F_SKIP_RH
;
283 bool recalc_csum
= true;
285 if (ipv6_ext_hdr(nh
->nexthdr
))
286 recalc_csum
= ipv6_find_hdr(skb
, &offset
,
287 NEXTHDR_ROUTING
, NULL
,
288 &flags
) != NEXTHDR_ROUTING
;
290 set_ipv6_addr(skb
, ipv6_key
->ipv6_proto
, daddr
,
291 ipv6_key
->ipv6_dst
, recalc_csum
);
294 set_ipv6_tc(nh
, ipv6_key
->ipv6_tclass
);
295 set_ipv6_fl(nh
, ntohl(ipv6_key
->ipv6_label
));
296 nh
->hop_limit
= ipv6_key
->ipv6_hlimit
;
301 /* Must follow make_writable() since that can move the skb data. */
302 static void set_tp_port(struct sk_buff
*skb
, __be16
*port
,
303 __be16 new_port
, __sum16
*check
)
305 inet_proto_csum_replace2(check
, skb
, *port
, new_port
, 0);
307 skb_clear_rxhash(skb
);
310 static void set_udp_port(struct sk_buff
*skb
, __be16
*port
, __be16 new_port
)
312 struct udphdr
*uh
= udp_hdr(skb
);
314 if (uh
->check
&& get_ip_summed(skb
) != OVS_CSUM_PARTIAL
) {
315 set_tp_port(skb
, port
, new_port
, &uh
->check
);
318 uh
->check
= CSUM_MANGLED_0
;
321 skb_clear_rxhash(skb
);
325 static int set_udp(struct sk_buff
*skb
, const struct ovs_key_udp
*udp_port_key
)
330 err
= make_writable(skb
, skb_transport_offset(skb
) +
331 sizeof(struct udphdr
));
336 if (udp_port_key
->udp_src
!= uh
->source
)
337 set_udp_port(skb
, &uh
->source
, udp_port_key
->udp_src
);
339 if (udp_port_key
->udp_dst
!= uh
->dest
)
340 set_udp_port(skb
, &uh
->dest
, udp_port_key
->udp_dst
);
345 static int set_tcp(struct sk_buff
*skb
, const struct ovs_key_tcp
*tcp_port_key
)
350 err
= make_writable(skb
, skb_transport_offset(skb
) +
351 sizeof(struct tcphdr
));
356 if (tcp_port_key
->tcp_src
!= th
->source
)
357 set_tp_port(skb
, &th
->source
, tcp_port_key
->tcp_src
, &th
->check
);
359 if (tcp_port_key
->tcp_dst
!= th
->dest
)
360 set_tp_port(skb
, &th
->dest
, tcp_port_key
->tcp_dst
, &th
->check
);
365 static int set_sctp(struct sk_buff
*skb
,
366 const struct ovs_key_sctp
*sctp_port_key
)
370 unsigned int sctphoff
= skb_transport_offset(skb
);
372 err
= make_writable(skb
, sctphoff
+ sizeof(struct sctphdr
));
377 if (sctp_port_key
->sctp_src
!= sh
->source
||
378 sctp_port_key
->sctp_dst
!= sh
->dest
) {
379 __le32 old_correct_csum
, new_csum
, old_csum
;
381 old_csum
= sh
->checksum
;
382 old_correct_csum
= sctp_compute_cksum(skb
, sctphoff
);
384 sh
->source
= sctp_port_key
->sctp_src
;
385 sh
->dest
= sctp_port_key
->sctp_dst
;
387 new_csum
= sctp_compute_cksum(skb
, sctphoff
);
389 /* Carry any checksum errors through. */
390 sh
->checksum
= old_csum
^ old_correct_csum
^ new_csum
;
392 skb_clear_rxhash(skb
);
398 static int do_output(struct datapath
*dp
, struct sk_buff
*skb
, int out_port
)
405 vport
= ovs_vport_rcu(dp
, out_port
);
406 if (unlikely(!vport
)) {
411 ovs_vport_send(vport
, skb
);
415 static int output_userspace(struct datapath
*dp
, struct sk_buff
*skb
,
416 const struct nlattr
*attr
)
418 struct dp_upcall_info upcall
;
419 const struct nlattr
*a
;
422 BUG_ON(!OVS_CB(skb
)->pkt_key
);
424 upcall
.cmd
= OVS_PACKET_CMD_ACTION
;
425 upcall
.key
= OVS_CB(skb
)->pkt_key
;
426 upcall
.userdata
= NULL
;
429 for (a
= nla_data(attr
), rem
= nla_len(attr
); rem
> 0;
430 a
= nla_next(a
, &rem
)) {
431 switch (nla_type(a
)) {
432 case OVS_USERSPACE_ATTR_USERDATA
:
436 case OVS_USERSPACE_ATTR_PID
:
437 upcall
.portid
= nla_get_u32(a
);
442 return ovs_dp_upcall(dp
, skb
, &upcall
);
445 static int sample(struct datapath
*dp
, struct sk_buff
*skb
,
446 const struct nlattr
*attr
)
448 const struct nlattr
*acts_list
= NULL
;
449 const struct nlattr
*a
;
452 for (a
= nla_data(attr
), rem
= nla_len(attr
); rem
> 0;
453 a
= nla_next(a
, &rem
)) {
454 switch (nla_type(a
)) {
455 case OVS_SAMPLE_ATTR_PROBABILITY
:
456 if (net_random() >= nla_get_u32(a
))
460 case OVS_SAMPLE_ATTR_ACTIONS
:
466 return do_execute_actions(dp
, skb
, nla_data(acts_list
),
467 nla_len(acts_list
), true);
470 static int execute_set_action(struct sk_buff
*skb
,
471 const struct nlattr
*nested_attr
)
475 switch (nla_type(nested_attr
)) {
476 case OVS_KEY_ATTR_PRIORITY
:
477 skb
->priority
= nla_get_u32(nested_attr
);
480 case OVS_KEY_ATTR_SKB_MARK
:
481 skb
->mark
= nla_get_u32(nested_attr
);
484 case OVS_KEY_ATTR_IPV4_TUNNEL
:
485 OVS_CB(skb
)->tun_key
= nla_data(nested_attr
);
488 case OVS_KEY_ATTR_ETHERNET
:
489 err
= set_eth_addr(skb
, nla_data(nested_attr
));
492 case OVS_KEY_ATTR_IPV4
:
493 err
= set_ipv4(skb
, nla_data(nested_attr
));
496 case OVS_KEY_ATTR_IPV6
:
497 err
= set_ipv6(skb
, nla_data(nested_attr
));
500 case OVS_KEY_ATTR_TCP
:
501 err
= set_tcp(skb
, nla_data(nested_attr
));
504 case OVS_KEY_ATTR_UDP
:
505 err
= set_udp(skb
, nla_data(nested_attr
));
508 case OVS_KEY_ATTR_SCTP
:
509 err
= set_sctp(skb
, nla_data(nested_attr
));
516 /* Execute a list of actions against 'skb'. */
517 static int do_execute_actions(struct datapath
*dp
, struct sk_buff
*skb
,
518 const struct nlattr
*attr
, int len
, bool keep_skb
)
520 /* Every output action needs a separate clone of 'skb', but the common
521 * case is just a single output action, so that doing a clone and
522 * then freeing the original skbuff is wasteful. So the following code
523 * is slightly obscure just to avoid that. */
525 const struct nlattr
*a
;
528 for (a
= attr
, rem
= len
; rem
> 0;
529 a
= nla_next(a
, &rem
)) {
532 if (prev_port
!= -1) {
533 do_output(dp
, skb_clone(skb
, GFP_ATOMIC
), prev_port
);
537 switch (nla_type(a
)) {
538 case OVS_ACTION_ATTR_OUTPUT
:
539 prev_port
= nla_get_u32(a
);
542 case OVS_ACTION_ATTR_USERSPACE
:
543 output_userspace(dp
, skb
, a
);
546 case OVS_ACTION_ATTR_PUSH_VLAN
:
547 err
= push_vlan(skb
, nla_data(a
));
548 if (unlikely(err
)) /* skb already freed. */
552 case OVS_ACTION_ATTR_POP_VLAN
:
556 case OVS_ACTION_ATTR_SET
:
557 err
= execute_set_action(skb
, nla_data(a
));
560 case OVS_ACTION_ATTR_SAMPLE
:
561 err
= sample(dp
, skb
, a
);
571 if (prev_port
!= -1) {
573 skb
= skb_clone(skb
, GFP_ATOMIC
);
575 do_output(dp
, skb
, prev_port
);
576 } else if (!keep_skb
)
582 /* We limit the number of times that we pass into execute_actions()
583 * to avoid blowing out the stack in the event that we have a loop. */
586 struct loop_counter
{
587 u8 count
; /* Count. */
588 bool looping
; /* Loop detected? */
591 static DEFINE_PER_CPU(struct loop_counter
, loop_counters
);
593 static int loop_suppress(struct datapath
*dp
, struct sw_flow_actions
*actions
)
596 pr_warn("%s: flow looped %d times, dropping\n",
597 ovs_dp_name(dp
), MAX_LOOPS
);
598 actions
->actions_len
= 0;
602 /* Execute a list of actions against 'skb'. */
603 int ovs_execute_actions(struct datapath
*dp
, struct sk_buff
*skb
)
605 struct sw_flow_actions
*acts
= rcu_dereference(OVS_CB(skb
)->flow
->sf_acts
);
606 struct loop_counter
*loop
;
609 /* Check whether we've looped too much. */
610 loop
= &__get_cpu_var(loop_counters
);
611 if (unlikely(++loop
->count
> MAX_LOOPS
))
612 loop
->looping
= true;
613 if (unlikely(loop
->looping
)) {
614 error
= loop_suppress(dp
, acts
);
619 OVS_CB(skb
)->tun_key
= NULL
;
620 error
= do_execute_actions(dp
, skb
, acts
->actions
,
621 acts
->actions_len
, false);
623 /* Check whether sub-actions looped too much. */
624 if (unlikely(loop
->looping
))
625 error
= loop_suppress(dp
, acts
);
628 /* Decrement loop counter. */
630 loop
->looping
= false;