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>
41 static int do_execute_actions(struct datapath
*dp
, struct sk_buff
*skb
,
42 const struct nlattr
*attr
, int len
, bool keep_skb
);
44 static int make_writable(struct sk_buff
*skb
, int write_len
)
46 if (!skb_cloned(skb
) || skb_clone_writable(skb
, write_len
))
49 return pskb_expand_head(skb
, 0, 0, GFP_ATOMIC
);
52 /* remove VLAN header from packet and update csum accordingly. */
53 static int __pop_vlan_tci(struct sk_buff
*skb
, __be16
*current_tci
)
55 struct vlan_hdr
*vhdr
;
58 err
= make_writable(skb
, VLAN_ETH_HLEN
);
62 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
63 skb
->csum
= csum_sub(skb
->csum
, csum_partial(skb
->data
64 + (2 * ETH_ALEN
), VLAN_HLEN
, 0));
66 vhdr
= (struct vlan_hdr
*)(skb
->data
+ ETH_HLEN
);
67 *current_tci
= vhdr
->h_vlan_TCI
;
69 memmove(skb
->data
+ VLAN_HLEN
, skb
->data
, 2 * ETH_ALEN
);
70 __skb_pull(skb
, VLAN_HLEN
);
72 vlan_set_encap_proto(skb
, vhdr
);
73 skb
->mac_header
+= VLAN_HLEN
;
74 skb_reset_mac_len(skb
);
79 static int pop_vlan(struct sk_buff
*skb
)
84 if (likely(vlan_tx_tag_present(skb
))) {
87 if (unlikely(skb
->protocol
!= htons(ETH_P_8021Q
) ||
88 skb
->len
< VLAN_ETH_HLEN
))
91 err
= __pop_vlan_tci(skb
, &tci
);
95 /* move next vlan tag to hw accel tag */
96 if (likely(skb
->protocol
!= htons(ETH_P_8021Q
) ||
97 skb
->len
< VLAN_ETH_HLEN
))
100 err
= __pop_vlan_tci(skb
, &tci
);
104 __vlan_hwaccel_put_tag(skb
, htons(ETH_P_8021Q
), ntohs(tci
));
108 static int push_vlan(struct sk_buff
*skb
, const struct ovs_action_push_vlan
*vlan
)
110 if (unlikely(vlan_tx_tag_present(skb
))) {
113 /* push down current VLAN tag */
114 current_tag
= vlan_tx_tag_get(skb
);
116 if (!__vlan_put_tag(skb
, skb
->vlan_proto
, current_tag
))
119 if (skb
->ip_summed
== CHECKSUM_COMPLETE
)
120 skb
->csum
= csum_add(skb
->csum
, csum_partial(skb
->data
121 + (2 * ETH_ALEN
), VLAN_HLEN
, 0));
124 __vlan_hwaccel_put_tag(skb
, vlan
->vlan_tpid
, ntohs(vlan
->vlan_tci
) & ~VLAN_TAG_PRESENT
);
128 static int set_eth_addr(struct sk_buff
*skb
,
129 const struct ovs_key_ethernet
*eth_key
)
132 err
= make_writable(skb
, ETH_HLEN
);
136 skb_postpull_rcsum(skb
, eth_hdr(skb
), ETH_ALEN
* 2);
138 ether_addr_copy(eth_hdr(skb
)->h_source
, eth_key
->eth_src
);
139 ether_addr_copy(eth_hdr(skb
)->h_dest
, eth_key
->eth_dst
);
141 ovs_skb_postpush_rcsum(skb
, eth_hdr(skb
), ETH_ALEN
* 2);
146 static void set_ip_addr(struct sk_buff
*skb
, struct iphdr
*nh
,
147 __be32
*addr
, __be32 new_addr
)
149 int transport_len
= skb
->len
- skb_transport_offset(skb
);
151 if (nh
->protocol
== IPPROTO_TCP
) {
152 if (likely(transport_len
>= sizeof(struct tcphdr
)))
153 inet_proto_csum_replace4(&tcp_hdr(skb
)->check
, skb
,
155 } else if (nh
->protocol
== IPPROTO_UDP
) {
156 if (likely(transport_len
>= sizeof(struct udphdr
))) {
157 struct udphdr
*uh
= udp_hdr(skb
);
159 if (uh
->check
|| skb
->ip_summed
== CHECKSUM_PARTIAL
) {
160 inet_proto_csum_replace4(&uh
->check
, skb
,
163 uh
->check
= CSUM_MANGLED_0
;
168 csum_replace4(&nh
->check
, *addr
, new_addr
);
169 skb_clear_rxhash(skb
);
173 static void update_ipv6_checksum(struct sk_buff
*skb
, u8 l4_proto
,
174 __be32 addr
[4], const __be32 new_addr
[4])
176 int transport_len
= skb
->len
- skb_transport_offset(skb
);
178 if (l4_proto
== IPPROTO_TCP
) {
179 if (likely(transport_len
>= sizeof(struct tcphdr
)))
180 inet_proto_csum_replace16(&tcp_hdr(skb
)->check
, skb
,
182 } else if (l4_proto
== IPPROTO_UDP
) {
183 if (likely(transport_len
>= sizeof(struct udphdr
))) {
184 struct udphdr
*uh
= udp_hdr(skb
);
186 if (uh
->check
|| skb
->ip_summed
== CHECKSUM_PARTIAL
) {
187 inet_proto_csum_replace16(&uh
->check
, skb
,
190 uh
->check
= CSUM_MANGLED_0
;
196 static void set_ipv6_addr(struct sk_buff
*skb
, u8 l4_proto
,
197 __be32 addr
[4], const __be32 new_addr
[4],
198 bool recalculate_csum
)
200 if (recalculate_csum
)
201 update_ipv6_checksum(skb
, l4_proto
, addr
, new_addr
);
203 skb_clear_rxhash(skb
);
204 memcpy(addr
, new_addr
, sizeof(__be32
[4]));
207 static void set_ipv6_tc(struct ipv6hdr
*nh
, u8 tc
)
209 nh
->priority
= tc
>> 4;
210 nh
->flow_lbl
[0] = (nh
->flow_lbl
[0] & 0x0F) | ((tc
& 0x0F) << 4);
213 static void set_ipv6_fl(struct ipv6hdr
*nh
, u32 fl
)
215 nh
->flow_lbl
[0] = (nh
->flow_lbl
[0] & 0xF0) | (fl
& 0x000F0000) >> 16;
216 nh
->flow_lbl
[1] = (fl
& 0x0000FF00) >> 8;
217 nh
->flow_lbl
[2] = fl
& 0x000000FF;
220 static void set_ip_ttl(struct sk_buff
*skb
, struct iphdr
*nh
, u8 new_ttl
)
222 csum_replace2(&nh
->check
, htons(nh
->ttl
<< 8), htons(new_ttl
<< 8));
226 static int set_ipv4(struct sk_buff
*skb
, const struct ovs_key_ipv4
*ipv4_key
)
231 err
= make_writable(skb
, skb_network_offset(skb
) +
232 sizeof(struct iphdr
));
238 if (ipv4_key
->ipv4_src
!= nh
->saddr
)
239 set_ip_addr(skb
, nh
, &nh
->saddr
, ipv4_key
->ipv4_src
);
241 if (ipv4_key
->ipv4_dst
!= nh
->daddr
)
242 set_ip_addr(skb
, nh
, &nh
->daddr
, ipv4_key
->ipv4_dst
);
244 if (ipv4_key
->ipv4_tos
!= nh
->tos
)
245 ipv4_change_dsfield(nh
, 0, ipv4_key
->ipv4_tos
);
247 if (ipv4_key
->ipv4_ttl
!= nh
->ttl
)
248 set_ip_ttl(skb
, nh
, ipv4_key
->ipv4_ttl
);
253 static int set_ipv6(struct sk_buff
*skb
, const struct ovs_key_ipv6
*ipv6_key
)
260 err
= make_writable(skb
, skb_network_offset(skb
) +
261 sizeof(struct ipv6hdr
));
266 saddr
= (__be32
*)&nh
->saddr
;
267 daddr
= (__be32
*)&nh
->daddr
;
269 if (memcmp(ipv6_key
->ipv6_src
, saddr
, sizeof(ipv6_key
->ipv6_src
)))
270 set_ipv6_addr(skb
, ipv6_key
->ipv6_proto
, saddr
,
271 ipv6_key
->ipv6_src
, true);
273 if (memcmp(ipv6_key
->ipv6_dst
, daddr
, sizeof(ipv6_key
->ipv6_dst
))) {
274 unsigned int offset
= 0;
275 int flags
= OVS_IP6T_FH_F_SKIP_RH
;
276 bool recalc_csum
= true;
278 if (ipv6_ext_hdr(nh
->nexthdr
))
279 recalc_csum
= ipv6_find_hdr(skb
, &offset
,
280 NEXTHDR_ROUTING
, NULL
,
281 &flags
) != NEXTHDR_ROUTING
;
283 set_ipv6_addr(skb
, ipv6_key
->ipv6_proto
, daddr
,
284 ipv6_key
->ipv6_dst
, recalc_csum
);
287 set_ipv6_tc(nh
, ipv6_key
->ipv6_tclass
);
288 set_ipv6_fl(nh
, ntohl(ipv6_key
->ipv6_label
));
289 nh
->hop_limit
= ipv6_key
->ipv6_hlimit
;
294 /* Must follow make_writable() since that can move the skb data. */
295 static void set_tp_port(struct sk_buff
*skb
, __be16
*port
,
296 __be16 new_port
, __sum16
*check
)
298 inet_proto_csum_replace2(check
, skb
, *port
, new_port
, 0);
300 skb_clear_rxhash(skb
);
303 static void set_udp_port(struct sk_buff
*skb
, __be16
*port
, __be16 new_port
)
305 struct udphdr
*uh
= udp_hdr(skb
);
307 if (uh
->check
&& skb
->ip_summed
!= CHECKSUM_PARTIAL
) {
308 set_tp_port(skb
, port
, new_port
, &uh
->check
);
311 uh
->check
= CSUM_MANGLED_0
;
314 skb_clear_rxhash(skb
);
318 static int set_udp(struct sk_buff
*skb
, const struct ovs_key_udp
*udp_port_key
)
323 err
= make_writable(skb
, skb_transport_offset(skb
) +
324 sizeof(struct udphdr
));
329 if (udp_port_key
->udp_src
!= uh
->source
)
330 set_udp_port(skb
, &uh
->source
, udp_port_key
->udp_src
);
332 if (udp_port_key
->udp_dst
!= uh
->dest
)
333 set_udp_port(skb
, &uh
->dest
, udp_port_key
->udp_dst
);
338 static int set_tcp(struct sk_buff
*skb
, const struct ovs_key_tcp
*tcp_port_key
)
343 err
= make_writable(skb
, skb_transport_offset(skb
) +
344 sizeof(struct tcphdr
));
349 if (tcp_port_key
->tcp_src
!= th
->source
)
350 set_tp_port(skb
, &th
->source
, tcp_port_key
->tcp_src
, &th
->check
);
352 if (tcp_port_key
->tcp_dst
!= th
->dest
)
353 set_tp_port(skb
, &th
->dest
, tcp_port_key
->tcp_dst
, &th
->check
);
358 static int set_sctp(struct sk_buff
*skb
,
359 const struct ovs_key_sctp
*sctp_port_key
)
363 unsigned int sctphoff
= skb_transport_offset(skb
);
365 err
= make_writable(skb
, sctphoff
+ sizeof(struct sctphdr
));
370 if (sctp_port_key
->sctp_src
!= sh
->source
||
371 sctp_port_key
->sctp_dst
!= sh
->dest
) {
372 __le32 old_correct_csum
, new_csum
, old_csum
;
374 old_csum
= sh
->checksum
;
375 old_correct_csum
= sctp_compute_cksum(skb
, sctphoff
);
377 sh
->source
= sctp_port_key
->sctp_src
;
378 sh
->dest
= sctp_port_key
->sctp_dst
;
380 new_csum
= sctp_compute_cksum(skb
, sctphoff
);
382 /* Carry any checksum errors through. */
383 sh
->checksum
= old_csum
^ old_correct_csum
^ new_csum
;
385 skb_clear_rxhash(skb
);
391 static int do_output(struct datapath
*dp
, struct sk_buff
*skb
, int out_port
)
398 vport
= ovs_vport_rcu(dp
, out_port
);
399 if (unlikely(!vport
)) {
404 ovs_vport_send(vport
, skb
);
408 static int output_userspace(struct datapath
*dp
, struct sk_buff
*skb
,
409 const struct nlattr
*attr
)
411 struct dp_upcall_info upcall
;
412 const struct nlattr
*a
;
415 BUG_ON(!OVS_CB(skb
)->pkt_key
);
417 upcall
.cmd
= OVS_PACKET_CMD_ACTION
;
418 upcall
.key
= OVS_CB(skb
)->pkt_key
;
419 upcall
.userdata
= NULL
;
422 for (a
= nla_data(attr
), rem
= nla_len(attr
); rem
> 0;
423 a
= nla_next(a
, &rem
)) {
424 switch (nla_type(a
)) {
425 case OVS_USERSPACE_ATTR_USERDATA
:
429 case OVS_USERSPACE_ATTR_PID
:
430 upcall
.portid
= nla_get_u32(a
);
435 return ovs_dp_upcall(dp
, skb
, &upcall
);
438 static int sample(struct datapath
*dp
, struct sk_buff
*skb
,
439 const struct nlattr
*attr
)
441 const struct nlattr
*acts_list
= NULL
;
442 const struct nlattr
*a
;
445 for (a
= nla_data(attr
), rem
= nla_len(attr
); rem
> 0;
446 a
= nla_next(a
, &rem
)) {
447 switch (nla_type(a
)) {
448 case OVS_SAMPLE_ATTR_PROBABILITY
:
449 if (net_random() >= nla_get_u32(a
))
453 case OVS_SAMPLE_ATTR_ACTIONS
:
459 return do_execute_actions(dp
, skb
, nla_data(acts_list
),
460 nla_len(acts_list
), true);
463 static void execute_hash(struct sk_buff
*skb
, const struct nlattr
*attr
)
465 struct sw_flow_key
*key
= OVS_CB(skb
)->pkt_key
;
466 struct ovs_action_hash
*hash_act
= nla_data(attr
);
469 /* OVS_HASH_ALG_L4 is the only possible hash algorithm. */
470 hash
= skb_get_rxhash(skb
);
471 hash
= jhash_1word(hash
, hash_act
->hash_basis
);
475 key
->ovs_flow_hash
= hash
;
478 static int execute_set_action(struct sk_buff
*skb
,
479 const struct nlattr
*nested_attr
)
483 switch (nla_type(nested_attr
)) {
484 case OVS_KEY_ATTR_PRIORITY
:
485 skb
->priority
= nla_get_u32(nested_attr
);
488 case OVS_KEY_ATTR_SKB_MARK
:
489 skb
->mark
= nla_get_u32(nested_attr
);
492 case OVS_KEY_ATTR_IPV4_TUNNEL
:
493 OVS_CB(skb
)->tun_key
= nla_data(nested_attr
);
496 case OVS_KEY_ATTR_ETHERNET
:
497 err
= set_eth_addr(skb
, nla_data(nested_attr
));
500 case OVS_KEY_ATTR_IPV4
:
501 err
= set_ipv4(skb
, nla_data(nested_attr
));
504 case OVS_KEY_ATTR_IPV6
:
505 err
= set_ipv6(skb
, nla_data(nested_attr
));
508 case OVS_KEY_ATTR_TCP
:
509 err
= set_tcp(skb
, nla_data(nested_attr
));
512 case OVS_KEY_ATTR_UDP
:
513 err
= set_udp(skb
, nla_data(nested_attr
));
516 case OVS_KEY_ATTR_SCTP
:
517 err
= set_sctp(skb
, nla_data(nested_attr
));
524 /* Execute a list of actions against 'skb'. */
525 static int do_execute_actions(struct datapath
*dp
, struct sk_buff
*skb
,
526 const struct nlattr
*attr
, int len
, bool keep_skb
)
528 /* Every output action needs a separate clone of 'skb', but the common
529 * case is just a single output action, so that doing a clone and
530 * then freeing the original skbuff is wasteful. So the following code
531 * is slightly obscure just to avoid that. */
533 const struct nlattr
*a
;
536 for (a
= attr
, rem
= len
; rem
> 0;
537 a
= nla_next(a
, &rem
)) {
540 if (prev_port
!= -1) {
541 do_output(dp
, skb_clone(skb
, GFP_ATOMIC
), prev_port
);
545 switch (nla_type(a
)) {
546 case OVS_ACTION_ATTR_OUTPUT
:
547 prev_port
= nla_get_u32(a
);
550 case OVS_ACTION_ATTR_USERSPACE
:
551 output_userspace(dp
, skb
, a
);
554 case OVS_ACTION_ATTR_HASH
:
555 execute_hash(skb
, a
);
558 case OVS_ACTION_ATTR_PUSH_VLAN
:
559 err
= push_vlan(skb
, nla_data(a
));
560 if (unlikely(err
)) /* skb already freed. */
564 case OVS_ACTION_ATTR_POP_VLAN
:
568 case OVS_ACTION_ATTR_SET
:
569 err
= execute_set_action(skb
, nla_data(a
));
572 case OVS_ACTION_ATTR_SAMPLE
:
573 err
= sample(dp
, skb
, a
);
574 if (unlikely(err
)) /* skb already freed. */
585 if (prev_port
!= -1) {
587 skb
= skb_clone(skb
, GFP_ATOMIC
);
589 do_output(dp
, skb
, prev_port
);
590 } else if (!keep_skb
)
596 /* We limit the number of times that we pass into execute_actions()
597 * to avoid blowing out the stack in the event that we have a loop. */
600 struct loop_counter
{
601 u8 count
; /* Count. */
602 bool looping
; /* Loop detected? */
605 static DEFINE_PER_CPU(struct loop_counter
, loop_counters
);
607 static int loop_suppress(struct datapath
*dp
, struct sw_flow_actions
*actions
)
610 pr_warn("%s: flow looped %d times, dropping\n",
611 ovs_dp_name(dp
), MAX_LOOPS
);
612 actions
->actions_len
= 0;
616 /* Execute a list of actions against 'skb'. */
617 int ovs_execute_actions(struct datapath
*dp
, struct sk_buff
*skb
)
619 struct sw_flow_actions
*acts
= rcu_dereference(OVS_CB(skb
)->flow
->sf_acts
);
620 struct loop_counter
*loop
;
623 /* Check whether we've looped too much. */
624 loop
= &__get_cpu_var(loop_counters
);
625 if (unlikely(++loop
->count
> MAX_LOOPS
))
626 loop
->looping
= true;
627 if (unlikely(loop
->looping
)) {
628 error
= loop_suppress(dp
, acts
);
633 OVS_CB(skb
)->tun_key
= NULL
;
634 error
= do_execute_actions(dp
, skb
, acts
->actions
,
635 acts
->actions_len
, false);
637 /* Check whether sub-actions looped too much. */
638 if (unlikely(loop
->looping
))
639 error
= loop_suppress(dp
, acts
);
642 /* Decrement loop counter. */
644 loop
->looping
= false;