2 * Copyright (c) 2007-2014 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 #include <linux/uaccess.h>
20 #include <linux/netdevice.h>
21 #include <linux/etherdevice.h>
22 #include <linux/if_ether.h>
23 #include <linux/if_vlan.h>
24 #include <net/llc_pdu.h>
25 #include <linux/kernel.h>
26 #include <linux/jhash.h>
27 #include <linux/jiffies.h>
28 #include <linux/llc.h>
29 #include <linux/module.h>
31 #include <linux/rcupdate.h>
32 #include <linux/cpumask.h>
33 #include <linux/if_arp.h>
35 #include <linux/ipv6.h>
36 #include <linux/mpls.h>
37 #include <linux/sctp.h>
38 #include <linux/smp.h>
39 #include <linux/tcp.h>
40 #include <linux/udp.h>
41 #include <linux/icmp.h>
42 #include <linux/icmpv6.h>
43 #include <linux/rculist.h>
45 #include <net/ip_tunnels.h>
48 #include <net/ndisc.h>
50 #include "conntrack.h"
53 #include "flow_netlink.h"
56 u64
ovs_flow_used_time(unsigned long flow_jiffies
)
58 struct timespec cur_ts
;
61 ktime_get_ts(&cur_ts
);
62 idle_ms
= jiffies_to_msecs(jiffies
- flow_jiffies
);
63 cur_ms
= (u64
)cur_ts
.tv_sec
* MSEC_PER_SEC
+
64 cur_ts
.tv_nsec
/ NSEC_PER_MSEC
;
66 return cur_ms
- idle_ms
;
69 #define TCP_FLAGS_BE16(tp) (*(__be16 *)&tcp_flag_word(tp) & htons(0x0FFF))
71 void ovs_flow_stats_update(struct sw_flow
*flow
, __be16 tcp_flags
,
72 const struct sk_buff
*skb
)
74 struct flow_stats
*stats
;
75 int node
= numa_node_id();
76 int cpu
= smp_processor_id();
77 int len
= skb
->len
+ (skb_vlan_tag_present(skb
) ? VLAN_HLEN
: 0);
79 stats
= rcu_dereference(flow
->stats
[cpu
]);
81 /* Check if already have CPU-specific stats. */
83 spin_lock(&stats
->lock
);
84 /* Mark if we write on the pre-allocated stats. */
85 if (cpu
== 0 && unlikely(flow
->stats_last_writer
!= cpu
))
86 flow
->stats_last_writer
= cpu
;
88 stats
= rcu_dereference(flow
->stats
[0]); /* Pre-allocated. */
89 spin_lock(&stats
->lock
);
91 /* If the current CPU is the only writer on the
92 * pre-allocated stats keep using them.
94 if (unlikely(flow
->stats_last_writer
!= cpu
)) {
95 /* A previous locker may have already allocated the
96 * stats, so we need to check again. If CPU-specific
97 * stats were already allocated, we update the pre-
98 * allocated stats as we have already locked them.
100 if (likely(flow
->stats_last_writer
!= -1) &&
101 likely(!rcu_access_pointer(flow
->stats
[cpu
]))) {
102 /* Try to allocate CPU-specific stats. */
103 struct flow_stats
*new_stats
;
106 kmem_cache_alloc_node(flow_stats_cache
,
112 if (likely(new_stats
)) {
113 new_stats
->used
= jiffies
;
114 new_stats
->packet_count
= 1;
115 new_stats
->byte_count
= len
;
116 new_stats
->tcp_flags
= tcp_flags
;
117 spin_lock_init(&new_stats
->lock
);
119 rcu_assign_pointer(flow
->stats
[cpu
],
124 flow
->stats_last_writer
= cpu
;
128 stats
->used
= jiffies
;
129 stats
->packet_count
++;
130 stats
->byte_count
+= len
;
131 stats
->tcp_flags
|= tcp_flags
;
133 spin_unlock(&stats
->lock
);
136 /* Must be called with rcu_read_lock or ovs_mutex. */
137 void ovs_flow_stats_get(const struct sw_flow
*flow
,
138 struct ovs_flow_stats
*ovs_stats
,
139 unsigned long *used
, __be16
*tcp_flags
)
145 memset(ovs_stats
, 0, sizeof(*ovs_stats
));
147 /* We open code this to make sure cpu 0 is always considered */
148 for (cpu
= 0; cpu
< nr_cpu_ids
; cpu
= cpumask_next(cpu
, cpu_possible_mask
)) {
149 struct flow_stats
*stats
= rcu_dereference_ovsl(flow
->stats
[cpu
]);
152 /* Local CPU may write on non-local stats, so we must
153 * block bottom-halves here.
155 spin_lock_bh(&stats
->lock
);
156 if (!*used
|| time_after(stats
->used
, *used
))
158 *tcp_flags
|= stats
->tcp_flags
;
159 ovs_stats
->n_packets
+= stats
->packet_count
;
160 ovs_stats
->n_bytes
+= stats
->byte_count
;
161 spin_unlock_bh(&stats
->lock
);
166 /* Called with ovs_mutex. */
167 void ovs_flow_stats_clear(struct sw_flow
*flow
)
171 /* We open code this to make sure cpu 0 is always considered */
172 for (cpu
= 0; cpu
< nr_cpu_ids
; cpu
= cpumask_next(cpu
, cpu_possible_mask
)) {
173 struct flow_stats
*stats
= ovsl_dereference(flow
->stats
[cpu
]);
176 spin_lock_bh(&stats
->lock
);
178 stats
->packet_count
= 0;
179 stats
->byte_count
= 0;
180 stats
->tcp_flags
= 0;
181 spin_unlock_bh(&stats
->lock
);
186 static int check_header(struct sk_buff
*skb
, int len
)
188 if (unlikely(skb
->len
< len
))
190 if (unlikely(!pskb_may_pull(skb
, len
)))
195 static bool arphdr_ok(struct sk_buff
*skb
)
197 return pskb_may_pull(skb
, skb_network_offset(skb
) +
198 sizeof(struct arp_eth_header
));
201 static int check_iphdr(struct sk_buff
*skb
)
203 unsigned int nh_ofs
= skb_network_offset(skb
);
207 err
= check_header(skb
, nh_ofs
+ sizeof(struct iphdr
));
211 ip_len
= ip_hdrlen(skb
);
212 if (unlikely(ip_len
< sizeof(struct iphdr
) ||
213 skb
->len
< nh_ofs
+ ip_len
))
216 skb_set_transport_header(skb
, nh_ofs
+ ip_len
);
220 static bool tcphdr_ok(struct sk_buff
*skb
)
222 int th_ofs
= skb_transport_offset(skb
);
225 if (unlikely(!pskb_may_pull(skb
, th_ofs
+ sizeof(struct tcphdr
))))
228 tcp_len
= tcp_hdrlen(skb
);
229 if (unlikely(tcp_len
< sizeof(struct tcphdr
) ||
230 skb
->len
< th_ofs
+ tcp_len
))
236 static bool udphdr_ok(struct sk_buff
*skb
)
238 return pskb_may_pull(skb
, skb_transport_offset(skb
) +
239 sizeof(struct udphdr
));
242 static bool sctphdr_ok(struct sk_buff
*skb
)
244 return pskb_may_pull(skb
, skb_transport_offset(skb
) +
245 sizeof(struct sctphdr
));
248 static bool icmphdr_ok(struct sk_buff
*skb
)
250 return pskb_may_pull(skb
, skb_transport_offset(skb
) +
251 sizeof(struct icmphdr
));
254 static int parse_ipv6hdr(struct sk_buff
*skb
, struct sw_flow_key
*key
)
256 unsigned int nh_ofs
= skb_network_offset(skb
);
264 err
= check_header(skb
, nh_ofs
+ sizeof(*nh
));
269 nexthdr
= nh
->nexthdr
;
270 payload_ofs
= (u8
*)(nh
+ 1) - skb
->data
;
272 key
->ip
.proto
= NEXTHDR_NONE
;
273 key
->ip
.tos
= ipv6_get_dsfield(nh
);
274 key
->ip
.ttl
= nh
->hop_limit
;
275 key
->ipv6
.label
= *(__be32
*)nh
& htonl(IPV6_FLOWINFO_FLOWLABEL
);
276 key
->ipv6
.addr
.src
= nh
->saddr
;
277 key
->ipv6
.addr
.dst
= nh
->daddr
;
279 payload_ofs
= ipv6_skip_exthdr(skb
, payload_ofs
, &nexthdr
, &frag_off
);
282 if (frag_off
& htons(~0x7))
283 key
->ip
.frag
= OVS_FRAG_TYPE_LATER
;
285 key
->ip
.frag
= OVS_FRAG_TYPE_FIRST
;
287 key
->ip
.frag
= OVS_FRAG_TYPE_NONE
;
290 /* Delayed handling of error in ipv6_skip_exthdr() as it
291 * always sets frag_off to a valid value which may be
292 * used to set key->ip.frag above.
294 if (unlikely(payload_ofs
< 0))
297 nh_len
= payload_ofs
- nh_ofs
;
298 skb_set_transport_header(skb
, nh_ofs
+ nh_len
);
299 key
->ip
.proto
= nexthdr
;
303 static bool icmp6hdr_ok(struct sk_buff
*skb
)
305 return pskb_may_pull(skb
, skb_transport_offset(skb
) +
306 sizeof(struct icmp6hdr
));
310 * Parse vlan tag from vlan header.
311 * Returns ERROR on memory error.
312 * Returns 0 if it encounters a non-vlan or incomplete packet.
313 * Returns 1 after successfully parsing vlan tag.
315 static int parse_vlan_tag(struct sk_buff
*skb
, struct vlan_head
*key_vh
)
317 struct vlan_head
*vh
= (struct vlan_head
*)skb
->data
;
319 if (likely(!eth_type_vlan(vh
->tpid
)))
322 if (unlikely(skb
->len
< sizeof(struct vlan_head
) + sizeof(__be16
)))
325 if (unlikely(!pskb_may_pull(skb
, sizeof(struct vlan_head
) +
329 vh
= (struct vlan_head
*)skb
->data
;
330 key_vh
->tci
= vh
->tci
| htons(VLAN_TAG_PRESENT
);
331 key_vh
->tpid
= vh
->tpid
;
333 __skb_pull(skb
, sizeof(struct vlan_head
));
337 static void clear_vlan(struct sw_flow_key
*key
)
339 key
->eth
.vlan
.tci
= 0;
340 key
->eth
.vlan
.tpid
= 0;
341 key
->eth
.cvlan
.tci
= 0;
342 key
->eth
.cvlan
.tpid
= 0;
345 static int parse_vlan(struct sk_buff
*skb
, struct sw_flow_key
*key
)
349 if (skb_vlan_tag_present(skb
)) {
350 key
->eth
.vlan
.tci
= htons(skb
->vlan_tci
);
351 key
->eth
.vlan
.tpid
= skb
->vlan_proto
;
353 /* Parse outer vlan tag in the non-accelerated case. */
354 res
= parse_vlan_tag(skb
, &key
->eth
.vlan
);
359 /* Parse inner vlan tag. */
360 res
= parse_vlan_tag(skb
, &key
->eth
.cvlan
);
367 static __be16
parse_ethertype(struct sk_buff
*skb
)
369 struct llc_snap_hdr
{
370 u8 dsap
; /* Always 0xAA */
371 u8 ssap
; /* Always 0xAA */
376 struct llc_snap_hdr
*llc
;
379 proto
= *(__be16
*) skb
->data
;
380 __skb_pull(skb
, sizeof(__be16
));
382 if (eth_proto_is_802_3(proto
))
385 if (skb
->len
< sizeof(struct llc_snap_hdr
))
386 return htons(ETH_P_802_2
);
388 if (unlikely(!pskb_may_pull(skb
, sizeof(struct llc_snap_hdr
))))
391 llc
= (struct llc_snap_hdr
*) skb
->data
;
392 if (llc
->dsap
!= LLC_SAP_SNAP
||
393 llc
->ssap
!= LLC_SAP_SNAP
||
394 (llc
->oui
[0] | llc
->oui
[1] | llc
->oui
[2]) != 0)
395 return htons(ETH_P_802_2
);
397 __skb_pull(skb
, sizeof(struct llc_snap_hdr
));
399 if (eth_proto_is_802_3(llc
->ethertype
))
400 return llc
->ethertype
;
402 return htons(ETH_P_802_2
);
405 static int parse_icmpv6(struct sk_buff
*skb
, struct sw_flow_key
*key
,
408 struct icmp6hdr
*icmp
= icmp6_hdr(skb
);
410 /* The ICMPv6 type and code fields use the 16-bit transport port
411 * fields, so we need to store them in 16-bit network byte order.
413 key
->tp
.src
= htons(icmp
->icmp6_type
);
414 key
->tp
.dst
= htons(icmp
->icmp6_code
);
415 memset(&key
->ipv6
.nd
, 0, sizeof(key
->ipv6
.nd
));
417 if (icmp
->icmp6_code
== 0 &&
418 (icmp
->icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
||
419 icmp
->icmp6_type
== NDISC_NEIGHBOUR_ADVERTISEMENT
)) {
420 int icmp_len
= skb
->len
- skb_transport_offset(skb
);
424 /* In order to process neighbor discovery options, we need the
427 if (unlikely(icmp_len
< sizeof(*nd
)))
430 if (unlikely(skb_linearize(skb
)))
433 nd
= (struct nd_msg
*)skb_transport_header(skb
);
434 key
->ipv6
.nd
.target
= nd
->target
;
436 icmp_len
-= sizeof(*nd
);
438 while (icmp_len
>= 8) {
439 struct nd_opt_hdr
*nd_opt
=
440 (struct nd_opt_hdr
*)(nd
->opt
+ offset
);
441 int opt_len
= nd_opt
->nd_opt_len
* 8;
443 if (unlikely(!opt_len
|| opt_len
> icmp_len
))
446 /* Store the link layer address if the appropriate
447 * option is provided. It is considered an error if
448 * the same link layer option is specified twice.
450 if (nd_opt
->nd_opt_type
== ND_OPT_SOURCE_LL_ADDR
452 if (unlikely(!is_zero_ether_addr(key
->ipv6
.nd
.sll
)))
454 ether_addr_copy(key
->ipv6
.nd
.sll
,
455 &nd
->opt
[offset
+sizeof(*nd_opt
)]);
456 } else if (nd_opt
->nd_opt_type
== ND_OPT_TARGET_LL_ADDR
458 if (unlikely(!is_zero_ether_addr(key
->ipv6
.nd
.tll
)))
460 ether_addr_copy(key
->ipv6
.nd
.tll
,
461 &nd
->opt
[offset
+sizeof(*nd_opt
)]);
472 memset(&key
->ipv6
.nd
.target
, 0, sizeof(key
->ipv6
.nd
.target
));
473 memset(key
->ipv6
.nd
.sll
, 0, sizeof(key
->ipv6
.nd
.sll
));
474 memset(key
->ipv6
.nd
.tll
, 0, sizeof(key
->ipv6
.nd
.tll
));
480 * key_extract - extracts a flow key from an Ethernet frame.
481 * @skb: sk_buff that contains the frame, with skb->data pointing to the
483 * @key: output flow key
485 * The caller must ensure that skb->len >= ETH_HLEN.
487 * Returns 0 if successful, otherwise a negative errno value.
489 * Initializes @skb header fields as follows:
491 * - skb->mac_header: the L2 header.
493 * - skb->network_header: just past the L2 header, or just past the
494 * VLAN header, to the first byte of the L2 payload.
496 * - skb->transport_header: If key->eth.type is ETH_P_IP or ETH_P_IPV6
497 * on output, then just past the IP header, if one is present and
498 * of a correct length, otherwise the same as skb->network_header.
499 * For other key->eth.type values it is left untouched.
501 * - skb->protocol: the type of the data starting at skb->network_header.
502 * Equals to key->eth.type.
504 static int key_extract(struct sk_buff
*skb
, struct sw_flow_key
*key
)
509 /* Flags are always used as part of stats */
512 skb_reset_mac_header(skb
);
516 if (key
->mac_proto
== MAC_PROTO_NONE
) {
517 if (unlikely(eth_type_vlan(skb
->protocol
)))
520 skb_reset_network_header(skb
);
523 ether_addr_copy(key
->eth
.src
, eth
->h_source
);
524 ether_addr_copy(key
->eth
.dst
, eth
->h_dest
);
526 __skb_pull(skb
, 2 * ETH_ALEN
);
527 /* We are going to push all headers that we pull, so no need to
528 * update skb->csum here.
531 if (unlikely(parse_vlan(skb
, key
)))
534 skb
->protocol
= parse_ethertype(skb
);
535 if (unlikely(skb
->protocol
== htons(0)))
538 skb_reset_network_header(skb
);
539 __skb_push(skb
, skb
->data
- skb_mac_header(skb
));
541 skb_reset_mac_len(skb
);
542 key
->eth
.type
= skb
->protocol
;
545 if (key
->eth
.type
== htons(ETH_P_IP
)) {
549 error
= check_iphdr(skb
);
550 if (unlikely(error
)) {
551 memset(&key
->ip
, 0, sizeof(key
->ip
));
552 memset(&key
->ipv4
, 0, sizeof(key
->ipv4
));
553 if (error
== -EINVAL
) {
554 skb
->transport_header
= skb
->network_header
;
561 key
->ipv4
.addr
.src
= nh
->saddr
;
562 key
->ipv4
.addr
.dst
= nh
->daddr
;
564 key
->ip
.proto
= nh
->protocol
;
565 key
->ip
.tos
= nh
->tos
;
566 key
->ip
.ttl
= nh
->ttl
;
568 offset
= nh
->frag_off
& htons(IP_OFFSET
);
570 key
->ip
.frag
= OVS_FRAG_TYPE_LATER
;
573 if (nh
->frag_off
& htons(IP_MF
) ||
574 skb_shinfo(skb
)->gso_type
& SKB_GSO_UDP
)
575 key
->ip
.frag
= OVS_FRAG_TYPE_FIRST
;
577 key
->ip
.frag
= OVS_FRAG_TYPE_NONE
;
579 /* Transport layer. */
580 if (key
->ip
.proto
== IPPROTO_TCP
) {
581 if (tcphdr_ok(skb
)) {
582 struct tcphdr
*tcp
= tcp_hdr(skb
);
583 key
->tp
.src
= tcp
->source
;
584 key
->tp
.dst
= tcp
->dest
;
585 key
->tp
.flags
= TCP_FLAGS_BE16(tcp
);
587 memset(&key
->tp
, 0, sizeof(key
->tp
));
590 } else if (key
->ip
.proto
== IPPROTO_UDP
) {
591 if (udphdr_ok(skb
)) {
592 struct udphdr
*udp
= udp_hdr(skb
);
593 key
->tp
.src
= udp
->source
;
594 key
->tp
.dst
= udp
->dest
;
596 memset(&key
->tp
, 0, sizeof(key
->tp
));
598 } else if (key
->ip
.proto
== IPPROTO_SCTP
) {
599 if (sctphdr_ok(skb
)) {
600 struct sctphdr
*sctp
= sctp_hdr(skb
);
601 key
->tp
.src
= sctp
->source
;
602 key
->tp
.dst
= sctp
->dest
;
604 memset(&key
->tp
, 0, sizeof(key
->tp
));
606 } else if (key
->ip
.proto
== IPPROTO_ICMP
) {
607 if (icmphdr_ok(skb
)) {
608 struct icmphdr
*icmp
= icmp_hdr(skb
);
609 /* The ICMP type and code fields use the 16-bit
610 * transport port fields, so we need to store
611 * them in 16-bit network byte order. */
612 key
->tp
.src
= htons(icmp
->type
);
613 key
->tp
.dst
= htons(icmp
->code
);
615 memset(&key
->tp
, 0, sizeof(key
->tp
));
619 } else if (key
->eth
.type
== htons(ETH_P_ARP
) ||
620 key
->eth
.type
== htons(ETH_P_RARP
)) {
621 struct arp_eth_header
*arp
;
622 bool arp_available
= arphdr_ok(skb
);
624 arp
= (struct arp_eth_header
*)skb_network_header(skb
);
627 arp
->ar_hrd
== htons(ARPHRD_ETHER
) &&
628 arp
->ar_pro
== htons(ETH_P_IP
) &&
629 arp
->ar_hln
== ETH_ALEN
&&
632 /* We only match on the lower 8 bits of the opcode. */
633 if (ntohs(arp
->ar_op
) <= 0xff)
634 key
->ip
.proto
= ntohs(arp
->ar_op
);
638 memcpy(&key
->ipv4
.addr
.src
, arp
->ar_sip
, sizeof(key
->ipv4
.addr
.src
));
639 memcpy(&key
->ipv4
.addr
.dst
, arp
->ar_tip
, sizeof(key
->ipv4
.addr
.dst
));
640 ether_addr_copy(key
->ipv4
.arp
.sha
, arp
->ar_sha
);
641 ether_addr_copy(key
->ipv4
.arp
.tha
, arp
->ar_tha
);
643 memset(&key
->ip
, 0, sizeof(key
->ip
));
644 memset(&key
->ipv4
, 0, sizeof(key
->ipv4
));
646 } else if (eth_p_mpls(key
->eth
.type
)) {
647 size_t stack_len
= MPLS_HLEN
;
649 skb_set_inner_network_header(skb
, skb
->mac_len
);
653 error
= check_header(skb
, skb
->mac_len
+ stack_len
);
657 memcpy(&lse
, skb_inner_network_header(skb
), MPLS_HLEN
);
659 if (stack_len
== MPLS_HLEN
)
660 memcpy(&key
->mpls
.top_lse
, &lse
, MPLS_HLEN
);
662 skb_set_inner_network_header(skb
, skb
->mac_len
+ stack_len
);
663 if (lse
& htonl(MPLS_LS_S_MASK
))
666 stack_len
+= MPLS_HLEN
;
668 } else if (key
->eth
.type
== htons(ETH_P_IPV6
)) {
669 int nh_len
; /* IPv6 Header + Extensions */
671 nh_len
= parse_ipv6hdr(skb
, key
);
672 if (unlikely(nh_len
< 0)) {
675 memset(&key
->ip
, 0, sizeof(key
->ip
));
676 memset(&key
->ipv6
.addr
, 0, sizeof(key
->ipv6
.addr
));
679 skb
->transport_header
= skb
->network_header
;
688 if (key
->ip
.frag
== OVS_FRAG_TYPE_LATER
)
690 if (skb_shinfo(skb
)->gso_type
& SKB_GSO_UDP
)
691 key
->ip
.frag
= OVS_FRAG_TYPE_FIRST
;
693 /* Transport layer. */
694 if (key
->ip
.proto
== NEXTHDR_TCP
) {
695 if (tcphdr_ok(skb
)) {
696 struct tcphdr
*tcp
= tcp_hdr(skb
);
697 key
->tp
.src
= tcp
->source
;
698 key
->tp
.dst
= tcp
->dest
;
699 key
->tp
.flags
= TCP_FLAGS_BE16(tcp
);
701 memset(&key
->tp
, 0, sizeof(key
->tp
));
703 } else if (key
->ip
.proto
== NEXTHDR_UDP
) {
704 if (udphdr_ok(skb
)) {
705 struct udphdr
*udp
= udp_hdr(skb
);
706 key
->tp
.src
= udp
->source
;
707 key
->tp
.dst
= udp
->dest
;
709 memset(&key
->tp
, 0, sizeof(key
->tp
));
711 } else if (key
->ip
.proto
== NEXTHDR_SCTP
) {
712 if (sctphdr_ok(skb
)) {
713 struct sctphdr
*sctp
= sctp_hdr(skb
);
714 key
->tp
.src
= sctp
->source
;
715 key
->tp
.dst
= sctp
->dest
;
717 memset(&key
->tp
, 0, sizeof(key
->tp
));
719 } else if (key
->ip
.proto
== NEXTHDR_ICMP
) {
720 if (icmp6hdr_ok(skb
)) {
721 error
= parse_icmpv6(skb
, key
, nh_len
);
725 memset(&key
->tp
, 0, sizeof(key
->tp
));
732 int ovs_flow_key_update(struct sk_buff
*skb
, struct sw_flow_key
*key
)
734 return key_extract(skb
, key
);
737 static int key_extract_mac_proto(struct sk_buff
*skb
)
739 switch (skb
->dev
->type
) {
741 return MAC_PROTO_ETHERNET
;
743 if (skb
->protocol
== htons(ETH_P_TEB
))
744 return MAC_PROTO_ETHERNET
;
745 return MAC_PROTO_NONE
;
751 int ovs_flow_key_extract(const struct ip_tunnel_info
*tun_info
,
752 struct sk_buff
*skb
, struct sw_flow_key
*key
)
756 /* Extract metadata from packet. */
758 key
->tun_proto
= ip_tunnel_info_af(tun_info
);
759 memcpy(&key
->tun_key
, &tun_info
->key
, sizeof(key
->tun_key
));
761 if (tun_info
->options_len
) {
762 BUILD_BUG_ON((1 << (sizeof(tun_info
->options_len
) *
764 > sizeof(key
->tun_opts
));
766 ip_tunnel_info_opts_get(TUN_METADATA_OPTS(key
, tun_info
->options_len
),
768 key
->tun_opts_len
= tun_info
->options_len
;
770 key
->tun_opts_len
= 0;
774 key
->tun_opts_len
= 0;
775 memset(&key
->tun_key
, 0, sizeof(key
->tun_key
));
778 key
->phy
.priority
= skb
->priority
;
779 key
->phy
.in_port
= OVS_CB(skb
)->input_vport
->port_no
;
780 key
->phy
.skb_mark
= skb
->mark
;
781 ovs_ct_fill_key(skb
, key
);
782 key
->ovs_flow_hash
= 0;
783 res
= key_extract_mac_proto(skb
);
786 key
->mac_proto
= res
;
789 return key_extract(skb
, key
);
792 int ovs_flow_key_extract_userspace(struct net
*net
, const struct nlattr
*attr
,
794 struct sw_flow_key
*key
, bool log
)
798 /* Extract metadata from netlink attributes. */
799 err
= ovs_nla_get_flow_metadata(net
, attr
, key
, log
);
803 if (ovs_key_mac_proto(key
) == MAC_PROTO_NONE
) {
804 /* key_extract assumes that skb->protocol is set-up for
805 * layer 3 packets which is the case for other callers,
806 * in particular packets recieved from the network stack.
807 * Here the correct value can be set from the metadata
810 skb
->protocol
= key
->eth
.type
;
814 skb_reset_mac_header(skb
);
817 /* Normally, setting the skb 'protocol' field would be
818 * handled by a call to eth_type_trans(), but it assumes
819 * there's a sending device, which we may not have.
821 if (eth_proto_is_802_3(eth
->h_proto
))
822 skb
->protocol
= eth
->h_proto
;
824 skb
->protocol
= htons(ETH_P_802_2
);
827 return key_extract(skb
, key
);