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
21 #include <linux/uaccess.h>
22 #include <linux/netdevice.h>
23 #include <linux/etherdevice.h>
24 #include <linux/if_ether.h>
25 #include <linux/if_vlan.h>
26 #include <net/llc_pdu.h>
27 #include <linux/kernel.h>
28 #include <linux/jhash.h>
29 #include <linux/jiffies.h>
30 #include <linux/llc.h>
31 #include <linux/module.h>
33 #include <linux/rcupdate.h>
34 #include <linux/if_arp.h>
36 #include <linux/ipv6.h>
37 #include <linux/sctp.h>
38 #include <linux/tcp.h>
39 #include <linux/udp.h>
40 #include <linux/icmp.h>
41 #include <linux/icmpv6.h>
42 #include <linux/rculist.h>
45 #include <net/ndisc.h>
49 u64
ovs_flow_used_time(unsigned long flow_jiffies
)
51 struct timespec cur_ts
;
54 ktime_get_ts(&cur_ts
);
55 idle_ms
= jiffies_to_msecs(jiffies
- flow_jiffies
);
56 cur_ms
= (u64
)cur_ts
.tv_sec
* MSEC_PER_SEC
+
57 cur_ts
.tv_nsec
/ NSEC_PER_MSEC
;
59 return cur_ms
- idle_ms
;
62 #define TCP_FLAGS_OFFSET 13
63 #define TCP_FLAG_MASK 0x3f
65 void ovs_flow_used(struct sw_flow
*flow
, struct sk_buff
*skb
)
69 if ((flow
->key
.eth
.type
== htons(ETH_P_IP
) ||
70 flow
->key
.eth
.type
== htons(ETH_P_IPV6
)) &&
71 flow
->key
.ip
.proto
== IPPROTO_TCP
&&
72 likely(skb
->len
>= skb_transport_offset(skb
) + sizeof(struct tcphdr
))) {
73 u8
*tcp
= (u8
*)tcp_hdr(skb
);
74 tcp_flags
= *(tcp
+ TCP_FLAGS_OFFSET
) & TCP_FLAG_MASK
;
77 spin_lock(&flow
->lock
);
80 flow
->byte_count
+= skb
->len
;
81 flow
->tcp_flags
|= tcp_flags
;
82 spin_unlock(&flow
->lock
);
85 static int check_header(struct sk_buff
*skb
, int len
)
87 if (unlikely(skb
->len
< len
))
89 if (unlikely(!pskb_may_pull(skb
, len
)))
94 static bool arphdr_ok(struct sk_buff
*skb
)
96 return pskb_may_pull(skb
, skb_network_offset(skb
) +
97 sizeof(struct arp_eth_header
));
100 static int check_iphdr(struct sk_buff
*skb
)
102 unsigned int nh_ofs
= skb_network_offset(skb
);
106 err
= check_header(skb
, nh_ofs
+ sizeof(struct iphdr
));
110 ip_len
= ip_hdrlen(skb
);
111 if (unlikely(ip_len
< sizeof(struct iphdr
) ||
112 skb
->len
< nh_ofs
+ ip_len
))
115 skb_set_transport_header(skb
, nh_ofs
+ ip_len
);
119 static bool tcphdr_ok(struct sk_buff
*skb
)
121 int th_ofs
= skb_transport_offset(skb
);
124 if (unlikely(!pskb_may_pull(skb
, th_ofs
+ sizeof(struct tcphdr
))))
127 tcp_len
= tcp_hdrlen(skb
);
128 if (unlikely(tcp_len
< sizeof(struct tcphdr
) ||
129 skb
->len
< th_ofs
+ tcp_len
))
135 static bool udphdr_ok(struct sk_buff
*skb
)
137 return pskb_may_pull(skb
, skb_transport_offset(skb
) +
138 sizeof(struct udphdr
));
141 static bool sctphdr_ok(struct sk_buff
*skb
)
143 return pskb_may_pull(skb
, skb_transport_offset(skb
) +
144 sizeof(struct sctphdr
));
147 static bool icmphdr_ok(struct sk_buff
*skb
)
149 return pskb_may_pull(skb
, skb_transport_offset(skb
) +
150 sizeof(struct icmphdr
));
153 static int parse_ipv6hdr(struct sk_buff
*skb
, struct sw_flow_key
*key
)
155 unsigned int nh_ofs
= skb_network_offset(skb
);
163 err
= check_header(skb
, nh_ofs
+ sizeof(*nh
));
168 nexthdr
= nh
->nexthdr
;
169 payload_ofs
= (u8
*)(nh
+ 1) - skb
->data
;
171 key
->ip
.proto
= NEXTHDR_NONE
;
172 key
->ip
.tos
= ipv6_get_dsfield(nh
);
173 key
->ip
.ttl
= nh
->hop_limit
;
174 key
->ipv6
.label
= *(__be32
*)nh
& htonl(IPV6_FLOWINFO_FLOWLABEL
);
175 key
->ipv6
.addr
.src
= nh
->saddr
;
176 key
->ipv6
.addr
.dst
= nh
->daddr
;
178 payload_ofs
= ipv6_skip_exthdr(skb
, payload_ofs
, &nexthdr
, &frag_off
);
179 if (unlikely(payload_ofs
< 0))
183 if (frag_off
& htons(~0x7))
184 key
->ip
.frag
= OVS_FRAG_TYPE_LATER
;
186 key
->ip
.frag
= OVS_FRAG_TYPE_FIRST
;
189 nh_len
= payload_ofs
- nh_ofs
;
190 skb_set_transport_header(skb
, nh_ofs
+ nh_len
);
191 key
->ip
.proto
= nexthdr
;
195 static bool icmp6hdr_ok(struct sk_buff
*skb
)
197 return pskb_may_pull(skb
, skb_transport_offset(skb
) +
198 sizeof(struct icmp6hdr
));
201 static int parse_vlan(struct sk_buff
*skb
, struct sw_flow_key
*key
)
204 __be16 eth_type
; /* ETH_P_8021Q */
207 struct qtag_prefix
*qp
;
209 if (unlikely(skb
->len
< sizeof(struct qtag_prefix
) + sizeof(__be16
)))
212 if (unlikely(!pskb_may_pull(skb
, sizeof(struct qtag_prefix
) +
216 qp
= (struct qtag_prefix
*) skb
->data
;
217 key
->eth
.tci
= qp
->tci
| htons(VLAN_TAG_PRESENT
);
218 __skb_pull(skb
, sizeof(struct qtag_prefix
));
223 static __be16
parse_ethertype(struct sk_buff
*skb
)
225 struct llc_snap_hdr
{
226 u8 dsap
; /* Always 0xAA */
227 u8 ssap
; /* Always 0xAA */
232 struct llc_snap_hdr
*llc
;
235 proto
= *(__be16
*) skb
->data
;
236 __skb_pull(skb
, sizeof(__be16
));
238 if (ntohs(proto
) >= ETH_P_802_3_MIN
)
241 if (skb
->len
< sizeof(struct llc_snap_hdr
))
242 return htons(ETH_P_802_2
);
244 if (unlikely(!pskb_may_pull(skb
, sizeof(struct llc_snap_hdr
))))
247 llc
= (struct llc_snap_hdr
*) skb
->data
;
248 if (llc
->dsap
!= LLC_SAP_SNAP
||
249 llc
->ssap
!= LLC_SAP_SNAP
||
250 (llc
->oui
[0] | llc
->oui
[1] | llc
->oui
[2]) != 0)
251 return htons(ETH_P_802_2
);
253 __skb_pull(skb
, sizeof(struct llc_snap_hdr
));
255 if (ntohs(llc
->ethertype
) >= ETH_P_802_3_MIN
)
256 return llc
->ethertype
;
258 return htons(ETH_P_802_2
);
261 static int parse_icmpv6(struct sk_buff
*skb
, struct sw_flow_key
*key
,
264 struct icmp6hdr
*icmp
= icmp6_hdr(skb
);
266 /* The ICMPv6 type and code fields use the 16-bit transport port
267 * fields, so we need to store them in 16-bit network byte order.
269 key
->ipv6
.tp
.src
= htons(icmp
->icmp6_type
);
270 key
->ipv6
.tp
.dst
= htons(icmp
->icmp6_code
);
272 if (icmp
->icmp6_code
== 0 &&
273 (icmp
->icmp6_type
== NDISC_NEIGHBOUR_SOLICITATION
||
274 icmp
->icmp6_type
== NDISC_NEIGHBOUR_ADVERTISEMENT
)) {
275 int icmp_len
= skb
->len
- skb_transport_offset(skb
);
279 /* In order to process neighbor discovery options, we need the
282 if (unlikely(icmp_len
< sizeof(*nd
)))
285 if (unlikely(skb_linearize(skb
)))
288 nd
= (struct nd_msg
*)skb_transport_header(skb
);
289 key
->ipv6
.nd
.target
= nd
->target
;
291 icmp_len
-= sizeof(*nd
);
293 while (icmp_len
>= 8) {
294 struct nd_opt_hdr
*nd_opt
=
295 (struct nd_opt_hdr
*)(nd
->opt
+ offset
);
296 int opt_len
= nd_opt
->nd_opt_len
* 8;
298 if (unlikely(!opt_len
|| opt_len
> icmp_len
))
301 /* Store the link layer address if the appropriate
302 * option is provided. It is considered an error if
303 * the same link layer option is specified twice.
305 if (nd_opt
->nd_opt_type
== ND_OPT_SOURCE_LL_ADDR
307 if (unlikely(!is_zero_ether_addr(key
->ipv6
.nd
.sll
)))
309 memcpy(key
->ipv6
.nd
.sll
,
310 &nd
->opt
[offset
+sizeof(*nd_opt
)], ETH_ALEN
);
311 } else if (nd_opt
->nd_opt_type
== ND_OPT_TARGET_LL_ADDR
313 if (unlikely(!is_zero_ether_addr(key
->ipv6
.nd
.tll
)))
315 memcpy(key
->ipv6
.nd
.tll
,
316 &nd
->opt
[offset
+sizeof(*nd_opt
)], ETH_ALEN
);
327 memset(&key
->ipv6
.nd
.target
, 0, sizeof(key
->ipv6
.nd
.target
));
328 memset(key
->ipv6
.nd
.sll
, 0, sizeof(key
->ipv6
.nd
.sll
));
329 memset(key
->ipv6
.nd
.tll
, 0, sizeof(key
->ipv6
.nd
.tll
));
335 * ovs_flow_extract - extracts a flow key from an Ethernet frame.
336 * @skb: sk_buff that contains the frame, with skb->data pointing to the
338 * @in_port: port number on which @skb was received.
339 * @key: output flow key
341 * The caller must ensure that skb->len >= ETH_HLEN.
343 * Returns 0 if successful, otherwise a negative errno value.
345 * Initializes @skb header pointers as follows:
347 * - skb->mac_header: the Ethernet header.
349 * - skb->network_header: just past the Ethernet header, or just past the
350 * VLAN header, to the first byte of the Ethernet payload.
352 * - skb->transport_header: If key->eth.type is ETH_P_IP or ETH_P_IPV6
353 * on output, then just past the IP header, if one is present and
354 * of a correct length, otherwise the same as skb->network_header.
355 * For other key->eth.type values it is left untouched.
357 int ovs_flow_extract(struct sk_buff
*skb
, u16 in_port
, struct sw_flow_key
*key
)
362 memset(key
, 0, sizeof(*key
));
364 key
->phy
.priority
= skb
->priority
;
365 if (OVS_CB(skb
)->tun_key
)
366 memcpy(&key
->tun_key
, OVS_CB(skb
)->tun_key
, sizeof(key
->tun_key
));
367 key
->phy
.in_port
= in_port
;
368 key
->phy
.skb_mark
= skb
->mark
;
370 skb_reset_mac_header(skb
);
372 /* Link layer. We are guaranteed to have at least the 14 byte Ethernet
373 * header in the linear data area.
376 memcpy(key
->eth
.src
, eth
->h_source
, ETH_ALEN
);
377 memcpy(key
->eth
.dst
, eth
->h_dest
, ETH_ALEN
);
379 __skb_pull(skb
, 2 * ETH_ALEN
);
380 /* We are going to push all headers that we pull, so no need to
381 * update skb->csum here. */
383 if (vlan_tx_tag_present(skb
))
384 key
->eth
.tci
= htons(vlan_get_tci(skb
));
385 else if (eth
->h_proto
== htons(ETH_P_8021Q
))
386 if (unlikely(parse_vlan(skb
, key
)))
389 key
->eth
.type
= parse_ethertype(skb
);
390 if (unlikely(key
->eth
.type
== htons(0)))
393 skb_reset_network_header(skb
);
394 __skb_push(skb
, skb
->data
- skb_mac_header(skb
));
397 if (key
->eth
.type
== htons(ETH_P_IP
)) {
401 error
= check_iphdr(skb
);
402 if (unlikely(error
)) {
403 if (error
== -EINVAL
) {
404 skb
->transport_header
= skb
->network_header
;
411 key
->ipv4
.addr
.src
= nh
->saddr
;
412 key
->ipv4
.addr
.dst
= nh
->daddr
;
414 key
->ip
.proto
= nh
->protocol
;
415 key
->ip
.tos
= nh
->tos
;
416 key
->ip
.ttl
= nh
->ttl
;
418 offset
= nh
->frag_off
& htons(IP_OFFSET
);
420 key
->ip
.frag
= OVS_FRAG_TYPE_LATER
;
423 if (nh
->frag_off
& htons(IP_MF
) ||
424 skb_shinfo(skb
)->gso_type
& SKB_GSO_UDP
)
425 key
->ip
.frag
= OVS_FRAG_TYPE_FIRST
;
427 /* Transport layer. */
428 if (key
->ip
.proto
== IPPROTO_TCP
) {
429 if (tcphdr_ok(skb
)) {
430 struct tcphdr
*tcp
= tcp_hdr(skb
);
431 key
->ipv4
.tp
.src
= tcp
->source
;
432 key
->ipv4
.tp
.dst
= tcp
->dest
;
434 } else if (key
->ip
.proto
== IPPROTO_UDP
) {
435 if (udphdr_ok(skb
)) {
436 struct udphdr
*udp
= udp_hdr(skb
);
437 key
->ipv4
.tp
.src
= udp
->source
;
438 key
->ipv4
.tp
.dst
= udp
->dest
;
440 } else if (key
->ip
.proto
== IPPROTO_SCTP
) {
441 if (sctphdr_ok(skb
)) {
442 struct sctphdr
*sctp
= sctp_hdr(skb
);
443 key
->ipv4
.tp
.src
= sctp
->source
;
444 key
->ipv4
.tp
.dst
= sctp
->dest
;
446 } else if (key
->ip
.proto
== IPPROTO_ICMP
) {
447 if (icmphdr_ok(skb
)) {
448 struct icmphdr
*icmp
= icmp_hdr(skb
);
449 /* The ICMP type and code fields use the 16-bit
450 * transport port fields, so we need to store
451 * them in 16-bit network byte order. */
452 key
->ipv4
.tp
.src
= htons(icmp
->type
);
453 key
->ipv4
.tp
.dst
= htons(icmp
->code
);
457 } else if ((key
->eth
.type
== htons(ETH_P_ARP
) ||
458 key
->eth
.type
== htons(ETH_P_RARP
)) && arphdr_ok(skb
)) {
459 struct arp_eth_header
*arp
;
461 arp
= (struct arp_eth_header
*)skb_network_header(skb
);
463 if (arp
->ar_hrd
== htons(ARPHRD_ETHER
)
464 && arp
->ar_pro
== htons(ETH_P_IP
)
465 && arp
->ar_hln
== ETH_ALEN
466 && arp
->ar_pln
== 4) {
468 /* We only match on the lower 8 bits of the opcode. */
469 if (ntohs(arp
->ar_op
) <= 0xff)
470 key
->ip
.proto
= ntohs(arp
->ar_op
);
471 memcpy(&key
->ipv4
.addr
.src
, arp
->ar_sip
, sizeof(key
->ipv4
.addr
.src
));
472 memcpy(&key
->ipv4
.addr
.dst
, arp
->ar_tip
, sizeof(key
->ipv4
.addr
.dst
));
473 memcpy(key
->ipv4
.arp
.sha
, arp
->ar_sha
, ETH_ALEN
);
474 memcpy(key
->ipv4
.arp
.tha
, arp
->ar_tha
, ETH_ALEN
);
476 } else if (key
->eth
.type
== htons(ETH_P_IPV6
)) {
477 int nh_len
; /* IPv6 Header + Extensions */
479 nh_len
= parse_ipv6hdr(skb
, key
);
480 if (unlikely(nh_len
< 0)) {
481 if (nh_len
== -EINVAL
) {
482 skb
->transport_header
= skb
->network_header
;
490 if (key
->ip
.frag
== OVS_FRAG_TYPE_LATER
)
492 if (skb_shinfo(skb
)->gso_type
& SKB_GSO_UDP
)
493 key
->ip
.frag
= OVS_FRAG_TYPE_FIRST
;
495 /* Transport layer. */
496 if (key
->ip
.proto
== NEXTHDR_TCP
) {
497 if (tcphdr_ok(skb
)) {
498 struct tcphdr
*tcp
= tcp_hdr(skb
);
499 key
->ipv6
.tp
.src
= tcp
->source
;
500 key
->ipv6
.tp
.dst
= tcp
->dest
;
502 } else if (key
->ip
.proto
== NEXTHDR_UDP
) {
503 if (udphdr_ok(skb
)) {
504 struct udphdr
*udp
= udp_hdr(skb
);
505 key
->ipv6
.tp
.src
= udp
->source
;
506 key
->ipv6
.tp
.dst
= udp
->dest
;
508 } else if (key
->ip
.proto
== NEXTHDR_SCTP
) {
509 if (sctphdr_ok(skb
)) {
510 struct sctphdr
*sctp
= sctp_hdr(skb
);
511 key
->ipv6
.tp
.src
= sctp
->source
;
512 key
->ipv6
.tp
.dst
= sctp
->dest
;
514 } else if (key
->ip
.proto
== NEXTHDR_ICMP
) {
515 if (icmp6hdr_ok(skb
)) {
516 error
= parse_icmpv6(skb
, key
, nh_len
);